3 * Server support for kernels:
4 2.6.18-238.12.1.el5 (RHEL5)
5 2.6.32-131.2.1.el6 (RHEL6)
6 * Client support for unpatched kernels:
7 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
8 2.6.18-238.12.1.el5 (RHEL5)
9 2.6.32-131.2.1.el6 (RHEL6)
10 2.6.29.4-167.fc11 (FC11)
11 2.6.16.60-0.69.1 (SLES10)
13 * Recommended e2fsprogs version: 1.41.10-sun2
14 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
15 Lustre filesystem with 4K stack may cause a stack overflow. For more
16 information, please refer to bugzilla 17630.
18 Severity : enhancement
19 Description: Update RHEL5.6 kernel to 2.6.18-238.9.1.el5
22 Frequency : only with SLES10
24 Description: Use OFED "KMP" provided by Novell
25 Details : SLES10 SP3 ships with OFED in a separate "KMP" package.
26 Lustre is now built against this package. That means you need to
27 install the ofed-kmp package from Novell for the patchless client.
28 Note that the ofed-kmp that Novell ships may not exactly match the
29 kernel version but should still be compatible.
31 Severity : enhancement
33 Description: Update RHEL5.5 kernel to 2.6.18-194.17.1.el5
34 Update OEL5.5 kernel to 2.6.18-194.17.1.0.1.el5.
36 --------------------------------------------------------------------------------
38 2010-07-15 Oracle, Inc.
40 * Server support for kernels:
41 2.6.18-164.11.1.el5 (RHEL 5)
42 2.6.18-164.11.1.0.1.el5 (OEL 5)
43 * Client support for unpatched kernels:
44 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
45 2.6.18-164.11.1.el5 (RHEL 5),
46 2.6.18-164.11.1.0.1.el5 (OEL 5)
47 2.6.16.60-0.42.8 (SLES 10),
49 2.6.29.4-167.fc11 (FC11)
50 * Recommended e2fsprogs version: 1.41.10-sun2
52 Severity : enhancement
54 Description: Distribute lfs_migrate script with lustre
55 Details : The lfs_migrate mentioned in the manual is now included in the
56 lustre tree in lustre/scripts and will distributed in lustre RPMs.
58 Severity : enhancement
60 Description: Failover nid list added to import proc file.
62 Severity : enhancement
64 Description: "som_preview" mount option required to use SOM
65 Details : The clients will need to use som_preview mount option
69 Description: Lustre does not do 1MB IOs to HW RAID
70 Details : Bump MAX_PHYS/HW_SEGMENTS and SG_ALL to 256 in the RHEL5 kernel.
72 Severity : enhancement
74 Description: lfs quota output is not very convenient for awk/sed-parsing
75 Details : Some positions in lfs quota output table could be empty or
76 non-empty which made it hard to parse it with scripts, now
77 a dash is put instead of space where there is not supposed
82 Description: Update to OFED 1.5.1
83 Details : OFED 1.5.1 will be used in environments that don't use
88 Description: Remove lnet. prefix from lctl params display
89 Details : Also renamed lnet.memused to lnet_memused to
94 Description: Use conf_param -d to remove permanant settings.
98 Description: Add list_param -R
99 Details : lctl list_param now supports recursive mode.
102 Bugzilla : 21829, 22312
103 Description: Fix llobdstat.
104 Details : 1. Now updating the write_bytes counter
105 2. Port script fixes from 1.8.
109 Description: decrease the usage of memory on clients.
110 Details : 1. On clients, recycle dentries and inodes unused.
111 2. Delete the code related to ll_deathrow(att 6215 in bug 1443). It
116 Description: llobdstat fix and enhancement
117 Details : add a counter to set a limit to how many samples will be returned
118 fix a wildcard in the path to limit to obdfilter stats only
122 Description: lfs check only allowed for root
123 Details : Code cleanups and added a regresion test of non-root use of
128 Description: MDS Oopses in ptlrpc_dispatch_difficult_reply()/__list_add()
129 Details : Fix hr->hr_index overload.
133 Description: Fix FIEMAP functionality.
137 Description: FC11 patchless client support added
141 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and
142 OEL5.4 kernel to 2.6.18-164.11.1.0.1.el5.
146 Description: New acc-sm script and various test-framework enhancments.
150 Description: Previously landed patch was reverted.
154 Description: Update SLES10 kernel to 2.6.16.60-0.42.8 (patchless)
158 Description: Add ldiskfs maxdirsize mount option.
162 Description: Fix mkfs.lustre support for 16TB LUNs.
166 Description: Print mmp_check_interval
167 Details : print mmp_check_interval and make it possible to abort
168 mount operation in case it takes too long.
172 Description: Add cascading_rw to lustre mpi tests.
176 Description: Store SOM data in EA.
177 Details : Also several SOM related cleanups.
181 Description: Add refcount to md_open_data
185 Description: Use refcount on lov_request_set to prevent memory corruption.
189 Description: Statahead: use parent dir's i_mutex for synchronization.
193 Description: loadgen improvements.
197 Description: Enable peer health detection in ksocklnd and ko2iblnd.
201 Description: Control DCACHE_LUSTRE_INVALID flag with MDS_INODE_LOOKCUP lock
205 Description: Fix race on obd_devcie:md_stats
209 Description: Fix SLES 11 client builds.
210 Details : Resolved uninitalized pointer gcc warnings.
214 Description: Allow non-root access for "lfs check".
215 Details : Added a check in obd_class_ioctl() for OBD_IOC_PING_TARGET.
219 Description: Clear imp_force_reconnect correctly in ptlrpc_connect_interpret()
221 Severity : enhancement
223 Description: Update kernel to RHEL5.4 2.6.18-164.6.1.el5 and
224 OEL5 2.6.18-164.6.1.0.1.el5(Both in-kernel OFED enabled).
228 Description: Sanity test_27p fails.
229 Details : OBD_FAIL_OST_ENOSPC has been moved to filter_statfs().
230 Sanity tests 27 were cleaned up and improved.
234 Description: Unify req format on client/servers
235 Details : Use new req_capsule API [almost] everywhere instead of old PTLRPC
236 buffers and swabbers approach..
241 Description: drop join file stuff.
243 Severity : enhancement
245 Description: Add 2.6.27 and SLES11 (patchless client) support
249 Description: Fix in ptlrpc_expire_one_request() to print the signed time
252 Severity : enhancement
254 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
257 Severity : enhancement
259 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
261 Severity : enhancement
263 Description: correctly handle big reply message.
264 Details : send LNet event if reply is bigger then buffer and adjust this buffer
269 Description: Drop unnecessary __GFP_NOMEMALLOC flag from filter_get_page()
271 Severity : enhancement
273 Description: Add support for OEL5.
277 Description: Conf-sanity.sh 50g test - deactivated OST should not cause a panic.
279 Severity : enhancement
281 Description: "weak-modules" support
282 Details : Implement "weak-modules" support which enables kernel modules
283 to be used with any kernel that implements the same kABI. In
284 order to achieve this modules are now installed in
285 /lib/modules/$(uname -r)/updates/kernel on all distributions.
