3 * See http://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
4 * Server support for kernels:
5 2.6.32-220.13.1.el6 (RHEL6)
6 * Client support for unpatched kernels:
7 2.6.32-220.13.1.el6 (RHEL6)
9 2.6.18-238.19.1.el5 (RHEL5)
10 2.6.16.60-0.69.1 (SLES10)
11 * Recommended e2fsprogs version: 1.41.90.wc4
12 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
13 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
14 stack overflow. For more information, please refer to bugzilla 17630.
15 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
16 the re-exporting nodes to be mounted with "32bitapi" mount option
18 Severity : enhancement
20 Description: Update RHEL6.2 kernel to 2.6.32-220.13.1.el6
22 Severity : enhancement
24 Description: Update RHEL6.2 kernel to 2.6.32-220.7.1.el6
26 --------------------------------------------------------------------------------
28 04-30-2012 Whamcloud, Inc.
30 * See http://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
31 * Server support for kernels:
32 2.6.32-220.4.2.el6 (RHEL6)
33 * Client support for unpatched kernels:
34 2.6.32-220.4.2.el6 (RHEL6)
36 2.6.18-238.19.1.el5 (RHEL5)
37 2.6.16.60-0.69.1 (SLES10)
38 * Recommended e2fsprogs version: 1.41.90.wc4
39 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
40 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
41 stack overflow. For more information, please refer to bugzilla 17630.
42 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
43 the re-exporting nodes to be mounted with "32bitapi" mount option
45 Severity : enhancement
47 Description: Update RHEL6.2 kernel to 2.6.32-220.4.2.el6
49 Severity : enhancement
51 Description: Update RHEL6.2 kernel to 2.6.32-220.4.1.el6
53 Severity : enhancement
55 Description: Update RHEL6.2 kernel to 2.6.32-220.el6
57 Severity : enhancement
59 Description: Update RHEL6.1 kernel to 2.6.32-131.17.1.el6
61 --------------------------------------------------------------------------------
63 09-30-2011 Whamcloud, Inc.
65 * Server support for kernels:
66 2.6.18-238.19.1.el5 (RHEL5)
67 2.6.32-131.6.1.el6 (RHEL6)
68 * Client support for unpatched kernels:
69 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
70 2.6.18-238.19.1.el5 (RHEL5)
71 2.6.32-131.6.1.el6 (RHEL6)
72 2.6.29.4-167.fc11 (FC11)
73 2.6.16.60-0.69.1 (SLES10)
75 * Recommended e2fsprogs version: 1.41.90.wc3
76 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
77 Lustre filesystem with 4K stack may cause a stack overflow. For more
78 information, please refer to bugzilla 17630.
79 * NFSv4 reexport requires Lustre client on the re-exporting nodes to be
80 mounted with "32bitapi" mount option
82 Severity : enhancement
83 Description: Update RHEL6.1 kernel to 2.6.32-131.6.1.el6
85 Severity : enhancement
86 Description: Update RHEL5.6 kernel to 2.6.18-238.19.1.el5
88 Severity : enhancement
89 Description: Update RHEL5.6 kernel to 2.6.18-238.9.1.el5
92 Frequency : only with SLES10
94 Description: Use OFED "KMP" provided by Novell
95 Details : SLES10 SP3 ships with OFED in a separate "KMP" package.
96 Lustre is now built against this package. That means you need to
97 install the ofed-kmp package from Novell for the patchless client.
98 Note that the ofed-kmp that Novell ships may not exactly match the
99 kernel version but should still be compatible.
101 Severity : enhancement
103 Description: Update RHEL5.5 kernel to 2.6.18-194.17.1.el5
104 Update OEL5.5 kernel to 2.6.18-194.17.1.0.1.el5.
106 --------------------------------------------------------------------------------
108 2010-07-15 Oracle, Inc.
110 * Server support for kernels:
111 2.6.18-164.11.1.el5 (RHEL 5)
112 2.6.18-164.11.1.0.1.el5 (OEL 5)
113 * Client support for unpatched kernels:
114 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
115 2.6.18-164.11.1.el5 (RHEL 5),
116 2.6.18-164.11.1.0.1.el5 (OEL 5)
117 2.6.16.60-0.42.8 (SLES 10),
119 2.6.29.4-167.fc11 (FC11)
120 * Recommended e2fsprogs version: 1.41.10-sun2
122 Severity : enhancement
124 Description: Distribute lfs_migrate script with lustre
125 Details : The lfs_migrate mentioned in the manual is now included in the
126 lustre tree in lustre/scripts and will distributed in lustre RPMs.
128 Severity : enhancement
130 Description: Failover nid list added to import proc file.
132 Severity : enhancement
134 Description: "som_preview" mount option required to use SOM
135 Details : The clients will need to use som_preview mount option
139 Description: Lustre does not do 1MB IOs to HW RAID
140 Details : Bump MAX_PHYS/HW_SEGMENTS and SG_ALL to 256 in the RHEL5 kernel.
142 Severity : enhancement
144 Description: lfs quota output is not very convenient for awk/sed-parsing
145 Details : Some positions in lfs quota output table could be empty or
146 non-empty which made it hard to parse it with scripts, now
147 a dash is put instead of space where there is not supposed
152 Description: Update to OFED 1.5.1
153 Details : OFED 1.5.1 will be used in environments that don't use
154 vendor supplied OFED.
158 Description: Remove lnet. prefix from lctl params display
159 Details : Also renamed lnet.memused to lnet_memused to
164 Description: Use conf_param -d to remove permanant settings.
168 Description: Add list_param -R
169 Details : lctl list_param now supports recursive mode.
172 Bugzilla : 21829, 22312
173 Description: Fix llobdstat.
174 Details : 1. Now updating the write_bytes counter
175 2. Port script fixes from 1.8.
179 Description: decrease the usage of memory on clients.
180 Details : 1. On clients, recycle dentries and inodes unused.
181 2. Delete the code related to ll_deathrow(att 6215 in bug 1443). It
186 Description: llobdstat fix and enhancement
187 Details : add a counter to set a limit to how many samples will be returned
188 fix a wildcard in the path to limit to obdfilter stats only
192 Description: lfs check only allowed for root
193 Details : Code cleanups and added a regresion test of non-root use of
198 Description: MDS Oopses in ptlrpc_dispatch_difficult_reply()/__list_add()
199 Details : Fix hr->hr_index overload.
203 Description: Fix FIEMAP functionality.
207 Description: FC11 patchless client support added
211 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and
212 OEL5.4 kernel to 2.6.18-164.11.1.0.1.el5.
216 Description: New acc-sm script and various test-framework enhancments.
220 Description: Previously landed patch was reverted.
224 Description: Update SLES10 kernel to 2.6.16.60-0.42.8 (patchless)
228 Description: Add ldiskfs maxdirsize mount option.
232 Description: Fix mkfs.lustre support for 16TB LUNs.
236 Description: Print mmp_check_interval
237 Details : print mmp_check_interval and make it possible to abort
238 mount operation in case it takes too long.
242 Description: Add cascading_rw to lustre mpi tests.
246 Description: Store SOM data in EA.
247 Details : Also several SOM related cleanups.
251 Description: Add refcount to md_open_data
255 Description: Use refcount on lov_request_set to prevent memory corruption.
259 Description: Statahead: use parent dir's i_mutex for synchronization.
263 Description: loadgen improvements.
267 Description: Enable peer health detection in ksocklnd and ko2iblnd.
271 Description: Control DCACHE_LUSTRE_INVALID flag with MDS_INODE_LOOKCUP lock
275 Description: Fix race on obd_devcie:md_stats
279 Description: Fix SLES 11 client builds.
280 Details : Resolved uninitalized pointer gcc warnings.
284 Description: Allow non-root access for "lfs check".
285 Details : Added a check in obd_class_ioctl() for OBD_IOC_PING_TARGET.
289 Description: Clear imp_force_reconnect correctly in ptlrpc_connect_interpret()
291 Severity : enhancement
293 Description: Update kernel to RHEL5.4 2.6.18-164.6.1.el5 and
294 OEL5 2.6.18-164.6.1.0.1.el5(Both in-kernel OFED enabled).
298 Description: Sanity test_27p fails.
299 Details : OBD_FAIL_OST_ENOSPC has been moved to filter_statfs().
300 Sanity tests 27 were cleaned up and improved.
304 Description: Unify req format on client/servers
305 Details : Use new req_capsule API [almost] everywhere instead of old PTLRPC
306 buffers and swabbers approach..
311 Description: drop join file stuff.
313 Severity : enhancement
315 Description: Add 2.6.27 and SLES11 (patchless client) support
319 Description: Fix in ptlrpc_expire_one_request() to print the signed time
322 Severity : enhancement
324 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
327 Severity : enhancement
329 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
331 Severity : enhancement
333 Description: correctly handle big reply message.
334 Details : send LNet event if reply is bigger then buffer and adjust this buffer
339 Description: Drop unnecessary __GFP_NOMEMALLOC flag from filter_get_page()
341 Severity : enhancement
343 Description: Add support for OEL5.
347 Description: Conf-sanity.sh 50g test - deactivated OST should not cause a panic.
349 Severity : enhancement
351 Description: "weak-modules" support
352 Details : Implement "weak-modules" support which enables kernel modules
353 to be used with any kernel that implements the same kABI. In
354 order to achieve this modules are now installed in
355 /lib/modules/$(uname -r)/updates/kernel on all distributions.
359 Description: client could not reconnect to OST because of an active request.
360 Details : The client now retries to reconnect to the same server, if a
361 connect request failed with EBUSY or -EAGAIN. Test case 35b
362 conf-sanity.sh has been created.
366 Description: Don't allow make backward step on assignin osc next id.
367 Details : race between allocation next id and ll_sync thread can be cause
368 of set wrong osc next id and can be kill valid ost objects.
372 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
376 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
380 Description: Error handling in osc_statfs_interpret() has been improved.
381 Details : Check in osc_statfs_interpret() for EBADR.
385 Description: Do not send statfs() requests to OSTs disabled by administrator.
386 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
388 Severity : enhancement
390 Description: Added sanity-quota test to for lfs(1) to work with numeric uid/gid.
392 Severity : enhancement
394 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
396 Severity : enhancement
398 Description: Enhancement for lfs(1) command to use numeric uid/gid.
402 Description: Improved cleanup procedure for conf-sanity test_46a.
406 Description: Do not start bulk transfer, if client tried to reconnect.
407 Details : Do not start bulk transfer, if client tried to reconnect,
408 but failed due to active rpcs. Check for exp_abort_active_req
409 has been implemented.
413 Description: Prevent inconsistences between linux and lustre mount structures.
414 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
415 Make the sleep interruptible.
417 Severity : enhancement
419 Description: provide server to client comms path
420 Details : server to client communications path via new LDLM_SET_INFO rpc
424 Description: truncate starts GFP_FS allocation under transaction causing deadlock
425 Details : ldiskfs_truncate calls grab_cache_page which may start page
426 allocation under an open transaction. This may lead to
427 calling prune_icache with consequent lustre reentrance.
431 Description: send correctly lsm on open replay
432 Details : MDS is trust to LSM size on replay open, but client can set wrong size
437 Description: enable client interface failover
438 Details : When a child reconnects from another NID, properly update export
439 nid hash position and ldlm reverse import.
441 Severity : enhancement
443 Description: Communicate OST degraded/readonly state via statfs to MDS
444 Details : Flags in the statfs returned from OSTs indicate whether the
445 OST is in a degraded RAID state, or if the filesystem has
446 turned read-only after a filesystem error is detected.
451 Description: don't panic if EPROTO was hit when reading symlink
452 Details : correctly handling request reference in error cases.
454 Severity : enhancement
456 Description: Add LustreNetLink, a kernel-userspace communcation path.
458 Severity : enhancement
460 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
465 Description: access to llog context before init.
466 Details : move handling CATALOGS file at osc layer and forbid access to llog
470 Frequency : rare, only when client is writing to OST on local node
472 Description: Avoid deadlock for local client writes
473 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
474 memory freeing context. This allows OST threads to set the
475 PF_MEMALLOC flag on task structures in order to allocate memory
476 from reserved pools and complete IO.
477 Use GFP_HIGHUSER for OST allocations for non-local client writes,
478 so that the OST threads generate memory pressure and allow
479 inactive pages to be reclaimed.
481 Severity : enhancement
483 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
488 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
489 Details : move ldlm namespace creation in setup phase to avoid grab
490 _lprocfs_lock with cli_sem held.
494 Description: Temporarily disable grant shrink.
495 Details : Disable the feature for debugging.
498 Frequency : only when formatting test filesystems
500 Description: Unable to run several mkfs.lustre on loop devices at the same time
501 Details : mkfs.lustre returns error 256 on the concurrent loop devices
502 formatting. The solution is to proper handle the error.
504 Severity : enhancement
506 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
508 Severity : enhancement
510 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
515 Description: resolve race between obd_disconnect and class_disconnect_exports
516 Details : if obd_disconnect will be called to already disconnected export he
517 forget release one reference and osc module can't unloaded.
519 Severity : enhancement
521 Description: Allow tuning service thread via /proc
522 Details : For each service a new
523 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
524 created that can be used to set min/max thread counts, and get the
525 current number of running threads.
530 Descriptoin: don't return error if have partially created objects for file.
531 Details : lov_update_create_set uses set->set_success as index for created
532 objects, so if some requests will be failed, they will have hole
533 at end of array and qos_shrink_lsm for allocate correct lsm.
535 Severity : enhancement
537 Description: Update OFED support to 1.4.1
542 Description: Reduce small size read RPC
543 Details : Set read-ahead limit for every file and only do read-ahead when
544 available read-ahead pages are bigger than 1M to avoid small size
547 Severity : enhancement
549 Description: move AT tunable parameters for more consistent usage
550 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
552 Severity : enhancement
554 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
555 Details : allow skip disconnected ost for send statfs request and hide error
561 Description: fix racy locking of mballoc block bitmaps causing BUG
562 Details : The locking of the mballoc buddy bitmap and the in-memory
563 block bitmap was using two different spin locks in some
564 cases. This made it possible to incorrectly access the
565 mballoc bitmap while another process was modifying it,
566 causing a sanity assertion to fail. While no on-disk corruption
567 was reported, there was some risk of this happening.
570 Frequency : rare, on llog test 6
572 Descriptoin: don't allow connect to already connected import
573 Details : allowing connect to already connected import is hide connecting problem.
578 Descriptoin: allow kill process which wait statahead result
579 Details : for some reasons 'ls' can stick in waiting result from statahead,
580 in this case need way for kill this process.
582 Severity : enhancement
584 Description: Add state history info file, enhance import info file
585 Details : Track import connection state changes in a new osc/mdc proc file;
586 add overview-type data to the osc/mdc import proc file.
588 Severity : enhancement
590 Description: MDS create should not wait for statfs RPC while holding DLM lock.
592 Severity : enhancement
594 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
599 Descriptoin: OST grant too much space to client even there are not enough space.
600 Details : Client will shrink its grant cache to OST if there are no write
601 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
602 this grant cache if there are already not enough avaible space
603 (left_space < total_clients * 32M).
606 Frequency : start MDS on uncleanly shutdowned MDS device
608 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
609 Details : stay in waiting mds<>ost recovery finished produce random bugs
610 due race between two ll_sync thread for one lov target. send
611 ACTIVATE event only if connect realy finished and import have
615 Frequency : rare, connect and disconnect target at same time
617 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
618 Details : don't call obd_disconnect under lov_lock. this long time
619 operation and can block ptlrpcd which answer to connect request.
624 Descriptoin: don't lose wakeup for imp_recovery_waitq
625 Details : recover_import_no_retry or invalidate_import and import_close can
626 both sleep on imp_recovery_waitq, but we was send only one wakeup
630 Frequency : always with long access acl
632 Descriptoin: mds can't pack reply with long acl.
633 Details : mds don't control size of acl but they limited by reint/getattr
636 Severity : enhancement
638 Description: Update to SLES10 kernel-2.6.16.60-0.33.
640 Severity : enhancement
642 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
645 Frequency : start MDS on uncleanly shutdowned MDS device
647 Descriptoin: aborting recovery hang on MDS
648 Details : don't throttle destroy RPCs for the MDT.
651 Frequency : on remount
653 Description: external journal device not working after the remount
654 Details : clear dev_rdonly flag for external journal devices in
660 Description: shutdown vs evict race
661 Details : client_disconnect_export vs connect request race.
662 if client will evicted at this time - we start invalidate
663 thread without referece to import and import can be freed
667 Frequency : rare, need acl's on inode.
669 Description: client can't handle ost additional correctly
670 Details : if ost was added after client connected to mds client can have
671 hit lnet_try_match_md ... to big messages to wide striped files.
672 in this case need teach client to handle config events about add
673 lov target and update client max ea size at that event.
675 Severity : enhancement
677 Description: Changelogs
678 Details : Changelogs are a lightweight mechanism to track filesystem
679 metadata and namespace changes. The changelog is recorded
680 permanently on the MDTs, and is periodically "consumed" / purged
681 when records are no longer needed.
683 Severity : enhancement
685 Description: compact fld format with extents
686 Details : Store range of seq rather than every seq in FLD. Seq
687 controller update FLD rather than clients. In Case of CMD, mdt0
688 has FLD, all other metadata server act as non persistent proxy
689 for FLD queries and cache fld entries in fld cache.
694 Description: don't skip ost target if they assigned to file
695 Details : Drop slow OSCs if we can, but not for requested start idx.
696 This means "if OSC is slow and it is not the requested
697 start OST, then it can be skipped, otherwise skip it only
698 if it is inactive/recovering/out-of-space.
702 Description: more cleanup in mds_lov
703 Details : not send LOV EA under replay, we can't know about they size at this
704 time. Don't allow client connect to mds before any ost connected,
705 for avoid problems with LOV EA size and returning EIO to client.
707 Severity : enhancement
709 Description: Interoperability at server side (Disk interoperability)
711 Severity : enhancement
713 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
715 Severity : enhancement
717 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
719 Severity : enhancement
721 Description: New client IO stack (CLIO).
723 Severity : enhancement
725 Description: Commit on sharing. Eliminate inter-client dependencies between
726 uncommitted transactions by doing transaction commits.
727 Thereby clients may recovery independently.
730 Frequency : Create a symlink file with a very long name
732 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
733 Details : If there is no extra space in the request for early cancels,
734 ldlm_req_handles_avail() returns 0 instead of a negative value.
736 Severity : enhancement
738 Description: Add /proc entry for import status
739 Details : The mdc, osc, and mgc import directories now have
740 an import directory that contains useful import data for debugging
743 Severity : enhancement
745 Description: Re-disable certain /proc logging
746 Details : Enable and disable client's offset_stats, extents_stats and
747 extents_stats_per_process stats logging on the fly.
750 Frequency : Only on FC kernels 2.6.22+
752 Description: oops in statahead
753 Details : Do not drop reference count for the dentry from VFS when lookup,
754 VFS will do that by itself.
756 Severity : enhancement
758 Description: Generic /proc file permissions
759 Details : Set /Proc file permissions in a more generic way to enable non-
760 root users operate on some /proc files.
764 Description: Hitting mdc_commit_close() ASSERTION
765 Details : Properly handle request reference release in
766 ll_release_openhandle().
770 Description: quota recovery deadlock during mds failover
771 Details : This patch includes att18982, att18236, att18237 in bz14840.
773 1. fix osts hang when mds does failover with quotaon
774 2. prevent watchdog storm when osts threads wait for the
779 Frequency : only patchless client
780 Description: add workaround for race between add/remove dentry from hash
782 Severity : enhancement
784 Description: Allow OST glimpses to return PW locks
788 Description: LBUG when llog conf file is full
789 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
793 Description: Prevent import from entering FULL state when server in recovery
797 Description: service mount cannot take device name with ":"
798 Details : Only when device name contains ":/" will mount treat it as
804 Description: replace ptlrpcd with the statahead thread to interpret the async
805 statahead RPC callback
809 Frequency : on recovery
810 Description: I/O failures after umount during fail back
811 Details : if client reconnected to restarted server we need join to recovery
812 instead of find server handler is changed and process self
813 eviction with cancel all locks.
815 Severity : enhancement
817 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
819 Severity : enhancement
821 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
823 Severity : enhancement
825 Description: Upcall on Lustre log has been dumped
826 Details : Allow for a user mode script to be called once a Lustre log has
827 been dumped. It passes the filename of the dumped log to the
828 script, the location of the script can be specified via
829 /proc/sys/lnet/debug_log_upcall.
834 Description: avoid idr_remove called for id which is not allocated.
835 Details : Move assigment s_dev for clustered nfs to end of initialization,
836 to avoid problem with error handling.
841 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
842 Details : When connection is reused this not moved from CONN_UNUSED_HASH
843 into CONN_USED_HASH and this prodice warning when put connection
844 again in unused hash.
847 Severity : enhancement
849 Description: File striping can now be set to use an arbitrary pool of OSTs.
851 Severity : enhancement
853 Description: Export bytes_read/bytes_write count on OSC/OST.
857 Description: Early reply size mismatch, MGC loses connection
858 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
859 the connect flags are properly negotiated.
864 Description: quotas are not honored with O_DIRECT
865 Details : all writes with the flag O_DIRECT will use grants which leads to
866 this problem. Now using OBD_BRW_SYNC to guard this.
870 Description: add quota statistics
871 Details : 1. sort out quota proc entries and proc code.
872 2. add quota statistics
874 Severity : enhancement
876 Description: enable quota support for HEAD.
880 Description: Properly propagate oinfo flags from lov to osc for statfs
881 Details : restore missing copy oi_flags to lov requests.
883 Severity : enhancement
885 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
886 : l_getgroups(8), lst(8), routerstat(8)
887 Details : included man pages for llobdstat(8), llstat(8),
888 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
890 Severity : enhancement
892 Description: configure's --enable-quota should check the
893 : kernel .config for CONFIG_QUOTA
894 Details : configure is terminated if --enable-quota is passed but
895 : no quota support is in kernel
899 Description: Remove portals compatibility
900 Details : Remove portals compatibility, not interoperable with releases
905 Description: Resolve device initialization race
906 Details : Prevent proc handler from accessing devices added to the
907 obd_devs array but yet be intialized.
909 Severity : enhancement
911 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
913 Severity : enhancement
915 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
920 Description: Using wrong pointer in osc_brw_prep_request
921 Details : Access to array[-1] can produce panic if kernel compiled with
922 CONFIG_PAGE_ALLOC enabled
926 Description: Client runs out of low memory
927 Details : Consider only lowmem when counting initial number of llap pages
931 Description: *optional* service tags registration
932 Details : if the "service tags" package is installed on a Lustre node
933 When the filesystem is mounted, a local-node service tag will
934 be created. See http://inventory.sun.com/ for more information
935 about the Service Tags asset management system.
939 Description: Kernel BUG tries to release flock
940 Details : Lustre does not destroy flock lock before last reference goes
941 away. So always drop flock locks when client is evicted and
942 perform unlock regardless of successfulness of speaking to MDS.
946 Description: add refcount protection for osc callbacks, avoid panic on shutdown
950 Description: sanity test 65a fails if stripecount of -1 is set
951 Details : handle -1 striping on filesystem in ll_dirstripe_verify
956 Description: ASSERTION(CheckWriteback(page,cmd)) failed
957 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
960 Severity : enhancement
962 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
966 Description: do not process already freed flock
967 Details : flock can possibly be freed by another thread before it reaches
968 to ldlm_flock_completion_ast.
972 Description: LBUG during stress test
973 Details : Need properly lock accesses the flock deadlock detection list.
977 Description: oops in page fault handler
978 Details : kernel page fault handler can return two special 'pages' in error
979 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
983 Description: timeout with invalidate import.
984 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
985 should be handled by ptlrpcd. This produce long age waiting and
986 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
988 Severity : enhancement
990 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
994 Description: enable MGS and MDT services start separately
995 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
996 a co-located MGS without starting the MGS, which is a complement
997 to 'nosvc' mount option.
