3 * See http://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
4 * Server support for kernels:
5 2.6.32-358.6.1.el6 (RHEL6)
6 * Client support for unpatched kernels:
7 2.6.32-358.6.1.el6 (RHEL6)
9 2.6.18-348.3.1.el5 (RHEL5)
10 2.6.16.60-0.69.1 (SLES10)
12 * Recommended e2fsprogs version: 1.42.7.wc1
13 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
14 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
15 stack overflow. For more information, please refer to bugzilla 17630.
16 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
17 the re-exporting nodes to be mounted with "32bitapi" mount option
19 --------------------------------------------------------------------------------
21 9-30-2012 Whamcloud, Inc.
23 * See http://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
24 * Server support for kernels:
25 2.6.32-279.5.1.el6 (RHEL6)
26 * Client support for unpatched kernels:
27 2.6.32-279.5.1.el6 (RHEL6)
29 2.6.18-238.19.1.el5 (RHEL5)
30 2.6.16.60-0.69.1 (SLES10)
31 * Recommended e2fsprogs version: 1.41.90.wc4
32 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
33 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
34 stack overflow. For more information, please refer to bugzilla 17630.
35 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
36 the re-exporting nodes to be mounted with "32bitapi" mount option
38 Severity : enhancement
40 Description: Update RHEL6.3 kernel to 2.6.32-279.5.1.el6.
42 Severity : enhancement
44 Description: Update RHEL6.3 kernel to 2.6.32-279.2.1.el6.
46 Severity : enhancement
48 Description: Update RHEL6.2 kernel to 2.6.32-220.17.1.el6
50 Severity : enhancement
52 Description: Update RHEL6.2 kernel to 2.6.32-220.13.1.el6
54 Severity : enhancement
56 Description: Update RHEL6.2 kernel to 2.6.32-220.7.1.el6
58 --------------------------------------------------------------------------------
60 04-30-2012 Whamcloud, Inc.
62 * See http://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
63 * Server support for kernels:
64 2.6.32-220.4.2.el6 (RHEL6)
65 * Client support for unpatched kernels:
66 2.6.32-220.4.2.el6 (RHEL6)
68 2.6.18-238.19.1.el5 (RHEL5)
69 2.6.16.60-0.69.1 (SLES10)
70 * Recommended e2fsprogs version: 1.41.90.wc4
71 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
72 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
73 stack overflow. For more information, please refer to bugzilla 17630.
74 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
75 the re-exporting nodes to be mounted with "32bitapi" mount option
77 Severity : enhancement
79 Description: Update RHEL6.2 kernel to 2.6.32-220.4.2.el6
81 Severity : enhancement
83 Description: Update RHEL6.2 kernel to 2.6.32-220.4.1.el6
85 Severity : enhancement
87 Description: Update RHEL6.2 kernel to 2.6.32-220.el6
89 Severity : enhancement
91 Description: Update RHEL6.1 kernel to 2.6.32-131.17.1.el6
93 --------------------------------------------------------------------------------
95 09-30-2011 Whamcloud, Inc.
97 * Server support for kernels:
98 2.6.18-238.19.1.el5 (RHEL5)
99 2.6.32-131.6.1.el6 (RHEL6)
100 * Client support for unpatched kernels:
101 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
102 2.6.18-238.19.1.el5 (RHEL5)
103 2.6.32-131.6.1.el6 (RHEL6)
104 2.6.29.4-167.fc11 (FC11)
105 2.6.16.60-0.69.1 (SLES10)
107 * Recommended e2fsprogs version: 1.41.90.wc3
108 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
109 Lustre filesystem with 4K stack may cause a stack overflow. For more
110 information, please refer to bugzilla 17630.
111 * NFSv4 reexport requires Lustre client on the re-exporting nodes to be
112 mounted with "32bitapi" mount option
114 Severity : enhancement
115 Description: Update RHEL6.1 kernel to 2.6.32-131.6.1.el6
117 Severity : enhancement
118 Description: Update RHEL5.6 kernel to 2.6.18-238.19.1.el5
120 Severity : enhancement
121 Description: Update RHEL5.6 kernel to 2.6.18-238.9.1.el5
124 Frequency : only with SLES10
126 Description: Use OFED "KMP" provided by Novell
127 Details : SLES10 SP3 ships with OFED in a separate "KMP" package.
128 Lustre is now built against this package. That means you need to
129 install the ofed-kmp package from Novell for the patchless client.
130 Note that the ofed-kmp that Novell ships may not exactly match the
131 kernel version but should still be compatible.
133 Severity : enhancement
135 Description: Update RHEL5.5 kernel to 2.6.18-194.17.1.el5
136 Update OEL5.5 kernel to 2.6.18-194.17.1.0.1.el5.
138 --------------------------------------------------------------------------------
140 2010-07-15 Oracle, Inc.
142 * Server support for kernels:
143 2.6.18-164.11.1.el5 (RHEL 5)
144 2.6.18-164.11.1.0.1.el5 (OEL 5)
145 * Client support for unpatched kernels:
146 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
147 2.6.18-164.11.1.el5 (RHEL 5),
148 2.6.18-164.11.1.0.1.el5 (OEL 5)
149 2.6.16.60-0.42.8 (SLES 10),
151 2.6.29.4-167.fc11 (FC11)
152 * Recommended e2fsprogs version: 1.41.10-sun2
154 Severity : enhancement
156 Description: Distribute lfs_migrate script with lustre
157 Details : The lfs_migrate mentioned in the manual is now included in the
158 lustre tree in lustre/scripts and will distributed in lustre RPMs.
160 Severity : enhancement
162 Description: Failover nid list added to import proc file.
164 Severity : enhancement
166 Description: "som_preview" mount option required to use SOM
167 Details : The clients will need to use som_preview mount option
171 Description: Lustre does not do 1MB IOs to HW RAID
172 Details : Bump MAX_PHYS/HW_SEGMENTS and SG_ALL to 256 in the RHEL5 kernel.
174 Severity : enhancement
176 Description: lfs quota output is not very convenient for awk/sed-parsing
177 Details : Some positions in lfs quota output table could be empty or
178 non-empty which made it hard to parse it with scripts, now
179 a dash is put instead of space where there is not supposed
184 Description: Update to OFED 1.5.1
185 Details : OFED 1.5.1 will be used in environments that don't use
186 vendor supplied OFED.
190 Description: Remove lnet. prefix from lctl params display
191 Details : Also renamed lnet.memused to lnet_memused to
196 Description: Use conf_param -d to remove permanant settings.
200 Description: Add list_param -R
201 Details : lctl list_param now supports recursive mode.
204 Bugzilla : 21829, 22312
205 Description: Fix llobdstat.
206 Details : 1. Now updating the write_bytes counter
207 2. Port script fixes from 1.8.
211 Description: decrease the usage of memory on clients.
212 Details : 1. On clients, recycle dentries and inodes unused.
213 2. Delete the code related to ll_deathrow(att 6215 in bug 1443). It
218 Description: llobdstat fix and enhancement
219 Details : add a counter to set a limit to how many samples will be returned
220 fix a wildcard in the path to limit to obdfilter stats only
224 Description: lfs check only allowed for root
225 Details : Code cleanups and added a regresion test of non-root use of
230 Description: MDS Oopses in ptlrpc_dispatch_difficult_reply()/__list_add()
231 Details : Fix hr->hr_index overload.
235 Description: Fix FIEMAP functionality.
239 Description: FC11 patchless client support added
243 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and
244 OEL5.4 kernel to 2.6.18-164.11.1.0.1.el5.
248 Description: New acc-sm script and various test-framework enhancments.
252 Description: Previously landed patch was reverted.
256 Description: Update SLES10 kernel to 2.6.16.60-0.42.8 (patchless)
260 Description: Add ldiskfs maxdirsize mount option.
264 Description: Fix mkfs.lustre support for 16TB LUNs.
268 Description: Print mmp_check_interval
269 Details : print mmp_check_interval and make it possible to abort
270 mount operation in case it takes too long.
274 Description: Add cascading_rw to lustre mpi tests.
278 Description: Store SOM data in EA.
279 Details : Also several SOM related cleanups.
283 Description: Add refcount to md_open_data
287 Description: Use refcount on lov_request_set to prevent memory corruption.
291 Description: Statahead: use parent dir's i_mutex for synchronization.
295 Description: loadgen improvements.
299 Description: Enable peer health detection in ksocklnd and ko2iblnd.
303 Description: Control DCACHE_LUSTRE_INVALID flag with MDS_INODE_LOOKCUP lock
307 Description: Fix race on obd_devcie:md_stats
311 Description: Fix SLES 11 client builds.
312 Details : Resolved uninitalized pointer gcc warnings.
316 Description: Allow non-root access for "lfs check".
317 Details : Added a check in obd_class_ioctl() for OBD_IOC_PING_TARGET.
321 Description: Clear imp_force_reconnect correctly in ptlrpc_connect_interpret()
323 Severity : enhancement
325 Description: Update kernel to RHEL5.4 2.6.18-164.6.1.el5 and
326 OEL5 2.6.18-164.6.1.0.1.el5(Both in-kernel OFED enabled).
330 Description: Sanity test_27p fails.
331 Details : OBD_FAIL_OST_ENOSPC has been moved to filter_statfs().
332 Sanity tests 27 were cleaned up and improved.
336 Description: Unify req format on client/servers
337 Details : Use new req_capsule API [almost] everywhere instead of old PTLRPC
338 buffers and swabbers approach..
343 Description: drop join file stuff.
345 Severity : enhancement
347 Description: Add 2.6.27 and SLES11 (patchless client) support
351 Description: Fix in ptlrpc_expire_one_request() to print the signed time
354 Severity : enhancement
356 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
359 Severity : enhancement
361 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
363 Severity : enhancement
365 Description: correctly handle big reply message.
366 Details : send LNet event if reply is bigger then buffer and adjust this buffer
371 Description: Drop unnecessary __GFP_NOMEMALLOC flag from filter_get_page()
373 Severity : enhancement
375 Description: Add support for OEL5.
379 Description: Conf-sanity.sh 50g test - deactivated OST should not cause a panic.
381 Severity : enhancement
383 Description: "weak-modules" support
384 Details : Implement "weak-modules" support which enables kernel modules
385 to be used with any kernel that implements the same kABI. In
386 order to achieve this modules are now installed in
387 /lib/modules/$(uname -r)/updates/kernel on all distributions.
391 Description: client could not reconnect to OST because of an active request.
392 Details : The client now retries to reconnect to the same server, if a
393 connect request failed with EBUSY or -EAGAIN. Test case 35b
394 conf-sanity.sh has been created.
398 Description: Don't allow make backward step on assignin osc next id.
399 Details : race between allocation next id and ll_sync thread can be cause
400 of set wrong osc next id and can be kill valid ost objects.
404 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
408 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
412 Description: Error handling in osc_statfs_interpret() has been improved.
413 Details : Check in osc_statfs_interpret() for EBADR.
417 Description: Do not send statfs() requests to OSTs disabled by administrator.
418 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
420 Severity : enhancement
422 Description: Added sanity-quota test to for lfs(1) to work with numeric uid/gid.
424 Severity : enhancement
426 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
428 Severity : enhancement
430 Description: Enhancement for lfs(1) command to use numeric uid/gid.
434 Description: Improved cleanup procedure for conf-sanity test_46a.
438 Description: Do not start bulk transfer, if client tried to reconnect.
439 Details : Do not start bulk transfer, if client tried to reconnect,
440 but failed due to active rpcs. Check for exp_abort_active_req
441 has been implemented.
445 Description: Prevent inconsistences between linux and lustre mount structures.
446 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
447 Make the sleep interruptible.
449 Severity : enhancement
451 Description: provide server to client comms path
452 Details : server to client communications path via new LDLM_SET_INFO rpc
456 Description: truncate starts GFP_FS allocation under transaction causing deadlock
457 Details : ldiskfs_truncate calls grab_cache_page which may start page
458 allocation under an open transaction. This may lead to
459 calling prune_icache with consequent lustre reentrance.
463 Description: send correctly lsm on open replay
464 Details : MDS is trust to LSM size on replay open, but client can set wrong size
469 Description: enable client interface failover
470 Details : When a child reconnects from another NID, properly update export
471 nid hash position and ldlm reverse import.
473 Severity : enhancement
475 Description: Communicate OST degraded/readonly state via statfs to MDS
476 Details : Flags in the statfs returned from OSTs indicate whether the
477 OST is in a degraded RAID state, or if the filesystem has
478 turned read-only after a filesystem error is detected.
483 Description: don't panic if EPROTO was hit when reading symlink
484 Details : correctly handling request reference in error cases.
486 Severity : enhancement
488 Description: Add LustreNetLink, a kernel-userspace communcation path.
490 Severity : enhancement
492 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
497 Description: access to llog context before init.
498 Details : move handling CATALOGS file at osc layer and forbid access to llog
502 Frequency : rare, only when client is writing to OST on local node
504 Description: Avoid deadlock for local client writes
505 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
506 memory freeing context. This allows OST threads to set the
507 PF_MEMALLOC flag on task structures in order to allocate memory
508 from reserved pools and complete IO.
509 Use GFP_HIGHUSER for OST allocations for non-local client writes,
510 so that the OST threads generate memory pressure and allow
511 inactive pages to be reclaimed.
513 Severity : enhancement
515 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
520 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
521 Details : move ldlm namespace creation in setup phase to avoid grab
522 _lprocfs_lock with cli_sem held.
526 Description: Temporarily disable grant shrink.
527 Details : Disable the feature for debugging.
530 Frequency : only when formatting test filesystems
532 Description: Unable to run several mkfs.lustre on loop devices at the same time
533 Details : mkfs.lustre returns error 256 on the concurrent loop devices
534 formatting. The solution is to proper handle the error.
536 Severity : enhancement
538 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
540 Severity : enhancement
542 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
547 Description: resolve race between obd_disconnect and class_disconnect_exports
548 Details : if obd_disconnect will be called to already disconnected export he
549 forget release one reference and osc module can't unloaded.
551 Severity : enhancement
553 Description: Allow tuning service thread via /proc
554 Details : For each service a new
555 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
556 created that can be used to set min/max thread counts, and get the
557 current number of running threads.
562 Descriptoin: don't return error if have partially created objects for file.
563 Details : lov_update_create_set uses set->set_success as index for created
564 objects, so if some requests will be failed, they will have hole
565 at end of array and qos_shrink_lsm for allocate correct lsm.
567 Severity : enhancement
569 Description: Update OFED support to 1.4.1
574 Description: Reduce small size read RPC
575 Details : Set read-ahead limit for every file and only do read-ahead when
576 available read-ahead pages are bigger than 1M to avoid small size
579 Severity : enhancement
581 Description: move AT tunable parameters for more consistent usage
582 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
584 Severity : enhancement
586 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
587 Details : allow skip disconnected ost for send statfs request and hide error
593 Description: fix racy locking of mballoc block bitmaps causing BUG
594 Details : The locking of the mballoc buddy bitmap and the in-memory
595 block bitmap was using two different spin locks in some
596 cases. This made it possible to incorrectly access the
597 mballoc bitmap while another process was modifying it,
598 causing a sanity assertion to fail. While no on-disk corruption
599 was reported, there was some risk of this happening.
602 Frequency : rare, on llog test 6
604 Descriptoin: don't allow connect to already connected import
605 Details : allowing connect to already connected import is hide connecting problem.
610 Descriptoin: allow kill process which wait statahead result
611 Details : for some reasons 'ls' can stick in waiting result from statahead,
612 in this case need way for kill this process.
614 Severity : enhancement
616 Description: Add state history info file, enhance import info file
617 Details : Track import connection state changes in a new osc/mdc proc file;
618 add overview-type data to the osc/mdc import proc file.
620 Severity : enhancement
622 Description: MDS create should not wait for statfs RPC while holding DLM lock.
624 Severity : enhancement
626 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
631 Descriptoin: OST grant too much space to client even there are not enough space.
632 Details : Client will shrink its grant cache to OST if there are no write
633 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
634 this grant cache if there are already not enough avaible space
635 (left_space < total_clients * 32M).
638 Frequency : start MDS on uncleanly shutdowned MDS device
640 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
641 Details : stay in waiting mds<>ost recovery finished produce random bugs
642 due race between two ll_sync thread for one lov target. send
643 ACTIVATE event only if connect realy finished and import have
647 Frequency : rare, connect and disconnect target at same time
649 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
650 Details : don't call obd_disconnect under lov_lock. this long time
651 operation and can block ptlrpcd which answer to connect request.
656 Descriptoin: don't lose wakeup for imp_recovery_waitq
657 Details : recover_import_no_retry or invalidate_import and import_close can
658 both sleep on imp_recovery_waitq, but we was send only one wakeup
662 Frequency : always with long access acl
664 Descriptoin: mds can't pack reply with long acl.
665 Details : mds don't control size of acl but they limited by reint/getattr
668 Severity : enhancement
670 Description: Update to SLES10 kernel-2.6.16.60-0.33.
672 Severity : enhancement
674 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
677 Frequency : start MDS on uncleanly shutdowned MDS device
679 Descriptoin: aborting recovery hang on MDS
680 Details : don't throttle destroy RPCs for the MDT.
683 Frequency : on remount
685 Description: external journal device not working after the remount
686 Details : clear dev_rdonly flag for external journal devices in
692 Description: shutdown vs evict race
693 Details : client_disconnect_export vs connect request race.
694 if client will evicted at this time - we start invalidate
695 thread without referece to import and import can be freed
699 Frequency : rare, need acl's on inode.
701 Description: client can't handle ost additional correctly
702 Details : if ost was added after client connected to mds client can have
703 hit lnet_try_match_md ... to big messages to wide striped files.
704 in this case need teach client to handle config events about add
705 lov target and update client max ea size at that event.
707 Severity : enhancement
709 Description: Changelogs
710 Details : Changelogs are a lightweight mechanism to track filesystem
711 metadata and namespace changes. The changelog is recorded
712 permanently on the MDTs, and is periodically "consumed" / purged
713 when records are no longer needed.
715 Severity : enhancement
717 Description: compact fld format with extents
718 Details : Store range of seq rather than every seq in FLD. Seq
719 controller update FLD rather than clients. In Case of CMD, mdt0
720 has FLD, all other metadata server act as non persistent proxy
721 for FLD queries and cache fld entries in fld cache.
726 Description: don't skip ost target if they assigned to file
727 Details : Drop slow OSCs if we can, but not for requested start idx.
728 This means "if OSC is slow and it is not the requested
729 start OST, then it can be skipped, otherwise skip it only
730 if it is inactive/recovering/out-of-space.
734 Description: more cleanup in mds_lov
735 Details : not send LOV EA under replay, we can't know about they size at this
736 time. Don't allow client connect to mds before any ost connected,
737 for avoid problems with LOV EA size and returning EIO to client.
739 Severity : enhancement
741 Description: Interoperability at server side (Disk interoperability)
743 Severity : enhancement
745 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
747 Severity : enhancement
749 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
751 Severity : enhancement
753 Description: New client IO stack (CLIO).
755 Severity : enhancement
757 Description: Commit on sharing. Eliminate inter-client dependencies between
758 uncommitted transactions by doing transaction commits.
759 Thereby clients may recovery independently.
762 Frequency : Create a symlink file with a very long name
764 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
765 Details : If there is no extra space in the request for early cancels,
766 ldlm_req_handles_avail() returns 0 instead of a negative value.
768 Severity : enhancement
770 Description: Add /proc entry for import status
771 Details : The mdc, osc, and mgc import directories now have
772 an import directory that contains useful import data for debugging
775 Severity : enhancement
777 Description: Re-disable certain /proc logging
778 Details : Enable and disable client's offset_stats, extents_stats and
779 extents_stats_per_process stats logging on the fly.
782 Frequency : Only on FC kernels 2.6.22+
784 Description: oops in statahead
785 Details : Do not drop reference count for the dentry from VFS when lookup,
786 VFS will do that by itself.
788 Severity : enhancement
790 Description: Generic /proc file permissions
791 Details : Set /Proc file permissions in a more generic way to enable non-
792 root users operate on some /proc files.
796 Description: Hitting mdc_commit_close() ASSERTION
797 Details : Properly handle request reference release in
798 ll_release_openhandle().
802 Description: quota recovery deadlock during mds failover
803 Details : This patch includes att18982, att18236, att18237 in bz14840.
805 1. fix osts hang when mds does failover with quotaon
806 2. prevent watchdog storm when osts threads wait for the
811 Frequency : only patchless client
812 Description: add workaround for race between add/remove dentry from hash
814 Severity : enhancement
816 Description: Allow OST glimpses to return PW locks
820 Description: LBUG when llog conf file is full
821 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
825 Description: Prevent import from entering FULL state when server in recovery
829 Description: service mount cannot take device name with ":"
830 Details : Only when device name contains ":/" will mount treat it as
836 Description: replace ptlrpcd with the statahead thread to interpret the async
837 statahead RPC callback
841 Frequency : on recovery
842 Description: I/O failures after umount during fail back
843 Details : if client reconnected to restarted server we need join to recovery
844 instead of find server handler is changed and process self
845 eviction with cancel all locks.
847 Severity : enhancement
849 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
851 Severity : enhancement
853 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
855 Severity : enhancement
857 Description: Upcall on Lustre log has been dumped
858 Details : Allow for a user mode script to be called once a Lustre log has
859 been dumped. It passes the filename of the dumped log to the
860 script, the location of the script can be specified via
861 /proc/sys/lnet/debug_log_upcall.
866 Description: avoid idr_remove called for id which is not allocated.
867 Details : Move assigment s_dev for clustered nfs to end of initialization,
868 to avoid problem with error handling.
873 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
874 Details : When connection is reused this not moved from CONN_UNUSED_HASH
875 into CONN_USED_HASH and this prodice warning when put connection
876 again in unused hash.
879 Severity : enhancement
881 Description: File striping can now be set to use an arbitrary pool of OSTs.
883 Severity : enhancement
885 Description: Export bytes_read/bytes_write count on OSC/OST.
889 Description: Early reply size mismatch, MGC loses connection
890 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
891 the connect flags are properly negotiated.
896 Description: quotas are not honored with O_DIRECT
897 Details : all writes with the flag O_DIRECT will use grants which leads to
898 this problem. Now using OBD_BRW_SYNC to guard this.
902 Description: add quota statistics
903 Details : 1. sort out quota proc entries and proc code.
904 2. add quota statistics
906 Severity : enhancement
908 Description: enable quota support for HEAD.
912 Description: Properly propagate oinfo flags from lov to osc for statfs
913 Details : restore missing copy oi_flags to lov requests.
915 Severity : enhancement
917 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
918 : l_getgroups(8), lst(8), routerstat(8)
919 Details : included man pages for llobdstat(8), llstat(8),
920 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
922 Severity : enhancement
924 Description: configure's --enable-quota should check the
925 : kernel .config for CONFIG_QUOTA
926 Details : configure is terminated if --enable-quota is passed but
927 : no quota support is in kernel
931 Description: Remove portals compatibility
932 Details : Remove portals compatibility, not interoperable with releases
937 Description: Resolve device initialization race
938 Details : Prevent proc handler from accessing devices added to the
939 obd_devs array but yet be intialized.
941 Severity : enhancement
943 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
945 Severity : enhancement
947 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
952 Description: Using wrong pointer in osc_brw_prep_request
953 Details : Access to array[-1] can produce panic if kernel compiled with
954 CONFIG_PAGE_ALLOC enabled
958 Description: Client runs out of low memory
959 Details : Consider only lowmem when counting initial number of llap pages
963 Description: *optional* service tags registration
964 Details : if the "service tags" package is installed on a Lustre node
965 When the filesystem is mounted, a local-node service tag will
966 be created. See http://inventory.sun.com/ for more information
967 about the Service Tags asset management system.
971 Description: Kernel BUG tries to release flock
972 Details : Lustre does not destroy flock lock before last reference goes
973 away. So always drop flock locks when client is evicted and
974 perform unlock regardless of successfulness of speaking to MDS.