289 Description: client could not reconnect to OST because of an active request.
290 Details : The client now retries to reconnect to the same server, if a
291 connect request failed with EBUSY or -EAGAIN. Test case 35b
292 conf-sanity.sh has been created.
296 Description: Don't allow make backward step on assignin osc next id.
297 Details : race between allocation next id and ll_sync thread can be cause
298 of set wrong osc next id and can be kill valid ost objects.
302 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
306 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
310 Description: Error handling in osc_statfs_interpret() has been improved.
311 Details : Check in osc_statfs_interpret() for EBADR.
315 Description: Do not send statfs() requests to OSTs disabled by administrator.
316 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
318 Severity : enhancement
320 Description: Added sanity-quota test to for lfs(1) to work with numeric uid/gid.
322 Severity : enhancement
324 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
326 Severity : enhancement
328 Description: Enhancement for lfs(1) command to use numeric uid/gid.
332 Description: Improved cleanup procedure for conf-sanity test_46a.
336 Description: Do not start bulk transfer, if client tried to reconnect.
337 Details : Do not start bulk transfer, if client tried to reconnect,
338 but failed due to active rpcs. Check for exp_abort_active_req
339 has been implemented.
343 Description: Prevent inconsistences between linux and lustre mount structures.
344 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
345 Make the sleep interruptible.
347 Severity : enhancement
349 Description: provide server to client comms path
350 Details : server to client communications path via new LDLM_SET_INFO rpc
354 Description: truncate starts GFP_FS allocation under transaction causing deadlock
355 Details : ldiskfs_truncate calls grab_cache_page which may start page
356 allocation under an open transaction. This may lead to
357 calling prune_icache with consequent lustre reentrance.
361 Description: send correctly lsm on open replay
362 Details : MDS is trust to LSM size on replay open, but client can set wrong size
367 Description: enable client interface failover
368 Details : When a child reconnects from another NID, properly update export
369 nid hash position and ldlm reverse import.
371 Severity : enhancement
373 Description: Communicate OST degraded/readonly state via statfs to MDS
374 Details : Flags in the statfs returned from OSTs indicate whether the
375 OST is in a degraded RAID state, or if the filesystem has
376 turned read-only after a filesystem error is detected.
381 Description: don't panic if EPROTO was hit when reading symlink
382 Details : correctly handling request reference in error cases.
384 Severity : enhancement
386 Description: Add LustreNetLink, a kernel-userspace communcation path.
388 Severity : enhancement
390 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
395 Description: access to llog context before init.
396 Details : move handling CATALOGS file at osc layer and forbid access to llog
400 Frequency : rare, only when client is writing to OST on local node
402 Description: Avoid deadlock for local client writes
403 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
404 memory freeing context. This allows OST threads to set the
405 PF_MEMALLOC flag on task structures in order to allocate memory
406 from reserved pools and complete IO.
407 Use GFP_HIGHUSER for OST allocations for non-local client writes,
408 so that the OST threads generate memory pressure and allow
409 inactive pages to be reclaimed.
411 Severity : enhancement
413 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
418 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
419 Details : move ldlm namespace creation in setup phase to avoid grab
420 _lprocfs_lock with cli_sem held.
424 Description: Temporarily disable grant shrink.
425 Details : Disable the feature for debugging.
428 Frequency : only when formatting test filesystems
430 Description: Unable to run several mkfs.lustre on loop devices at the same time
431 Details : mkfs.lustre returns error 256 on the concurrent loop devices
432 formatting. The solution is to proper handle the error.
434 Severity : enhancement
436 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
438 Severity : enhancement
440 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
445 Description: resolve race between obd_disconnect and class_disconnect_exports
446 Details : if obd_disconnect will be called to already disconnected export he
447 forget release one reference and osc module can't unloaded.
449 Severity : enhancement
451 Description: Allow tuning service thread via /proc
452 Details : For each service a new
453 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
454 created that can be used to set min/max thread counts, and get the
455 current number of running threads.
460 Descriptoin: don't return error if have partially created objects for file.
461 Details : lov_update_create_set uses set->set_success as index for created
462 objects, so if some requests will be failed, they will have hole
463 at end of array and qos_shrink_lsm for allocate correct lsm.
465 Severity : enhancement
467 Description: Update OFED support to 1.4.1
472 Description: Reduce small size read RPC
473 Details : Set read-ahead limit for every file and only do read-ahead when
474 available read-ahead pages are bigger than 1M to avoid small size
477 Severity : enhancement
479 Description: move AT tunable parameters for more consistent usage
480 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
482 Severity : enhancement
484 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
485 Details : allow skip disconnected ost for send statfs request and hide error
491 Description: fix racy locking of mballoc block bitmaps causing BUG
492 Details : The locking of the mballoc buddy bitmap and the in-memory
493 block bitmap was using two different spin locks in some
494 cases. This made it possible to incorrectly access the
495 mballoc bitmap while another process was modifying it,
496 causing a sanity assertion to fail. While no on-disk corruption
497 was reported, there was some risk of this happening.
500 Frequency : rare, on llog test 6
502 Descriptoin: don't allow connect to already connected import
503 Details : allowing connect to already connected import is hide connecting problem.
508 Descriptoin: allow kill process which wait statahead result
509 Details : for some reasons 'ls' can stick in waiting result from statahead,
510 in this case need way for kill this process.
512 Severity : enhancement
514 Description: Add state history info file, enhance import info file
515 Details : Track import connection state changes in a new osc/mdc proc file;
516 add overview-type data to the osc/mdc import proc file.
518 Severity : enhancement
520 Description: MDS create should not wait for statfs RPC while holding DLM lock.
522 Severity : enhancement
524 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
529 Descriptoin: OST grant too much space to client even there are not enough space.
530 Details : Client will shrink its grant cache to OST if there are no write
531 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
532 this grant cache if there are already not enough avaible space
533 (left_space < total_clients * 32M).
536 Frequency : start MDS on uncleanly shutdowned MDS device
538 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
539 Details : stay in waiting mds<>ost recovery finished produce random bugs
540 due race between two ll_sync thread for one lov target. send
541 ACTIVATE event only if connect realy finished and import have
545 Frequency : rare, connect and disconnect target at same time
547 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
548 Details : don't call obd_disconnect under lov_lock. this long time
549 operation and can block ptlrpcd which answer to connect request.
554 Descriptoin: don't lose wakeup for imp_recovery_waitq
555 Details : recover_import_no_retry or invalidate_import and import_close can
556 both sleep on imp_recovery_waitq, but we was send only one wakeup
560 Frequency : always with long access acl
562 Descriptoin: mds can't pack reply with long acl.
563 Details : mds don't control size of acl but they limited by reint/getattr
566 Severity : enhancement
568 Description: Update to SLES10 kernel-2.6.16.60-0.33.
570 Severity : enhancement
572 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
575 Frequency : start MDS on uncleanly shutdowned MDS device
577 Descriptoin: aborting recovery hang on MDS
578 Details : don't throttle destroy RPCs for the MDT.
581 Frequency : on remount
583 Description: external journal device not working after the remount
584 Details : clear dev_rdonly flag for external journal devices in
590 Description: shutdown vs evict race
591 Details : client_disconnect_export vs connect request race.
592 if client will evicted at this time - we start invalidate
593 thread without referece to import and import can be freed
597 Frequency : rare, need acl's on inode.
599 Description: client can't handle ost additional correctly
600 Details : if ost was added after client connected to mds client can have
601 hit lnet_try_match_md ... to big messages to wide striped files.