1001 Frequency : after recovery
1002 Description: precreate to many object's after del orphan.
1003 Details : del orphan st in oscc last_id == next_id and this triger growing
1004 count of precreated objects. Set flag LOW to skip increase count
1005 of precreated objects.
1009 Frequency : rare, on clear nid stats
1010 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
1011 Details : when clean nid stats sometimes try destroy live entry,
1012 and this produce panic in free.
1016 Description: Stack overflow during MDS log replay
1017 ease stack pressure by using a thread dealing llog_process.
1021 Description: wait until IO finished before start new when do lock cancel.
1022 Details : VM protocol want old IO finished before start new, in this case
1023 need wait until PG_writeback is cleared until check dirty flag and
1024 call writepages in lock cancel callback.
1026 Severity : enhancement
1028 Description: using special macro for print time and cleanup in includes.
1032 Description: mds_mfd_close() ASSERTION(rc == 0)
1033 Details : In mds_mfd_close(), we need protect inode's writecount change
1034 within its orphan write semaphore to prevent possible races.
1038 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
1042 Frequency : rare, on shutdown ost
1043 Description: don't hit live lock with umount ost.
1044 Details : shrink_dcache_parent can be in long loop with destroy dentries,
1045 use shrink_dcache_sb instead.
1049 Description: don't panic with use echo client
1050 Details : echo client pass NULL as client nid pointer and this produce null
1051 pointer dereference.
1055 Description: fix build on ppc32
1056 Details : compile code with -m64 flag produce wrong object file for ppc32.
1060 Description: add message levels for liblustreapi
1064 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
1065 Details : Move the 'good_osts' check before the 'total_bavail' check. This
1066 will result in an -EAGAIN and in the exit call path we call
1067 alloc_rr() which will with increasing aggressiveness attempt to
1068 aquire precreated objects on the minimum number of required OSCs.
1072 Description: Use old size assignment to avoid deadlock
1073 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
1074 the scheduling while holding a spinlock. We do not need locking
1075 for size in ll_update_inode() because size is only updated from
1076 the MDS for directories or files without objects, so there is no
1077 other place to do the update, and concurrent access to such inodes
1078 are protected by the inode lock.
1082 Description: resolve "_IOWR redefined" build error on SLES10
1086 Description: dump the memory debugging after all modules are unloaded to
1087 suppress false negative in conf_sanity test 39
1089 Severity : enhancement
1091 Description: build kernel-ib packages for OFED 1.3 in our release cycle
1096 Description: fix SLES kernel versioning
1097 Details : the kernel version for our SLES 10 kernel did not include a "-"
1098 before the "smp" at the end. while this was not a problem in
1099 general, it did mean that software trying to use the kernel
1100 version to try to detect a vendor specific kernel would fail.
1101 this was most evident by the OFED build scripts.
1105 Description: Don't update lov_desc members until making sure they are valid
1106 Details : When updating lov_desc members via proc fs, need fix their
1107 validities before doing the real update.
1111 Description: don't put request into delay list while invalidate in flight.
1112 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
1113 import in flight. this produce timeout for invalidate and sometimes
1114 can cause stale data.
1118 Frequency : on ppc only
1119 Description: not convert ost objects for directory because it's not exist.
1120 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
1124 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
1126 Severity : enhancement
1128 Description: Update to support 2.6.22.14 vanilla kernel.
1132 Frequency : rare, on recovery
1133 Description: read procfs can produce deadlock in some situation
1134 Details : Holding lprocfs lock which send rpc can produce block for destroy
1135 obd objects and this also block reconnect with -EALREADY. This
1136 isn't fix all lprocfs bugs - but make it rare.
1138 Severity : enhancement
1140 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
1143 Frequency : frequent on X2 node
1145 Description: mdc_set_open_replay_data LBUG
1146 Details : Set replay data for requests that are eligible for replay.
1150 Description: lustre_mgs: operation 101 on unconnected MGS
1151 Details : When MGC is disconnected from MGS long enough, MGS will evict the
1152 MGC, and late on MGC cannot successfully connect to MGS and a lot
1153 of the error messages complaining that MGS is not connected.
1157 Frequency : on network error
1158 Description: panic with double free request if network error
1159 Details : mdc_finish_enqueue is finish request if any network error ocuring,
1160 but it's true only for synchronus enqueue, for async enqueue
1161 (via ptlrpcd) this incorrect and ptlrpcd want finish request
1164 Severity : enhancement
1166 Description: client-side metadata stat-ahead during readdir(directory readahead)
1167 Details : perform client-side metadata stat-ahead when the client detects
1168 readdir and sequential stat of dir entries therein
1171 Frequency : on start mds
1173 Description: Implement get_info(last_id) in obdfilter.
1176 Frequency : occasional
1178 Description: Correctly check stale fid, not start epoch if ost not support SOM
1179 Details : open with flag O_CREATE need set old fid in op_fid3 because
1180 op_fid2 was overwritten with new generated fid, but mds can answer
1181 with one of these two fids and both is not stale. Setattr
1182 incorrectly started an epoch and assume will be called
1183 done_writing, but without SOM done_writing ever being called.
1186 Frequency : rare, depends on device drivers and load
1188 Description: MDS or OSS nodes crash due to stack overflow
1189 Details : Code changes in 1.8.0 increased the stack usage of some functions.
1190 In some cases, in conjunction with device drivers that use a lot
1191 of stack the MDS (or possibly OSS) service threads could overflow
1192 the stack. One change which was identified to consume additional
1193 stack has been reworked to avoid the extra stack usage.
1196 Frequency : occasional
1198 Description: Do not fail import if osc_interpret_create gets -EAGAIN
1199 Details : If osc_interpret_create got -EAGAIN it immediately exits and
1200 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
1201 oscc_has_objects and see OSC has no objests and call
1202 oscc_internal_create to resend create request.
1204 Severity : enhancement
1206 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
1208 Severity : enhancement
1210 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
1213 Frequency : very rare
1215 Description: Fix replay if there is an un-replied request and open
1216 Details : In some cases, older replay request will revert the
1217 mcd->mcd_last_xid on MDS which is used to record the client's
1218 latest sent request.
1220 Severity : enhancement
1222 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
1224 Severity : enhancement
1226 Description: Add rhel5 support to HEAD.
1228 Serverity : enhancement
1230 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
1235 Description: Don't allow skipping OSTs if index has been specified.
1236 Details : Don't allow skipping OSTs if index has been specified, make
1237 locking in internal create lots better.
1241 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
1242 Details : access to bitfield in structure is always rounded to long
1243 and this produce problem with not atomic change any bit.
1247 Description: Lustre make rpms failed.
1248 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
1249 builds Lustre rpms from tarball.
1252 Frequency : rare on shutdown ost
1254 Description: If llog cancel was not send before clean_exports phase, this can
1255 produce deadlock in llog code.
1256 Details : If llog thread has last reference to obd and call class_import_put
1257 this produce deadlock because llog_cleanup_commit_master wait when
1258 last llog_commit_thread exited, but this never success because was
1259 called from llog_commit_thread.
1263 Description: allow userland application know is lost one of stripes.
1264 Details : fill lvb_blocks with error code on ost and return it to
1265 application if error flag found.
1269 Description: NULL lov_tgts causing MDS oops
1270 Details : more safe checks for NULL lov_tgts for avoid oops.
1272 Severity : enhancement
1274 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
1278 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
1279 Details : If an OST is down the MDS will hang indefinitely in
1280 obd_statfs_rqset() waiting for the statfs data. While for
1281 MDS QOS usage of statfs, it should not stuck in waiting.
1283 Severity : enhancement
1285 Description: remote_acl support
1286 Details : Support ACL-based permission check for remote user.
1287 Support setfacl/getfacl for remote user with the utils
1288 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
1289 the system "{s,g}etfacl" utils.
1291 Severity : enhancement
1293 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
1295 Severity : enhancement
1297 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
1301 Description: make mgs_setparam() handle fsname containing dash
1302 Details : fsname containing a dash does not work with lctl conf_param
1304 Severity : enhancement
1306 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
1308 Severity : enhancement
1310 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
1314 Frequency : rare, at shutdown
1315 Description: access already free / zero obd_namespace.
1316 Details : if client_disconnect_export was called without force flag set,
1317 and exist connect request in flight, this can produce access to
1318 NULL pointer (or already free pointer) when connect_interpret
1319 store ocd flags in obd_namespace.
1323 Frequency : only at startup
1324 Description: not alloc memory with spinlock held.
1325 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
1326 if any spinlock held.
1328 Severity : enhancement
1330 Description: make lustre randomly fail allocating memory
1331 Details : Make lustre randomly failed allocating memory for testing purpose.
1333 Severity : enhancement
1335 Description: lost problems with lov objid file
1336 Details : Fixes some scability and access to not inited memory problems
1337 in work with lov objdid file.
1342 Description: lfs find does not continue on file error
1343 Details : Continue other files processing when a file/dir is absent.
1347 Description: Inconsistent usage of lustre_pack_reply()
1348 Details : Standardize the usage of lustre_pack_reply() such that it
1349 always generate a CERROR on failure.
1352 Frequency : occasional
1354 Description: MDS hang or stay in waiting lock
1355 Details : If client receive lock with CBPENDING flag ldlm need send lock
1356 cancel as separate rpc, to avoid situation when cancel request
1357 can't processed due all i/o threads stay in wait lock.
1361 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1365 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1369 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1373 Description: Update to RHEL4 latest kernel.
1375 Severity : enhancement
1377 Description: Build SLES10 patchless client fails
1378 Details : The configure was broken by run ./configure with
1379 --with-linux-obj=.... argument for patchless client. When the
1380 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1381 can't search header adequately. Use absolute path such as
1382 -I($LINUX)/include instead.
1386 Description: interrupt oig_wait produce painc on resend.
1387 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1388 private set, and this produce situation when rq_ptlrpcd_data not
1389 copyed to new allocated request and triggered LBUG on assert
1390 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1391 ptlrpcd set if request is changed and can be safety changed to use
1396 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1397 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1398 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1399 these macros and replace them with their difinitions instead.
1403 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1404 Details : Update which_patch & target file for SLES10 latest kernel.
1407 Description: Remove client patches from SLES 10 kernel.
1408 Details : This causes SLES 10 clients to behave as patchless clients
1409 even on a Lustre-patched (server) kernel.
1411 Severity : enhancement
1413 Description: self-adjustable client's lru lists
1414 Details : use adaptive algorithm for managing client cached locks lru
1415 lists according to current server load, other client's work
1416 pattern, memory activities, etc. Both, server and client
1417 side namespaces provide number of proc tunables for controlling
1420 Severity : enhancement
1422 Description: light-weight GSS support
1423 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
1424 and reduce performance overhead.
1426 Severity : enhancement
1428 Description: Linux keyring support
1429 Details : Support using service of Linux keyring for Lustre GSS internal
1430 context refresh/cache mechanism.
1434 Description: Fix errors in lfs documentation
1435 Details : Fixes man pages
1439 Description: don't use GFP_* in generic Lustre code.
1440 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
1444 Description: obdclass is limited by single OBD_ALLOC(idarray)
1445 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
1449 Description: warnings with build patchless client with vanila 2.6.19 and up
1450 Details : change old ctl_table style and replace ctl_table/ctl_table_header
1451 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
1455 Frequency : Only for SLES
1456 Description: sanity_quota fail test_1
1457 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
1458 file, which makes TSTID[2] inunique.
1462 Frequency : Only for SLES9
1463 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
1464 Details : Update target/ChangeLog/which_patch .
1468 Frequency : Only for test_10 in sanity-quota.sh
1469 Description: a bug in quota test code
1470 Details : checking a test flag in a wrong place
1474 Frequency : Only for test_13 in sanity-quota.sh
1475 Description: a bug in quota test script
1476 Details : shouldn't check overfull allocated quota
1480 Description: jbd statistics
1481 Details : Port older jbd statistics patch for sles10
1483 Severity : enhancement
1485 Description: Add printing inode info into message about error in writepage.
1490 Description: reply_lock_interpret crash due to race with it and lock cancel.
1491 Details : Do not replay locks that are being cancelled. Do not reference
1492 locks by their address during replay, just by their handle.
1496 Frequency : When flocks are used.
1497 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
1498 Details : Flock locks might destroy just granted lock if it could be merged
1499 with another existing flock, this is done in completion handler,
1500 so teach ldlm_cli_enquque_fini that this is a valid case for
1506 Description: Oops in read and write path when failing to allocate lock.
1507 Details : Check if lock allocation failed and return error back.
1511 Description: lstripe command fails for valid OST index
1512 Details : The stripe offset is compared to lov->desc.ld_tgt_count
1513 instead of lov->desc.ld_active_tgt_count.
1517 Description: sanity.sh failed test 103
1518 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
1523 Description: df doesn't work properly if diskfs blocksize != 4K
1524 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
1528 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
1529 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
1531 Severity : enhancement
1533 Description: Adaptive timeouts
1534 Details : RPC timeouts adapt to changing server load and network
1535 conditions to reduce resend attempts and improve recovery time.
1539 Description: llapi_file_create() does not allow some changes
1540 Details : add llapi_file_open() that allows specifying the mode and
1541 open flags, and also returns an open file handle.
1545 Description: merge and cleanup kernel patches.
1546 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
1550 Description: Add journal checksum support.(Kernel part)
1551 Details : The journal checksum feature adds two new flags i.e
1552 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
1553 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
1554 indicates that the commit block contains the checksum for
1555 the blocks described by the descriptor blocks. Now commit
1556 record can be sent to disk without waiting for descriptor
1557 blocks to be written to disk. This behavior is controlled
1558 using JBD2_FEATURE_ASYNC_COMMIT flag.
1562 Description: OSS needs mutliple precreate threads
1563 Details : Add ability to start more than one create thread per OSS.
1567 Description: Sanity.sh test_65k failed: setstripe should have succeeded
1568 Details : execute lfs setstripe on client
1572 Description: mds_obd_create error creating tmp object
1573 Details : When the user sets quota on root, llog will be affected and can't
1574 create files and write files.
1577 Frequency : Always on ia64 patchless client, and possibly others.
1579 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
1580 Details : This allows the patchless client to be loaded on architectures
1581 without this export.
1585 Description: RedHat Update kernel for RHEL5
1586 Details : Add the kernel config file for RHEL5.
1590 Description: RedHat Update kernel for RHEL5
1591 Details : Modify the kernel config file more closer RHEL5.
1595 Description: Build failure against Centos5 (RHEL5)
1596 Details : Define PAGE_SIZE when it isn't present.
1600 Description: test_7 on sanity-quota.sh can't be used on separate machines
1601 Details : add facet to handle it
1605 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
1606 Details : with NFS, the anon dentry's parent was set to itself in
1607 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
1608 corresponding dentry other than use rec->ur_name.
1610 Severity : enhancement
1612 Description: lfs setstripe enhancement
1613 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
1617 Description: enable data checksumming by default
1618 Details : enable checksum by default, allow --disable-checksum
1619 configure option and "-o nochecksum" mount option. Checksums
1620 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
1624 Description: lustre support for RHEL5
1625 Details : Add support for RHEL5.
1629 Description: Client eviction due to failover config
1630 Details : after a connection loss, the lustre client should attempt to
1631 reconnect to the last active server first before trying the
1632 other potential connections.
1636 Description: when mds and osts use different quota unit(32bit and 64bit),
1637 quota will be released repeatly.
1638 Details : void sending multiple quota reqs to mds, which will keep the
1639 status between the reqs.
1643 Description: osts not allocated evenly to files
1644 Details : change the condition to increase offset_idx
1647 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
1649 Description: Data corruption for OSTs that are formatted larger than 2TB
1651 Details : When generating the bio request for lustre file writes the
1652 sector number would overflow a temporary variable before being
1653 used for the IO. The data reads correctly from Lustre (which
1654 will overflow in a similar manner) but other file data or
1655 filesystem metadata may be corrupted in some cases.
1659 Description: Tune the kernel for good SCSI performance.
1660 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
1661 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
1666 Description: rewrite ext2-derived code in obdclass/uuid.c
1667 Details : rewrite inherited code (uuid parsing code from ext2 utils)
1668 from scratch preserving functionality.
1672 Description: "lfs find -obd UUID" prints directories
1673 Details : "lfs find -obd UUID" will return all directory names instead
1674 of just file names. It is incorrect because the directories
1675 do not reside on the OSTs.
1679 Description: use i_size_read and i_size_write in 2.6 port
1680 Details : replace inode->i_size access with i_size_read/write()
1683 Frequency : when removing large files
1685 Description: scheduling issue during removal of large Lustre files
1686 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
1687 It causes scheduling issues when removing large files (17TB in the
1691 Frequency : only with liblustre clients on XT3
1693 Description: evictions taking too long
1694 Details : allow llrd to evict clients directly on OSTs
1697 Frequency : only on ppc
1699 Description: /proc/fs/lustre/devices broken on ppc
1700 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
1701 We should make sure the type of 'index' is loff_t and then cast
1702 explicitly as needed below. Do not assign an explicitly cast
1706 Frequency : only for rhel5
1708 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
1709 Details : Modify the target file & which_kernel.
1711 Severity : enhancement
1713 Description: omit set fsid for export NFS
1714 Details : fix set/restore device id for avoid EMFILE error and mark lustre
1715 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
1719 Frequency : Always, for kernels after 2.6.16
1720 Description: Fix warning idr_remove called for id=.. which is not allocated.
1721 Details : Last kernels save old s_dev before kill super and not allow
1722 to restore from callback - restore it before call kill_anon_super.
1726 Description: buffer overruns could theoretically occur
1727 Details : llapi_semantic_traverse() modifies the "path" argument by
1728 appending values to the end of the origin string, and a
1729 overrun may occur. Adding buffer overrun check in liblustreapi.
1733 Description: Fix error on 'ls .' at the top of the Lustre mount.
1734 Details : Don't revalidate dentry if it is a root dentry.
1738 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
1739 Details : Modify vm-tunables-rhel4.patch.
1743 Description: Kernel config for 2.6.18-vanilla.
1744 Details : Modify targets/2.6-vanilla.target.in.
1745 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
1746 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
1747 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
1748 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
1752 Description: improve handling recoverable errors
1753 Details : if request processig with error which can be recoverable on server
1754 request should be resend, otherwise page released from cache and
1759 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
1760 Details : Modify target file & which_patch.
1761 A flaw was found in the IA32 system call emulation provided
1762 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
1763 value could be stored in the %RAX register, which could trigger an
1764 out-of-bounds system call table access. An untrusted local user
1765 could exploit this flaw to run code in the kernel
1766 (ie a root privilege escalation). (CVE-2007-4573).
1770 Description: O_DIRECT bypasses client statistics.
1771 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
1772 read_ahead_stats were updated. copied the stats section from
1773 osc_send_oap_rpc() into async_internal().
1777 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
1781 Description: change order of libsysio includes
1782 Details : '#include sysio.h' should always come before '#include xtio.h'
1786 Description: handle "serious error: objid * already exists" more gracefully
1787 Details : If LAST_ID value on disk is smaller than the objects existing in
1788 the O/0/d* directories, it indicates disk corruption and causes an
1789 LBUG(). If the object is 0-length, then we should use the existing
1790 object. This will help to avoid a full fsck in most cases.
1792 Severity : enhancement
1794 Description: adapt the lustre_config script to support the upgrade case
1795 Details : Add "-u" option for lustre_config script to support upgrading 1.4
1796 server targets to 1.6 in parallel.
1800 Description: To avoid grant space > avaible space when the disk is almost
1801 full. Without this patch you might see the error "grant XXXX >
1802 available" or some LBUG about grant, when the disk is almost
1804 Details : In filter_check_grant, for non_grant cache write, we should
1805 check the left space by if (*left > ungranted + bytes), instead
1806 of (*left > ungranted), because only we are sure the left space
1807 is enough for another "bytes", then the ungrant space should be
1808 increase. In client, we should update cl_avail_grant only there
1809 is OBD_MD_FLGRANT in the reply.
1813 Description: Update RHEL 4 kernel to fix local root privilege escalation.
1814 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
1815 described in CVE-2007-4573. This problem could allow untrusted
1816 local users to gain root access.
1819 Frequency : when using O_DIRECT and quotas
1821 Description: Incorrect file ownership on O_DIRECT output files
1822 Details : block usage reported by 'lfs quota' does not take into account
1823 files that have been written with O_DIRECT.
1828 Description: touch file failed when fs is not full
1829 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
1830 otherwise we can get ENOSP while fs is not full.
1834 Description: parallel lock callbacks
1835 Details : Instead of sending blocking and completion callbacks as separated
1836 requests, adding them to a set and sending in parallel.
1839 Frequency : only for Cray XT3
1840 Bugzilla : 12829/13455
1841 Description: Changing primary group doesn't change the group lustre assigns to
1843 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
1844 overridden with the primary group provided by the group upcall,
1845 whereas the supplied fsgid can be trusted if it is in the list of
1846 supplementary groups returned by the group upcall.
1848 Severity : enhancement
1850 Description: Allow masking D_WARNING, D_ERROR messages from console
1851 Details : Console messages can now be disabled via lnet.printk.
1855 Description: User code with malformed file open parameter crashes client node
1856 Details : Before packing join_file req, all the related reference should be
1857 checked carefully in case some malformed flags cause fake
1858 join_file req on client.
1862 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
1863 Details : ldlm_completion_ast() assumes that a lock is granted when the req
1864 mode is equal to the granted mode. However, it should also check
1865 that LDLM_FL_CP_REQD is not set.
1869 Description: Heavy nfs access might result in deadlocks
1870 Details : After ELC code landed, it is now improper to enqueue any mds
1871 locks under och_sem, because enqueue might want to decide to
1872 cancel open locks for same inode we are holding och_sem for.
1876 Description: Client eviction while running blogbench
1877 Details : A lot of unlink operations with concurrent I/O can lead to a
1878 deadlock causing evictions. To address the problem, the number of
1879 oustanding OST_DESTROY requests is now throttled to
1880 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
1881 callbacks are processed in priority.
1885 Description: enable ACLs on MDS by default
1886 Details : ACLs must be enabled on MDS by default.
1889 Frequency : PPC/PPC64 only
1891 Description: conflicts between asm-ppc64/types.h and lustre_types.h
1892 Details : fix duplicated definitions between asm-ppc64/types.h and
1893 lustre_types.h on PPC.
1896 Frequency : PPC/PPC64 only
1898 Description: asm-ppc/segment.h does not exist
1899 Details : fix compile issue on PPC.
1903 Description: better handle error messages in extents code
1906 Frequency : RHEL4 only
1908 Description: mkfs is very slow on IA64/RHEL4
1909 Details : A performance regression has been discovered in the MPT Fusion
1910 driver between versions 3.02.73rh and 3.02.99.00rh. As a
1911 consequence, we have downgraded the MPT Fusion driver in the RHEL4
1912 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
1914 Severity : enhancement
1916 Description: SNMP support enhancement
1917 Details : Adding total number of sampled request for an MDS node in snmp
1920 Severity : enhancement
1922 Description: Optimize ldlm waiting list processing for PR extent locks
1923 Details : When processing waiting list for read extent lock and meeting read
1924 lock that is same or wider to it that is not contended, skip
1925 processing rest of the list and immediatelly return current
1926 status of conflictness, since we are guaranteed there are no
1927 conflicting locks in the rest of the list.
1931 Description: Time out and refuse to reconnect
1932 Details : When the failover node is the primary node, it is possible
1933 to have two identical connections in imp_conn_list. We must
1934 compare not conn's pointers but NIDs, otherwise we can defeat
1935 connection throttling.
1939 Description: port llog fixes from b1_6 into HEAD
1940 Details : Port llog reference couting and some llog cleanups from b1_6
1941 (bug 10800) into HEAD, for protect from panic and access to already
1942 free llog structures.
1946 Description: Detect stride IO mode in read-ahead
1947 Details : When a client does stride read, read-ahead should detect that and
1948 read-ahead pages according to the detected stride pattern.