978 Description: add refcount protection for osc callbacks, avoid panic on shutdown
982 Description: sanity test 65a fails if stripecount of -1 is set
983 Details : handle -1 striping on filesystem in ll_dirstripe_verify
988 Description: ASSERTION(CheckWriteback(page,cmd)) failed
989 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
992 Severity : enhancement
994 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
998 Description: do not process already freed flock
999 Details : flock can possibly be freed by another thread before it reaches
1000 to ldlm_flock_completion_ast.
1004 Description: LBUG during stress test
1005 Details : Need properly lock accesses the flock deadlock detection list.
1009 Description: oops in page fault handler
1010 Details : kernel page fault handler can return two special 'pages' in error
1011 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
1015 Description: timeout with invalidate import.
1016 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
1017 should be handled by ptlrpcd. This produce long age waiting and
1018 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
1020 Severity : enhancement
1022 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
1026 Description: enable MGS and MDT services start separately
1027 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
1028 a co-located MGS without starting the MGS, which is a complement
1029 to 'nosvc' mount option.
1033 Frequency : after recovery
1034 Description: precreate to many object's after del orphan.
1035 Details : del orphan st in oscc last_id == next_id and this triger growing
1036 count of precreated objects. Set flag LOW to skip increase count
1037 of precreated objects.
1041 Frequency : rare, on clear nid stats
1042 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
1043 Details : when clean nid stats sometimes try destroy live entry,
1044 and this produce panic in free.
1048 Description: Stack overflow during MDS log replay
1049 ease stack pressure by using a thread dealing llog_process.
1053 Description: wait until IO finished before start new when do lock cancel.
1054 Details : VM protocol want old IO finished before start new, in this case
1055 need wait until PG_writeback is cleared until check dirty flag and
1056 call writepages in lock cancel callback.
1058 Severity : enhancement
1060 Description: using special macro for print time and cleanup in includes.
1064 Description: mds_mfd_close() ASSERTION(rc == 0)
1065 Details : In mds_mfd_close(), we need protect inode's writecount change
1066 within its orphan write semaphore to prevent possible races.
1070 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
1074 Frequency : rare, on shutdown ost
1075 Description: don't hit live lock with umount ost.
1076 Details : shrink_dcache_parent can be in long loop with destroy dentries,
1077 use shrink_dcache_sb instead.
1081 Description: don't panic with use echo client
1082 Details : echo client pass NULL as client nid pointer and this produce null
1083 pointer dereference.
1087 Description: fix build on ppc32
1088 Details : compile code with -m64 flag produce wrong object file for ppc32.
1092 Description: add message levels for liblustreapi
1096 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
1097 Details : Move the 'good_osts' check before the 'total_bavail' check. This
1098 will result in an -EAGAIN and in the exit call path we call
1099 alloc_rr() which will with increasing aggressiveness attempt to
1100 aquire precreated objects on the minimum number of required OSCs.
1104 Description: Use old size assignment to avoid deadlock
1105 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
1106 the scheduling while holding a spinlock. We do not need locking
1107 for size in ll_update_inode() because size is only updated from
1108 the MDS for directories or files without objects, so there is no
1109 other place to do the update, and concurrent access to such inodes
1110 are protected by the inode lock.
1114 Description: resolve "_IOWR redefined" build error on SLES10
1118 Description: dump the memory debugging after all modules are unloaded to
1119 suppress false negative in conf_sanity test 39
1121 Severity : enhancement
1123 Description: build kernel-ib packages for OFED 1.3 in our release cycle
1128 Description: fix SLES kernel versioning
1129 Details : the kernel version for our SLES 10 kernel did not include a "-"
1130 before the "smp" at the end. while this was not a problem in
1131 general, it did mean that software trying to use the kernel
1132 version to try to detect a vendor specific kernel would fail.
1133 this was most evident by the OFED build scripts.
1137 Description: Don't update lov_desc members until making sure they are valid
1138 Details : When updating lov_desc members via proc fs, need fix their
1139 validities before doing the real update.
1143 Description: don't put request into delay list while invalidate in flight.
1144 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
1145 import in flight. this produce timeout for invalidate and sometimes
1146 can cause stale data.
1150 Frequency : on ppc only
1151 Description: not convert ost objects for directory because it's not exist.
1152 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
1156 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
1158 Severity : enhancement
1160 Description: Update to support 2.6.22.14 vanilla kernel.
1164 Frequency : rare, on recovery
1165 Description: read procfs can produce deadlock in some situation
1166 Details : Holding lprocfs lock which send rpc can produce block for destroy
1167 obd objects and this also block reconnect with -EALREADY. This
1168 isn't fix all lprocfs bugs - but make it rare.
1170 Severity : enhancement
1172 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
1175 Frequency : frequent on X2 node
1177 Description: mdc_set_open_replay_data LBUG
1178 Details : Set replay data for requests that are eligible for replay.
1182 Description: lustre_mgs: operation 101 on unconnected MGS
1183 Details : When MGC is disconnected from MGS long enough, MGS will evict the
1184 MGC, and late on MGC cannot successfully connect to MGS and a lot
1185 of the error messages complaining that MGS is not connected.
1189 Frequency : on network error
1190 Description: panic with double free request if network error
1191 Details : mdc_finish_enqueue is finish request if any network error ocuring,
1192 but it's true only for synchronus enqueue, for async enqueue
1193 (via ptlrpcd) this incorrect and ptlrpcd want finish request
1196 Severity : enhancement
1198 Description: client-side metadata stat-ahead during readdir(directory readahead)
1199 Details : perform client-side metadata stat-ahead when the client detects
1200 readdir and sequential stat of dir entries therein
1203 Frequency : on start mds
1205 Description: Implement get_info(last_id) in obdfilter.
1208 Frequency : occasional
1210 Description: Correctly check stale fid, not start epoch if ost not support SOM
1211 Details : open with flag O_CREATE need set old fid in op_fid3 because
1212 op_fid2 was overwritten with new generated fid, but mds can answer
1213 with one of these two fids and both is not stale. Setattr
1214 incorrectly started an epoch and assume will be called
1215 done_writing, but without SOM done_writing ever being called.
1218 Frequency : rare, depends on device drivers and load
1220 Description: MDS or OSS nodes crash due to stack overflow
1221 Details : Code changes in 1.8.0 increased the stack usage of some functions.
1222 In some cases, in conjunction with device drivers that use a lot
1223 of stack the MDS (or possibly OSS) service threads could overflow
1224 the stack. One change which was identified to consume additional
1225 stack has been reworked to avoid the extra stack usage.
1228 Frequency : occasional
1230 Description: Do not fail import if osc_interpret_create gets -EAGAIN
1231 Details : If osc_interpret_create got -EAGAIN it immediately exits and
1232 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
1233 oscc_has_objects and see OSC has no objests and call
1234 oscc_internal_create to resend create request.
1236 Severity : enhancement
1238 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
1240 Severity : enhancement
1242 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
1245 Frequency : very rare
1247 Description: Fix replay if there is an un-replied request and open
1248 Details : In some cases, older replay request will revert the
1249 mcd->mcd_last_xid on MDS which is used to record the client's
1250 latest sent request.
1252 Severity : enhancement
1254 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
1256 Severity : enhancement
1258 Description: Add rhel5 support to HEAD.
1260 Serverity : enhancement
1262 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
1267 Description: Don't allow skipping OSTs if index has been specified.
1268 Details : Don't allow skipping OSTs if index has been specified, make
1269 locking in internal create lots better.
1273 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
1274 Details : access to bitfield in structure is always rounded to long
1275 and this produce problem with not atomic change any bit.
1279 Description: Lustre make rpms failed.
1280 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
1281 builds Lustre rpms from tarball.
1284 Frequency : rare on shutdown ost
1286 Description: If llog cancel was not send before clean_exports phase, this can
1287 produce deadlock in llog code.
1288 Details : If llog thread has last reference to obd and call class_import_put
1289 this produce deadlock because llog_cleanup_commit_master wait when
1290 last llog_commit_thread exited, but this never success because was
1291 called from llog_commit_thread.
1295 Description: allow userland application know is lost one of stripes.
1296 Details : fill lvb_blocks with error code on ost and return it to
1297 application if error flag found.
1301 Description: NULL lov_tgts causing MDS oops
1302 Details : more safe checks for NULL lov_tgts for avoid oops.
1304 Severity : enhancement
1306 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
1310 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
1311 Details : If an OST is down the MDS will hang indefinitely in
1312 obd_statfs_rqset() waiting for the statfs data. While for
1313 MDS QOS usage of statfs, it should not stuck in waiting.
1315 Severity : enhancement
1317 Description: remote_acl support
1318 Details : Support ACL-based permission check for remote user.
1319 Support setfacl/getfacl for remote user with the utils
1320 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
1321 the system "{s,g}etfacl" utils.
1323 Severity : enhancement
1325 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
1327 Severity : enhancement
1329 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
1333 Description: make mgs_setparam() handle fsname containing dash
1334 Details : fsname containing a dash does not work with lctl conf_param
1336 Severity : enhancement
1338 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
1340 Severity : enhancement
1342 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
1346 Frequency : rare, at shutdown
1347 Description: access already free / zero obd_namespace.
1348 Details : if client_disconnect_export was called without force flag set,
1349 and exist connect request in flight, this can produce access to
1350 NULL pointer (or already free pointer) when connect_interpret
1351 store ocd flags in obd_namespace.
1355 Frequency : only at startup
1356 Description: not alloc memory with spinlock held.
1357 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
1358 if any spinlock held.
1360 Severity : enhancement
1362 Description: make lustre randomly fail allocating memory
1363 Details : Make lustre randomly failed allocating memory for testing purpose.
1365 Severity : enhancement
1367 Description: lost problems with lov objid file
1368 Details : Fixes some scability and access to not inited memory problems
1369 in work with lov objdid file.
1374 Description: lfs find does not continue on file error
1375 Details : Continue other files processing when a file/dir is absent.
1379 Description: Inconsistent usage of lustre_pack_reply()
1380 Details : Standardize the usage of lustre_pack_reply() such that it
1381 always generate a CERROR on failure.
1384 Frequency : occasional
1386 Description: MDS hang or stay in waiting lock
1387 Details : If client receive lock with CBPENDING flag ldlm need send lock
1388 cancel as separate rpc, to avoid situation when cancel request
1389 can't processed due all i/o threads stay in wait lock.
1393 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1397 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1401 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1405 Description: Update to RHEL4 latest kernel.
1407 Severity : enhancement
1409 Description: Build SLES10 patchless client fails
1410 Details : The configure was broken by run ./configure with
1411 --with-linux-obj=.... argument for patchless client. When the
1412 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1413 can't search header adequately. Use absolute path such as
1414 -I($LINUX)/include instead.
1418 Description: interrupt oig_wait produce painc on resend.
1419 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1420 private set, and this produce situation when rq_ptlrpcd_data not
1421 copyed to new allocated request and triggered LBUG on assert
1422 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1423 ptlrpcd set if request is changed and can be safety changed to use
1428 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1429 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1430 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1431 these macros and replace them with their difinitions instead.
1435 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1436 Details : Update which_patch & target file for SLES10 latest kernel.
1439 Description: Remove client patches from SLES 10 kernel.
1440 Details : This causes SLES 10 clients to behave as patchless clients
1441 even on a Lustre-patched (server) kernel.
1443 Severity : enhancement
1445 Description: self-adjustable client's lru lists
1446 Details : use adaptive algorithm for managing client cached locks lru
1447 lists according to current server load, other client's work
1448 pattern, memory activities, etc. Both, server and client
1449 side namespaces provide number of proc tunables for controlling
1452 Severity : enhancement
1454 Description: light-weight GSS support
1455 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
1456 and reduce performance overhead.
1458 Severity : enhancement
1460 Description: Linux keyring support
1461 Details : Support using service of Linux keyring for Lustre GSS internal
1462 context refresh/cache mechanism.
1466 Description: Fix errors in lfs documentation
1467 Details : Fixes man pages
1471 Description: don't use GFP_* in generic Lustre code.
1472 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
1476 Description: obdclass is limited by single OBD_ALLOC(idarray)
1477 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
1481 Description: warnings with build patchless client with vanila 2.6.19 and up
1482 Details : change old ctl_table style and replace ctl_table/ctl_table_header
1483 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
1487 Frequency : Only for SLES
1488 Description: sanity_quota fail test_1
1489 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
1490 file, which makes TSTID[2] inunique.
1494 Frequency : Only for SLES9
1495 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
1496 Details : Update target/ChangeLog/which_patch .
1500 Frequency : Only for test_10 in sanity-quota.sh
1501 Description: a bug in quota test code
1502 Details : checking a test flag in a wrong place
1506 Frequency : Only for test_13 in sanity-quota.sh
1507 Description: a bug in quota test script
1508 Details : shouldn't check overfull allocated quota
1512 Description: jbd statistics
1513 Details : Port older jbd statistics patch for sles10
1515 Severity : enhancement
1517 Description: Add printing inode info into message about error in writepage.
1522 Description: reply_lock_interpret crash due to race with it and lock cancel.
1523 Details : Do not replay locks that are being cancelled. Do not reference
1524 locks by their address during replay, just by their handle.
1528 Frequency : When flocks are used.
1529 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
1530 Details : Flock locks might destroy just granted lock if it could be merged
1531 with another existing flock, this is done in completion handler,
1532 so teach ldlm_cli_enquque_fini that this is a valid case for
1538 Description: Oops in read and write path when failing to allocate lock.
1539 Details : Check if lock allocation failed and return error back.
1543 Description: lstripe command fails for valid OST index
1544 Details : The stripe offset is compared to lov->desc.ld_tgt_count
1545 instead of lov->desc.ld_active_tgt_count.
1549 Description: sanity.sh failed test 103
1550 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
1555 Description: df doesn't work properly if diskfs blocksize != 4K
1556 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
1560 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
1561 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
1563 Severity : enhancement
1565 Description: Adaptive timeouts
1566 Details : RPC timeouts adapt to changing server load and network
1567 conditions to reduce resend attempts and improve recovery time.
1571 Description: llapi_file_create() does not allow some changes
1572 Details : add llapi_file_open() that allows specifying the mode and
1573 open flags, and also returns an open file handle.
1577 Description: merge and cleanup kernel patches.
1578 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
1582 Description: Add journal checksum support.(Kernel part)
1583 Details : The journal checksum feature adds two new flags i.e
1584 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
1585 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
1586 indicates that the commit block contains the checksum for
1587 the blocks described by the descriptor blocks. Now commit
1588 record can be sent to disk without waiting for descriptor
1589 blocks to be written to disk. This behavior is controlled
1590 using JBD2_FEATURE_ASYNC_COMMIT flag.
1594 Description: OSS needs mutliple precreate threads
1595 Details : Add ability to start more than one create thread per OSS.
1599 Description: Sanity.sh test_65k failed: setstripe should have succeeded
1600 Details : execute lfs setstripe on client
1604 Description: mds_obd_create error creating tmp object
1605 Details : When the user sets quota on root, llog will be affected and can't
1606 create files and write files.
1609 Frequency : Always on ia64 patchless client, and possibly others.
1611 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
1612 Details : This allows the patchless client to be loaded on architectures
1613 without this export.
1617 Description: RedHat Update kernel for RHEL5
1618 Details : Add the kernel config file for RHEL5.
1622 Description: RedHat Update kernel for RHEL5
1623 Details : Modify the kernel config file more closer RHEL5.
1627 Description: Build failure against Centos5 (RHEL5)
1628 Details : Define PAGE_SIZE when it isn't present.
1632 Description: test_7 on sanity-quota.sh can't be used on separate machines
1633 Details : add facet to handle it
1637 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
1638 Details : with NFS, the anon dentry's parent was set to itself in
1639 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
1640 corresponding dentry other than use rec->ur_name.
1642 Severity : enhancement
1644 Description: lfs setstripe enhancement
1645 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
1649 Description: enable data checksumming by default
1650 Details : enable checksum by default, allow --disable-checksum
1651 configure option and "-o nochecksum" mount option. Checksums
1652 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
1656 Description: lustre support for RHEL5
1657 Details : Add support for RHEL5.
1661 Description: Client eviction due to failover config
1662 Details : after a connection loss, the lustre client should attempt to
1663 reconnect to the last active server first before trying the
1664 other potential connections.
1668 Description: when mds and osts use different quota unit(32bit and 64bit),
1669 quota will be released repeatly.
1670 Details : void sending multiple quota reqs to mds, which will keep the
1671 status between the reqs.
1675 Description: osts not allocated evenly to files
1676 Details : change the condition to increase offset_idx
1679 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
1681 Description: Data corruption for OSTs that are formatted larger than 2TB
1683 Details : When generating the bio request for lustre file writes the
1684 sector number would overflow a temporary variable before being
1685 used for the IO. The data reads correctly from Lustre (which
1686 will overflow in a similar manner) but other file data or
1687 filesystem metadata may be corrupted in some cases.
1691 Description: Tune the kernel for good SCSI performance.
1692 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
1693 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
1698 Description: rewrite ext2-derived code in obdclass/uuid.c
1699 Details : rewrite inherited code (uuid parsing code from ext2 utils)
1700 from scratch preserving functionality.
1704 Description: "lfs find -obd UUID" prints directories
1705 Details : "lfs find -obd UUID" will return all directory names instead
1706 of just file names. It is incorrect because the directories
1707 do not reside on the OSTs.
1711 Description: use i_size_read and i_size_write in 2.6 port
1712 Details : replace inode->i_size access with i_size_read/write()
1715 Frequency : when removing large files
1717 Description: scheduling issue during removal of large Lustre files
1718 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
1719 It causes scheduling issues when removing large files (17TB in the
1723 Frequency : only with liblustre clients on XT3
1725 Description: evictions taking too long
1726 Details : allow llrd to evict clients directly on OSTs
1729 Frequency : only on ppc
1731 Description: /proc/fs/lustre/devices broken on ppc
1732 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
1733 We should make sure the type of 'index' is loff_t and then cast
1734 explicitly as needed below. Do not assign an explicitly cast
1738 Frequency : only for rhel5
1740 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
1741 Details : Modify the target file & which_kernel.
1743 Severity : enhancement
1745 Description: omit set fsid for export NFS
1746 Details : fix set/restore device id for avoid EMFILE error and mark lustre
1747 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
1751 Frequency : Always, for kernels after 2.6.16
1752 Description: Fix warning idr_remove called for id=.. which is not allocated.
1753 Details : Last kernels save old s_dev before kill super and not allow
1754 to restore from callback - restore it before call kill_anon_super.
1758 Description: buffer overruns could theoretically occur
1759 Details : llapi_semantic_traverse() modifies the "path" argument by
1760 appending values to the end of the origin string, and a
1761 overrun may occur. Adding buffer overrun check in liblustreapi.
1765 Description: Fix error on 'ls .' at the top of the Lustre mount.
1766 Details : Don't revalidate dentry if it is a root dentry.
1770 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
1771 Details : Modify vm-tunables-rhel4.patch.
1775 Description: Kernel config for 2.6.18-vanilla.
1776 Details : Modify targets/2.6-vanilla.target.in.
1777 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
1778 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
1779 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
1780 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
1784 Description: improve handling recoverable errors
1785 Details : if request processig with error which can be recoverable on server
1786 request should be resend, otherwise page released from cache and
1791 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
1792 Details : Modify target file & which_patch.
1793 A flaw was found in the IA32 system call emulation provided
1794 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
1795 value could be stored in the %RAX register, which could trigger an
1796 out-of-bounds system call table access. An untrusted local user
1797 could exploit this flaw to run code in the kernel
1798 (ie a root privilege escalation). (CVE-2007-4573).
1802 Description: O_DIRECT bypasses client statistics.
1803 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
1804 read_ahead_stats were updated. copied the stats section from
1805 osc_send_oap_rpc() into async_internal().
1809 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
1813 Description: change order of libsysio includes
1814 Details : '#include sysio.h' should always come before '#include xtio.h'
1818 Description: handle "serious error: objid * already exists" more gracefully
1819 Details : If LAST_ID value on disk is smaller than the objects existing in
1820 the O/0/d* directories, it indicates disk corruption and causes an
1821 LBUG(). If the object is 0-length, then we should use the existing
1822 object. This will help to avoid a full fsck in most cases.
1824 Severity : enhancement
1826 Description: adapt the lustre_config script to support the upgrade case
1827 Details : Add "-u" option for lustre_config script to support upgrading 1.4
1828 server targets to 1.6 in parallel.
1832 Description: To avoid grant space > avaible space when the disk is almost
1833 full. Without this patch you might see the error "grant XXXX >
1834 available" or some LBUG about grant, when the disk is almost
1836 Details : In filter_check_grant, for non_grant cache write, we should
1837 check the left space by if (*left > ungranted + bytes), instead
1838 of (*left > ungranted), because only we are sure the left space
1839 is enough for another "bytes", then the ungrant space should be
1840 increase. In client, we should update cl_avail_grant only there
1841 is OBD_MD_FLGRANT in the reply.
1845 Description: Update RHEL 4 kernel to fix local root privilege escalation.
1846 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
1847 described in CVE-2007-4573. This problem could allow untrusted
1848 local users to gain root access.
1851 Frequency : when using O_DIRECT and quotas
1853 Description: Incorrect file ownership on O_DIRECT output files
1854 Details : block usage reported by 'lfs quota' does not take into account
1855 files that have been written with O_DIRECT.
1860 Description: touch file failed when fs is not full
1861 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
1862 otherwise we can get ENOSP while fs is not full.
1866 Description: parallel lock callbacks
1867 Details : Instead of sending blocking and completion callbacks as separated
1868 requests, adding them to a set and sending in parallel.
1871 Frequency : only for Cray XT3
1872 Bugzilla : 12829/13455
1873 Description: Changing primary group doesn't change the group lustre assigns to
1875 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
1876 overridden with the primary group provided by the group upcall,
1877 whereas the supplied fsgid can be trusted if it is in the list of
1878 supplementary groups returned by the group upcall.
1880 Severity : enhancement
1882 Description: Allow masking D_WARNING, D_ERROR messages from console
1883 Details : Console messages can now be disabled via lnet.printk.
1887 Description: User code with malformed file open parameter crashes client node
1888 Details : Before packing join_file req, all the related reference should be
1889 checked carefully in case some malformed flags cause fake
1890 join_file req on client.
1894 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
1895 Details : ldlm_completion_ast() assumes that a lock is granted when the req
1896 mode is equal to the granted mode. However, it should also check
1897 that LDLM_FL_CP_REQD is not set.
1901 Description: Heavy nfs access might result in deadlocks
1902 Details : After ELC code landed, it is now improper to enqueue any mds
1903 locks under och_sem, because enqueue might want to decide to
1904 cancel open locks for same inode we are holding och_sem for.
1908 Description: Client eviction while running blogbench
1909 Details : A lot of unlink operations with concurrent I/O can lead to a
1910 deadlock causing evictions. To address the problem, the number of
1911 oustanding OST_DESTROY requests is now throttled to
1912 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
1913 callbacks are processed in priority.
1917 Description: enable ACLs on MDS by default
1918 Details : ACLs must be enabled on MDS by default.
1921 Frequency : PPC/PPC64 only
1923 Description: conflicts between asm-ppc64/types.h and lustre_types.h
1924 Details : fix duplicated definitions between asm-ppc64/types.h and
1925 lustre_types.h on PPC.
1928 Frequency : PPC/PPC64 only
1930 Description: asm-ppc/segment.h does not exist
1931 Details : fix compile issue on PPC.
1935 Description: better handle error messages in extents code
1938 Frequency : RHEL4 only
1940 Description: mkfs is very slow on IA64/RHEL4
1941 Details : A performance regression has been discovered in the MPT Fusion
1942 driver between versions 3.02.73rh and 3.02.99.00rh. As a
1943 consequence, we have downgraded the MPT Fusion driver in the RHEL4
1944 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
1946 Severity : enhancement
1948 Description: SNMP support enhancement
1949 Details : Adding total number of sampled request for an MDS node in snmp
1952 Severity : enhancement
1954 Description: Optimize ldlm waiting list processing for PR extent locks
1955 Details : When processing waiting list for read extent lock and meeting read
1956 lock that is same or wider to it that is not contended, skip
1957 processing rest of the list and immediatelly return current
1958 status of conflictness, since we are guaranteed there are no
1959 conflicting locks in the rest of the list.
