3 * See http://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
4 * Server support for kernels:
5 2.6.32-220.4.2.el6 (RHEL6)
6 * Client support for unpatched kernels:
7 2.6.32-220.4.2.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 2.6.32-220.4.2.el6
22 Severity : enhancement
24 Description: Update RHEL6.2 kernel to 2.6.32-220.4.1.el6
26 Severity : enhancement
28 Description: Update RHEL6.2 kernel to 2.6.32-220.el6
30 Severity : enhancement
32 Description: Update RHEL6.1 kernel to 2.6.32-131.17.1.el6
34 --------------------------------------------------------------------------------
36 09-30-2011 Whamcloud, Inc.
38 * Server support for kernels:
39 2.6.18-238.19.1.el5 (RHEL5)
40 2.6.32-131.6.1.el6 (RHEL6)
41 * Client support for unpatched kernels:
42 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
43 2.6.18-238.19.1.el5 (RHEL5)
44 2.6.32-131.6.1.el6 (RHEL6)
45 2.6.29.4-167.fc11 (FC11)
46 2.6.16.60-0.69.1 (SLES10)
48 * Recommended e2fsprogs version: 1.41.90.wc3
49 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
50 Lustre filesystem with 4K stack may cause a stack overflow. For more
51 information, please refer to bugzilla 17630.
52 * NFSv4 reexport requires Lustre client on the re-exporting nodes to be
53 mounted with "32bitapi" mount option
55 Severity : enhancement
56 Description: Update RHEL6.1 kernel to 2.6.32-131.6.1.el6
58 Severity : enhancement
59 Description: Update RHEL5.6 kernel to 2.6.18-238.19.1.el5
61 Severity : enhancement
62 Description: Update RHEL5.6 kernel to 2.6.18-238.9.1.el5
65 Frequency : only with SLES10
67 Description: Use OFED "KMP" provided by Novell
68 Details : SLES10 SP3 ships with OFED in a separate "KMP" package.
69 Lustre is now built against this package. That means you need to
70 install the ofed-kmp package from Novell for the patchless client.
71 Note that the ofed-kmp that Novell ships may not exactly match the
72 kernel version but should still be compatible.
74 Severity : enhancement
76 Description: Update RHEL5.5 kernel to 2.6.18-194.17.1.el5
77 Update OEL5.5 kernel to 2.6.18-194.17.1.0.1.el5.
79 --------------------------------------------------------------------------------
81 2010-07-15 Oracle, Inc.
83 * Server support for kernels:
84 2.6.18-164.11.1.el5 (RHEL 5)
85 2.6.18-164.11.1.0.1.el5 (OEL 5)
86 * Client support for unpatched kernels:
87 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
88 2.6.18-164.11.1.el5 (RHEL 5),
89 2.6.18-164.11.1.0.1.el5 (OEL 5)
90 2.6.16.60-0.42.8 (SLES 10),
92 2.6.29.4-167.fc11 (FC11)
93 * Recommended e2fsprogs version: 1.41.10-sun2
95 Severity : enhancement
97 Description: Distribute lfs_migrate script with lustre
98 Details : The lfs_migrate mentioned in the manual is now included in the
99 lustre tree in lustre/scripts and will distributed in lustre RPMs.
101 Severity : enhancement
103 Description: Failover nid list added to import proc file.
105 Severity : enhancement
107 Description: "som_preview" mount option required to use SOM
108 Details : The clients will need to use som_preview mount option
112 Description: Lustre does not do 1MB IOs to HW RAID
113 Details : Bump MAX_PHYS/HW_SEGMENTS and SG_ALL to 256 in the RHEL5 kernel.
115 Severity : enhancement
117 Description: lfs quota output is not very convenient for awk/sed-parsing
118 Details : Some positions in lfs quota output table could be empty or
119 non-empty which made it hard to parse it with scripts, now
120 a dash is put instead of space where there is not supposed
125 Description: Update to OFED 1.5.1
126 Details : OFED 1.5.1 will be used in environments that don't use
127 vendor supplied OFED.
131 Description: Remove lnet. prefix from lctl params display
132 Details : Also renamed lnet.memused to lnet_memused to
137 Description: Use conf_param -d to remove permanant settings.
141 Description: Add list_param -R
142 Details : lctl list_param now supports recursive mode.
145 Bugzilla : 21829, 22312
146 Description: Fix llobdstat.
147 Details : 1. Now updating the write_bytes counter
148 2. Port script fixes from 1.8.
152 Description: decrease the usage of memory on clients.
153 Details : 1. On clients, recycle dentries and inodes unused.
154 2. Delete the code related to ll_deathrow(att 6215 in bug 1443). It
159 Description: llobdstat fix and enhancement
160 Details : add a counter to set a limit to how many samples will be returned
161 fix a wildcard in the path to limit to obdfilter stats only
165 Description: lfs check only allowed for root
166 Details : Code cleanups and added a regresion test of non-root use of
171 Description: MDS Oopses in ptlrpc_dispatch_difficult_reply()/__list_add()
172 Details : Fix hr->hr_index overload.
176 Description: Fix FIEMAP functionality.
180 Description: FC11 patchless client support added
184 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and
185 OEL5.4 kernel to 2.6.18-164.11.1.0.1.el5.
189 Description: New acc-sm script and various test-framework enhancments.
193 Description: Previously landed patch was reverted.
197 Description: Update SLES10 kernel to 2.6.16.60-0.42.8 (patchless)
201 Description: Add ldiskfs maxdirsize mount option.
205 Description: Fix mkfs.lustre support for 16TB LUNs.
209 Description: Print mmp_check_interval
210 Details : print mmp_check_interval and make it possible to abort
211 mount operation in case it takes too long.
215 Description: Add cascading_rw to lustre mpi tests.
219 Description: Store SOM data in EA.
220 Details : Also several SOM related cleanups.
224 Description: Add refcount to md_open_data
228 Description: Use refcount on lov_request_set to prevent memory corruption.
232 Description: Statahead: use parent dir's i_mutex for synchronization.
236 Description: loadgen improvements.
240 Description: Enable peer health detection in ksocklnd and ko2iblnd.
244 Description: Control DCACHE_LUSTRE_INVALID flag with MDS_INODE_LOOKCUP lock
248 Description: Fix race on obd_devcie:md_stats
252 Description: Fix SLES 11 client builds.
253 Details : Resolved uninitalized pointer gcc warnings.
257 Description: Allow non-root access for "lfs check".
258 Details : Added a check in obd_class_ioctl() for OBD_IOC_PING_TARGET.
262 Description: Clear imp_force_reconnect correctly in ptlrpc_connect_interpret()
264 Severity : enhancement
266 Description: Update kernel to RHEL5.4 2.6.18-164.6.1.el5 and
267 OEL5 2.6.18-164.6.1.0.1.el5(Both in-kernel OFED enabled).
271 Description: Sanity test_27p fails.
272 Details : OBD_FAIL_OST_ENOSPC has been moved to filter_statfs().
273 Sanity tests 27 were cleaned up and improved.
277 Description: Unify req format on client/servers
278 Details : Use new req_capsule API [almost] everywhere instead of old PTLRPC
279 buffers and swabbers approach..
284 Description: drop join file stuff.
286 Severity : enhancement
288 Description: Add 2.6.27 and SLES11 (patchless client) support
292 Description: Fix in ptlrpc_expire_one_request() to print the signed time
295 Severity : enhancement
297 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
300 Severity : enhancement
302 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
304 Severity : enhancement
306 Description: correctly handle big reply message.
307 Details : send LNet event if reply is bigger then buffer and adjust this buffer
312 Description: Drop unnecessary __GFP_NOMEMALLOC flag from filter_get_page()
314 Severity : enhancement
316 Description: Add support for OEL5.
320 Description: Conf-sanity.sh 50g test - deactivated OST should not cause a panic.
322 Severity : enhancement
324 Description: "weak-modules" support
325 Details : Implement "weak-modules" support which enables kernel modules
326 to be used with any kernel that implements the same kABI. In
327 order to achieve this modules are now installed in
328 /lib/modules/$(uname -r)/updates/kernel on all distributions.
332 Description: client could not reconnect to OST because of an active request.
333 Details : The client now retries to reconnect to the same server, if a
334 connect request failed with EBUSY or -EAGAIN. Test case 35b
335 conf-sanity.sh has been created.
339 Description: Don't allow make backward step on assignin osc next id.
340 Details : race between allocation next id and ll_sync thread can be cause
341 of set wrong osc next id and can be kill valid ost objects.
345 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
349 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
353 Description: Error handling in osc_statfs_interpret() has been improved.
354 Details : Check in osc_statfs_interpret() for EBADR.
358 Description: Do not send statfs() requests to OSTs disabled by administrator.
359 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
361 Severity : enhancement
363 Description: Added sanity-quota test to for lfs(1) to work with numeric uid/gid.
365 Severity : enhancement
367 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
369 Severity : enhancement
371 Description: Enhancement for lfs(1) command to use numeric uid/gid.
375 Description: Improved cleanup procedure for conf-sanity test_46a.
379 Description: Do not start bulk transfer, if client tried to reconnect.
380 Details : Do not start bulk transfer, if client tried to reconnect,
381 but failed due to active rpcs. Check for exp_abort_active_req
382 has been implemented.
386 Description: Prevent inconsistences between linux and lustre mount structures.
387 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
388 Make the sleep interruptible.
390 Severity : enhancement
392 Description: provide server to client comms path
393 Details : server to client communications path via new LDLM_SET_INFO rpc
397 Description: truncate starts GFP_FS allocation under transaction causing deadlock
398 Details : ldiskfs_truncate calls grab_cache_page which may start page
399 allocation under an open transaction. This may lead to
400 calling prune_icache with consequent lustre reentrance.
404 Description: send correctly lsm on open replay
405 Details : MDS is trust to LSM size on replay open, but client can set wrong size
410 Description: enable client interface failover
411 Details : When a child reconnects from another NID, properly update export
412 nid hash position and ldlm reverse import.
414 Severity : enhancement
416 Description: Communicate OST degraded/readonly state via statfs to MDS
417 Details : Flags in the statfs returned from OSTs indicate whether the
418 OST is in a degraded RAID state, or if the filesystem has
419 turned read-only after a filesystem error is detected.
424 Description: don't panic if EPROTO was hit when reading symlink
425 Details : correctly handling request reference in error cases.
427 Severity : enhancement
429 Description: Add LustreNetLink, a kernel-userspace communcation path.
431 Severity : enhancement
433 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
438 Description: access to llog context before init.
439 Details : move handling CATALOGS file at osc layer and forbid access to llog
443 Frequency : rare, only when client is writing to OST on local node
445 Description: Avoid deadlock for local client writes
446 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
447 memory freeing context. This allows OST threads to set the
448 PF_MEMALLOC flag on task structures in order to allocate memory
449 from reserved pools and complete IO.
450 Use GFP_HIGHUSER for OST allocations for non-local client writes,
451 so that the OST threads generate memory pressure and allow
452 inactive pages to be reclaimed.
454 Severity : enhancement
456 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
461 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
462 Details : move ldlm namespace creation in setup phase to avoid grab
463 _lprocfs_lock with cli_sem held.
467 Description: Temporarily disable grant shrink.
468 Details : Disable the feature for debugging.
471 Frequency : only when formatting test filesystems
473 Description: Unable to run several mkfs.lustre on loop devices at the same time
474 Details : mkfs.lustre returns error 256 on the concurrent loop devices
475 formatting. The solution is to proper handle the error.
477 Severity : enhancement
479 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
481 Severity : enhancement
483 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
488 Description: resolve race between obd_disconnect and class_disconnect_exports
489 Details : if obd_disconnect will be called to already disconnected export he
490 forget release one reference and osc module can't unloaded.
492 Severity : enhancement
494 Description: Allow tuning service thread via /proc
495 Details : For each service a new
496 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
497 created that can be used to set min/max thread counts, and get the
498 current number of running threads.
503 Descriptoin: don't return error if have partially created objects for file.
504 Details : lov_update_create_set uses set->set_success as index for created
505 objects, so if some requests will be failed, they will have hole
506 at end of array and qos_shrink_lsm for allocate correct lsm.
508 Severity : enhancement
510 Description: Update OFED support to 1.4.1
515 Description: Reduce small size read RPC
516 Details : Set read-ahead limit for every file and only do read-ahead when
517 available read-ahead pages are bigger than 1M to avoid small size
520 Severity : enhancement
522 Description: move AT tunable parameters for more consistent usage
523 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
525 Severity : enhancement
527 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
528 Details : allow skip disconnected ost for send statfs request and hide error
534 Description: fix racy locking of mballoc block bitmaps causing BUG
535 Details : The locking of the mballoc buddy bitmap and the in-memory
536 block bitmap was using two different spin locks in some
537 cases. This made it possible to incorrectly access the
538 mballoc bitmap while another process was modifying it,
539 causing a sanity assertion to fail. While no on-disk corruption
540 was reported, there was some risk of this happening.
543 Frequency : rare, on llog test 6
545 Descriptoin: don't allow connect to already connected import
546 Details : allowing connect to already connected import is hide connecting problem.
551 Descriptoin: allow kill process which wait statahead result
552 Details : for some reasons 'ls' can stick in waiting result from statahead,
553 in this case need way for kill this process.
555 Severity : enhancement
557 Description: Add state history info file, enhance import info file
558 Details : Track import connection state changes in a new osc/mdc proc file;
559 add overview-type data to the osc/mdc import proc file.
561 Severity : enhancement
563 Description: MDS create should not wait for statfs RPC while holding DLM lock.
565 Severity : enhancement
567 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
572 Descriptoin: OST grant too much space to client even there are not enough space.
573 Details : Client will shrink its grant cache to OST if there are no write
574 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
575 this grant cache if there are already not enough avaible space
576 (left_space < total_clients * 32M).
579 Frequency : start MDS on uncleanly shutdowned MDS device
581 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
582 Details : stay in waiting mds<>ost recovery finished produce random bugs
583 due race between two ll_sync thread for one lov target. send
584 ACTIVATE event only if connect realy finished and import have
588 Frequency : rare, connect and disconnect target at same time
590 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
591 Details : don't call obd_disconnect under lov_lock. this long time
592 operation and can block ptlrpcd which answer to connect request.
597 Descriptoin: don't lose wakeup for imp_recovery_waitq
598 Details : recover_import_no_retry or invalidate_import and import_close can
599 both sleep on imp_recovery_waitq, but we was send only one wakeup
603 Frequency : always with long access acl
605 Descriptoin: mds can't pack reply with long acl.
606 Details : mds don't control size of acl but they limited by reint/getattr
609 Severity : enhancement
611 Description: Update to SLES10 kernel-2.6.16.60-0.33.
613 Severity : enhancement
615 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
618 Frequency : start MDS on uncleanly shutdowned MDS device
620 Descriptoin: aborting recovery hang on MDS
621 Details : don't throttle destroy RPCs for the MDT.
624 Frequency : on remount
626 Description: external journal device not working after the remount
627 Details : clear dev_rdonly flag for external journal devices in
633 Description: shutdown vs evict race
634 Details : client_disconnect_export vs connect request race.
635 if client will evicted at this time - we start invalidate
636 thread without referece to import and import can be freed
640 Frequency : rare, need acl's on inode.
642 Description: client can't handle ost additional correctly
643 Details : if ost was added after client connected to mds client can have
644 hit lnet_try_match_md ... to big messages to wide striped files.
645 in this case need teach client to handle config events about add
646 lov target and update client max ea size at that event.
648 Severity : enhancement
650 Description: Changelogs
651 Details : Changelogs are a lightweight mechanism to track filesystem
652 metadata and namespace changes. The changelog is recorded
653 permanently on the MDTs, and is periodically "consumed" / purged
654 when records are no longer needed.
656 Severity : enhancement
658 Description: compact fld format with extents
659 Details : Store range of seq rather than every seq in FLD. Seq
660 controller update FLD rather than clients. In Case of CMD, mdt0
661 has FLD, all other metadata server act as non persistent proxy
662 for FLD queries and cache fld entries in fld cache.
667 Description: don't skip ost target if they assigned to file
668 Details : Drop slow OSCs if we can, but not for requested start idx.
669 This means "if OSC is slow and it is not the requested
670 start OST, then it can be skipped, otherwise skip it only
671 if it is inactive/recovering/out-of-space.
675 Description: more cleanup in mds_lov
676 Details : not send LOV EA under replay, we can't know about they size at this
677 time. Don't allow client connect to mds before any ost connected,
678 for avoid problems with LOV EA size and returning EIO to client.
680 Severity : enhancement
682 Description: Interoperability at server side (Disk interoperability)
684 Severity : enhancement
686 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
688 Severity : enhancement
690 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
692 Severity : enhancement
694 Description: New client IO stack (CLIO).
696 Severity : enhancement
698 Description: Commit on sharing. Eliminate inter-client dependencies between
699 uncommitted transactions by doing transaction commits.
700 Thereby clients may recovery independently.
703 Frequency : Create a symlink file with a very long name
705 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
706 Details : If there is no extra space in the request for early cancels,
707 ldlm_req_handles_avail() returns 0 instead of a negative value.
709 Severity : enhancement
711 Description: Add /proc entry for import status
712 Details : The mdc, osc, and mgc import directories now have
713 an import directory that contains useful import data for debugging
716 Severity : enhancement
718 Description: Re-disable certain /proc logging
719 Details : Enable and disable client's offset_stats, extents_stats and
720 extents_stats_per_process stats logging on the fly.
723 Frequency : Only on FC kernels 2.6.22+
725 Description: oops in statahead
726 Details : Do not drop reference count for the dentry from VFS when lookup,
727 VFS will do that by itself.
729 Severity : enhancement
731 Description: Generic /proc file permissions
732 Details : Set /Proc file permissions in a more generic way to enable non-
733 root users operate on some /proc files.
737 Description: Hitting mdc_commit_close() ASSERTION
738 Details : Properly handle request reference release in
739 ll_release_openhandle().
743 Description: quota recovery deadlock during mds failover
744 Details : This patch includes att18982, att18236, att18237 in bz14840.
746 1. fix osts hang when mds does failover with quotaon
747 2. prevent watchdog storm when osts threads wait for the
752 Frequency : only patchless client
753 Description: add workaround for race between add/remove dentry from hash
755 Severity : enhancement
757 Description: Allow OST glimpses to return PW locks
761 Description: LBUG when llog conf file is full
762 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
766 Description: Prevent import from entering FULL state when server in recovery
770 Description: service mount cannot take device name with ":"
771 Details : Only when device name contains ":/" will mount treat it as
777 Description: replace ptlrpcd with the statahead thread to interpret the async
778 statahead RPC callback
782 Frequency : on recovery
783 Description: I/O failures after umount during fail back
784 Details : if client reconnected to restarted server we need join to recovery
785 instead of find server handler is changed and process self
786 eviction with cancel all locks.
788 Severity : enhancement
790 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
792 Severity : enhancement
794 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
796 Severity : enhancement
798 Description: Upcall on Lustre log has been dumped
799 Details : Allow for a user mode script to be called once a Lustre log has
800 been dumped. It passes the filename of the dumped log to the
801 script, the location of the script can be specified via
802 /proc/sys/lnet/debug_log_upcall.
807 Description: avoid idr_remove called for id which is not allocated.
808 Details : Move assigment s_dev for clustered nfs to end of initialization,
809 to avoid problem with error handling.
814 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
815 Details : When connection is reused this not moved from CONN_UNUSED_HASH
816 into CONN_USED_HASH and this prodice warning when put connection
817 again in unused hash.
820 Severity : enhancement
822 Description: File striping can now be set to use an arbitrary pool of OSTs.
824 Severity : enhancement
826 Description: Export bytes_read/bytes_write count on OSC/OST.
830 Description: Early reply size mismatch, MGC loses connection
831 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
832 the connect flags are properly negotiated.
837 Description: quotas are not honored with O_DIRECT
838 Details : all writes with the flag O_DIRECT will use grants which leads to
839 this problem. Now using OBD_BRW_SYNC to guard this.
843 Description: add quota statistics
844 Details : 1. sort out quota proc entries and proc code.
845 2. add quota statistics
847 Severity : enhancement
849 Description: enable quota support for HEAD.
853 Description: Properly propagate oinfo flags from lov to osc for statfs
854 Details : restore missing copy oi_flags to lov requests.
856 Severity : enhancement
858 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
859 : l_getgroups(8), lst(8), routerstat(8)
860 Details : included man pages for llobdstat(8), llstat(8),
861 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
863 Severity : enhancement
865 Description: configure's --enable-quota should check the
866 : kernel .config for CONFIG_QUOTA
867 Details : configure is terminated if --enable-quota is passed but
868 : no quota support is in kernel
872 Description: Remove portals compatibility
873 Details : Remove portals compatibility, not interoperable with releases
878 Description: Resolve device initialization race
879 Details : Prevent proc handler from accessing devices added to the
880 obd_devs array but yet be intialized.
882 Severity : enhancement
884 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
886 Severity : enhancement
888 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
893 Description: Using wrong pointer in osc_brw_prep_request
894 Details : Access to array[-1] can produce panic if kernel compiled with
895 CONFIG_PAGE_ALLOC enabled
899 Description: Client runs out of low memory
900 Details : Consider only lowmem when counting initial number of llap pages
904 Description: *optional* service tags registration
905 Details : if the "service tags" package is installed on a Lustre node
906 When the filesystem is mounted, a local-node service tag will
907 be created. See http://inventory.sun.com/ for more information
908 about the Service Tags asset management system.
912 Description: Kernel BUG tries to release flock
913 Details : Lustre does not destroy flock lock before last reference goes
914 away. So always drop flock locks when client is evicted and
915 perform unlock regardless of successfulness of speaking to MDS.
919 Description: add refcount protection for osc callbacks, avoid panic on shutdown
923 Description: sanity test 65a fails if stripecount of -1 is set
924 Details : handle -1 striping on filesystem in ll_dirstripe_verify
929 Description: ASSERTION(CheckWriteback(page,cmd)) failed
930 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
933 Severity : enhancement
935 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
939 Description: do not process already freed flock
940 Details : flock can possibly be freed by another thread before it reaches
941 to ldlm_flock_completion_ast.
945 Description: LBUG during stress test
946 Details : Need properly lock accesses the flock deadlock detection list.
950 Description: oops in page fault handler
951 Details : kernel page fault handler can return two special 'pages' in error
952 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
956 Description: timeout with invalidate import.
957 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
958 should be handled by ptlrpcd. This produce long age waiting and
959 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
961 Severity : enhancement
963 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
967 Description: enable MGS and MDT services start separately
968 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
969 a co-located MGS without starting the MGS, which is a complement
970 to 'nosvc' mount option.
974 Frequency : after recovery
975 Description: precreate to many object's after del orphan.
976 Details : del orphan st in oscc last_id == next_id and this triger growing
977 count of precreated objects. Set flag LOW to skip increase count
978 of precreated objects.
982 Frequency : rare, on clear nid stats
983 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
984 Details : when clean nid stats sometimes try destroy live entry,
985 and this produce panic in free.
989 Description: Stack overflow during MDS log replay
990 ease stack pressure by using a thread dealing llog_process.
994 Description: wait until IO finished before start new when do lock cancel.
995 Details : VM protocol want old IO finished before start new, in this case
996 need wait until PG_writeback is cleared until check dirty flag and
997 call writepages in lock cancel callback.
999 Severity : enhancement
1001 Description: using special macro for print time and cleanup in includes.
1005 Description: mds_mfd_close() ASSERTION(rc == 0)
1006 Details : In mds_mfd_close(), we need protect inode's writecount change
1007 within its orphan write semaphore to prevent possible races.
1011 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
1015 Frequency : rare, on shutdown ost
1016 Description: don't hit live lock with umount ost.
1017 Details : shrink_dcache_parent can be in long loop with destroy dentries,
1018 use shrink_dcache_sb instead.