1952 Description: data checksumming impacts single node performance
1953 Details : add support for several checksum algorithm. Currently, only CRC32
1954 and Adler-32 are supported. The checksum type can be changed on
1955 the fly via /proc/fs/lustre/osc/*/checksum_type.
1959 Description: use adler32 for page checksums
1960 Details : when available, use the Adler-32 algorithm instead of CRC32 for
1965 Description: build for x2 fails
1966 Details : fix compile issue on Cray systems.
1970 Description: Properly match for duplicate locks
1971 Details : Due to different lock order from skiplists code, we need to
1972 traverse entire list for now
1975 Frequency : only on PPC/SLES10
1977 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
1978 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
1979 be defined. Add a hack in mkfs_lustre.c to work around this compile
1984 Description: LASSERT on MDS when client holding flock lock dies
1985 Details : ldlm pool logic depends on number of granted locks equal to
1986 number of released locks which is not true for flock locks, so
1987 just exclude such locks from consideration.
1991 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
1992 Details : Use fsfilt_sync() for both the whole filesystem sync and
1993 individual file sync to eliminate dangerous inode locking
1994 with I_LOCK that can lead to a deadlock.
1998 Description: performance in 1.6.3
1999 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
2000 sure that 1MB requests are not fragmented by the block layer.
2002 Severity : enhancement
2004 Description: organize the server-side client stats on per-nid basis
2005 Details : Change the structure of stats under obdfilter and mds to
2013 The "uuid"s file would list the uuids of _active_ exports.
2014 And the clear entry is to clear all stats and stale nids.
2016 Severity : enhancement
2018 Description: eliminate client locks in face of contention
2019 Details : file contention detection and lockless i/o implementation
2020 for contended files.
2024 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
2027 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
2029 Description: MDS may lose file striping (and hence file data) in some cases
2030 Details : If there are additional extended attributes stored on the MDS,
2031 in particular ACLs, SELinux, or user attributes (if user_xattr
2032 is specified for the client mount options) then there is a risk
2033 of attribute loss. Additionally, the Lustre file striping
2034 needs to be larger than default (e.g. striped over all OSTs),
2035 and an additional attribute must be stored initially in the
2036 inode and then increase in size enough to be moved to the
2037 external attribute block (e.g. ACL growing in size) for file
2042 Description: skiplist implementation simplification
2043 Details : skiplists are used to group compatible locks on granted list
2044 that was implemented as tracking first and last lock of each
2045 lock group the patch changes that to using doubly linked lists
2049 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
2050 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
2051 because LASSERT on both the data supplied by a client, and the
2052 data on disk is dangerous and incorrect.
2054 Severity : enhancement
2056 Description: Slow truncate/writes to huge files at high offsets.
2057 Details : Directly associate cached pages to lock that protect those pages,
2058 this allows us to quickly find what pages to write and remove
2059 once lock callback is received.
2063 Description: more ldlm soft lockups
2064 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
2065 starve other threads from the resource lock for a long time in
2066 case of long waiting queue, so change the debug level from
2067 D_OTHER to the less frequently used D_INFO.
2069 Severity : enhancement
2071 Description: add -gid, -group, -uid, -user options to lfs find
2075 Description: Hung threads in invalidate_inode_pages2_range
2076 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
2077 once one is completed. As a result, some RPCs are stuck in the
2078 queue and are never sent.
2082 Description: filter threads hungs on waiting journal commit
2083 Details : Cleanup filter group llog code, then only filter group llog will
2084 be only created in the MDS/OST syncing process.
2088 Description: Procfs and llog threads access destoryed import sometimes.
2089 Details : Sync the import destoryed process with procfs and llog threads by
2090 the import refcount and semaphore.
2092 Severity : enhancement
2094 Description: openlock cache of b1_6 port to HEAD
2099 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
2100 Details : If insertion of an extent fails, then discard the inode
2101 preallocation and free data blocks else it can lead to duplicate
2106 Description: don't always update ctime in ext3_xattr_set_handle()
2107 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
2108 In some cases the ctime should not be updated, for example for
2109 2.0->1.8 compatibility it is necessary to delete an xattr and it
2110 should not update the ctime.
2114 Bugzilla : 15713/16362
2115 Description: Assertion in iopen_connect_dentry in 1.6.3
2116 Details : looking up an inode via iopen with the wrong generation number can
2117 populate the dcache with a disconneced dentry while the inode
2118 number is in the process of being reallocated. This causes an
2119 assertion failure in iopen since the inode's dentry list contains
2120 both a connected and disconnected dentry.
2124 Description: assertion failure in ldlm_handle2lock()
2125 Details : fix a race between class_handle_unhash() and class_handle2object()
2126 introduced in lustre 1.6.5 by bug 13622.
2131 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
2132 Details : remove the limit of 256 scsi disks in the sd_iostat patch
2137 Description: oops in sd_iostats_seq_show()
2138 Details : unloading/reloading the scsi low level driver triggers a kernel
2139 bug when trying to access the sd iostat file.
2144 Description: Kernel panics during QLogic driver reload
2145 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
2146 patch, causing memory corruption.
2151 Description: journal_dev option does not work in b1_6
2152 Details : pass mount option during pre-mount.
2154 Severity : enhancement
2156 Description: Add a FIEMAP(FIle Extent MAP) ioctl
2157 Details : FIEMAP ioctl will allow an application to efficiently fetch the
2158 extent information of a file. It can be used to map logical blocks
2159 in a file to physical blocks in the block device.
2163 Description: LDLM soft lockups - improvement
2164 Details : It is be possible to send the lock handle along with each read
2165 or write request because the client is already doing a lock match
2166 itself so there isn't any reason the OST should have to re-do that
2170 Frequency : only on Cray X2
2172 Description: X2 build failures
2173 Details : fix build failures on Cray X2.
2177 Description: xid & resent requests
2178 Details : Initialize RPC XID from clock at startup (randomly if clock is
2181 Severity : enhancement
2183 Description: Add lustre_start utility to start or stop multiple Lustre servers
2188 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
2189 Details : In case of memory pressure, list_del() can be called twice on
2190 req->rq_history_list, causing a kernel oops.
2194 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
2195 Details : fix stack overflow in the distributed lock manager by defering
2196 export eviction after a failed AST to the elt thread instead of
2197 handling it in the dlm interpret routine.
2201 Description: Convert some comments to new format.
2202 Details : Update documenting comments to match doxygen conventions.
2206 Description: Grammar fixes.
2207 Details : A couple of trivial spelling fixes.
2211 Description: OSD_COUNTERS-mandatory
2212 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
2213 to simplify the code.
2217 Description: simplify cmm_device freeing logic.
2218 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
2222 Description: Add lockdep support to dt_object_operations locking interface.
2223 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
2224 parameter indicating lock ordering. Update mdd code to use new
2229 Description: Introduce failloc constants for lockless IO tests.
2230 Details : Add two new failloc constants to test lockless IO. Only one of
2231 them in implemented---another is checked in yet to be landed
2236 Description: Add lockdep support for inode mutex.
2237 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
2238 in the situations where Lustre has to lock more than one inode
2243 Description: Add optional invariants checking support.
2244 Details : Add new LINVRNT() macro, optional on new --enable-invariants
2245 configure switch. This macro is to be used for consistency and
2246 sanity checks that are too expensive to be left in `production'
2251 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
2252 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
2257 Description: Add lockdep support for ldlm_lock and ldlm_resource.
2258 Details : Use spin_lock_nested() in (the only) situation where more than
2259 one ldlm_lock is locked simultaneously. Also, fix possible
2260 dead-lock in ldlm_lock_change_resource() by enforcing particular
2265 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
2266 Details : Instead of specifying each ldlm_lock call-back through separate
2267 parameter, wrap them into struct ldlm_callback_suite.
2271 Description: Kill join_lru obd method and its callers.
2272 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
2273 in memory---a requirement implemented through ->o_join_lru() obd
2274 method in HEAD. Remove this method and its users.
2278 Description: Add asynchronous ldlm ENQUEUE completion handler.
2279 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
2280 case that stock ldlm_completion_ast() cannot handle as it waits
2281 until lock is granted. Introduce new ldlm_completion_ast_async()
2282 for this. Also comment ldlm_completion_ast().
2286 Description: ldlm_error <-> errno conversion.
2287 Details : Add functions to map (rather arbitrary) between LDLM error codes
2288 and standard errno values. CLIO needs this to prevent LDLM specific
2289 constants from escaping ldlm and osc.
2293 Description: Kill unused ldlm_handle2lock_ns() function.
2294 Details : Kill unused ldlm_handle2lock_ns() function.
2298 Description: Add lu_ref support to ldlm_lock
2299 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
2300 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
2301 and ldlm_resource. LDLM interface has to be changed, because code that
2302 releases a reference on a lock, has to "know" what reference this is.
2303 In the most frequent case
2305 lock = ldlm_handle2lock(handle);
2307 LDLM_LOCK_PUT(lock);
2309 no changes are required. When any other reference (received _not_ from
2310 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
2311 instead of LDLM_LOCK_PUT().
2313 Arguably, changes are pervasive, and interface requires some discipline
2314 for proper use. On the other hand, it was very instrumental in finding
2315 a few leaked lock references.
2319 Description: Add ldlm_lock_addref_try().
2320 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
2321 attempts to addref a lock that might be being canceled
2326 Description: Add ldlm_weigh_callback().
2327 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
2328 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
2329 hard-coded `number of pages' logic.
2333 Description: Add lockdep annotations to llog code.
2334 Details : Use appropriately tagged _nested() locking calls in the places
2335 where llog takes more than one ->lgh_lock lock.
2339 Description: Add loi_kms_set().
2340 Details : Wrap kms updates into a helper function.
2344 Description: Constify instances of struct lsm_operations.
2345 Details : Constify instances of struct lsm_operations.
2349 Description: lu_conf support.
2350 Details : On a server, a file system object is uniquely identified
2351 by a fid, which is sufficient to locate and load all object
2352 state (inode). On a client, on the other hand, more data are
2353 necessary instantiate an object. Change lu_object_find() and
2354 friends to take additional `lu_conf' argument describing object.
2355 Typically this includes layout information.
2359 Description: lu_context fixes.
2360 Details : Introduce new lu_context functions that are needed on the client
2361 side, where some system threads (ptlrpcd) are shared by multiple
2362 modules, and so cannot be stopped during module shutdown.
2366 Description: Add start and stop methods to lu_device_type_operations.
2367 Details : Introduce two new methods in lu_device_type_operations, that are
2368 invoked when first instance of a given type is created and last one
2369 is destroyed respectively. This is need by CLIO.
2373 Description: Add lu_ref support to struct lu_device.
2374 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
2375 track leaked references.
2379 Description: Introduce lu_kmem_descr.
2380 Details : lu_kmem_descr and its companion interface allow to create
2381 and destroy a number of kmem caches at once.
2385 Description: Fix lu_object finalization race.
2386 Details : Fix a race between lu_object_find() finding an object and its
2387 concurrent finalization. This race is (most likely) not possible
2388 on the server, but might happen on the client.
2392 Description: Introduce lu_ref interface.
2393 Details : lu_ref is a debugging module allowing to track references to
2394 a given object. It is quite cpu expensive, and has to be
2395 explicitly enabled with --enable-lu_ref. See usage description
2400 Description: Factor lu_site procfs stats into a separate function.
2401 Details : Separate lu_site stats printing code into a separate function
2402 that can be reused on a client.
2406 Description: Constify instances of struct {lu,dt,md}_device_operations.
2407 Details : Constify instances of struct {lu,dt,md}_device_operations.
2411 Description: Introduce struct md_site and move meta-data specific parts of
2412 struct lu_site here.
2413 Details : Move md-specific fields out of struct lu_site into special struct
2414 md_site, so that lu_site can be used on a client.
2418 Description: Kill mdd_lov_destroy().
2419 Details : Remove unused mdd code.
2423 Description: Add st_block checking to multistat.c.
2424 Details : Add st_block checking to multistat.c.
2428 Description: Add lu_ref support to struct obd_device.
2429 Details : Add lu_ref tracking to obd_device.
2433 Description: Kill obd_set_fail_loc().
2434 Details : Remove unused code.
2438 Description: Add special type for ptlrpc_request interpret functions.
2439 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
2440 there. Actual usage will be in CLIO.
2444 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
2445 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
2446 doesn't work with lockdep.
2450 Description: Add rwv.c test program.
2451 Details : New testing program exercising readv(2) and writev(2) (Qian).
2455 Description: Add sendfile.c test program.
2456 Details : New testing program exercising sendfile(2) (Jay).
2460 Description: Ratelimit a message that can be very frequent.
2461 Details : Ratelimit a memory allocation failure message that can
2466 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2467 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2471 Description: Liblustre build fixes.
2472 Details : Liblustre build fixes.
2476 Description: libcfs: add cfs_{need,cond}_resched() interface.
2477 Details : libcfs: add cfs_{need,cond}_resched() definition and
2478 implementations for Linux, NT, and liblustre.
2480 Severity : enhancement
2482 Description: More exported tunables for mballoc
2483 Details : Add support for tunable preallocation window and new tunables for
2484 large/small requests
2488 Description: Detect corruption of block bitmap and checking for preallocations
2489 Details : Checks validity of on-disk block bitmap. Also it does better
2490 checking of number of applied preallocations. When corruption is
2491 found, it turns filesystem readonly to prevent further corruptions.
2495 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
2496 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
2497 window to 0. If it does hit cache miss, and read-pattern is still
2498 stride-io mode, does not reset the stride window, but also does not
2499 increase the stride window length in this case.
2503 Frequency : only for big-endian servers
2504 Description: Check if system is big-endian while mounting fs with extents feature
2505 Details : Mounting a filesystem with extents feature will fail on big-endian
2506 systems since ext3-based ldiskfs is not supported on big-endian
2507 systems. This can be over-riden with "bigendian_extents" mount option.
2509 Severity : enhancement
2511 Description: The root squash functionality
2512 Details : A security feature, which is to prevent users from being able
2513 to mount lustre on their desktop, run as root, and delete
2514 all of the files in the filesystem. The goal is accomplished by
2515 remapping user id (UID) and group id (GID) of the root user to
2516 a UID and GID specified by the system administartor via Lustre
2517 configuration management server (MGS). The functionality also
2518 allows to specify sets of clients for which the remapping does
2523 Description: Excessive recovery window
2524 Details : With AT enabled, the recovery window can be excessively long (6000+
2525 seconds). To address this problem, we no longer use
2526 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
2527 timeout no longer depends on the service time, it is set to
2528 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
2529 time via pb_service_time.
2533 Description: Watchdog triggered on MDS failover
2534 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
2535 the OSTs know that the MDS "UUID" can be reused for the same export
2536 from a different NID, so we do not need to wait for the export to be
2540 Frequency : rare, only if using MMP with Linux RAID
2542 Description: MMP doesn't work with Linux RAID
2543 Details : While using HA for Lustre servers with Linux RAID, it is possible
2544 that MMP will not detect multiple mounts. To make this work we
2545 need to unplug the device queue in RAID when the MMP block is being
2546 written. Also while reading the MMP block, we should read it from
2547 disk and not the cached one.
2549 Severity : enhancement
2551 Description: open file using fid
2552 Details : A file can be opened using just its fid, like
2553 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
2556 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
2559 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
2560 Details : add proper locking for creation of jbd_4k slab cache
2564 Description: MMP check in ext3_remount() fails without displaying any error
2565 Details : When multiple mount protection fails during remount, proper error
2568 Severity : enhancement
2570 Description: Allow stripe size to be up to 4G-64k
2571 Details : Fix math logic to allow large stripe sizes.
2575 Description: add check for >8TB ldiskfs filesystems
2576 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
2577 Don't allow >8TB ldiskfs filesystems to be mounted without
2578 force_over_8tb mount option
2583 Descriptoin: OSS complains "too many blocks" and crashes
2584 Details : In filter_commitrw_write():
2585 1. lquota_chkquota();
2586 2. obdo_from_inode();
2587 3. lquota_pending_commit();
2588 "2." has changed uid and gid in this case, then caused
2589 lquota_chkquota() and lquota_pending_commit() to use
2590 different uid and gid. That is the root of the bug.
2594 Descriptoin: getstripe doesn't report stripe count = -1 sanely
2595 Details : Detect and print out the 'stripe count = -1' case properly.
2599 Description: Out or order replies might be lost on replay
2600 Details : In ptlrpc_retain_replayable_request if we cannot find retained
2601 request with tid smaller then one currently being added, add it
2602 to the start, not end of the list.
2604 --------------------------------------------------------------------------------
2606 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
2608 * Support for kernels:
2609 2.6.5-7.283 (SLES 9),
2610 2.6.9-55.EL (RHEL 4),
2611 2.6.16.46-0.14 (SLES 10),
2612 2.6.18.8 vanilla (kernel.org)
2613 * Client support for unpatched kernels:
2614 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2615 2.6.16 - 2.6.22 vanilla (kernel.org)
2616 * Due to recently discovered recovery problems, we do not recommend
2617 using patchless RHEL 4 clients with this or any earlier release.
2618 * Recommended e2fsprogs version: 1.39.cfs8
2619 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2620 * Starting with this release, the ldiskfs backing filesystem required
2621 by Lustre is now in its own package, lustre-ldiskfs. This package
2622 should be installed. It is versioned separately from Lustre and
2623 may be released separately in future.
2627 Description: block reactivating mgc import until all deactivates complete
2628 Details : Fix race when failing back MDT/MGS to itself (testing)
2631 Frequency : at statup only
2633 Description: mds_lov_synchronize race leads to various problems
2634 Details : simultaneous MDT->OST connections at startup can cause the
2635 sync to abort, leaving the OSC in a bad state.
2637 Severity : enhancement
2639 Description: add optional extra BUILD_VERSION info
2640 Details : add a new environment variable (namely LUSTRE_VERS) which allows
2641 to override the lustre version.
2644 Frequency : 2.6.18 servers only
2646 Description: ll_kern_mount() doesn't release the module reference
2647 Details : The ldiskfs module reference count never drops down to 0
2648 because ll_kern_mount() doesn't release the module reference.
2653 Description: server LBUG when using old ost_num_threads parameter
2654 Details : Accept the old ost_num_threads parameter but warn that it
2655 is deprecated, and fix an off-by-one error that caused an LBUG.
2660 Description: Transient SCSI error results in persistent IO issue
2661 Details : iobuf->dr_error is not reinitialized to 0 between two
2665 Frequency : sometimes when underlying device returns I/O errors
2667 Description: OSTs not going read-only during write failures
2668 Details : OSTs are not remounted read-only when the journal commit threads
2669 get I/O errors because fsfilt_ext3 calls journal_start/stop()
2670 instead of the ext3 wrappers.
2674 Description: poor connect scaling with increasing client count
2675 Details : Don't run filter_grant_sanity_check for more than 100 exports
2676 to improve scaling for large numbers of clients.
2679 Frequency : SLES10 only
2681 Description: sanity-quota.sh quotacheck failed: rc = -22
2682 Details : Quotas cannot be enabled on SLES10.
2685 Frequency : liblustre clients only
2687 Description: getdirentries does not give error when run on compute nodes
2688 Details : getdirentries does not fail when the size specified as an argument
2689 is too small to contain at least one entry
2691 Severity : enhancement
2693 Description: Add LNET router traceability for debug purposes
2694 Details : If a checksum failure occurs with a router as part of the
2695 IO path, the NID of the last router that forwarded the bulk data
2696 is printed so it can be identified.
2701 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
2702 Details : Due to a race condition, liblustre clients were occasionally
2703 evicted incorrectly.
2705 Severity : enhancement
2707 Description: lfs setstripe use optional parameters instead of postional
2710 Severity : enhancement
2712 Description: Nanosecond timestamp support for ldiskfs
2713 Details : The on-disk ldiskfs filesystem has added support for nanosecond
2714 resolution timestamps. There is not yet support for this at
2715 the Lustre filesystem level.
2718 Frequency : during server recovery
2720 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
2721 Details : request with rq_no_resend flag not awake l_wait_event if they get
2725 Frequency : nfs export on patchless client
2727 Description: connectathon hang when test nfs export over patchless client
2728 Details : Disconnected dentry cannot be found with lookup, so we do not need
2729 to unhash it or make it invalid
2732 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
2733 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
2734 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
2739 Description: open req refcounting wrong on reconnect
2740 Details : If reconnect happened between getting open reply from server and
2741 call to mdc_set_replay_data in ll_file_open, we will schedule
2742 replay for unreferenced request that we are about to free.
2743 Subsequent close will crash in variety of ways.
2744 Check that request is still eligible for replay in
2745 mdc_set_replay_data().
2750 Description: disable writes to filesystem when reading health_check file
2751 Details : the default for reading the health_check proc file has changed
2752 to NOT do a journal transaction and write to disk, because this
2753 can cause reads of the /proc file to hang and block HA state
2754 checking on a healthy but otherwise heavily loaded system. It
2755 is possible to return to the previous behaviour during configure
2756 with --enable-health-write.
2758 Severity : enhancement
2760 Description: 64-bit inode version
2761 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
2762 to the inode. This will be required for version-based recovery.
2767 Description: MDS fails to start if a duplicate client export is detected
2768 Details : in some rare cases it was possible for a client to connect to
2769 an MDS multiple times. Upon recovery the MDS would detect this
2770 and fail during startup. Handle this more gracefully.
2772 Severity : enhancement
2774 Description: Add -o localflock option to simulate old noflock
2776 Details : This will achieve local-only flock/fcntl locks
2782 Description: log_commit_thread vs filter_destroy race leads to crash
2783 Details : Take import reference before releasing llog record semaphore
2788 Description: Wrong request locking in request set processing
2789 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
2790 imp_delayed_list, in this place should be used imp_lock.
2793 Frequency : when reconnection
2795 Description: Grant Leak when osc reconnect to OST
2796 Details : When osc reconnect ost, OST(filter) should check whether it
2797 should grant more space to client by comparing fed_grant and
2798 cl_avail_grant, and return the granted space to client instead
2799 of "new granted" space, because client will call osc_init_grant
2800 to update the client grant space info.
2803 Frequency : when client reconnect to OST
2805 Description: Grant Leak when osc do resend and replay bulk write
2806 Details : When osc reconnect to OST, OST(filter)should clear grant info of
2807 bulk write request, because the grant info while be sync between
2808 OSC and OST when reconnect, and we should ignore the grant info
2809 these of resend/replay write req.
2814 Description: Grant space more than avaiable left space sometimes.
2815 Details : When then OST is about to be full, if two bulk writing from
2816 different clients came to OST. Accord the avaliable space of the
2817 OST, the first req should be permitted, and the second one
2818 should be denied by ENOSPC. But if the seconde arrived before
2819 the first one is commited. The OST might wrongly permit second
2820 writing, which will cause grant space > avaiable space.
2823 Frequency : when client is evicted
2825 Description: Grant might be wrongly erased when osc is evicted by OST
2826 Details : when the import is evicted by server, it will fork another
2827 thread ptlrpc_invalidate_import_thread to invalidate the
2828 import, where the grant will be set to 0. While the original
2829 thread will update the grant it got when connecting. So if
2830 the former happened latter, the grant will be wrongly errased
2831 because of this race.
2836 Description: Checking Stale with correct fid
2837 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
2838 to check whether it is stale, because sometimes, we want the
2839 enqueue happened anyway, and op_data.fid2 will not be initialized.
2841 Severity : enhancement
2843 Description: update patchless client
2844 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
2847 Frequency : only with 2.4 kernel
2849 Description: random memory corruption
2850 Details : size of struct ll_inode_info is to big for union inode.u and this
2851 can be cause of random memory corruption.
2856 Description: Memory leak in recovery
2857 Details : Lov_mds_md was not free in an error handler in mds_create_object.
2858 It should also check obd_fail before fsfilt_start, otherwise if
2859 fsfilt_start return -EROFS,(failover mds during mds recovery).
2860 then the req will return with repmsg->transno = 0 and rc = EROFS.