1963 Description: Time out and refuse to reconnect
1964 Details : When the failover node is the primary node, it is possible
1965 to have two identical connections in imp_conn_list. We must
1966 compare not conn's pointers but NIDs, otherwise we can defeat
1967 connection throttling.
1971 Description: port llog fixes from b1_6 into HEAD
1972 Details : Port llog reference couting and some llog cleanups from b1_6
1973 (bug 10800) into HEAD, for protect from panic and access to already
1974 free llog structures.
1978 Description: Detect stride IO mode in read-ahead
1979 Details : When a client does stride read, read-ahead should detect that and
1980 read-ahead pages according to the detected stride pattern.
1984 Description: data checksumming impacts single node performance
1985 Details : add support for several checksum algorithm. Currently, only CRC32
1986 and Adler-32 are supported. The checksum type can be changed on
1987 the fly via /proc/fs/lustre/osc/*/checksum_type.
1991 Description: use adler32 for page checksums
1992 Details : when available, use the Adler-32 algorithm instead of CRC32 for
1997 Description: build for x2 fails
1998 Details : fix compile issue on Cray systems.
2002 Description: Properly match for duplicate locks
2003 Details : Due to different lock order from skiplists code, we need to
2004 traverse entire list for now
2007 Frequency : only on PPC/SLES10
2009 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
2010 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
2011 be defined. Add a hack in mkfs_lustre.c to work around this compile
2016 Description: LASSERT on MDS when client holding flock lock dies
2017 Details : ldlm pool logic depends on number of granted locks equal to
2018 number of released locks which is not true for flock locks, so
2019 just exclude such locks from consideration.
2023 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
2024 Details : Use fsfilt_sync() for both the whole filesystem sync and
2025 individual file sync to eliminate dangerous inode locking
2026 with I_LOCK that can lead to a deadlock.
2030 Description: performance in 1.6.3
2031 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
2032 sure that 1MB requests are not fragmented by the block layer.
2034 Severity : enhancement
2036 Description: organize the server-side client stats on per-nid basis
2037 Details : Change the structure of stats under obdfilter and mds to
2045 The "uuid"s file would list the uuids of _active_ exports.
2046 And the clear entry is to clear all stats and stale nids.
2048 Severity : enhancement
2050 Description: eliminate client locks in face of contention
2051 Details : file contention detection and lockless i/o implementation
2052 for contended files.
2056 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
2059 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
2061 Description: MDS may lose file striping (and hence file data) in some cases
2062 Details : If there are additional extended attributes stored on the MDS,
2063 in particular ACLs, SELinux, or user attributes (if user_xattr
2064 is specified for the client mount options) then there is a risk
2065 of attribute loss. Additionally, the Lustre file striping
2066 needs to be larger than default (e.g. striped over all OSTs),
2067 and an additional attribute must be stored initially in the
2068 inode and then increase in size enough to be moved to the
2069 external attribute block (e.g. ACL growing in size) for file
2074 Description: skiplist implementation simplification
2075 Details : skiplists are used to group compatible locks on granted list
2076 that was implemented as tracking first and last lock of each
2077 lock group the patch changes that to using doubly linked lists
2081 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
2082 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
2083 because LASSERT on both the data supplied by a client, and the
2084 data on disk is dangerous and incorrect.
2086 Severity : enhancement
2088 Description: Slow truncate/writes to huge files at high offsets.
2089 Details : Directly associate cached pages to lock that protect those pages,
2090 this allows us to quickly find what pages to write and remove
2091 once lock callback is received.
2095 Description: more ldlm soft lockups
2096 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
2097 starve other threads from the resource lock for a long time in
2098 case of long waiting queue, so change the debug level from
2099 D_OTHER to the less frequently used D_INFO.
2101 Severity : enhancement
2103 Description: add -gid, -group, -uid, -user options to lfs find
2107 Description: Hung threads in invalidate_inode_pages2_range
2108 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
2109 once one is completed. As a result, some RPCs are stuck in the
2110 queue and are never sent.
2114 Description: filter threads hungs on waiting journal commit
2115 Details : Cleanup filter group llog code, then only filter group llog will
2116 be only created in the MDS/OST syncing process.
2120 Description: Procfs and llog threads access destoryed import sometimes.
2121 Details : Sync the import destoryed process with procfs and llog threads by
2122 the import refcount and semaphore.
2124 Severity : enhancement
2126 Description: openlock cache of b1_6 port to HEAD
2131 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
2132 Details : If insertion of an extent fails, then discard the inode
2133 preallocation and free data blocks else it can lead to duplicate
2138 Description: don't always update ctime in ext3_xattr_set_handle()
2139 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
2140 In some cases the ctime should not be updated, for example for
2141 2.0->1.8 compatibility it is necessary to delete an xattr and it
2142 should not update the ctime.
2146 Bugzilla : 15713/16362
2147 Description: Assertion in iopen_connect_dentry in 1.6.3
2148 Details : looking up an inode via iopen with the wrong generation number can
2149 populate the dcache with a disconneced dentry while the inode
2150 number is in the process of being reallocated. This causes an
2151 assertion failure in iopen since the inode's dentry list contains
2152 both a connected and disconnected dentry.
2156 Description: assertion failure in ldlm_handle2lock()
2157 Details : fix a race between class_handle_unhash() and class_handle2object()
2158 introduced in lustre 1.6.5 by bug 13622.
2163 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
2164 Details : remove the limit of 256 scsi disks in the sd_iostat patch
2169 Description: oops in sd_iostats_seq_show()
2170 Details : unloading/reloading the scsi low level driver triggers a kernel
2171 bug when trying to access the sd iostat file.
2176 Description: Kernel panics during QLogic driver reload
2177 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
2178 patch, causing memory corruption.
2183 Description: journal_dev option does not work in b1_6
2184 Details : pass mount option during pre-mount.
2186 Severity : enhancement
2188 Description: Add a FIEMAP(FIle Extent MAP) ioctl
2189 Details : FIEMAP ioctl will allow an application to efficiently fetch the
2190 extent information of a file. It can be used to map logical blocks
2191 in a file to physical blocks in the block device.
2195 Description: LDLM soft lockups - improvement
2196 Details : It is be possible to send the lock handle along with each read
2197 or write request because the client is already doing a lock match
2198 itself so there isn't any reason the OST should have to re-do that
2202 Frequency : only on Cray X2
2204 Description: X2 build failures
2205 Details : fix build failures on Cray X2.
2209 Description: xid & resent requests
2210 Details : Initialize RPC XID from clock at startup (randomly if clock is
2213 Severity : enhancement
2215 Description: Add lustre_start utility to start or stop multiple Lustre servers
2220 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
2221 Details : In case of memory pressure, list_del() can be called twice on
2222 req->rq_history_list, causing a kernel oops.
2226 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
2227 Details : fix stack overflow in the distributed lock manager by defering
2228 export eviction after a failed AST to the elt thread instead of
2229 handling it in the dlm interpret routine.
2233 Description: Convert some comments to new format.
2234 Details : Update documenting comments to match doxygen conventions.
2238 Description: Grammar fixes.
2239 Details : A couple of trivial spelling fixes.
2243 Description: OSD_COUNTERS-mandatory
2244 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
2245 to simplify the code.
2249 Description: simplify cmm_device freeing logic.
2250 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
2254 Description: Add lockdep support to dt_object_operations locking interface.
2255 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
2256 parameter indicating lock ordering. Update mdd code to use new
2261 Description: Introduce failloc constants for lockless IO tests.
2262 Details : Add two new failloc constants to test lockless IO. Only one of
2263 them in implemented---another is checked in yet to be landed
2268 Description: Add lockdep support for inode mutex.
2269 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
2270 in the situations where Lustre has to lock more than one inode
2275 Description: Add optional invariants checking support.
2276 Details : Add new LINVRNT() macro, optional on new --enable-invariants
2277 configure switch. This macro is to be used for consistency and
2278 sanity checks that are too expensive to be left in `production'
2283 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
2284 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
2289 Description: Add lockdep support for ldlm_lock and ldlm_resource.
2290 Details : Use spin_lock_nested() in (the only) situation where more than
2291 one ldlm_lock is locked simultaneously. Also, fix possible
2292 dead-lock in ldlm_lock_change_resource() by enforcing particular
2297 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
2298 Details : Instead of specifying each ldlm_lock call-back through separate
2299 parameter, wrap them into struct ldlm_callback_suite.
2303 Description: Kill join_lru obd method and its callers.
2304 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
2305 in memory---a requirement implemented through ->o_join_lru() obd
2306 method in HEAD. Remove this method and its users.
2310 Description: Add asynchronous ldlm ENQUEUE completion handler.
2311 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
2312 case that stock ldlm_completion_ast() cannot handle as it waits
2313 until lock is granted. Introduce new ldlm_completion_ast_async()
2314 for this. Also comment ldlm_completion_ast().
2318 Description: ldlm_error <-> errno conversion.
2319 Details : Add functions to map (rather arbitrary) between LDLM error codes
2320 and standard errno values. CLIO needs this to prevent LDLM specific
2321 constants from escaping ldlm and osc.
2325 Description: Kill unused ldlm_handle2lock_ns() function.
2326 Details : Kill unused ldlm_handle2lock_ns() function.
2330 Description: Add lu_ref support to ldlm_lock
2331 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
2332 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
2333 and ldlm_resource. LDLM interface has to be changed, because code that
2334 releases a reference on a lock, has to "know" what reference this is.
2335 In the most frequent case
2337 lock = ldlm_handle2lock(handle);
2339 LDLM_LOCK_PUT(lock);
2341 no changes are required. When any other reference (received _not_ from
2342 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
2343 instead of LDLM_LOCK_PUT().
2345 Arguably, changes are pervasive, and interface requires some discipline
2346 for proper use. On the other hand, it was very instrumental in finding
2347 a few leaked lock references.
2351 Description: Add ldlm_lock_addref_try().
2352 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
2353 attempts to addref a lock that might be being canceled
2358 Description: Add ldlm_weigh_callback().
2359 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
2360 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
2361 hard-coded `number of pages' logic.
2365 Description: Add lockdep annotations to llog code.
2366 Details : Use appropriately tagged _nested() locking calls in the places
2367 where llog takes more than one ->lgh_lock lock.
2371 Description: Add loi_kms_set().
2372 Details : Wrap kms updates into a helper function.
2376 Description: Constify instances of struct lsm_operations.
2377 Details : Constify instances of struct lsm_operations.
2381 Description: lu_conf support.
2382 Details : On a server, a file system object is uniquely identified
2383 by a fid, which is sufficient to locate and load all object
2384 state (inode). On a client, on the other hand, more data are
2385 necessary instantiate an object. Change lu_object_find() and
2386 friends to take additional `lu_conf' argument describing object.
2387 Typically this includes layout information.
2391 Description: lu_context fixes.
2392 Details : Introduce new lu_context functions that are needed on the client
2393 side, where some system threads (ptlrpcd) are shared by multiple
2394 modules, and so cannot be stopped during module shutdown.
2398 Description: Add start and stop methods to lu_device_type_operations.
2399 Details : Introduce two new methods in lu_device_type_operations, that are
2400 invoked when first instance of a given type is created and last one
2401 is destroyed respectively. This is need by CLIO.
2405 Description: Add lu_ref support to struct lu_device.
2406 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
2407 track leaked references.
2411 Description: Introduce lu_kmem_descr.
2412 Details : lu_kmem_descr and its companion interface allow to create
2413 and destroy a number of kmem caches at once.
2417 Description: Fix lu_object finalization race.
2418 Details : Fix a race between lu_object_find() finding an object and its
2419 concurrent finalization. This race is (most likely) not possible
2420 on the server, but might happen on the client.
2424 Description: Introduce lu_ref interface.
2425 Details : lu_ref is a debugging module allowing to track references to
2426 a given object. It is quite cpu expensive, and has to be
2427 explicitly enabled with --enable-lu_ref. See usage description
2432 Description: Factor lu_site procfs stats into a separate function.
2433 Details : Separate lu_site stats printing code into a separate function
2434 that can be reused on a client.
2438 Description: Constify instances of struct {lu,dt,md}_device_operations.
2439 Details : Constify instances of struct {lu,dt,md}_device_operations.
2443 Description: Introduce struct md_site and move meta-data specific parts of
2444 struct lu_site here.
2445 Details : Move md-specific fields out of struct lu_site into special struct
2446 md_site, so that lu_site can be used on a client.
2450 Description: Kill mdd_lov_destroy().
2451 Details : Remove unused mdd code.
2455 Description: Add st_block checking to multistat.c.
2456 Details : Add st_block checking to multistat.c.
2460 Description: Add lu_ref support to struct obd_device.
2461 Details : Add lu_ref tracking to obd_device.
2465 Description: Kill obd_set_fail_loc().
2466 Details : Remove unused code.
2470 Description: Add special type for ptlrpc_request interpret functions.
2471 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
2472 there. Actual usage will be in CLIO.
2476 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
2477 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
2478 doesn't work with lockdep.
2482 Description: Add rwv.c test program.
2483 Details : New testing program exercising readv(2) and writev(2) (Qian).
2487 Description: Add sendfile.c test program.
2488 Details : New testing program exercising sendfile(2) (Jay).
2492 Description: Ratelimit a message that can be very frequent.
2493 Details : Ratelimit a memory allocation failure message that can
2498 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2499 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2503 Description: Liblustre build fixes.
2504 Details : Liblustre build fixes.
2508 Description: libcfs: add cfs_{need,cond}_resched() interface.
2509 Details : libcfs: add cfs_{need,cond}_resched() definition and
2510 implementations for Linux, NT, and liblustre.
2512 Severity : enhancement
2514 Description: More exported tunables for mballoc
2515 Details : Add support for tunable preallocation window and new tunables for
2516 large/small requests
2520 Description: Detect corruption of block bitmap and checking for preallocations
2521 Details : Checks validity of on-disk block bitmap. Also it does better
2522 checking of number of applied preallocations. When corruption is
2523 found, it turns filesystem readonly to prevent further corruptions.
2527 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
2528 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
2529 window to 0. If it does hit cache miss, and read-pattern is still
2530 stride-io mode, does not reset the stride window, but also does not
2531 increase the stride window length in this case.
2535 Frequency : only for big-endian servers
2536 Description: Check if system is big-endian while mounting fs with extents feature
2537 Details : Mounting a filesystem with extents feature will fail on big-endian
2538 systems since ext3-based ldiskfs is not supported on big-endian
2539 systems. This can be over-riden with "bigendian_extents" mount option.
2541 Severity : enhancement
2543 Description: The root squash functionality
2544 Details : A security feature, which is to prevent users from being able
2545 to mount lustre on their desktop, run as root, and delete
2546 all of the files in the filesystem. The goal is accomplished by
2547 remapping user id (UID) and group id (GID) of the root user to
2548 a UID and GID specified by the system administartor via Lustre
2549 configuration management server (MGS). The functionality also
2550 allows to specify sets of clients for which the remapping does
2555 Description: Excessive recovery window
2556 Details : With AT enabled, the recovery window can be excessively long (6000+
2557 seconds). To address this problem, we no longer use
2558 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
2559 timeout no longer depends on the service time, it is set to
2560 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
2561 time via pb_service_time.
2565 Description: Watchdog triggered on MDS failover
2566 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
2567 the OSTs know that the MDS "UUID" can be reused for the same export
2568 from a different NID, so we do not need to wait for the export to be
2572 Frequency : rare, only if using MMP with Linux RAID
2574 Description: MMP doesn't work with Linux RAID
2575 Details : While using HA for Lustre servers with Linux RAID, it is possible
2576 that MMP will not detect multiple mounts. To make this work we
2577 need to unplug the device queue in RAID when the MMP block is being
2578 written. Also while reading the MMP block, we should read it from
2579 disk and not the cached one.
2581 Severity : enhancement
2583 Description: open file using fid
2584 Details : A file can be opened using just its fid, like
2585 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
2588 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
2591 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
2592 Details : add proper locking for creation of jbd_4k slab cache
2596 Description: MMP check in ext3_remount() fails without displaying any error
2597 Details : When multiple mount protection fails during remount, proper error
2600 Severity : enhancement
2602 Description: Allow stripe size to be up to 4G-64k
2603 Details : Fix math logic to allow large stripe sizes.
2607 Description: add check for >8TB ldiskfs filesystems
2608 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
2609 Don't allow >8TB ldiskfs filesystems to be mounted without
2610 force_over_8tb mount option
2615 Descriptoin: OSS complains "too many blocks" and crashes
2616 Details : In filter_commitrw_write():
2617 1. lquota_chkquota();
2618 2. obdo_from_inode();
2619 3. lquota_pending_commit();
2620 "2." has changed uid and gid in this case, then caused
2621 lquota_chkquota() and lquota_pending_commit() to use
2622 different uid and gid. That is the root of the bug.
2626 Descriptoin: getstripe doesn't report stripe count = -1 sanely
2627 Details : Detect and print out the 'stripe count = -1' case properly.
2631 Description: Out or order replies might be lost on replay
2632 Details : In ptlrpc_retain_replayable_request if we cannot find retained
2633 request with tid smaller then one currently being added, add it
2634 to the start, not end of the list.
2636 --------------------------------------------------------------------------------
2638 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
2640 * Support for kernels:
2641 2.6.5-7.283 (SLES 9),
2642 2.6.9-55.EL (RHEL 4),
2643 2.6.16.46-0.14 (SLES 10),
2644 2.6.18.8 vanilla (kernel.org)
2645 * Client support for unpatched kernels:
2646 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2647 2.6.16 - 2.6.22 vanilla (kernel.org)
2648 * Due to recently discovered recovery problems, we do not recommend
2649 using patchless RHEL 4 clients with this or any earlier release.
2650 * Recommended e2fsprogs version: 1.39.cfs8
2651 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2652 * Starting with this release, the ldiskfs backing filesystem required
2653 by Lustre is now in its own package, lustre-ldiskfs. This package
2654 should be installed. It is versioned separately from Lustre and
2655 may be released separately in future.
2659 Description: block reactivating mgc import until all deactivates complete
2660 Details : Fix race when failing back MDT/MGS to itself (testing)
2663 Frequency : at statup only
2665 Description: mds_lov_synchronize race leads to various problems
2666 Details : simultaneous MDT->OST connections at startup can cause the
2667 sync to abort, leaving the OSC in a bad state.
2669 Severity : enhancement
2671 Description: add optional extra BUILD_VERSION info
2672 Details : add a new environment variable (namely LUSTRE_VERS) which allows
2673 to override the lustre version.
2676 Frequency : 2.6.18 servers only
2678 Description: ll_kern_mount() doesn't release the module reference
2679 Details : The ldiskfs module reference count never drops down to 0
2680 because ll_kern_mount() doesn't release the module reference.
2685 Description: server LBUG when using old ost_num_threads parameter
2686 Details : Accept the old ost_num_threads parameter but warn that it
2687 is deprecated, and fix an off-by-one error that caused an LBUG.
2692 Description: Transient SCSI error results in persistent IO issue
2693 Details : iobuf->dr_error is not reinitialized to 0 between two
2697 Frequency : sometimes when underlying device returns I/O errors
2699 Description: OSTs not going read-only during write failures
2700 Details : OSTs are not remounted read-only when the journal commit threads
2701 get I/O errors because fsfilt_ext3 calls journal_start/stop()
2702 instead of the ext3 wrappers.
2706 Description: poor connect scaling with increasing client count
2707 Details : Don't run filter_grant_sanity_check for more than 100 exports
2708 to improve scaling for large numbers of clients.
2711 Frequency : SLES10 only
2713 Description: sanity-quota.sh quotacheck failed: rc = -22
2714 Details : Quotas cannot be enabled on SLES10.
2717 Frequency : liblustre clients only
2719 Description: getdirentries does not give error when run on compute nodes
2720 Details : getdirentries does not fail when the size specified as an argument
2721 is too small to contain at least one entry
2723 Severity : enhancement
2725 Description: Add LNET router traceability for debug purposes
2726 Details : If a checksum failure occurs with a router as part of the
2727 IO path, the NID of the last router that forwarded the bulk data
2728 is printed so it can be identified.
2733 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
2734 Details : Due to a race condition, liblustre clients were occasionally
2735 evicted incorrectly.
2737 Severity : enhancement
2739 Description: lfs setstripe use optional parameters instead of postional
2742 Severity : enhancement
2744 Description: Nanosecond timestamp support for ldiskfs
2745 Details : The on-disk ldiskfs filesystem has added support for nanosecond
2746 resolution timestamps. There is not yet support for this at
2747 the Lustre filesystem level.
2750 Frequency : during server recovery
2752 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
2753 Details : request with rq_no_resend flag not awake l_wait_event if they get
2757 Frequency : nfs export on patchless client
2759 Description: connectathon hang when test nfs export over patchless client
2760 Details : Disconnected dentry cannot be found with lookup, so we do not need
2761 to unhash it or make it invalid
2764 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
2765 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
2766 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
2771 Description: open req refcounting wrong on reconnect
2772 Details : If reconnect happened between getting open reply from server and
2773 call to mdc_set_replay_data in ll_file_open, we will schedule
2774 replay for unreferenced request that we are about to free.
2775 Subsequent close will crash in variety of ways.
2776 Check that request is still eligible for replay in
2777 mdc_set_replay_data().
2782 Description: disable writes to filesystem when reading health_check file
2783 Details : the default for reading the health_check proc file has changed
2784 to NOT do a journal transaction and write to disk, because this
2785 can cause reads of the /proc file to hang and block HA state
2786 checking on a healthy but otherwise heavily loaded system. It
2787 is possible to return to the previous behaviour during configure
2788 with --enable-health-write.
2790 Severity : enhancement
2792 Description: 64-bit inode version
2793 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
2794 to the inode. This will be required for version-based recovery.
2799 Description: MDS fails to start if a duplicate client export is detected
2800 Details : in some rare cases it was possible for a client to connect to
2801 an MDS multiple times. Upon recovery the MDS would detect this
2802 and fail during startup. Handle this more gracefully.
2804 Severity : enhancement
2806 Description: Add -o localflock option to simulate old noflock
2808 Details : This will achieve local-only flock/fcntl locks
2814 Description: log_commit_thread vs filter_destroy race leads to crash
2815 Details : Take import reference before releasing llog record semaphore
2820 Description: Wrong request locking in request set processing
2821 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
2822 imp_delayed_list, in this place should be used imp_lock.
2825 Frequency : when reconnection
2827 Description: Grant Leak when osc reconnect to OST
2828 Details : When osc reconnect ost, OST(filter) should check whether it
2829 should grant more space to client by comparing fed_grant and
2830 cl_avail_grant, and return the granted space to client instead
2831 of "new granted" space, because client will call osc_init_grant
2832 to update the client grant space info.
2835 Frequency : when client reconnect to OST
2837 Description: Grant Leak when osc do resend and replay bulk write
2838 Details : When osc reconnect to OST, OST(filter)should clear grant info of
2839 bulk write request, because the grant info while be sync between
2840 OSC and OST when reconnect, and we should ignore the grant info
2841 these of resend/replay write req.
2846 Description: Grant space more than avaiable left space sometimes.
2847 Details : When then OST is about to be full, if two bulk writing from
2848 different clients came to OST. Accord the avaliable space of the
2849 OST, the first req should be permitted, and the second one
2850 should be denied by ENOSPC. But if the seconde arrived before
2851 the first one is commited. The OST might wrongly permit second
2852 writing, which will cause grant space > avaiable space.
2855 Frequency : when client is evicted
2857 Description: Grant might be wrongly erased when osc is evicted by OST
2858 Details : when the import is evicted by server, it will fork another
2859 thread ptlrpc_invalidate_import_thread to invalidate the
2860 import, where the grant will be set to 0. While the original
2861 thread will update the grant it got when connecting. So if
2862 the former happened latter, the grant will be wrongly errased
2863 because of this race.
2868 Description: Checking Stale with correct fid
2869 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
2870 to check whether it is stale, because sometimes, we want the
2871 enqueue happened anyway, and op_data.fid2 will not be initialized.