1022 Description: don't panic with use echo client
1023 Details : echo client pass NULL as client nid pointer and this produce null
1024 pointer dereference.
1028 Description: fix build on ppc32
1029 Details : compile code with -m64 flag produce wrong object file for ppc32.
1033 Description: add message levels for liblustreapi
1037 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
1038 Details : Move the 'good_osts' check before the 'total_bavail' check. This
1039 will result in an -EAGAIN and in the exit call path we call
1040 alloc_rr() which will with increasing aggressiveness attempt to
1041 aquire precreated objects on the minimum number of required OSCs.
1045 Description: Use old size assignment to avoid deadlock
1046 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
1047 the scheduling while holding a spinlock. We do not need locking
1048 for size in ll_update_inode() because size is only updated from
1049 the MDS for directories or files without objects, so there is no
1050 other place to do the update, and concurrent access to such inodes
1051 are protected by the inode lock.
1055 Description: resolve "_IOWR redefined" build error on SLES10
1059 Description: dump the memory debugging after all modules are unloaded to
1060 suppress false negative in conf_sanity test 39
1062 Severity : enhancement
1064 Description: build kernel-ib packages for OFED 1.3 in our release cycle
1069 Description: fix SLES kernel versioning
1070 Details : the kernel version for our SLES 10 kernel did not include a "-"
1071 before the "smp" at the end. while this was not a problem in
1072 general, it did mean that software trying to use the kernel
1073 version to try to detect a vendor specific kernel would fail.
1074 this was most evident by the OFED build scripts.
1078 Description: Don't update lov_desc members until making sure they are valid
1079 Details : When updating lov_desc members via proc fs, need fix their
1080 validities before doing the real update.
1084 Description: don't put request into delay list while invalidate in flight.
1085 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
1086 import in flight. this produce timeout for invalidate and sometimes
1087 can cause stale data.
1091 Frequency : on ppc only
1092 Description: not convert ost objects for directory because it's not exist.
1093 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
1097 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
1099 Severity : enhancement
1101 Description: Update to support 2.6.22.14 vanilla kernel.
1105 Frequency : rare, on recovery
1106 Description: read procfs can produce deadlock in some situation
1107 Details : Holding lprocfs lock which send rpc can produce block for destroy
1108 obd objects and this also block reconnect with -EALREADY. This
1109 isn't fix all lprocfs bugs - but make it rare.
1111 Severity : enhancement
1113 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
1116 Frequency : frequent on X2 node
1118 Description: mdc_set_open_replay_data LBUG
1119 Details : Set replay data for requests that are eligible for replay.
1123 Description: lustre_mgs: operation 101 on unconnected MGS
1124 Details : When MGC is disconnected from MGS long enough, MGS will evict the
1125 MGC, and late on MGC cannot successfully connect to MGS and a lot
1126 of the error messages complaining that MGS is not connected.
1130 Frequency : on network error
1131 Description: panic with double free request if network error
1132 Details : mdc_finish_enqueue is finish request if any network error ocuring,
1133 but it's true only for synchronus enqueue, for async enqueue
1134 (via ptlrpcd) this incorrect and ptlrpcd want finish request
1137 Severity : enhancement
1139 Description: client-side metadata stat-ahead during readdir(directory readahead)
1140 Details : perform client-side metadata stat-ahead when the client detects
1141 readdir and sequential stat of dir entries therein
1144 Frequency : on start mds
1146 Description: Implement get_info(last_id) in obdfilter.
1149 Frequency : occasional
1151 Description: Correctly check stale fid, not start epoch if ost not support SOM
1152 Details : open with flag O_CREATE need set old fid in op_fid3 because
1153 op_fid2 was overwritten with new generated fid, but mds can answer
1154 with one of these two fids and both is not stale. Setattr
1155 incorrectly started an epoch and assume will be called
1156 done_writing, but without SOM done_writing ever being called.
1159 Frequency : rare, depends on device drivers and load
1161 Description: MDS or OSS nodes crash due to stack overflow
1162 Details : Code changes in 1.8.0 increased the stack usage of some functions.
1163 In some cases, in conjunction with device drivers that use a lot
1164 of stack the MDS (or possibly OSS) service threads could overflow
1165 the stack. One change which was identified to consume additional
1166 stack has been reworked to avoid the extra stack usage.
1169 Frequency : occasional
1171 Description: Do not fail import if osc_interpret_create gets -EAGAIN
1172 Details : If osc_interpret_create got -EAGAIN it immediately exits and
1173 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
1174 oscc_has_objects and see OSC has no objests and call
1175 oscc_internal_create to resend create request.
1177 Severity : enhancement
1179 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
1181 Severity : enhancement
1183 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
1186 Frequency : very rare
1188 Description: Fix replay if there is an un-replied request and open
1189 Details : In some cases, older replay request will revert the
1190 mcd->mcd_last_xid on MDS which is used to record the client's
1191 latest sent request.
1193 Severity : enhancement
1195 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
1197 Severity : enhancement
1199 Description: Add rhel5 support to HEAD.
1201 Serverity : enhancement
1203 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
1208 Description: Don't allow skipping OSTs if index has been specified.
1209 Details : Don't allow skipping OSTs if index has been specified, make
1210 locking in internal create lots better.
1214 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
1215 Details : access to bitfield in structure is always rounded to long
1216 and this produce problem with not atomic change any bit.
1220 Description: Lustre make rpms failed.
1221 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
1222 builds Lustre rpms from tarball.
1225 Frequency : rare on shutdown ost
1227 Description: If llog cancel was not send before clean_exports phase, this can
1228 produce deadlock in llog code.
1229 Details : If llog thread has last reference to obd and call class_import_put
1230 this produce deadlock because llog_cleanup_commit_master wait when
1231 last llog_commit_thread exited, but this never success because was
1232 called from llog_commit_thread.
1236 Description: allow userland application know is lost one of stripes.
1237 Details : fill lvb_blocks with error code on ost and return it to
1238 application if error flag found.
1242 Description: NULL lov_tgts causing MDS oops
1243 Details : more safe checks for NULL lov_tgts for avoid oops.
1245 Severity : enhancement
1247 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
1251 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
1252 Details : If an OST is down the MDS will hang indefinitely in
1253 obd_statfs_rqset() waiting for the statfs data. While for
1254 MDS QOS usage of statfs, it should not stuck in waiting.
1256 Severity : enhancement
1258 Description: remote_acl support
1259 Details : Support ACL-based permission check for remote user.
1260 Support setfacl/getfacl for remote user with the utils
1261 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
1262 the system "{s,g}etfacl" utils.
1264 Severity : enhancement
1266 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
1268 Severity : enhancement
1270 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
1274 Description: make mgs_setparam() handle fsname containing dash
1275 Details : fsname containing a dash does not work with lctl conf_param
1277 Severity : enhancement
1279 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
1281 Severity : enhancement
1283 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
1287 Frequency : rare, at shutdown
1288 Description: access already free / zero obd_namespace.
1289 Details : if client_disconnect_export was called without force flag set,
1290 and exist connect request in flight, this can produce access to
1291 NULL pointer (or already free pointer) when connect_interpret
1292 store ocd flags in obd_namespace.
1296 Frequency : only at startup
1297 Description: not alloc memory with spinlock held.
1298 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
1299 if any spinlock held.
1301 Severity : enhancement
1303 Description: make lustre randomly fail allocating memory
1304 Details : Make lustre randomly failed allocating memory for testing purpose.
1306 Severity : enhancement
1308 Description: lost problems with lov objid file
1309 Details : Fixes some scability and access to not inited memory problems
1310 in work with lov objdid file.
1315 Description: lfs find does not continue on file error
1316 Details : Continue other files processing when a file/dir is absent.
1320 Description: Inconsistent usage of lustre_pack_reply()
1321 Details : Standardize the usage of lustre_pack_reply() such that it
1322 always generate a CERROR on failure.
1325 Frequency : occasional
1327 Description: MDS hang or stay in waiting lock
1328 Details : If client receive lock with CBPENDING flag ldlm need send lock
1329 cancel as separate rpc, to avoid situation when cancel request
1330 can't processed due all i/o threads stay in wait lock.
1334 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1338 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1342 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1346 Description: Update to RHEL4 latest kernel.
1348 Severity : enhancement
1350 Description: Build SLES10 patchless client fails
1351 Details : The configure was broken by run ./configure with
1352 --with-linux-obj=.... argument for patchless client. When the
1353 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1354 can't search header adequately. Use absolute path such as
1355 -I($LINUX)/include instead.
1359 Description: interrupt oig_wait produce painc on resend.
1360 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1361 private set, and this produce situation when rq_ptlrpcd_data not
1362 copyed to new allocated request and triggered LBUG on assert
1363 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1364 ptlrpcd set if request is changed and can be safety changed to use
1369 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1370 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1371 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1372 these macros and replace them with their difinitions instead.
1376 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1377 Details : Update which_patch & target file for SLES10 latest kernel.
1380 Description: Remove client patches from SLES 10 kernel.
1381 Details : This causes SLES 10 clients to behave as patchless clients
1382 even on a Lustre-patched (server) kernel.
1384 Severity : enhancement
1386 Description: self-adjustable client's lru lists
1387 Details : use adaptive algorithm for managing client cached locks lru
1388 lists according to current server load, other client's work
1389 pattern, memory activities, etc. Both, server and client
1390 side namespaces provide number of proc tunables for controlling
1393 Severity : enhancement
1395 Description: light-weight GSS support
1396 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
1397 and reduce performance overhead.
1399 Severity : enhancement
1401 Description: Linux keyring support
1402 Details : Support using service of Linux keyring for Lustre GSS internal
1403 context refresh/cache mechanism.
1407 Description: Fix errors in lfs documentation
1408 Details : Fixes man pages
1412 Description: don't use GFP_* in generic Lustre code.
1413 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
1417 Description: obdclass is limited by single OBD_ALLOC(idarray)
1418 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
1422 Description: warnings with build patchless client with vanila 2.6.19 and up
1423 Details : change old ctl_table style and replace ctl_table/ctl_table_header
1424 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
1428 Frequency : Only for SLES
1429 Description: sanity_quota fail test_1
1430 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
1431 file, which makes TSTID[2] inunique.
1435 Frequency : Only for SLES9
1436 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
1437 Details : Update target/ChangeLog/which_patch .
1441 Frequency : Only for test_10 in sanity-quota.sh
1442 Description: a bug in quota test code
1443 Details : checking a test flag in a wrong place
1447 Frequency : Only for test_13 in sanity-quota.sh
1448 Description: a bug in quota test script
1449 Details : shouldn't check overfull allocated quota
1453 Description: jbd statistics
1454 Details : Port older jbd statistics patch for sles10
1456 Severity : enhancement
1458 Description: Add printing inode info into message about error in writepage.
1463 Description: reply_lock_interpret crash due to race with it and lock cancel.
1464 Details : Do not replay locks that are being cancelled. Do not reference
1465 locks by their address during replay, just by their handle.
1469 Frequency : When flocks are used.
1470 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
1471 Details : Flock locks might destroy just granted lock if it could be merged
1472 with another existing flock, this is done in completion handler,
1473 so teach ldlm_cli_enquque_fini that this is a valid case for
1479 Description: Oops in read and write path when failing to allocate lock.
1480 Details : Check if lock allocation failed and return error back.
1484 Description: lstripe command fails for valid OST index
1485 Details : The stripe offset is compared to lov->desc.ld_tgt_count
1486 instead of lov->desc.ld_active_tgt_count.
1490 Description: sanity.sh failed test 103
1491 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
1496 Description: df doesn't work properly if diskfs blocksize != 4K
1497 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
1501 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
1502 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
1504 Severity : enhancement
1506 Description: Adaptive timeouts
1507 Details : RPC timeouts adapt to changing server load and network
1508 conditions to reduce resend attempts and improve recovery time.
1512 Description: llapi_file_create() does not allow some changes
1513 Details : add llapi_file_open() that allows specifying the mode and
1514 open flags, and also returns an open file handle.
1518 Description: merge and cleanup kernel patches.
1519 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
1523 Description: Add journal checksum support.(Kernel part)
1524 Details : The journal checksum feature adds two new flags i.e
1525 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
1526 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
1527 indicates that the commit block contains the checksum for
1528 the blocks described by the descriptor blocks. Now commit
1529 record can be sent to disk without waiting for descriptor
1530 blocks to be written to disk. This behavior is controlled
1531 using JBD2_FEATURE_ASYNC_COMMIT flag.
1535 Description: OSS needs mutliple precreate threads
1536 Details : Add ability to start more than one create thread per OSS.
1540 Description: Sanity.sh test_65k failed: setstripe should have succeeded
1541 Details : execute lfs setstripe on client
1545 Description: mds_obd_create error creating tmp object
1546 Details : When the user sets quota on root, llog will be affected and can't
1547 create files and write files.
1550 Frequency : Always on ia64 patchless client, and possibly others.
1552 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
1553 Details : This allows the patchless client to be loaded on architectures
1554 without this export.
1558 Description: RedHat Update kernel for RHEL5
1559 Details : Add the kernel config file for RHEL5.
1563 Description: RedHat Update kernel for RHEL5
1564 Details : Modify the kernel config file more closer RHEL5.
1568 Description: Build failure against Centos5 (RHEL5)
1569 Details : Define PAGE_SIZE when it isn't present.
1573 Description: test_7 on sanity-quota.sh can't be used on separate machines
1574 Details : add facet to handle it
1578 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
1579 Details : with NFS, the anon dentry's parent was set to itself in
1580 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
1581 corresponding dentry other than use rec->ur_name.
1583 Severity : enhancement
1585 Description: lfs setstripe enhancement
1586 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
1590 Description: enable data checksumming by default
1591 Details : enable checksum by default, allow --disable-checksum
1592 configure option and "-o nochecksum" mount option. Checksums
1593 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
1597 Description: lustre support for RHEL5
1598 Details : Add support for RHEL5.
1602 Description: Client eviction due to failover config
1603 Details : after a connection loss, the lustre client should attempt to
1604 reconnect to the last active server first before trying the
1605 other potential connections.
1609 Description: when mds and osts use different quota unit(32bit and 64bit),
1610 quota will be released repeatly.
1611 Details : void sending multiple quota reqs to mds, which will keep the
1612 status between the reqs.
1616 Description: osts not allocated evenly to files
1617 Details : change the condition to increase offset_idx
1620 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
1622 Description: Data corruption for OSTs that are formatted larger than 2TB
1624 Details : When generating the bio request for lustre file writes the
1625 sector number would overflow a temporary variable before being
1626 used for the IO. The data reads correctly from Lustre (which
1627 will overflow in a similar manner) but other file data or
1628 filesystem metadata may be corrupted in some cases.
1632 Description: Tune the kernel for good SCSI performance.
1633 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
1634 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
1639 Description: rewrite ext2-derived code in obdclass/uuid.c
1640 Details : rewrite inherited code (uuid parsing code from ext2 utils)
1641 from scratch preserving functionality.
1645 Description: "lfs find -obd UUID" prints directories
1646 Details : "lfs find -obd UUID" will return all directory names instead
1647 of just file names. It is incorrect because the directories
1648 do not reside on the OSTs.
1652 Description: use i_size_read and i_size_write in 2.6 port
1653 Details : replace inode->i_size access with i_size_read/write()
1656 Frequency : when removing large files
1658 Description: scheduling issue during removal of large Lustre files
1659 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
1660 It causes scheduling issues when removing large files (17TB in the
1664 Frequency : only with liblustre clients on XT3
1666 Description: evictions taking too long
1667 Details : allow llrd to evict clients directly on OSTs
1670 Frequency : only on ppc
1672 Description: /proc/fs/lustre/devices broken on ppc
1673 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
1674 We should make sure the type of 'index' is loff_t and then cast
1675 explicitly as needed below. Do not assign an explicitly cast
1679 Frequency : only for rhel5
1681 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
1682 Details : Modify the target file & which_kernel.
1684 Severity : enhancement
1686 Description: omit set fsid for export NFS
1687 Details : fix set/restore device id for avoid EMFILE error and mark lustre
1688 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
1692 Frequency : Always, for kernels after 2.6.16
1693 Description: Fix warning idr_remove called for id=.. which is not allocated.
1694 Details : Last kernels save old s_dev before kill super and not allow
1695 to restore from callback - restore it before call kill_anon_super.
1699 Description: buffer overruns could theoretically occur
1700 Details : llapi_semantic_traverse() modifies the "path" argument by
1701 appending values to the end of the origin string, and a
1702 overrun may occur. Adding buffer overrun check in liblustreapi.
1706 Description: Fix error on 'ls .' at the top of the Lustre mount.
1707 Details : Don't revalidate dentry if it is a root dentry.
1711 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
1712 Details : Modify vm-tunables-rhel4.patch.
1716 Description: Kernel config for 2.6.18-vanilla.
1717 Details : Modify targets/2.6-vanilla.target.in.
1718 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
1719 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
1720 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
1721 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
1725 Description: improve handling recoverable errors
1726 Details : if request processig with error which can be recoverable on server
1727 request should be resend, otherwise page released from cache and
1732 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
1733 Details : Modify target file & which_patch.
1734 A flaw was found in the IA32 system call emulation provided
1735 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
1736 value could be stored in the %RAX register, which could trigger an
1737 out-of-bounds system call table access. An untrusted local user
1738 could exploit this flaw to run code in the kernel
1739 (ie a root privilege escalation). (CVE-2007-4573).
1743 Description: O_DIRECT bypasses client statistics.
1744 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
1745 read_ahead_stats were updated. copied the stats section from
1746 osc_send_oap_rpc() into async_internal().
1750 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
1754 Description: change order of libsysio includes
1755 Details : '#include sysio.h' should always come before '#include xtio.h'
1759 Description: handle "serious error: objid * already exists" more gracefully
1760 Details : If LAST_ID value on disk is smaller than the objects existing in
1761 the O/0/d* directories, it indicates disk corruption and causes an
1762 LBUG(). If the object is 0-length, then we should use the existing
1763 object. This will help to avoid a full fsck in most cases.
1765 Severity : enhancement
1767 Description: adapt the lustre_config script to support the upgrade case
1768 Details : Add "-u" option for lustre_config script to support upgrading 1.4
1769 server targets to 1.6 in parallel.
1773 Description: To avoid grant space > avaible space when the disk is almost
1774 full. Without this patch you might see the error "grant XXXX >
1775 available" or some LBUG about grant, when the disk is almost
1777 Details : In filter_check_grant, for non_grant cache write, we should
1778 check the left space by if (*left > ungranted + bytes), instead
1779 of (*left > ungranted), because only we are sure the left space
1780 is enough for another "bytes", then the ungrant space should be
1781 increase. In client, we should update cl_avail_grant only there
1782 is OBD_MD_FLGRANT in the reply.
1786 Description: Update RHEL 4 kernel to fix local root privilege escalation.
1787 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
1788 described in CVE-2007-4573. This problem could allow untrusted
1789 local users to gain root access.
1792 Frequency : when using O_DIRECT and quotas
1794 Description: Incorrect file ownership on O_DIRECT output files
1795 Details : block usage reported by 'lfs quota' does not take into account
1796 files that have been written with O_DIRECT.
1801 Description: touch file failed when fs is not full
1802 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
1803 otherwise we can get ENOSP while fs is not full.
1807 Description: parallel lock callbacks
1808 Details : Instead of sending blocking and completion callbacks as separated
1809 requests, adding them to a set and sending in parallel.
1812 Frequency : only for Cray XT3
1813 Bugzilla : 12829/13455
1814 Description: Changing primary group doesn't change the group lustre assigns to
1816 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
1817 overridden with the primary group provided by the group upcall,
1818 whereas the supplied fsgid can be trusted if it is in the list of
1819 supplementary groups returned by the group upcall.
1821 Severity : enhancement
1823 Description: Allow masking D_WARNING, D_ERROR messages from console
1824 Details : Console messages can now be disabled via lnet.printk.
1828 Description: User code with malformed file open parameter crashes client node
1829 Details : Before packing join_file req, all the related reference should be
1830 checked carefully in case some malformed flags cause fake
1831 join_file req on client.
1835 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
1836 Details : ldlm_completion_ast() assumes that a lock is granted when the req
1837 mode is equal to the granted mode. However, it should also check
1838 that LDLM_FL_CP_REQD is not set.
1842 Description: Heavy nfs access might result in deadlocks
1843 Details : After ELC code landed, it is now improper to enqueue any mds
1844 locks under och_sem, because enqueue might want to decide to
1845 cancel open locks for same inode we are holding och_sem for.
1849 Description: Client eviction while running blogbench
1850 Details : A lot of unlink operations with concurrent I/O can lead to a
1851 deadlock causing evictions. To address the problem, the number of
1852 oustanding OST_DESTROY requests is now throttled to
1853 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
1854 callbacks are processed in priority.
1858 Description: enable ACLs on MDS by default
1859 Details : ACLs must be enabled on MDS by default.
1862 Frequency : PPC/PPC64 only
1864 Description: conflicts between asm-ppc64/types.h and lustre_types.h
1865 Details : fix duplicated definitions between asm-ppc64/types.h and
1866 lustre_types.h on PPC.
1869 Frequency : PPC/PPC64 only
1871 Description: asm-ppc/segment.h does not exist
1872 Details : fix compile issue on PPC.
1876 Description: better handle error messages in extents code
1879 Frequency : RHEL4 only
1881 Description: mkfs is very slow on IA64/RHEL4
1882 Details : A performance regression has been discovered in the MPT Fusion
1883 driver between versions 3.02.73rh and 3.02.99.00rh. As a
1884 consequence, we have downgraded the MPT Fusion driver in the RHEL4
1885 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
1887 Severity : enhancement
1889 Description: SNMP support enhancement
1890 Details : Adding total number of sampled request for an MDS node in snmp
1893 Severity : enhancement
1895 Description: Optimize ldlm waiting list processing for PR extent locks
1896 Details : When processing waiting list for read extent lock and meeting read
1897 lock that is same or wider to it that is not contended, skip
1898 processing rest of the list and immediatelly return current
1899 status of conflictness, since we are guaranteed there are no
1900 conflicting locks in the rest of the list.
1904 Description: Time out and refuse to reconnect
1905 Details : When the failover node is the primary node, it is possible
1906 to have two identical connections in imp_conn_list. We must
1907 compare not conn's pointers but NIDs, otherwise we can defeat
1908 connection throttling.
1912 Description: port llog fixes from b1_6 into HEAD
1913 Details : Port llog reference couting and some llog cleanups from b1_6
1914 (bug 10800) into HEAD, for protect from panic and access to already
1915 free llog structures.
1919 Description: Detect stride IO mode in read-ahead
1920 Details : When a client does stride read, read-ahead should detect that and
1921 read-ahead pages according to the detected stride pattern.
1925 Description: data checksumming impacts single node performance
1926 Details : add support for several checksum algorithm. Currently, only CRC32
1927 and Adler-32 are supported. The checksum type can be changed on
1928 the fly via /proc/fs/lustre/osc/*/checksum_type.
1932 Description: use adler32 for page checksums
1933 Details : when available, use the Adler-32 algorithm instead of CRC32 for
1938 Description: build for x2 fails
1939 Details : fix compile issue on Cray systems.
1943 Description: Properly match for duplicate locks
1944 Details : Due to different lock order from skiplists code, we need to
1945 traverse entire list for now
1948 Frequency : only on PPC/SLES10
1950 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
1951 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
1952 be defined. Add a hack in mkfs_lustre.c to work around this compile
1957 Description: LASSERT on MDS when client holding flock lock dies
1958 Details : ldlm pool logic depends on number of granted locks equal to
1959 number of released locks which is not true for flock locks, so
1960 just exclude such locks from consideration.