2861 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
2862 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
2863 be freed no matter whether fsfilt_commit success or not.
2866 Frequency : only with huge count clients
2868 Description: Prevents from taking the superblock lock in llap_from_page for
2870 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
2871 for llap_from_page call in ll_removepage prevents from taking the
2872 superblock lock for a soon died page.
2877 Description: Not check open intent error before release open handle
2878 Details : in some rare cases, the open intent error is not checked before
2879 release open handle, which may cause
2880 ASSERTION(open_req->rq_transno != 0), because it tries to release
2881 the failed open handle.
2886 Description: Set cat log bitmap only after create log success.
2887 Details : in some rare cases, the cat log bitmap is set too early. and it
2888 should be set only after create log success.
2892 Description: Accessing a block bevice can re-enable I/O when Lustre is
2893 tearing down a device.
2894 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
2899 Description: service threads may hog cpus when there are a lot of requests
2900 Details : Insert cond_resched to give other threads a chance to use some CPU
2905 Description: the cat log was not initialized in recovery
2906 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
2907 unlink log on mds will not be initialized until mds post
2908 recovery. And also in mds post recovery, the unlink log will
2909 initialization will be done asynchronausly, so there will be race
2910 between add unlink log and unlink log initialization.
2914 Description: brw_stats were being printed incorrectly
2915 Details : brw_stats were being printed as log2 but all of them were not
2916 recorded as log2. Also remove some code duplication arising from
2917 filter_tally_{read,write}.
2921 Frequency : rare, only in recovery.
2922 Description: ASSERTION(req->rq_type != LI_POISON) failed
2923 Details : imp_lock should be held while iterating over imp_sending_list for
2924 prevent destroy request after get timeout in ptlrpc_queue_wait.
2928 Description: replay-single.sh test 52 fails
2929 Details : A lock's skiplist need to be cleanup when it being unlinked
2930 from its resource list.
2934 Description: Short directio read returns full requested size rather than
2936 Details : Direct I/O operations should return actual amount of bytes
2937 transferred rather than requested size.
2939 Severity : enhancement
2941 Description: metadata RPC reduction (e.g. for rm performance)
2942 Details : decrease the amount of synchronous RPC between clients and servers
2943 by canceling conflicing lock before the operation on the client
2944 and packing thier handles into the main operation RPC to server.
2946 Severity : enhancement
2948 Description: Async OSC create to avoid the blocking unnecessarily.
2949 Details : If a OST has no remain object, system will block on the creating
2950 when need to create a new object on this OST. Now, ways use
2951 pre-created objects when available, instead of blocking on an
2952 empty osc while others are not empty. If we must block, we block
2953 for the shortest possible period of time.
2957 Description: improve handling recoverable errors
2958 Details : if request processig with error which can be recoverable on server
2959 request should be resend, otherwise page released from cache and
2962 Severity : enhancement
2964 Description: refine locking for avoid write wrong info into lov_objid file
2965 Details : fix possible races with add new target and write/update data in
2968 --------------------------------------------------------------------------------
2970 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
2975 Frequency : on some architectures
2977 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
2978 Details : Uninitialized flags sometimes cause configuration commands to
2982 Frequency : patchless clients only
2984 Description: missing __iget() symbol export
2985 Details : The __iget() symbol export is missing. To avoid the need for
2986 this on patchless clients the deathrow inode reaper is turned
2987 off, and we depend on the VM to clean up old inodes. This
2988 dependency was during via the fix for bug 12181.
2990 --------------------------------------------------------------------------------
2992 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
2994 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
2995 INTEROPERATE with older versions automatically. In many cases a
2996 special upgrade step is needed. Please read the
2997 user documentation before upgrading any part of a live system.
2998 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
2999 version will not interoperate with 1.6 betas before beta5 (1.5.95).
3000 * WARNING: Lustre configuration and startup changes are required with
3001 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
3005 Severity : enhancement
3007 Description: MountConf
3008 Details : Lustre configuration is now managed via mkfs and mount
3009 commands instead of lmc and lconf. New obd types (MGS, MGC)
3010 are added for dynamic configuration management. See
3011 https://mail.clusterfs.com/wikis/lustre/MountConf for
3014 Severity : enhancement
3016 Description: dynamic OST addition
3017 Details : OSTs can now be added to a live filesystem
3019 Severity : enhancement
3021 Description: startup order invariance
3022 Details : MDTs and OSTs can be started in any order. Clients only
3023 require the MDT to complete startup.
3025 Severity : enhancement
3027 Description: parallel, asynchronous orphan cleanup
3028 Details : orphan cleanup is now performed in separate threads for each
3029 OST, allowing parallel non-blocking operation.
3031 Severity : enhancement
3033 Description: optimized stripe assignment
3034 Details : stripe assignments are now made based on ost space available,
3035 ost previous usage, and OSS previous usage, in order to try
3036 to optimize storage space and networking resources.
3038 Severity : enhancement
3040 Description: Permanently set tunables
3041 Details : All writable /proc/fs/lustre tunables can now be permanently
3042 set on a per-server basis, at mkfs time or on a live system.
3044 Severity : enhancement
3046 Description: Lustre message v2
3047 Details : Add lustre message format v2.
3049 Severity : enhancement
3051 Description: client OST exclusion list
3052 Details : Clients can be started with a list of OSTs that should be
3053 declared "inactive" for known non-responsive OSTs.
3057 Description: SPEC SFS validation failure on NFS v2 over lustre.
3058 Details : Changes the blocksize for regular files to be 2x RPC size,
3059 and not depend on stripe size.
3061 Severity : enhancement
3063 Description: Multiple MD RPCs in flight.
3064 Details : Further unserialise some read-only MDS RPCs - learn about intents.
3065 To avoid overly-overloading MDS, introduce a limit on number of
3066 MDS RPCs in flight for a single client and add /proc controls
3067 to adjust this limit.
3069 Severity : enhancement
3071 Description: client read/write statistics
3072 Details : Add client read/write call usage stats for performance
3073 analysis of user processes.
3074 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
3075 file access. extents_stats shows chunk size distribution.
3076 extents_stats_per_process show chunk size distribution per
3079 Severity : enhancement
3081 Description: mds statistics
3082 Details : Add detailed mds operations statistics in
3083 /proc/fs/lustre/mds/*/stats.
3087 Description: Failure of copying files with lustre special EAs.
3088 Details : Client side always return success for setxattr call for lustre
3089 special xattr (currently only "trusted.lov").
3094 Description: Refcount LNET uuids
3095 Details : The global LNET uuid list grew linearly with every startup;
3096 refcount repeated list entries instead of always adding to
3099 Severity : enhancement
3101 Description: Dynamic service threads
3102 Details : Within a small range, start extra service threads
3103 automatically when the request queue builds up.
3106 Frequency : mixed-endian client/server environments
3108 Description: mixed-endian crashes
3109 Details : The new msg_v2 system had some failures in mixed-endian
3112 Severity : enhancement
3114 Description: Easy OST removal
3115 Details : OSTs can be permanently deactivated with e.g. 'lctl
3116 conf_param lustre-OST0001.osc.active=0'
3118 Severity : enhancement
3120 Description: MGS proc entries
3121 Details : Added basic proc entries for the MGS showing what filesystems
3124 Severity : enhancement
3126 Description: provide MGS failover
3127 Details : Added config lock reacquisition after MGS server failover.
3129 Severity : enhancement
3131 Description: add Linux 2.4 support
3132 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
3136 Description: a large application tries to do I/O to the same resource and dies
3137 in the middle of it.
3138 Details : Check the req->rq_arrival time after the call to
3139 ost_brw_lock_get(), but before we do anything about
3140 processing it & sending the BULK transfer request. This
3141 should help move old stale pending locks off the queue as
3142 quickly as obd_timeout.
3145 Frequency : when an incorrect nid is specified during startup
3147 Description: ptlrpc connect to non-existant node causes kernel crash
3148 Details : LNET can't be re-entered from an event callback, which
3149 happened when we expire a message after the export has been
3150 cleaned up. Instead, hand the zombie cleanup off to another
3153 Severity : enhancement
3155 Description: plain/inodebits lock performance improvement
3156 Details : Grouping plain/inodebits in granted list by their request modes
3157 and bits policy, thus improving the performance of search through
3161 Frequency : only if OST filesystem is corrupted
3163 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3164 Details : for a short time RPCs with bulk IO are in the replay list,
3165 but replay of bulk IOs is unimplemented. If the OST filesystem
3166 is corrupted due to disk cache incoherency and then replay is
3167 started it is possible to trip an assertion. Avoid putting
3168 committed RPCs into the replay list at all to avoid this issue.
3171 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3173 Description: System hang on startup
3174 Details : This bug allowed the liblustre (e.g. catamount) client to
3175 return to the app before handling all startup RPCs. This
3176 could leave the node unresponsive to lustre network traffic
3177 and manifested as a server ptllnd timeout.
3179 Severity : enhancement
3181 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3182 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
3183 boolean to control whether to print peer debug info when a
3184 client's RPC times out.
3187 Frequency : only for kernels with patches from Lustre below 1.4.3
3189 Description: Remove old rdonly API
3190 Details : Remove old rdonly API which unused from at least lustre 1.4.3
3193 Frequency : only for devices with external journals
3195 Description: Set external device read-only also
3196 Details : During a commanded failover stop, we set the disk device
3197 read-only while the server shuts down. We now also set any
3198 external journal device read-only at the same time.
3201 Frequency : when upgrading from 1.4 while trying to change parameters
3203 Description: The wrong (new) MDC name was used when setting parameters for
3204 upgraded MDT's. Also allows changing of OSC (and MDC)
3205 parameters if --writeconf is specified at tunefs upgrade time.
3208 Frequency : when setting specific ost indicies
3210 Description: QOS code breaks on skipped indicies
3211 Details : Add checks for missing OST indicies in the QOS code, so OSTs
3212 created with --index need not be sequential.
3214 Severity : enhancement
3216 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3217 Details : The uninit_groups feature works in conjunction with the kernel
3218 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3219 up the pass1 processing of e2fsck. This is a read-only feature
3220 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3221 mount filesystems that have had this feature enabled.
3223 Severity : enhancement
3225 Description: Improve multi-block allocation algorithm to avoid fragmentation
3226 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3227 allocation locality and avoid filesystem fragmentation.
3229 ------------------------------------------------------------------------------
3231 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
3233 * Support for kernels:
3234 2.4.21-47.0.1.EL (RHEL 3)
3235 2.6.5-7.283 (SLES 9)
3236 2.6.9-42.0.10.EL (RHEL 4)
3237 2.6.12.6 vanilla (kernel.org)
3238 2.6.16.27-0.9 (SLES 10)
3239 * Recommended e2fsprogs version: 1.39.cfs5
3241 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3245 Frequency : occasional, depends on client load and configuration
3246 Bugzilla : 12181, 12203
3247 Description: data loss for recently-modified files
3249 Details : In some cases it is possible that recently written or created
3250 files may not be written to disk in a timely manner (this should
3251 normally be within 30s unless client IO load is very high).
3252 The problem appears as zero-length files or files that are a
3253 multiple of 1MB in size after a client crash or client eviction
3254 that are missing data at the end of the file.
3256 This problem is more likely to be hit on clients where files are
3257 repeatedly created and unlinked in the same directory, clients
3258 have a large amount of RAM, have many CPUs, the filesystem has
3259 many OSTs, the clients are rebooted frequently, and/or the files
3260 are not accessed by other nodes after being written.
3262 The presence of the problem can be detected by looking at
3263 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
3264 smaller than the second (nr_unused) then dirty files will not
3265 be flushed automatically to disk. "sync; sleep 10" should be
3266 run several times on the node before unmounting it to update
3267 Lustre (this is also safe to run on nodes without this problem).
3269 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
3270 that can cause this same problem, so customers using that kernel
3271 also need to update the kernel in addition to Lustre. In order
3272 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
3274 It is normal that files written just before a client crash (less
3275 than 30s) may not yet have been flushed to disk, even for local
3279 Frequency : frequent on thin XT3 nodes
3281 Description: UUID collision on thin XT3 Linux nodes
3282 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
3283 randomly, since we relied on an insufficiently-seeded PRNG.
3288 Description: OSS hangs after "All ost request buffers busy"
3289 Details : A deadlock between quota and journal operations caused OSS
3290 hangs after printing "All ost request buffers busy."
3293 Frequency : always on liblustre builds
3295 Description: Cleanup compiler warnings on liblustre
3298 Frequency : always on liblustre builds on XT3
3300 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
3304 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
3305 > 32000 subdirectories
3306 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
3307 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
3308 subdirectory count crosses 32000. This will aid e2fsck to
3309 correctly handle more than 32000 subdirectories.
3312 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3314 Description: System hang on startup
3315 Details : This bug allowed the liblustre (e.g. catamount) client to
3316 return to the app before handling all startup RPCs. This
3317 could leave the node unresponsive to lustre network traffic
3318 and manifested as a server ptllnd timeout.
3320 Severity : enhancement
3322 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3323 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
3324 boolean to control whether to print peer debug info when a
3325 client's RPC times out.
3330 Description: make O_SYNC working on 2.6 kernels
3331 Details : 2.6 kernels use different method for mark pages for write,
3332 so need add a code to lustre for O_SYNC work.
3337 Description: Failure to close file and release space on NFS
3338 Details : Put inode details into lock acquired in ll_intent_file_open.
3339 Use mdc_intent_lock in ll_intent_open to properly
3340 detect all kind of errors unhandled by mdc_enqueue.
3345 Description: proc file read during shutdown sometimes raced obd removal,
3347 Details : Add lock to prevent obd access after proc file removal.
3350 Frequency : Only for files larger than 4GB on 32-bit clients.
3352 Description: improperly doing page alignment of locks
3353 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
3354 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3359 Description: RPCs being resent when they shouldn't be
3360 Details : Some RPCs that should not be resent are being resent. This
3361 can cause inconsistencies in the RPC state machine. Do not
3362 resend such requests.
3365 Frequency : rare, only with NFS export
3367 Description: Crash on NFS re-export node
3368 Details : under very unusual load conditions an assertion is hit in
3369 ll_intent_file_open()
3372 Frequency : only if OST filesystem is corrupted
3374 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3375 Details : for a short time RPCs with bulk IO are in the replay list,
3376 but replay of bulk IOs is unimplemented. If the OST filesystem
3377 is corrupted due to disk cache incoherency and then replay is
3378 started it is possible to trip an assertion. Avoid putting
3379 committed RPCs into the replay list at all to avoid this issue.
3384 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3385 Details : Large single O_DIRECT read and write calls can fail to allocate
3386 a sufficiently large buffer to process the request. In case of
3387 allocation failure the allocation is retried with a smaller
3388 buffer and broken into smaller requests.
3390 Severity : enhancement
3392 Description: Add -o localflock option to simulate old noflock behaviour.
3393 Details : This will achieve local-only flock/fcntl locks coherentness.
3398 Description: versioning check is incomplete
3399 Details : Checking the version difference of client vs. server, report
3400 error if the gap is too big.
3405 Description: add support PG_writeback bit
3406 Details : add support for PG_writeback bit for Lustre, for more carefull
3407 work with page cache in 2.6 kernel. This also fix some deadlocks
3408 and remove hack for work O_SYNC with 2.6 kernel.
3410 Severity : enhancement
3412 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3413 Details : The uninit_groups feature works in conjunction with the kernel
3414 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3415 up the pass1 processing of e2fsck. This is a read-only feature
3416 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3417 mount filesystems that have had this feature enabled.
3419 Severity : enhancement
3421 Description: Improve multi-block allocation algorithm to avoid fragmentation
3422 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3423 allocation locality and avoid filesystem fragmentation.
3425 ------------------------------------------------------------------------------
3427 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3429 * Support for kernels:
3430 2.6.9-42.0.3.EL (RHEL 4)
3431 2.6.5-7.276 (SLES 9)
3432 2.4.21-47.0.1.EL (RHEL 3)
3433 2.6.12.6 vanilla (kernel.org)
3434 2.6.16.21-0.8 (SLES10)
3435 * Recommended e2fsprogs version: 1.39.cfs2-0
3437 * The backwards-compatible /proc/sys/portals symlink has been removed
3438 in this release. Before upgrading, please ensure that you change
3439 any configuration scripts or /etc/sysctl.conf files that access
3440 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3441 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3442 in advance of the upgrade on any system running Lustre 1.4.6 or
3443 newer, since /proc/sys/lnet was added in that version.
3444 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3448 Frequency : only when quota is used
3450 Description: avoid scanning export list for quota master
3451 Details : Change the algorithms to avoid scanning export list in order
3452 to improve the efficiency.
3455 Frequency : MDS failover only, very rarely
3457 Description: "went back in time" messages on mds failover
3458 Details : The greatest transno may be lost when the current operation
3459 finishes with an error (transno==0) and the client's last_rcvd
3460 record is over-written. Save the greatest transno in the
3461 mds_last_transno for this case.
3464 Frequency : always for specific kernels and striping counts
3466 Description: client may get "Matching packet too big" without ACL support
3467 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
3468 when trying to access files in certain configurations. The
3469 clients should in fact be denied when mounting because they do
3470 not understand ACLs.
3473 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
3475 Description: many clients connecting with IO in progress causes connect timeouts
3476 Details : Avoid synchronous journal commits to avoid delays caused by many
3477 clients connecting/disconnecting when bulk IO is in progress.
3478 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
3479 OST_IO_PORTAL to avoid delays behind potentially many pending
3483 Frequency : occasionally with multiple writers to a single file
3485 Description: shared writes to file may result in wrong size reported by stat()
3486 Details : Allow growing of kms when extent lock is cancelled
3489 Frequency : always with random mmap IO to multi-striped file
3491 Description: mmap write might be lost if we are writing to a 'hole' in stripe
3492 Details : Only if the hole is at the end of OST object so that kms is too
3493 small. Fix is to increase kms accordingly in ll_nopage.
3496 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
3498 Description: writes to a missing object would leak memory on the OST
3499 Details : If there is an inconsistency between the MDS and OST filesystems,
3500 such that the MDS references an object that doesn't exist, writes
3501 to that object will leak memory due to incorrect cleanup in the
3502 error handling path, eventually running out of memory on the OST.
3507 Description: Creating too long symlink causes lustre errors
3508 Details : Check symlink and name lengths before sending requests to MDS.
3511 Frequency : only if flock is enabled (not on by default)
3513 Description: posix locks not released on fd closure on 2.6.9+
3514 Details : We failed to add posix locks to list of inode locks on 2.6.9+
3515 kernels, this caused such locks not to be released on fd close and
3516 then assertions on fs unmount about still used locks.
3519 Frequency : MDS failover only, very rarely
3521 Description: clients may get ASSERTION(granted_lock != NULL)
3522 Details : When request was taking a long time, and a client was resending
3523 a getattr by name lock request. The were multiple lock requests
3524 with the same client lock handle and
3525 mds_getattr_name->fixup_handle_for_resent_request found one of the
3526 lock handles but later failed with ASSERTION(granted_lock != NULL).
3531 Description: handle->h_buffer_credits > 0, assertion failure
3532 Details : h_buffer_credits is zero after truncate, causing assertion
3533 failure. This patch extends the transaction or creates a new
3537 Frequency : NFS re-export or patchless client
3538 Bugzilla : 11179, 10796
3539 Description: Crash on NFS re-export node (__d_move)
3540 Details : We do not want to hash the dentry if we don't have a lock.
3541 But if this dentry is later used in d_move, we'd hit uninitialised
3542 list head d_hash, so we just do this to init d_hash field but
3543 leave dentry unhashed.
3546 Frequency : NFS re-export or patchless client
3548 Description: NFS exports has problem with symbolic link
3549 Details : lustre client didn't properly install dentry when re-exported
3550 to NFS or running patchless client.
3553 Frequency : NFS re-export or patchless client
3555 Description: Various nfs/patchless fixes.
3556 Details : fixes reuse disconected alias for lookup process - this fixes
3557 warning "find_exported_dentry: npd != pd",
3558 fix permission error with open files at nfs.
3559 fix apply umask when do revalidate.
3562 Frequency : occasional
3564 Description: Crash on NFS re-export node
3565 Details : calling clear_page() on the wrong pointer triggered oops in
3566 generic_mapping_read().
3569 Frequency : rarely, using O_DIRECT IO
3571 Description: unaligned directio crashes client with LASSERT
3572 Details : check for unaligned buffers before trying any requests.
3575 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
3577 Description: stale data returned from RAID cache
3578 Details : If only a small amount of IO is done to the RAID device before
3579 reading it again it is possible to get stale data from the RAID
3580 cache instead of reading it from disk.
3583 Frequency : always for sles10 kernel
3585 Description: sles10 support
3586 Details : ll_follow_link: compile fixes and using of nd_set_link
3587 under newer kernels.
3590 Frequency : depends on arch, kernel and compiler version, always on sles10
3593 Description: recursive or deep enough symlinks cause stack overflow
3594 Details : getting rid of large stack-allocated variable in
3598 Frequency : depends on hardware
3600 Description: lustre write performance loss in the SLES10 kernel
3601 Details : the performance loss is caused by using of write barriers in the
3602 ext3 code. The SLES10 kernel turns barrier support on by
3603 default. The fix is to undo that change for ldiskfs.
3605 ------------------------------------------------------------------------------
3607 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
3609 * Support for kernels:
3610 2.6.9-42.0.3EL (RHEL 4)
3611 2.6.5-7.276 (SLES 9)
3612 2.4.21-47.0.1.EL (RHEL 3)
3613 2.6.12.6 vanilla (kernel.org)
3617 Frequency : quota enabled and large files being deleted
3619 Description: releasing more than 4GB of quota at once hangs OST
3620 Details : If a user deletes more than 4GB of files on a single OST it
3621 will cause the OST to spin in an infinite loop. Release
3622 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3627 Description: statfs data retrieved from /proc may be stale or zero
3628 Details : When reading per-device statfs data from /proc, in the
3629 {kbytes,files}_{total,free,avail} files, it may appear
3630 as zero or be out of date.
3633 Frequency : systems with MD RAID1 external journal devices
3635 Description: lconf's call to blkid is confused by RAID1 journal devices
3636 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3637 of returning all block devices that match the journal UUID.
3640 Frequency : always, for aggregate stripe size over 4GB
3642 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
3643 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
3644 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
3647 Frequency : always if "lfs find" used on a local file/directory
3649 Description: "lfs find" segfaults if used on a local file/directory
3650 Details : The case where a directory component was not specified wasn't
3651 handled correctly. Handle this properly.
3654 Frequency : always on ppc64
3656 Description: the write to an ext3 filesystem mounted with mballoc got stuck
3657 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3658 perform endianness conversion.
3661 Frequency : rarely (truncate to non-zero file size after write under load)
3662 Bugzilla : 10730, 10687
3663 Description: Files padded with zeros to next 4K multiple
3664 Details : With filesystems mounted using the "extents" option (2.6 kernels)
3665 it is possible that files that are truncated to a non-zero size
3666 immediately after being written are filled with zero bytes beyond
3667 the truncated size. No file data is lost.
3669 Severity : enhancement
3671 Description: Allow recovery/failover for liblustre clients.
3672 Details : liblustre clients were unaware of failover configurations until
3675 Severity : enhancement
3677 Description: user file locks should fail when not mounting with flock option
3678 Details : Set up an error-returning stub in ll_file_operations.lock field
3679 to prevent incorrect behaviour when client is mounted without
3680 flock option. Also, set up properly f_op->flock field for
3684 Frequency : always on ia64
3686 Description: "lfs df" loops on printing out MDS statfs information
3687 Details : The obd_ioctl_data was not initialized and in some systems
3688 this caused a failure during the ioctl that did not return
3689 an error. Initialize the struct and return an error on failure.
3692 Frequency : SLES 9 only
3694 Description: Error of copying files with lustre special EAs as root
3695 Details : Client side always return success for setxattr call for lustre
3696 special xattr (currently only "trusted.lov").