2873 Severity : enhancement
2875 Description: update patchless client
2876 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
2879 Frequency : only with 2.4 kernel
2881 Description: random memory corruption
2882 Details : size of struct ll_inode_info is to big for union inode.u and this
2883 can be cause of random memory corruption.
2888 Description: Memory leak in recovery
2889 Details : Lov_mds_md was not free in an error handler in mds_create_object.
2890 It should also check obd_fail before fsfilt_start, otherwise if
2891 fsfilt_start return -EROFS,(failover mds during mds recovery).
2892 then the req will return with repmsg->transno = 0 and rc = EROFS.
2893 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
2894 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
2895 be freed no matter whether fsfilt_commit success or not.
2898 Frequency : only with huge count clients
2900 Description: Prevents from taking the superblock lock in llap_from_page for
2902 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
2903 for llap_from_page call in ll_removepage prevents from taking the
2904 superblock lock for a soon died page.
2909 Description: Not check open intent error before release open handle
2910 Details : in some rare cases, the open intent error is not checked before
2911 release open handle, which may cause
2912 ASSERTION(open_req->rq_transno != 0), because it tries to release
2913 the failed open handle.
2918 Description: Set cat log bitmap only after create log success.
2919 Details : in some rare cases, the cat log bitmap is set too early. and it
2920 should be set only after create log success.
2924 Description: Accessing a block bevice can re-enable I/O when Lustre is
2925 tearing down a device.
2926 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
2931 Description: service threads may hog cpus when there are a lot of requests
2932 Details : Insert cond_resched to give other threads a chance to use some CPU
2937 Description: the cat log was not initialized in recovery
2938 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
2939 unlink log on mds will not be initialized until mds post
2940 recovery. And also in mds post recovery, the unlink log will
2941 initialization will be done asynchronausly, so there will be race
2942 between add unlink log and unlink log initialization.
2946 Description: brw_stats were being printed incorrectly
2947 Details : brw_stats were being printed as log2 but all of them were not
2948 recorded as log2. Also remove some code duplication arising from
2949 filter_tally_{read,write}.
2953 Frequency : rare, only in recovery.
2954 Description: ASSERTION(req->rq_type != LI_POISON) failed
2955 Details : imp_lock should be held while iterating over imp_sending_list for
2956 prevent destroy request after get timeout in ptlrpc_queue_wait.
2960 Description: replay-single.sh test 52 fails
2961 Details : A lock's skiplist need to be cleanup when it being unlinked
2962 from its resource list.
2966 Description: Short directio read returns full requested size rather than
2968 Details : Direct I/O operations should return actual amount of bytes
2969 transferred rather than requested size.
2971 Severity : enhancement
2973 Description: metadata RPC reduction (e.g. for rm performance)
2974 Details : decrease the amount of synchronous RPC between clients and servers
2975 by canceling conflicing lock before the operation on the client
2976 and packing thier handles into the main operation RPC to server.
2978 Severity : enhancement
2980 Description: Async OSC create to avoid the blocking unnecessarily.
2981 Details : If a OST has no remain object, system will block on the creating
2982 when need to create a new object on this OST. Now, ways use
2983 pre-created objects when available, instead of blocking on an
2984 empty osc while others are not empty. If we must block, we block
2985 for the shortest possible period of time.
2989 Description: improve handling recoverable errors
2990 Details : if request processig with error which can be recoverable on server
2991 request should be resend, otherwise page released from cache and
2994 Severity : enhancement
2996 Description: refine locking for avoid write wrong info into lov_objid file
2997 Details : fix possible races with add new target and write/update data in
3000 --------------------------------------------------------------------------------
3002 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
3007 Frequency : on some architectures
3009 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
3010 Details : Uninitialized flags sometimes cause configuration commands to
3014 Frequency : patchless clients only
3016 Description: missing __iget() symbol export
3017 Details : The __iget() symbol export is missing. To avoid the need for
3018 this on patchless clients the deathrow inode reaper is turned
3019 off, and we depend on the VM to clean up old inodes. This
3020 dependency was during via the fix for bug 12181.
3022 --------------------------------------------------------------------------------
3024 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
3026 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
3027 INTEROPERATE with older versions automatically. In many cases a
3028 special upgrade step is needed. Please read the
3029 user documentation before upgrading any part of a live system.
3030 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
3031 version will not interoperate with 1.6 betas before beta5 (1.5.95).
3032 * WARNING: Lustre configuration and startup changes are required with
3033 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
3037 Severity : enhancement
3039 Description: MountConf
3040 Details : Lustre configuration is now managed via mkfs and mount
3041 commands instead of lmc and lconf. New obd types (MGS, MGC)
3042 are added for dynamic configuration management. See
3043 https://mail.clusterfs.com/wikis/lustre/MountConf for
3046 Severity : enhancement
3048 Description: dynamic OST addition
3049 Details : OSTs can now be added to a live filesystem
3051 Severity : enhancement
3053 Description: startup order invariance
3054 Details : MDTs and OSTs can be started in any order. Clients only
3055 require the MDT to complete startup.
3057 Severity : enhancement
3059 Description: parallel, asynchronous orphan cleanup
3060 Details : orphan cleanup is now performed in separate threads for each
3061 OST, allowing parallel non-blocking operation.
3063 Severity : enhancement
3065 Description: optimized stripe assignment
3066 Details : stripe assignments are now made based on ost space available,
3067 ost previous usage, and OSS previous usage, in order to try
3068 to optimize storage space and networking resources.
3070 Severity : enhancement
3072 Description: Permanently set tunables
3073 Details : All writable /proc/fs/lustre tunables can now be permanently
3074 set on a per-server basis, at mkfs time or on a live system.
3076 Severity : enhancement
3078 Description: Lustre message v2
3079 Details : Add lustre message format v2.
3081 Severity : enhancement
3083 Description: client OST exclusion list
3084 Details : Clients can be started with a list of OSTs that should be
3085 declared "inactive" for known non-responsive OSTs.
3089 Description: SPEC SFS validation failure on NFS v2 over lustre.
3090 Details : Changes the blocksize for regular files to be 2x RPC size,
3091 and not depend on stripe size.
3093 Severity : enhancement
3095 Description: Multiple MD RPCs in flight.
3096 Details : Further unserialise some read-only MDS RPCs - learn about intents.
3097 To avoid overly-overloading MDS, introduce a limit on number of
3098 MDS RPCs in flight for a single client and add /proc controls
3099 to adjust this limit.
3101 Severity : enhancement
3103 Description: client read/write statistics
3104 Details : Add client read/write call usage stats for performance
3105 analysis of user processes.
3106 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
3107 file access. extents_stats shows chunk size distribution.
3108 extents_stats_per_process show chunk size distribution per
3111 Severity : enhancement
3113 Description: mds statistics
3114 Details : Add detailed mds operations statistics in
3115 /proc/fs/lustre/mds/*/stats.
3119 Description: Failure of copying files with lustre special EAs.
3120 Details : Client side always return success for setxattr call for lustre
3121 special xattr (currently only "trusted.lov").
3126 Description: Refcount LNET uuids
3127 Details : The global LNET uuid list grew linearly with every startup;
3128 refcount repeated list entries instead of always adding to
3131 Severity : enhancement
3133 Description: Dynamic service threads
3134 Details : Within a small range, start extra service threads
3135 automatically when the request queue builds up.
3138 Frequency : mixed-endian client/server environments
3140 Description: mixed-endian crashes
3141 Details : The new msg_v2 system had some failures in mixed-endian
3144 Severity : enhancement
3146 Description: Easy OST removal
3147 Details : OSTs can be permanently deactivated with e.g. 'lctl
3148 conf_param lustre-OST0001.osc.active=0'
3150 Severity : enhancement
3152 Description: MGS proc entries
3153 Details : Added basic proc entries for the MGS showing what filesystems
3156 Severity : enhancement
3158 Description: provide MGS failover
3159 Details : Added config lock reacquisition after MGS server failover.
3161 Severity : enhancement
3163 Description: add Linux 2.4 support
3164 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
3168 Description: a large application tries to do I/O to the same resource and dies
3169 in the middle of it.
3170 Details : Check the req->rq_arrival time after the call to
3171 ost_brw_lock_get(), but before we do anything about
3172 processing it & sending the BULK transfer request. This
3173 should help move old stale pending locks off the queue as
3174 quickly as obd_timeout.
3177 Frequency : when an incorrect nid is specified during startup
3179 Description: ptlrpc connect to non-existant node causes kernel crash
3180 Details : LNET can't be re-entered from an event callback, which
3181 happened when we expire a message after the export has been
3182 cleaned up. Instead, hand the zombie cleanup off to another
3185 Severity : enhancement
3187 Description: plain/inodebits lock performance improvement
3188 Details : Grouping plain/inodebits in granted list by their request modes
3189 and bits policy, thus improving the performance of search through
3193 Frequency : only if OST filesystem is corrupted
3195 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3196 Details : for a short time RPCs with bulk IO are in the replay list,
3197 but replay of bulk IOs is unimplemented. If the OST filesystem
3198 is corrupted due to disk cache incoherency and then replay is
3199 started it is possible to trip an assertion. Avoid putting
3200 committed RPCs into the replay list at all to avoid this issue.
3203 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3205 Description: System hang on startup
3206 Details : This bug allowed the liblustre (e.g. catamount) client to
3207 return to the app before handling all startup RPCs. This
3208 could leave the node unresponsive to lustre network traffic
3209 and manifested as a server ptllnd timeout.
3211 Severity : enhancement
3213 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3214 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
3215 boolean to control whether to print peer debug info when a
3216 client's RPC times out.
3219 Frequency : only for kernels with patches from Lustre below 1.4.3
3221 Description: Remove old rdonly API
3222 Details : Remove old rdonly API which unused from at least lustre 1.4.3
3225 Frequency : only for devices with external journals
3227 Description: Set external device read-only also
3228 Details : During a commanded failover stop, we set the disk device
3229 read-only while the server shuts down. We now also set any
3230 external journal device read-only at the same time.
3233 Frequency : when upgrading from 1.4 while trying to change parameters
3235 Description: The wrong (new) MDC name was used when setting parameters for
3236 upgraded MDT's. Also allows changing of OSC (and MDC)
3237 parameters if --writeconf is specified at tunefs upgrade time.
3240 Frequency : when setting specific ost indicies
3242 Description: QOS code breaks on skipped indicies
3243 Details : Add checks for missing OST indicies in the QOS code, so OSTs
3244 created with --index need not be sequential.
3246 Severity : enhancement
3248 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3249 Details : The uninit_groups feature works in conjunction with the kernel
3250 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3251 up the pass1 processing of e2fsck. This is a read-only feature
3252 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3253 mount filesystems that have had this feature enabled.
3255 Severity : enhancement
3257 Description: Improve multi-block allocation algorithm to avoid fragmentation
3258 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3259 allocation locality and avoid filesystem fragmentation.
3261 ------------------------------------------------------------------------------
3263 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
3265 * Support for kernels:
3266 2.4.21-47.0.1.EL (RHEL 3)
3267 2.6.5-7.283 (SLES 9)
3268 2.6.9-42.0.10.EL (RHEL 4)
3269 2.6.12.6 vanilla (kernel.org)
3270 2.6.16.27-0.9 (SLES 10)
3271 * Recommended e2fsprogs version: 1.39.cfs5
3273 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3277 Frequency : occasional, depends on client load and configuration
3278 Bugzilla : 12181, 12203
3279 Description: data loss for recently-modified files
3281 Details : In some cases it is possible that recently written or created
3282 files may not be written to disk in a timely manner (this should
3283 normally be within 30s unless client IO load is very high).
3284 The problem appears as zero-length files or files that are a
3285 multiple of 1MB in size after a client crash or client eviction
3286 that are missing data at the end of the file.
3288 This problem is more likely to be hit on clients where files are
3289 repeatedly created and unlinked in the same directory, clients
3290 have a large amount of RAM, have many CPUs, the filesystem has
3291 many OSTs, the clients are rebooted frequently, and/or the files
3292 are not accessed by other nodes after being written.
3294 The presence of the problem can be detected by looking at
3295 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
3296 smaller than the second (nr_unused) then dirty files will not
3297 be flushed automatically to disk. "sync; sleep 10" should be
3298 run several times on the node before unmounting it to update
3299 Lustre (this is also safe to run on nodes without this problem).
3301 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
3302 that can cause this same problem, so customers using that kernel
3303 also need to update the kernel in addition to Lustre. In order
3304 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
3306 It is normal that files written just before a client crash (less
3307 than 30s) may not yet have been flushed to disk, even for local
3311 Frequency : frequent on thin XT3 nodes
3313 Description: UUID collision on thin XT3 Linux nodes
3314 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
3315 randomly, since we relied on an insufficiently-seeded PRNG.
3320 Description: OSS hangs after "All ost request buffers busy"
3321 Details : A deadlock between quota and journal operations caused OSS
3322 hangs after printing "All ost request buffers busy."
3325 Frequency : always on liblustre builds
3327 Description: Cleanup compiler warnings on liblustre
3330 Frequency : always on liblustre builds on XT3
3332 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
3336 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
3337 > 32000 subdirectories
3338 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
3339 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
3340 subdirectory count crosses 32000. This will aid e2fsck to
3341 correctly handle more than 32000 subdirectories.
3344 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3346 Description: System hang on startup
3347 Details : This bug allowed the liblustre (e.g. catamount) client to
3348 return to the app before handling all startup RPCs. This
3349 could leave the node unresponsive to lustre network traffic
3350 and manifested as a server ptllnd timeout.
3352 Severity : enhancement
3354 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3355 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
3356 boolean to control whether to print peer debug info when a
3357 client's RPC times out.
3362 Description: make O_SYNC working on 2.6 kernels
3363 Details : 2.6 kernels use different method for mark pages for write,
3364 so need add a code to lustre for O_SYNC work.
3369 Description: Failure to close file and release space on NFS
3370 Details : Put inode details into lock acquired in ll_intent_file_open.
3371 Use mdc_intent_lock in ll_intent_open to properly
3372 detect all kind of errors unhandled by mdc_enqueue.
3377 Description: proc file read during shutdown sometimes raced obd removal,
3379 Details : Add lock to prevent obd access after proc file removal.
3382 Frequency : Only for files larger than 4GB on 32-bit clients.
3384 Description: improperly doing page alignment of locks
3385 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
3386 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3391 Description: RPCs being resent when they shouldn't be
3392 Details : Some RPCs that should not be resent are being resent. This
3393 can cause inconsistencies in the RPC state machine. Do not
3394 resend such requests.
3397 Frequency : rare, only with NFS export
3399 Description: Crash on NFS re-export node
3400 Details : under very unusual load conditions an assertion is hit in
3401 ll_intent_file_open()
3404 Frequency : only if OST filesystem is corrupted
3406 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3407 Details : for a short time RPCs with bulk IO are in the replay list,
3408 but replay of bulk IOs is unimplemented. If the OST filesystem
3409 is corrupted due to disk cache incoherency and then replay is
3410 started it is possible to trip an assertion. Avoid putting
3411 committed RPCs into the replay list at all to avoid this issue.
3416 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3417 Details : Large single O_DIRECT read and write calls can fail to allocate
3418 a sufficiently large buffer to process the request. In case of
3419 allocation failure the allocation is retried with a smaller
3420 buffer and broken into smaller requests.
3422 Severity : enhancement
3424 Description: Add -o localflock option to simulate old noflock behaviour.
3425 Details : This will achieve local-only flock/fcntl locks coherentness.
3430 Description: versioning check is incomplete
3431 Details : Checking the version difference of client vs. server, report
3432 error if the gap is too big.
3437 Description: add support PG_writeback bit
3438 Details : add support for PG_writeback bit for Lustre, for more carefull
3439 work with page cache in 2.6 kernel. This also fix some deadlocks
3440 and remove hack for work O_SYNC with 2.6 kernel.
3442 Severity : enhancement
3444 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3445 Details : The uninit_groups feature works in conjunction with the kernel
3446 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3447 up the pass1 processing of e2fsck. This is a read-only feature
3448 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3449 mount filesystems that have had this feature enabled.
3451 Severity : enhancement
3453 Description: Improve multi-block allocation algorithm to avoid fragmentation
3454 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3455 allocation locality and avoid filesystem fragmentation.
3457 ------------------------------------------------------------------------------
3459 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3461 * Support for kernels:
3462 2.6.9-42.0.3.EL (RHEL 4)
3463 2.6.5-7.276 (SLES 9)
3464 2.4.21-47.0.1.EL (RHEL 3)
3465 2.6.12.6 vanilla (kernel.org)
3466 2.6.16.21-0.8 (SLES10)
3467 * Recommended e2fsprogs version: 1.39.cfs2-0
3469 * The backwards-compatible /proc/sys/portals symlink has been removed
3470 in this release. Before upgrading, please ensure that you change
3471 any configuration scripts or /etc/sysctl.conf files that access
3472 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3473 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3474 in advance of the upgrade on any system running Lustre 1.4.6 or
3475 newer, since /proc/sys/lnet was added in that version.
3476 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3480 Frequency : only when quota is used
3482 Description: avoid scanning export list for quota master
3483 Details : Change the algorithms to avoid scanning export list in order
3484 to improve the efficiency.
3487 Frequency : MDS failover only, very rarely
3489 Description: "went back in time" messages on mds failover
3490 Details : The greatest transno may be lost when the current operation
3491 finishes with an error (transno==0) and the client's last_rcvd
3492 record is over-written. Save the greatest transno in the
3493 mds_last_transno for this case.
3496 Frequency : always for specific kernels and striping counts
3498 Description: client may get "Matching packet too big" without ACL support
3499 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
3500 when trying to access files in certain configurations. The
3501 clients should in fact be denied when mounting because they do
3502 not understand ACLs.
3505 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
3507 Description: many clients connecting with IO in progress causes connect timeouts
3508 Details : Avoid synchronous journal commits to avoid delays caused by many
3509 clients connecting/disconnecting when bulk IO is in progress.
3510 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
3511 OST_IO_PORTAL to avoid delays behind potentially many pending
3515 Frequency : occasionally with multiple writers to a single file
3517 Description: shared writes to file may result in wrong size reported by stat()
3518 Details : Allow growing of kms when extent lock is cancelled
3521 Frequency : always with random mmap IO to multi-striped file
3523 Description: mmap write might be lost if we are writing to a 'hole' in stripe
3524 Details : Only if the hole is at the end of OST object so that kms is too
3525 small. Fix is to increase kms accordingly in ll_nopage.
3528 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
3530 Description: writes to a missing object would leak memory on the OST
3531 Details : If there is an inconsistency between the MDS and OST filesystems,
3532 such that the MDS references an object that doesn't exist, writes
3533 to that object will leak memory due to incorrect cleanup in the
3534 error handling path, eventually running out of memory on the OST.
3539 Description: Creating too long symlink causes lustre errors
3540 Details : Check symlink and name lengths before sending requests to MDS.
3543 Frequency : only if flock is enabled (not on by default)
3545 Description: posix locks not released on fd closure on 2.6.9+
3546 Details : We failed to add posix locks to list of inode locks on 2.6.9+
3547 kernels, this caused such locks not to be released on fd close and
3548 then assertions on fs unmount about still used locks.
3551 Frequency : MDS failover only, very rarely
3553 Description: clients may get ASSERTION(granted_lock != NULL)
3554 Details : When request was taking a long time, and a client was resending
3555 a getattr by name lock request. The were multiple lock requests
3556 with the same client lock handle and
3557 mds_getattr_name->fixup_handle_for_resent_request found one of the
3558 lock handles but later failed with ASSERTION(granted_lock != NULL).
3563 Description: handle->h_buffer_credits > 0, assertion failure
3564 Details : h_buffer_credits is zero after truncate, causing assertion
3565 failure. This patch extends the transaction or creates a new
3569 Frequency : NFS re-export or patchless client
3570 Bugzilla : 11179, 10796
3571 Description: Crash on NFS re-export node (__d_move)
3572 Details : We do not want to hash the dentry if we don't have a lock.
3573 But if this dentry is later used in d_move, we'd hit uninitialised
3574 list head d_hash, so we just do this to init d_hash field but
3575 leave dentry unhashed.
3578 Frequency : NFS re-export or patchless client
3580 Description: NFS exports has problem with symbolic link
3581 Details : lustre client didn't properly install dentry when re-exported
3582 to NFS or running patchless client.
3585 Frequency : NFS re-export or patchless client
3587 Description: Various nfs/patchless fixes.
3588 Details : fixes reuse disconected alias for lookup process - this fixes
3589 warning "find_exported_dentry: npd != pd",
3590 fix permission error with open files at nfs.
3591 fix apply umask when do revalidate.
3594 Frequency : occasional
3596 Description: Crash on NFS re-export node
3597 Details : calling clear_page() on the wrong pointer triggered oops in
3598 generic_mapping_read().
3601 Frequency : rarely, using O_DIRECT IO
3603 Description: unaligned directio crashes client with LASSERT
3604 Details : check for unaligned buffers before trying any requests.
3607 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
3609 Description: stale data returned from RAID cache
3610 Details : If only a small amount of IO is done to the RAID device before
3611 reading it again it is possible to get stale data from the RAID
3612 cache instead of reading it from disk.
3615 Frequency : always for sles10 kernel
3617 Description: sles10 support
3618 Details : ll_follow_link: compile fixes and using of nd_set_link
3619 under newer kernels.
3622 Frequency : depends on arch, kernel and compiler version, always on sles10
3625 Description: recursive or deep enough symlinks cause stack overflow
3626 Details : getting rid of large stack-allocated variable in
3630 Frequency : depends on hardware
3632 Description: lustre write performance loss in the SLES10 kernel
3633 Details : the performance loss is caused by using of write barriers in the
3634 ext3 code. The SLES10 kernel turns barrier support on by
3635 default. The fix is to undo that change for ldiskfs.
3637 ------------------------------------------------------------------------------
3639 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
3641 * Support for kernels:
3642 2.6.9-42.0.3EL (RHEL 4)
3643 2.6.5-7.276 (SLES 9)
3644 2.4.21-47.0.1.EL (RHEL 3)
3645 2.6.12.6 vanilla (kernel.org)
3649 Frequency : quota enabled and large files being deleted
3651 Description: releasing more than 4GB of quota at once hangs OST
3652 Details : If a user deletes more than 4GB of files on a single OST it
3653 will cause the OST to spin in an infinite loop. Release
3654 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3659 Description: statfs data retrieved from /proc may be stale or zero
3660 Details : When reading per-device statfs data from /proc, in the
3661 {kbytes,files}_{total,free,avail} files, it may appear
3662 as zero or be out of date.
3665 Frequency : systems with MD RAID1 external journal devices
3667 Description: lconf's call to blkid is confused by RAID1 journal devices
3668 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3669 of returning all block devices that match the journal UUID.
3672 Frequency : always, for aggregate stripe size over 4GB
3674 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
3675 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
3676 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
3679 Frequency : always if "lfs find" used on a local file/directory
3681 Description: "lfs find" segfaults if used on a local file/directory
3682 Details : The case where a directory component was not specified wasn't
3683 handled correctly. Handle this properly.
3686 Frequency : always on ppc64
3688 Description: the write to an ext3 filesystem mounted with mballoc got stuck
3689 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3690 perform endianness conversion.
3693 Frequency : rarely (truncate to non-zero file size after write under load)
3694 Bugzilla : 10730, 10687
3695 Description: Files padded with zeros to next 4K multiple
3696 Details : With filesystems mounted using the "extents" option (2.6 kernels)
3697 it is possible that files that are truncated to a non-zero size
3698 immediately after being written are filled with zero bytes beyond
3699 the truncated size. No file data is lost.
3701 Severity : enhancement
3703 Description: Allow recovery/failover for liblustre clients.
3704 Details : liblustre clients were unaware of failover configurations until
3707 Severity : enhancement
3709 Description: user file locks should fail when not mounting with flock option
3710 Details : Set up an error-returning stub in ll_file_operations.lock field
3711 to prevent incorrect behaviour when client is mounted without
3712 flock option. Also, set up properly f_op->flock field for
3716 Frequency : always on ia64
3718 Description: "lfs df" loops on printing out MDS statfs information
3719 Details : The obd_ioctl_data was not initialized and in some systems
3720 this caused a failure during the ioctl that did not return
3721 an error. Initialize the struct and return an error on failure.