1964 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
1965 Details : Use fsfilt_sync() for both the whole filesystem sync and
1966 individual file sync to eliminate dangerous inode locking
1967 with I_LOCK that can lead to a deadlock.
1971 Description: performance in 1.6.3
1972 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
1973 sure that 1MB requests are not fragmented by the block layer.
1975 Severity : enhancement
1977 Description: organize the server-side client stats on per-nid basis
1978 Details : Change the structure of stats under obdfilter and mds to
1986 The "uuid"s file would list the uuids of _active_ exports.
1987 And the clear entry is to clear all stats and stale nids.
1989 Severity : enhancement
1991 Description: eliminate client locks in face of contention
1992 Details : file contention detection and lockless i/o implementation
1993 for contended files.
1997 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
2000 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
2002 Description: MDS may lose file striping (and hence file data) in some cases
2003 Details : If there are additional extended attributes stored on the MDS,
2004 in particular ACLs, SELinux, or user attributes (if user_xattr
2005 is specified for the client mount options) then there is a risk
2006 of attribute loss. Additionally, the Lustre file striping
2007 needs to be larger than default (e.g. striped over all OSTs),
2008 and an additional attribute must be stored initially in the
2009 inode and then increase in size enough to be moved to the
2010 external attribute block (e.g. ACL growing in size) for file
2015 Description: skiplist implementation simplification
2016 Details : skiplists are used to group compatible locks on granted list
2017 that was implemented as tracking first and last lock of each
2018 lock group the patch changes that to using doubly linked lists
2022 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
2023 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
2024 because LASSERT on both the data supplied by a client, and the
2025 data on disk is dangerous and incorrect.
2027 Severity : enhancement
2029 Description: Slow truncate/writes to huge files at high offsets.
2030 Details : Directly associate cached pages to lock that protect those pages,
2031 this allows us to quickly find what pages to write and remove
2032 once lock callback is received.
2036 Description: more ldlm soft lockups
2037 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
2038 starve other threads from the resource lock for a long time in
2039 case of long waiting queue, so change the debug level from
2040 D_OTHER to the less frequently used D_INFO.
2042 Severity : enhancement
2044 Description: add -gid, -group, -uid, -user options to lfs find
2048 Description: Hung threads in invalidate_inode_pages2_range
2049 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
2050 once one is completed. As a result, some RPCs are stuck in the
2051 queue and are never sent.
2055 Description: filter threads hungs on waiting journal commit
2056 Details : Cleanup filter group llog code, then only filter group llog will
2057 be only created in the MDS/OST syncing process.
2061 Description: Procfs and llog threads access destoryed import sometimes.
2062 Details : Sync the import destoryed process with procfs and llog threads by
2063 the import refcount and semaphore.
2065 Severity : enhancement
2067 Description: openlock cache of b1_6 port to HEAD
2072 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
2073 Details : If insertion of an extent fails, then discard the inode
2074 preallocation and free data blocks else it can lead to duplicate
2079 Description: don't always update ctime in ext3_xattr_set_handle()
2080 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
2081 In some cases the ctime should not be updated, for example for
2082 2.0->1.8 compatibility it is necessary to delete an xattr and it
2083 should not update the ctime.
2087 Bugzilla : 15713/16362
2088 Description: Assertion in iopen_connect_dentry in 1.6.3
2089 Details : looking up an inode via iopen with the wrong generation number can
2090 populate the dcache with a disconneced dentry while the inode
2091 number is in the process of being reallocated. This causes an
2092 assertion failure in iopen since the inode's dentry list contains
2093 both a connected and disconnected dentry.
2097 Description: assertion failure in ldlm_handle2lock()
2098 Details : fix a race between class_handle_unhash() and class_handle2object()
2099 introduced in lustre 1.6.5 by bug 13622.
2104 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
2105 Details : remove the limit of 256 scsi disks in the sd_iostat patch
2110 Description: oops in sd_iostats_seq_show()
2111 Details : unloading/reloading the scsi low level driver triggers a kernel
2112 bug when trying to access the sd iostat file.
2117 Description: Kernel panics during QLogic driver reload
2118 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
2119 patch, causing memory corruption.
2124 Description: journal_dev option does not work in b1_6
2125 Details : pass mount option during pre-mount.
2127 Severity : enhancement
2129 Description: Add a FIEMAP(FIle Extent MAP) ioctl
2130 Details : FIEMAP ioctl will allow an application to efficiently fetch the
2131 extent information of a file. It can be used to map logical blocks
2132 in a file to physical blocks in the block device.
2136 Description: LDLM soft lockups - improvement
2137 Details : It is be possible to send the lock handle along with each read
2138 or write request because the client is already doing a lock match
2139 itself so there isn't any reason the OST should have to re-do that
2143 Frequency : only on Cray X2
2145 Description: X2 build failures
2146 Details : fix build failures on Cray X2.
2150 Description: xid & resent requests
2151 Details : Initialize RPC XID from clock at startup (randomly if clock is
2154 Severity : enhancement
2156 Description: Add lustre_start utility to start or stop multiple Lustre servers
2161 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
2162 Details : In case of memory pressure, list_del() can be called twice on
2163 req->rq_history_list, causing a kernel oops.
2167 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
2168 Details : fix stack overflow in the distributed lock manager by defering
2169 export eviction after a failed AST to the elt thread instead of
2170 handling it in the dlm interpret routine.
2174 Description: Convert some comments to new format.
2175 Details : Update documenting comments to match doxygen conventions.
2179 Description: Grammar fixes.
2180 Details : A couple of trivial spelling fixes.
2184 Description: OSD_COUNTERS-mandatory
2185 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
2186 to simplify the code.
2190 Description: simplify cmm_device freeing logic.
2191 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
2195 Description: Add lockdep support to dt_object_operations locking interface.
2196 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
2197 parameter indicating lock ordering. Update mdd code to use new
2202 Description: Introduce failloc constants for lockless IO tests.
2203 Details : Add two new failloc constants to test lockless IO. Only one of
2204 them in implemented---another is checked in yet to be landed
2209 Description: Add lockdep support for inode mutex.
2210 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
2211 in the situations where Lustre has to lock more than one inode
2216 Description: Add optional invariants checking support.
2217 Details : Add new LINVRNT() macro, optional on new --enable-invariants
2218 configure switch. This macro is to be used for consistency and
2219 sanity checks that are too expensive to be left in `production'
2224 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
2225 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
2230 Description: Add lockdep support for ldlm_lock and ldlm_resource.
2231 Details : Use spin_lock_nested() in (the only) situation where more than
2232 one ldlm_lock is locked simultaneously. Also, fix possible
2233 dead-lock in ldlm_lock_change_resource() by enforcing particular
2238 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
2239 Details : Instead of specifying each ldlm_lock call-back through separate
2240 parameter, wrap them into struct ldlm_callback_suite.
2244 Description: Kill join_lru obd method and its callers.
2245 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
2246 in memory---a requirement implemented through ->o_join_lru() obd
2247 method in HEAD. Remove this method and its users.
2251 Description: Add asynchronous ldlm ENQUEUE completion handler.
2252 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
2253 case that stock ldlm_completion_ast() cannot handle as it waits
2254 until lock is granted. Introduce new ldlm_completion_ast_async()
2255 for this. Also comment ldlm_completion_ast().
2259 Description: ldlm_error <-> errno conversion.
2260 Details : Add functions to map (rather arbitrary) between LDLM error codes
2261 and standard errno values. CLIO needs this to prevent LDLM specific
2262 constants from escaping ldlm and osc.
2266 Description: Kill unused ldlm_handle2lock_ns() function.
2267 Details : Kill unused ldlm_handle2lock_ns() function.
2271 Description: Add lu_ref support to ldlm_lock
2272 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
2273 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
2274 and ldlm_resource. LDLM interface has to be changed, because code that
2275 releases a reference on a lock, has to "know" what reference this is.
2276 In the most frequent case
2278 lock = ldlm_handle2lock(handle);
2280 LDLM_LOCK_PUT(lock);
2282 no changes are required. When any other reference (received _not_ from
2283 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
2284 instead of LDLM_LOCK_PUT().
2286 Arguably, changes are pervasive, and interface requires some discipline
2287 for proper use. On the other hand, it was very instrumental in finding
2288 a few leaked lock references.
2292 Description: Add ldlm_lock_addref_try().
2293 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
2294 attempts to addref a lock that might be being canceled
2299 Description: Add ldlm_weigh_callback().
2300 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
2301 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
2302 hard-coded `number of pages' logic.
2306 Description: Add lockdep annotations to llog code.
2307 Details : Use appropriately tagged _nested() locking calls in the places
2308 where llog takes more than one ->lgh_lock lock.
2312 Description: Add loi_kms_set().
2313 Details : Wrap kms updates into a helper function.
2317 Description: Constify instances of struct lsm_operations.
2318 Details : Constify instances of struct lsm_operations.
2322 Description: lu_conf support.
2323 Details : On a server, a file system object is uniquely identified
2324 by a fid, which is sufficient to locate and load all object
2325 state (inode). On a client, on the other hand, more data are
2326 necessary instantiate an object. Change lu_object_find() and
2327 friends to take additional `lu_conf' argument describing object.
2328 Typically this includes layout information.
2332 Description: lu_context fixes.
2333 Details : Introduce new lu_context functions that are needed on the client
2334 side, where some system threads (ptlrpcd) are shared by multiple
2335 modules, and so cannot be stopped during module shutdown.
2339 Description: Add start and stop methods to lu_device_type_operations.
2340 Details : Introduce two new methods in lu_device_type_operations, that are
2341 invoked when first instance of a given type is created and last one
2342 is destroyed respectively. This is need by CLIO.
2346 Description: Add lu_ref support to struct lu_device.
2347 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
2348 track leaked references.
2352 Description: Introduce lu_kmem_descr.
2353 Details : lu_kmem_descr and its companion interface allow to create
2354 and destroy a number of kmem caches at once.
2358 Description: Fix lu_object finalization race.
2359 Details : Fix a race between lu_object_find() finding an object and its
2360 concurrent finalization. This race is (most likely) not possible
2361 on the server, but might happen on the client.
2365 Description: Introduce lu_ref interface.
2366 Details : lu_ref is a debugging module allowing to track references to
2367 a given object. It is quite cpu expensive, and has to be
2368 explicitly enabled with --enable-lu_ref. See usage description
2373 Description: Factor lu_site procfs stats into a separate function.
2374 Details : Separate lu_site stats printing code into a separate function
2375 that can be reused on a client.
2379 Description: Constify instances of struct {lu,dt,md}_device_operations.
2380 Details : Constify instances of struct {lu,dt,md}_device_operations.
2384 Description: Introduce struct md_site and move meta-data specific parts of
2385 struct lu_site here.
2386 Details : Move md-specific fields out of struct lu_site into special struct
2387 md_site, so that lu_site can be used on a client.
2391 Description: Kill mdd_lov_destroy().
2392 Details : Remove unused mdd code.
2396 Description: Add st_block checking to multistat.c.
2397 Details : Add st_block checking to multistat.c.
2401 Description: Add lu_ref support to struct obd_device.
2402 Details : Add lu_ref tracking to obd_device.
2406 Description: Kill obd_set_fail_loc().
2407 Details : Remove unused code.
2411 Description: Add special type for ptlrpc_request interpret functions.
2412 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
2413 there. Actual usage will be in CLIO.
2417 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
2418 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
2419 doesn't work with lockdep.
2423 Description: Add rwv.c test program.
2424 Details : New testing program exercising readv(2) and writev(2) (Qian).
2428 Description: Add sendfile.c test program.
2429 Details : New testing program exercising sendfile(2) (Jay).
2433 Description: Ratelimit a message that can be very frequent.
2434 Details : Ratelimit a memory allocation failure message that can
2439 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2440 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2444 Description: Liblustre build fixes.
2445 Details : Liblustre build fixes.
2449 Description: libcfs: add cfs_{need,cond}_resched() interface.
2450 Details : libcfs: add cfs_{need,cond}_resched() definition and
2451 implementations for Linux, NT, and liblustre.
2453 Severity : enhancement
2455 Description: More exported tunables for mballoc
2456 Details : Add support for tunable preallocation window and new tunables for
2457 large/small requests
2461 Description: Detect corruption of block bitmap and checking for preallocations
2462 Details : Checks validity of on-disk block bitmap. Also it does better
2463 checking of number of applied preallocations. When corruption is
2464 found, it turns filesystem readonly to prevent further corruptions.
2468 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
2469 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
2470 window to 0. If it does hit cache miss, and read-pattern is still
2471 stride-io mode, does not reset the stride window, but also does not
2472 increase the stride window length in this case.
2476 Frequency : only for big-endian servers
2477 Description: Check if system is big-endian while mounting fs with extents feature
2478 Details : Mounting a filesystem with extents feature will fail on big-endian
2479 systems since ext3-based ldiskfs is not supported on big-endian
2480 systems. This can be over-riden with "bigendian_extents" mount option.
2482 Severity : enhancement
2484 Description: The root squash functionality
2485 Details : A security feature, which is to prevent users from being able
2486 to mount lustre on their desktop, run as root, and delete
2487 all of the files in the filesystem. The goal is accomplished by
2488 remapping user id (UID) and group id (GID) of the root user to
2489 a UID and GID specified by the system administartor via Lustre
2490 configuration management server (MGS). The functionality also
2491 allows to specify sets of clients for which the remapping does
2496 Description: Excessive recovery window
2497 Details : With AT enabled, the recovery window can be excessively long (6000+
2498 seconds). To address this problem, we no longer use
2499 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
2500 timeout no longer depends on the service time, it is set to
2501 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
2502 time via pb_service_time.
2506 Description: Watchdog triggered on MDS failover
2507 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
2508 the OSTs know that the MDS "UUID" can be reused for the same export
2509 from a different NID, so we do not need to wait for the export to be
2513 Frequency : rare, only if using MMP with Linux RAID
2515 Description: MMP doesn't work with Linux RAID
2516 Details : While using HA for Lustre servers with Linux RAID, it is possible
2517 that MMP will not detect multiple mounts. To make this work we
2518 need to unplug the device queue in RAID when the MMP block is being
2519 written. Also while reading the MMP block, we should read it from
2520 disk and not the cached one.
2522 Severity : enhancement
2524 Description: open file using fid
2525 Details : A file can be opened using just its fid, like
2526 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
2529 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
2532 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
2533 Details : add proper locking for creation of jbd_4k slab cache
2537 Description: MMP check in ext3_remount() fails without displaying any error
2538 Details : When multiple mount protection fails during remount, proper error
2541 Severity : enhancement
2543 Description: Allow stripe size to be up to 4G-64k
2544 Details : Fix math logic to allow large stripe sizes.
2548 Description: add check for >8TB ldiskfs filesystems
2549 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
2550 Don't allow >8TB ldiskfs filesystems to be mounted without
2551 force_over_8tb mount option
2556 Descriptoin: OSS complains "too many blocks" and crashes
2557 Details : In filter_commitrw_write():
2558 1. lquota_chkquota();
2559 2. obdo_from_inode();
2560 3. lquota_pending_commit();
2561 "2." has changed uid and gid in this case, then caused
2562 lquota_chkquota() and lquota_pending_commit() to use
2563 different uid and gid. That is the root of the bug.
2567 Descriptoin: getstripe doesn't report stripe count = -1 sanely
2568 Details : Detect and print out the 'stripe count = -1' case properly.
2572 Description: Out or order replies might be lost on replay
2573 Details : In ptlrpc_retain_replayable_request if we cannot find retained
2574 request with tid smaller then one currently being added, add it
2575 to the start, not end of the list.
2577 --------------------------------------------------------------------------------
2579 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
2581 * Support for kernels:
2582 2.6.5-7.283 (SLES 9),
2583 2.6.9-55.EL (RHEL 4),
2584 2.6.16.46-0.14 (SLES 10),
2585 2.6.18.8 vanilla (kernel.org)
2586 * Client support for unpatched kernels:
2587 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2588 2.6.16 - 2.6.22 vanilla (kernel.org)
2589 * Due to recently discovered recovery problems, we do not recommend
2590 using patchless RHEL 4 clients with this or any earlier release.
2591 * Recommended e2fsprogs version: 1.39.cfs8
2592 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2593 * Starting with this release, the ldiskfs backing filesystem required
2594 by Lustre is now in its own package, lustre-ldiskfs. This package
2595 should be installed. It is versioned separately from Lustre and
2596 may be released separately in future.
2600 Description: block reactivating mgc import until all deactivates complete
2601 Details : Fix race when failing back MDT/MGS to itself (testing)
2604 Frequency : at statup only
2606 Description: mds_lov_synchronize race leads to various problems
2607 Details : simultaneous MDT->OST connections at startup can cause the
2608 sync to abort, leaving the OSC in a bad state.
2610 Severity : enhancement
2612 Description: add optional extra BUILD_VERSION info
2613 Details : add a new environment variable (namely LUSTRE_VERS) which allows
2614 to override the lustre version.
2617 Frequency : 2.6.18 servers only
2619 Description: ll_kern_mount() doesn't release the module reference
2620 Details : The ldiskfs module reference count never drops down to 0
2621 because ll_kern_mount() doesn't release the module reference.
2626 Description: server LBUG when using old ost_num_threads parameter
2627 Details : Accept the old ost_num_threads parameter but warn that it
2628 is deprecated, and fix an off-by-one error that caused an LBUG.
2633 Description: Transient SCSI error results in persistent IO issue
2634 Details : iobuf->dr_error is not reinitialized to 0 between two
2638 Frequency : sometimes when underlying device returns I/O errors
2640 Description: OSTs not going read-only during write failures
2641 Details : OSTs are not remounted read-only when the journal commit threads
2642 get I/O errors because fsfilt_ext3 calls journal_start/stop()
2643 instead of the ext3 wrappers.
2647 Description: poor connect scaling with increasing client count
2648 Details : Don't run filter_grant_sanity_check for more than 100 exports
2649 to improve scaling for large numbers of clients.
2652 Frequency : SLES10 only
2654 Description: sanity-quota.sh quotacheck failed: rc = -22
2655 Details : Quotas cannot be enabled on SLES10.
2658 Frequency : liblustre clients only
2660 Description: getdirentries does not give error when run on compute nodes
2661 Details : getdirentries does not fail when the size specified as an argument
2662 is too small to contain at least one entry
2664 Severity : enhancement
2666 Description: Add LNET router traceability for debug purposes
2667 Details : If a checksum failure occurs with a router as part of the
2668 IO path, the NID of the last router that forwarded the bulk data
2669 is printed so it can be identified.
2674 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
2675 Details : Due to a race condition, liblustre clients were occasionally
2676 evicted incorrectly.
2678 Severity : enhancement
2680 Description: lfs setstripe use optional parameters instead of postional
2683 Severity : enhancement
2685 Description: Nanosecond timestamp support for ldiskfs
2686 Details : The on-disk ldiskfs filesystem has added support for nanosecond
2687 resolution timestamps. There is not yet support for this at
2688 the Lustre filesystem level.
2691 Frequency : during server recovery
2693 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
2694 Details : request with rq_no_resend flag not awake l_wait_event if they get
2698 Frequency : nfs export on patchless client
2700 Description: connectathon hang when test nfs export over patchless client
2701 Details : Disconnected dentry cannot be found with lookup, so we do not need
2702 to unhash it or make it invalid
2705 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
2706 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
2707 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
2712 Description: open req refcounting wrong on reconnect
2713 Details : If reconnect happened between getting open reply from server and
2714 call to mdc_set_replay_data in ll_file_open, we will schedule
2715 replay for unreferenced request that we are about to free.
2716 Subsequent close will crash in variety of ways.
2717 Check that request is still eligible for replay in
2718 mdc_set_replay_data().
2723 Description: disable writes to filesystem when reading health_check file
2724 Details : the default for reading the health_check proc file has changed
2725 to NOT do a journal transaction and write to disk, because this
2726 can cause reads of the /proc file to hang and block HA state
2727 checking on a healthy but otherwise heavily loaded system. It
2728 is possible to return to the previous behaviour during configure
2729 with --enable-health-write.
2731 Severity : enhancement
2733 Description: 64-bit inode version
2734 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
2735 to the inode. This will be required for version-based recovery.
2740 Description: MDS fails to start if a duplicate client export is detected
2741 Details : in some rare cases it was possible for a client to connect to
2742 an MDS multiple times. Upon recovery the MDS would detect this
2743 and fail during startup. Handle this more gracefully.
2745 Severity : enhancement
2747 Description: Add -o localflock option to simulate old noflock
2749 Details : This will achieve local-only flock/fcntl locks
2755 Description: log_commit_thread vs filter_destroy race leads to crash
2756 Details : Take import reference before releasing llog record semaphore
2761 Description: Wrong request locking in request set processing
2762 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
2763 imp_delayed_list, in this place should be used imp_lock.
2766 Frequency : when reconnection
2768 Description: Grant Leak when osc reconnect to OST
2769 Details : When osc reconnect ost, OST(filter) should check whether it
2770 should grant more space to client by comparing fed_grant and
2771 cl_avail_grant, and return the granted space to client instead
2772 of "new granted" space, because client will call osc_init_grant
2773 to update the client grant space info.
2776 Frequency : when client reconnect to OST
2778 Description: Grant Leak when osc do resend and replay bulk write
2779 Details : When osc reconnect to OST, OST(filter)should clear grant info of
2780 bulk write request, because the grant info while be sync between
2781 OSC and OST when reconnect, and we should ignore the grant info
2782 these of resend/replay write req.
2787 Description: Grant space more than avaiable left space sometimes.
2788 Details : When then OST is about to be full, if two bulk writing from
2789 different clients came to OST. Accord the avaliable space of the
2790 OST, the first req should be permitted, and the second one
2791 should be denied by ENOSPC. But if the seconde arrived before
2792 the first one is commited. The OST might wrongly permit second
2793 writing, which will cause grant space > avaiable space.
2796 Frequency : when client is evicted
2798 Description: Grant might be wrongly erased when osc is evicted by OST
2799 Details : when the import is evicted by server, it will fork another
2800 thread ptlrpc_invalidate_import_thread to invalidate the
2801 import, where the grant will be set to 0. While the original
2802 thread will update the grant it got when connecting. So if
2803 the former happened latter, the grant will be wrongly errased
2804 because of this race.
2809 Description: Checking Stale with correct fid
2810 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
2811 to check whether it is stale, because sometimes, we want the
2812 enqueue happened anyway, and op_data.fid2 will not be initialized.
2814 Severity : enhancement
2816 Description: update patchless client
2817 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
2820 Frequency : only with 2.4 kernel
2822 Description: random memory corruption
2823 Details : size of struct ll_inode_info is to big for union inode.u and this
2824 can be cause of random memory corruption.
2829 Description: Memory leak in recovery
2830 Details : Lov_mds_md was not free in an error handler in mds_create_object.
2831 It should also check obd_fail before fsfilt_start, otherwise if
2832 fsfilt_start return -EROFS,(failover mds during mds recovery).
2833 then the req will return with repmsg->transno = 0 and rc = EROFS.
2834 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
2835 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
2836 be freed no matter whether fsfilt_commit success or not.
2839 Frequency : only with huge count clients
2841 Description: Prevents from taking the superblock lock in llap_from_page for
2843 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
2844 for llap_from_page call in ll_removepage prevents from taking the
2845 superblock lock for a soon died page.
2850 Description: Not check open intent error before release open handle
2851 Details : in some rare cases, the open intent error is not checked before
2852 release open handle, which may cause
2853 ASSERTION(open_req->rq_transno != 0), because it tries to release
2854 the failed open handle.
2859 Description: Set cat log bitmap only after create log success.
2860 Details : in some rare cases, the cat log bitmap is set too early. and it
2861 should be set only after create log success.
2865 Description: Accessing a block bevice can re-enable I/O when Lustre is
2866 tearing down a device.