3699 Frequency : rarely on clusters with both ia64+i386 clients
3701 Description: ia64+i686 clients doing shared IO on the same file may LBUG
3702 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
3703 clients are doing concurrent writes to the same file it is
3704 possible that the ia64 clients may LASSERT because the OST
3705 extent locks are not PAGE_SIZE aligned. Ensure that grown
3706 locks are always aligned on the request boundary.
3709 Frequency : specific use, occasional
3711 Description: Overwriting in use executable truncates on-disk binary image
3712 Details : If one node attempts to overwrite an executable in use by
3713 another node, we now correctly return ETXTBSY instead of
3714 truncating the file.
3716 Severity : enhancement
3718 Description: Async OSC create to avoid the blocking unnecessarily.
3719 Details : If a OST has no remain object, system will block on the creating
3720 when need to create a new object on this OST. Now, ways use
3721 pre-created objects when available, instead of blocking on an
3722 empty osc while others are not empty. If we must block, we block
3723 for the shortest possible period of time.
3728 Description: chmod on Lustre root is propagated to other clients
3729 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
3730 invalid by the follow_mount time.
3733 Frequency : liblustre clients only
3735 Description: Race in 'instant cancel' lock handling could lead to such locks
3736 never to be granted in case of SMP MDS
3737 Details : Do not destroy not yet granted but cbpending locks in
3741 Frequency : replay/resend of open
3743 Description: non null lock assetion failure in mds_intent_policy
3744 Details : Trying to replay/resend lockless open requests resulted in
3745 mds_open() returning 0 with no lock. Now it sets a flag if
3746 a lock is going to be returned.
3748 Severity : enhancement
3750 Description: Checksum enhancements
3751 Details : New checksum enhancements allow for resending RPCs that failed
3754 Severity : enhancement
3756 Description: Tunables on number of dirty pages in cacche
3757 Details : Allow to set limit on number of dirty pages cached.
3762 Description: client crash on unmount - lock still has references
3763 Details : In some error handling cases it was possible to leak a lock
3764 reference on a client while accessing a file. This was not
3765 harmful to the client during operation, but would cause the
3766 client to crash when the filesystem is unmounted.
3769 Frequency : specific case, rare
3771 Description: ETXTBSY on mds though file not in use by client
3772 Details : ETXTBSY is no longer incorrectly returned when attempting to
3773 chmod or chown a directory that the user previously tried to
3774 execute or a currently-executing binary.
3777 Frequency : extremely rare except on liblustre-based clients
3779 Description: Lustre space not freed when files are deleted
3780 Details : Clean up open-unlinked files after client eviction. Previously
3781 the unlink was skipped and the files remained as orphans.
3786 Description: OST failure "would be an LBUG" in waiting_locks_callback()
3787 Details : In some cases it was possible to send a blocking callback to
3788 a client doing a glimpse, even though that client didn't get
3789 a lock granted. When the glimpse lock is cancelled on the OST
3790 the freed lock is left on the waiting list and corrupted the list.
3793 Frequency : all core dumps
3795 Description: Broke core dumps to lustre
3796 Details : Negative dentry may be unhashed if parent does not have UPDATE
3797 lock, but some callers, e.g. do_coredump, expect dentry to be
3798 hashed after successful create, hash it in ll_create_it.
3800 ------------------------------------------------------------------------------
3802 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
3804 * Support for kernels:
3805 2.6.9-42.0.2.EL (RHEL 4)
3806 2.6.5-7.276 (SLES 9)
3807 2.4.21-40.EL (RHEL 3)
3808 2.6.12.6 vanilla (kernel.org)
3812 Frequency : always on RHEL 3
3814 Description: Number of open files grows over time
3815 Details : The number of open files grows over time, whether or not
3816 Lustre is started. This was due to a filp leak introduced
3817 by one of our kernel patches.
3819 ------------------------------------------------------------------------------
3821 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
3823 * Support for kernels:
3824 2.6.9-42.EL (RHEL 4)
3825 2.6.5-7.276 (SLES 9)
3826 2.4.21-40.EL (RHEL 3)
3827 2.6.12.6 vanilla (kernel.org)
3832 Bugzilla : 5719, 9635, 9792, 9684
3833 Description: OST (or MDS) trips assertions in (re)connection under heavy load
3834 Details : If a server is under heavy load and cannot reply to new
3835 connection requests before the client resends the (re)connect,
3836 the connection handling code can behave badly if two service
3837 threads are concurrently handing separate (re)connections from
3838 the same client. Add better locking to the connection handling
3839 code, and ensure that only a single connection will be processed
3840 for a given client UUID, even if the lock is dropped.
3842 Severity : enhancement
3844 Description: add TCP zero-copy support to kernel
3845 Details : Add support to the kernel TCP stack to allow zero-copy bulk
3846 sends if the hardware supports scatter-gather and checksumming.
3847 This allows socklnd to do client-write and server-read more
3848 efficiently and reduce CPU utilization from skbuf copying.
3851 Frequency : only if NFS exporting from client
3853 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
3854 Details : If setting attributes on a file created under NFS that had
3855 never been opened it would be possible to oops the client
3856 if the file had no objects.
3859 Frequency : always for liblustre
3861 Description: liblustre client does MDS+OSTs setattr RPC for each write
3862 Details : When doing a write from a liblustre client, the client
3863 incorrectly issued an RPC to the MDS and each OST the file was
3864 striped over in order to update the timestamps. When writing
3865 with small chunks and many clients this could overwhelm the MDS
3866 with RPCs. In all cases it would slow down the write because
3867 these RPCs are unnecessary.
3869 Severity : enhancement
3871 Description: allow number of MDS service threads to be changed at module load
3872 Details : It is now possible to change the number of MDS service threads
3873 running. Adding "options mds mds_num_threads={N}" to the MDS's
3874 /etc/modprobe.conf will set the number of threads for the next
3875 time Lustre is restarted (assuming the "mds" module is also
3876 reloaded at that time). The default number of threads will
3877 stay the same, 32 for most systems.
3882 Description: OST crash if filesystem is unformatted or corrupt
3883 Details : If an OST is started on a device that has never been formatted
3884 or if the filesystem is corrupt and cannot even mount then the
3885 error handling cleanup routines would dereference a NULL pointer.
3890 Description: NULL pointer deref in llap_from_page.
3891 Details : get_cache_page_nowait can return a page with NULL (or otherwise
3892 incorrect) mapping if the page was truncated/reclaimed while it was
3893 searched for. Check for this condition and skip such pages when
3894 doing readahead. Introduce extra check to llap_from_page() to
3895 verify page->mapping->host is non-NULL (so page is not anonymous).
3898 Frequency : Sometimes when using sys_sendfile
3900 Description: "page not covered by a lock" warnings from ll_readpage
3901 Details : sendfile called ll_readpage without right page locks present.
3902 Now we introduced ll_file_sendfile that does necessary locking
3903 around call to generic_file_sendfile() much like we do in
3907 Frequency : with certain MDS communication failures at client mount time
3909 Description: NULL pointer deref after failed client mount
3910 Details : a client connection request may delayed by the network layer
3911 and not be sent until after the PTLRPC layer has timed out the
3912 request. If the client fails the mount immediately it will try
3913 to clean up before the network times out the request. Add a
3914 reference from the request import to the obd device and delay
3915 the cleanup until the network drops the request.
3918 Frequency : occasionally during client (re)connect
3920 Description: assertion failure during client (re)connect
3921 Details : processing a client connection request may be delayed by the
3922 client or server longer than the client connect timeout. This
3923 causes the client to resend the connection request. If the
3924 original connection request is replied in this interval, the
3925 client may trip an assertion failure in ptlrpc_connect_interpret()
3926 which thought it would be the only running connect process.
3929 Frequency : only with obd_echo servers and clients that are rebooted
3931 Description: kernel BUG accessing uninitialized data structure
3932 Details : When running an obd_echo server it did not start the ping_evictor
3933 thread, and when a client was evicted an uninitialized data
3934 structure was accessed. Start the ping_evictor in the RPC
3935 service startup instead of the OBD startup.
3937 Severity : enhancement
3938 Bugzilla : 10193 (patchless)
3939 Description: Remove dependency on various unexported kernel interfaces.
3940 Details : No longer need reparent_to_init, exit_mm, exit_files,
3941 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
3944 Frequency : rare (only users of deprecated and unsupported LDAP config)
3946 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
3947 Details : LDAP apparently contains 'lustreName' attributes instead of
3948 'name'. A simple remapping of the name is sufficient.
3951 Frequency : rare (only with non-default dump_on_timeout debug enabled)
3953 Description: waiting_locks_callback trips kernel BUG if client is evicted
3954 Details : Running with the dump_on_timeout debug flag turned on makes
3955 it possible that the waiting_locks_callback() can try to dump
3956 the Lustre kernel debug logs from an interrupt handler. Defer
3957 this log dumping to the expired_lock_main() thread.
3959 Severity : enhancement
3961 Description: Support NFS exporting on 2.6 kernels.
3962 Details : Implement non-rawops metadata methods for NFS server to use without
3963 changing NFS server code.
3966 Frequency : very rare (synthetic metadata workload only)
3968 Description: two racing renames might cause an MDS thread to deadlock
3969 Details : Running the "racer" program may cause one MDS thread to rename
3970 a file from being the source of a rename to being the target of
3971 a rename at exactly the same time that another thread is doing
3972 so, and the second thread has already enqueued these locks after
3973 doing a lookup of the target and is trying to relock them in
3974 order. Ensure that we don't try to re-lock the same resource.
3977 Frequency : only very large systems with liblustre clients
3979 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
3980 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
3981 all OSTs in parallel. This allows the eviction of stale liblustre
3982 clients to proceed much faster than if they were done in series,
3983 and also offers similar improvements for other set_info RPCs.
3988 Description: excessive CPU usage during initial read phase on client
3989 Details : During the initial read phase on a client, it would agressively
3990 retry readahead on the file, consuming too much CPU and impacting
3991 performance (since 1.4.5.8). Improve the readahead algorithm
3992 to avoid this, and also improve some other common cases (read
3993 of small files in particular, where "small" is files smaller than
3994 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
3999 Description: MDS crash when receiving packet with unknown intent.
4000 Details : Do not LBUG in unknown intent case, just return -EFAULT
4002 Severity : enhancement
4003 Bugzilla : 9293, 9385
4004 Description: MDS RPCs are serialised on client. This is unnecessary for some.
4005 Details : Do not serialize getattr (non-intent version) and statfs.
4008 Frequency : occasional, when OST network is overloaded/intermittent
4010 Description: client evicted by OST after bulk IO timeout
4011 Details : If a client sends a bulk IO request (read or write) the OST
4012 may evict the client if it is unresposive to its data GET/PUT
4013 request. This is incorrect if the network is overloaded (takes
4014 too long to transfer the RPC data) or dropped the OST GET/PUT
4015 request. There is no need to evict the client at all, since
4016 the pinger and/or lock callbacks will handle this, and the
4017 client can restart the bulk request.
4020 Frequency : Always when mmapping file with no objects
4022 Description: client crashes when mmapping file with no objects
4023 Details : Check that we actually have objects in a file before doing any
4024 operations on objects in ll_vm_open, ll_vm_close and
4030 Description: Request leak when working with deleted CWD
4031 Details : Introduce advanced request refcount tracking for requests
4032 referenced from lustre intent.
4034 Severity : Enhancement
4036 Description: Cache open file handles on client.
4037 Details : MDS now will return special lock along with openhandle, if
4038 requested and client is allowed to hold openhandle, even if unused,
4039 until such a lock is revoked. Helps NFS a lot, since NFS is opening
4040 closing files for every read/write openration.
4042 Severity : Enhancement
4044 Description: Cache open negative dentries on client when possible.
4045 Details : Guard negative dentries with UPDATE lock on parent dir, drop
4046 negative dentries on lock revocation.
4051 Description: Remounting a client read-only wasn't possible with a zconf mount
4052 Details : It wasn't possible to remount a client read-only with llmount.
4054 Severity : enhancement
4055 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
4056 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
4057 MPICH is included to add Lustre-specific ADIO interfaces.
4058 This is based closely on the UFS ADIO layer and only differs
4059 in file creation, in order to allow the OST striping to be set.
4060 This is user-contributed code and not supported by CFS.
4065 Description: extended inode attributes (immutable, append-only) work improperly
4066 when 2.4 and 2.6 kernels are used on client/server or vice versa
4067 Details : Introduce kernel-independent values for these flags.
4069 Severity : enhancement
4072 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
4073 Details : Many of the /proc/ tunables can only be tuned at a megabyte
4074 granularity. Now, Fractional MB granularity is be supported,
4075 this is very useful for low memory system.
4077 Severity : enhancement
4079 Description: Getattr by fid
4080 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
4081 locks, avoids extra getattr rpc requests to MDS, allows '/' to
4082 have locks and avoids getattr rpc requests for it on every stat.
4085 Frequency : Always, for filesystems larger than 2TB
4087 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4088 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
4089 filesystems larger than 2TB to be mounted with mballoc enabled.
4092 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
4094 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4095 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
4096 filesystems larger than 2TB is mounted then the kernel will
4097 silently corrupt the start of the filesystem. CONFIG_LBD is
4098 enabled for all CFS-supported kernels, but the possibility of
4099 this happening with a modified kernel config exists.
4101 Severity : enhancement
4103 Description: add client O_DIRECT support for 2.6 kernels
4104 Details : It is now possible to do O_DIRECT reads and writes to files
4105 in the Lustre client mountpoint on 2.6 kernel clients.
4107 Severity : enhancement
4109 Description: parallel glimpse, setattr, statfs, punch, destroy requests
4110 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
4111 parallel, not waiting for response from every OST before sending
4112 a rpc to the next OST.
4117 Description: setattr vs write race when updating file timestamps
4118 Details : Client processes that update a file timestamp into the past
4119 right after writing to the file (e.g. tar) it is possible that
4120 the updated file modification time can be reset to the current
4121 time due to a race between processing the setattr and write RPC.
4123 Severity : enhancement
4125 Description: Bring 'lfs find' closer in line with regular Linux find.
4126 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
4127 -print0 options and obtains all the needed info through the lustre
4130 Severity : enhancement
4132 Description: support up to 1024 configured devices on one node
4133 Details : change obd_dev array from statically allocated to dynamically
4134 allocated structs as they are first used to reduce memory usage
4139 Description: Flush dirty partially truncated pages during truncate
4140 Details : Immediatelly flush partially truncated pages in filter_setattr,
4141 this way we completely avoid having any pages in page cache on OST
4142 and can retire ugly workarounds during writes to flush such pages.
4147 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
4148 Details : protect inode from truncation within vfs_unlink() context
4149 just take a reference before calling vfs_unlink() and release it
4150 when parent's i_sem is free.
4155 Description: last_id value checked outside lock on OST caused LASSERT failure
4156 Details : If there were multiple MDS->OST object precreate requests in
4157 flight, it was possible that the OST's last object id was checked
4158 outside a lock and incorrectly tripped an assertion. Move checks
4159 inside locks, and discard old precreate requests.
4162 Frequency : always, if extents are used on OSTs
4164 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
4165 Details : OSTs using the extents format would not zero the high 16 bits of
4166 the index physical block number. This is not a problem for any
4167 OST filesystems smaller than 16TB, and no kernels support ext3
4168 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
4169 new/modified files) and can be fixed for existing filesystems
4170 with e2fsprogs-1.39-cfs1.
4175 Description: import connection selection may be incorrect if timer wraps
4176 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
4177 import connections to be ignored if the 32-bit jiffies counter
4178 wraps. Use a 64-bit jiffies counter.
4181 Frequency : very large clusters immediately after boot
4183 Description: LNET request buffers exhausted under heavy short-term load
4184 Details : If a large number of client requests are generated on a service
4185 that has previously never seen so many requests it is possible
4186 that the request buffer growth cannot keep up with the spike in
4187 demand. Instead of dropping incoming requests, they are held in
4188 the LND until the RPC service can accept more requests.
4191 Frequency : Sometimes during replay
4193 Description: Assertion failure in ll_local_open after replay.
4194 Details : If replay happened on an open request reply before we were able
4195 to set replay handler, reply will become not swabbed tripping the
4196 assertion in ll_local_open. Now we set the handler right after
4197 recognising of open request
4200 Frequency : very rare
4202 Description: kernel reports "badness in vsnprintf"
4203 Details : Reading from the "recovery_status" /proc file in small chunks
4204 may cause a negative length in lprocfs_obd_rd_recovery_status()
4205 call to vsnprintf() (which is otherwise harmless). Exit early
4206 if there is no more space in the output buffer.
4208 Severity : enhancement
4210 Description: clear OBD RPC statistics by writing to them
4211 Details : It is now possible to clear the OBD RPC statistics by writing
4212 to the "stats" file.
4217 Description: Client mtime is not the same on different clients after utimes
4218 Details : In some cases, the client was using the utimes() syscall on
4219 a file cached on another node. The clients now validate the
4220 ctime from the MDS + OSTs to determine which one is right.
4225 Description: Inability to activate failout mode
4226 Details : lconf script incorrectly assumed that in pythong string's numeric
4227 value is used in comparisons.
4230 Frequency : always with multiple stripes per file
4232 Description: Inefficient object allocation for mutli-stripe files
4233 Details : When selecting which OSTs to stripe files over, for files with
4234 a stripe count that divides evenly into the number of OSTs,
4235 the MDS is always picking the same starting OST for each file.
4236 Return the OST selection heuristic to the original design.
4241 Description: mount failures may take full timeout to return an error
4242 Details : Under some heavy load conditions it is possible that a
4243 failed mount can wait for the full obd_timeout interval,
4244 possibly several minutes, before reporting an error.
4245 Instead return an error as soon as the status is known.
4247 Frequency : quota enabled and large files being deleted
4249 Description: releasing more than 4GB of quota at once hangs OST
4250 Details : If a user deletes more than 4GB of files on a single OST it
4251 will cause the OST to spin in an infinite loop. Release
4252 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4257 Description: statfs data retrieved from /proc may be stale or zero
4258 Details : When reading per-device statfs data from /proc, in the
4259 {kbytes,files}_{total,free,avail} files, it may appear
4260 as zero or be out of date.
4263 Frequency : systems with MD RAID1 external journal devices
4265 Description: lconf's call to blkid is confused by RAID1 journal devices
4266 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4267 of returning all block devices that match the journal UUID.
4270 Frequency : always, for aggregate stripe size over 4GB
4272 Description: assertion fails when trying to use 4GB stripe size
4273 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
4274 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
4277 Frequency : always on ppc64
4279 Description: the first write on an ext3 filesystem with mballoc got stuck
4280 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4281 perform endianness conversion.
4283 ------------------------------------------------------------------------------
4285 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4287 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
4288 INTEROPERATE with older versions automatically. Please read the
4289 user documentation before upgrading any part of a live system.
4290 * WARNING: Lustre networking configuration changes are required with
4291 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
4294 * Support for kernels:
4295 2.6.9-22.0.2.EL (RHEL 4)
4296 2.6.5-7.244 (SLES 9)
4297 2.6.12.6 vanilla (kernel.org)
4300 Severity : enhancement
4301 Bugzilla : 7981/8208
4302 Description: Introduced Lustre Networking (LNET)
4303 Details : LNET is new networking infrastructure for Lustre, it includes
4304 a reorganized network configuration mode (see the user
4305 documentation for full details) as well as support for routing
4306 between different network fabrics. Lustre Networking Devices
4307 (LNDs) for the supported network fabrics have also been
4308 created for this new infrastructure.
4310 Severity : enhancement
4311 Description: Introduced Access control lists
4312 Details : clients can set ACLs on files and directories in order to have
4313 more fine-grained permissions than the standard Unix UGO+RWX.
4314 The MDS must be started with the "-o acl" mount option.
4316 Severity : enhancement
4317 Description: Introduced filesystem quotas
4318 Details : Administrators may now establish per-user quotas on the
4321 Severity : enhancement
4323 Description: Configuration change for the XT3
4324 The PTLLND is now used to run Lustre over Portals on the XT3
4325 The configure option(s) --with-cray-portals are no longer used.
4326 Rather --with-portals=<path-to-portals-includes> is used to
4327 enable building on the XT3. In addition to enable XT3 specific
4328 features the option --enable-cray-xt3 must be used.
4333 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
4334 Details : A race between allocating a new llcd and re-getting the llcd_lock
4335 allowed another thread to grab newly-allocated llcd.
4337 Severity : enhancement
4339 Description: 2.6 OST async journal commit and locking fix to improve performance
4340 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
4341 2.6 kernels are now async as they already were in 2.4 kernels so
4342 that they can commit concurrently with the network bulk transfer.
4343 For block-allocated files the filter allocation semaphore is held
4344 to avoid filesystem fragmentation during allocation. BKL lock
4345 removed for 2.6 xattr operations where it is no longer needed.
4350 Description: lconf incorrectly determined whether two IP networks could talk
4351 Details : In some more complicated routing and multiple-network
4352 configurations, lconf will avoid trying to make a network
4353 connection to a disjoint part of the IP space. It was doing the
4354 math incorrectly for one set of cases.
4359 Description: Fix for potential infinite loop processing records in an llog.
4360 Details : If an llog record is corrupted/zeroed, it is possible to loop
4361 forever in llog_process(). Validate the llog record length
4362 and skip the remainder of the block on error.
4365 Frequency : occasional (liblustre only)
4367 Description: liblustre could not open files whose last component is a symlink
4368 Details : sysio_path_walk() would incorrectly pass the open intent to
4369 intermediate path components.
4372 Frequency : rare (liblustre only with non-standard tuning)
4373 Bugzilla : 7201 (7350)
4374 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4375 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4376 after it has been released, only do this for referenced locks
4378 Severity : enhancement
4380 Description: specifying an (invalid) directory default stripe_size of -1
4381 would reset the directory default striping
4382 Details : stripe_size -1 was used internally to signal directory stripe
4383 removal, now use "all default" to signal dir stripe removal
4384 as a directory striping of "all default" is not useful
4387 Frequency : common for large clusters running liblustre clients
4389 Description: doing an ls when liblustre clients are running is slow
4390 Details : sending a glimpse AST to a liblustre client waits for every AST
4391 to time out, as liblustre clients will not respond. Since they
4392 cannot cache data we refresh the OST lock LVB from disk instead.
4394 Severity : enhancement
4396 Description: doing an ls at the same time as file IO can be slow
4397 Details : enqueue and other "small" requests can be blocked behind many
4398 large IO requests. Create a new OST IO portal for non-IO
4399 requests so they can be processed faster.
4402 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4404 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4405 created file with mode 000
4406 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4407 allow an ftruncate() as a non-root user to a file with mode 000.
4408 Lustre now respects this flag to disable mode checks when
4409 truncating a file owned by the user
4412 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4415 Description: Revoking locks from clients that went dead or catatonic might take
4417 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4418 cancellation of such locks instant on servers without waiting for
4419 any reply from clients. Clients drops these locks when cancel
4420 notification from server is received without replying.
4423 Frequency : liblustre-only, when liblustre client dies or becomes busy
4425 Description: Doing ls on Linux clients can take a long time with active
4427 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4428 granting such locks to it in the first place if possible. Locks
4429 are taken by proxy on the OST during the read or write and
4430 dropped immediately afterward. Add connect flags handling, do
4431 not grant locks to liblustre clients for glimpse ASTs.
4433 Severity : enhancement
4435 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4436 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4437 sequential reads and behaves badly with applications doing random
4438 reads. Improve it by reading ahead at least read region, and
4439 avoiding excessive large RPC for small reads.
4441 Severity : enhancement
4443 Description: Creating more than 1000 files for a single job may cause a load
4444 imbalance on the OSTs if there are also a large number of OSTs.
4445 Details : qos_prep_create() uses an OST index reseed value that is an
4446 even multiple of the number of available OSTs so that if the
4447 reseed happens in the middle of the object allocation it will
4448 still utilize the OSTs as uniformly as possible.