3724 Frequency : SLES 9 only
3726 Description: Error of copying files with lustre special EAs as root
3727 Details : Client side always return success for setxattr call for lustre
3728 special xattr (currently only "trusted.lov").
3731 Frequency : rarely on clusters with both ia64+i386 clients
3733 Description: ia64+i686 clients doing shared IO on the same file may LBUG
3734 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
3735 clients are doing concurrent writes to the same file it is
3736 possible that the ia64 clients may LASSERT because the OST
3737 extent locks are not PAGE_SIZE aligned. Ensure that grown
3738 locks are always aligned on the request boundary.
3741 Frequency : specific use, occasional
3743 Description: Overwriting in use executable truncates on-disk binary image
3744 Details : If one node attempts to overwrite an executable in use by
3745 another node, we now correctly return ETXTBSY instead of
3746 truncating the file.
3748 Severity : enhancement
3750 Description: Async OSC create to avoid the blocking unnecessarily.
3751 Details : If a OST has no remain object, system will block on the creating
3752 when need to create a new object on this OST. Now, ways use
3753 pre-created objects when available, instead of blocking on an
3754 empty osc while others are not empty. If we must block, we block
3755 for the shortest possible period of time.
3760 Description: chmod on Lustre root is propagated to other clients
3761 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
3762 invalid by the follow_mount time.
3765 Frequency : liblustre clients only
3767 Description: Race in 'instant cancel' lock handling could lead to such locks
3768 never to be granted in case of SMP MDS
3769 Details : Do not destroy not yet granted but cbpending locks in
3773 Frequency : replay/resend of open
3775 Description: non null lock assetion failure in mds_intent_policy
3776 Details : Trying to replay/resend lockless open requests resulted in
3777 mds_open() returning 0 with no lock. Now it sets a flag if
3778 a lock is going to be returned.
3780 Severity : enhancement
3782 Description: Checksum enhancements
3783 Details : New checksum enhancements allow for resending RPCs that failed
3786 Severity : enhancement
3788 Description: Tunables on number of dirty pages in cacche
3789 Details : Allow to set limit on number of dirty pages cached.
3794 Description: client crash on unmount - lock still has references
3795 Details : In some error handling cases it was possible to leak a lock
3796 reference on a client while accessing a file. This was not
3797 harmful to the client during operation, but would cause the
3798 client to crash when the filesystem is unmounted.
3801 Frequency : specific case, rare
3803 Description: ETXTBSY on mds though file not in use by client
3804 Details : ETXTBSY is no longer incorrectly returned when attempting to
3805 chmod or chown a directory that the user previously tried to
3806 execute or a currently-executing binary.
3809 Frequency : extremely rare except on liblustre-based clients
3811 Description: Lustre space not freed when files are deleted
3812 Details : Clean up open-unlinked files after client eviction. Previously
3813 the unlink was skipped and the files remained as orphans.
3818 Description: OST failure "would be an LBUG" in waiting_locks_callback()
3819 Details : In some cases it was possible to send a blocking callback to
3820 a client doing a glimpse, even though that client didn't get
3821 a lock granted. When the glimpse lock is cancelled on the OST
3822 the freed lock is left on the waiting list and corrupted the list.
3825 Frequency : all core dumps
3827 Description: Broke core dumps to lustre
3828 Details : Negative dentry may be unhashed if parent does not have UPDATE
3829 lock, but some callers, e.g. do_coredump, expect dentry to be
3830 hashed after successful create, hash it in ll_create_it.
3832 ------------------------------------------------------------------------------
3834 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
3836 * Support for kernels:
3837 2.6.9-42.0.2.EL (RHEL 4)
3838 2.6.5-7.276 (SLES 9)
3839 2.4.21-40.EL (RHEL 3)
3840 2.6.12.6 vanilla (kernel.org)
3844 Frequency : always on RHEL 3
3846 Description: Number of open files grows over time
3847 Details : The number of open files grows over time, whether or not
3848 Lustre is started. This was due to a filp leak introduced
3849 by one of our kernel patches.
3851 ------------------------------------------------------------------------------
3853 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
3855 * Support for kernels:
3856 2.6.9-42.EL (RHEL 4)
3857 2.6.5-7.276 (SLES 9)
3858 2.4.21-40.EL (RHEL 3)
3859 2.6.12.6 vanilla (kernel.org)
3864 Bugzilla : 5719, 9635, 9792, 9684
3865 Description: OST (or MDS) trips assertions in (re)connection under heavy load
3866 Details : If a server is under heavy load and cannot reply to new
3867 connection requests before the client resends the (re)connect,
3868 the connection handling code can behave badly if two service
3869 threads are concurrently handing separate (re)connections from
3870 the same client. Add better locking to the connection handling
3871 code, and ensure that only a single connection will be processed
3872 for a given client UUID, even if the lock is dropped.
3874 Severity : enhancement
3876 Description: add TCP zero-copy support to kernel
3877 Details : Add support to the kernel TCP stack to allow zero-copy bulk
3878 sends if the hardware supports scatter-gather and checksumming.
3879 This allows socklnd to do client-write and server-read more
3880 efficiently and reduce CPU utilization from skbuf copying.
3883 Frequency : only if NFS exporting from client
3885 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
3886 Details : If setting attributes on a file created under NFS that had
3887 never been opened it would be possible to oops the client
3888 if the file had no objects.
3891 Frequency : always for liblustre
3893 Description: liblustre client does MDS+OSTs setattr RPC for each write
3894 Details : When doing a write from a liblustre client, the client
3895 incorrectly issued an RPC to the MDS and each OST the file was
3896 striped over in order to update the timestamps. When writing
3897 with small chunks and many clients this could overwhelm the MDS
3898 with RPCs. In all cases it would slow down the write because
3899 these RPCs are unnecessary.
3901 Severity : enhancement
3903 Description: allow number of MDS service threads to be changed at module load
3904 Details : It is now possible to change the number of MDS service threads
3905 running. Adding "options mds mds_num_threads={N}" to the MDS's
3906 /etc/modprobe.conf will set the number of threads for the next
3907 time Lustre is restarted (assuming the "mds" module is also
3908 reloaded at that time). The default number of threads will
3909 stay the same, 32 for most systems.
3914 Description: OST crash if filesystem is unformatted or corrupt
3915 Details : If an OST is started on a device that has never been formatted
3916 or if the filesystem is corrupt and cannot even mount then the
3917 error handling cleanup routines would dereference a NULL pointer.
3922 Description: NULL pointer deref in llap_from_page.
3923 Details : get_cache_page_nowait can return a page with NULL (or otherwise
3924 incorrect) mapping if the page was truncated/reclaimed while it was
3925 searched for. Check for this condition and skip such pages when
3926 doing readahead. Introduce extra check to llap_from_page() to
3927 verify page->mapping->host is non-NULL (so page is not anonymous).
3930 Frequency : Sometimes when using sys_sendfile
3932 Description: "page not covered by a lock" warnings from ll_readpage
3933 Details : sendfile called ll_readpage without right page locks present.
3934 Now we introduced ll_file_sendfile that does necessary locking
3935 around call to generic_file_sendfile() much like we do in
3939 Frequency : with certain MDS communication failures at client mount time
3941 Description: NULL pointer deref after failed client mount
3942 Details : a client connection request may delayed by the network layer
3943 and not be sent until after the PTLRPC layer has timed out the
3944 request. If the client fails the mount immediately it will try
3945 to clean up before the network times out the request. Add a
3946 reference from the request import to the obd device and delay
3947 the cleanup until the network drops the request.
3950 Frequency : occasionally during client (re)connect
3952 Description: assertion failure during client (re)connect
3953 Details : processing a client connection request may be delayed by the
3954 client or server longer than the client connect timeout. This
3955 causes the client to resend the connection request. If the
3956 original connection request is replied in this interval, the
3957 client may trip an assertion failure in ptlrpc_connect_interpret()
3958 which thought it would be the only running connect process.
3961 Frequency : only with obd_echo servers and clients that are rebooted
3963 Description: kernel BUG accessing uninitialized data structure
3964 Details : When running an obd_echo server it did not start the ping_evictor
3965 thread, and when a client was evicted an uninitialized data
3966 structure was accessed. Start the ping_evictor in the RPC
3967 service startup instead of the OBD startup.
3969 Severity : enhancement
3970 Bugzilla : 10193 (patchless)
3971 Description: Remove dependency on various unexported kernel interfaces.
3972 Details : No longer need reparent_to_init, exit_mm, exit_files,
3973 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
3976 Frequency : rare (only users of deprecated and unsupported LDAP config)
3978 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
3979 Details : LDAP apparently contains 'lustreName' attributes instead of
3980 'name'. A simple remapping of the name is sufficient.
3983 Frequency : rare (only with non-default dump_on_timeout debug enabled)
3985 Description: waiting_locks_callback trips kernel BUG if client is evicted
3986 Details : Running with the dump_on_timeout debug flag turned on makes
3987 it possible that the waiting_locks_callback() can try to dump
3988 the Lustre kernel debug logs from an interrupt handler. Defer
3989 this log dumping to the expired_lock_main() thread.
3991 Severity : enhancement
3993 Description: Support NFS exporting on 2.6 kernels.
3994 Details : Implement non-rawops metadata methods for NFS server to use without
3995 changing NFS server code.
3998 Frequency : very rare (synthetic metadata workload only)
4000 Description: two racing renames might cause an MDS thread to deadlock
4001 Details : Running the "racer" program may cause one MDS thread to rename
4002 a file from being the source of a rename to being the target of
4003 a rename at exactly the same time that another thread is doing
4004 so, and the second thread has already enqueued these locks after
4005 doing a lookup of the target and is trying to relock them in
4006 order. Ensure that we don't try to re-lock the same resource.
4009 Frequency : only very large systems with liblustre clients
4011 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
4012 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
4013 all OSTs in parallel. This allows the eviction of stale liblustre
4014 clients to proceed much faster than if they were done in series,
4015 and also offers similar improvements for other set_info RPCs.
4020 Description: excessive CPU usage during initial read phase on client
4021 Details : During the initial read phase on a client, it would agressively
4022 retry readahead on the file, consuming too much CPU and impacting
4023 performance (since 1.4.5.8). Improve the readahead algorithm
4024 to avoid this, and also improve some other common cases (read
4025 of small files in particular, where "small" is files smaller than
4026 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
4031 Description: MDS crash when receiving packet with unknown intent.
4032 Details : Do not LBUG in unknown intent case, just return -EFAULT
4034 Severity : enhancement
4035 Bugzilla : 9293, 9385
4036 Description: MDS RPCs are serialised on client. This is unnecessary for some.
4037 Details : Do not serialize getattr (non-intent version) and statfs.
4040 Frequency : occasional, when OST network is overloaded/intermittent
4042 Description: client evicted by OST after bulk IO timeout
4043 Details : If a client sends a bulk IO request (read or write) the OST
4044 may evict the client if it is unresposive to its data GET/PUT
4045 request. This is incorrect if the network is overloaded (takes
4046 too long to transfer the RPC data) or dropped the OST GET/PUT
4047 request. There is no need to evict the client at all, since
4048 the pinger and/or lock callbacks will handle this, and the
4049 client can restart the bulk request.
4052 Frequency : Always when mmapping file with no objects
4054 Description: client crashes when mmapping file with no objects
4055 Details : Check that we actually have objects in a file before doing any
4056 operations on objects in ll_vm_open, ll_vm_close and
4062 Description: Request leak when working with deleted CWD
4063 Details : Introduce advanced request refcount tracking for requests
4064 referenced from lustre intent.
4066 Severity : Enhancement
4068 Description: Cache open file handles on client.
4069 Details : MDS now will return special lock along with openhandle, if
4070 requested and client is allowed to hold openhandle, even if unused,
4071 until such a lock is revoked. Helps NFS a lot, since NFS is opening
4072 closing files for every read/write openration.
4074 Severity : Enhancement
4076 Description: Cache open negative dentries on client when possible.
4077 Details : Guard negative dentries with UPDATE lock on parent dir, drop
4078 negative dentries on lock revocation.
4083 Description: Remounting a client read-only wasn't possible with a zconf mount
4084 Details : It wasn't possible to remount a client read-only with llmount.
4086 Severity : enhancement
4087 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
4088 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
4089 MPICH is included to add Lustre-specific ADIO interfaces.
4090 This is based closely on the UFS ADIO layer and only differs
4091 in file creation, in order to allow the OST striping to be set.
4092 This is user-contributed code and not supported by CFS.
4097 Description: extended inode attributes (immutable, append-only) work improperly
4098 when 2.4 and 2.6 kernels are used on client/server or vice versa
4099 Details : Introduce kernel-independent values for these flags.
4101 Severity : enhancement
4104 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
4105 Details : Many of the /proc/ tunables can only be tuned at a megabyte
4106 granularity. Now, Fractional MB granularity is be supported,
4107 this is very useful for low memory system.
4109 Severity : enhancement
4111 Description: Getattr by fid
4112 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
4113 locks, avoids extra getattr rpc requests to MDS, allows '/' to
4114 have locks and avoids getattr rpc requests for it on every stat.
4117 Frequency : Always, for filesystems larger than 2TB
4119 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4120 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
4121 filesystems larger than 2TB to be mounted with mballoc enabled.
4124 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
4126 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4127 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
4128 filesystems larger than 2TB is mounted then the kernel will
4129 silently corrupt the start of the filesystem. CONFIG_LBD is
4130 enabled for all CFS-supported kernels, but the possibility of
4131 this happening with a modified kernel config exists.
4133 Severity : enhancement
4135 Description: add client O_DIRECT support for 2.6 kernels
4136 Details : It is now possible to do O_DIRECT reads and writes to files
4137 in the Lustre client mountpoint on 2.6 kernel clients.
4139 Severity : enhancement
4141 Description: parallel glimpse, setattr, statfs, punch, destroy requests
4142 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
4143 parallel, not waiting for response from every OST before sending
4144 a rpc to the next OST.
4149 Description: setattr vs write race when updating file timestamps
4150 Details : Client processes that update a file timestamp into the past
4151 right after writing to the file (e.g. tar) it is possible that
4152 the updated file modification time can be reset to the current
4153 time due to a race between processing the setattr and write RPC.
4155 Severity : enhancement
4157 Description: Bring 'lfs find' closer in line with regular Linux find.
4158 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
4159 -print0 options and obtains all the needed info through the lustre
4162 Severity : enhancement
4164 Description: support up to 1024 configured devices on one node
4165 Details : change obd_dev array from statically allocated to dynamically
4166 allocated structs as they are first used to reduce memory usage
4171 Description: Flush dirty partially truncated pages during truncate
4172 Details : Immediatelly flush partially truncated pages in filter_setattr,
4173 this way we completely avoid having any pages in page cache on OST
4174 and can retire ugly workarounds during writes to flush such pages.
4179 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
4180 Details : protect inode from truncation within vfs_unlink() context
4181 just take a reference before calling vfs_unlink() and release it
4182 when parent's i_sem is free.
4187 Description: last_id value checked outside lock on OST caused LASSERT failure
4188 Details : If there were multiple MDS->OST object precreate requests in
4189 flight, it was possible that the OST's last object id was checked
4190 outside a lock and incorrectly tripped an assertion. Move checks
4191 inside locks, and discard old precreate requests.
4194 Frequency : always, if extents are used on OSTs
4196 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
4197 Details : OSTs using the extents format would not zero the high 16 bits of
4198 the index physical block number. This is not a problem for any
4199 OST filesystems smaller than 16TB, and no kernels support ext3
4200 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
4201 new/modified files) and can be fixed for existing filesystems
4202 with e2fsprogs-1.39-cfs1.
4207 Description: import connection selection may be incorrect if timer wraps
4208 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
4209 import connections to be ignored if the 32-bit jiffies counter
4210 wraps. Use a 64-bit jiffies counter.
4213 Frequency : very large clusters immediately after boot
4215 Description: LNET request buffers exhausted under heavy short-term load
4216 Details : If a large number of client requests are generated on a service
4217 that has previously never seen so many requests it is possible
4218 that the request buffer growth cannot keep up with the spike in
4219 demand. Instead of dropping incoming requests, they are held in
4220 the LND until the RPC service can accept more requests.
4223 Frequency : Sometimes during replay
4225 Description: Assertion failure in ll_local_open after replay.
4226 Details : If replay happened on an open request reply before we were able
4227 to set replay handler, reply will become not swabbed tripping the
4228 assertion in ll_local_open. Now we set the handler right after
4229 recognising of open request
4232 Frequency : very rare
4234 Description: kernel reports "badness in vsnprintf"
4235 Details : Reading from the "recovery_status" /proc file in small chunks
4236 may cause a negative length in lprocfs_obd_rd_recovery_status()
4237 call to vsnprintf() (which is otherwise harmless). Exit early
4238 if there is no more space in the output buffer.
4240 Severity : enhancement
4242 Description: clear OBD RPC statistics by writing to them
4243 Details : It is now possible to clear the OBD RPC statistics by writing
4244 to the "stats" file.
4249 Description: Client mtime is not the same on different clients after utimes
4250 Details : In some cases, the client was using the utimes() syscall on
4251 a file cached on another node. The clients now validate the
4252 ctime from the MDS + OSTs to determine which one is right.
4257 Description: Inability to activate failout mode
4258 Details : lconf script incorrectly assumed that in pythong string's numeric
4259 value is used in comparisons.
4262 Frequency : always with multiple stripes per file
4264 Description: Inefficient object allocation for mutli-stripe files
4265 Details : When selecting which OSTs to stripe files over, for files with
4266 a stripe count that divides evenly into the number of OSTs,
4267 the MDS is always picking the same starting OST for each file.
4268 Return the OST selection heuristic to the original design.
4273 Description: mount failures may take full timeout to return an error
4274 Details : Under some heavy load conditions it is possible that a
4275 failed mount can wait for the full obd_timeout interval,
4276 possibly several minutes, before reporting an error.
4277 Instead return an error as soon as the status is known.
4279 Frequency : quota enabled and large files being deleted
4281 Description: releasing more than 4GB of quota at once hangs OST
4282 Details : If a user deletes more than 4GB of files on a single OST it
4283 will cause the OST to spin in an infinite loop. Release
4284 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4289 Description: statfs data retrieved from /proc may be stale or zero
4290 Details : When reading per-device statfs data from /proc, in the
4291 {kbytes,files}_{total,free,avail} files, it may appear
4292 as zero or be out of date.
4295 Frequency : systems with MD RAID1 external journal devices
4297 Description: lconf's call to blkid is confused by RAID1 journal devices
4298 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4299 of returning all block devices that match the journal UUID.
4302 Frequency : always, for aggregate stripe size over 4GB
4304 Description: assertion fails when trying to use 4GB stripe size
4305 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
4306 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
4309 Frequency : always on ppc64
4311 Description: the first write on an ext3 filesystem with mballoc got stuck
4312 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4313 perform endianness conversion.
4315 ------------------------------------------------------------------------------
4317 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4319 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
4320 INTEROPERATE with older versions automatically. Please read the
4321 user documentation before upgrading any part of a live system.
4322 * WARNING: Lustre networking configuration changes are required with
4323 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
4326 * Support for kernels:
4327 2.6.9-22.0.2.EL (RHEL 4)
4328 2.6.5-7.244 (SLES 9)
4329 2.6.12.6 vanilla (kernel.org)
4332 Severity : enhancement
4333 Bugzilla : 7981/8208
4334 Description: Introduced Lustre Networking (LNET)
4335 Details : LNET is new networking infrastructure for Lustre, it includes
4336 a reorganized network configuration mode (see the user
4337 documentation for full details) as well as support for routing
4338 between different network fabrics. Lustre Networking Devices
4339 (LNDs) for the supported network fabrics have also been
4340 created for this new infrastructure.
4342 Severity : enhancement
4343 Description: Introduced Access control lists
4344 Details : clients can set ACLs on files and directories in order to have
4345 more fine-grained permissions than the standard Unix UGO+RWX.
4346 The MDS must be started with the "-o acl" mount option.
4348 Severity : enhancement
4349 Description: Introduced filesystem quotas
4350 Details : Administrators may now establish per-user quotas on the
4353 Severity : enhancement
4355 Description: Configuration change for the XT3
4356 The PTLLND is now used to run Lustre over Portals on the XT3
4357 The configure option(s) --with-cray-portals are no longer used.
4358 Rather --with-portals=<path-to-portals-includes> is used to
4359 enable building on the XT3. In addition to enable XT3 specific
4360 features the option --enable-cray-xt3 must be used.
4365 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
4366 Details : A race between allocating a new llcd and re-getting the llcd_lock
4367 allowed another thread to grab newly-allocated llcd.
4369 Severity : enhancement
4371 Description: 2.6 OST async journal commit and locking fix to improve performance
4372 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
4373 2.6 kernels are now async as they already were in 2.4 kernels so
4374 that they can commit concurrently with the network bulk transfer.
4375 For block-allocated files the filter allocation semaphore is held
4376 to avoid filesystem fragmentation during allocation. BKL lock
4377 removed for 2.6 xattr operations where it is no longer needed.
4382 Description: lconf incorrectly determined whether two IP networks could talk
4383 Details : In some more complicated routing and multiple-network
4384 configurations, lconf will avoid trying to make a network
4385 connection to a disjoint part of the IP space. It was doing the
4386 math incorrectly for one set of cases.
4391 Description: Fix for potential infinite loop processing records in an llog.
4392 Details : If an llog record is corrupted/zeroed, it is possible to loop
4393 forever in llog_process(). Validate the llog record length
4394 and skip the remainder of the block on error.
4397 Frequency : occasional (liblustre only)
4399 Description: liblustre could not open files whose last component is a symlink
4400 Details : sysio_path_walk() would incorrectly pass the open intent to
4401 intermediate path components.
4404 Frequency : rare (liblustre only with non-standard tuning)
4405 Bugzilla : 7201 (7350)
4406 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4407 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4408 after it has been released, only do this for referenced locks
4410 Severity : enhancement
4412 Description: specifying an (invalid) directory default stripe_size of -1
4413 would reset the directory default striping
4414 Details : stripe_size -1 was used internally to signal directory stripe
4415 removal, now use "all default" to signal dir stripe removal
4416 as a directory striping of "all default" is not useful
4419 Frequency : common for large clusters running liblustre clients
4421 Description: doing an ls when liblustre clients are running is slow
4422 Details : sending a glimpse AST to a liblustre client waits for every AST
4423 to time out, as liblustre clients will not respond. Since they
4424 cannot cache data we refresh the OST lock LVB from disk instead.
4426 Severity : enhancement
4428 Description: doing an ls at the same time as file IO can be slow
4429 Details : enqueue and other "small" requests can be blocked behind many
4430 large IO requests. Create a new OST IO portal for non-IO
4431 requests so they can be processed faster.
4434 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4436 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4437 created file with mode 000
4438 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4439 allow an ftruncate() as a non-root user to a file with mode 000.
4440 Lustre now respects this flag to disable mode checks when
4441 truncating a file owned by the user
4444 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4447 Description: Revoking locks from clients that went dead or catatonic might take
4449 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4450 cancellation of such locks instant on servers without waiting for
4451 any reply from clients. Clients drops these locks when cancel
4452 notification from server is received without replying.
4455 Frequency : liblustre-only, when liblustre client dies or becomes busy
4457 Description: Doing ls on Linux clients can take a long time with active
4459 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4460 granting such locks to it in the first place if possible. Locks
4461 are taken by proxy on the OST during the read or write and
4462 dropped immediately afterward. Add connect flags handling, do
4463 not grant locks to liblustre clients for glimpse ASTs.
4465 Severity : enhancement
4467 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4468 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4469 sequential reads and behaves badly with applications doing random
4470 reads. Improve it by reading ahead at least read region, and
4471 avoiding excessive large RPC for small reads.