2867 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
2872 Description: service threads may hog cpus when there are a lot of requests
2873 Details : Insert cond_resched to give other threads a chance to use some CPU
2878 Description: the cat log was not initialized in recovery
2879 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
2880 unlink log on mds will not be initialized until mds post
2881 recovery. And also in mds post recovery, the unlink log will
2882 initialization will be done asynchronausly, so there will be race
2883 between add unlink log and unlink log initialization.
2887 Description: brw_stats were being printed incorrectly
2888 Details : brw_stats were being printed as log2 but all of them were not
2889 recorded as log2. Also remove some code duplication arising from
2890 filter_tally_{read,write}.
2894 Frequency : rare, only in recovery.
2895 Description: ASSERTION(req->rq_type != LI_POISON) failed
2896 Details : imp_lock should be held while iterating over imp_sending_list for
2897 prevent destroy request after get timeout in ptlrpc_queue_wait.
2901 Description: replay-single.sh test 52 fails
2902 Details : A lock's skiplist need to be cleanup when it being unlinked
2903 from its resource list.
2907 Description: Short directio read returns full requested size rather than
2909 Details : Direct I/O operations should return actual amount of bytes
2910 transferred rather than requested size.
2912 Severity : enhancement
2914 Description: metadata RPC reduction (e.g. for rm performance)
2915 Details : decrease the amount of synchronous RPC between clients and servers
2916 by canceling conflicing lock before the operation on the client
2917 and packing thier handles into the main operation RPC to server.
2919 Severity : enhancement
2921 Description: Async OSC create to avoid the blocking unnecessarily.
2922 Details : If a OST has no remain object, system will block on the creating
2923 when need to create a new object on this OST. Now, ways use
2924 pre-created objects when available, instead of blocking on an
2925 empty osc while others are not empty. If we must block, we block
2926 for the shortest possible period of time.
2930 Description: improve handling recoverable errors
2931 Details : if request processig with error which can be recoverable on server
2932 request should be resend, otherwise page released from cache and
2935 Severity : enhancement
2937 Description: refine locking for avoid write wrong info into lov_objid file
2938 Details : fix possible races with add new target and write/update data in
2941 --------------------------------------------------------------------------------
2943 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
2948 Frequency : on some architectures
2950 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
2951 Details : Uninitialized flags sometimes cause configuration commands to
2955 Frequency : patchless clients only
2957 Description: missing __iget() symbol export
2958 Details : The __iget() symbol export is missing. To avoid the need for
2959 this on patchless clients the deathrow inode reaper is turned
2960 off, and we depend on the VM to clean up old inodes. This
2961 dependency was during via the fix for bug 12181.
2963 --------------------------------------------------------------------------------
2965 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
2967 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
2968 INTEROPERATE with older versions automatically. In many cases a
2969 special upgrade step is needed. Please read the
2970 user documentation before upgrading any part of a live system.
2971 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
2972 version will not interoperate with 1.6 betas before beta5 (1.5.95).
2973 * WARNING: Lustre configuration and startup changes are required with
2974 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
2978 Severity : enhancement
2980 Description: MountConf
2981 Details : Lustre configuration is now managed via mkfs and mount
2982 commands instead of lmc and lconf. New obd types (MGS, MGC)
2983 are added for dynamic configuration management. See
2984 https://mail.clusterfs.com/wikis/lustre/MountConf for
2987 Severity : enhancement
2989 Description: dynamic OST addition
2990 Details : OSTs can now be added to a live filesystem
2992 Severity : enhancement
2994 Description: startup order invariance
2995 Details : MDTs and OSTs can be started in any order. Clients only
2996 require the MDT to complete startup.
2998 Severity : enhancement
3000 Description: parallel, asynchronous orphan cleanup
3001 Details : orphan cleanup is now performed in separate threads for each
3002 OST, allowing parallel non-blocking operation.
3004 Severity : enhancement
3006 Description: optimized stripe assignment
3007 Details : stripe assignments are now made based on ost space available,
3008 ost previous usage, and OSS previous usage, in order to try
3009 to optimize storage space and networking resources.
3011 Severity : enhancement
3013 Description: Permanently set tunables
3014 Details : All writable /proc/fs/lustre tunables can now be permanently
3015 set on a per-server basis, at mkfs time or on a live system.
3017 Severity : enhancement
3019 Description: Lustre message v2
3020 Details : Add lustre message format v2.
3022 Severity : enhancement
3024 Description: client OST exclusion list
3025 Details : Clients can be started with a list of OSTs that should be
3026 declared "inactive" for known non-responsive OSTs.
3030 Description: SPEC SFS validation failure on NFS v2 over lustre.
3031 Details : Changes the blocksize for regular files to be 2x RPC size,
3032 and not depend on stripe size.
3034 Severity : enhancement
3036 Description: Multiple MD RPCs in flight.
3037 Details : Further unserialise some read-only MDS RPCs - learn about intents.
3038 To avoid overly-overloading MDS, introduce a limit on number of
3039 MDS RPCs in flight for a single client and add /proc controls
3040 to adjust this limit.
3042 Severity : enhancement
3044 Description: client read/write statistics
3045 Details : Add client read/write call usage stats for performance
3046 analysis of user processes.
3047 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
3048 file access. extents_stats shows chunk size distribution.
3049 extents_stats_per_process show chunk size distribution per
3052 Severity : enhancement
3054 Description: mds statistics
3055 Details : Add detailed mds operations statistics in
3056 /proc/fs/lustre/mds/*/stats.
3060 Description: Failure of copying files with lustre special EAs.
3061 Details : Client side always return success for setxattr call for lustre
3062 special xattr (currently only "trusted.lov").
3067 Description: Refcount LNET uuids
3068 Details : The global LNET uuid list grew linearly with every startup;
3069 refcount repeated list entries instead of always adding to
3072 Severity : enhancement
3074 Description: Dynamic service threads
3075 Details : Within a small range, start extra service threads
3076 automatically when the request queue builds up.
3079 Frequency : mixed-endian client/server environments
3081 Description: mixed-endian crashes
3082 Details : The new msg_v2 system had some failures in mixed-endian
3085 Severity : enhancement
3087 Description: Easy OST removal
3088 Details : OSTs can be permanently deactivated with e.g. 'lctl
3089 conf_param lustre-OST0001.osc.active=0'
3091 Severity : enhancement
3093 Description: MGS proc entries
3094 Details : Added basic proc entries for the MGS showing what filesystems
3097 Severity : enhancement
3099 Description: provide MGS failover
3100 Details : Added config lock reacquisition after MGS server failover.
3102 Severity : enhancement
3104 Description: add Linux 2.4 support
3105 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
3109 Description: a large application tries to do I/O to the same resource and dies
3110 in the middle of it.
3111 Details : Check the req->rq_arrival time after the call to
3112 ost_brw_lock_get(), but before we do anything about
3113 processing it & sending the BULK transfer request. This
3114 should help move old stale pending locks off the queue as
3115 quickly as obd_timeout.
3118 Frequency : when an incorrect nid is specified during startup
3120 Description: ptlrpc connect to non-existant node causes kernel crash
3121 Details : LNET can't be re-entered from an event callback, which
3122 happened when we expire a message after the export has been
3123 cleaned up. Instead, hand the zombie cleanup off to another
3126 Severity : enhancement
3128 Description: plain/inodebits lock performance improvement
3129 Details : Grouping plain/inodebits in granted list by their request modes
3130 and bits policy, thus improving the performance of search through
3134 Frequency : only if OST filesystem is corrupted
3136 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3137 Details : for a short time RPCs with bulk IO are in the replay list,
3138 but replay of bulk IOs is unimplemented. If the OST filesystem
3139 is corrupted due to disk cache incoherency and then replay is
3140 started it is possible to trip an assertion. Avoid putting
3141 committed RPCs into the replay list at all to avoid this issue.
3144 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3146 Description: System hang on startup
3147 Details : This bug allowed the liblustre (e.g. catamount) client to
3148 return to the app before handling all startup RPCs. This
3149 could leave the node unresponsive to lustre network traffic
3150 and manifested as a server ptllnd timeout.
3152 Severity : enhancement
3154 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3155 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
3156 boolean to control whether to print peer debug info when a
3157 client's RPC times out.
3160 Frequency : only for kernels with patches from Lustre below 1.4.3
3162 Description: Remove old rdonly API
3163 Details : Remove old rdonly API which unused from at least lustre 1.4.3
3166 Frequency : only for devices with external journals
3168 Description: Set external device read-only also
3169 Details : During a commanded failover stop, we set the disk device
3170 read-only while the server shuts down. We now also set any
3171 external journal device read-only at the same time.
3174 Frequency : when upgrading from 1.4 while trying to change parameters
3176 Description: The wrong (new) MDC name was used when setting parameters for
3177 upgraded MDT's. Also allows changing of OSC (and MDC)
3178 parameters if --writeconf is specified at tunefs upgrade time.
3181 Frequency : when setting specific ost indicies
3183 Description: QOS code breaks on skipped indicies
3184 Details : Add checks for missing OST indicies in the QOS code, so OSTs
3185 created with --index need not be sequential.
3187 Severity : enhancement
3189 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3190 Details : The uninit_groups feature works in conjunction with the kernel
3191 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3192 up the pass1 processing of e2fsck. This is a read-only feature
3193 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3194 mount filesystems that have had this feature enabled.
3196 Severity : enhancement
3198 Description: Improve multi-block allocation algorithm to avoid fragmentation
3199 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3200 allocation locality and avoid filesystem fragmentation.
3202 ------------------------------------------------------------------------------
3204 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
3206 * Support for kernels:
3207 2.4.21-47.0.1.EL (RHEL 3)
3208 2.6.5-7.283 (SLES 9)
3209 2.6.9-42.0.10.EL (RHEL 4)
3210 2.6.12.6 vanilla (kernel.org)
3211 2.6.16.27-0.9 (SLES 10)
3212 * Recommended e2fsprogs version: 1.39.cfs5
3214 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3218 Frequency : occasional, depends on client load and configuration
3219 Bugzilla : 12181, 12203
3220 Description: data loss for recently-modified files
3222 Details : In some cases it is possible that recently written or created
3223 files may not be written to disk in a timely manner (this should
3224 normally be within 30s unless client IO load is very high).
3225 The problem appears as zero-length files or files that are a
3226 multiple of 1MB in size after a client crash or client eviction
3227 that are missing data at the end of the file.
3229 This problem is more likely to be hit on clients where files are
3230 repeatedly created and unlinked in the same directory, clients
3231 have a large amount of RAM, have many CPUs, the filesystem has
3232 many OSTs, the clients are rebooted frequently, and/or the files
3233 are not accessed by other nodes after being written.
3235 The presence of the problem can be detected by looking at
3236 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
3237 smaller than the second (nr_unused) then dirty files will not
3238 be flushed automatically to disk. "sync; sleep 10" should be
3239 run several times on the node before unmounting it to update
3240 Lustre (this is also safe to run on nodes without this problem).
3242 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
3243 that can cause this same problem, so customers using that kernel
3244 also need to update the kernel in addition to Lustre. In order
3245 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
3247 It is normal that files written just before a client crash (less
3248 than 30s) may not yet have been flushed to disk, even for local
3252 Frequency : frequent on thin XT3 nodes
3254 Description: UUID collision on thin XT3 Linux nodes
3255 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
3256 randomly, since we relied on an insufficiently-seeded PRNG.
3261 Description: OSS hangs after "All ost request buffers busy"
3262 Details : A deadlock between quota and journal operations caused OSS
3263 hangs after printing "All ost request buffers busy."
3266 Frequency : always on liblustre builds
3268 Description: Cleanup compiler warnings on liblustre
3271 Frequency : always on liblustre builds on XT3
3273 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
3277 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
3278 > 32000 subdirectories
3279 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
3280 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
3281 subdirectory count crosses 32000. This will aid e2fsck to
3282 correctly handle more than 32000 subdirectories.
3285 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3287 Description: System hang on startup
3288 Details : This bug allowed the liblustre (e.g. catamount) client to
3289 return to the app before handling all startup RPCs. This
3290 could leave the node unresponsive to lustre network traffic
3291 and manifested as a server ptllnd timeout.
3293 Severity : enhancement
3295 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3296 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
3297 boolean to control whether to print peer debug info when a
3298 client's RPC times out.
3303 Description: make O_SYNC working on 2.6 kernels
3304 Details : 2.6 kernels use different method for mark pages for write,
3305 so need add a code to lustre for O_SYNC work.
3310 Description: Failure to close file and release space on NFS
3311 Details : Put inode details into lock acquired in ll_intent_file_open.
3312 Use mdc_intent_lock in ll_intent_open to properly
3313 detect all kind of errors unhandled by mdc_enqueue.
3318 Description: proc file read during shutdown sometimes raced obd removal,
3320 Details : Add lock to prevent obd access after proc file removal.
3323 Frequency : Only for files larger than 4GB on 32-bit clients.
3325 Description: improperly doing page alignment of locks
3326 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
3327 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3332 Description: RPCs being resent when they shouldn't be
3333 Details : Some RPCs that should not be resent are being resent. This
3334 can cause inconsistencies in the RPC state machine. Do not
3335 resend such requests.
3338 Frequency : rare, only with NFS export
3340 Description: Crash on NFS re-export node
3341 Details : under very unusual load conditions an assertion is hit in
3342 ll_intent_file_open()
3345 Frequency : only if OST filesystem is corrupted
3347 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3348 Details : for a short time RPCs with bulk IO are in the replay list,
3349 but replay of bulk IOs is unimplemented. If the OST filesystem
3350 is corrupted due to disk cache incoherency and then replay is
3351 started it is possible to trip an assertion. Avoid putting
3352 committed RPCs into the replay list at all to avoid this issue.
3357 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3358 Details : Large single O_DIRECT read and write calls can fail to allocate
3359 a sufficiently large buffer to process the request. In case of
3360 allocation failure the allocation is retried with a smaller
3361 buffer and broken into smaller requests.
3363 Severity : enhancement
3365 Description: Add -o localflock option to simulate old noflock behaviour.
3366 Details : This will achieve local-only flock/fcntl locks coherentness.
3371 Description: versioning check is incomplete
3372 Details : Checking the version difference of client vs. server, report
3373 error if the gap is too big.
3378 Description: add support PG_writeback bit
3379 Details : add support for PG_writeback bit for Lustre, for more carefull
3380 work with page cache in 2.6 kernel. This also fix some deadlocks
3381 and remove hack for work O_SYNC with 2.6 kernel.
3383 Severity : enhancement
3385 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3386 Details : The uninit_groups feature works in conjunction with the kernel
3387 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3388 up the pass1 processing of e2fsck. This is a read-only feature
3389 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3390 mount filesystems that have had this feature enabled.
3392 Severity : enhancement
3394 Description: Improve multi-block allocation algorithm to avoid fragmentation
3395 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3396 allocation locality and avoid filesystem fragmentation.
3398 ------------------------------------------------------------------------------
3400 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3402 * Support for kernels:
3403 2.6.9-42.0.3.EL (RHEL 4)
3404 2.6.5-7.276 (SLES 9)
3405 2.4.21-47.0.1.EL (RHEL 3)
3406 2.6.12.6 vanilla (kernel.org)
3407 2.6.16.21-0.8 (SLES10)
3408 * Recommended e2fsprogs version: 1.39.cfs2-0
3410 * The backwards-compatible /proc/sys/portals symlink has been removed
3411 in this release. Before upgrading, please ensure that you change
3412 any configuration scripts or /etc/sysctl.conf files that access
3413 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3414 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3415 in advance of the upgrade on any system running Lustre 1.4.6 or
3416 newer, since /proc/sys/lnet was added in that version.
3417 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3421 Frequency : only when quota is used
3423 Description: avoid scanning export list for quota master
3424 Details : Change the algorithms to avoid scanning export list in order
3425 to improve the efficiency.
3428 Frequency : MDS failover only, very rarely
3430 Description: "went back in time" messages on mds failover
3431 Details : The greatest transno may be lost when the current operation
3432 finishes with an error (transno==0) and the client's last_rcvd
3433 record is over-written. Save the greatest transno in the
3434 mds_last_transno for this case.
3437 Frequency : always for specific kernels and striping counts
3439 Description: client may get "Matching packet too big" without ACL support
3440 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
3441 when trying to access files in certain configurations. The
3442 clients should in fact be denied when mounting because they do
3443 not understand ACLs.
3446 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
3448 Description: many clients connecting with IO in progress causes connect timeouts
3449 Details : Avoid synchronous journal commits to avoid delays caused by many
3450 clients connecting/disconnecting when bulk IO is in progress.
3451 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
3452 OST_IO_PORTAL to avoid delays behind potentially many pending
3456 Frequency : occasionally with multiple writers to a single file
3458 Description: shared writes to file may result in wrong size reported by stat()
3459 Details : Allow growing of kms when extent lock is cancelled
3462 Frequency : always with random mmap IO to multi-striped file
3464 Description: mmap write might be lost if we are writing to a 'hole' in stripe
3465 Details : Only if the hole is at the end of OST object so that kms is too
3466 small. Fix is to increase kms accordingly in ll_nopage.
3469 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
3471 Description: writes to a missing object would leak memory on the OST
3472 Details : If there is an inconsistency between the MDS and OST filesystems,
3473 such that the MDS references an object that doesn't exist, writes
3474 to that object will leak memory due to incorrect cleanup in the
3475 error handling path, eventually running out of memory on the OST.
3480 Description: Creating too long symlink causes lustre errors
3481 Details : Check symlink and name lengths before sending requests to MDS.
3484 Frequency : only if flock is enabled (not on by default)
3486 Description: posix locks not released on fd closure on 2.6.9+
3487 Details : We failed to add posix locks to list of inode locks on 2.6.9+
3488 kernels, this caused such locks not to be released on fd close and
3489 then assertions on fs unmount about still used locks.
3492 Frequency : MDS failover only, very rarely
3494 Description: clients may get ASSERTION(granted_lock != NULL)
3495 Details : When request was taking a long time, and a client was resending
3496 a getattr by name lock request. The were multiple lock requests
3497 with the same client lock handle and
3498 mds_getattr_name->fixup_handle_for_resent_request found one of the
3499 lock handles but later failed with ASSERTION(granted_lock != NULL).
3504 Description: handle->h_buffer_credits > 0, assertion failure
3505 Details : h_buffer_credits is zero after truncate, causing assertion
3506 failure. This patch extends the transaction or creates a new
3510 Frequency : NFS re-export or patchless client
3511 Bugzilla : 11179, 10796
3512 Description: Crash on NFS re-export node (__d_move)
3513 Details : We do not want to hash the dentry if we don't have a lock.
3514 But if this dentry is later used in d_move, we'd hit uninitialised
3515 list head d_hash, so we just do this to init d_hash field but
3516 leave dentry unhashed.
3519 Frequency : NFS re-export or patchless client
3521 Description: NFS exports has problem with symbolic link
3522 Details : lustre client didn't properly install dentry when re-exported
3523 to NFS or running patchless client.
3526 Frequency : NFS re-export or patchless client
3528 Description: Various nfs/patchless fixes.
3529 Details : fixes reuse disconected alias for lookup process - this fixes
3530 warning "find_exported_dentry: npd != pd",
3531 fix permission error with open files at nfs.
3532 fix apply umask when do revalidate.
3535 Frequency : occasional
3537 Description: Crash on NFS re-export node
3538 Details : calling clear_page() on the wrong pointer triggered oops in
3539 generic_mapping_read().
3542 Frequency : rarely, using O_DIRECT IO
3544 Description: unaligned directio crashes client with LASSERT
3545 Details : check for unaligned buffers before trying any requests.
3548 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
3550 Description: stale data returned from RAID cache
3551 Details : If only a small amount of IO is done to the RAID device before
3552 reading it again it is possible to get stale data from the RAID
3553 cache instead of reading it from disk.
3556 Frequency : always for sles10 kernel
3558 Description: sles10 support
3559 Details : ll_follow_link: compile fixes and using of nd_set_link
3560 under newer kernels.
3563 Frequency : depends on arch, kernel and compiler version, always on sles10
3566 Description: recursive or deep enough symlinks cause stack overflow
3567 Details : getting rid of large stack-allocated variable in
3571 Frequency : depends on hardware
3573 Description: lustre write performance loss in the SLES10 kernel
3574 Details : the performance loss is caused by using of write barriers in the
3575 ext3 code. The SLES10 kernel turns barrier support on by
3576 default. The fix is to undo that change for ldiskfs.
3578 ------------------------------------------------------------------------------
3580 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
3582 * Support for kernels:
3583 2.6.9-42.0.3EL (RHEL 4)
3584 2.6.5-7.276 (SLES 9)
3585 2.4.21-47.0.1.EL (RHEL 3)
3586 2.6.12.6 vanilla (kernel.org)
3590 Frequency : quota enabled and large files being deleted
3592 Description: releasing more than 4GB of quota at once hangs OST
3593 Details : If a user deletes more than 4GB of files on a single OST it
3594 will cause the OST to spin in an infinite loop. Release
3595 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3600 Description: statfs data retrieved from /proc may be stale or zero
3601 Details : When reading per-device statfs data from /proc, in the
3602 {kbytes,files}_{total,free,avail} files, it may appear
3603 as zero or be out of date.
3606 Frequency : systems with MD RAID1 external journal devices
3608 Description: lconf's call to blkid is confused by RAID1 journal devices
3609 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3610 of returning all block devices that match the journal UUID.
3613 Frequency : always, for aggregate stripe size over 4GB
3615 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
3616 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
3617 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
3620 Frequency : always if "lfs find" used on a local file/directory
3622 Description: "lfs find" segfaults if used on a local file/directory
3623 Details : The case where a directory component was not specified wasn't
3624 handled correctly. Handle this properly.
3627 Frequency : always on ppc64
3629 Description: the write to an ext3 filesystem mounted with mballoc got stuck
3630 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3631 perform endianness conversion.
3634 Frequency : rarely (truncate to non-zero file size after write under load)
3635 Bugzilla : 10730, 10687
3636 Description: Files padded with zeros to next 4K multiple
3637 Details : With filesystems mounted using the "extents" option (2.6 kernels)
3638 it is possible that files that are truncated to a non-zero size
3639 immediately after being written are filled with zero bytes beyond
3640 the truncated size. No file data is lost.
3642 Severity : enhancement
3644 Description: Allow recovery/failover for liblustre clients.
3645 Details : liblustre clients were unaware of failover configurations until
3648 Severity : enhancement
3650 Description: user file locks should fail when not mounting with flock option
3651 Details : Set up an error-returning stub in ll_file_operations.lock field
3652 to prevent incorrect behaviour when client is mounted without
3653 flock option. Also, set up properly f_op->flock field for
3657 Frequency : always on ia64
3659 Description: "lfs df" loops on printing out MDS statfs information
3660 Details : The obd_ioctl_data was not initialized and in some systems
3661 this caused a failure during the ioctl that did not return
3662 an error. Initialize the struct and return an error on failure.
3665 Frequency : SLES 9 only
3667 Description: Error of copying files with lustre special EAs as root
3668 Details : Client side always return success for setxattr call for lustre
3669 special xattr (currently only "trusted.lov").
3672 Frequency : rarely on clusters with both ia64+i386 clients
3674 Description: ia64+i686 clients doing shared IO on the same file may LBUG
3675 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
3676 clients are doing concurrent writes to the same file it is
3677 possible that the ia64 clients may LASSERT because the OST
3678 extent locks are not PAGE_SIZE aligned. Ensure that grown
3679 locks are always aligned on the request boundary.
3682 Frequency : specific use, occasional
3684 Description: Overwriting in use executable truncates on-disk binary image
3685 Details : If one node attempts to overwrite an executable in use by
3686 another node, we now correctly return ETXTBSY instead of
3687 truncating the file.