602 in this case need teach client to handle config events about add
603 lov target and update client max ea size at that event.
605 Severity : enhancement
607 Description: Changelogs
608 Details : Changelogs are a lightweight mechanism to track filesystem
609 metadata and namespace changes. The changelog is recorded
610 permanently on the MDTs, and is periodically "consumed" / purged
611 when records are no longer needed.
613 Severity : enhancement
615 Description: compact fld format with extents
616 Details : Store range of seq rather than every seq in FLD. Seq
617 controller update FLD rather than clients. In Case of CMD, mdt0
618 has FLD, all other metadata server act as non persistent proxy
619 for FLD queries and cache fld entries in fld cache.
624 Description: don't skip ost target if they assigned to file
625 Details : Drop slow OSCs if we can, but not for requested start idx.
626 This means "if OSC is slow and it is not the requested
627 start OST, then it can be skipped, otherwise skip it only
628 if it is inactive/recovering/out-of-space.
632 Description: more cleanup in mds_lov
633 Details : not send LOV EA under replay, we can't know about they size at this
634 time. Don't allow client connect to mds before any ost connected,
635 for avoid problems with LOV EA size and returning EIO to client.
637 Severity : enhancement
639 Description: Interoperability at server side (Disk interoperability)
641 Severity : enhancement
643 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
645 Severity : enhancement
647 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
649 Severity : enhancement
651 Description: New client IO stack (CLIO).
653 Severity : enhancement
655 Description: Commit on sharing. Eliminate inter-client dependencies between
656 uncommitted transactions by doing transaction commits.
657 Thereby clients may recovery independently.
660 Frequency : Create a symlink file with a very long name
662 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
663 Details : If there is no extra space in the request for early cancels,
664 ldlm_req_handles_avail() returns 0 instead of a negative value.
666 Severity : enhancement
668 Description: Add /proc entry for import status
669 Details : The mdc, osc, and mgc import directories now have
670 an import directory that contains useful import data for debugging
673 Severity : enhancement
675 Description: Re-disable certain /proc logging
676 Details : Enable and disable client's offset_stats, extents_stats and
677 extents_stats_per_process stats logging on the fly.
680 Frequency : Only on FC kernels 2.6.22+
682 Description: oops in statahead
683 Details : Do not drop reference count for the dentry from VFS when lookup,
684 VFS will do that by itself.
686 Severity : enhancement
688 Description: Generic /proc file permissions
689 Details : Set /Proc file permissions in a more generic way to enable non-
690 root users operate on some /proc files.
694 Description: Hitting mdc_commit_close() ASSERTION
695 Details : Properly handle request reference release in
696 ll_release_openhandle().
700 Description: quota recovery deadlock during mds failover
701 Details : This patch includes att18982, att18236, att18237 in bz14840.
703 1. fix osts hang when mds does failover with quotaon
704 2. prevent watchdog storm when osts threads wait for the
709 Frequency : only patchless client
710 Description: add workaround for race between add/remove dentry from hash
712 Severity : enhancement
714 Description: Allow OST glimpses to return PW locks
718 Description: LBUG when llog conf file is full
719 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
723 Description: Prevent import from entering FULL state when server in recovery
727 Description: service mount cannot take device name with ":"
728 Details : Only when device name contains ":/" will mount treat it as
734 Description: replace ptlrpcd with the statahead thread to interpret the async
735 statahead RPC callback
739 Frequency : on recovery
740 Description: I/O failures after umount during fail back
741 Details : if client reconnected to restarted server we need join to recovery
742 instead of find server handler is changed and process self
743 eviction with cancel all locks.
745 Severity : enhancement
747 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
749 Severity : enhancement
751 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
753 Severity : enhancement
755 Description: Upcall on Lustre log has been dumped
756 Details : Allow for a user mode script to be called once a Lustre log has
757 been dumped. It passes the filename of the dumped log to the
758 script, the location of the script can be specified via
759 /proc/sys/lnet/debug_log_upcall.
764 Description: avoid idr_remove called for id which is not allocated.
765 Details : Move assigment s_dev for clustered nfs to end of initialization,
766 to avoid problem with error handling.
771 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
772 Details : When connection is reused this not moved from CONN_UNUSED_HASH
773 into CONN_USED_HASH and this prodice warning when put connection
774 again in unused hash.
777 Severity : enhancement
779 Description: File striping can now be set to use an arbitrary pool of OSTs.
781 Severity : enhancement
783 Description: Export bytes_read/bytes_write count on OSC/OST.
787 Description: Early reply size mismatch, MGC loses connection
788 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
789 the connect flags are properly negotiated.
794 Description: quotas are not honored with O_DIRECT
795 Details : all writes with the flag O_DIRECT will use grants which leads to
796 this problem. Now using OBD_BRW_SYNC to guard this.
800 Description: add quota statistics
801 Details : 1. sort out quota proc entries and proc code.
802 2. add quota statistics
804 Severity : enhancement
806 Description: enable quota support for HEAD.
810 Description: Properly propagate oinfo flags from lov to osc for statfs
811 Details : restore missing copy oi_flags to lov requests.
813 Severity : enhancement
815 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
816 : l_getgroups(8), lst(8), routerstat(8)
817 Details : included man pages for llobdstat(8), llstat(8),
818 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
820 Severity : enhancement
822 Description: configure's --enable-quota should check the
823 : kernel .config for CONFIG_QUOTA
824 Details : configure is terminated if --enable-quota is passed but
825 : no quota support is in kernel
829 Description: Remove portals compatibility
830 Details : Remove portals compatibility, not interoperable with releases
835 Description: Resolve device initialization race
836 Details : Prevent proc handler from accessing devices added to the
837 obd_devs array but yet be intialized.
839 Severity : enhancement
841 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
843 Severity : enhancement
845 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
850 Description: Using wrong pointer in osc_brw_prep_request
851 Details : Access to array[-1] can produce panic if kernel compiled with
852 CONFIG_PAGE_ALLOC enabled
856 Description: Client runs out of low memory
857 Details : Consider only lowmem when counting initial number of llap pages
861 Description: *optional* service tags registration
862 Details : if the "service tags" package is installed on a Lustre node
863 When the filesystem is mounted, a local-node service tag will
864 be created. See http://inventory.sun.com/ for more information
865 about the Service Tags asset management system.
869 Description: Kernel BUG tries to release flock
870 Details : Lustre does not destroy flock lock before last reference goes
871 away. So always drop flock locks when client is evicted and
872 perform unlock regardless of successfulness of speaking to MDS.
876 Description: add refcount protection for osc callbacks, avoid panic on shutdown
880 Description: sanity test 65a fails if stripecount of -1 is set
881 Details : handle -1 striping on filesystem in ll_dirstripe_verify
886 Description: ASSERTION(CheckWriteback(page,cmd)) failed
887 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
890 Severity : enhancement
892 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
896 Description: do not process already freed flock
897 Details : flock can possibly be freed by another thread before it reaches
898 to ldlm_flock_completion_ast.