4453 Description: OST or MDS may oops in ping_evictor_main()
4454 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
4455 but doesn't restart at beginning of obd_exports_timed list
4458 Severity : enhancement
4460 Description: improve by-nid export eviction on the MDS and OST
4461 Details : allow multiple exports with the same NID to be evicted at one
4462 time without re-searching the exports list.
4465 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
4467 Description: MDS may oops in groups_free()
4468 Details : in rare race conditions a newly allocated group_info struct is
4469 freed again, and this can be NULL. The 2.4 compatibility code
4470 for groups_free() checked for a NULL pointer, but 2.6 did not.
4473 Frequency : common for liblustre clients doing little filesystem IO
4474 Bugzilla : 9352, 7313
4475 Description: server may evict liblustre clients accessing contended locks
4476 Details : if a client is granted a lock or receives a completion AST
4477 with a blocking AST already set it would not reply to the AST
4478 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
4481 Frequency : lfs setstripe, only systems with more than 160 OSTs
4483 Description: unable to set striping with a starting offset beyond OST 160
4484 Details : llapi_create_file() incorrectly limited the starting stripe
4485 index to the maximum single-file stripe count.
4488 Frequency : LDAP users only
4490 Description: lconf did not handle in-kernel recovery with LDAP properly
4491 Details : lconf/LustreDB get_refs() is searching the wrong namespace
4493 Severity : enhancement
4495 Description: bind OST threads to NUMA nodes to improve performance
4496 Details : all OST threads are uniformly bound to CPUs on a single NUMA
4497 node and do their allocations there to localize memory access
4499 Severity : enhancement
4501 Description: llmount can determine client NID directly from Myrinet (GM)
4502 Details : the client NID code from gmnalnid was moved directly into
4503 llmount, removing the need to use this or specifying the
4504 client NID explicitly when mounting GM clients with zeroconf
4507 Frequency : if client is started with down MDS
4509 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
4510 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
4511 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
4512 ptlrpc_import_delayed_req() because we didn't check if the
4513 request was interrupted, and we also didn't break out of the
4514 event loop if there was no timeout
4519 Description: data loss during non-page-aligned writes to a single file from
4520 both multiple nodes and multiple threads on one node at same time
4521 Details : updates to KMS and lsm weren't protected by common lock. Resulting
4522 inconsistency led to false short-reads, that were cached and later
4523 used by ->prepare_write() to fill in partially written page,
4524 leading to data loss.
4527 Frequency : always, if lconf --abort_recovery used
4529 Description: lconf --abort_recovery fails with 'Operation not supported'
4530 Details : lconf was attempting to abort recovery on the MDT device and not
4533 Severity : enhancement
4535 Description: remove cleanup logs
4536 Details : replace lconf-generated cleanup logs with lustre internal
4537 cleanup routines. Eliminates the need for client-cleanup and
4540 Severity : enhancement
4542 Description: add support for EAs (user and system) on lustre filesystems
4543 Details : it is now possible to store extended attributes in the Lustre
4544 client filesystem, and with the user_xattr mount option it
4545 is possible to allow users to store EAs on their files also
4547 Severity : enhancement
4549 Description: Add possibility (config option) to show minimal available OST free
4551 Details : When compiled with --enable-mindf configure option, statfs(2)
4552 (and so, df) will return least minimal free space available from
4553 all OSTs as amount of free space on FS, instead of summary of
4554 free spaces of all OSTs.
4556 Severity : enhancement
4558 Description: do not expand extent locks acquired on OST-side
4559 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
4560 by server: they are not cached anyway.
4563 Frequency : when mmap is used/binaries executed from Lustre
4565 Description: Unmmap pages before throwing them away from read cache.
4566 Details : llap_shrink cache now attempts to unmap pages before discarding
4567 them (if unmapping failed - do not discard). SLES9 kernel has
4568 extra checks that trigger if this unmapping is not done first.
4573 Description: lconf didn't resolve symlinks before checking to see whether a
4574 given mountpoint was already in use
4577 Frequency : when migrating failover services
4578 Bugzilla : 6395, 9514
4579 Description: When migrating a subset of services from a node (e.g. failback
4580 from a failover service node) the remaining services would
4581 time out and evict clients.
4582 Details : lconf --force (implied by --failover) sets the global obd_timeout
4583 to 5 seconds in order to quickly disconnect, but this caused
4584 other RPCs to time out too quickly. Do not change the global
4585 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
4587 Severity : enhancement
4588 Frequency : if MDS is started with down OST
4589 Bugzilla : 9439,5706
4590 Description: Allow startup/shutdown of an MDS without depending on the
4591 availability of the OSTs.
4592 Details : Asynchronously call mds_lov_synchronize during MDS startup.
4593 Add appropriate locking and lov-osc refcounts for safe
4594 cleaning. Add osc abort_inflight calls in case the
4595 synchronize never started.
4598 Frequency : occasional (Cray XT3 only)
4600 Description: root not authorized to access files in CRAY_PORTALS environment
4601 Details : The client process capabilities were not honoured on the MDS in
4602 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
4603 been accessed by an authorized user then root was able to access
4604 the file on the local client also. The root user capabilities
4605 are now allowed on the MDS, as this environment has secure UID.
4608 Frequency : occasional
4610 Description: ldiskfs "too long searching" message happens too often
4611 Details : A debugging message (otherwise harmless) prints too often on
4612 the OST console. This has been reduced to only happen when
4613 there are fragmentation problems on the filesystem.
4618 Description: Division by zero in statfs when all OSCs are inactive
4619 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
4620 lov_statfs divides by value returned by lov_get_stripecnt().
4624 Bugzilla : 9489, 3273
4625 Description: First write from each client to each OST was only 4kB in size,
4626 to initialize client writeback cache, which caused sub-optimal
4627 RPCs and poor layout on disk for the first writen file.
4628 Details : Clients now request an initial cache grant at (re)connect time
4629 and so that they can start streaming writes to the cache right
4630 away and always do full-sized RPCs if there is enough data.
4631 If the OST is rebooted the client also re-establishes its grant
4632 so that client cached writes will be honoured under the grant.
4637 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
4638 Details : Now I/O RPCs go to different portal number and (presumably) fast
4639 lock requests (and glimses) and other RPCs get their own service
4640 threads pool that should be able to service those RPCs
4643 Severity : enhancement
4645 Description: Ability to exchange lustre version between client and servers and
4646 issue warnings at client side if client is too old. Also for
4647 liblustre clients there is ability to refuse connection of too old
4649 Details : New 'version' field is added to connect data structure that is
4650 filled with version info. That info is later checked by server and
4654 Frequency : rare, liblustre only.
4655 Bugzilla : 9296, 9581
4656 Description: Two simultaneous writes from liblustre at offset within same page
4657 might proceed at the same time overwriting eachother with stale
4659 Details : I/O lock withing llu_file_prwv was released too early, before data
4660 actually was hitting the wire. Extended lock-holding time until
4661 server acknowledges receiving data.
4664 Frequency : extremely rare. Never observed in practice.
4666 Description: avoid generating lustre_handle cookie of 0.
4667 Details : class_handle_hash() generates handle cookies by incrementing
4668 global counter, and can hit 0 occasionaly (this is unlikely, but
4669 not impossible, because initial value of cookie counter is
4670 selected randonly). Value of 0 is used as a sentinel meaning
4671 "unassigned handle" --- avoid it. Also coalesce two critical
4672 sections in this function into one.
4674 Severity : enhancement
4676 Description: allow liblustre clients to delegate truncate locking to OST
4677 Details : To avoid overhead of locking, liblustre client instructs OST to
4678 take extent lock in ost_punch() on client's behalf. New connection
4679 flag is added to handle backward compatibility.
4681 Severity : enhancement
4682 Bugzilla : 4928, 7341, 9758
4683 Description: allow number of OST service threads to be specified
4684 Details : a module parameter allows the number of OST service threads
4685 to be specified via "options ost ost_num_threads={N}" in the
4686 OSS's /etc/modules.conf or /etc/modprobe.conf.
4690 Bugzilla : 6146, 9635, 9895
4691 Description: servers crash with bad pointer in target_handle_connect()
4692 Details : In rare cases when a client is reconnecting it was possible that
4693 the connection request was the last reference for that export.
4694 We would temporarily drop the export reference and get a new
4695 one, but this may have been the last reference and the export
4696 was just destroyed. Get new reference before dropping old one.
4698 Severity : enhancement
4699 Frequency : if client is started with failover MDS
4701 Description: Allow multiple MDS hostnames in the mount command
4702 Details : Try to read the configuration from all specified MDS
4703 hostnames during a client mount in case the "primary"
4706 Severity : enhancement
4708 Description: Stop sending data to evicted clients as soon as possible.
4709 Details : Check if the client we are about to send or are sending data to
4710 was evicted already. (Check is done every second of waiting,
4711 for which l_wait_event interface was extended to allow checking
4712 of exit condition at specified intervals).
4715 Frequency : rare, normally only when NFS exporting is done from client
4717 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
4719 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
4722 Frequency : always, if srand() is called before liblustre initialization
4724 Description: Liblustre uses system PRNG disturbing its usage by user application
4725 Details : Introduce internal to lustre fast and high-quality PRNG for
4726 lustre usage and make liblustre and some other places in generic
4727 lustre code to use it.
4729 Severity : enhancement
4730 Bugzilla : 9477, 9557, 9870
4731 Description: Verify that the MDS configuration logs are updated when xml is
4732 Details : Check if the .xml configuration logs are newer than the config
4733 logs stored on the MDS and report an error if this is the case.
4734 Request --write-conf, or allow starting with --old_conf.
4736 Severity : enhancement
4738 Description: Handle symlinks in the path when checking if Lustre is mounted.
4739 Details : Resolve intermediate symlinks when checking if a client has
4740 mounted a filesystem to avoid duplicate client mounts.
4745 Description: lconf can hit an error exception but still return success.
4746 Details : The lconf command catches the Command error exception at the top
4747 level script context and will exit with the associated exit
4748 status, but doesn't ensure that this exit status is non-zero.
4753 Description: failure of ptlrpc thread startup can cause oops
4754 Details : Starting a ptlrpc service thread can fail if there are a large
4755 number of threads or the server memory is very fragmented.
4756 Handle this without oopsing.
4759 Frequency : always, only if liblustre and non-default acceptor port was used
4761 Description: liblustre cannot connect to servers with non-default acceptor port
4762 Details : tcpnal_set_default_params() was not called and was therefore
4763 ignoring the environment varaible TCPNAL_PORT, as well as other
4764 TCPNAL_ environment variables
4769 Description: two objects could be created on the same OST for a single file
4770 Details : If an OST is down, in some cases it was possible to create two
4771 objects on a single OST for a single file. No problems other
4772 than potential performance impact and spurious error messages.
4776 Bugzilla : 5681, 9562
4777 Description: Client may oops in ll_unhash_aliases
4778 Details : Client dcache may become inconsistent in race condition.
4779 In some cases "getcwd" can fail if the current directory is
4785 Description: Inode refcounting problems in NFS export code
4786 Details : link_raw functions used to call d_instantiate without obtaining
4787 extra inode reference first.
4791 Bugzilla : 9942, 9903
4792 Description: Referencing freed requests leading to crash, memleaks with NFS.
4793 Details : We used to require that call to ll_revalidate_it was always
4794 followed by ll_lookup_it. Also with revalidate_special() it is
4795 possible to call ll_revalidate_it() twice for the same dentry
4796 even if first occurence returned success. This fix changes semantic
4797 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
4798 reference on a request referred from the intent.
4799 ll_intent_release() then releases such a request.
4802 Frequency : rare, normally benchmark loads only
4804 Description: unlinked inodes were kept in memory on the client
4805 Details : If a client is repeatedly creating and unlinking files it
4806 can accumulate a lot of stale inodes in the inode slab cache.
4807 If there is no other client load running this can cause the
4808 client node to run out of memory. Instead flush old inodes
4809 from client cache that have the same inode number as a new inode.
4812 Frequency : SLES9 2.6.5 kernel and long filenames only
4813 Bugzilla : 9969, 10379
4814 Description: utime reports stale NFS file handle
4815 Details : SLES9 uses out-of-dentry names in some cases, which confused
4816 the lustre dentry revalidation. Change it to always use the
4820 Frequency : rare, unless heavy write-truncate concurrency is continuous
4821 Bugzilla : 4180, 6984, 7171, 9963, 9331
4822 Description: OST becomes very slow and/or deadlocked during object unlink
4823 Details : filter_destroy() was holding onto the parent directory lock
4824 while truncating+unlinking objects. For very large objects this
4825 may block other threads for a long time and slow overall OST
4826 responsiveness. It may also be possible to get a lock ordering
4827 deadlock in this case, or run out of journal credits because of
4828 the combined truncate+unlink. Solution is to do object truncate
4829 first in one transaction without parent lock, and then do the
4830 final unlink in a new transaction with the parent lock. This
4831 reduces the lock hold time dramatically.
4834 Frequency : rare, 2.4 kernels only
4836 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
4837 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
4838 before unmounting their filesystem in order to avoid deadlock.
4839 The kernel_locked() function in 2.4 kernels only checks whether
4840 the kernel lock is held, not whether it is this process that is
4841 holding it as 2.6 kernels do.
4846 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
4847 Details : The client ptlrpc code may be trying to reconnect to a down
4848 server before a previous connection attempt has timed out.
4849 Increase the reconnect interval to be longer than the connection
4850 timeout interval to avoid sending duplicate connections to
4854 Frequency : echo_client brw_test command
4856 Description: fix echo_client to work with OST preallocated code
4857 Details : OST preallocation code (5137) didn't take echo_client IO path
4858 into account: echo_client calls filter methods outside of any
4859 OST thread and, hence, there is no per-thread preallocated
4860 pages and buffers to use. Solution: hijack pga pages for IO. As
4861 a byproduct, this avoids unnecessary data copying.
4865 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
4866 Description: Client can oops in mdc_commit_close() after open replay
4867 Details : It was possible for the MDS to return an open request with no
4868 transaction number in mds_finish_transno() if the client was
4869 evicted, but without actually returning an error. Clients
4870 would later try to replay that open and may trip an assertion
4871 Simplify the client close codepath, and always return an error
4872 from the MDS in case the open is not successful.
4875 Frequency : rare, 2.6 OSTs only
4877 Description: OST may deadlock under high load on fragmented files
4878 Details : If there was a heavy load and highly-fragmented OST filesystems
4879 it was possible to have all the OST threads deadlock waiting on
4880 allocation of biovecs, because the biovecs were not released
4881 until the entire RPC IO was completed. Instead, release biovecs
4882 as soon as they are complete to ensure forward IO progress.
4884 Severity : enhancement
4886 Description: Support for specifying external journal device at mount
4887 Details : If an OST or MDS device is formatted with an external journal
4888 device, this device major/minor is stored in the ext3 superblock
4889 and may not be valid for failover. Allow detecting and
4890 specifying the external journal at mount time.
4895 Description: Mounting an MDS with pending unlinked files may cause oops
4896 Details : target_finish_recovery() calls mds_postrecov() which returned
4897 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
4898 considers this an error and immediately begins cleaning up the
4899 lov, just after starting the mds_lov process
4901 Severity : enhancement
4903 Description: Implement 'lfs df' to report actual free space on per-OST basis
4904 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
4905 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
4906 usage of objects; '-h' to report in human readable format.
4908 ------------------------------------------------------------------------------
4910 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
4917 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
4918 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
4919 at startup during orphan unlinking. Instead, skip buddy update
4920 before setup, as it will be regenerated after recovery is complete.
4925 Description: If an OST is inactive, its locks might reference stale inodes.
4926 Details : lov_change_cbdata() must iterate over all namespaces, even if
4927 they are inactive to clear inode references from the lock.
4929 Severity : enhancement
4930 Frequency : occasional, if non-standard max_dirty_mb used
4932 Description: Client will block write RPCs if not enough grant
4933 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
4934 then the client will block writes while waiting for another RPC
4935 to complete instead of consuming its dirty limit. With change
4936 we get improved performance when max_dirty_mb is small.
4938 Severity : enhancement
4939 Bugzilla : 3389, 6253
4940 Description: Add support for supplementary groups on the MDS.
4941 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
4942 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
4943 lookups for user supplementary groups into a cache.
4947 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
4948 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
4949 for liblustre O_EXCL is enforced by MDS.
4953 Description: readdir never returns NULL in liblustre.
4954 Details : Corrected llu_iop_getdirentries logic, to return offset of next
4955 dentry in struct dirent.
4959 Frequency : liblustre only, depends on application IO pattern
4960 Description: liblustre clients evicted if not contacting servers
4961 Details : Don't put liblustre clients into the ping_evictor list, so
4962 they will not be evicted by the pinger ever.
4964 Severity : enhancement
4966 Description: Add ability to evict clients by NID from MDS.
4967 Details : By echoing "nid:$NID" string into
4968 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
4969 $NID would be instantly evicted from this MDS and from all active
4970 OSTs connected to it.
4974 Description: Do not query file size twice, somewhat slowing stat(2) calls.
4975 Details : lookup_it_finish() used to query file size from OSTs that was not
4980 Description: service threads change working directory to that of init
4981 Details : Starting lustre service threads may pin the working directory
4982 of the parent thread, making that filesystem busy. Threads
4983 now change to the working directory of init to avoid this.
4987 Frequency : during shutdown only
4988 Description: shutdown with a failed MDS or OST can cause unmount to hang
4989 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
4994 Frequency : 2.6 only
4995 Description: chmod/chown may include an extra supplementary group
4996 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
4997 supplementary group and if none is specified this is used.
5000 Bugzilla : 5479 (6816)
5002 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
5003 Details : If lookup is in progress on a file that is unlinked we might try
5004 to revalidate the inode and fail in revalidate after lookup is
5005 complete and ll_file_open() enqueues the open again but
5006 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
5009 Frequency : always, if lconf --abort_recovery used
5011 Description: lconf --abort_recovery fails with 'Operation not supported'
5012 Details : lconf was attempting to abort recovery on the MDT device and not
5015 ------------------------------------------------------------------------------
5017 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
5022 Frequency : rare (only unsupported configurations with a node running as an
5024 Bugzilla : 6514, 5137
5025 Description: Mounting a Lustre file system on a node running as an OST could
5027 Details : OSTs now preallocates memory needed to write out data at
5028 startup, instead of when needed, to avoid having to
5029 allocate memory in possibly low memory situations.
5030 Specifically, if the file system is mounted on on OST,
5031 memory pressure could force it to try to write out data,
5032 which it needed to allocate memory to do. Due to the low
5033 memory, it would be unable to do so and the node would
5034 become unresponsive.
5036 Severity : enhancement
5038 Description: Addition of lconf --service command line option
5039 Details : lconf now accepts a '--service <arg>' option, which is
5040 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
5042 Severity : enhancement
5044 Description: Failover mode is now the default for OSTs.
5045 Details : By default, OSTs will now run in failover mode. To return to
5046 the old behaviour, add '--failout' to the lmc line for OSTs.
5048 Severity : enhancement
5050 Description: Health checks are now provided for MDS and OSTs
5051 Details : Additional detailed health check information on MSD and OSTs
5052 is now provided through the procfs health_check value.
5055 Frequency : occasional, depends on IO load
5057 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
5058 after numerous create/delete cycles
5059 Details : The ext3 inode allocation policy would not allocate new inodes
5060 very well on the OSTs because there are no new directories
5061 being created. Instead we look for groups with free space if
5062 the parent directories are nearly full.
5067 Description: Network or server problems during mount may cause partially
5068 mounted clients instead of returning an error.
5069 Details : The config llog parsing code may overwrite the error return
5070 code during mount error handling, returning success instead
5076 Description: MDS can fail to allocate large reply buffers
5077 Details : After long uptimes the MDS can fail to allocate large reply
5078 buffers (e.g. zconf client mount config records) due to memory
5079 fragmentation or consumption by the buffer cache. Preallocate
5080 some large reply buffers so that these replies can be sent even
5081 under memory pressure.
5085 Frequency : rare (liblustre)
5086 Description: fsx running with liblustre complained that using truncate() to
5087 extend the file doesn't work. This patch corrects that issue.
5088 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
5089 ATTR_SIZE and lsm use in llu_setattr_raw.
5093 Frequency : rare, only 2.6 kernels
5094 Description: Unusual file access patterns on the MDS may result in inode
5095 data being lost in very rare circumstances.
5096 Details : Bad interaction between the ea-in-inode patch and the "no-read"
5097 code in the 2.6 kernel caused the inode and/or EA data not to
5098 be read from disk, causing single-file corruption.
5102 Frequency : rare, only 2.6 filesystems using extents
5103 Description: Heavy concurrent write and delete load may cause data corruption.
5104 Details : It was possible under high-load situations to have an extent
5105 metadata block in the block device cache from a just-unlinked
5106 file overwrite a newly-allocated data block. We now unmap any
5107 metadata buffers that alias just-allocated data blocks.
5111 Frequency : filesystems with default stripe_count larger than 77
5112 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
5113 Details : lconf specifies an inode size of 4096 bytes when the default
5114 stripe_count is larger than 77. This conflicts with the default
5115 inode density of 1 per 4096 bytes. Allocate smaller inodes in
5116 this case to avoid pinning too much memory for large EAs.
5118 ------------------------------------------------------------------------------
5120 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
5125 Frequency : rare (extremely heavy IO load with hundreds of clients)
5127 Description: Client is evicted, gets IO error writing to file
5128 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
5129 caused clients to be evicted for AST timeouts. The fixes in
5130 bug 5192 mean we no longer need to have such short AST timeouts
5131 so ldlm_timeout has been increased.
5134 Frequency : occasional during --force or --failover shutdown under load
5135 Bugzilla : 5949, 4834
5136 Description: Server oops/LBUG if stopped with --force or --failover under load
5137 Details : a collection of import/export refcount and cleanup ordering
5138 issues fixed for safer force cleanup
5141 Frequency : only filesystems larger than 120 OSTs
5142 Bugzilla : 5990, 6223
5143 Description: lfs getstripe would oops on a very large filesystem
5144 Details : lov_getconfig used kfree on vmalloc'd memory
5147 Frequency : only filesystems exporting via NFS to Solaris 10 clients
5148 Bugzilla : 6242, 6243
5149 Description: reading from files that had been truncated to a non-zero size
5150 but never opened returned no data
5151 Details : ll_file_read() reads zeros from no-object files to EOF
5156 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
5157 Details : There's little harm in aborting MDS/OSS recovery and letting it
5158 try again, so I removed the LASSERT and return an error instead.
5160 Severity : enhancement
5162 Description: New debugging infrastructure for tracking down data corruption
5163 Details : The I/O checksum code was replaced to: (a) control it at runtime,
5164 (b) cover more of the client-side code path, and (c) try to narrow
5165 down where problems occurred
5169 Bugzilla : 3819, 4364, 4397, 6313
5170 Description: Racing close and eviction MDS could cause assertion in mds_close
5171 Details : It was possible to get multiple mfd references during close and
5172 client eviction, leading to one thread referencing a freed mfd.
5174 Severity: : enhancement
5175 Bugzilla : 3262, 6359
5176 Description: Attempts to reconnect to servers are now more aggressive.
5177 Details : This builds on the enhanced upcall-less recovery that was added
5178 in 1.4.2. When trying to reconnect to servers, clients will
5179 now try each server in the failover group every 10 seconds. By
5180 default, clients would previously try one server every 25 seconds.
5185 Description: After recovery, certain operations trigger a failed
5186 assertion on a client.
5187 Details : Failing over an mds, using lconf -d --failover, while a
5188 client was doing a readdir() call would cause the client to
5189 LBUG after recovery completed and the readdir() was resent.
5191 Severity : enhancement
5193 Description: Default groups are now added by lconf
5194 Details : You can now run lconf --group <servicename> without having to
5195 manually add groups with lmc.