3689 Severity : enhancement
3691 Description: Async OSC create to avoid the blocking unnecessarily.
3692 Details : If a OST has no remain object, system will block on the creating
3693 when need to create a new object on this OST. Now, ways use
3694 pre-created objects when available, instead of blocking on an
3695 empty osc while others are not empty. If we must block, we block
3696 for the shortest possible period of time.
3701 Description: chmod on Lustre root is propagated to other clients
3702 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
3703 invalid by the follow_mount time.
3706 Frequency : liblustre clients only
3708 Description: Race in 'instant cancel' lock handling could lead to such locks
3709 never to be granted in case of SMP MDS
3710 Details : Do not destroy not yet granted but cbpending locks in
3714 Frequency : replay/resend of open
3716 Description: non null lock assetion failure in mds_intent_policy
3717 Details : Trying to replay/resend lockless open requests resulted in
3718 mds_open() returning 0 with no lock. Now it sets a flag if
3719 a lock is going to be returned.
3721 Severity : enhancement
3723 Description: Checksum enhancements
3724 Details : New checksum enhancements allow for resending RPCs that failed
3727 Severity : enhancement
3729 Description: Tunables on number of dirty pages in cacche
3730 Details : Allow to set limit on number of dirty pages cached.
3735 Description: client crash on unmount - lock still has references
3736 Details : In some error handling cases it was possible to leak a lock
3737 reference on a client while accessing a file. This was not
3738 harmful to the client during operation, but would cause the
3739 client to crash when the filesystem is unmounted.
3742 Frequency : specific case, rare
3744 Description: ETXTBSY on mds though file not in use by client
3745 Details : ETXTBSY is no longer incorrectly returned when attempting to
3746 chmod or chown a directory that the user previously tried to
3747 execute or a currently-executing binary.
3750 Frequency : extremely rare except on liblustre-based clients
3752 Description: Lustre space not freed when files are deleted
3753 Details : Clean up open-unlinked files after client eviction. Previously
3754 the unlink was skipped and the files remained as orphans.
3759 Description: OST failure "would be an LBUG" in waiting_locks_callback()
3760 Details : In some cases it was possible to send a blocking callback to
3761 a client doing a glimpse, even though that client didn't get
3762 a lock granted. When the glimpse lock is cancelled on the OST
3763 the freed lock is left on the waiting list and corrupted the list.
3766 Frequency : all core dumps
3768 Description: Broke core dumps to lustre
3769 Details : Negative dentry may be unhashed if parent does not have UPDATE
3770 lock, but some callers, e.g. do_coredump, expect dentry to be
3771 hashed after successful create, hash it in ll_create_it.
3773 ------------------------------------------------------------------------------
3775 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
3777 * Support for kernels:
3778 2.6.9-42.0.2.EL (RHEL 4)
3779 2.6.5-7.276 (SLES 9)
3780 2.4.21-40.EL (RHEL 3)
3781 2.6.12.6 vanilla (kernel.org)
3785 Frequency : always on RHEL 3
3787 Description: Number of open files grows over time
3788 Details : The number of open files grows over time, whether or not
3789 Lustre is started. This was due to a filp leak introduced
3790 by one of our kernel patches.
3792 ------------------------------------------------------------------------------
3794 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
3796 * Support for kernels:
3797 2.6.9-42.EL (RHEL 4)
3798 2.6.5-7.276 (SLES 9)
3799 2.4.21-40.EL (RHEL 3)
3800 2.6.12.6 vanilla (kernel.org)
3805 Bugzilla : 5719, 9635, 9792, 9684
3806 Description: OST (or MDS) trips assertions in (re)connection under heavy load
3807 Details : If a server is under heavy load and cannot reply to new
3808 connection requests before the client resends the (re)connect,
3809 the connection handling code can behave badly if two service
3810 threads are concurrently handing separate (re)connections from
3811 the same client. Add better locking to the connection handling
3812 code, and ensure that only a single connection will be processed
3813 for a given client UUID, even if the lock is dropped.
3815 Severity : enhancement
3817 Description: add TCP zero-copy support to kernel
3818 Details : Add support to the kernel TCP stack to allow zero-copy bulk
3819 sends if the hardware supports scatter-gather and checksumming.
3820 This allows socklnd to do client-write and server-read more
3821 efficiently and reduce CPU utilization from skbuf copying.
3824 Frequency : only if NFS exporting from client
3826 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
3827 Details : If setting attributes on a file created under NFS that had
3828 never been opened it would be possible to oops the client
3829 if the file had no objects.
3832 Frequency : always for liblustre
3834 Description: liblustre client does MDS+OSTs setattr RPC for each write
3835 Details : When doing a write from a liblustre client, the client
3836 incorrectly issued an RPC to the MDS and each OST the file was
3837 striped over in order to update the timestamps. When writing
3838 with small chunks and many clients this could overwhelm the MDS
3839 with RPCs. In all cases it would slow down the write because
3840 these RPCs are unnecessary.
3842 Severity : enhancement
3844 Description: allow number of MDS service threads to be changed at module load
3845 Details : It is now possible to change the number of MDS service threads
3846 running. Adding "options mds mds_num_threads={N}" to the MDS's
3847 /etc/modprobe.conf will set the number of threads for the next
3848 time Lustre is restarted (assuming the "mds" module is also
3849 reloaded at that time). The default number of threads will
3850 stay the same, 32 for most systems.
3855 Description: OST crash if filesystem is unformatted or corrupt
3856 Details : If an OST is started on a device that has never been formatted
3857 or if the filesystem is corrupt and cannot even mount then the
3858 error handling cleanup routines would dereference a NULL pointer.
3863 Description: NULL pointer deref in llap_from_page.
3864 Details : get_cache_page_nowait can return a page with NULL (or otherwise
3865 incorrect) mapping if the page was truncated/reclaimed while it was
3866 searched for. Check for this condition and skip such pages when
3867 doing readahead. Introduce extra check to llap_from_page() to
3868 verify page->mapping->host is non-NULL (so page is not anonymous).
3871 Frequency : Sometimes when using sys_sendfile
3873 Description: "page not covered by a lock" warnings from ll_readpage
3874 Details : sendfile called ll_readpage without right page locks present.
3875 Now we introduced ll_file_sendfile that does necessary locking
3876 around call to generic_file_sendfile() much like we do in
3880 Frequency : with certain MDS communication failures at client mount time
3882 Description: NULL pointer deref after failed client mount
3883 Details : a client connection request may delayed by the network layer
3884 and not be sent until after the PTLRPC layer has timed out the
3885 request. If the client fails the mount immediately it will try
3886 to clean up before the network times out the request. Add a
3887 reference from the request import to the obd device and delay
3888 the cleanup until the network drops the request.
3891 Frequency : occasionally during client (re)connect
3893 Description: assertion failure during client (re)connect
3894 Details : processing a client connection request may be delayed by the
3895 client or server longer than the client connect timeout. This
3896 causes the client to resend the connection request. If the
3897 original connection request is replied in this interval, the
3898 client may trip an assertion failure in ptlrpc_connect_interpret()
3899 which thought it would be the only running connect process.
3902 Frequency : only with obd_echo servers and clients that are rebooted
3904 Description: kernel BUG accessing uninitialized data structure
3905 Details : When running an obd_echo server it did not start the ping_evictor
3906 thread, and when a client was evicted an uninitialized data
3907 structure was accessed. Start the ping_evictor in the RPC
3908 service startup instead of the OBD startup.
3910 Severity : enhancement
3911 Bugzilla : 10193 (patchless)
3912 Description: Remove dependency on various unexported kernel interfaces.
3913 Details : No longer need reparent_to_init, exit_mm, exit_files,
3914 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
3917 Frequency : rare (only users of deprecated and unsupported LDAP config)
3919 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
3920 Details : LDAP apparently contains 'lustreName' attributes instead of
3921 'name'. A simple remapping of the name is sufficient.
3924 Frequency : rare (only with non-default dump_on_timeout debug enabled)
3926 Description: waiting_locks_callback trips kernel BUG if client is evicted
3927 Details : Running with the dump_on_timeout debug flag turned on makes
3928 it possible that the waiting_locks_callback() can try to dump
3929 the Lustre kernel debug logs from an interrupt handler. Defer
3930 this log dumping to the expired_lock_main() thread.
3932 Severity : enhancement
3934 Description: Support NFS exporting on 2.6 kernels.
3935 Details : Implement non-rawops metadata methods for NFS server to use without
3936 changing NFS server code.
3939 Frequency : very rare (synthetic metadata workload only)
3941 Description: two racing renames might cause an MDS thread to deadlock
3942 Details : Running the "racer" program may cause one MDS thread to rename
3943 a file from being the source of a rename to being the target of
3944 a rename at exactly the same time that another thread is doing
3945 so, and the second thread has already enqueued these locks after
3946 doing a lookup of the target and is trying to relock them in
3947 order. Ensure that we don't try to re-lock the same resource.
3950 Frequency : only very large systems with liblustre clients
3952 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
3953 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
3954 all OSTs in parallel. This allows the eviction of stale liblustre
3955 clients to proceed much faster than if they were done in series,
3956 and also offers similar improvements for other set_info RPCs.
3961 Description: excessive CPU usage during initial read phase on client
3962 Details : During the initial read phase on a client, it would agressively
3963 retry readahead on the file, consuming too much CPU and impacting
3964 performance (since 1.4.5.8). Improve the readahead algorithm
3965 to avoid this, and also improve some other common cases (read
3966 of small files in particular, where "small" is files smaller than
3967 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
3972 Description: MDS crash when receiving packet with unknown intent.
3973 Details : Do not LBUG in unknown intent case, just return -EFAULT
3975 Severity : enhancement
3976 Bugzilla : 9293, 9385
3977 Description: MDS RPCs are serialised on client. This is unnecessary for some.
3978 Details : Do not serialize getattr (non-intent version) and statfs.
3981 Frequency : occasional, when OST network is overloaded/intermittent
3983 Description: client evicted by OST after bulk IO timeout
3984 Details : If a client sends a bulk IO request (read or write) the OST
3985 may evict the client if it is unresposive to its data GET/PUT
3986 request. This is incorrect if the network is overloaded (takes
3987 too long to transfer the RPC data) or dropped the OST GET/PUT
3988 request. There is no need to evict the client at all, since
3989 the pinger and/or lock callbacks will handle this, and the
3990 client can restart the bulk request.
3993 Frequency : Always when mmapping file with no objects
3995 Description: client crashes when mmapping file with no objects
3996 Details : Check that we actually have objects in a file before doing any
3997 operations on objects in ll_vm_open, ll_vm_close and
4003 Description: Request leak when working with deleted CWD
4004 Details : Introduce advanced request refcount tracking for requests
4005 referenced from lustre intent.
4007 Severity : Enhancement
4009 Description: Cache open file handles on client.
4010 Details : MDS now will return special lock along with openhandle, if
4011 requested and client is allowed to hold openhandle, even if unused,
4012 until such a lock is revoked. Helps NFS a lot, since NFS is opening
4013 closing files for every read/write openration.
4015 Severity : Enhancement
4017 Description: Cache open negative dentries on client when possible.
4018 Details : Guard negative dentries with UPDATE lock on parent dir, drop
4019 negative dentries on lock revocation.
4024 Description: Remounting a client read-only wasn't possible with a zconf mount
4025 Details : It wasn't possible to remount a client read-only with llmount.
4027 Severity : enhancement
4028 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
4029 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
4030 MPICH is included to add Lustre-specific ADIO interfaces.
4031 This is based closely on the UFS ADIO layer and only differs
4032 in file creation, in order to allow the OST striping to be set.
4033 This is user-contributed code and not supported by CFS.
4038 Description: extended inode attributes (immutable, append-only) work improperly
4039 when 2.4 and 2.6 kernels are used on client/server or vice versa
4040 Details : Introduce kernel-independent values for these flags.
4042 Severity : enhancement
4045 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
4046 Details : Many of the /proc/ tunables can only be tuned at a megabyte
4047 granularity. Now, Fractional MB granularity is be supported,
4048 this is very useful for low memory system.
4050 Severity : enhancement
4052 Description: Getattr by fid
4053 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
4054 locks, avoids extra getattr rpc requests to MDS, allows '/' to
4055 have locks and avoids getattr rpc requests for it on every stat.
4058 Frequency : Always, for filesystems larger than 2TB
4060 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4061 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
4062 filesystems larger than 2TB to be mounted with mballoc enabled.
4065 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
4067 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4068 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
4069 filesystems larger than 2TB is mounted then the kernel will
4070 silently corrupt the start of the filesystem. CONFIG_LBD is
4071 enabled for all CFS-supported kernels, but the possibility of
4072 this happening with a modified kernel config exists.
4074 Severity : enhancement
4076 Description: add client O_DIRECT support for 2.6 kernels
4077 Details : It is now possible to do O_DIRECT reads and writes to files
4078 in the Lustre client mountpoint on 2.6 kernel clients.
4080 Severity : enhancement
4082 Description: parallel glimpse, setattr, statfs, punch, destroy requests
4083 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
4084 parallel, not waiting for response from every OST before sending
4085 a rpc to the next OST.
4090 Description: setattr vs write race when updating file timestamps
4091 Details : Client processes that update a file timestamp into the past
4092 right after writing to the file (e.g. tar) it is possible that
4093 the updated file modification time can be reset to the current
4094 time due to a race between processing the setattr and write RPC.
4096 Severity : enhancement
4098 Description: Bring 'lfs find' closer in line with regular Linux find.
4099 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
4100 -print0 options and obtains all the needed info through the lustre
4103 Severity : enhancement
4105 Description: support up to 1024 configured devices on one node
4106 Details : change obd_dev array from statically allocated to dynamically
4107 allocated structs as they are first used to reduce memory usage
4112 Description: Flush dirty partially truncated pages during truncate
4113 Details : Immediatelly flush partially truncated pages in filter_setattr,
4114 this way we completely avoid having any pages in page cache on OST
4115 and can retire ugly workarounds during writes to flush such pages.
4120 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
4121 Details : protect inode from truncation within vfs_unlink() context
4122 just take a reference before calling vfs_unlink() and release it
4123 when parent's i_sem is free.
4128 Description: last_id value checked outside lock on OST caused LASSERT failure
4129 Details : If there were multiple MDS->OST object precreate requests in
4130 flight, it was possible that the OST's last object id was checked
4131 outside a lock and incorrectly tripped an assertion. Move checks
4132 inside locks, and discard old precreate requests.
4135 Frequency : always, if extents are used on OSTs
4137 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
4138 Details : OSTs using the extents format would not zero the high 16 bits of
4139 the index physical block number. This is not a problem for any
4140 OST filesystems smaller than 16TB, and no kernels support ext3
4141 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
4142 new/modified files) and can be fixed for existing filesystems
4143 with e2fsprogs-1.39-cfs1.
4148 Description: import connection selection may be incorrect if timer wraps
4149 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
4150 import connections to be ignored if the 32-bit jiffies counter
4151 wraps. Use a 64-bit jiffies counter.
4154 Frequency : very large clusters immediately after boot
4156 Description: LNET request buffers exhausted under heavy short-term load
4157 Details : If a large number of client requests are generated on a service
4158 that has previously never seen so many requests it is possible
4159 that the request buffer growth cannot keep up with the spike in
4160 demand. Instead of dropping incoming requests, they are held in
4161 the LND until the RPC service can accept more requests.
4164 Frequency : Sometimes during replay
4166 Description: Assertion failure in ll_local_open after replay.
4167 Details : If replay happened on an open request reply before we were able
4168 to set replay handler, reply will become not swabbed tripping the
4169 assertion in ll_local_open. Now we set the handler right after
4170 recognising of open request
4173 Frequency : very rare
4175 Description: kernel reports "badness in vsnprintf"
4176 Details : Reading from the "recovery_status" /proc file in small chunks
4177 may cause a negative length in lprocfs_obd_rd_recovery_status()
4178 call to vsnprintf() (which is otherwise harmless). Exit early
4179 if there is no more space in the output buffer.
4181 Severity : enhancement
4183 Description: clear OBD RPC statistics by writing to them
4184 Details : It is now possible to clear the OBD RPC statistics by writing
4185 to the "stats" file.
4190 Description: Client mtime is not the same on different clients after utimes
4191 Details : In some cases, the client was using the utimes() syscall on
4192 a file cached on another node. The clients now validate the
4193 ctime from the MDS + OSTs to determine which one is right.
4198 Description: Inability to activate failout mode
4199 Details : lconf script incorrectly assumed that in pythong string's numeric
4200 value is used in comparisons.
4203 Frequency : always with multiple stripes per file
4205 Description: Inefficient object allocation for mutli-stripe files
4206 Details : When selecting which OSTs to stripe files over, for files with
4207 a stripe count that divides evenly into the number of OSTs,
4208 the MDS is always picking the same starting OST for each file.
4209 Return the OST selection heuristic to the original design.
4214 Description: mount failures may take full timeout to return an error
4215 Details : Under some heavy load conditions it is possible that a
4216 failed mount can wait for the full obd_timeout interval,
4217 possibly several minutes, before reporting an error.
4218 Instead return an error as soon as the status is known.
4220 Frequency : quota enabled and large files being deleted
4222 Description: releasing more than 4GB of quota at once hangs OST
4223 Details : If a user deletes more than 4GB of files on a single OST it
4224 will cause the OST to spin in an infinite loop. Release
4225 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4230 Description: statfs data retrieved from /proc may be stale or zero
4231 Details : When reading per-device statfs data from /proc, in the
4232 {kbytes,files}_{total,free,avail} files, it may appear
4233 as zero or be out of date.
4236 Frequency : systems with MD RAID1 external journal devices
4238 Description: lconf's call to blkid is confused by RAID1 journal devices
4239 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4240 of returning all block devices that match the journal UUID.
4243 Frequency : always, for aggregate stripe size over 4GB
4245 Description: assertion fails when trying to use 4GB stripe size
4246 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
4247 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
4250 Frequency : always on ppc64
4252 Description: the first write on an ext3 filesystem with mballoc got stuck
4253 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4254 perform endianness conversion.
4256 ------------------------------------------------------------------------------
4258 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4260 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
4261 INTEROPERATE with older versions automatically. Please read the
4262 user documentation before upgrading any part of a live system.
4263 * WARNING: Lustre networking configuration changes are required with
4264 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
4267 * Support for kernels:
4268 2.6.9-22.0.2.EL (RHEL 4)
4269 2.6.5-7.244 (SLES 9)
4270 2.6.12.6 vanilla (kernel.org)
4273 Severity : enhancement
4274 Bugzilla : 7981/8208
4275 Description: Introduced Lustre Networking (LNET)
4276 Details : LNET is new networking infrastructure for Lustre, it includes
4277 a reorganized network configuration mode (see the user
4278 documentation for full details) as well as support for routing
4279 between different network fabrics. Lustre Networking Devices
4280 (LNDs) for the supported network fabrics have also been
4281 created for this new infrastructure.
4283 Severity : enhancement
4284 Description: Introduced Access control lists
4285 Details : clients can set ACLs on files and directories in order to have
4286 more fine-grained permissions than the standard Unix UGO+RWX.
4287 The MDS must be started with the "-o acl" mount option.
4289 Severity : enhancement
4290 Description: Introduced filesystem quotas
4291 Details : Administrators may now establish per-user quotas on the
4294 Severity : enhancement
4296 Description: Configuration change for the XT3
4297 The PTLLND is now used to run Lustre over Portals on the XT3
4298 The configure option(s) --with-cray-portals are no longer used.
4299 Rather --with-portals=<path-to-portals-includes> is used to
4300 enable building on the XT3. In addition to enable XT3 specific
4301 features the option --enable-cray-xt3 must be used.
4306 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
4307 Details : A race between allocating a new llcd and re-getting the llcd_lock
4308 allowed another thread to grab newly-allocated llcd.
4310 Severity : enhancement
4312 Description: 2.6 OST async journal commit and locking fix to improve performance
4313 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
4314 2.6 kernels are now async as they already were in 2.4 kernels so
4315 that they can commit concurrently with the network bulk transfer.
4316 For block-allocated files the filter allocation semaphore is held
4317 to avoid filesystem fragmentation during allocation. BKL lock
4318 removed for 2.6 xattr operations where it is no longer needed.
4323 Description: lconf incorrectly determined whether two IP networks could talk
4324 Details : In some more complicated routing and multiple-network
4325 configurations, lconf will avoid trying to make a network
4326 connection to a disjoint part of the IP space. It was doing the
4327 math incorrectly for one set of cases.
4332 Description: Fix for potential infinite loop processing records in an llog.
4333 Details : If an llog record is corrupted/zeroed, it is possible to loop
4334 forever in llog_process(). Validate the llog record length
4335 and skip the remainder of the block on error.
4338 Frequency : occasional (liblustre only)
4340 Description: liblustre could not open files whose last component is a symlink
4341 Details : sysio_path_walk() would incorrectly pass the open intent to
4342 intermediate path components.
4345 Frequency : rare (liblustre only with non-standard tuning)
4346 Bugzilla : 7201 (7350)
4347 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4348 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4349 after it has been released, only do this for referenced locks
4351 Severity : enhancement
4353 Description: specifying an (invalid) directory default stripe_size of -1
4354 would reset the directory default striping
4355 Details : stripe_size -1 was used internally to signal directory stripe
4356 removal, now use "all default" to signal dir stripe removal
4357 as a directory striping of "all default" is not useful
4360 Frequency : common for large clusters running liblustre clients
4362 Description: doing an ls when liblustre clients are running is slow
4363 Details : sending a glimpse AST to a liblustre client waits for every AST
4364 to time out, as liblustre clients will not respond. Since they
4365 cannot cache data we refresh the OST lock LVB from disk instead.
4367 Severity : enhancement
4369 Description: doing an ls at the same time as file IO can be slow
4370 Details : enqueue and other "small" requests can be blocked behind many
4371 large IO requests. Create a new OST IO portal for non-IO
4372 requests so they can be processed faster.
4375 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4377 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4378 created file with mode 000
4379 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4380 allow an ftruncate() as a non-root user to a file with mode 000.
4381 Lustre now respects this flag to disable mode checks when
4382 truncating a file owned by the user
4385 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4388 Description: Revoking locks from clients that went dead or catatonic might take
4390 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4391 cancellation of such locks instant on servers without waiting for
4392 any reply from clients. Clients drops these locks when cancel
4393 notification from server is received without replying.
4396 Frequency : liblustre-only, when liblustre client dies or becomes busy
4398 Description: Doing ls on Linux clients can take a long time with active
4400 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4401 granting such locks to it in the first place if possible. Locks
4402 are taken by proxy on the OST during the read or write and
4403 dropped immediately afterward. Add connect flags handling, do
4404 not grant locks to liblustre clients for glimpse ASTs.
4406 Severity : enhancement
4408 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4409 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4410 sequential reads and behaves badly with applications doing random
4411 reads. Improve it by reading ahead at least read region, and
4412 avoiding excessive large RPC for small reads.
4414 Severity : enhancement
4416 Description: Creating more than 1000 files for a single job may cause a load
4417 imbalance on the OSTs if there are also a large number of OSTs.
4418 Details : qos_prep_create() uses an OST index reseed value that is an
4419 even multiple of the number of available OSTs so that if the
4420 reseed happens in the middle of the object allocation it will
4421 still utilize the OSTs as uniformly as possible.
4426 Description: OST or MDS may oops in ping_evictor_main()
4427 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
4428 but doesn't restart at beginning of obd_exports_timed list
4431 Severity : enhancement
4433 Description: improve by-nid export eviction on the MDS and OST
4434 Details : allow multiple exports with the same NID to be evicted at one
4435 time without re-searching the exports list.
4438 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
4440 Description: MDS may oops in groups_free()
4441 Details : in rare race conditions a newly allocated group_info struct is
4442 freed again, and this can be NULL. The 2.4 compatibility code
4443 for groups_free() checked for a NULL pointer, but 2.6 did not.