902 Description: LBUG during stress test
903 Details : Need properly lock accesses the flock deadlock detection list.
907 Description: oops in page fault handler
908 Details : kernel page fault handler can return two special 'pages' in error
909 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
913 Description: timeout with invalidate import.
914 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
915 should be handled by ptlrpcd. This produce long age waiting and
916 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
918 Severity : enhancement
920 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
924 Description: enable MGS and MDT services start separately
925 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
926 a co-located MGS without starting the MGS, which is a complement
927 to 'nosvc' mount option.
931 Frequency : after recovery
932 Description: precreate to many object's after del orphan.
933 Details : del orphan st in oscc last_id == next_id and this triger growing
934 count of precreated objects. Set flag LOW to skip increase count
935 of precreated objects.
939 Frequency : rare, on clear nid stats
940 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
941 Details : when clean nid stats sometimes try destroy live entry,
942 and this produce panic in free.
946 Description: Stack overflow during MDS log replay
947 ease stack pressure by using a thread dealing llog_process.
951 Description: wait until IO finished before start new when do lock cancel.
952 Details : VM protocol want old IO finished before start new, in this case
953 need wait until PG_writeback is cleared until check dirty flag and
954 call writepages in lock cancel callback.
956 Severity : enhancement
958 Description: using special macro for print time and cleanup in includes.
962 Description: mds_mfd_close() ASSERTION(rc == 0)
963 Details : In mds_mfd_close(), we need protect inode's writecount change
964 within its orphan write semaphore to prevent possible races.
968 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
972 Frequency : rare, on shutdown ost
973 Description: don't hit live lock with umount ost.
974 Details : shrink_dcache_parent can be in long loop with destroy dentries,
975 use shrink_dcache_sb instead.
979 Description: don't panic with use echo client
980 Details : echo client pass NULL as client nid pointer and this produce null
985 Description: fix build on ppc32
986 Details : compile code with -m64 flag produce wrong object file for ppc32.
990 Description: add message levels for liblustreapi
994 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
995 Details : Move the 'good_osts' check before the 'total_bavail' check. This
996 will result in an -EAGAIN and in the exit call path we call
997 alloc_rr() which will with increasing aggressiveness attempt to
998 aquire precreated objects on the minimum number of required OSCs.
1002 Description: Use old size assignment to avoid deadlock
1003 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
1004 the scheduling while holding a spinlock. We do not need locking
1005 for size in ll_update_inode() because size is only updated from
1006 the MDS for directories or files without objects, so there is no
1007 other place to do the update, and concurrent access to such inodes
1008 are protected by the inode lock.
1012 Description: resolve "_IOWR redefined" build error on SLES10
1016 Description: dump the memory debugging after all modules are unloaded to
1017 suppress false negative in conf_sanity test 39
1019 Severity : enhancement
1021 Description: build kernel-ib packages for OFED 1.3 in our release cycle
1026 Description: fix SLES kernel versioning
1027 Details : the kernel version for our SLES 10 kernel did not include a "-"
1028 before the "smp" at the end. while this was not a problem in
1029 general, it did mean that software trying to use the kernel
1030 version to try to detect a vendor specific kernel would fail.
1031 this was most evident by the OFED build scripts.
1035 Description: Don't update lov_desc members until making sure they are valid
1036 Details : When updating lov_desc members via proc fs, need fix their
1037 validities before doing the real update.
1041 Description: don't put request into delay list while invalidate in flight.
1042 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
1043 import in flight. this produce timeout for invalidate and sometimes
1044 can cause stale data.
1048 Frequency : on ppc only
1049 Description: not convert ost objects for directory because it's not exist.
1050 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
1054 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
1056 Severity : enhancement
1058 Description: Update to support 2.6.22.14 vanilla kernel.
1062 Frequency : rare, on recovery
1063 Description: read procfs can produce deadlock in some situation
1064 Details : Holding lprocfs lock which send rpc can produce block for destroy
1065 obd objects and this also block reconnect with -EALREADY. This
1066 isn't fix all lprocfs bugs - but make it rare.
1068 Severity : enhancement
1070 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
1073 Frequency : frequent on X2 node
1075 Description: mdc_set_open_replay_data LBUG
1076 Details : Set replay data for requests that are eligible for replay.
1080 Description: lustre_mgs: operation 101 on unconnected MGS
1081 Details : When MGC is disconnected from MGS long enough, MGS will evict the
1082 MGC, and late on MGC cannot successfully connect to MGS and a lot
1083 of the error messages complaining that MGS is not connected.
1087 Frequency : on network error
1088 Description: panic with double free request if network error
1089 Details : mdc_finish_enqueue is finish request if any network error ocuring,
1090 but it's true only for synchronus enqueue, for async enqueue
1091 (via ptlrpcd) this incorrect and ptlrpcd want finish request
1094 Severity : enhancement
1096 Description: client-side metadata stat-ahead during readdir(directory readahead)
1097 Details : perform client-side metadata stat-ahead when the client detects
1098 readdir and sequential stat of dir entries therein
1101 Frequency : on start mds
1103 Description: Implement get_info(last_id) in obdfilter.
1106 Frequency : occasional
1108 Description: Correctly check stale fid, not start epoch if ost not support SOM
1109 Details : open with flag O_CREATE need set old fid in op_fid3 because
1110 op_fid2 was overwritten with new generated fid, but mds can answer
1111 with one of these two fids and both is not stale. Setattr
1112 incorrectly started an epoch and assume will be called
1113 done_writing, but without SOM done_writing ever being called.
1116 Frequency : rare, depends on device drivers and load
1118 Description: MDS or OSS nodes crash due to stack overflow
1119 Details : Code changes in 1.8.0 increased the stack usage of some functions.
1120 In some cases, in conjunction with device drivers that use a lot
1121 of stack the MDS (or possibly OSS) service threads could overflow
1122 the stack. One change which was identified to consume additional
1123 stack has been reworked to avoid the extra stack usage.
1126 Frequency : occasional
1128 Description: Do not fail import if osc_interpret_create gets -EAGAIN
1129 Details : If osc_interpret_create got -EAGAIN it immediately exits and
1130 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
1131 oscc_has_objects and see OSC has no objests and call
1132 oscc_internal_create to resend create request.
1134 Severity : enhancement
1136 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
1138 Severity : enhancement
1140 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
1143 Frequency : very rare
1145 Description: Fix replay if there is an un-replied request and open
1146 Details : In some cases, older replay request will revert the
1147 mcd->mcd_last_xid on MDS which is used to record the client's
1148 latest sent request.
1150 Severity : enhancement
1152 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
1154 Severity : enhancement
1156 Description: Add rhel5 support to HEAD.
1158 Serverity : enhancement
1160 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
1165 Description: Don't allow skipping OSTs if index has been specified.
1166 Details : Don't allow skipping OSTs if index has been specified, make
1167 locking in internal create lots better.