4473 Severity : enhancement
4475 Description: Creating more than 1000 files for a single job may cause a load
4476 imbalance on the OSTs if there are also a large number of OSTs.
4477 Details : qos_prep_create() uses an OST index reseed value that is an
4478 even multiple of the number of available OSTs so that if the
4479 reseed happens in the middle of the object allocation it will
4480 still utilize the OSTs as uniformly as possible.
4485 Description: OST or MDS may oops in ping_evictor_main()
4486 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
4487 but doesn't restart at beginning of obd_exports_timed list
4490 Severity : enhancement
4492 Description: improve by-nid export eviction on the MDS and OST
4493 Details : allow multiple exports with the same NID to be evicted at one
4494 time without re-searching the exports list.
4497 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
4499 Description: MDS may oops in groups_free()
4500 Details : in rare race conditions a newly allocated group_info struct is
4501 freed again, and this can be NULL. The 2.4 compatibility code
4502 for groups_free() checked for a NULL pointer, but 2.6 did not.
4505 Frequency : common for liblustre clients doing little filesystem IO
4506 Bugzilla : 9352, 7313
4507 Description: server may evict liblustre clients accessing contended locks
4508 Details : if a client is granted a lock or receives a completion AST
4509 with a blocking AST already set it would not reply to the AST
4510 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
4513 Frequency : lfs setstripe, only systems with more than 160 OSTs
4515 Description: unable to set striping with a starting offset beyond OST 160
4516 Details : llapi_create_file() incorrectly limited the starting stripe
4517 index to the maximum single-file stripe count.
4520 Frequency : LDAP users only
4522 Description: lconf did not handle in-kernel recovery with LDAP properly
4523 Details : lconf/LustreDB get_refs() is searching the wrong namespace
4525 Severity : enhancement
4527 Description: bind OST threads to NUMA nodes to improve performance
4528 Details : all OST threads are uniformly bound to CPUs on a single NUMA
4529 node and do their allocations there to localize memory access
4531 Severity : enhancement
4533 Description: llmount can determine client NID directly from Myrinet (GM)
4534 Details : the client NID code from gmnalnid was moved directly into
4535 llmount, removing the need to use this or specifying the
4536 client NID explicitly when mounting GM clients with zeroconf
4539 Frequency : if client is started with down MDS
4541 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
4542 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
4543 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
4544 ptlrpc_import_delayed_req() because we didn't check if the
4545 request was interrupted, and we also didn't break out of the
4546 event loop if there was no timeout
4551 Description: data loss during non-page-aligned writes to a single file from
4552 both multiple nodes and multiple threads on one node at same time
4553 Details : updates to KMS and lsm weren't protected by common lock. Resulting
4554 inconsistency led to false short-reads, that were cached and later
4555 used by ->prepare_write() to fill in partially written page,
4556 leading to data loss.
4559 Frequency : always, if lconf --abort_recovery used
4561 Description: lconf --abort_recovery fails with 'Operation not supported'
4562 Details : lconf was attempting to abort recovery on the MDT device and not
4565 Severity : enhancement
4567 Description: remove cleanup logs
4568 Details : replace lconf-generated cleanup logs with lustre internal
4569 cleanup routines. Eliminates the need for client-cleanup and
4572 Severity : enhancement
4574 Description: add support for EAs (user and system) on lustre filesystems
4575 Details : it is now possible to store extended attributes in the Lustre
4576 client filesystem, and with the user_xattr mount option it
4577 is possible to allow users to store EAs on their files also
4579 Severity : enhancement
4581 Description: Add possibility (config option) to show minimal available OST free
4583 Details : When compiled with --enable-mindf configure option, statfs(2)
4584 (and so, df) will return least minimal free space available from
4585 all OSTs as amount of free space on FS, instead of summary of
4586 free spaces of all OSTs.
4588 Severity : enhancement
4590 Description: do not expand extent locks acquired on OST-side
4591 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
4592 by server: they are not cached anyway.
4595 Frequency : when mmap is used/binaries executed from Lustre
4597 Description: Unmmap pages before throwing them away from read cache.
4598 Details : llap_shrink cache now attempts to unmap pages before discarding
4599 them (if unmapping failed - do not discard). SLES9 kernel has
4600 extra checks that trigger if this unmapping is not done first.
4605 Description: lconf didn't resolve symlinks before checking to see whether a
4606 given mountpoint was already in use
4609 Frequency : when migrating failover services
4610 Bugzilla : 6395, 9514
4611 Description: When migrating a subset of services from a node (e.g. failback
4612 from a failover service node) the remaining services would
4613 time out and evict clients.
4614 Details : lconf --force (implied by --failover) sets the global obd_timeout
4615 to 5 seconds in order to quickly disconnect, but this caused
4616 other RPCs to time out too quickly. Do not change the global
4617 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
4619 Severity : enhancement
4620 Frequency : if MDS is started with down OST
4621 Bugzilla : 9439,5706
4622 Description: Allow startup/shutdown of an MDS without depending on the
4623 availability of the OSTs.
4624 Details : Asynchronously call mds_lov_synchronize during MDS startup.
4625 Add appropriate locking and lov-osc refcounts for safe
4626 cleaning. Add osc abort_inflight calls in case the
4627 synchronize never started.
4630 Frequency : occasional (Cray XT3 only)
4632 Description: root not authorized to access files in CRAY_PORTALS environment
4633 Details : The client process capabilities were not honoured on the MDS in
4634 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
4635 been accessed by an authorized user then root was able to access
4636 the file on the local client also. The root user capabilities
4637 are now allowed on the MDS, as this environment has secure UID.
4640 Frequency : occasional
4642 Description: ldiskfs "too long searching" message happens too often
4643 Details : A debugging message (otherwise harmless) prints too often on
4644 the OST console. This has been reduced to only happen when
4645 there are fragmentation problems on the filesystem.
4650 Description: Division by zero in statfs when all OSCs are inactive
4651 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
4652 lov_statfs divides by value returned by lov_get_stripecnt().
4656 Bugzilla : 9489, 3273
4657 Description: First write from each client to each OST was only 4kB in size,
4658 to initialize client writeback cache, which caused sub-optimal
4659 RPCs and poor layout on disk for the first writen file.
4660 Details : Clients now request an initial cache grant at (re)connect time
4661 and so that they can start streaming writes to the cache right
4662 away and always do full-sized RPCs if there is enough data.
4663 If the OST is rebooted the client also re-establishes its grant
4664 so that client cached writes will be honoured under the grant.
4669 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
4670 Details : Now I/O RPCs go to different portal number and (presumably) fast
4671 lock requests (and glimses) and other RPCs get their own service
4672 threads pool that should be able to service those RPCs
4675 Severity : enhancement
4677 Description: Ability to exchange lustre version between client and servers and
4678 issue warnings at client side if client is too old. Also for
4679 liblustre clients there is ability to refuse connection of too old
4681 Details : New 'version' field is added to connect data structure that is
4682 filled with version info. That info is later checked by server and
4686 Frequency : rare, liblustre only.
4687 Bugzilla : 9296, 9581
4688 Description: Two simultaneous writes from liblustre at offset within same page
4689 might proceed at the same time overwriting eachother with stale
4691 Details : I/O lock withing llu_file_prwv was released too early, before data
4692 actually was hitting the wire. Extended lock-holding time until
4693 server acknowledges receiving data.
4696 Frequency : extremely rare. Never observed in practice.
4698 Description: avoid generating lustre_handle cookie of 0.
4699 Details : class_handle_hash() generates handle cookies by incrementing
4700 global counter, and can hit 0 occasionaly (this is unlikely, but
4701 not impossible, because initial value of cookie counter is
4702 selected randonly). Value of 0 is used as a sentinel meaning
4703 "unassigned handle" --- avoid it. Also coalesce two critical
4704 sections in this function into one.
4706 Severity : enhancement
4708 Description: allow liblustre clients to delegate truncate locking to OST
4709 Details : To avoid overhead of locking, liblustre client instructs OST to
4710 take extent lock in ost_punch() on client's behalf. New connection
4711 flag is added to handle backward compatibility.
4713 Severity : enhancement
4714 Bugzilla : 4928, 7341, 9758
4715 Description: allow number of OST service threads to be specified
4716 Details : a module parameter allows the number of OST service threads
4717 to be specified via "options ost ost_num_threads={N}" in the
4718 OSS's /etc/modules.conf or /etc/modprobe.conf.
4722 Bugzilla : 6146, 9635, 9895
4723 Description: servers crash with bad pointer in target_handle_connect()
4724 Details : In rare cases when a client is reconnecting it was possible that
4725 the connection request was the last reference for that export.
4726 We would temporarily drop the export reference and get a new
4727 one, but this may have been the last reference and the export
4728 was just destroyed. Get new reference before dropping old one.
4730 Severity : enhancement
4731 Frequency : if client is started with failover MDS
4733 Description: Allow multiple MDS hostnames in the mount command
4734 Details : Try to read the configuration from all specified MDS
4735 hostnames during a client mount in case the "primary"
4738 Severity : enhancement
4740 Description: Stop sending data to evicted clients as soon as possible.
4741 Details : Check if the client we are about to send or are sending data to
4742 was evicted already. (Check is done every second of waiting,
4743 for which l_wait_event interface was extended to allow checking
4744 of exit condition at specified intervals).
4747 Frequency : rare, normally only when NFS exporting is done from client
4749 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
4751 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
4754 Frequency : always, if srand() is called before liblustre initialization
4756 Description: Liblustre uses system PRNG disturbing its usage by user application
4757 Details : Introduce internal to lustre fast and high-quality PRNG for
4758 lustre usage and make liblustre and some other places in generic
4759 lustre code to use it.
4761 Severity : enhancement
4762 Bugzilla : 9477, 9557, 9870
4763 Description: Verify that the MDS configuration logs are updated when xml is
4764 Details : Check if the .xml configuration logs are newer than the config
4765 logs stored on the MDS and report an error if this is the case.
4766 Request --write-conf, or allow starting with --old_conf.
4768 Severity : enhancement
4770 Description: Handle symlinks in the path when checking if Lustre is mounted.
4771 Details : Resolve intermediate symlinks when checking if a client has
4772 mounted a filesystem to avoid duplicate client mounts.
4777 Description: lconf can hit an error exception but still return success.
4778 Details : The lconf command catches the Command error exception at the top
4779 level script context and will exit with the associated exit
4780 status, but doesn't ensure that this exit status is non-zero.
4785 Description: failure of ptlrpc thread startup can cause oops
4786 Details : Starting a ptlrpc service thread can fail if there are a large
4787 number of threads or the server memory is very fragmented.
4788 Handle this without oopsing.
4791 Frequency : always, only if liblustre and non-default acceptor port was used
4793 Description: liblustre cannot connect to servers with non-default acceptor port
4794 Details : tcpnal_set_default_params() was not called and was therefore
4795 ignoring the environment varaible TCPNAL_PORT, as well as other
4796 TCPNAL_ environment variables
4801 Description: two objects could be created on the same OST for a single file
4802 Details : If an OST is down, in some cases it was possible to create two
4803 objects on a single OST for a single file. No problems other
4804 than potential performance impact and spurious error messages.
4808 Bugzilla : 5681, 9562
4809 Description: Client may oops in ll_unhash_aliases
4810 Details : Client dcache may become inconsistent in race condition.
4811 In some cases "getcwd" can fail if the current directory is
4817 Description: Inode refcounting problems in NFS export code
4818 Details : link_raw functions used to call d_instantiate without obtaining
4819 extra inode reference first.
4823 Bugzilla : 9942, 9903
4824 Description: Referencing freed requests leading to crash, memleaks with NFS.
4825 Details : We used to require that call to ll_revalidate_it was always
4826 followed by ll_lookup_it. Also with revalidate_special() it is
4827 possible to call ll_revalidate_it() twice for the same dentry
4828 even if first occurence returned success. This fix changes semantic
4829 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
4830 reference on a request referred from the intent.
4831 ll_intent_release() then releases such a request.
4834 Frequency : rare, normally benchmark loads only
4836 Description: unlinked inodes were kept in memory on the client
4837 Details : If a client is repeatedly creating and unlinking files it
4838 can accumulate a lot of stale inodes in the inode slab cache.
4839 If there is no other client load running this can cause the
4840 client node to run out of memory. Instead flush old inodes
4841 from client cache that have the same inode number as a new inode.
4844 Frequency : SLES9 2.6.5 kernel and long filenames only
4845 Bugzilla : 9969, 10379
4846 Description: utime reports stale NFS file handle
4847 Details : SLES9 uses out-of-dentry names in some cases, which confused
4848 the lustre dentry revalidation. Change it to always use the
4852 Frequency : rare, unless heavy write-truncate concurrency is continuous
4853 Bugzilla : 4180, 6984, 7171, 9963, 9331
4854 Description: OST becomes very slow and/or deadlocked during object unlink
4855 Details : filter_destroy() was holding onto the parent directory lock
4856 while truncating+unlinking objects. For very large objects this
4857 may block other threads for a long time and slow overall OST
4858 responsiveness. It may also be possible to get a lock ordering
4859 deadlock in this case, or run out of journal credits because of
4860 the combined truncate+unlink. Solution is to do object truncate
4861 first in one transaction without parent lock, and then do the
4862 final unlink in a new transaction with the parent lock. This
4863 reduces the lock hold time dramatically.
4866 Frequency : rare, 2.4 kernels only
4868 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
4869 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
4870 before unmounting their filesystem in order to avoid deadlock.
4871 The kernel_locked() function in 2.4 kernels only checks whether
4872 the kernel lock is held, not whether it is this process that is
4873 holding it as 2.6 kernels do.
4878 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
4879 Details : The client ptlrpc code may be trying to reconnect to a down
4880 server before a previous connection attempt has timed out.
4881 Increase the reconnect interval to be longer than the connection
4882 timeout interval to avoid sending duplicate connections to
4886 Frequency : echo_client brw_test command
4888 Description: fix echo_client to work with OST preallocated code
4889 Details : OST preallocation code (5137) didn't take echo_client IO path
4890 into account: echo_client calls filter methods outside of any
4891 OST thread and, hence, there is no per-thread preallocated
4892 pages and buffers to use. Solution: hijack pga pages for IO. As
4893 a byproduct, this avoids unnecessary data copying.
4897 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
4898 Description: Client can oops in mdc_commit_close() after open replay
4899 Details : It was possible for the MDS to return an open request with no
4900 transaction number in mds_finish_transno() if the client was
4901 evicted, but without actually returning an error. Clients
4902 would later try to replay that open and may trip an assertion
4903 Simplify the client close codepath, and always return an error
4904 from the MDS in case the open is not successful.
4907 Frequency : rare, 2.6 OSTs only
4909 Description: OST may deadlock under high load on fragmented files
4910 Details : If there was a heavy load and highly-fragmented OST filesystems
4911 it was possible to have all the OST threads deadlock waiting on
4912 allocation of biovecs, because the biovecs were not released
4913 until the entire RPC IO was completed. Instead, release biovecs
4914 as soon as they are complete to ensure forward IO progress.
4916 Severity : enhancement
4918 Description: Support for specifying external journal device at mount
4919 Details : If an OST or MDS device is formatted with an external journal
4920 device, this device major/minor is stored in the ext3 superblock
4921 and may not be valid for failover. Allow detecting and
4922 specifying the external journal at mount time.
4927 Description: Mounting an MDS with pending unlinked files may cause oops
4928 Details : target_finish_recovery() calls mds_postrecov() which returned
4929 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
4930 considers this an error and immediately begins cleaning up the
4931 lov, just after starting the mds_lov process
4933 Severity : enhancement
4935 Description: Implement 'lfs df' to report actual free space on per-OST basis
4936 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
4937 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
4938 usage of objects; '-h' to report in human readable format.
4940 ------------------------------------------------------------------------------
4942 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
4949 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
4950 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
4951 at startup during orphan unlinking. Instead, skip buddy update
4952 before setup, as it will be regenerated after recovery is complete.
4957 Description: If an OST is inactive, its locks might reference stale inodes.
4958 Details : lov_change_cbdata() must iterate over all namespaces, even if
4959 they are inactive to clear inode references from the lock.
4961 Severity : enhancement
4962 Frequency : occasional, if non-standard max_dirty_mb used
4964 Description: Client will block write RPCs if not enough grant
4965 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
4966 then the client will block writes while waiting for another RPC
4967 to complete instead of consuming its dirty limit. With change
4968 we get improved performance when max_dirty_mb is small.
4970 Severity : enhancement
4971 Bugzilla : 3389, 6253
4972 Description: Add support for supplementary groups on the MDS.
4973 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
4974 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
4975 lookups for user supplementary groups into a cache.
4979 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
4980 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
4981 for liblustre O_EXCL is enforced by MDS.
4985 Description: readdir never returns NULL in liblustre.
4986 Details : Corrected llu_iop_getdirentries logic, to return offset of next
4987 dentry in struct dirent.
4991 Frequency : liblustre only, depends on application IO pattern
4992 Description: liblustre clients evicted if not contacting servers
4993 Details : Don't put liblustre clients into the ping_evictor list, so
4994 they will not be evicted by the pinger ever.
4996 Severity : enhancement
4998 Description: Add ability to evict clients by NID from MDS.
4999 Details : By echoing "nid:$NID" string into
5000 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
5001 $NID would be instantly evicted from this MDS and from all active
5002 OSTs connected to it.
5006 Description: Do not query file size twice, somewhat slowing stat(2) calls.
5007 Details : lookup_it_finish() used to query file size from OSTs that was not
5012 Description: service threads change working directory to that of init
5013 Details : Starting lustre service threads may pin the working directory
5014 of the parent thread, making that filesystem busy. Threads
5015 now change to the working directory of init to avoid this.
5019 Frequency : during shutdown only
5020 Description: shutdown with a failed MDS or OST can cause unmount to hang
5021 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
5026 Frequency : 2.6 only
5027 Description: chmod/chown may include an extra supplementary group
5028 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
5029 supplementary group and if none is specified this is used.
5032 Bugzilla : 5479 (6816)
5034 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
5035 Details : If lookup is in progress on a file that is unlinked we might try
5036 to revalidate the inode and fail in revalidate after lookup is
5037 complete and ll_file_open() enqueues the open again but
5038 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
5041 Frequency : always, if lconf --abort_recovery used
5043 Description: lconf --abort_recovery fails with 'Operation not supported'
5044 Details : lconf was attempting to abort recovery on the MDT device and not
5047 ------------------------------------------------------------------------------
5049 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
5054 Frequency : rare (only unsupported configurations with a node running as an
5056 Bugzilla : 6514, 5137
5057 Description: Mounting a Lustre file system on a node running as an OST could
5059 Details : OSTs now preallocates memory needed to write out data at
5060 startup, instead of when needed, to avoid having to
5061 allocate memory in possibly low memory situations.
5062 Specifically, if the file system is mounted on on OST,
5063 memory pressure could force it to try to write out data,
5064 which it needed to allocate memory to do. Due to the low
5065 memory, it would be unable to do so and the node would
5066 become unresponsive.
5068 Severity : enhancement
5070 Description: Addition of lconf --service command line option
5071 Details : lconf now accepts a '--service <arg>' option, which is
5072 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
5074 Severity : enhancement
5076 Description: Failover mode is now the default for OSTs.
5077 Details : By default, OSTs will now run in failover mode. To return to
5078 the old behaviour, add '--failout' to the lmc line for OSTs.
5080 Severity : enhancement
5082 Description: Health checks are now provided for MDS and OSTs
5083 Details : Additional detailed health check information on MSD and OSTs
5084 is now provided through the procfs health_check value.
5087 Frequency : occasional, depends on IO load
5089 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
5090 after numerous create/delete cycles
5091 Details : The ext3 inode allocation policy would not allocate new inodes
5092 very well on the OSTs because there are no new directories
5093 being created. Instead we look for groups with free space if
5094 the parent directories are nearly full.
5099 Description: Network or server problems during mount may cause partially
5100 mounted clients instead of returning an error.
5101 Details : The config llog parsing code may overwrite the error return
5102 code during mount error handling, returning success instead
5108 Description: MDS can fail to allocate large reply buffers
5109 Details : After long uptimes the MDS can fail to allocate large reply
5110 buffers (e.g. zconf client mount config records) due to memory
5111 fragmentation or consumption by the buffer cache. Preallocate
5112 some large reply buffers so that these replies can be sent even
5113 under memory pressure.
5117 Frequency : rare (liblustre)
5118 Description: fsx running with liblustre complained that using truncate() to
5119 extend the file doesn't work. This patch corrects that issue.
5120 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
5121 ATTR_SIZE and lsm use in llu_setattr_raw.
5125 Frequency : rare, only 2.6 kernels
5126 Description: Unusual file access patterns on the MDS may result in inode
5127 data being lost in very rare circumstances.
5128 Details : Bad interaction between the ea-in-inode patch and the "no-read"
5129 code in the 2.6 kernel caused the inode and/or EA data not to
5130 be read from disk, causing single-file corruption.
5134 Frequency : rare, only 2.6 filesystems using extents
5135 Description: Heavy concurrent write and delete load may cause data corruption.
5136 Details : It was possible under high-load situations to have an extent
5137 metadata block in the block device cache from a just-unlinked
5138 file overwrite a newly-allocated data block. We now unmap any
5139 metadata buffers that alias just-allocated data blocks.
5143 Frequency : filesystems with default stripe_count larger than 77
5144 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
5145 Details : lconf specifies an inode size of 4096 bytes when the default
5146 stripe_count is larger than 77. This conflicts with the default
5147 inode density of 1 per 4096 bytes. Allocate smaller inodes in
5148 this case to avoid pinning too much memory for large EAs.
5150 ------------------------------------------------------------------------------
5152 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
5157 Frequency : rare (extremely heavy IO load with hundreds of clients)
5159 Description: Client is evicted, gets IO error writing to file
5160 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
5161 caused clients to be evicted for AST timeouts. The fixes in
5162 bug 5192 mean we no longer need to have such short AST timeouts
5163 so ldlm_timeout has been increased.
5166 Frequency : occasional during --force or --failover shutdown under load
5167 Bugzilla : 5949, 4834
5168 Description: Server oops/LBUG if stopped with --force or --failover under load
5169 Details : a collection of import/export refcount and cleanup ordering
5170 issues fixed for safer force cleanup
5173 Frequency : only filesystems larger than 120 OSTs
5174 Bugzilla : 5990, 6223
5175 Description: lfs getstripe would oops on a very large filesystem
5176 Details : lov_getconfig used kfree on vmalloc'd memory
5179 Frequency : only filesystems exporting via NFS to Solaris 10 clients
5180 Bugzilla : 6242, 6243
5181 Description: reading from files that had been truncated to a non-zero size
5182 but never opened returned no data
5183 Details : ll_file_read() reads zeros from no-object files to EOF
5188 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
5189 Details : There's little harm in aborting MDS/OSS recovery and letting it
5190 try again, so I removed the LASSERT and return an error instead.
5192 Severity : enhancement
5194 Description: New debugging infrastructure for tracking down data corruption
5195 Details : The I/O checksum code was replaced to: (a) control it at runtime,
5196 (b) cover more of the client-side code path, and (c) try to narrow
5197 down where problems occurred
5201 Bugzilla : 3819, 4364, 4397, 6313
5202 Description: Racing close and eviction MDS could cause assertion in mds_close
5203 Details : It was possible to get multiple mfd references during close and
5204 client eviction, leading to one thread referencing a freed mfd.
5206 Severity: : enhancement
5207 Bugzilla : 3262, 6359
5208 Description: Attempts to reconnect to servers are now more aggressive.
5209 Details : This builds on the enhanced upcall-less recovery that was added
5210 in 1.4.2. When trying to reconnect to servers, clients will
5211 now try each server in the failover group every 10 seconds. By
5212 default, clients would previously try one server every 25 seconds.
5217 Description: After recovery, certain operations trigger a failed
5218 assertion on a client.
5219 Details : Failing over an mds, using lconf -d --failover, while a
5220 client was doing a readdir() call would cause the client to
5221 LBUG after recovery completed and the readdir() was resent.
5223 Severity : enhancement
5225 Description: Default groups are now added by lconf
5226 Details : You can now run lconf --group <servicename> without having to
5227 manually add groups with lmc.