4446 Frequency : common for liblustre clients doing little filesystem IO
4447 Bugzilla : 9352, 7313
4448 Description: server may evict liblustre clients accessing contended locks
4449 Details : if a client is granted a lock or receives a completion AST
4450 with a blocking AST already set it would not reply to the AST
4451 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
4454 Frequency : lfs setstripe, only systems with more than 160 OSTs
4456 Description: unable to set striping with a starting offset beyond OST 160
4457 Details : llapi_create_file() incorrectly limited the starting stripe
4458 index to the maximum single-file stripe count.
4461 Frequency : LDAP users only
4463 Description: lconf did not handle in-kernel recovery with LDAP properly
4464 Details : lconf/LustreDB get_refs() is searching the wrong namespace
4466 Severity : enhancement
4468 Description: bind OST threads to NUMA nodes to improve performance
4469 Details : all OST threads are uniformly bound to CPUs on a single NUMA
4470 node and do their allocations there to localize memory access
4472 Severity : enhancement
4474 Description: llmount can determine client NID directly from Myrinet (GM)
4475 Details : the client NID code from gmnalnid was moved directly into
4476 llmount, removing the need to use this or specifying the
4477 client NID explicitly when mounting GM clients with zeroconf
4480 Frequency : if client is started with down MDS
4482 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
4483 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
4484 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
4485 ptlrpc_import_delayed_req() because we didn't check if the
4486 request was interrupted, and we also didn't break out of the
4487 event loop if there was no timeout
4492 Description: data loss during non-page-aligned writes to a single file from
4493 both multiple nodes and multiple threads on one node at same time
4494 Details : updates to KMS and lsm weren't protected by common lock. Resulting
4495 inconsistency led to false short-reads, that were cached and later
4496 used by ->prepare_write() to fill in partially written page,
4497 leading to data loss.
4500 Frequency : always, if lconf --abort_recovery used
4502 Description: lconf --abort_recovery fails with 'Operation not supported'
4503 Details : lconf was attempting to abort recovery on the MDT device and not
4506 Severity : enhancement
4508 Description: remove cleanup logs
4509 Details : replace lconf-generated cleanup logs with lustre internal
4510 cleanup routines. Eliminates the need for client-cleanup and
4513 Severity : enhancement
4515 Description: add support for EAs (user and system) on lustre filesystems
4516 Details : it is now possible to store extended attributes in the Lustre
4517 client filesystem, and with the user_xattr mount option it
4518 is possible to allow users to store EAs on their files also
4520 Severity : enhancement
4522 Description: Add possibility (config option) to show minimal available OST free
4524 Details : When compiled with --enable-mindf configure option, statfs(2)
4525 (and so, df) will return least minimal free space available from
4526 all OSTs as amount of free space on FS, instead of summary of
4527 free spaces of all OSTs.
4529 Severity : enhancement
4531 Description: do not expand extent locks acquired on OST-side
4532 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
4533 by server: they are not cached anyway.
4536 Frequency : when mmap is used/binaries executed from Lustre
4538 Description: Unmmap pages before throwing them away from read cache.
4539 Details : llap_shrink cache now attempts to unmap pages before discarding
4540 them (if unmapping failed - do not discard). SLES9 kernel has
4541 extra checks that trigger if this unmapping is not done first.
4546 Description: lconf didn't resolve symlinks before checking to see whether a
4547 given mountpoint was already in use
4550 Frequency : when migrating failover services
4551 Bugzilla : 6395, 9514
4552 Description: When migrating a subset of services from a node (e.g. failback
4553 from a failover service node) the remaining services would
4554 time out and evict clients.
4555 Details : lconf --force (implied by --failover) sets the global obd_timeout
4556 to 5 seconds in order to quickly disconnect, but this caused
4557 other RPCs to time out too quickly. Do not change the global
4558 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
4560 Severity : enhancement
4561 Frequency : if MDS is started with down OST
4562 Bugzilla : 9439,5706
4563 Description: Allow startup/shutdown of an MDS without depending on the
4564 availability of the OSTs.
4565 Details : Asynchronously call mds_lov_synchronize during MDS startup.
4566 Add appropriate locking and lov-osc refcounts for safe
4567 cleaning. Add osc abort_inflight calls in case the
4568 synchronize never started.
4571 Frequency : occasional (Cray XT3 only)
4573 Description: root not authorized to access files in CRAY_PORTALS environment
4574 Details : The client process capabilities were not honoured on the MDS in
4575 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
4576 been accessed by an authorized user then root was able to access
4577 the file on the local client also. The root user capabilities
4578 are now allowed on the MDS, as this environment has secure UID.
4581 Frequency : occasional
4583 Description: ldiskfs "too long searching" message happens too often
4584 Details : A debugging message (otherwise harmless) prints too often on
4585 the OST console. This has been reduced to only happen when
4586 there are fragmentation problems on the filesystem.
4591 Description: Division by zero in statfs when all OSCs are inactive
4592 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
4593 lov_statfs divides by value returned by lov_get_stripecnt().
4597 Bugzilla : 9489, 3273
4598 Description: First write from each client to each OST was only 4kB in size,
4599 to initialize client writeback cache, which caused sub-optimal
4600 RPCs and poor layout on disk for the first writen file.
4601 Details : Clients now request an initial cache grant at (re)connect time
4602 and so that they can start streaming writes to the cache right
4603 away and always do full-sized RPCs if there is enough data.
4604 If the OST is rebooted the client also re-establishes its grant
4605 so that client cached writes will be honoured under the grant.
4610 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
4611 Details : Now I/O RPCs go to different portal number and (presumably) fast
4612 lock requests (and glimses) and other RPCs get their own service
4613 threads pool that should be able to service those RPCs
4616 Severity : enhancement
4618 Description: Ability to exchange lustre version between client and servers and
4619 issue warnings at client side if client is too old. Also for
4620 liblustre clients there is ability to refuse connection of too old
4622 Details : New 'version' field is added to connect data structure that is
4623 filled with version info. That info is later checked by server and
4627 Frequency : rare, liblustre only.
4628 Bugzilla : 9296, 9581
4629 Description: Two simultaneous writes from liblustre at offset within same page
4630 might proceed at the same time overwriting eachother with stale
4632 Details : I/O lock withing llu_file_prwv was released too early, before data
4633 actually was hitting the wire. Extended lock-holding time until
4634 server acknowledges receiving data.
4637 Frequency : extremely rare. Never observed in practice.
4639 Description: avoid generating lustre_handle cookie of 0.
4640 Details : class_handle_hash() generates handle cookies by incrementing
4641 global counter, and can hit 0 occasionaly (this is unlikely, but
4642 not impossible, because initial value of cookie counter is
4643 selected randonly). Value of 0 is used as a sentinel meaning
4644 "unassigned handle" --- avoid it. Also coalesce two critical
4645 sections in this function into one.
4647 Severity : enhancement
4649 Description: allow liblustre clients to delegate truncate locking to OST
4650 Details : To avoid overhead of locking, liblustre client instructs OST to
4651 take extent lock in ost_punch() on client's behalf. New connection
4652 flag is added to handle backward compatibility.
4654 Severity : enhancement
4655 Bugzilla : 4928, 7341, 9758
4656 Description: allow number of OST service threads to be specified
4657 Details : a module parameter allows the number of OST service threads
4658 to be specified via "options ost ost_num_threads={N}" in the
4659 OSS's /etc/modules.conf or /etc/modprobe.conf.
4663 Bugzilla : 6146, 9635, 9895
4664 Description: servers crash with bad pointer in target_handle_connect()
4665 Details : In rare cases when a client is reconnecting it was possible that
4666 the connection request was the last reference for that export.
4667 We would temporarily drop the export reference and get a new
4668 one, but this may have been the last reference and the export
4669 was just destroyed. Get new reference before dropping old one.
4671 Severity : enhancement
4672 Frequency : if client is started with failover MDS
4674 Description: Allow multiple MDS hostnames in the mount command
4675 Details : Try to read the configuration from all specified MDS
4676 hostnames during a client mount in case the "primary"
4679 Severity : enhancement
4681 Description: Stop sending data to evicted clients as soon as possible.
4682 Details : Check if the client we are about to send or are sending data to
4683 was evicted already. (Check is done every second of waiting,
4684 for which l_wait_event interface was extended to allow checking
4685 of exit condition at specified intervals).
4688 Frequency : rare, normally only when NFS exporting is done from client
4690 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
4692 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
4695 Frequency : always, if srand() is called before liblustre initialization
4697 Description: Liblustre uses system PRNG disturbing its usage by user application
4698 Details : Introduce internal to lustre fast and high-quality PRNG for
4699 lustre usage and make liblustre and some other places in generic
4700 lustre code to use it.
4702 Severity : enhancement
4703 Bugzilla : 9477, 9557, 9870
4704 Description: Verify that the MDS configuration logs are updated when xml is
4705 Details : Check if the .xml configuration logs are newer than the config
4706 logs stored on the MDS and report an error if this is the case.
4707 Request --write-conf, or allow starting with --old_conf.
4709 Severity : enhancement
4711 Description: Handle symlinks in the path when checking if Lustre is mounted.
4712 Details : Resolve intermediate symlinks when checking if a client has
4713 mounted a filesystem to avoid duplicate client mounts.
4718 Description: lconf can hit an error exception but still return success.
4719 Details : The lconf command catches the Command error exception at the top
4720 level script context and will exit with the associated exit
4721 status, but doesn't ensure that this exit status is non-zero.
4726 Description: failure of ptlrpc thread startup can cause oops
4727 Details : Starting a ptlrpc service thread can fail if there are a large
4728 number of threads or the server memory is very fragmented.
4729 Handle this without oopsing.
4732 Frequency : always, only if liblustre and non-default acceptor port was used
4734 Description: liblustre cannot connect to servers with non-default acceptor port
4735 Details : tcpnal_set_default_params() was not called and was therefore
4736 ignoring the environment varaible TCPNAL_PORT, as well as other
4737 TCPNAL_ environment variables
4742 Description: two objects could be created on the same OST for a single file
4743 Details : If an OST is down, in some cases it was possible to create two
4744 objects on a single OST for a single file. No problems other
4745 than potential performance impact and spurious error messages.
4749 Bugzilla : 5681, 9562
4750 Description: Client may oops in ll_unhash_aliases
4751 Details : Client dcache may become inconsistent in race condition.
4752 In some cases "getcwd" can fail if the current directory is
4758 Description: Inode refcounting problems in NFS export code
4759 Details : link_raw functions used to call d_instantiate without obtaining
4760 extra inode reference first.
4764 Bugzilla : 9942, 9903
4765 Description: Referencing freed requests leading to crash, memleaks with NFS.
4766 Details : We used to require that call to ll_revalidate_it was always
4767 followed by ll_lookup_it. Also with revalidate_special() it is
4768 possible to call ll_revalidate_it() twice for the same dentry
4769 even if first occurence returned success. This fix changes semantic
4770 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
4771 reference on a request referred from the intent.
4772 ll_intent_release() then releases such a request.
4775 Frequency : rare, normally benchmark loads only
4777 Description: unlinked inodes were kept in memory on the client
4778 Details : If a client is repeatedly creating and unlinking files it
4779 can accumulate a lot of stale inodes in the inode slab cache.
4780 If there is no other client load running this can cause the
4781 client node to run out of memory. Instead flush old inodes
4782 from client cache that have the same inode number as a new inode.
4785 Frequency : SLES9 2.6.5 kernel and long filenames only
4786 Bugzilla : 9969, 10379
4787 Description: utime reports stale NFS file handle
4788 Details : SLES9 uses out-of-dentry names in some cases, which confused
4789 the lustre dentry revalidation. Change it to always use the
4793 Frequency : rare, unless heavy write-truncate concurrency is continuous
4794 Bugzilla : 4180, 6984, 7171, 9963, 9331
4795 Description: OST becomes very slow and/or deadlocked during object unlink
4796 Details : filter_destroy() was holding onto the parent directory lock
4797 while truncating+unlinking objects. For very large objects this
4798 may block other threads for a long time and slow overall OST
4799 responsiveness. It may also be possible to get a lock ordering
4800 deadlock in this case, or run out of journal credits because of
4801 the combined truncate+unlink. Solution is to do object truncate
4802 first in one transaction without parent lock, and then do the
4803 final unlink in a new transaction with the parent lock. This
4804 reduces the lock hold time dramatically.
4807 Frequency : rare, 2.4 kernels only
4809 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
4810 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
4811 before unmounting their filesystem in order to avoid deadlock.
4812 The kernel_locked() function in 2.4 kernels only checks whether
4813 the kernel lock is held, not whether it is this process that is
4814 holding it as 2.6 kernels do.
4819 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
4820 Details : The client ptlrpc code may be trying to reconnect to a down
4821 server before a previous connection attempt has timed out.
4822 Increase the reconnect interval to be longer than the connection
4823 timeout interval to avoid sending duplicate connections to
4827 Frequency : echo_client brw_test command
4829 Description: fix echo_client to work with OST preallocated code
4830 Details : OST preallocation code (5137) didn't take echo_client IO path
4831 into account: echo_client calls filter methods outside of any
4832 OST thread and, hence, there is no per-thread preallocated
4833 pages and buffers to use. Solution: hijack pga pages for IO. As
4834 a byproduct, this avoids unnecessary data copying.
4838 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
4839 Description: Client can oops in mdc_commit_close() after open replay
4840 Details : It was possible for the MDS to return an open request with no
4841 transaction number in mds_finish_transno() if the client was
4842 evicted, but without actually returning an error. Clients
4843 would later try to replay that open and may trip an assertion
4844 Simplify the client close codepath, and always return an error
4845 from the MDS in case the open is not successful.
4848 Frequency : rare, 2.6 OSTs only
4850 Description: OST may deadlock under high load on fragmented files
4851 Details : If there was a heavy load and highly-fragmented OST filesystems
4852 it was possible to have all the OST threads deadlock waiting on
4853 allocation of biovecs, because the biovecs were not released
4854 until the entire RPC IO was completed. Instead, release biovecs
4855 as soon as they are complete to ensure forward IO progress.
4857 Severity : enhancement
4859 Description: Support for specifying external journal device at mount
4860 Details : If an OST or MDS device is formatted with an external journal
4861 device, this device major/minor is stored in the ext3 superblock
4862 and may not be valid for failover. Allow detecting and
4863 specifying the external journal at mount time.
4868 Description: Mounting an MDS with pending unlinked files may cause oops
4869 Details : target_finish_recovery() calls mds_postrecov() which returned
4870 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
4871 considers this an error and immediately begins cleaning up the
4872 lov, just after starting the mds_lov process
4874 Severity : enhancement
4876 Description: Implement 'lfs df' to report actual free space on per-OST basis
4877 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
4878 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
4879 usage of objects; '-h' to report in human readable format.
4881 ------------------------------------------------------------------------------
4883 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
4890 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
4891 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
4892 at startup during orphan unlinking. Instead, skip buddy update
4893 before setup, as it will be regenerated after recovery is complete.
4898 Description: If an OST is inactive, its locks might reference stale inodes.
4899 Details : lov_change_cbdata() must iterate over all namespaces, even if
4900 they are inactive to clear inode references from the lock.
4902 Severity : enhancement
4903 Frequency : occasional, if non-standard max_dirty_mb used
4905 Description: Client will block write RPCs if not enough grant
4906 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
4907 then the client will block writes while waiting for another RPC
4908 to complete instead of consuming its dirty limit. With change
4909 we get improved performance when max_dirty_mb is small.
4911 Severity : enhancement
4912 Bugzilla : 3389, 6253
4913 Description: Add support for supplementary groups on the MDS.
4914 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
4915 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
4916 lookups for user supplementary groups into a cache.
4920 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
4921 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
4922 for liblustre O_EXCL is enforced by MDS.
4926 Description: readdir never returns NULL in liblustre.
4927 Details : Corrected llu_iop_getdirentries logic, to return offset of next
4928 dentry in struct dirent.
4932 Frequency : liblustre only, depends on application IO pattern
4933 Description: liblustre clients evicted if not contacting servers
4934 Details : Don't put liblustre clients into the ping_evictor list, so
4935 they will not be evicted by the pinger ever.
4937 Severity : enhancement
4939 Description: Add ability to evict clients by NID from MDS.
4940 Details : By echoing "nid:$NID" string into
4941 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
4942 $NID would be instantly evicted from this MDS and from all active
4943 OSTs connected to it.
4947 Description: Do not query file size twice, somewhat slowing stat(2) calls.
4948 Details : lookup_it_finish() used to query file size from OSTs that was not
4953 Description: service threads change working directory to that of init
4954 Details : Starting lustre service threads may pin the working directory
4955 of the parent thread, making that filesystem busy. Threads
4956 now change to the working directory of init to avoid this.
4960 Frequency : during shutdown only
4961 Description: shutdown with a failed MDS or OST can cause unmount to hang
4962 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
4967 Frequency : 2.6 only
4968 Description: chmod/chown may include an extra supplementary group
4969 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
4970 supplementary group and if none is specified this is used.
4973 Bugzilla : 5479 (6816)
4975 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
4976 Details : If lookup is in progress on a file that is unlinked we might try
4977 to revalidate the inode and fail in revalidate after lookup is
4978 complete and ll_file_open() enqueues the open again but
4979 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
4982 Frequency : always, if lconf --abort_recovery used
4984 Description: lconf --abort_recovery fails with 'Operation not supported'
4985 Details : lconf was attempting to abort recovery on the MDT device and not
4988 ------------------------------------------------------------------------------
4990 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
4995 Frequency : rare (only unsupported configurations with a node running as an
4997 Bugzilla : 6514, 5137
4998 Description: Mounting a Lustre file system on a node running as an OST could
5000 Details : OSTs now preallocates memory needed to write out data at
5001 startup, instead of when needed, to avoid having to
5002 allocate memory in possibly low memory situations.
5003 Specifically, if the file system is mounted on on OST,
5004 memory pressure could force it to try to write out data,
5005 which it needed to allocate memory to do. Due to the low
5006 memory, it would be unable to do so and the node would
5007 become unresponsive.
5009 Severity : enhancement
5011 Description: Addition of lconf --service command line option
5012 Details : lconf now accepts a '--service <arg>' option, which is
5013 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
5015 Severity : enhancement
5017 Description: Failover mode is now the default for OSTs.
5018 Details : By default, OSTs will now run in failover mode. To return to
5019 the old behaviour, add '--failout' to the lmc line for OSTs.
5021 Severity : enhancement
5023 Description: Health checks are now provided for MDS and OSTs
5024 Details : Additional detailed health check information on MSD and OSTs
5025 is now provided through the procfs health_check value.
5028 Frequency : occasional, depends on IO load
5030 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
5031 after numerous create/delete cycles
5032 Details : The ext3 inode allocation policy would not allocate new inodes
5033 very well on the OSTs because there are no new directories
5034 being created. Instead we look for groups with free space if
5035 the parent directories are nearly full.
5040 Description: Network or server problems during mount may cause partially
5041 mounted clients instead of returning an error.
5042 Details : The config llog parsing code may overwrite the error return
5043 code during mount error handling, returning success instead
5049 Description: MDS can fail to allocate large reply buffers
5050 Details : After long uptimes the MDS can fail to allocate large reply
5051 buffers (e.g. zconf client mount config records) due to memory
5052 fragmentation or consumption by the buffer cache. Preallocate
5053 some large reply buffers so that these replies can be sent even
5054 under memory pressure.
5058 Frequency : rare (liblustre)
5059 Description: fsx running with liblustre complained that using truncate() to
5060 extend the file doesn't work. This patch corrects that issue.
5061 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
5062 ATTR_SIZE and lsm use in llu_setattr_raw.
5066 Frequency : rare, only 2.6 kernels
5067 Description: Unusual file access patterns on the MDS may result in inode
5068 data being lost in very rare circumstances.
5069 Details : Bad interaction between the ea-in-inode patch and the "no-read"
5070 code in the 2.6 kernel caused the inode and/or EA data not to
5071 be read from disk, causing single-file corruption.
5075 Frequency : rare, only 2.6 filesystems using extents
5076 Description: Heavy concurrent write and delete load may cause data corruption.
5077 Details : It was possible under high-load situations to have an extent
5078 metadata block in the block device cache from a just-unlinked
5079 file overwrite a newly-allocated data block. We now unmap any
5080 metadata buffers that alias just-allocated data blocks.
5084 Frequency : filesystems with default stripe_count larger than 77
5085 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
5086 Details : lconf specifies an inode size of 4096 bytes when the default
5087 stripe_count is larger than 77. This conflicts with the default
5088 inode density of 1 per 4096 bytes. Allocate smaller inodes in
5089 this case to avoid pinning too much memory for large EAs.
5091 ------------------------------------------------------------------------------
5093 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
5098 Frequency : rare (extremely heavy IO load with hundreds of clients)
5100 Description: Client is evicted, gets IO error writing to file
5101 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
5102 caused clients to be evicted for AST timeouts. The fixes in
5103 bug 5192 mean we no longer need to have such short AST timeouts
5104 so ldlm_timeout has been increased.
5107 Frequency : occasional during --force or --failover shutdown under load
5108 Bugzilla : 5949, 4834
5109 Description: Server oops/LBUG if stopped with --force or --failover under load
5110 Details : a collection of import/export refcount and cleanup ordering
5111 issues fixed for safer force cleanup
5114 Frequency : only filesystems larger than 120 OSTs
5115 Bugzilla : 5990, 6223
5116 Description: lfs getstripe would oops on a very large filesystem
5117 Details : lov_getconfig used kfree on vmalloc'd memory
5120 Frequency : only filesystems exporting via NFS to Solaris 10 clients
5121 Bugzilla : 6242, 6243
5122 Description: reading from files that had been truncated to a non-zero size
5123 but never opened returned no data
5124 Details : ll_file_read() reads zeros from no-object files to EOF
5129 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
5130 Details : There's little harm in aborting MDS/OSS recovery and letting it
5131 try again, so I removed the LASSERT and return an error instead.
5133 Severity : enhancement
5135 Description: New debugging infrastructure for tracking down data corruption
5136 Details : The I/O checksum code was replaced to: (a) control it at runtime,
5137 (b) cover more of the client-side code path, and (c) try to narrow
5138 down where problems occurred
5142 Bugzilla : 3819, 4364, 4397, 6313
5143 Description: Racing close and eviction MDS could cause assertion in mds_close
5144 Details : It was possible to get multiple mfd references during close and
5145 client eviction, leading to one thread referencing a freed mfd.
5147 Severity: : enhancement
5148 Bugzilla : 3262, 6359
5149 Description: Attempts to reconnect to servers are now more aggressive.
5150 Details : This builds on the enhanced upcall-less recovery that was added
5151 in 1.4.2. When trying to reconnect to servers, clients will
5152 now try each server in the failover group every 10 seconds. By
5153 default, clients would previously try one server every 25 seconds.
5158 Description: After recovery, certain operations trigger a failed
5159 assertion on a client.
5160 Details : Failing over an mds, using lconf -d --failover, while a
5161 client was doing a readdir() call would cause the client to
5162 LBUG after recovery completed and the readdir() was resent.
5164 Severity : enhancement
5166 Description: Default groups are now added by lconf
5167 Details : You can now run lconf --group <servicename> without having to
5168 manually add groups with lmc.
5171 Frequency : occasional
5173 Description: Nodes with an elan id of 0 trigger a failed assertion
5176 Frequency : always when accessing e.g. tty/console device nodes
5178 Description: tty and some other devices nodes cannot be used on lustre
5179 Details : file's private_data field is used by device data and lustre
5180 values in there got lost. New field was added to struct file to
5181 store fs-specific private data.
5184 Frequency : when exporting Lustre via NFS
5186 Description: NFSD failed occasionally when looking up a path component
5187 Details : NFSD is looking up ".." which was broken in ext3 directories
5188 that had grown large enough to become hashed.