1171 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
1172 Details : access to bitfield in structure is always rounded to long
1173 and this produce problem with not atomic change any bit.
1177 Description: Lustre make rpms failed.
1178 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
1179 builds Lustre rpms from tarball.
1182 Frequency : rare on shutdown ost
1184 Description: If llog cancel was not send before clean_exports phase, this can
1185 produce deadlock in llog code.
1186 Details : If llog thread has last reference to obd and call class_import_put
1187 this produce deadlock because llog_cleanup_commit_master wait when
1188 last llog_commit_thread exited, but this never success because was
1189 called from llog_commit_thread.
1193 Description: allow userland application know is lost one of stripes.
1194 Details : fill lvb_blocks with error code on ost and return it to
1195 application if error flag found.
1199 Description: NULL lov_tgts causing MDS oops
1200 Details : more safe checks for NULL lov_tgts for avoid oops.
1202 Severity : enhancement
1204 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
1208 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
1209 Details : If an OST is down the MDS will hang indefinitely in
1210 obd_statfs_rqset() waiting for the statfs data. While for
1211 MDS QOS usage of statfs, it should not stuck in waiting.
1213 Severity : enhancement
1215 Description: remote_acl support
1216 Details : Support ACL-based permission check for remote user.
1217 Support setfacl/getfacl for remote user with the utils
1218 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
1219 the system "{s,g}etfacl" utils.
1221 Severity : enhancement
1223 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
1225 Severity : enhancement
1227 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
1231 Description: make mgs_setparam() handle fsname containing dash
1232 Details : fsname containing a dash does not work with lctl conf_param
1234 Severity : enhancement
1236 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
1238 Severity : enhancement
1240 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
1244 Frequency : rare, at shutdown
1245 Description: access already free / zero obd_namespace.
1246 Details : if client_disconnect_export was called without force flag set,
1247 and exist connect request in flight, this can produce access to
1248 NULL pointer (or already free pointer) when connect_interpret
1249 store ocd flags in obd_namespace.
1253 Frequency : only at startup
1254 Description: not alloc memory with spinlock held.
1255 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
1256 if any spinlock held.
1258 Severity : enhancement
1260 Description: make lustre randomly fail allocating memory
1261 Details : Make lustre randomly failed allocating memory for testing purpose.
1263 Severity : enhancement
1265 Description: lost problems with lov objid file
1266 Details : Fixes some scability and access to not inited memory problems
1267 in work with lov objdid file.
1272 Description: lfs find does not continue on file error
1273 Details : Continue other files processing when a file/dir is absent.
1277 Description: Inconsistent usage of lustre_pack_reply()
1278 Details : Standardize the usage of lustre_pack_reply() such that it
1279 always generate a CERROR on failure.
1282 Frequency : occasional
1284 Description: MDS hang or stay in waiting lock
1285 Details : If client receive lock with CBPENDING flag ldlm need send lock
1286 cancel as separate rpc, to avoid situation when cancel request
1287 can't processed due all i/o threads stay in wait lock.
1291 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1295 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1299 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1303 Description: Update to RHEL4 latest kernel.
1305 Severity : enhancement
1307 Description: Build SLES10 patchless client fails
1308 Details : The configure was broken by run ./configure with
1309 --with-linux-obj=.... argument for patchless client. When the
1310 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1311 can't search header adequately. Use absolute path such as
1312 -I($LINUX)/include instead.
1316 Description: interrupt oig_wait produce painc on resend.
1317 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1318 private set, and this produce situation when rq_ptlrpcd_data not
1319 copyed to new allocated request and triggered LBUG on assert
1320 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1321 ptlrpcd set if request is changed and can be safety changed to use
1326 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1327 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1328 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1329 these macros and replace them with their difinitions instead.
1333 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1334 Details : Update which_patch & target file for SLES10 latest kernel.
1337 Description: Remove client patches from SLES 10 kernel.
1338 Details : This causes SLES 10 clients to behave as patchless clients
1339 even on a Lustre-patched (server) kernel.
1341 Severity : enhancement
1343 Description: self-adjustable client's lru lists
1344 Details : use adaptive algorithm for managing client cached locks lru
1345 lists according to current server load, other client's work
1346 pattern, memory activities, etc. Both, server and client
1347 side namespaces provide number of proc tunables for controlling
1350 Severity : enhancement
1352 Description: light-weight GSS support
1353 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
1354 and reduce performance overhead.
1356 Severity : enhancement
1358 Description: Linux keyring support
1359 Details : Support using service of Linux keyring for Lustre GSS internal
1360 context refresh/cache mechanism.
1364 Description: Fix errors in lfs documentation
1365 Details : Fixes man pages
1369 Description: don't use GFP_* in generic Lustre code.
1370 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
1374 Description: obdclass is limited by single OBD_ALLOC(idarray)
1375 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
1379 Description: warnings with build patchless client with vanila 2.6.19 and up
1380 Details : change old ctl_table style and replace ctl_table/ctl_table_header
1381 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
1385 Frequency : Only for SLES
1386 Description: sanity_quota fail test_1
1387 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
1388 file, which makes TSTID[2] inunique.
1392 Frequency : Only for SLES9
1393 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
1394 Details : Update target/ChangeLog/which_patch .
1398 Frequency : Only for test_10 in sanity-quota.sh
1399 Description: a bug in quota test code
1400 Details : checking a test flag in a wrong place
1404 Frequency : Only for test_13 in sanity-quota.sh
1405 Description: a bug in quota test script
1406 Details : shouldn't check overfull allocated quota
1410 Description: jbd statistics
1411 Details : Port older jbd statistics patch for sles10
1413 Severity : enhancement
1415 Description: Add printing inode info into message about error in writepage.
1420 Description: reply_lock_interpret crash due to race with it and lock cancel.
1421 Details : Do not replay locks that are being cancelled. Do not reference
1422 locks by their address during replay, just by their handle.
1426 Frequency : When flocks are used.
1427 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
1428 Details : Flock locks might destroy just granted lock if it could be merged
1429 with another existing flock, this is done in completion handler,
1430 so teach ldlm_cli_enquque_fini that this is a valid case for
1436 Description: Oops in read and write path when failing to allocate lock.
1437 Details : Check if lock allocation failed and return error back.
1441 Description: lstripe command fails for valid OST index
1442 Details : The stripe offset is compared to lov->desc.ld_tgt_count
1443 instead of lov->desc.ld_active_tgt_count.
1447 Description: sanity.sh failed test 103
1448 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
1453 Description: df doesn't work properly if diskfs blocksize != 4K
1454 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
1458 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
1459 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
1461 Severity : enhancement
1463 Description: Adaptive timeouts
1464 Details : RPC timeouts adapt to changing server load and network
1465 conditions to reduce resend attempts and improve recovery time.
1469 Description: llapi_file_create() does not allow some changes
1470 Details : add llapi_file_open() that allows specifying the mode and
1471 open flags, and also returns an open file handle.
1475 Description: merge and cleanup kernel patches.
1476 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
1480 Description: Add journal checksum support.(Kernel part)
1481 Details : The journal checksum feature adds two new flags i.e
1482 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
1483 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
1484 indicates that the commit block contains the checksum for
1485 the blocks described by the descriptor blocks. Now commit
1486 record can be sent to disk without waiting for descriptor
1487 blocks to be written to disk. This behavior is controlled
1488 using JBD2_FEATURE_ASYNC_COMMIT flag.