5230 Frequency : occasional
5232 Description: Nodes with an elan id of 0 trigger a failed assertion
5235 Frequency : always when accessing e.g. tty/console device nodes
5237 Description: tty and some other devices nodes cannot be used on lustre
5238 Details : file's private_data field is used by device data and lustre
5239 values in there got lost. New field was added to struct file to
5240 store fs-specific private data.
5243 Frequency : when exporting Lustre via NFS
5245 Description: NFSD failed occasionally when looking up a path component
5246 Details : NFSD is looking up ".." which was broken in ext3 directories
5247 that had grown large enough to become hashed.
5250 Frequency : Clusters with multiple interfaces not on the same subnet
5252 Description: Nodes will repeatedly try to reconnect to an interface which it
5253 cannot reach and report an error to the log.
5254 Details : Extra peer list entries will be created by lconf with some peers
5255 unreachable. lconf now validates the peer before adding it.
5258 Frequency : Only if a default stripe is set on the filesystem root.
5260 Description: Setting a default stripe on the filesystem root prevented the
5261 filesystem from being remounted.
5262 Details : The client was sending extra request flags in the root getattr
5263 request and did not allocate a reply buffer for the dir EA.
5266 Frequency : occasional, higher if lots of files are accessed by one client
5267 Bugzilla : 6159, 6097
5268 Description: Client trips assertion regarding lsm mismatch/magic
5269 Details : While revalidating inodes the VFS looks up inodes with ifind()
5270 and in rare cases can find an inode that is being freed.
5271 The ll_test_inode() code will free the lsm during ifind()
5272 when it finds an existing inode and then the VFS later attaches
5273 this free lsm to a new inode.
5277 Bugzilla : 6422, 7030
5278 Description: MDS deadlock between mkdir and client eviction
5279 Details : Creating a new file via mkdir or mknod (starting a transaction
5280 and getting the ns lock) can deadlock with client eviction
5281 (gets ns lock and trying to finish a synchronous transaction).
5284 Frequency : occasional
5285 Description: While starting a server, the fsfilt_ext3 module could not be
5287 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
5288 kernels. Previously, lconf would still use the ext3 name
5289 when trying to load modules. Now, it will correctly use
5290 ext3 on 2.4 and ldiskfs on 2.6.
5292 Severity : enhancement
5293 Description: The default stripe count has been changed to 1
5294 Details : The interpretation of the default stripe count (0, to lfs
5295 or lmc) has been changed to mean striping across a single
5296 OST, rather than all available. For general usage we have
5297 found a stripe count of 1 or 2 works best.
5299 Severity : enhancement
5300 Description: Add support for compiling against Cray portals.
5301 Details : Conditional compiling for some areas that are different
5305 Frequency : occasional
5306 Bugzilla : 6409, 6834
5307 Description: Creating files with an explicit stripe count may lead to
5308 a failed assertion on the MDS
5309 Details : If some OSTs are full or unavailable, creating files may
5310 trigger a failed assertion on the MDS. Now, Lustre will
5311 try to use other servers or return an error to the
5315 Frequency : occasional
5317 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
5318 caused lock timeout during readahead (since 1.4.2).
5319 Details : Processes doing readahead might match a lock that hasn't been
5320 granted yet if there are overlapping and conflicting lock
5321 requests. The readahead process waits on ungranted lock
5322 (original lock is CBPENDING), while OST waits for that process
5323 to cancel CBPENDING read lock and eventually evicts client.
5325 Severity : enhancement
5327 Description: Initial enabling of flock support for clients
5328 Details : Implements fcntl advisory locking and file status functions.
5329 This feature is provided as an optional mount flag (default
5330 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
5331 locking are implemented yet, and those that are are not guaranteed
5332 to be completely correct in production environments.
5333 mount -t lustre -o [flock|noflock] ...
5336 Frequency : occasional
5338 Description: OSTs running 2.4 kernels but with extents enabled might trip an
5339 assertion in the ext3 JBD (journaling) layer.
5340 Details : The b_committed_data struct is protected by the big kernel lock
5341 in 2.4 kernels, serializing journal_commit_transaction() and
5342 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
5343 access to this struct. In 2.6 kernels there is finer grained
5344 locking to improve SMP performance of the JBD layer.
5348 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
5353 Description: merge and cleanup kernel patches.
5356 -----------------------------------------------------------------------------
5358 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
5360 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
5361 versions of Lustre. Please update all servers and clients to
5362 version 1.4.2 or later at the same time. You must also run
5363 "lconf --write-conf {config}.xml" on the MDS while it is stopped
5364 to update the configuration logs.
5366 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
5367 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
5368 - swab llog records of type '0' so we get proper header size/idx (5861)
5369 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
5370 - fix rename of one directory over another leaking an inode (5953)
5371 - avoid SetPageDirty on 2.6 (5981)
5372 - don't re-add just-being-destroyed locks to the waiting list (5653)
5373 - when creating new directories, inherit the parent's custom
5374 striping settings if present parent (3048)
5375 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
5376 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
5377 - don't hold client locks on temporary worklist from l_lru (5666)
5378 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
5379 - automatically evict dead clients (5921)
5380 - Update file size properly in create+truncate+fstat case (6196)
5381 - Do not unhash mountpoint dentries, do not allow removal of
5383 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
5384 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
5385 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
5386 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
5387 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
5388 - Correctly handle reads of files with no objects (6243)
5389 - lctl recover will also mark a device active if deactivate used (5933)
5391 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
5392 - allow --write-conf on an MDS with different nettype than client (5619)
5393 - don't write config llogs to MDS for mounts not from that MDS (5617)
5394 - lconf should create multiple TCP connections from a client (5201)
5395 - init scripts are now turned off by default; run chkconfig --on
5396 lustre and chkconfig --on lustrefs to use them
5397 - upcalls are no longer needed for clients to recover to failover
5399 - add --abort-recovery option to lconf to abort recovery on device
5401 - add support for an arbitrary number of OSTs (3026)
5402 - Quota support protocol changes.
5403 - forward compatibility changes to wire structs (6007)
5404 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5405 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5406 - improved "lustre status" script
5407 - initialize blocksize for non-regular files (6062)
5408 - added --disable-server and --disable-client configure options (5782)
5409 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5410 - Vanilla 2.4.29 support
5411 - increase maximum number of obd devices to 520 (6242)
5412 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5413 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5414 SLES 9 2.6.5 kernels
5415 - limit stripes per file to 160 (the maximum EA size) (6093)
5417 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5420 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5421 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5422 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5423 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5424 - small ext3 extents cleanups and fixes (5733)
5425 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5426 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5427 - avoid needless client->OST connect, fix handle mismatch (5317)
5428 - fix DLM error path that led to out-of-sync client, long delays (5779)
5429 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5430 - fix several locking issues related to i_size (5492,5624,5654,5672)
5431 - don't move pending lock onto export if it is already evicted (5683)
5432 - fix kernel oops when creating .foo in unlinked directory (5548)
5433 - fix deadlock in obdfilter statistics vs. object create (5811)
5434 - use time_{before,after} to avoid timer jiffies wrap (5882)
5435 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5436 - Do not leak request if server was not able to process it (5154)
5437 - If mds_open unable to find parent dir, make that negative lookup(5154)
5438 - don't create new directories with extent-mapping (5909, 5936)
5440 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5441 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5442 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5443 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5444 - support for RHEL4 kernel on i686 (5773)
5445 - provide error messages when incompatible logs are encountered (5898)
5447 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5448 * version 1.4.0.10 (1.4.1 release candidate 1)
5450 - don't keep a lock reference when lock is not granted (4238)
5451 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5452 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5453 - drop import inflight refcount on signal_completed_replay error (5255)
5454 - unlock page after async write error during send (3677)
5455 - handle missing objects in filter_preprw_read properly (5265)
5456 - no transno return for symlink open, don't save no-trasno open (3440)
5457 - don't try to complete elan receive that already failed (4012)
5458 - free RPC server reply state on error (5406)
5459 - clean up thread from ptlrpc_start_thread() on error (5160)
5460 - readahead could read extra page into cache that wasn't ejected (5388)
5461 - prevent races in class_attach/setup/cleanup/detach (5260)
5462 - don't dereference de->d_inode after l_dput of de (5458)
5463 - use "int" for stripe value returned from lock_to_stripe (5544)
5464 - mballoc allocation and error-checking fixes in 2.6 (5504)
5465 - block device patches to fix I/O request sizes in 2.6 (5482)
5466 - look up hostnames for IB nals (5602)
5467 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5468 - don't start multiple acceptors for the same port (5277)
5469 - fix incorrect LASSERT in mds_getattr_name (5635)
5470 - export a proc file for general "ping" checking (5628)
5471 - fix "lfs check" to not block when the MDS is down (5628)
5473 - service request history (4965)
5474 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5475 - create an "evict_client" /proc entry on OSTs, like the MDS has
5476 - fix mount usage message, return errors per mount(8) (5168)
5477 - change grep [] to grep "[]" in tests so they work in more UMLs
5478 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5479 - remove ancient LOV_MAGIC_V0 EA support (5047)
5480 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5481 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5482 - allow readahead allocations to fail when low on memory (5383)
5483 - mmap locking landed again, after considerable improvement (2828)
5484 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
5486 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
5489 - send OST transaction number in read/write reply to free req (4966)
5490 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5491 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5492 - the watchdog thread now runs as interruptible (5246)
5493 - flock/lockf fixes (but it's still disabled, pending 5135)
5494 - don't use EXT3 constants in llite code (5094)
5495 - memory shortage at startup could cause assertion (5176)
5497 - reorganization of lov code
5498 - single portals codebase
5500 - add extents/mballoc support (5025)
5501 - direct I/O reads in the obdfilter (4048)
5502 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
5504 tbd Cluster File Systems, Inc. <info@clusterfs.com>
5507 - send OST transaction number in read/write reply to free req (4966)
5508 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5509 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5510 - the watchdog thread now runs as interruptible (5246)
5511 - handle missing objects in filter_preprw_read properly (5265)
5512 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5513 - drop import inflight refcount on signal_completed_replay error (5255)
5514 - unlock page after async write error during send (3677)
5515 - return original error code on reconstructed replies (3761)
5516 - no transno return for symlink open, don't save no-trasno open (3440)
5518 - add pid to ldlm debugging output (4922)
5519 - bump the watchdog timeouts -- we can't handle 30sec yet
5520 - extra debugging for orphan dentry/inode bug (5259)
5522 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
5525 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
5526 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
5527 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
5528 - fix NULL dereference in /proc/sys/portals/routes (4827)
5529 - allow failed mdc_close() operations to be interrupted (4561)
5530 - stop precreate on OST before MDS would time out on it (4778)
5531 - don't send partial-page writes before EOF from client (4410)
5532 - discard client grant for sub-page writes on large-page clients (4520)
5533 - don't free dentries not owned by NFS code, check generation (4806)
5534 - fix lsm leak if mds_create_objects() fails (4801)
5535 - limit debug_daemon file size, always print CERROR messages (4789)
5536 - use transno after validating reply (3892)
5537 - process timed out requests if import state changes (3754)
5538 - update mtime on OST during writes, return in glimpse (4829)
5539 - add mkfsoptions to LDAP (4679)
5540 - use ->max_readahead method instead of zapping global ra (5039)
5541 - don't interrupt __l_wait_event() during strace
5543 - add software watchdogs to catch hung threads quickly (4941)
5544 - make lustrefs init script start after nfs is mounted
5545 - fix CWARN/ERROR duplication (4930)
5546 - return async write errors to application if possible (2248)
5547 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
5548 - print NAL number in %x format (4645)
5549 - update barely-supported suse-2.4.21-171 series (4842)
5550 - support for sles 9 %post scripts
5551 - support for building 2.6 kernel-source packages
5552 - support for sles km_* packages
5554 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
5557 - ignore -ENOENT errors in osc_destroy (3639)
5558 - notify osc create thread that OSC is being cleaned up (4600)
5559 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
5560 - reconstruct ost_handle() like mds_handle() (4657)
5561 - create a new thread to do import eviction to avoid deadlock (3969)
5562 - let lconf resolve symlinked-to devices (4629)
5563 - don't unlink "objects" from directory with default EA (4554)
5564 - hold socknal file ref over connect in case target is down (4394)
5565 - allow more than 32000 subdirectories in a single directory (3244)
5566 - fix blocks count for O_DIRECT writes (3751)
5567 - OST returns ENOSPC from object create when no space left (4539)
5568 - don't send truncate RPC if file size isn't changing (4410)
5569 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
5570 - bind to privileged port in socknal and tcpnal (3689)
5572 - rate limit CERROR/CWARN console message to avoid overload (4519)
5573 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
5574 - basic mmap support (3918)
5575 - kernel patch series update from b1_4 (4711)
5577 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
5580 - avoid crash during MDS cleanup with OST shut down (2775)
5581 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
5582 - don't use bad inodes on the MDS (3744)
5583 - dynamic object preallocation to improve recovery speed (4236)
5584 - don't hold spinlock over lock dumping or change debug flags (4401)
5585 - don't zero obd_dev when it is force cleaned (3651)
5586 - print grants to console if they go negative (4431)
5587 - "lctl deactivate" will stop automatic recovery attempts (3406)
5588 - look for existing locks in ldlm_handle_enqueue() (3764)
5589 - don't resolve lock handle twice in recovery avoiding race (4401)
5590 - revalidate should check working dir is a directory (4134)
5592 - don't always mark "slow" obdfilter messages as errors (4418)
5594 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
5597 - don't close LustreDB during write_conf until it is done (3860)
5598 - fix typo in lconf for_each_profile (3821)
5599 - allow dumping logs from multiple threads at one time (3820)
5600 - don't allow multiple threads in OSC recovery (3812)
5601 - fix debug_size parameters (3864)
5602 - fix mds_postrecov to initialize import for llog ctxt (3121)
5603 - replace config semaphore with spinlock (3306)
5604 - be sure to send a reply for a CANCEL rpc with bad export (3863)
5605 - don't allow enqueue to complete on a destroyed export (3822)
5606 - down write_lock before checking llog header bitmap (3825)
5607 - recover from lock replay timeout (3764)
5608 - up llog sem before sending rpc (3652)
5609 - reduce ns lock hold times when setting kms (3267)
5610 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
5611 - fix NULL deref and obd_dev leak on setup error (3312)
5612 - replace some LBUG about llog ops with error handling (3841)
5613 - don't match INVALID dentries from d_lookup and spin (3784)
5614 - hold dcache_lock while marking dentries INVALID and hashing (4255)
5615 - fix invalid assertion in ptlrpc_set_wait (3880)
5617 - add libwrap support for the TCP acceptor (3996)
5618 - add /proc/sys/portals/routes for non-root route listing (3994)
5619 - allow setting MDS UUID in .xml (2580)
5620 - print the stack of a process that LBUGs (4228)
5622 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
5625 - don't cleanup request in ll_file_open() on failed MDS open (3430)
5626 - make sure to unset replay flag from failed open requests (3440)
5627 - if default stripe count is 0, use OST count for inode size (3636)
5628 - update parent mtime/ctime on client for create/unlink (2611)
5629 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
5630 - free recovery state on server during a forced cleanup (3571)
5631 - unregister_reply for resent reqs (3063)
5632 - loop back devices mounting and status check on 2.6 (3563)
5633 - fix resource-creation race that can provoke i_size == 0 (3513)
5634 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
5635 - more debugging for page-accounting assertion (3746)
5636 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
5637 - avoid deadlock after precreation failure (3758)
5638 - fix race and lock order deadlock in orphan handling (3450, 3750)
5639 - add validity checks when grabbing inodes from l_ast_data (3599)
5641 - add /proc/.../recovery_status to obdfilter (3428)
5642 - lightweight CDEBUG infrastructure, debug daemon (3668)
5643 - change default OSC RPC parameters to be better on small clusters
5644 - turn off OST read cache for files smaller than 32MB
5645 - install man pages and include them in rpms (3100)
5646 - add new init script for (un)mounting lustre filesystems (2593)
5647 - run chkconfig in %post for init scripts (3701)
5648 - drop scimac NAL (unmaintained)
5650 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
5653 - clean kiobufs before and after use (3485)
5654 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
5655 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
5656 - change init script to fail more clearly if not run as root (1528)
5657 - allow clients to reconnect during replay (1742)
5658 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
5659 - don't do DNS lookups on NIDs too small for IP addresses (3442)
5660 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
5661 - fix cond_resched (3554)
5662 - only evict unfinished clients after recovery (3515)
5663 - allow bulk resend, prevent data loss (3570)
5664 - dynamic ptlrpc request buffer allocation (2102)
5665 - don't allow unlinking open directory if it isn't empty (2904)
5666 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
5667 - remove extraneous obd dereference causing LASSERT failure (3334)
5668 - don't use get_cycles() when creating temp. files on the mds (3156)
5669 - hold i_sem when setting i_size in ll_extent_lock() (3564)
5670 - handle EEXIST for set-stripe, set proper directory name (3336)
5672 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
5673 - fix ksocknal_fmb_callback() error messages (2918)
5675 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
5678 - don't copy lvb into (possibly NULL) reply on error (2983)
5679 - don't deref dentry after dput, don't free lvb on error (2922)
5680 - use the kms to determine writeback rpc length (2947)
5681 - increment oti_logcookies when osc is inactive (2948)
5682 - update client's i_blocks count via lvb messages (2543)
5683 - handle intent open/close of special files properly (1557)
5684 - mount MDS with errors=remount-ro, like obdfilter (2009)
5685 - initialize lock handle to avoid ASSERT on error cleanup (3057)
5686 - don't use cancelling-locks' kms values (2947)
5687 - use highest lock extent for kms, not last one (2925)
5688 - don't dereference ERR_PTR() dentry in error handling path (3107)
5689 - fix thread race in portals_debug_dumplog() (3122)
5690 - create lprocfs device entries at setup instead of at attach (1519)
5691 - common AST error handler, don't evict client on completion race (3145)
5692 - zero nameidata in detach_mnt in 2.6 (3118)
5693 - verify d_inode after revalidate_special is valid in 2.6 (3116)
5694 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
5695 - initialize RPC timeout timer earlier for 2.6 (3219)
5696 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
5697 - print nal/nid for unknown nid (3258)
5698 - additional checks for oscc recovery before doing precreate (3284)
5699 - fix ll_extent_lock() error return code for 64-bit systems (3043)
5700 - don't crash in mdc_close for bad permissions on open (3285)
5701 - zero i_rdev for non-device files (3147)
5702 - clear page->private before handing to FS, better assertion (3119)
5703 - tune the read pipeline (3236)
5704 - fix incorrect decref of invalidated dentry (2350)
5705 - provide read-ahead stats and refine rpc in flight stats (3328)
5706 - don't hold journal transaction open across create RPC (3313)
5707 - update atime on MDS at close time (3265)
5708 - close LDAP connection when recovering to avoid server load (3315)
5709 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
5710 - don't leak open file on MDS after open resend (3325)
5711 - serialize filter_precreate and filter_destroy_precreated (3329)
5712 - loop device shouldn't call sync_dev() for nul device (3092)
5713 - clear page cache after eviction (2766)
5714 - resynchronize MDS->OST in background (2824)
5715 - refuse to mount the same filesystem twice on same mountpoint (3394)
5716 - allow llmount to create routes for mounting behind routers (3320)
5717 - push lock cancellation to blocking thread for glimpse ASTs (3409)
5718 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
5719 - fix rare problem with rename on htree directories (3417)
5721 - allow default OST striping configuration per directory (1414)
5722 - fix compilation for qswnal for 2.6 kernels (3125)
5723 - increase maximum number of MDS request buffers for large systems
5724 - change liblustreapi to be useful for external progs like lfsck (3098)
5725 - increase local configuration timeout for slow disks (3353)
5726 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
5728 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5731 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
5732 - don't overwrite extent policy data in reply if lock was blocked (2901)
5733 - drop filter export grants atomically with removal from device (2663)
5734 - del obd_self_export from work_list in class_disconnect_exports (2908)
5735 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
5736 - swab reply message in mdc_close, other PPC fixes (2464)
5737 - fix destroying of named logs (2325)
5738 - overwrite old logs when running lconf --write_conf (2264)
5739 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
5740 - fix race in target_handle_connect (2898)
5741 - mds_reint_create() should take same inode create lock (2926)
5742 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
5743 - don't close files for self_export to avoid uninitialized obd (2936)
5744 - allow MDS with the same name as client node (2939)
5745 - hold dentry reference for closed log files for unlink (2325)
5746 - reserve space for all logs during transactions (2059)
5747 - don't evict page beyond end of stripe extent (2925)
5748 - don't oops on a deleted current working directory (2399)
5749 - handle hard links to targets without a parent properly (2517)
5750 - don't dereference NULL lock when racing during eviction (2867)
5751 - don't grow lock extents when lots of conflicting locks (2919)
5753 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
5756 - account for cache space usage on clients to avoid data loss (974)
5757 - lfsck support in lustre kernel code (2349)
5758 - reduce journal credits needed for BRW writes (2370)
5759 - orphan handling to avoid losing space on client/server crashes
5760 - ptlrpcd can be blocked, stopping ALL progress (2477)
5761 - use lock value blocks to assist in proper KMS, faster stat (1021)
5762 - takes i_sem instead of DLM locks internally on obdfilter (2720)
5763 - recovery for initial connections (2355)
5764 - fixes for mds_cleanup_orphans (1934)
5765 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
5766 - block all file creations until orphan recovery completes (1901)
5767 - client remove rq_connection from request struct (2423)
5768 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
5769 - recovery timer race (2670)
5770 - mdc_close recovey bug (2532)
5771 - ptlrpc cleanup bug (2710)
5772 - mds timeout on local locks (2588)
5773 - namespace lock held during RPCs (2431)
5774 - handle interrupted sync write properly (2503)
5775 - don't try to handle a message that hasn't been replied to (2699)
5776 - client assert failure during cleanup after abort recovery (2701)
5777 - leak mdc device after failed mount (2712)
5778 - ptlrpc_check_set allows timedout requests to complete (2714)
5779 - wait for inflight reqs when ptlrpcd finishes (2710)
5780 - make sure unregistered services are removed from the srv_list
5781 - reset bulk XID's when resending them (caught by 1138 test)
5782 - unregister_bulk after timeout
5783 - fix lconf error (2694)
5784 - handle write after unfinished setstripe, stripe-only getstripe (2388)
5785 - readahead locks pages, leaves pending causing memory pressure (2673)
5786 - increase OST request buffers to 4096 on large machines (2729)
5787 - fix up permission of existing directories in simple_mkdir (2661)
5788 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
5789 - don't assign transno to errored transactions (2742)
5790 - don't delete objects on OST if given a bogus objid from MDS (2751)
5791 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
5792 - if rq_no_resend, then timeout request after recovery (2432)
5793 - fix MDS llog_logid record size, 64-bit array alignment (2733)
5794 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
5795 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
5796 - MDS recovery shouldn't delete objects that it has given out (2730)
5797 - if enqueue arrives after completion, don't clobber LVB (2819)
5798 - don't unlock pages twice when trigger_group_io returns error (2814)