5191 Frequency : Clusters with multiple interfaces not on the same subnet
5193 Description: Nodes will repeatedly try to reconnect to an interface which it
5194 cannot reach and report an error to the log.
5195 Details : Extra peer list entries will be created by lconf with some peers
5196 unreachable. lconf now validates the peer before adding it.
5199 Frequency : Only if a default stripe is set on the filesystem root.
5201 Description: Setting a default stripe on the filesystem root prevented the
5202 filesystem from being remounted.
5203 Details : The client was sending extra request flags in the root getattr
5204 request and did not allocate a reply buffer for the dir EA.
5207 Frequency : occasional, higher if lots of files are accessed by one client
5208 Bugzilla : 6159, 6097
5209 Description: Client trips assertion regarding lsm mismatch/magic
5210 Details : While revalidating inodes the VFS looks up inodes with ifind()
5211 and in rare cases can find an inode that is being freed.
5212 The ll_test_inode() code will free the lsm during ifind()
5213 when it finds an existing inode and then the VFS later attaches
5214 this free lsm to a new inode.
5218 Bugzilla : 6422, 7030
5219 Description: MDS deadlock between mkdir and client eviction
5220 Details : Creating a new file via mkdir or mknod (starting a transaction
5221 and getting the ns lock) can deadlock with client eviction
5222 (gets ns lock and trying to finish a synchronous transaction).
5225 Frequency : occasional
5226 Description: While starting a server, the fsfilt_ext3 module could not be
5228 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
5229 kernels. Previously, lconf would still use the ext3 name
5230 when trying to load modules. Now, it will correctly use
5231 ext3 on 2.4 and ldiskfs on 2.6.
5233 Severity : enhancement
5234 Description: The default stripe count has been changed to 1
5235 Details : The interpretation of the default stripe count (0, to lfs
5236 or lmc) has been changed to mean striping across a single
5237 OST, rather than all available. For general usage we have
5238 found a stripe count of 1 or 2 works best.
5240 Severity : enhancement
5241 Description: Add support for compiling against Cray portals.
5242 Details : Conditional compiling for some areas that are different
5246 Frequency : occasional
5247 Bugzilla : 6409, 6834
5248 Description: Creating files with an explicit stripe count may lead to
5249 a failed assertion on the MDS
5250 Details : If some OSTs are full or unavailable, creating files may
5251 trigger a failed assertion on the MDS. Now, Lustre will
5252 try to use other servers or return an error to the
5256 Frequency : occasional
5258 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
5259 caused lock timeout during readahead (since 1.4.2).
5260 Details : Processes doing readahead might match a lock that hasn't been
5261 granted yet if there are overlapping and conflicting lock
5262 requests. The readahead process waits on ungranted lock
5263 (original lock is CBPENDING), while OST waits for that process
5264 to cancel CBPENDING read lock and eventually evicts client.
5266 Severity : enhancement
5268 Description: Initial enabling of flock support for clients
5269 Details : Implements fcntl advisory locking and file status functions.
5270 This feature is provided as an optional mount flag (default
5271 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
5272 locking are implemented yet, and those that are are not guaranteed
5273 to be completely correct in production environments.
5274 mount -t lustre -o [flock|noflock] ...
5277 Frequency : occasional
5279 Description: OSTs running 2.4 kernels but with extents enabled might trip an
5280 assertion in the ext3 JBD (journaling) layer.
5281 Details : The b_committed_data struct is protected by the big kernel lock
5282 in 2.4 kernels, serializing journal_commit_transaction() and
5283 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
5284 access to this struct. In 2.6 kernels there is finer grained
5285 locking to improve SMP performance of the JBD layer.
5289 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
5294 Description: merge and cleanup kernel patches.
5297 -----------------------------------------------------------------------------
5299 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
5301 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
5302 versions of Lustre. Please update all servers and clients to
5303 version 1.4.2 or later at the same time. You must also run
5304 "lconf --write-conf {config}.xml" on the MDS while it is stopped
5305 to update the configuration logs.
5307 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
5308 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
5309 - swab llog records of type '0' so we get proper header size/idx (5861)
5310 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
5311 - fix rename of one directory over another leaking an inode (5953)
5312 - avoid SetPageDirty on 2.6 (5981)
5313 - don't re-add just-being-destroyed locks to the waiting list (5653)
5314 - when creating new directories, inherit the parent's custom
5315 striping settings if present parent (3048)
5316 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
5317 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
5318 - don't hold client locks on temporary worklist from l_lru (5666)
5319 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
5320 - automatically evict dead clients (5921)
5321 - Update file size properly in create+truncate+fstat case (6196)
5322 - Do not unhash mountpoint dentries, do not allow removal of
5324 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
5325 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
5326 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
5327 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
5328 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
5329 - Correctly handle reads of files with no objects (6243)
5330 - lctl recover will also mark a device active if deactivate used (5933)
5332 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
5333 - allow --write-conf on an MDS with different nettype than client (5619)
5334 - don't write config llogs to MDS for mounts not from that MDS (5617)
5335 - lconf should create multiple TCP connections from a client (5201)
5336 - init scripts are now turned off by default; run chkconfig --on
5337 lustre and chkconfig --on lustrefs to use them
5338 - upcalls are no longer needed for clients to recover to failover
5340 - add --abort-recovery option to lconf to abort recovery on device
5342 - add support for an arbitrary number of OSTs (3026)
5343 - Quota support protocol changes.
5344 - forward compatibility changes to wire structs (6007)
5345 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5346 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5347 - improved "lustre status" script
5348 - initialize blocksize for non-regular files (6062)
5349 - added --disable-server and --disable-client configure options (5782)
5350 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5351 - Vanilla 2.4.29 support
5352 - increase maximum number of obd devices to 520 (6242)
5353 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5354 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5355 SLES 9 2.6.5 kernels
5356 - limit stripes per file to 160 (the maximum EA size) (6093)
5358 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5361 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5362 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5363 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5364 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5365 - small ext3 extents cleanups and fixes (5733)
5366 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5367 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5368 - avoid needless client->OST connect, fix handle mismatch (5317)
5369 - fix DLM error path that led to out-of-sync client, long delays (5779)
5370 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5371 - fix several locking issues related to i_size (5492,5624,5654,5672)
5372 - don't move pending lock onto export if it is already evicted (5683)
5373 - fix kernel oops when creating .foo in unlinked directory (5548)
5374 - fix deadlock in obdfilter statistics vs. object create (5811)
5375 - use time_{before,after} to avoid timer jiffies wrap (5882)
5376 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5377 - Do not leak request if server was not able to process it (5154)
5378 - If mds_open unable to find parent dir, make that negative lookup(5154)
5379 - don't create new directories with extent-mapping (5909, 5936)
5381 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5382 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5383 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5384 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5385 - support for RHEL4 kernel on i686 (5773)
5386 - provide error messages when incompatible logs are encountered (5898)
5388 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5389 * version 1.4.0.10 (1.4.1 release candidate 1)
5391 - don't keep a lock reference when lock is not granted (4238)
5392 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5393 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5394 - drop import inflight refcount on signal_completed_replay error (5255)
5395 - unlock page after async write error during send (3677)
5396 - handle missing objects in filter_preprw_read properly (5265)
5397 - no transno return for symlink open, don't save no-trasno open (3440)
5398 - don't try to complete elan receive that already failed (4012)
5399 - free RPC server reply state on error (5406)
5400 - clean up thread from ptlrpc_start_thread() on error (5160)
5401 - readahead could read extra page into cache that wasn't ejected (5388)
5402 - prevent races in class_attach/setup/cleanup/detach (5260)
5403 - don't dereference de->d_inode after l_dput of de (5458)
5404 - use "int" for stripe value returned from lock_to_stripe (5544)
5405 - mballoc allocation and error-checking fixes in 2.6 (5504)
5406 - block device patches to fix I/O request sizes in 2.6 (5482)
5407 - look up hostnames for IB nals (5602)
5408 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5409 - don't start multiple acceptors for the same port (5277)
5410 - fix incorrect LASSERT in mds_getattr_name (5635)
5411 - export a proc file for general "ping" checking (5628)
5412 - fix "lfs check" to not block when the MDS is down (5628)
5414 - service request history (4965)
5415 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5416 - create an "evict_client" /proc entry on OSTs, like the MDS has
5417 - fix mount usage message, return errors per mount(8) (5168)
5418 - change grep [] to grep "[]" in tests so they work in more UMLs
5419 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5420 - remove ancient LOV_MAGIC_V0 EA support (5047)
5421 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5422 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5423 - allow readahead allocations to fail when low on memory (5383)
5424 - mmap locking landed again, after considerable improvement (2828)
5425 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
5427 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
5430 - send OST transaction number in read/write reply to free req (4966)
5431 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5432 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5433 - the watchdog thread now runs as interruptible (5246)
5434 - flock/lockf fixes (but it's still disabled, pending 5135)
5435 - don't use EXT3 constants in llite code (5094)
5436 - memory shortage at startup could cause assertion (5176)
5438 - reorganization of lov code
5439 - single portals codebase
5441 - add extents/mballoc support (5025)
5442 - direct I/O reads in the obdfilter (4048)
5443 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
5445 tbd Cluster File Systems, Inc. <info@clusterfs.com>
5448 - send OST transaction number in read/write reply to free req (4966)
5449 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5450 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5451 - the watchdog thread now runs as interruptible (5246)
5452 - handle missing objects in filter_preprw_read properly (5265)
5453 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5454 - drop import inflight refcount on signal_completed_replay error (5255)
5455 - unlock page after async write error during send (3677)
5456 - return original error code on reconstructed replies (3761)
5457 - no transno return for symlink open, don't save no-trasno open (3440)
5459 - add pid to ldlm debugging output (4922)
5460 - bump the watchdog timeouts -- we can't handle 30sec yet
5461 - extra debugging for orphan dentry/inode bug (5259)
5463 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
5466 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
5467 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
5468 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
5469 - fix NULL dereference in /proc/sys/portals/routes (4827)
5470 - allow failed mdc_close() operations to be interrupted (4561)
5471 - stop precreate on OST before MDS would time out on it (4778)
5472 - don't send partial-page writes before EOF from client (4410)
5473 - discard client grant for sub-page writes on large-page clients (4520)
5474 - don't free dentries not owned by NFS code, check generation (4806)
5475 - fix lsm leak if mds_create_objects() fails (4801)
5476 - limit debug_daemon file size, always print CERROR messages (4789)
5477 - use transno after validating reply (3892)
5478 - process timed out requests if import state changes (3754)
5479 - update mtime on OST during writes, return in glimpse (4829)
5480 - add mkfsoptions to LDAP (4679)
5481 - use ->max_readahead method instead of zapping global ra (5039)
5482 - don't interrupt __l_wait_event() during strace
5484 - add software watchdogs to catch hung threads quickly (4941)
5485 - make lustrefs init script start after nfs is mounted
5486 - fix CWARN/ERROR duplication (4930)
5487 - return async write errors to application if possible (2248)
5488 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
5489 - print NAL number in %x format (4645)
5490 - update barely-supported suse-2.4.21-171 series (4842)
5491 - support for sles 9 %post scripts
5492 - support for building 2.6 kernel-source packages
5493 - support for sles km_* packages
5495 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
5498 - ignore -ENOENT errors in osc_destroy (3639)
5499 - notify osc create thread that OSC is being cleaned up (4600)
5500 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
5501 - reconstruct ost_handle() like mds_handle() (4657)
5502 - create a new thread to do import eviction to avoid deadlock (3969)
5503 - let lconf resolve symlinked-to devices (4629)
5504 - don't unlink "objects" from directory with default EA (4554)
5505 - hold socknal file ref over connect in case target is down (4394)
5506 - allow more than 32000 subdirectories in a single directory (3244)
5507 - fix blocks count for O_DIRECT writes (3751)
5508 - OST returns ENOSPC from object create when no space left (4539)
5509 - don't send truncate RPC if file size isn't changing (4410)
5510 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
5511 - bind to privileged port in socknal and tcpnal (3689)
5513 - rate limit CERROR/CWARN console message to avoid overload (4519)
5514 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
5515 - basic mmap support (3918)
5516 - kernel patch series update from b1_4 (4711)
5518 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
5521 - avoid crash during MDS cleanup with OST shut down (2775)
5522 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
5523 - don't use bad inodes on the MDS (3744)
5524 - dynamic object preallocation to improve recovery speed (4236)
5525 - don't hold spinlock over lock dumping or change debug flags (4401)
5526 - don't zero obd_dev when it is force cleaned (3651)
5527 - print grants to console if they go negative (4431)
5528 - "lctl deactivate" will stop automatic recovery attempts (3406)
5529 - look for existing locks in ldlm_handle_enqueue() (3764)
5530 - don't resolve lock handle twice in recovery avoiding race (4401)
5531 - revalidate should check working dir is a directory (4134)
5533 - don't always mark "slow" obdfilter messages as errors (4418)
5535 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
5538 - don't close LustreDB during write_conf until it is done (3860)
5539 - fix typo in lconf for_each_profile (3821)
5540 - allow dumping logs from multiple threads at one time (3820)
5541 - don't allow multiple threads in OSC recovery (3812)
5542 - fix debug_size parameters (3864)
5543 - fix mds_postrecov to initialize import for llog ctxt (3121)
5544 - replace config semaphore with spinlock (3306)
5545 - be sure to send a reply for a CANCEL rpc with bad export (3863)
5546 - don't allow enqueue to complete on a destroyed export (3822)
5547 - down write_lock before checking llog header bitmap (3825)
5548 - recover from lock replay timeout (3764)
5549 - up llog sem before sending rpc (3652)
5550 - reduce ns lock hold times when setting kms (3267)
5551 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
5552 - fix NULL deref and obd_dev leak on setup error (3312)
5553 - replace some LBUG about llog ops with error handling (3841)
5554 - don't match INVALID dentries from d_lookup and spin (3784)
5555 - hold dcache_lock while marking dentries INVALID and hashing (4255)
5556 - fix invalid assertion in ptlrpc_set_wait (3880)
5558 - add libwrap support for the TCP acceptor (3996)
5559 - add /proc/sys/portals/routes for non-root route listing (3994)
5560 - allow setting MDS UUID in .xml (2580)
5561 - print the stack of a process that LBUGs (4228)
5563 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
5566 - don't cleanup request in ll_file_open() on failed MDS open (3430)
5567 - make sure to unset replay flag from failed open requests (3440)
5568 - if default stripe count is 0, use OST count for inode size (3636)
5569 - update parent mtime/ctime on client for create/unlink (2611)
5570 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
5571 - free recovery state on server during a forced cleanup (3571)
5572 - unregister_reply for resent reqs (3063)
5573 - loop back devices mounting and status check on 2.6 (3563)
5574 - fix resource-creation race that can provoke i_size == 0 (3513)
5575 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
5576 - more debugging for page-accounting assertion (3746)
5577 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
5578 - avoid deadlock after precreation failure (3758)
5579 - fix race and lock order deadlock in orphan handling (3450, 3750)
5580 - add validity checks when grabbing inodes from l_ast_data (3599)
5582 - add /proc/.../recovery_status to obdfilter (3428)
5583 - lightweight CDEBUG infrastructure, debug daemon (3668)
5584 - change default OSC RPC parameters to be better on small clusters
5585 - turn off OST read cache for files smaller than 32MB
5586 - install man pages and include them in rpms (3100)
5587 - add new init script for (un)mounting lustre filesystems (2593)
5588 - run chkconfig in %post for init scripts (3701)
5589 - drop scimac NAL (unmaintained)
5591 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
5594 - clean kiobufs before and after use (3485)
5595 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
5596 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
5597 - change init script to fail more clearly if not run as root (1528)
5598 - allow clients to reconnect during replay (1742)
5599 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
5600 - don't do DNS lookups on NIDs too small for IP addresses (3442)
5601 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
5602 - fix cond_resched (3554)
5603 - only evict unfinished clients after recovery (3515)
5604 - allow bulk resend, prevent data loss (3570)
5605 - dynamic ptlrpc request buffer allocation (2102)
5606 - don't allow unlinking open directory if it isn't empty (2904)
5607 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
5608 - remove extraneous obd dereference causing LASSERT failure (3334)
5609 - don't use get_cycles() when creating temp. files on the mds (3156)
5610 - hold i_sem when setting i_size in ll_extent_lock() (3564)
5611 - handle EEXIST for set-stripe, set proper directory name (3336)
5613 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
5614 - fix ksocknal_fmb_callback() error messages (2918)
5616 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
5619 - don't copy lvb into (possibly NULL) reply on error (2983)
5620 - don't deref dentry after dput, don't free lvb on error (2922)
5621 - use the kms to determine writeback rpc length (2947)
5622 - increment oti_logcookies when osc is inactive (2948)
5623 - update client's i_blocks count via lvb messages (2543)
5624 - handle intent open/close of special files properly (1557)
5625 - mount MDS with errors=remount-ro, like obdfilter (2009)
5626 - initialize lock handle to avoid ASSERT on error cleanup (3057)
5627 - don't use cancelling-locks' kms values (2947)
5628 - use highest lock extent for kms, not last one (2925)
5629 - don't dereference ERR_PTR() dentry in error handling path (3107)
5630 - fix thread race in portals_debug_dumplog() (3122)
5631 - create lprocfs device entries at setup instead of at attach (1519)
5632 - common AST error handler, don't evict client on completion race (3145)
5633 - zero nameidata in detach_mnt in 2.6 (3118)
5634 - verify d_inode after revalidate_special is valid in 2.6 (3116)
5635 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
5636 - initialize RPC timeout timer earlier for 2.6 (3219)
5637 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
5638 - print nal/nid for unknown nid (3258)
5639 - additional checks for oscc recovery before doing precreate (3284)
5640 - fix ll_extent_lock() error return code for 64-bit systems (3043)
5641 - don't crash in mdc_close for bad permissions on open (3285)
5642 - zero i_rdev for non-device files (3147)
5643 - clear page->private before handing to FS, better assertion (3119)
5644 - tune the read pipeline (3236)
5645 - fix incorrect decref of invalidated dentry (2350)
5646 - provide read-ahead stats and refine rpc in flight stats (3328)
5647 - don't hold journal transaction open across create RPC (3313)
5648 - update atime on MDS at close time (3265)
5649 - close LDAP connection when recovering to avoid server load (3315)
5650 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
5651 - don't leak open file on MDS after open resend (3325)
5652 - serialize filter_precreate and filter_destroy_precreated (3329)
5653 - loop device shouldn't call sync_dev() for nul device (3092)
5654 - clear page cache after eviction (2766)
5655 - resynchronize MDS->OST in background (2824)
5656 - refuse to mount the same filesystem twice on same mountpoint (3394)
5657 - allow llmount to create routes for mounting behind routers (3320)
5658 - push lock cancellation to blocking thread for glimpse ASTs (3409)
5659 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
5660 - fix rare problem with rename on htree directories (3417)
5662 - allow default OST striping configuration per directory (1414)
5663 - fix compilation for qswnal for 2.6 kernels (3125)
5664 - increase maximum number of MDS request buffers for large systems
5665 - change liblustreapi to be useful for external progs like lfsck (3098)
5666 - increase local configuration timeout for slow disks (3353)
5667 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
5669 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5672 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
5673 - don't overwrite extent policy data in reply if lock was blocked (2901)
5674 - drop filter export grants atomically with removal from device (2663)
5675 - del obd_self_export from work_list in class_disconnect_exports (2908)
5676 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
5677 - swab reply message in mdc_close, other PPC fixes (2464)
5678 - fix destroying of named logs (2325)
5679 - overwrite old logs when running lconf --write_conf (2264)
5680 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
5681 - fix race in target_handle_connect (2898)
5682 - mds_reint_create() should take same inode create lock (2926)
5683 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
5684 - don't close files for self_export to avoid uninitialized obd (2936)
5685 - allow MDS with the same name as client node (2939)
5686 - hold dentry reference for closed log files for unlink (2325)
5687 - reserve space for all logs during transactions (2059)
5688 - don't evict page beyond end of stripe extent (2925)
5689 - don't oops on a deleted current working directory (2399)
5690 - handle hard links to targets without a parent properly (2517)
5691 - don't dereference NULL lock when racing during eviction (2867)
5692 - don't grow lock extents when lots of conflicting locks (2919)
5694 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
5697 - account for cache space usage on clients to avoid data loss (974)
5698 - lfsck support in lustre kernel code (2349)
5699 - reduce journal credits needed for BRW writes (2370)
5700 - orphan handling to avoid losing space on client/server crashes
5701 - ptlrpcd can be blocked, stopping ALL progress (2477)
5702 - use lock value blocks to assist in proper KMS, faster stat (1021)
5703 - takes i_sem instead of DLM locks internally on obdfilter (2720)
5704 - recovery for initial connections (2355)
5705 - fixes for mds_cleanup_orphans (1934)
5706 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
5707 - block all file creations until orphan recovery completes (1901)
5708 - client remove rq_connection from request struct (2423)
5709 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
5710 - recovery timer race (2670)
5711 - mdc_close recovey bug (2532)
5712 - ptlrpc cleanup bug (2710)
5713 - mds timeout on local locks (2588)
5714 - namespace lock held during RPCs (2431)
5715 - handle interrupted sync write properly (2503)
5716 - don't try to handle a message that hasn't been replied to (2699)
5717 - client assert failure during cleanup after abort recovery (2701)
5718 - leak mdc device after failed mount (2712)
5719 - ptlrpc_check_set allows timedout requests to complete (2714)
5720 - wait for inflight reqs when ptlrpcd finishes (2710)
5721 - make sure unregistered services are removed from the srv_list
5722 - reset bulk XID's when resending them (caught by 1138 test)
5723 - unregister_bulk after timeout
5724 - fix lconf error (2694)
5725 - handle write after unfinished setstripe, stripe-only getstripe (2388)
5726 - readahead locks pages, leaves pending causing memory pressure (2673)
5727 - increase OST request buffers to 4096 on large machines (2729)
5728 - fix up permission of existing directories in simple_mkdir (2661)
5729 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
5730 - don't assign transno to errored transactions (2742)
5731 - don't delete objects on OST if given a bogus objid from MDS (2751)
5732 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
5733 - if rq_no_resend, then timeout request after recovery (2432)
5734 - fix MDS llog_logid record size, 64-bit array alignment (2733)
5735 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
5736 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
5737 - MDS recovery shouldn't delete objects that it has given out (2730)
5738 - if enqueue arrives after completion, don't clobber LVB (2819)
5739 - don't unlock pages twice when trigger_group_io returns error (2814)
5740 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
5741 - don't write pages to disk if there was an error (1450)
5742 - don't ping imports that have recovery disabled (2676)
5743 - take buffered bytes into account when balancing socknal conn (2817)
5744 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