1492 Description: OSS needs mutliple precreate threads
1493 Details : Add ability to start more than one create thread per OSS.
1497 Description: Sanity.sh test_65k failed: setstripe should have succeeded
1498 Details : execute lfs setstripe on client
1502 Description: mds_obd_create error creating tmp object
1503 Details : When the user sets quota on root, llog will be affected and can't
1504 create files and write files.
1507 Frequency : Always on ia64 patchless client, and possibly others.
1509 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
1510 Details : This allows the patchless client to be loaded on architectures
1511 without this export.
1515 Description: RedHat Update kernel for RHEL5
1516 Details : Add the kernel config file for RHEL5.
1520 Description: RedHat Update kernel for RHEL5
1521 Details : Modify the kernel config file more closer RHEL5.
1525 Description: Build failure against Centos5 (RHEL5)
1526 Details : Define PAGE_SIZE when it isn't present.
1530 Description: test_7 on sanity-quota.sh can't be used on separate machines
1531 Details : add facet to handle it
1535 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
1536 Details : with NFS, the anon dentry's parent was set to itself in
1537 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
1538 corresponding dentry other than use rec->ur_name.
1540 Severity : enhancement
1542 Description: lfs setstripe enhancement
1543 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
1547 Description: enable data checksumming by default
1548 Details : enable checksum by default, allow --disable-checksum
1549 configure option and "-o nochecksum" mount option. Checksums
1550 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
1554 Description: lustre support for RHEL5
1555 Details : Add support for RHEL5.
1559 Description: Client eviction due to failover config
1560 Details : after a connection loss, the lustre client should attempt to
1561 reconnect to the last active server first before trying the
1562 other potential connections.
1566 Description: when mds and osts use different quota unit(32bit and 64bit),
1567 quota will be released repeatly.
1568 Details : void sending multiple quota reqs to mds, which will keep the
1569 status between the reqs.
1573 Description: osts not allocated evenly to files
1574 Details : change the condition to increase offset_idx
1577 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
1579 Description: Data corruption for OSTs that are formatted larger than 2TB
1581 Details : When generating the bio request for lustre file writes the
1582 sector number would overflow a temporary variable before being
1583 used for the IO. The data reads correctly from Lustre (which
1584 will overflow in a similar manner) but other file data or
1585 filesystem metadata may be corrupted in some cases.
1589 Description: Tune the kernel for good SCSI performance.
1590 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
1591 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
1596 Description: rewrite ext2-derived code in obdclass/uuid.c
1597 Details : rewrite inherited code (uuid parsing code from ext2 utils)
1598 from scratch preserving functionality.
1602 Description: "lfs find -obd UUID" prints directories
1603 Details : "lfs find -obd UUID" will return all directory names instead
1604 of just file names. It is incorrect because the directories
1605 do not reside on the OSTs.
1609 Description: use i_size_read and i_size_write in 2.6 port
1610 Details : replace inode->i_size access with i_size_read/write()
1613 Frequency : when removing large files
1615 Description: scheduling issue during removal of large Lustre files
1616 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
1617 It causes scheduling issues when removing large files (17TB in the
1621 Frequency : only with liblustre clients on XT3
1623 Description: evictions taking too long
1624 Details : allow llrd to evict clients directly on OSTs
1627 Frequency : only on ppc
1629 Description: /proc/fs/lustre/devices broken on ppc
1630 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
1631 We should make sure the type of 'index' is loff_t and then cast
1632 explicitly as needed below. Do not assign an explicitly cast
1636 Frequency : only for rhel5
1638 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
1639 Details : Modify the target file & which_kernel.
1641 Severity : enhancement
1643 Description: omit set fsid for export NFS
1644 Details : fix set/restore device id for avoid EMFILE error and mark lustre
1645 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
1649 Frequency : Always, for kernels after 2.6.16
1650 Description: Fix warning idr_remove called for id=.. which is not allocated.
1651 Details : Last kernels save old s_dev before kill super and not allow
1652 to restore from callback - restore it before call kill_anon_super.
1656 Description: buffer overruns could theoretically occur
1657 Details : llapi_semantic_traverse() modifies the "path" argument by
1658 appending values to the end of the origin string, and a
1659 overrun may occur. Adding buffer overrun check in liblustreapi.
1663 Description: Fix error on 'ls .' at the top of the Lustre mount.
1664 Details : Don't revalidate dentry if it is a root dentry.
1668 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
1669 Details : Modify vm-tunables-rhel4.patch.
1673 Description: Kernel config for 2.6.18-vanilla.
1674 Details : Modify targets/2.6-vanilla.target.in.
1675 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
1676 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
1677 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
1678 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
1682 Description: improve handling recoverable errors
1683 Details : if request processig with error which can be recoverable on server
1684 request should be resend, otherwise page released from cache and
1689 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
1690 Details : Modify target file & which_patch.
1691 A flaw was found in the IA32 system call emulation provided
1692 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
1693 value could be stored in the %RAX register, which could trigger an
1694 out-of-bounds system call table access. An untrusted local user
1695 could exploit this flaw to run code in the kernel
1696 (ie a root privilege escalation). (CVE-2007-4573).
1700 Description: O_DIRECT bypasses client statistics.
1701 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
1702 read_ahead_stats were updated. copied the stats section from
1703 osc_send_oap_rpc() into async_internal().
1707 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
1711 Description: change order of libsysio includes
1712 Details : '#include sysio.h' should always come before '#include xtio.h'
1716 Description: handle "serious error: objid * already exists" more gracefully
1717 Details : If LAST_ID value on disk is smaller than the objects existing in
1718 the O/0/d* directories, it indicates disk corruption and causes an
1719 LBUG(). If the object is 0-length, then we should use the existing
1720 object. This will help to avoid a full fsck in most cases.
1722 Severity : enhancement
1724 Description: adapt the lustre_config script to support the upgrade case
1725 Details : Add "-u" option for lustre_config script to support upgrading 1.4
1726 server targets to 1.6 in parallel.
1730 Description: To avoid grant space > avaible space when the disk is almost
1731 full. Without this patch you might see the error "grant XXXX >
1732 available" or some LBUG about grant, when the disk is almost
1734 Details : In filter_check_grant, for non_grant cache write, we should
1735 check the left space by if (*left > ungranted + bytes), instead
1736 of (*left > ungranted), because only we are sure the left space
1737 is enough for another "bytes", then the ungrant space should be
1738 increase. In client, we should update cl_avail_grant only there
1739 is OBD_MD_FLGRANT in the reply.
1743 Description: Update RHEL 4 kernel to fix local root privilege escalation.
1744 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
1745 described in CVE-2007-4573. This problem could allow untrusted
1746 local users to gain root access.
1749 Frequency : when using O_DIRECT and quotas
1751 Description: Incorrect file ownership on O_DIRECT output files
1752 Details : block usage reported by 'lfs quota' does not take into account
1753 files that have been written with O_DIRECT.