5799 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
5800 - don't write pages to disk if there was an error (1450)
5801 - don't ping imports that have recovery disabled (2676)
5802 - take buffered bytes into account when balancing socknal conn (2817)
5803 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
5804 - reconnect unlink llog connection after MDS reconnects to OST (2816)
5805 - remove little-endian swabbing of llog records (1987)
5806 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
5807 - retry reposting request buffers if they fail (1191)
5808 - grow extent at grant time to avoid granting a revoked lock (2809)
5809 - lock revoke doesn't evict page if covered by a second lock (2765)
5810 - disable VM readahead to avoid reading outside lock extents (2805)
5812 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
5813 - updated kernel patches for hp-2.4.20 kernel (2681)
5815 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
5818 - fix truncated write corruption (2366)
5819 - fix for failed assertion in iopen_connect_dentry (1792,2517)
5821 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
5822 - ensure len(uuid) < 37 in lmc (1171)
5823 - fix ia64 OOPS in llog_test (2255)
5824 - zero end of page at obdfilter for partial page writes (2648)
5825 - don't leave stale dentries around after renames (bug 2428)
5826 - fix timeouts when evicting a client with a single lock held (2642)
5827 - set deadline for the initial HELLO message to drain (2634)
5828 - print out dotted-quad IP addresses in the socknal (2302)
5830 - additional debugging for MDS client eviction problem (2443)
5831 - fix mkfsoptions support for osts (2603, 2604)
5833 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
5836 - add series for the vanilla 2.6.0 kernel
5837 - add series for the vanilla 2.4.24 kernel
5838 - add series for a cray x86/64 UL kernel drop
5839 - fix xattr patches for the vanilla 2.4.19 series
5841 - generate true UUIDs in lmc (1171)
5842 - have portals stack dumping break in UML (2466)
5843 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
5844 - allocate with _NOFS in ldlm to avoid deadlock (1933)
5845 - wake callback waiting threads on client eviction (2460)
5846 - Add --ptldebug and --subsystem to lmc (1719)
5847 - update assertion to allow safe interrupt allocation
5848 - set rq_no_resend for cancel requests (2432)
5849 - recalculate ptlrpcd timeout after resend (2494)
5850 - call vfs_rmdir when removing pending directories (2368)
5851 - fix renaming a file to itself (2429)
5852 - lmc creates a default one-stripe lov (2454)
5853 - expand procfs space to handle large clusters (2326)
5854 - increase UML stack to avoid overflow
5855 - update lconf's list of debug and subsystem masks
5856 - fix lfs find --obd (2510)
5857 - /proc tunable for disabling filter read caching (2591)
5858 - stop rpm packages from altering slapd.conf (2301)
5859 - disable nagle in the socknal under 0conf (2578)
5860 - choose mds inode size based on stripe count (2572)
5861 - fix kernel-source rpm problems (2516)
5863 - add --disable-doc to avoid pdf generation (2421)
5864 - update documentation, tests, type-os, comments
5865 - avoid format warnings on ia64
5866 - remove the TOE NAL
5867 - tiny code cleanups by removing unused fields
5869 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
5872 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
5873 - remove the most blatant lies from BUILDING file (2371)
5874 - change default debug level to reasonable production setting
5875 - reduce client side cache size to reduce cache flush time
5876 - reduce max RPCs in flight to avoid unnecessary file fragmentation
5877 - make TCP zerocopy and pinger support enabled by default (2476)
5878 - sync writes completed after process exits caused crashes (2319)
5879 - maintain correct mount count on the MDS (2356)
5880 - backout 1557, because 2316 wasn't really fixed
5881 - better file I/O statistics gathering in /proc
5882 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
5883 - another kernel patch to fix zero-copy TCP function export
5884 - don't take duplicate lock when processing re-sent getattr (2420)
5885 - lctl uses obd_self_export instead of creating new conn (2353)
5886 - MDS/OST recovery case which requires object creation asserted (2425)
5887 - move lfs from /usr/sbin to /usr/bin in packages
5888 - fix race between mds_client_add and mds_client_free (2417)
5889 - use kmalloc instead of slabs in portals (2430)
5890 - don't create duplicate records when a failover MDS is present (2442)
5891 - remove unnecessary mount age check (2332)
5892 - don't remove directory inodes from locks prematurely (2451)
5893 - don't break if MDS service name is the same as hostname (2103)
5894 - fix races in client write RPC generation when cache full (2482)
5896 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
5899 - remove now-unused request->rq_obd (278)
5900 - if an allocation fails, print out how much memory we've used (1933)
5901 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
5902 - add the "configurable stack size" patch to most series files (1256)
5903 - ability to write large log records, for 100+ OST configs (2306)
5904 - fix NULL deref when filter_prep fails (2314)
5905 - fix operator precedence error in filter_sync
5906 - dynamic allocation of socknal TX descriptors (2315)
5907 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
5908 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
5909 - ensure that CWARN messages in Portals always get to the syslog
5910 - __init/__exit are not for prototype decls (ldlm_init/exit)
5911 - x86-64 compile warning fixes
5912 - fix gateway LMC keyword conflict (2318)
5913 - fix MDS lock inversions in getattr/reint paths (1844)
5914 - fix a rare lock re-ordering bug, which caused deadlock (2322)
5915 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
5916 - DLM race condition prevented some lock evictions (2328)
5917 - ENOMEM detection and retry on socknal sends (2230)
5918 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
5919 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
5920 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
5921 - better error messages when a client is rejected during recovery (1505)
5922 - avoid cancelling locks which were never granted, after failure (2330)
5923 - fix i_sem/journal inversion in mds_client_add (2333)
5924 - fix truncate/getattr lock cycle deadlock (2334)
5925 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
5926 - fix two rare exit paths which could leak an l_lock() ref (2321)
5927 - fencepost error in MDS/OST orphan recovery (2226)
5928 - make log record alignment 8 bytes (1988)
5929 - lstripe now fails when requested offset > ost_count (2237)
5930 - ensure that all kernel series have a complete list.h (1607)
5931 - fix crashes in special-file operations (2316)
5932 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
5934 - allow configurable automake binary, for testing new versions
5935 - small update to the lfs documentation
5937 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
5939 * fix negative export reference count in fsfilt_sync (2312)
5941 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
5942 * release candidate 0.9.1
5944 - orphans are moved into the PENDING directory for possible recovery
5945 - replayed opens now open by fid for orphan/rename safety (1042)
5946 - last close of an orphan inode generates a transno (683)
5947 - chdir() and mount() now pin the directory entry (1020)
5948 - avoid CERROR in normal ll_setattr_raw() error case (1500)
5949 - discard very old requests without processing them (1502)
5950 - remove some common, well-understood CERRORs (1505)
5951 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
5952 - clear "grant" flags in OST replies until OST grant code lands (1644)
5953 - fix read performance by not clobbering i_blksize on client (1598)
5954 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
5955 - make LRU size a /proc tunable, clears locks when reduced (707)
5956 - fix some lprocfs rot that prevented ptlbd from loading (1732)
5957 - server locks take references on exports now (1558)
5958 - build fixes for 2.4.20-rh trees (1663)
5959 - return an error from lov_create if all OSCs are inactive (1751)
5960 - fix import levels when a reconnect happens without a timeout (1597)
5961 - exit early from mds_open if we get a lookup error (1749)
5962 - partial page read at EOF wouldn't wait for disk before sending (1642)
5963 - avoid NULL deref in obdfilter when reading page past EOF (1592)
5964 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
5965 - fix LBUG in ll_it_open_error with rc = -2 (1861)
5966 - write/truncate lock inversion (1639)
5967 - Don't auto-load obdclass, portals modules during cleanup (1495)
5968 - fix timestamps from jumping to "now" (1763)
5969 - extra journal assertions (1648)
5970 - add an extra multiunlink test (1771)
5971 - fix read_record/write_record API (1776)
5972 - fix leak of offset_extent, possible incorrect i_size later (1772)
5973 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
5974 - Debugging for the kqswnal_get_idle_tx problems (1820)
5975 - Allow recovery to be attempted multiple times (1536)
5976 - Write out MDS last_rcvd file after it is first created (1600)
5977 - Fix tx_descriptor leak in failed transmit situations (1827)
5978 - ext3 journaling fixes for assertion failure after IO error (1871)
5979 - class_export_put() on freed export after completion AST error (1896)
5980 - Fix revalidate looping in VFS (1322)
5981 - Don't access a freed export during MDS_REINT timeout (1521)
5982 - Add open-unlink recovery support on the MDS (1673,1764)
5983 - Return an error if no MDS data was read from last_rcvd (1946)
5984 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
5985 - Don't setup a disk device that doesn't match exported UUID (317)
5986 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
5987 - avoid committing NULL handle in force close
5988 - local.sh is now a one-stripe LOV configuration
5989 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
5990 - fix ext3 htree duplicate directory entry corruption (1516)
5991 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
5992 - update to new LOV EA format (2097)
5993 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
5995 2003-06-15 Phil Schwan <phil@clusterfs.com>
5998 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
5999 - per-import/export recovery handling (958, 931, 959)
6000 - multiple last-rcvd slots, for serving multiple FSes (949)
6001 - connections are again shared between multiple imp/exports (963, 964)
6002 - "umount -f" would hang if any requests needed to be sent (393, 978)
6003 - avoid pinning large req buffer by copying for queued messages (989)
6004 - add "uuid" to "lctl device" command to help upcalls (991)
6005 - "open" RPCs with transnos would confuse recovery counters (1037)
6006 - do proper endian conversion of all wire messages (288, 340, 891)
6007 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
6008 - call waiting locks callback from LDLM recovery thread (1127, 1151)
6009 - fix ptlrpc_connection leak in target_handle_connect (1174)
6010 - fix import refcounting bug in OST and MDS cleanup (1134)
6011 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
6012 - fix very unlikely obd_types race condition (501)
6013 - remove osc_open hack for echo_client (1187)
6014 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
6015 - a failure in read_inode2 leads to deadlock (1139)
6016 - cancel ack-locks as soon as transaction is committed (1072)
6017 - fix major leaks and crashes in the bulk I/O path (937, 1057)
6018 - make sure to commitrw after any preprw to avoid deadlock (1162)
6019 - failing to execute a file in a lustre FS would lock inode (1203)
6020 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
6021 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
6022 - obd_uuid2tgt would walk off the end of the list (1255)
6023 - on IA64 the osc would give portals incorrect bulk size (1258)
6024 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
6025 - fix lock inversion caused by new llite matching code (1282)
6026 - limit the number of dirty pages on a client to 10MB (1286)
6027 - timed out locks were not being corrected cancelled (1289)
6028 - fix O_DIRECT above 4GB on IA-32 (1292)
6029 * major user-visible changes
6030 - fail out/fail over policy now controlled by the upcall (993)
6032 - add OBD_PING to check server availability and failure (954)
6033 - lustre messages are now sent in sending host order (288, 340, 891)
6034 - add eadatalen to MDS getattr reply (340)
6035 - OST read replies may contain second buffer, with per-page status (593)
6037 2003-03-11 Phil Schwan <phil@clusterfs.com>
6040 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
6041 - failed open()s could cause deadlock; fixed (867, 869)
6042 - stop cancelling OST locks when files are closed (481)
6043 - overlapping XID spaces caused network corruption (851, 853)
6044 - fix unsafe fsfilt counter arithmetic; change to atomic_t
6045 - setattr_raw added, to do single-RPC, server-side setattrs
6046 - lmc/lconf syntax change for OST UUIDs
6047 - fix crashy race condition between ptlrpc_free_req and osc_close
6048 - don't use request in mdc_enqueue if we hit a timeout (889)
6049 - don't set the inode i_size for regular files from the MDS (896)
6050 - handle out of order completion AST (842)
6051 - don't LBUG if a lock request times out after receiving AST (913)
6052 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
6053 - if a bad lock AST arrives, send an error instead of dropping entirely
6054 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
6055 - fix leak in bulk IO when only partially completed (899, 900, 926)
6056 - fix O_DIRECT for ia64 (55)
6057 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
6058 - C-z after timeout could hang a process forever; fixed (977)
6060 - client-side I/O cache (678, 924, 929, 941, 970)
6062 - READPAGE and SETATTRs which don't take server-side locks get
6065 2003-02-11 Phil Schwan <phil@clusterfs.com>
6068 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
6069 - Fix fsfilt_extN_readpage() to read a full page of directory
6070 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
6071 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
6072 - Fix imbalanced LOV object allocation and out-of-bound access (469)
6073 - Most intent operations were removed, in favour of a new RPC mode
6074 that does a single RPC to the server and bypasses most of the VFS
6075 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
6076 - Aggressively cancel local locks on DLM servers
6077 - mds_reint_unlink sends EA to the client if it's the last nlink.
6078 client uses that EA to unlink OST objects.
6079 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
6080 - recursive symlinks were fixed (439)
6081 - fixed NULL deref in DEBUG_REQ
6082 - filter_update_lastobjid no longer calls sync, which annoyed extN
6083 - fixed multi-client small-writes to a single file problem (445)
6084 - fixed mtime updates during file writes (607)
6085 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
6086 - fixed bug in obd_brw_read/write() (under guise of testing 367)
6087 - fixed Linux OST size reporting problem (444, 656)
6088 - OST now updates object mtime with writes or setattr (607, 619)
6089 - client verifies file size before zeroing page past EOF (445)
6090 - OST now writes last allocated objid to disk with allocation (108)
6091 - LOV on echo now works (409)
6093 - mds_reint_unlink sends a new buffer, with the EA included. this
6094 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
6095 set if a regular file was being unlinked, and it was the last link
6096 - use PtlGet from the target for bulk writes (315)
6097 - OST now updates object mtime with writes or setattr (607, 619)
6098 - LDLM now has a grant-time callback to revalidate locked items, if
6100 - Many MDS operations were reorganized to combat race conditions
6102 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
6103 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
6105 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
6108 - Fully reactivate OST imports after reconnection (512, others)
6109 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
6110 - More graceful error handling for truncating on dead OST (515)
6111 - Don't error out unless we're actually accessing dead stripes (474)
6112 - Fix garbage sizes when stripes are missing (410)
6113 - LRU counters were broken, causing constant lock purge (433, 432)
6114 - garbage on read from stripes with failed OSTs (441)
6115 - mark OSCs as active before reconnecting during recovery (438)
6116 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
6117 - lfind did not preserve OST order in output (443)
6118 - symlinks cause hung clients, incorrect data (439)
6119 - stop dereferencing request after dropping refcount (457)
6120 - don't LASSERT(spin_is_locked) on non-SMP (455)
6121 - fixes for many rename() bugs
6122 - fstat didn't correctly synchronize attributes (399)
6123 - server must handle lock cancellation during blocking AST prep (487)
6124 - bulk descriptors were free()d too soon (511)
6125 - fix paths in lconf, which would load incorrect modules (451, 507)
6126 - fix confusing lconf 'host not found' error message (386)
6127 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
6128 - fix race condition in mdc_blocking_ast() for inode access (526)
6129 - fix lov_unpackmd() unpacking wrong number of stripes (537)
6130 - fix lov_set_osc_active() marking wrong OSC inactive (440)
6131 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
6132 - fix multiple writes of stripe MD to MDS (358, maybe 519)
6133 - fix lstripe in several ways (kernel side) (527)
6134 - fix request leak in ldlm_cli_enqueue (262)
6135 - incorrect OSC was marked inactive after OST failure
6136 - call mds_fs_cleanup before unmounting filesystem (524)
6137 - fix races between taking ns_lock and ldlm_lock_change_resource
6138 - fix races updating LOV export open file list
6139 - fix lov_enqueue error path, avoid decref-ing bad lock handle
6140 - fix recovery NULL deref in ldlm_cli_cancel_unused
6141 - fix some DLM races by using new hash table for lock handles (419)
6142 - permit the client to specify desired inodes, at replay
6143 - duplicate requests when we queue them for replay reintegration
6144 - fix last_rcvd offset calculation
6145 - sync after each recovered transaction, so we always make progress
6146 - never, not always, ERESTART requests without transnos
6147 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
6149 - skip replay if the MDS says that the client is already connected
6150 - don't check for a recovery-enabled export to match lctl's UUID
6151 - don't INC_USE_COUNT for phantom exports
6152 - don't crash when cleaning up phantom exports (567)
6153 - don't double-finish or set replay data for errored mdc_open requests
6154 - abort requests when they time out, so we don't get old replies
6155 - send/receive replies for AST messages again
6156 - if the client says that it doesn't have the lock, cancel it on the
6158 - if we timeout during I/O, don't try to cancel an in-use lock; instead
6159 mark it as destroyed, it will all work out when decref is called
6160 - fix module use counts (22, 581)
6162 - ASTs now expect a reply (server cancels lock on error reply)
6164 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
6167 - fix many simultaneous client startup (392)
6168 - fix dentry->d_it clobbering
6169 - credentials weren't being shipped for readdir/getattr operations
6170 - remove invalid assertions triggered during some concurrent MD
6172 - proper Lustre versions added (336, 389)
6173 - fix memory leak for create error case (398)
6174 - fix LOV locking bug that would get cli/srv out of sync
6175 - fix echo client over LOV (409)
6176 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
6177 - fix double-O_EXCL intent crash (424)
6178 - avoid sending multiple lock CANCELs (352)
6180 - MDS can do multi-client recovery (modulo bugs in new code)
6182 - many updates, edits, cleanups
6184 2002-11-18 Phil Schwan <phil@clusterfs.com>
6187 - fix null d_it dereference (346)
6188 - fix full OST/dbench hang (333)
6189 - fix permission problem with file removal (286)
6190 - fix removal of OSCs from LOV when they fail
6191 - fix NULL deref during bulk timeout (214)
6192 - fix problems related to multiple filesystems on one MDS (241)
6193 - fixed serious subtle metadata locking bugs
6194 - free locks on clients when inodes are removed due to memory
6196 - fix inode pointer in lock data (285)
6197 - partial support for multiple MDS on a single host (241)
6198 - data locks weren't cancelled at clear_inode time (290, 311)
6199 - intent locks could lead to unbounded lock growth (205)
6200 - added a maximum lock count, an LRU list, and a flusher
6201 - fix multiple rename (365)
6202 - properly abstracted the echo client
6203 - OSC locked 1 byte too many; fixed
6204 - rewrote brw callback code:
6205 - fixed recovery bugs related to LOVs (306)
6206 - fixed too-many-pages-in-one-write crash (191)
6207 - fixed (again) crash in sync_io_timeout (214)
6208 - probably fixed callback-related race (385)
6210 - Add capability to MDS protocol
6211 - LDLM cancellations and callbacks on different portals
6213 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
6216 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
6217 - defer open object destruction to close time (601981, 138)
6218 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
6219 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
6220 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
6221 - don't call obd_setattr in ll_file_release for destroyed objects
6222 * protocol change to lustre_msg: move |version| and add |flags|
6223 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
6224 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
6225 * added replay of create, unlink, link and rename operations during
6226 MDS failover; recovery should be much more robust now
6227 * remove failed OSCs from LOVs (only lov_create uses this so far)
6228 * the lustre-HOWTO was brought (more) up to date (582544)
6230 2002-10-23 Phil Schwan <phil@clusterfs.com>
6233 - in-use dentries weren't being reused properly (617851)
6234 - prevent multiple LDLM setup (599178)
6235 - fix LOV size calculations for truncate (617853)
6236 - fix client handling of MDS intent errors (POSIX)
6237 - fix permission bug in lovstripe.c test (624321)
6238 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
6239 - truncate past end of file could corrupt data
6240 - proper cleanup after timeouts, crashes, etc (592524, 550815)
6241 - a race in recovery could return ETIMEDOUT to apps (623947)
6242 - building outside the source directory was fixed
6243 * the lustre-HOWTO was brought (more) up to date (582544)
6244 * major progress was made on recovery functionality
6246 2002-10-10 Phil Schwan <phil@clusterfs.com>
6249 - recovery deadlock fix
6250 - rm -rf causes LBUG fix (617817)
6251 - file open by multiple tasks fix (618962)
6252 - directory permissions bugs (602707 and 620007)
6253 - journal_stop fixed with locking (611313)
6254 - O_APPEND failures resolved (618273, perhaps 614459)
6255 - lconf PATH fix (619770)
6256 - IA64 build fix (621450)
6257 - RPC buffer sizes scale with amount of memory
6259 2002-10-01 Phil Schwan <phil@clusterfs.com>
6262 - locks would be cancelled without throwing away data pages,
6263 resulting in inconsistent data (605627)
6264 - inode attributes were not always being refreshed (605627, 612449)
6265 - lconf now continues to cleanup after lctl reports an error
6266 - MDS now enforces user permissions (602707)
6267 - lprocfs cleanup fixed, but not yet enabled (614157)
6268 - fixed infinite server hang, should a client not respond to an AST
6269 - avoid going into recovery if user calls readlink() with a buffer
6270 that's too small (613941)
6271 - AST RPCs no longer require replies (614867) -- this may be changed
6272 - don't crash server if client sends an IOV that's too big (611336)
6273 - fixed lock conversion deadlock (611892)
6274 - fixed the following of symlinks (614622)
6275 * recovery: the server can remove locks from a client that dies, other
6276 clients can make progress
6277 * more extN patch fixes
6278 * compile-time configurable ptlrpc buffer allocations
6280 - collaborative read cache document
6281 - Lustre Lite Performance CDR document-in-progress
6283 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6286 - fix typo in patch-2.4.18
6288 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6291 - clear ptlrpc request each time in handle_incoming_request()
6292 - unlink of files now destroys the object on the OST
6294 2002-09-19 Peter Braam <braam@clusterfs.com>
6296 * add hard link support
6297 * change obdfile creation method
6298 * kernel patch changed
6300 2002-09-19 Peter Braam <braam@clusterfs.com>
6303 - stack overflow bug in extN fixed
6305 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
6307 * documentation updates
6308 - add man pages for config tools
6309 - update tests/README to describe testing with new config tools
6310 - finish metadata API descriptions
6311 * bug fixes and cleanups
6312 - statfs workaround for 16TB limit
6313 - LOV stripe allocation improved, can stripe on subset of OSTs
6314 - LOV file size/IO offset was wrong for files > 4GB in size
6315 - object EA data was being dropped, caused files to be unreadable
6316 - memory overflow with non-LOV OST caused memory corruption
6317 - fixed regression tests to work with new config tools, obdfilter
6318 - fixed bug when directory size became larger than 1 block
6319 - fixed bug (for single client case) when PWD was deleted
6320 - invalidate local directory pages when doing intent-based ops
6321 - avoid LDLM oops when lock callback contained bad data
6323 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
6325 * documentation updates
6326 * bug fixes and cleanups
6327 - configuration tools
6330 - 64-bit compile warnings
6331 - 64-bit internal statfs data
6333 * test_brw on persistent OST devices
6335 * lprocfs (disabled)
6337 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
6339 * documentation updates
6340 * bug fixes and cleanups
6341 * configuration tools
6343 2002-08-30 Peter J. Braam <braam@clusterfs.com>
6346 * many small fixes to 0_5_4
6347 * io/network handling
6348 * thinkos in MDS operations
6350 2002-08-24 Peter J. Braam <braam@clusterfs.com>
6353 * crucial basic fixes to 0.5.3
6354 * IOR, Iozone work over Elan
6357 2002-08-07 Phil Schwan <phil@clusterfs.com>
6358 * version 0_5_3, our first alpha
6359 * we use the new Portals iovs
6360 * documentation updates
6361 * bug fixes and cleanups
6362 * small changes in the DLM wire protocol
6364 2002-07-25 Peter J. Braam <braam@clusterfs.com>
6365 * version 0_5_1 with some initial stability,
6366 * locking on MD and file I/O.
6367 * documentation updates
6368 * several bug fixes since 0.5.0
6369 * small changes in wire protocol
6371 2002-07-18 Phil Schwan <phil@clusterfs.com>
6373 * delivered as Lustre Light Alpha
6374 * fixed a crash after handling invalid MDS requests
6375 * fixed directory pages for architectures with non-4k pages sizes
6377 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
6378 * release version v0_4_4
6379 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
6381 * Fixes a number of interruption problems with OST operations.
6382 * Update documentation for portals header changes
6383 * Move all wire protocol structs/defines to lustre_idl.h
6384 * Fixes symlink length bug.
6385 * Add tcpdump to repository.
6387 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
6388 * release version v0_4_3
6389 * Fixes statfs for inodes on extN.
6390 * Fixes bug in runtests which would delete /etc/hosts.
6391 * Use 64-bit object IDs wherever possible (not into VFS though)
6392 Remove ost_get_info, which is unused by lustre, and out of date.
6394 2002-07-03 Peter Braam <braam@clusterfs.com>
6395 * release version v0_4_2 Fixes a lookup error (type not passed)
6396 * move forward to head of Portals
6397 * move forward to latest Lustre kernel
6399 2002-06-25 Peter Braam <braam@clusterfs.com>
6400 * release version v0_4_1. Hopefully stable on single node use.