5198 Frequency : occasional
5200 Description: Nodes with an elan id of 0 trigger a failed assertion
5203 Frequency : always when accessing e.g. tty/console device nodes
5205 Description: tty and some other devices nodes cannot be used on lustre
5206 Details : file's private_data field is used by device data and lustre
5207 values in there got lost. New field was added to struct file to
5208 store fs-specific private data.
5211 Frequency : when exporting Lustre via NFS
5213 Description: NFSD failed occasionally when looking up a path component
5214 Details : NFSD is looking up ".." which was broken in ext3 directories
5215 that had grown large enough to become hashed.
5218 Frequency : Clusters with multiple interfaces not on the same subnet
5220 Description: Nodes will repeatedly try to reconnect to an interface which it
5221 cannot reach and report an error to the log.
5222 Details : Extra peer list entries will be created by lconf with some peers
5223 unreachable. lconf now validates the peer before adding it.
5226 Frequency : Only if a default stripe is set on the filesystem root.
5228 Description: Setting a default stripe on the filesystem root prevented the
5229 filesystem from being remounted.
5230 Details : The client was sending extra request flags in the root getattr
5231 request and did not allocate a reply buffer for the dir EA.
5234 Frequency : occasional, higher if lots of files are accessed by one client
5235 Bugzilla : 6159, 6097
5236 Description: Client trips assertion regarding lsm mismatch/magic
5237 Details : While revalidating inodes the VFS looks up inodes with ifind()
5238 and in rare cases can find an inode that is being freed.
5239 The ll_test_inode() code will free the lsm during ifind()
5240 when it finds an existing inode and then the VFS later attaches
5241 this free lsm to a new inode.
5245 Bugzilla : 6422, 7030
5246 Description: MDS deadlock between mkdir and client eviction
5247 Details : Creating a new file via mkdir or mknod (starting a transaction
5248 and getting the ns lock) can deadlock with client eviction
5249 (gets ns lock and trying to finish a synchronous transaction).
5252 Frequency : occasional
5253 Description: While starting a server, the fsfilt_ext3 module could not be
5255 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
5256 kernels. Previously, lconf would still use the ext3 name
5257 when trying to load modules. Now, it will correctly use
5258 ext3 on 2.4 and ldiskfs on 2.6.
5260 Severity : enhancement
5261 Description: The default stripe count has been changed to 1
5262 Details : The interpretation of the default stripe count (0, to lfs
5263 or lmc) has been changed to mean striping across a single
5264 OST, rather than all available. For general usage we have
5265 found a stripe count of 1 or 2 works best.
5267 Severity : enhancement
5268 Description: Add support for compiling against Cray portals.
5269 Details : Conditional compiling for some areas that are different
5273 Frequency : occasional
5274 Bugzilla : 6409, 6834
5275 Description: Creating files with an explicit stripe count may lead to
5276 a failed assertion on the MDS
5277 Details : If some OSTs are full or unavailable, creating files may
5278 trigger a failed assertion on the MDS. Now, Lustre will
5279 try to use other servers or return an error to the
5283 Frequency : occasional
5285 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
5286 caused lock timeout during readahead (since 1.4.2).
5287 Details : Processes doing readahead might match a lock that hasn't been
5288 granted yet if there are overlapping and conflicting lock
5289 requests. The readahead process waits on ungranted lock
5290 (original lock is CBPENDING), while OST waits for that process
5291 to cancel CBPENDING read lock and eventually evicts client.
5293 Severity : enhancement
5295 Description: Initial enabling of flock support for clients
5296 Details : Implements fcntl advisory locking and file status functions.
5297 This feature is provided as an optional mount flag (default
5298 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
5299 locking are implemented yet, and those that are are not guaranteed
5300 to be completely correct in production environments.
5301 mount -t lustre -o [flock|noflock] ...
5304 Frequency : occasional
5306 Description: OSTs running 2.4 kernels but with extents enabled might trip an
5307 assertion in the ext3 JBD (journaling) layer.
5308 Details : The b_committed_data struct is protected by the big kernel lock
5309 in 2.4 kernels, serializing journal_commit_transaction() and
5310 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
5311 access to this struct. In 2.6 kernels there is finer grained
5312 locking to improve SMP performance of the JBD layer.
5316 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
5321 Description: merge and cleanup kernel patches.
5324 -----------------------------------------------------------------------------
5326 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
5328 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
5329 versions of Lustre. Please update all servers and clients to
5330 version 1.4.2 or later at the same time. You must also run
5331 "lconf --write-conf {config}.xml" on the MDS while it is stopped
5332 to update the configuration logs.
5334 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
5335 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
5336 - swab llog records of type '0' so we get proper header size/idx (5861)
5337 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
5338 - fix rename of one directory over another leaking an inode (5953)
5339 - avoid SetPageDirty on 2.6 (5981)
5340 - don't re-add just-being-destroyed locks to the waiting list (5653)
5341 - when creating new directories, inherit the parent's custom
5342 striping settings if present parent (3048)
5343 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
5344 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
5345 - don't hold client locks on temporary worklist from l_lru (5666)
5346 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
5347 - automatically evict dead clients (5921)
5348 - Update file size properly in create+truncate+fstat case (6196)
5349 - Do not unhash mountpoint dentries, do not allow removal of
5351 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
5352 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
5353 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
5354 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
5355 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
5356 - Correctly handle reads of files with no objects (6243)
5357 - lctl recover will also mark a device active if deactivate used (5933)
5359 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
5360 - allow --write-conf on an MDS with different nettype than client (5619)
5361 - don't write config llogs to MDS for mounts not from that MDS (5617)
5362 - lconf should create multiple TCP connections from a client (5201)
5363 - init scripts are now turned off by default; run chkconfig --on
5364 lustre and chkconfig --on lustrefs to use them
5365 - upcalls are no longer needed for clients to recover to failover
5367 - add --abort-recovery option to lconf to abort recovery on device
5369 - add support for an arbitrary number of OSTs (3026)
5370 - Quota support protocol changes.
5371 - forward compatibility changes to wire structs (6007)
5372 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5373 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5374 - improved "lustre status" script
5375 - initialize blocksize for non-regular files (6062)
5376 - added --disable-server and --disable-client configure options (5782)
5377 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5378 - Vanilla 2.4.29 support
5379 - increase maximum number of obd devices to 520 (6242)
5380 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5381 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5382 SLES 9 2.6.5 kernels
5383 - limit stripes per file to 160 (the maximum EA size) (6093)
5385 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5388 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5389 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5390 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5391 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5392 - small ext3 extents cleanups and fixes (5733)
5393 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5394 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5395 - avoid needless client->OST connect, fix handle mismatch (5317)
5396 - fix DLM error path that led to out-of-sync client, long delays (5779)
5397 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5398 - fix several locking issues related to i_size (5492,5624,5654,5672)
5399 - don't move pending lock onto export if it is already evicted (5683)
5400 - fix kernel oops when creating .foo in unlinked directory (5548)
5401 - fix deadlock in obdfilter statistics vs. object create (5811)
5402 - use time_{before,after} to avoid timer jiffies wrap (5882)
5403 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5404 - Do not leak request if server was not able to process it (5154)
5405 - If mds_open unable to find parent dir, make that negative lookup(5154)
5406 - don't create new directories with extent-mapping (5909, 5936)
5408 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5409 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5410 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5411 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5412 - support for RHEL4 kernel on i686 (5773)
5413 - provide error messages when incompatible logs are encountered (5898)
5415 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5416 * version 1.4.0.10 (1.4.1 release candidate 1)
5418 - don't keep a lock reference when lock is not granted (4238)
5419 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5420 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5421 - drop import inflight refcount on signal_completed_replay error (5255)
5422 - unlock page after async write error during send (3677)
5423 - handle missing objects in filter_preprw_read properly (5265)
5424 - no transno return for symlink open, don't save no-trasno open (3440)
5425 - don't try to complete elan receive that already failed (4012)
5426 - free RPC server reply state on error (5406)
5427 - clean up thread from ptlrpc_start_thread() on error (5160)
5428 - readahead could read extra page into cache that wasn't ejected (5388)
5429 - prevent races in class_attach/setup/cleanup/detach (5260)
5430 - don't dereference de->d_inode after l_dput of de (5458)
5431 - use "int" for stripe value returned from lock_to_stripe (5544)
5432 - mballoc allocation and error-checking fixes in 2.6 (5504)
5433 - block device patches to fix I/O request sizes in 2.6 (5482)
5434 - look up hostnames for IB nals (5602)
5435 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5436 - don't start multiple acceptors for the same port (5277)
5437 - fix incorrect LASSERT in mds_getattr_name (5635)
5438 - export a proc file for general "ping" checking (5628)
5439 - fix "lfs check" to not block when the MDS is down (5628)
5441 - service request history (4965)
5442 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5443 - create an "evict_client" /proc entry on OSTs, like the MDS has
5444 - fix mount usage message, return errors per mount(8) (5168)
5445 - change grep [] to grep "[]" in tests so they work in more UMLs
5446 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5447 - remove ancient LOV_MAGIC_V0 EA support (5047)
5448 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5449 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5450 - allow readahead allocations to fail when low on memory (5383)
5451 - mmap locking landed again, after considerable improvement (2828)
5452 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
5454 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
5457 - send OST transaction number in read/write reply to free req (4966)
5458 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5459 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5460 - the watchdog thread now runs as interruptible (5246)
5461 - flock/lockf fixes (but it's still disabled, pending 5135)
5462 - don't use EXT3 constants in llite code (5094)
5463 - memory shortage at startup could cause assertion (5176)
5465 - reorganization of lov code
5466 - single portals codebase
5468 - add extents/mballoc support (5025)
5469 - direct I/O reads in the obdfilter (4048)
5470 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
5472 tbd Cluster File Systems, Inc. <info@clusterfs.com>
5475 - send OST transaction number in read/write reply to free req (4966)
5476 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5477 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5478 - the watchdog thread now runs as interruptible (5246)
5479 - handle missing objects in filter_preprw_read properly (5265)
5480 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5481 - drop import inflight refcount on signal_completed_replay error (5255)
5482 - unlock page after async write error during send (3677)
5483 - return original error code on reconstructed replies (3761)
5484 - no transno return for symlink open, don't save no-trasno open (3440)
5486 - add pid to ldlm debugging output (4922)
5487 - bump the watchdog timeouts -- we can't handle 30sec yet
5488 - extra debugging for orphan dentry/inode bug (5259)
5490 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
5493 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
5494 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
5495 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
5496 - fix NULL dereference in /proc/sys/portals/routes (4827)
5497 - allow failed mdc_close() operations to be interrupted (4561)
5498 - stop precreate on OST before MDS would time out on it (4778)
5499 - don't send partial-page writes before EOF from client (4410)
5500 - discard client grant for sub-page writes on large-page clients (4520)
5501 - don't free dentries not owned by NFS code, check generation (4806)
5502 - fix lsm leak if mds_create_objects() fails (4801)
5503 - limit debug_daemon file size, always print CERROR messages (4789)
5504 - use transno after validating reply (3892)
5505 - process timed out requests if import state changes (3754)
5506 - update mtime on OST during writes, return in glimpse (4829)
5507 - add mkfsoptions to LDAP (4679)
5508 - use ->max_readahead method instead of zapping global ra (5039)
5509 - don't interrupt __l_wait_event() during strace
5511 - add software watchdogs to catch hung threads quickly (4941)
5512 - make lustrefs init script start after nfs is mounted
5513 - fix CWARN/ERROR duplication (4930)
5514 - return async write errors to application if possible (2248)
5515 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
5516 - print NAL number in %x format (4645)
5517 - update barely-supported suse-2.4.21-171 series (4842)
5518 - support for sles 9 %post scripts
5519 - support for building 2.6 kernel-source packages
5520 - support for sles km_* packages
5522 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
5525 - ignore -ENOENT errors in osc_destroy (3639)
5526 - notify osc create thread that OSC is being cleaned up (4600)
5527 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
5528 - reconstruct ost_handle() like mds_handle() (4657)
5529 - create a new thread to do import eviction to avoid deadlock (3969)
5530 - let lconf resolve symlinked-to devices (4629)
5531 - don't unlink "objects" from directory with default EA (4554)
5532 - hold socknal file ref over connect in case target is down (4394)
5533 - allow more than 32000 subdirectories in a single directory (3244)
5534 - fix blocks count for O_DIRECT writes (3751)
5535 - OST returns ENOSPC from object create when no space left (4539)
5536 - don't send truncate RPC if file size isn't changing (4410)
5537 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
5538 - bind to privileged port in socknal and tcpnal (3689)
5540 - rate limit CERROR/CWARN console message to avoid overload (4519)
5541 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
5542 - basic mmap support (3918)
5543 - kernel patch series update from b1_4 (4711)
5545 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
5548 - avoid crash during MDS cleanup with OST shut down (2775)
5549 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
5550 - don't use bad inodes on the MDS (3744)
5551 - dynamic object preallocation to improve recovery speed (4236)
5552 - don't hold spinlock over lock dumping or change debug flags (4401)
5553 - don't zero obd_dev when it is force cleaned (3651)
5554 - print grants to console if they go negative (4431)
5555 - "lctl deactivate" will stop automatic recovery attempts (3406)
5556 - look for existing locks in ldlm_handle_enqueue() (3764)
5557 - don't resolve lock handle twice in recovery avoiding race (4401)
5558 - revalidate should check working dir is a directory (4134)
5560 - don't always mark "slow" obdfilter messages as errors (4418)
5562 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
5565 - don't close LustreDB during write_conf until it is done (3860)
5566 - fix typo in lconf for_each_profile (3821)
5567 - allow dumping logs from multiple threads at one time (3820)
5568 - don't allow multiple threads in OSC recovery (3812)
5569 - fix debug_size parameters (3864)
5570 - fix mds_postrecov to initialize import for llog ctxt (3121)
5571 - replace config semaphore with spinlock (3306)
5572 - be sure to send a reply for a CANCEL rpc with bad export (3863)
5573 - don't allow enqueue to complete on a destroyed export (3822)
5574 - down write_lock before checking llog header bitmap (3825)
5575 - recover from lock replay timeout (3764)
5576 - up llog sem before sending rpc (3652)
5577 - reduce ns lock hold times when setting kms (3267)
5578 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
5579 - fix NULL deref and obd_dev leak on setup error (3312)
5580 - replace some LBUG about llog ops with error handling (3841)
5581 - don't match INVALID dentries from d_lookup and spin (3784)
5582 - hold dcache_lock while marking dentries INVALID and hashing (4255)
5583 - fix invalid assertion in ptlrpc_set_wait (3880)
5585 - add libwrap support for the TCP acceptor (3996)
5586 - add /proc/sys/portals/routes for non-root route listing (3994)
5587 - allow setting MDS UUID in .xml (2580)
5588 - print the stack of a process that LBUGs (4228)
5590 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
5593 - don't cleanup request in ll_file_open() on failed MDS open (3430)
5594 - make sure to unset replay flag from failed open requests (3440)
5595 - if default stripe count is 0, use OST count for inode size (3636)
5596 - update parent mtime/ctime on client for create/unlink (2611)
5597 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
5598 - free recovery state on server during a forced cleanup (3571)
5599 - unregister_reply for resent reqs (3063)
5600 - loop back devices mounting and status check on 2.6 (3563)
5601 - fix resource-creation race that can provoke i_size == 0 (3513)
5602 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
5603 - more debugging for page-accounting assertion (3746)
5604 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
5605 - avoid deadlock after precreation failure (3758)
5606 - fix race and lock order deadlock in orphan handling (3450, 3750)
5607 - add validity checks when grabbing inodes from l_ast_data (3599)
5609 - add /proc/.../recovery_status to obdfilter (3428)
5610 - lightweight CDEBUG infrastructure, debug daemon (3668)
5611 - change default OSC RPC parameters to be better on small clusters
5612 - turn off OST read cache for files smaller than 32MB
5613 - install man pages and include them in rpms (3100)
5614 - add new init script for (un)mounting lustre filesystems (2593)
5615 - run chkconfig in %post for init scripts (3701)
5616 - drop scimac NAL (unmaintained)
5618 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
5621 - clean kiobufs before and after use (3485)
5622 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
5623 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
5624 - change init script to fail more clearly if not run as root (1528)
5625 - allow clients to reconnect during replay (1742)
5626 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
5627 - don't do DNS lookups on NIDs too small for IP addresses (3442)
5628 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
5629 - fix cond_resched (3554)
5630 - only evict unfinished clients after recovery (3515)
5631 - allow bulk resend, prevent data loss (3570)
5632 - dynamic ptlrpc request buffer allocation (2102)
5633 - don't allow unlinking open directory if it isn't empty (2904)
5634 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
5635 - remove extraneous obd dereference causing LASSERT failure (3334)
5636 - don't use get_cycles() when creating temp. files on the mds (3156)
5637 - hold i_sem when setting i_size in ll_extent_lock() (3564)
5638 - handle EEXIST for set-stripe, set proper directory name (3336)
5640 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
5641 - fix ksocknal_fmb_callback() error messages (2918)
5643 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
5646 - don't copy lvb into (possibly NULL) reply on error (2983)
5647 - don't deref dentry after dput, don't free lvb on error (2922)
5648 - use the kms to determine writeback rpc length (2947)
5649 - increment oti_logcookies when osc is inactive (2948)
5650 - update client's i_blocks count via lvb messages (2543)
5651 - handle intent open/close of special files properly (1557)
5652 - mount MDS with errors=remount-ro, like obdfilter (2009)
5653 - initialize lock handle to avoid ASSERT on error cleanup (3057)
5654 - don't use cancelling-locks' kms values (2947)
5655 - use highest lock extent for kms, not last one (2925)
5656 - don't dereference ERR_PTR() dentry in error handling path (3107)
5657 - fix thread race in portals_debug_dumplog() (3122)
5658 - create lprocfs device entries at setup instead of at attach (1519)
5659 - common AST error handler, don't evict client on completion race (3145)
5660 - zero nameidata in detach_mnt in 2.6 (3118)
5661 - verify d_inode after revalidate_special is valid in 2.6 (3116)
5662 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
5663 - initialize RPC timeout timer earlier for 2.6 (3219)
5664 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
5665 - print nal/nid for unknown nid (3258)
5666 - additional checks for oscc recovery before doing precreate (3284)
5667 - fix ll_extent_lock() error return code for 64-bit systems (3043)
5668 - don't crash in mdc_close for bad permissions on open (3285)
5669 - zero i_rdev for non-device files (3147)
5670 - clear page->private before handing to FS, better assertion (3119)
5671 - tune the read pipeline (3236)
5672 - fix incorrect decref of invalidated dentry (2350)
5673 - provide read-ahead stats and refine rpc in flight stats (3328)
5674 - don't hold journal transaction open across create RPC (3313)
5675 - update atime on MDS at close time (3265)
5676 - close LDAP connection when recovering to avoid server load (3315)
5677 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
5678 - don't leak open file on MDS after open resend (3325)
5679 - serialize filter_precreate and filter_destroy_precreated (3329)
5680 - loop device shouldn't call sync_dev() for nul device (3092)
5681 - clear page cache after eviction (2766)
5682 - resynchronize MDS->OST in background (2824)
5683 - refuse to mount the same filesystem twice on same mountpoint (3394)
5684 - allow llmount to create routes for mounting behind routers (3320)
5685 - push lock cancellation to blocking thread for glimpse ASTs (3409)
5686 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
5687 - fix rare problem with rename on htree directories (3417)
5689 - allow default OST striping configuration per directory (1414)
5690 - fix compilation for qswnal for 2.6 kernels (3125)
5691 - increase maximum number of MDS request buffers for large systems
5692 - change liblustreapi to be useful for external progs like lfsck (3098)
5693 - increase local configuration timeout for slow disks (3353)
5694 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
5696 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5699 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
5700 - don't overwrite extent policy data in reply if lock was blocked (2901)
5701 - drop filter export grants atomically with removal from device (2663)
5702 - del obd_self_export from work_list in class_disconnect_exports (2908)
5703 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
5704 - swab reply message in mdc_close, other PPC fixes (2464)
5705 - fix destroying of named logs (2325)
5706 - overwrite old logs when running lconf --write_conf (2264)
5707 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
5708 - fix race in target_handle_connect (2898)
5709 - mds_reint_create() should take same inode create lock (2926)
5710 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
5711 - don't close files for self_export to avoid uninitialized obd (2936)
5712 - allow MDS with the same name as client node (2939)
5713 - hold dentry reference for closed log files for unlink (2325)
5714 - reserve space for all logs during transactions (2059)
5715 - don't evict page beyond end of stripe extent (2925)
5716 - don't oops on a deleted current working directory (2399)
5717 - handle hard links to targets without a parent properly (2517)
5718 - don't dereference NULL lock when racing during eviction (2867)
5719 - don't grow lock extents when lots of conflicting locks (2919)
5721 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
5724 - account for cache space usage on clients to avoid data loss (974)
5725 - lfsck support in lustre kernel code (2349)
5726 - reduce journal credits needed for BRW writes (2370)
5727 - orphan handling to avoid losing space on client/server crashes
5728 - ptlrpcd can be blocked, stopping ALL progress (2477)
5729 - use lock value blocks to assist in proper KMS, faster stat (1021)
5730 - takes i_sem instead of DLM locks internally on obdfilter (2720)
5731 - recovery for initial connections (2355)
5732 - fixes for mds_cleanup_orphans (1934)
5733 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
5734 - block all file creations until orphan recovery completes (1901)
5735 - client remove rq_connection from request struct (2423)
5736 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
5737 - recovery timer race (2670)
5738 - mdc_close recovey bug (2532)
5739 - ptlrpc cleanup bug (2710)
5740 - mds timeout on local locks (2588)
5741 - namespace lock held during RPCs (2431)
5742 - handle interrupted sync write properly (2503)
5743 - don't try to handle a message that hasn't been replied to (2699)
5744 - client assert failure during cleanup after abort recovery (2701)
5745 - leak mdc device after failed mount (2712)
5746 - ptlrpc_check_set allows timedout requests to complete (2714)
5747 - wait for inflight reqs when ptlrpcd finishes (2710)
5748 - make sure unregistered services are removed from the srv_list
5749 - reset bulk XID's when resending them (caught by 1138 test)
5750 - unregister_bulk after timeout
5751 - fix lconf error (2694)
5752 - handle write after unfinished setstripe, stripe-only getstripe (2388)
5753 - readahead locks pages, leaves pending causing memory pressure (2673)
5754 - increase OST request buffers to 4096 on large machines (2729)
5755 - fix up permission of existing directories in simple_mkdir (2661)
5756 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
5757 - don't assign transno to errored transactions (2742)
5758 - don't delete objects on OST if given a bogus objid from MDS (2751)
5759 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
5760 - if rq_no_resend, then timeout request after recovery (2432)
5761 - fix MDS llog_logid record size, 64-bit array alignment (2733)
5762 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
5763 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
5764 - MDS recovery shouldn't delete objects that it has given out (2730)
5765 - if enqueue arrives after completion, don't clobber LVB (2819)
5766 - don't unlock pages twice when trigger_group_io returns error (2814)