1758 Description: touch file failed when fs is not full
1759 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
1760 otherwise we can get ENOSP while fs is not full.
1764 Description: parallel lock callbacks
1765 Details : Instead of sending blocking and completion callbacks as separated
1766 requests, adding them to a set and sending in parallel.
1769 Frequency : only for Cray XT3
1770 Bugzilla : 12829/13455
1771 Description: Changing primary group doesn't change the group lustre assigns to
1773 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
1774 overridden with the primary group provided by the group upcall,
1775 whereas the supplied fsgid can be trusted if it is in the list of
1776 supplementary groups returned by the group upcall.
1778 Severity : enhancement
1780 Description: Allow masking D_WARNING, D_ERROR messages from console
1781 Details : Console messages can now be disabled via lnet.printk.
1785 Description: User code with malformed file open parameter crashes client node
1786 Details : Before packing join_file req, all the related reference should be
1787 checked carefully in case some malformed flags cause fake
1788 join_file req on client.
1792 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
1793 Details : ldlm_completion_ast() assumes that a lock is granted when the req
1794 mode is equal to the granted mode. However, it should also check
1795 that LDLM_FL_CP_REQD is not set.
1799 Description: Heavy nfs access might result in deadlocks
1800 Details : After ELC code landed, it is now improper to enqueue any mds
1801 locks under och_sem, because enqueue might want to decide to
1802 cancel open locks for same inode we are holding och_sem for.
1806 Description: Client eviction while running blogbench
1807 Details : A lot of unlink operations with concurrent I/O can lead to a
1808 deadlock causing evictions. To address the problem, the number of
1809 oustanding OST_DESTROY requests is now throttled to
1810 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
1811 callbacks are processed in priority.
1815 Description: enable ACLs on MDS by default
1816 Details : ACLs must be enabled on MDS by default.
1819 Frequency : PPC/PPC64 only
1821 Description: conflicts between asm-ppc64/types.h and lustre_types.h
1822 Details : fix duplicated definitions between asm-ppc64/types.h and
1823 lustre_types.h on PPC.
1826 Frequency : PPC/PPC64 only
1828 Description: asm-ppc/segment.h does not exist
1829 Details : fix compile issue on PPC.
1833 Description: better handle error messages in extents code
1836 Frequency : RHEL4 only
1838 Description: mkfs is very slow on IA64/RHEL4
1839 Details : A performance regression has been discovered in the MPT Fusion
1840 driver between versions 3.02.73rh and 3.02.99.00rh. As a
1841 consequence, we have downgraded the MPT Fusion driver in the RHEL4
1842 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
1844 Severity : enhancement
1846 Description: SNMP support enhancement
1847 Details : Adding total number of sampled request for an MDS node in snmp
1850 Severity : enhancement
1852 Description: Optimize ldlm waiting list processing for PR extent locks
1853 Details : When processing waiting list for read extent lock and meeting read
1854 lock that is same or wider to it that is not contended, skip
1855 processing rest of the list and immediatelly return current
1856 status of conflictness, since we are guaranteed there are no
1857 conflicting locks in the rest of the list.
1861 Description: Time out and refuse to reconnect
1862 Details : When the failover node is the primary node, it is possible
1863 to have two identical connections in imp_conn_list. We must
1864 compare not conn's pointers but NIDs, otherwise we can defeat
1865 connection throttling.
1869 Description: port llog fixes from b1_6 into HEAD
1870 Details : Port llog reference couting and some llog cleanups from b1_6
1871 (bug 10800) into HEAD, for protect from panic and access to already
1872 free llog structures.
1876 Description: Detect stride IO mode in read-ahead
1877 Details : When a client does stride read, read-ahead should detect that and
1878 read-ahead pages according to the detected stride pattern.
1882 Description: data checksumming impacts single node performance
1883 Details : add support for several checksum algorithm. Currently, only CRC32
1884 and Adler-32 are supported. The checksum type can be changed on
1885 the fly via /proc/fs/lustre/osc/*/checksum_type.
1889 Description: use adler32 for page checksums
1890 Details : when available, use the Adler-32 algorithm instead of CRC32 for
1895 Description: build for x2 fails
1896 Details : fix compile issue on Cray systems.
1900 Description: Properly match for duplicate locks
1901 Details : Due to different lock order from skiplists code, we need to
1902 traverse entire list for now
1905 Frequency : only on PPC/SLES10
1907 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
1908 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
1909 be defined. Add a hack in mkfs_lustre.c to work around this compile
1914 Description: LASSERT on MDS when client holding flock lock dies
1915 Details : ldlm pool logic depends on number of granted locks equal to
1916 number of released locks which is not true for flock locks, so
1917 just exclude such locks from consideration.
1921 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
1922 Details : Use fsfilt_sync() for both the whole filesystem sync and
1923 individual file sync to eliminate dangerous inode locking
1924 with I_LOCK that can lead to a deadlock.
1928 Description: performance in 1.6.3
1929 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
1930 sure that 1MB requests are not fragmented by the block layer.
1932 Severity : enhancement
1934 Description: organize the server-side client stats on per-nid basis
1935 Details : Change the structure of stats under obdfilter and mds to
1943 The "uuid"s file would list the uuids of _active_ exports.
1944 And the clear entry is to clear all stats and stale nids.
1946 Severity : enhancement
1948 Description: eliminate client locks in face of contention
1949 Details : file contention detection and lockless i/o implementation
1950 for contended files.
1954 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
1957 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
1959 Description: MDS may lose file striping (and hence file data) in some cases
1960 Details : If there are additional extended attributes stored on the MDS,
1961 in particular ACLs, SELinux, or user attributes (if user_xattr
1962 is specified for the client mount options) then there is a risk
1963 of attribute loss. Additionally, the Lustre file striping
1964 needs to be larger than default (e.g. striped over all OSTs),
1965 and an additional attribute must be stored initially in the
1966 inode and then increase in size enough to be moved to the
1967 external attribute block (e.g. ACL growing in size) for file
1972 Description: skiplist implementation simplification
1973 Details : skiplists are used to group compatible locks on granted list
1974 that was implemented as tracking first and last lock of each
1975 lock group the patch changes that to using doubly linked lists
1979 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
1980 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
1981 because LASSERT on both the data supplied by a client, and the
1982 data on disk is dangerous and incorrect.
1984 Severity : enhancement
1986 Description: Slow truncate/writes to huge files at high offsets.
1987 Details : Directly associate cached pages to lock that protect those pages,
1988 this allows us to quickly find what pages to write and remove
1989 once lock callback is received.
1993 Description: more ldlm soft lockups
1994 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
1995 starve other threads from the resource lock for a long time in
1996 case of long waiting queue, so change the debug level from
1997 D_OTHER to the less frequently used D_INFO.
1999 Severity : enhancement
2001 Description: add -gid, -group, -uid, -user options to lfs find
2005 Description: Hung threads in invalidate_inode_pages2_range
2006 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
2007 once one is completed. As a result, some RPCs are stuck in the
2008 queue and are never sent.
2012 Description: filter threads hungs on waiting journal commit
2013 Details : Cleanup filter group llog code, then only filter group llog will
2014 be only created in the MDS/OST syncing process.