5745 - reconnect unlink llog connection after MDS reconnects to OST (2816)
5746 - remove little-endian swabbing of llog records (1987)
5747 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
5748 - retry reposting request buffers if they fail (1191)
5749 - grow extent at grant time to avoid granting a revoked lock (2809)
5750 - lock revoke doesn't evict page if covered by a second lock (2765)
5751 - disable VM readahead to avoid reading outside lock extents (2805)
5753 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
5754 - updated kernel patches for hp-2.4.20 kernel (2681)
5756 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
5759 - fix truncated write corruption (2366)
5760 - fix for failed assertion in iopen_connect_dentry (1792,2517)
5762 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
5763 - ensure len(uuid) < 37 in lmc (1171)
5764 - fix ia64 OOPS in llog_test (2255)
5765 - zero end of page at obdfilter for partial page writes (2648)
5766 - don't leave stale dentries around after renames (bug 2428)
5767 - fix timeouts when evicting a client with a single lock held (2642)
5768 - set deadline for the initial HELLO message to drain (2634)
5769 - print out dotted-quad IP addresses in the socknal (2302)
5771 - additional debugging for MDS client eviction problem (2443)
5772 - fix mkfsoptions support for osts (2603, 2604)
5774 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
5777 - add series for the vanilla 2.6.0 kernel
5778 - add series for the vanilla 2.4.24 kernel
5779 - add series for a cray x86/64 UL kernel drop
5780 - fix xattr patches for the vanilla 2.4.19 series
5782 - generate true UUIDs in lmc (1171)
5783 - have portals stack dumping break in UML (2466)
5784 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
5785 - allocate with _NOFS in ldlm to avoid deadlock (1933)
5786 - wake callback waiting threads on client eviction (2460)
5787 - Add --ptldebug and --subsystem to lmc (1719)
5788 - update assertion to allow safe interrupt allocation
5789 - set rq_no_resend for cancel requests (2432)
5790 - recalculate ptlrpcd timeout after resend (2494)
5791 - call vfs_rmdir when removing pending directories (2368)
5792 - fix renaming a file to itself (2429)
5793 - lmc creates a default one-stripe lov (2454)
5794 - expand procfs space to handle large clusters (2326)
5795 - increase UML stack to avoid overflow
5796 - update lconf's list of debug and subsystem masks
5797 - fix lfs find --obd (2510)
5798 - /proc tunable for disabling filter read caching (2591)
5799 - stop rpm packages from altering slapd.conf (2301)
5800 - disable nagle in the socknal under 0conf (2578)
5801 - choose mds inode size based on stripe count (2572)
5802 - fix kernel-source rpm problems (2516)
5804 - add --disable-doc to avoid pdf generation (2421)
5805 - update documentation, tests, type-os, comments
5806 - avoid format warnings on ia64
5807 - remove the TOE NAL
5808 - tiny code cleanups by removing unused fields
5810 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
5813 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
5814 - remove the most blatant lies from BUILDING file (2371)
5815 - change default debug level to reasonable production setting
5816 - reduce client side cache size to reduce cache flush time
5817 - reduce max RPCs in flight to avoid unnecessary file fragmentation
5818 - make TCP zerocopy and pinger support enabled by default (2476)
5819 - sync writes completed after process exits caused crashes (2319)
5820 - maintain correct mount count on the MDS (2356)
5821 - backout 1557, because 2316 wasn't really fixed
5822 - better file I/O statistics gathering in /proc
5823 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
5824 - another kernel patch to fix zero-copy TCP function export
5825 - don't take duplicate lock when processing re-sent getattr (2420)
5826 - lctl uses obd_self_export instead of creating new conn (2353)
5827 - MDS/OST recovery case which requires object creation asserted (2425)
5828 - move lfs from /usr/sbin to /usr/bin in packages
5829 - fix race between mds_client_add and mds_client_free (2417)
5830 - use kmalloc instead of slabs in portals (2430)
5831 - don't create duplicate records when a failover MDS is present (2442)
5832 - remove unnecessary mount age check (2332)
5833 - don't remove directory inodes from locks prematurely (2451)
5834 - don't break if MDS service name is the same as hostname (2103)
5835 - fix races in client write RPC generation when cache full (2482)
5837 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
5840 - remove now-unused request->rq_obd (278)
5841 - if an allocation fails, print out how much memory we've used (1933)
5842 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
5843 - add the "configurable stack size" patch to most series files (1256)
5844 - ability to write large log records, for 100+ OST configs (2306)
5845 - fix NULL deref when filter_prep fails (2314)
5846 - fix operator precedence error in filter_sync
5847 - dynamic allocation of socknal TX descriptors (2315)
5848 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
5849 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
5850 - ensure that CWARN messages in Portals always get to the syslog
5851 - __init/__exit are not for prototype decls (ldlm_init/exit)
5852 - x86-64 compile warning fixes
5853 - fix gateway LMC keyword conflict (2318)
5854 - fix MDS lock inversions in getattr/reint paths (1844)
5855 - fix a rare lock re-ordering bug, which caused deadlock (2322)
5856 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
5857 - DLM race condition prevented some lock evictions (2328)
5858 - ENOMEM detection and retry on socknal sends (2230)
5859 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
5860 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
5861 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
5862 - better error messages when a client is rejected during recovery (1505)
5863 - avoid cancelling locks which were never granted, after failure (2330)
5864 - fix i_sem/journal inversion in mds_client_add (2333)
5865 - fix truncate/getattr lock cycle deadlock (2334)
5866 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
5867 - fix two rare exit paths which could leak an l_lock() ref (2321)
5868 - fencepost error in MDS/OST orphan recovery (2226)
5869 - make log record alignment 8 bytes (1988)
5870 - lstripe now fails when requested offset > ost_count (2237)
5871 - ensure that all kernel series have a complete list.h (1607)
5872 - fix crashes in special-file operations (2316)
5873 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
5875 - allow configurable automake binary, for testing new versions
5876 - small update to the lfs documentation
5878 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
5880 * fix negative export reference count in fsfilt_sync (2312)
5882 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
5883 * release candidate 0.9.1
5885 - orphans are moved into the PENDING directory for possible recovery
5886 - replayed opens now open by fid for orphan/rename safety (1042)
5887 - last close of an orphan inode generates a transno (683)
5888 - chdir() and mount() now pin the directory entry (1020)
5889 - avoid CERROR in normal ll_setattr_raw() error case (1500)
5890 - discard very old requests without processing them (1502)
5891 - remove some common, well-understood CERRORs (1505)
5892 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
5893 - clear "grant" flags in OST replies until OST grant code lands (1644)
5894 - fix read performance by not clobbering i_blksize on client (1598)
5895 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
5896 - make LRU size a /proc tunable, clears locks when reduced (707)
5897 - fix some lprocfs rot that prevented ptlbd from loading (1732)
5898 - server locks take references on exports now (1558)
5899 - build fixes for 2.4.20-rh trees (1663)
5900 - return an error from lov_create if all OSCs are inactive (1751)
5901 - fix import levels when a reconnect happens without a timeout (1597)
5902 - exit early from mds_open if we get a lookup error (1749)
5903 - partial page read at EOF wouldn't wait for disk before sending (1642)
5904 - avoid NULL deref in obdfilter when reading page past EOF (1592)
5905 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
5906 - fix LBUG in ll_it_open_error with rc = -2 (1861)
5907 - write/truncate lock inversion (1639)
5908 - Don't auto-load obdclass, portals modules during cleanup (1495)
5909 - fix timestamps from jumping to "now" (1763)
5910 - extra journal assertions (1648)
5911 - add an extra multiunlink test (1771)
5912 - fix read_record/write_record API (1776)
5913 - fix leak of offset_extent, possible incorrect i_size later (1772)
5914 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
5915 - Debugging for the kqswnal_get_idle_tx problems (1820)
5916 - Allow recovery to be attempted multiple times (1536)
5917 - Write out MDS last_rcvd file after it is first created (1600)
5918 - Fix tx_descriptor leak in failed transmit situations (1827)
5919 - ext3 journaling fixes for assertion failure after IO error (1871)
5920 - class_export_put() on freed export after completion AST error (1896)
5921 - Fix revalidate looping in VFS (1322)
5922 - Don't access a freed export during MDS_REINT timeout (1521)
5923 - Add open-unlink recovery support on the MDS (1673,1764)
5924 - Return an error if no MDS data was read from last_rcvd (1946)
5925 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
5926 - Don't setup a disk device that doesn't match exported UUID (317)
5927 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
5928 - avoid committing NULL handle in force close
5929 - local.sh is now a one-stripe LOV configuration
5930 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
5931 - fix ext3 htree duplicate directory entry corruption (1516)
5932 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
5933 - update to new LOV EA format (2097)
5934 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
5936 2003-06-15 Phil Schwan <phil@clusterfs.com>
5939 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
5940 - per-import/export recovery handling (958, 931, 959)
5941 - multiple last-rcvd slots, for serving multiple FSes (949)
5942 - connections are again shared between multiple imp/exports (963, 964)
5943 - "umount -f" would hang if any requests needed to be sent (393, 978)
5944 - avoid pinning large req buffer by copying for queued messages (989)
5945 - add "uuid" to "lctl device" command to help upcalls (991)
5946 - "open" RPCs with transnos would confuse recovery counters (1037)
5947 - do proper endian conversion of all wire messages (288, 340, 891)
5948 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
5949 - call waiting locks callback from LDLM recovery thread (1127, 1151)
5950 - fix ptlrpc_connection leak in target_handle_connect (1174)
5951 - fix import refcounting bug in OST and MDS cleanup (1134)
5952 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
5953 - fix very unlikely obd_types race condition (501)
5954 - remove osc_open hack for echo_client (1187)
5955 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
5956 - a failure in read_inode2 leads to deadlock (1139)
5957 - cancel ack-locks as soon as transaction is committed (1072)
5958 - fix major leaks and crashes in the bulk I/O path (937, 1057)
5959 - make sure to commitrw after any preprw to avoid deadlock (1162)
5960 - failing to execute a file in a lustre FS would lock inode (1203)
5961 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
5962 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
5963 - obd_uuid2tgt would walk off the end of the list (1255)
5964 - on IA64 the osc would give portals incorrect bulk size (1258)
5965 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
5966 - fix lock inversion caused by new llite matching code (1282)
5967 - limit the number of dirty pages on a client to 10MB (1286)
5968 - timed out locks were not being corrected cancelled (1289)
5969 - fix O_DIRECT above 4GB on IA-32 (1292)
5970 * major user-visible changes
5971 - fail out/fail over policy now controlled by the upcall (993)
5973 - add OBD_PING to check server availability and failure (954)
5974 - lustre messages are now sent in sending host order (288, 340, 891)
5975 - add eadatalen to MDS getattr reply (340)
5976 - OST read replies may contain second buffer, with per-page status (593)
5978 2003-03-11 Phil Schwan <phil@clusterfs.com>
5981 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
5982 - failed open()s could cause deadlock; fixed (867, 869)
5983 - stop cancelling OST locks when files are closed (481)
5984 - overlapping XID spaces caused network corruption (851, 853)
5985 - fix unsafe fsfilt counter arithmetic; change to atomic_t
5986 - setattr_raw added, to do single-RPC, server-side setattrs
5987 - lmc/lconf syntax change for OST UUIDs
5988 - fix crashy race condition between ptlrpc_free_req and osc_close
5989 - don't use request in mdc_enqueue if we hit a timeout (889)
5990 - don't set the inode i_size for regular files from the MDS (896)
5991 - handle out of order completion AST (842)
5992 - don't LBUG if a lock request times out after receiving AST (913)
5993 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
5994 - if a bad lock AST arrives, send an error instead of dropping entirely
5995 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
5996 - fix leak in bulk IO when only partially completed (899, 900, 926)
5997 - fix O_DIRECT for ia64 (55)
5998 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
5999 - C-z after timeout could hang a process forever; fixed (977)
6001 - client-side I/O cache (678, 924, 929, 941, 970)
6003 - READPAGE and SETATTRs which don't take server-side locks get
6006 2003-02-11 Phil Schwan <phil@clusterfs.com>
6009 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
6010 - Fix fsfilt_extN_readpage() to read a full page of directory
6011 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
6012 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
6013 - Fix imbalanced LOV object allocation and out-of-bound access (469)
6014 - Most intent operations were removed, in favour of a new RPC mode
6015 that does a single RPC to the server and bypasses most of the VFS
6016 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
6017 - Aggressively cancel local locks on DLM servers
6018 - mds_reint_unlink sends EA to the client if it's the last nlink.
6019 client uses that EA to unlink OST objects.
6020 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
6021 - recursive symlinks were fixed (439)
6022 - fixed NULL deref in DEBUG_REQ
6023 - filter_update_lastobjid no longer calls sync, which annoyed extN
6024 - fixed multi-client small-writes to a single file problem (445)
6025 - fixed mtime updates during file writes (607)
6026 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
6027 - fixed bug in obd_brw_read/write() (under guise of testing 367)
6028 - fixed Linux OST size reporting problem (444, 656)
6029 - OST now updates object mtime with writes or setattr (607, 619)
6030 - client verifies file size before zeroing page past EOF (445)
6031 - OST now writes last allocated objid to disk with allocation (108)
6032 - LOV on echo now works (409)
6034 - mds_reint_unlink sends a new buffer, with the EA included. this
6035 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
6036 set if a regular file was being unlinked, and it was the last link
6037 - use PtlGet from the target for bulk writes (315)
6038 - OST now updates object mtime with writes or setattr (607, 619)
6039 - LDLM now has a grant-time callback to revalidate locked items, if
6041 - Many MDS operations were reorganized to combat race conditions
6043 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
6044 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
6046 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
6049 - Fully reactivate OST imports after reconnection (512, others)
6050 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
6051 - More graceful error handling for truncating on dead OST (515)
6052 - Don't error out unless we're actually accessing dead stripes (474)
6053 - Fix garbage sizes when stripes are missing (410)
6054 - LRU counters were broken, causing constant lock purge (433, 432)
6055 - garbage on read from stripes with failed OSTs (441)
6056 - mark OSCs as active before reconnecting during recovery (438)
6057 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
6058 - lfind did not preserve OST order in output (443)
6059 - symlinks cause hung clients, incorrect data (439)
6060 - stop dereferencing request after dropping refcount (457)
6061 - don't LASSERT(spin_is_locked) on non-SMP (455)
6062 - fixes for many rename() bugs
6063 - fstat didn't correctly synchronize attributes (399)
6064 - server must handle lock cancellation during blocking AST prep (487)
6065 - bulk descriptors were free()d too soon (511)
6066 - fix paths in lconf, which would load incorrect modules (451, 507)
6067 - fix confusing lconf 'host not found' error message (386)
6068 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
6069 - fix race condition in mdc_blocking_ast() for inode access (526)
6070 - fix lov_unpackmd() unpacking wrong number of stripes (537)
6071 - fix lov_set_osc_active() marking wrong OSC inactive (440)
6072 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
6073 - fix multiple writes of stripe MD to MDS (358, maybe 519)
6074 - fix lstripe in several ways (kernel side) (527)
6075 - fix request leak in ldlm_cli_enqueue (262)
6076 - incorrect OSC was marked inactive after OST failure
6077 - call mds_fs_cleanup before unmounting filesystem (524)
6078 - fix races between taking ns_lock and ldlm_lock_change_resource
6079 - fix races updating LOV export open file list
6080 - fix lov_enqueue error path, avoid decref-ing bad lock handle
6081 - fix recovery NULL deref in ldlm_cli_cancel_unused
6082 - fix some DLM races by using new hash table for lock handles (419)
6083 - permit the client to specify desired inodes, at replay
6084 - duplicate requests when we queue them for replay reintegration
6085 - fix last_rcvd offset calculation
6086 - sync after each recovered transaction, so we always make progress
6087 - never, not always, ERESTART requests without transnos
6088 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
6090 - skip replay if the MDS says that the client is already connected
6091 - don't check for a recovery-enabled export to match lctl's UUID
6092 - don't INC_USE_COUNT for phantom exports
6093 - don't crash when cleaning up phantom exports (567)
6094 - don't double-finish or set replay data for errored mdc_open requests
6095 - abort requests when they time out, so we don't get old replies
6096 - send/receive replies for AST messages again
6097 - if the client says that it doesn't have the lock, cancel it on the
6099 - if we timeout during I/O, don't try to cancel an in-use lock; instead
6100 mark it as destroyed, it will all work out when decref is called
6101 - fix module use counts (22, 581)
6103 - ASTs now expect a reply (server cancels lock on error reply)
6105 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
6108 - fix many simultaneous client startup (392)
6109 - fix dentry->d_it clobbering
6110 - credentials weren't being shipped for readdir/getattr operations
6111 - remove invalid assertions triggered during some concurrent MD
6113 - proper Lustre versions added (336, 389)
6114 - fix memory leak for create error case (398)
6115 - fix LOV locking bug that would get cli/srv out of sync
6116 - fix echo client over LOV (409)
6117 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
6118 - fix double-O_EXCL intent crash (424)
6119 - avoid sending multiple lock CANCELs (352)
6121 - MDS can do multi-client recovery (modulo bugs in new code)
6123 - many updates, edits, cleanups
6125 2002-11-18 Phil Schwan <phil@clusterfs.com>
6128 - fix null d_it dereference (346)
6129 - fix full OST/dbench hang (333)
6130 - fix permission problem with file removal (286)
6131 - fix removal of OSCs from LOV when they fail
6132 - fix NULL deref during bulk timeout (214)
6133 - fix problems related to multiple filesystems on one MDS (241)
6134 - fixed serious subtle metadata locking bugs
6135 - free locks on clients when inodes are removed due to memory
6137 - fix inode pointer in lock data (285)
6138 - partial support for multiple MDS on a single host (241)
6139 - data locks weren't cancelled at clear_inode time (290, 311)
6140 - intent locks could lead to unbounded lock growth (205)
6141 - added a maximum lock count, an LRU list, and a flusher
6142 - fix multiple rename (365)
6143 - properly abstracted the echo client
6144 - OSC locked 1 byte too many; fixed
6145 - rewrote brw callback code:
6146 - fixed recovery bugs related to LOVs (306)
6147 - fixed too-many-pages-in-one-write crash (191)
6148 - fixed (again) crash in sync_io_timeout (214)
6149 - probably fixed callback-related race (385)
6151 - Add capability to MDS protocol
6152 - LDLM cancellations and callbacks on different portals
6154 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
6157 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
6158 - defer open object destruction to close time (601981, 138)
6159 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
6160 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
6161 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
6162 - don't call obd_setattr in ll_file_release for destroyed objects
6163 * protocol change to lustre_msg: move |version| and add |flags|
6164 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
6165 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
6166 * added replay of create, unlink, link and rename operations during
6167 MDS failover; recovery should be much more robust now
6168 * remove failed OSCs from LOVs (only lov_create uses this so far)
6169 * the lustre-HOWTO was brought (more) up to date (582544)
6171 2002-10-23 Phil Schwan <phil@clusterfs.com>
6174 - in-use dentries weren't being reused properly (617851)
6175 - prevent multiple LDLM setup (599178)
6176 - fix LOV size calculations for truncate (617853)
6177 - fix client handling of MDS intent errors (POSIX)
6178 - fix permission bug in lovstripe.c test (624321)
6179 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
6180 - truncate past end of file could corrupt data
6181 - proper cleanup after timeouts, crashes, etc (592524, 550815)
6182 - a race in recovery could return ETIMEDOUT to apps (623947)
6183 - building outside the source directory was fixed
6184 * the lustre-HOWTO was brought (more) up to date (582544)
6185 * major progress was made on recovery functionality
6187 2002-10-10 Phil Schwan <phil@clusterfs.com>
6190 - recovery deadlock fix
6191 - rm -rf causes LBUG fix (617817)
6192 - file open by multiple tasks fix (618962)
6193 - directory permissions bugs (602707 and 620007)
6194 - journal_stop fixed with locking (611313)
6195 - O_APPEND failures resolved (618273, perhaps 614459)
6196 - lconf PATH fix (619770)
6197 - IA64 build fix (621450)
6198 - RPC buffer sizes scale with amount of memory
6200 2002-10-01 Phil Schwan <phil@clusterfs.com>
6203 - locks would be cancelled without throwing away data pages,
6204 resulting in inconsistent data (605627)
6205 - inode attributes were not always being refreshed (605627, 612449)
6206 - lconf now continues to cleanup after lctl reports an error
6207 - MDS now enforces user permissions (602707)
6208 - lprocfs cleanup fixed, but not yet enabled (614157)
6209 - fixed infinite server hang, should a client not respond to an AST
6210 - avoid going into recovery if user calls readlink() with a buffer
6211 that's too small (613941)
6212 - AST RPCs no longer require replies (614867) -- this may be changed
6213 - don't crash server if client sends an IOV that's too big (611336)
6214 - fixed lock conversion deadlock (611892)
6215 - fixed the following of symlinks (614622)
6216 * recovery: the server can remove locks from a client that dies, other
6217 clients can make progress
6218 * more extN patch fixes
6219 * compile-time configurable ptlrpc buffer allocations
6221 - collaborative read cache document
6222 - Lustre Lite Performance CDR document-in-progress
6224 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6227 - fix typo in patch-2.4.18
6229 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6232 - clear ptlrpc request each time in handle_incoming_request()
6233 - unlink of files now destroys the object on the OST
6235 2002-09-19 Peter Braam <braam@clusterfs.com>
6237 * add hard link support
6238 * change obdfile creation method
6239 * kernel patch changed
6241 2002-09-19 Peter Braam <braam@clusterfs.com>
6244 - stack overflow bug in extN fixed
6246 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
6248 * documentation updates
6249 - add man pages for config tools
6250 - update tests/README to describe testing with new config tools
6251 - finish metadata API descriptions
6252 * bug fixes and cleanups
6253 - statfs workaround for 16TB limit
6254 - LOV stripe allocation improved, can stripe on subset of OSTs
6255 - LOV file size/IO offset was wrong for files > 4GB in size
6256 - object EA data was being dropped, caused files to be unreadable
6257 - memory overflow with non-LOV OST caused memory corruption
6258 - fixed regression tests to work with new config tools, obdfilter
6259 - fixed bug when directory size became larger than 1 block
6260 - fixed bug (for single client case) when PWD was deleted
6261 - invalidate local directory pages when doing intent-based ops
6262 - avoid LDLM oops when lock callback contained bad data
6264 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
6266 * documentation updates
6267 * bug fixes and cleanups
6268 - configuration tools
6271 - 64-bit compile warnings
6272 - 64-bit internal statfs data
6274 * test_brw on persistent OST devices
6276 * lprocfs (disabled)
6278 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
6280 * documentation updates
6281 * bug fixes and cleanups
6282 * configuration tools
6284 2002-08-30 Peter J. Braam <braam@clusterfs.com>
6287 * many small fixes to 0_5_4
6288 * io/network handling
6289 * thinkos in MDS operations
6291 2002-08-24 Peter J. Braam <braam@clusterfs.com>
6294 * crucial basic fixes to 0.5.3
6295 * IOR, Iozone work over Elan
6298 2002-08-07 Phil Schwan <phil@clusterfs.com>
6299 * version 0_5_3, our first alpha
6300 * we use the new Portals iovs
6301 * documentation updates
6302 * bug fixes and cleanups
6303 * small changes in the DLM wire protocol
6305 2002-07-25 Peter J. Braam <braam@clusterfs.com>
6306 * version 0_5_1 with some initial stability,
6307 * locking on MD and file I/O.
6308 * documentation updates
6309 * several bug fixes since 0.5.0
6310 * small changes in wire protocol
6312 2002-07-18 Phil Schwan <phil@clusterfs.com>
6314 * delivered as Lustre Light Alpha
6315 * fixed a crash after handling invalid MDS requests
6316 * fixed directory pages for architectures with non-4k pages sizes
6318 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
6319 * release version v0_4_4
6320 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
6322 * Fixes a number of interruption problems with OST operations.
6323 * Update documentation for portals header changes
6324 * Move all wire protocol structs/defines to lustre_idl.h
6325 * Fixes symlink length bug.
6326 * Add tcpdump to repository.
6328 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
6329 * release version v0_4_3
6330 * Fixes statfs for inodes on extN.
6331 * Fixes bug in runtests which would delete /etc/hosts.
6332 * Use 64-bit object IDs wherever possible (not into VFS though)
6333 Remove ost_get_info, which is unused by lustre, and out of date.
6335 2002-07-03 Peter Braam <braam@clusterfs.com>
6336 * release version v0_4_2 Fixes a lookup error (type not passed)
6337 * move forward to head of Portals
6338 * move forward to latest Lustre kernel
6340 2002-06-25 Peter Braam <braam@clusterfs.com>
6341 * release version v0_4_1. Hopefully stable on single node use.