5767 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
5768 - don't write pages to disk if there was an error (1450)
5769 - don't ping imports that have recovery disabled (2676)
5770 - take buffered bytes into account when balancing socknal conn (2817)
5771 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
5772 - reconnect unlink llog connection after MDS reconnects to OST (2816)
5773 - remove little-endian swabbing of llog records (1987)
5774 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
5775 - retry reposting request buffers if they fail (1191)
5776 - grow extent at grant time to avoid granting a revoked lock (2809)
5777 - lock revoke doesn't evict page if covered by a second lock (2765)
5778 - disable VM readahead to avoid reading outside lock extents (2805)
5780 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
5781 - updated kernel patches for hp-2.4.20 kernel (2681)
5783 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
5786 - fix truncated write corruption (2366)
5787 - fix for failed assertion in iopen_connect_dentry (1792,2517)
5789 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
5790 - ensure len(uuid) < 37 in lmc (1171)
5791 - fix ia64 OOPS in llog_test (2255)
5792 - zero end of page at obdfilter for partial page writes (2648)
5793 - don't leave stale dentries around after renames (bug 2428)
5794 - fix timeouts when evicting a client with a single lock held (2642)
5795 - set deadline for the initial HELLO message to drain (2634)
5796 - print out dotted-quad IP addresses in the socknal (2302)
5798 - additional debugging for MDS client eviction problem (2443)
5799 - fix mkfsoptions support for osts (2603, 2604)
5801 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
5804 - add series for the vanilla 2.6.0 kernel
5805 - add series for the vanilla 2.4.24 kernel
5806 - add series for a cray x86/64 UL kernel drop
5807 - fix xattr patches for the vanilla 2.4.19 series
5809 - generate true UUIDs in lmc (1171)
5810 - have portals stack dumping break in UML (2466)
5811 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
5812 - allocate with _NOFS in ldlm to avoid deadlock (1933)
5813 - wake callback waiting threads on client eviction (2460)
5814 - Add --ptldebug and --subsystem to lmc (1719)
5815 - update assertion to allow safe interrupt allocation
5816 - set rq_no_resend for cancel requests (2432)
5817 - recalculate ptlrpcd timeout after resend (2494)
5818 - call vfs_rmdir when removing pending directories (2368)
5819 - fix renaming a file to itself (2429)
5820 - lmc creates a default one-stripe lov (2454)
5821 - expand procfs space to handle large clusters (2326)
5822 - increase UML stack to avoid overflow
5823 - update lconf's list of debug and subsystem masks
5824 - fix lfs find --obd (2510)
5825 - /proc tunable for disabling filter read caching (2591)
5826 - stop rpm packages from altering slapd.conf (2301)
5827 - disable nagle in the socknal under 0conf (2578)
5828 - choose mds inode size based on stripe count (2572)
5829 - fix kernel-source rpm problems (2516)
5831 - add --disable-doc to avoid pdf generation (2421)
5832 - update documentation, tests, type-os, comments
5833 - avoid format warnings on ia64
5834 - remove the TOE NAL
5835 - tiny code cleanups by removing unused fields
5837 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
5840 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
5841 - remove the most blatant lies from BUILDING file (2371)
5842 - change default debug level to reasonable production setting
5843 - reduce client side cache size to reduce cache flush time
5844 - reduce max RPCs in flight to avoid unnecessary file fragmentation
5845 - make TCP zerocopy and pinger support enabled by default (2476)
5846 - sync writes completed after process exits caused crashes (2319)
5847 - maintain correct mount count on the MDS (2356)
5848 - backout 1557, because 2316 wasn't really fixed
5849 - better file I/O statistics gathering in /proc
5850 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
5851 - another kernel patch to fix zero-copy TCP function export
5852 - don't take duplicate lock when processing re-sent getattr (2420)
5853 - lctl uses obd_self_export instead of creating new conn (2353)
5854 - MDS/OST recovery case which requires object creation asserted (2425)
5855 - move lfs from /usr/sbin to /usr/bin in packages
5856 - fix race between mds_client_add and mds_client_free (2417)
5857 - use kmalloc instead of slabs in portals (2430)
5858 - don't create duplicate records when a failover MDS is present (2442)
5859 - remove unnecessary mount age check (2332)
5860 - don't remove directory inodes from locks prematurely (2451)
5861 - don't break if MDS service name is the same as hostname (2103)
5862 - fix races in client write RPC generation when cache full (2482)
5864 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
5867 - remove now-unused request->rq_obd (278)
5868 - if an allocation fails, print out how much memory we've used (1933)
5869 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
5870 - add the "configurable stack size" patch to most series files (1256)
5871 - ability to write large log records, for 100+ OST configs (2306)
5872 - fix NULL deref when filter_prep fails (2314)
5873 - fix operator precedence error in filter_sync
5874 - dynamic allocation of socknal TX descriptors (2315)
5875 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
5876 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
5877 - ensure that CWARN messages in Portals always get to the syslog
5878 - __init/__exit are not for prototype decls (ldlm_init/exit)
5879 - x86-64 compile warning fixes
5880 - fix gateway LMC keyword conflict (2318)
5881 - fix MDS lock inversions in getattr/reint paths (1844)
5882 - fix a rare lock re-ordering bug, which caused deadlock (2322)
5883 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
5884 - DLM race condition prevented some lock evictions (2328)
5885 - ENOMEM detection and retry on socknal sends (2230)
5886 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
5887 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
5888 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
5889 - better error messages when a client is rejected during recovery (1505)
5890 - avoid cancelling locks which were never granted, after failure (2330)
5891 - fix i_sem/journal inversion in mds_client_add (2333)
5892 - fix truncate/getattr lock cycle deadlock (2334)
5893 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
5894 - fix two rare exit paths which could leak an l_lock() ref (2321)
5895 - fencepost error in MDS/OST orphan recovery (2226)
5896 - make log record alignment 8 bytes (1988)
5897 - lstripe now fails when requested offset > ost_count (2237)
5898 - ensure that all kernel series have a complete list.h (1607)
5899 - fix crashes in special-file operations (2316)
5900 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
5902 - allow configurable automake binary, for testing new versions
5903 - small update to the lfs documentation
5905 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
5907 * fix negative export reference count in fsfilt_sync (2312)
5909 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
5910 * release candidate 0.9.1
5912 - orphans are moved into the PENDING directory for possible recovery
5913 - replayed opens now open by fid for orphan/rename safety (1042)
5914 - last close of an orphan inode generates a transno (683)
5915 - chdir() and mount() now pin the directory entry (1020)
5916 - avoid CERROR in normal ll_setattr_raw() error case (1500)
5917 - discard very old requests without processing them (1502)
5918 - remove some common, well-understood CERRORs (1505)
5919 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
5920 - clear "grant" flags in OST replies until OST grant code lands (1644)
5921 - fix read performance by not clobbering i_blksize on client (1598)
5922 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
5923 - make LRU size a /proc tunable, clears locks when reduced (707)
5924 - fix some lprocfs rot that prevented ptlbd from loading (1732)
5925 - server locks take references on exports now (1558)
5926 - build fixes for 2.4.20-rh trees (1663)
5927 - return an error from lov_create if all OSCs are inactive (1751)
5928 - fix import levels when a reconnect happens without a timeout (1597)
5929 - exit early from mds_open if we get a lookup error (1749)
5930 - partial page read at EOF wouldn't wait for disk before sending (1642)
5931 - avoid NULL deref in obdfilter when reading page past EOF (1592)
5932 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
5933 - fix LBUG in ll_it_open_error with rc = -2 (1861)
5934 - write/truncate lock inversion (1639)
5935 - Don't auto-load obdclass, portals modules during cleanup (1495)
5936 - fix timestamps from jumping to "now" (1763)
5937 - extra journal assertions (1648)
5938 - add an extra multiunlink test (1771)
5939 - fix read_record/write_record API (1776)
5940 - fix leak of offset_extent, possible incorrect i_size later (1772)
5941 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
5942 - Debugging for the kqswnal_get_idle_tx problems (1820)
5943 - Allow recovery to be attempted multiple times (1536)
5944 - Write out MDS last_rcvd file after it is first created (1600)
5945 - Fix tx_descriptor leak in failed transmit situations (1827)
5946 - ext3 journaling fixes for assertion failure after IO error (1871)
5947 - class_export_put() on freed export after completion AST error (1896)
5948 - Fix revalidate looping in VFS (1322)
5949 - Don't access a freed export during MDS_REINT timeout (1521)
5950 - Add open-unlink recovery support on the MDS (1673,1764)
5951 - Return an error if no MDS data was read from last_rcvd (1946)
5952 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
5953 - Don't setup a disk device that doesn't match exported UUID (317)
5954 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
5955 - avoid committing NULL handle in force close
5956 - local.sh is now a one-stripe LOV configuration
5957 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
5958 - fix ext3 htree duplicate directory entry corruption (1516)
5959 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
5960 - update to new LOV EA format (2097)
5961 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
5963 2003-06-15 Phil Schwan <phil@clusterfs.com>
5966 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
5967 - per-import/export recovery handling (958, 931, 959)
5968 - multiple last-rcvd slots, for serving multiple FSes (949)
5969 - connections are again shared between multiple imp/exports (963, 964)
5970 - "umount -f" would hang if any requests needed to be sent (393, 978)
5971 - avoid pinning large req buffer by copying for queued messages (989)
5972 - add "uuid" to "lctl device" command to help upcalls (991)
5973 - "open" RPCs with transnos would confuse recovery counters (1037)
5974 - do proper endian conversion of all wire messages (288, 340, 891)
5975 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
5976 - call waiting locks callback from LDLM recovery thread (1127, 1151)
5977 - fix ptlrpc_connection leak in target_handle_connect (1174)
5978 - fix import refcounting bug in OST and MDS cleanup (1134)
5979 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
5980 - fix very unlikely obd_types race condition (501)
5981 - remove osc_open hack for echo_client (1187)
5982 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
5983 - a failure in read_inode2 leads to deadlock (1139)
5984 - cancel ack-locks as soon as transaction is committed (1072)
5985 - fix major leaks and crashes in the bulk I/O path (937, 1057)
5986 - make sure to commitrw after any preprw to avoid deadlock (1162)
5987 - failing to execute a file in a lustre FS would lock inode (1203)
5988 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
5989 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
5990 - obd_uuid2tgt would walk off the end of the list (1255)
5991 - on IA64 the osc would give portals incorrect bulk size (1258)
5992 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
5993 - fix lock inversion caused by new llite matching code (1282)
5994 - limit the number of dirty pages on a client to 10MB (1286)
5995 - timed out locks were not being corrected cancelled (1289)
5996 - fix O_DIRECT above 4GB on IA-32 (1292)
5997 * major user-visible changes
5998 - fail out/fail over policy now controlled by the upcall (993)
6000 - add OBD_PING to check server availability and failure (954)
6001 - lustre messages are now sent in sending host order (288, 340, 891)
6002 - add eadatalen to MDS getattr reply (340)
6003 - OST read replies may contain second buffer, with per-page status (593)
6005 2003-03-11 Phil Schwan <phil@clusterfs.com>
6008 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
6009 - failed open()s could cause deadlock; fixed (867, 869)
6010 - stop cancelling OST locks when files are closed (481)
6011 - overlapping XID spaces caused network corruption (851, 853)
6012 - fix unsafe fsfilt counter arithmetic; change to atomic_t
6013 - setattr_raw added, to do single-RPC, server-side setattrs
6014 - lmc/lconf syntax change for OST UUIDs
6015 - fix crashy race condition between ptlrpc_free_req and osc_close
6016 - don't use request in mdc_enqueue if we hit a timeout (889)
6017 - don't set the inode i_size for regular files from the MDS (896)
6018 - handle out of order completion AST (842)
6019 - don't LBUG if a lock request times out after receiving AST (913)
6020 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
6021 - if a bad lock AST arrives, send an error instead of dropping entirely
6022 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
6023 - fix leak in bulk IO when only partially completed (899, 900, 926)
6024 - fix O_DIRECT for ia64 (55)
6025 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
6026 - C-z after timeout could hang a process forever; fixed (977)
6028 - client-side I/O cache (678, 924, 929, 941, 970)
6030 - READPAGE and SETATTRs which don't take server-side locks get
6033 2003-02-11 Phil Schwan <phil@clusterfs.com>
6036 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
6037 - Fix fsfilt_extN_readpage() to read a full page of directory
6038 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
6039 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
6040 - Fix imbalanced LOV object allocation and out-of-bound access (469)
6041 - Most intent operations were removed, in favour of a new RPC mode
6042 that does a single RPC to the server and bypasses most of the VFS
6043 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
6044 - Aggressively cancel local locks on DLM servers
6045 - mds_reint_unlink sends EA to the client if it's the last nlink.
6046 client uses that EA to unlink OST objects.
6047 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
6048 - recursive symlinks were fixed (439)
6049 - fixed NULL deref in DEBUG_REQ
6050 - filter_update_lastobjid no longer calls sync, which annoyed extN
6051 - fixed multi-client small-writes to a single file problem (445)
6052 - fixed mtime updates during file writes (607)
6053 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
6054 - fixed bug in obd_brw_read/write() (under guise of testing 367)
6055 - fixed Linux OST size reporting problem (444, 656)
6056 - OST now updates object mtime with writes or setattr (607, 619)
6057 - client verifies file size before zeroing page past EOF (445)
6058 - OST now writes last allocated objid to disk with allocation (108)
6059 - LOV on echo now works (409)
6061 - mds_reint_unlink sends a new buffer, with the EA included. this
6062 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
6063 set if a regular file was being unlinked, and it was the last link
6064 - use PtlGet from the target for bulk writes (315)
6065 - OST now updates object mtime with writes or setattr (607, 619)
6066 - LDLM now has a grant-time callback to revalidate locked items, if
6068 - Many MDS operations were reorganized to combat race conditions
6070 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
6071 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
6073 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
6076 - Fully reactivate OST imports after reconnection (512, others)
6077 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
6078 - More graceful error handling for truncating on dead OST (515)
6079 - Don't error out unless we're actually accessing dead stripes (474)
6080 - Fix garbage sizes when stripes are missing (410)
6081 - LRU counters were broken, causing constant lock purge (433, 432)
6082 - garbage on read from stripes with failed OSTs (441)
6083 - mark OSCs as active before reconnecting during recovery (438)
6084 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
6085 - lfind did not preserve OST order in output (443)
6086 - symlinks cause hung clients, incorrect data (439)
6087 - stop dereferencing request after dropping refcount (457)
6088 - don't LASSERT(spin_is_locked) on non-SMP (455)
6089 - fixes for many rename() bugs
6090 - fstat didn't correctly synchronize attributes (399)
6091 - server must handle lock cancellation during blocking AST prep (487)
6092 - bulk descriptors were free()d too soon (511)
6093 - fix paths in lconf, which would load incorrect modules (451, 507)
6094 - fix confusing lconf 'host not found' error message (386)
6095 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
6096 - fix race condition in mdc_blocking_ast() for inode access (526)
6097 - fix lov_unpackmd() unpacking wrong number of stripes (537)
6098 - fix lov_set_osc_active() marking wrong OSC inactive (440)
6099 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
6100 - fix multiple writes of stripe MD to MDS (358, maybe 519)
6101 - fix lstripe in several ways (kernel side) (527)
6102 - fix request leak in ldlm_cli_enqueue (262)
6103 - incorrect OSC was marked inactive after OST failure
6104 - call mds_fs_cleanup before unmounting filesystem (524)
6105 - fix races between taking ns_lock and ldlm_lock_change_resource
6106 - fix races updating LOV export open file list
6107 - fix lov_enqueue error path, avoid decref-ing bad lock handle
6108 - fix recovery NULL deref in ldlm_cli_cancel_unused
6109 - fix some DLM races by using new hash table for lock handles (419)
6110 - permit the client to specify desired inodes, at replay
6111 - duplicate requests when we queue them for replay reintegration
6112 - fix last_rcvd offset calculation
6113 - sync after each recovered transaction, so we always make progress
6114 - never, not always, ERESTART requests without transnos
6115 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
6117 - skip replay if the MDS says that the client is already connected
6118 - don't check for a recovery-enabled export to match lctl's UUID
6119 - don't INC_USE_COUNT for phantom exports
6120 - don't crash when cleaning up phantom exports (567)
6121 - don't double-finish or set replay data for errored mdc_open requests
6122 - abort requests when they time out, so we don't get old replies
6123 - send/receive replies for AST messages again
6124 - if the client says that it doesn't have the lock, cancel it on the
6126 - if we timeout during I/O, don't try to cancel an in-use lock; instead
6127 mark it as destroyed, it will all work out when decref is called
6128 - fix module use counts (22, 581)
6130 - ASTs now expect a reply (server cancels lock on error reply)
6132 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
6135 - fix many simultaneous client startup (392)
6136 - fix dentry->d_it clobbering
6137 - credentials weren't being shipped for readdir/getattr operations
6138 - remove invalid assertions triggered during some concurrent MD
6140 - proper Lustre versions added (336, 389)
6141 - fix memory leak for create error case (398)
6142 - fix LOV locking bug that would get cli/srv out of sync
6143 - fix echo client over LOV (409)
6144 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
6145 - fix double-O_EXCL intent crash (424)
6146 - avoid sending multiple lock CANCELs (352)
6148 - MDS can do multi-client recovery (modulo bugs in new code)
6150 - many updates, edits, cleanups
6152 2002-11-18 Phil Schwan <phil@clusterfs.com>
6155 - fix null d_it dereference (346)
6156 - fix full OST/dbench hang (333)
6157 - fix permission problem with file removal (286)
6158 - fix removal of OSCs from LOV when they fail
6159 - fix NULL deref during bulk timeout (214)
6160 - fix problems related to multiple filesystems on one MDS (241)
6161 - fixed serious subtle metadata locking bugs
6162 - free locks on clients when inodes are removed due to memory
6164 - fix inode pointer in lock data (285)
6165 - partial support for multiple MDS on a single host (241)
6166 - data locks weren't cancelled at clear_inode time (290, 311)
6167 - intent locks could lead to unbounded lock growth (205)
6168 - added a maximum lock count, an LRU list, and a flusher
6169 - fix multiple rename (365)
6170 - properly abstracted the echo client
6171 - OSC locked 1 byte too many; fixed
6172 - rewrote brw callback code:
6173 - fixed recovery bugs related to LOVs (306)
6174 - fixed too-many-pages-in-one-write crash (191)
6175 - fixed (again) crash in sync_io_timeout (214)
6176 - probably fixed callback-related race (385)
6178 - Add capability to MDS protocol
6179 - LDLM cancellations and callbacks on different portals
6181 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
6184 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
6185 - defer open object destruction to close time (601981, 138)
6186 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
6187 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
6188 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
6189 - don't call obd_setattr in ll_file_release for destroyed objects
6190 * protocol change to lustre_msg: move |version| and add |flags|
6191 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
6192 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
6193 * added replay of create, unlink, link and rename operations during
6194 MDS failover; recovery should be much more robust now
6195 * remove failed OSCs from LOVs (only lov_create uses this so far)
6196 * the lustre-HOWTO was brought (more) up to date (582544)
6198 2002-10-23 Phil Schwan <phil@clusterfs.com>
6201 - in-use dentries weren't being reused properly (617851)
6202 - prevent multiple LDLM setup (599178)
6203 - fix LOV size calculations for truncate (617853)
6204 - fix client handling of MDS intent errors (POSIX)
6205 - fix permission bug in lovstripe.c test (624321)
6206 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
6207 - truncate past end of file could corrupt data
6208 - proper cleanup after timeouts, crashes, etc (592524, 550815)
6209 - a race in recovery could return ETIMEDOUT to apps (623947)
6210 - building outside the source directory was fixed
6211 * the lustre-HOWTO was brought (more) up to date (582544)
6212 * major progress was made on recovery functionality
6214 2002-10-10 Phil Schwan <phil@clusterfs.com>
6217 - recovery deadlock fix
6218 - rm -rf causes LBUG fix (617817)
6219 - file open by multiple tasks fix (618962)
6220 - directory permissions bugs (602707 and 620007)
6221 - journal_stop fixed with locking (611313)
6222 - O_APPEND failures resolved (618273, perhaps 614459)
6223 - lconf PATH fix (619770)
6224 - IA64 build fix (621450)
6225 - RPC buffer sizes scale with amount of memory
6227 2002-10-01 Phil Schwan <phil@clusterfs.com>
6230 - locks would be cancelled without throwing away data pages,
6231 resulting in inconsistent data (605627)
6232 - inode attributes were not always being refreshed (605627, 612449)
6233 - lconf now continues to cleanup after lctl reports an error
6234 - MDS now enforces user permissions (602707)
6235 - lprocfs cleanup fixed, but not yet enabled (614157)
6236 - fixed infinite server hang, should a client not respond to an AST
6237 - avoid going into recovery if user calls readlink() with a buffer
6238 that's too small (613941)
6239 - AST RPCs no longer require replies (614867) -- this may be changed
6240 - don't crash server if client sends an IOV that's too big (611336)
6241 - fixed lock conversion deadlock (611892)
6242 - fixed the following of symlinks (614622)
6243 * recovery: the server can remove locks from a client that dies, other
6244 clients can make progress
6245 * more extN patch fixes
6246 * compile-time configurable ptlrpc buffer allocations
6248 - collaborative read cache document
6249 - Lustre Lite Performance CDR document-in-progress
6251 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6254 - fix typo in patch-2.4.18
6256 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6259 - clear ptlrpc request each time in handle_incoming_request()
6260 - unlink of files now destroys the object on the OST
6262 2002-09-19 Peter Braam <braam@clusterfs.com>
6264 * add hard link support
6265 * change obdfile creation method
6266 * kernel patch changed
6268 2002-09-19 Peter Braam <braam@clusterfs.com>
6271 - stack overflow bug in extN fixed
6273 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
6275 * documentation updates
6276 - add man pages for config tools
6277 - update tests/README to describe testing with new config tools
6278 - finish metadata API descriptions
6279 * bug fixes and cleanups
6280 - statfs workaround for 16TB limit
6281 - LOV stripe allocation improved, can stripe on subset of OSTs
6282 - LOV file size/IO offset was wrong for files > 4GB in size
6283 - object EA data was being dropped, caused files to be unreadable
6284 - memory overflow with non-LOV OST caused memory corruption
6285 - fixed regression tests to work with new config tools, obdfilter
6286 - fixed bug when directory size became larger than 1 block
6287 - fixed bug (for single client case) when PWD was deleted
6288 - invalidate local directory pages when doing intent-based ops
6289 - avoid LDLM oops when lock callback contained bad data
6291 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
6293 * documentation updates
6294 * bug fixes and cleanups
6295 - configuration tools
6298 - 64-bit compile warnings
6299 - 64-bit internal statfs data
6301 * test_brw on persistent OST devices
6303 * lprocfs (disabled)
6305 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
6307 * documentation updates
6308 * bug fixes and cleanups
6309 * configuration tools
6311 2002-08-30 Peter J. Braam <braam@clusterfs.com>
6314 * many small fixes to 0_5_4
6315 * io/network handling
6316 * thinkos in MDS operations
6318 2002-08-24 Peter J. Braam <braam@clusterfs.com>
6321 * crucial basic fixes to 0.5.3
6322 * IOR, Iozone work over Elan
6325 2002-08-07 Phil Schwan <phil@clusterfs.com>
6326 * version 0_5_3, our first alpha
6327 * we use the new Portals iovs
6328 * documentation updates
6329 * bug fixes and cleanups
6330 * small changes in the DLM wire protocol
6332 2002-07-25 Peter J. Braam <braam@clusterfs.com>
6333 * version 0_5_1 with some initial stability,
6334 * locking on MD and file I/O.
6335 * documentation updates
6336 * several bug fixes since 0.5.0
6337 * small changes in wire protocol
6339 2002-07-18 Phil Schwan <phil@clusterfs.com>
6341 * delivered as Lustre Light Alpha
6342 * fixed a crash after handling invalid MDS requests
6343 * fixed directory pages for architectures with non-4k pages sizes
6345 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
6346 * release version v0_4_4
6347 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
6349 * Fixes a number of interruption problems with OST operations.
6350 * Update documentation for portals header changes
6351 * Move all wire protocol structs/defines to lustre_idl.h
6352 * Fixes symlink length bug.
6353 * Add tcpdump to repository.
6355 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
6356 * release version v0_4_3
6357 * Fixes statfs for inodes on extN.
6358 * Fixes bug in runtests which would delete /etc/hosts.
6359 * Use 64-bit object IDs wherever possible (not into VFS though)
6360 Remove ost_get_info, which is unused by lustre, and out of date.
6362 2002-07-03 Peter Braam <braam@clusterfs.com>
6363 * release version v0_4_2 Fixes a lookup error (type not passed)
6364 * move forward to head of Portals
6365 * move forward to latest Lustre kernel
6367 2002-06-25 Peter Braam <braam@clusterfs.com>
6368 * release version v0_4_1. Hopefully stable on single node use.