2018 Description: Procfs and llog threads access destoryed import sometimes.
2019 Details : Sync the import destoryed process with procfs and llog threads by
2020 the import refcount and semaphore.
2022 Severity : enhancement
2024 Description: openlock cache of b1_6 port to HEAD
2029 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
2030 Details : If insertion of an extent fails, then discard the inode
2031 preallocation and free data blocks else it can lead to duplicate
2036 Description: don't always update ctime in ext3_xattr_set_handle()
2037 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
2038 In some cases the ctime should not be updated, for example for
2039 2.0->1.8 compatibility it is necessary to delete an xattr and it
2040 should not update the ctime.
2044 Bugzilla : 15713/16362
2045 Description: Assertion in iopen_connect_dentry in 1.6.3
2046 Details : looking up an inode via iopen with the wrong generation number can
2047 populate the dcache with a disconneced dentry while the inode
2048 number is in the process of being reallocated. This causes an
2049 assertion failure in iopen since the inode's dentry list contains
2050 both a connected and disconnected dentry.
2054 Description: assertion failure in ldlm_handle2lock()
2055 Details : fix a race between class_handle_unhash() and class_handle2object()
2056 introduced in lustre 1.6.5 by bug 13622.
2061 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
2062 Details : remove the limit of 256 scsi disks in the sd_iostat patch
2067 Description: oops in sd_iostats_seq_show()
2068 Details : unloading/reloading the scsi low level driver triggers a kernel
2069 bug when trying to access the sd iostat file.
2074 Description: Kernel panics during QLogic driver reload
2075 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
2076 patch, causing memory corruption.
2081 Description: journal_dev option does not work in b1_6
2082 Details : pass mount option during pre-mount.
2084 Severity : enhancement
2086 Description: Add a FIEMAP(FIle Extent MAP) ioctl
2087 Details : FIEMAP ioctl will allow an application to efficiently fetch the
2088 extent information of a file. It can be used to map logical blocks
2089 in a file to physical blocks in the block device.
2093 Description: LDLM soft lockups - improvement
2094 Details : It is be possible to send the lock handle along with each read
2095 or write request because the client is already doing a lock match
2096 itself so there isn't any reason the OST should have to re-do that
2100 Frequency : only on Cray X2
2102 Description: X2 build failures
2103 Details : fix build failures on Cray X2.
2107 Description: xid & resent requests
2108 Details : Initialize RPC XID from clock at startup (randomly if clock is
2111 Severity : enhancement
2113 Description: Add lustre_start utility to start or stop multiple Lustre servers
2118 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
2119 Details : In case of memory pressure, list_del() can be called twice on
2120 req->rq_history_list, causing a kernel oops.
2124 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
2125 Details : fix stack overflow in the distributed lock manager by defering
2126 export eviction after a failed AST to the elt thread instead of
2127 handling it in the dlm interpret routine.
2131 Description: Convert some comments to new format.
2132 Details : Update documenting comments to match doxygen conventions.
2136 Description: Grammar fixes.
2137 Details : A couple of trivial spelling fixes.
2141 Description: OSD_COUNTERS-mandatory
2142 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
2143 to simplify the code.
2147 Description: simplify cmm_device freeing logic.
2148 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
2152 Description: Add lockdep support to dt_object_operations locking interface.
2153 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
2154 parameter indicating lock ordering. Update mdd code to use new
2159 Description: Introduce failloc constants for lockless IO tests.
2160 Details : Add two new failloc constants to test lockless IO. Only one of
2161 them in implemented---another is checked in yet to be landed
2166 Description: Add lockdep support for inode mutex.
2167 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
2168 in the situations where Lustre has to lock more than one inode
2173 Description: Add optional invariants checking support.
2174 Details : Add new LINVRNT() macro, optional on new --enable-invariants
2175 configure switch. This macro is to be used for consistency and
2176 sanity checks that are too expensive to be left in `production'
2181 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
2182 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
2187 Description: Add lockdep support for ldlm_lock and ldlm_resource.
2188 Details : Use spin_lock_nested() in (the only) situation where more than
2189 one ldlm_lock is locked simultaneously. Also, fix possible
2190 dead-lock in ldlm_lock_change_resource() by enforcing particular
2195 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
2196 Details : Instead of specifying each ldlm_lock call-back through separate
2197 parameter, wrap them into struct ldlm_callback_suite.
2201 Description: Kill join_lru obd method and its callers.
2202 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
2203 in memory---a requirement implemented through ->o_join_lru() obd
2204 method in HEAD. Remove this method and its users.
2208 Description: Add asynchronous ldlm ENQUEUE completion handler.
2209 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
2210 case that stock ldlm_completion_ast() cannot handle as it waits
2211 until lock is granted. Introduce new ldlm_completion_ast_async()
2212 for this. Also comment ldlm_completion_ast().
2216 Description: ldlm_error <-> errno conversion.
2217 Details : Add functions to map (rather arbitrary) between LDLM error codes
2218 and standard errno values. CLIO needs this to prevent LDLM specific
2219 constants from escaping ldlm and osc.
2223 Description: Kill unused ldlm_handle2lock_ns() function.
2224 Details : Kill unused ldlm_handle2lock_ns() function.
2228 Description: Add lu_ref support to ldlm_lock
2229 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
2230 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
2231 and ldlm_resource. LDLM interface has to be changed, because code that
2232 releases a reference on a lock, has to "know" what reference this is.
2233 In the most frequent case
2235 lock = ldlm_handle2lock(handle);
2237 LDLM_LOCK_PUT(lock);
2239 no changes are required. When any other reference (received _not_ from
2240 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
2241 instead of LDLM_LOCK_PUT().
2243 Arguably, changes are pervasive, and interface requires some discipline
2244 for proper use. On the other hand, it was very instrumental in finding
2245 a few leaked lock references.
2249 Description: Add ldlm_lock_addref_try().
2250 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
2251 attempts to addref a lock that might be being canceled
2256 Description: Add ldlm_weigh_callback().
2257 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
2258 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
2259 hard-coded `number of pages' logic.
2263 Description: Add lockdep annotations to llog code.
2264 Details : Use appropriately tagged _nested() locking calls in the places
2265 where llog takes more than one ->lgh_lock lock.
2269 Description: Add loi_kms_set().
2270 Details : Wrap kms updates into a helper function.
2274 Description: Constify instances of struct lsm_operations.
2275 Details : Constify instances of struct lsm_operations.
2279 Description: lu_conf support.
2280 Details : On a server, a file system object is uniquely identified
2281 by a fid, which is sufficient to locate and load all object
2282 state (inode). On a client, on the other hand, more data are
2283 necessary instantiate an object. Change lu_object_find() and
2284 friends to take additional `lu_conf' argument describing object.
2285 Typically this includes layout information.
2289 Description: lu_context fixes.
2290 Details : Introduce new lu_context functions that are needed on the client
2291 side, where some system threads (ptlrpcd) are shared by multiple
2292 modules, and so cannot be stopped during module shutdown.
2296 Description: Add start and stop methods to lu_device_type_operations.
2297 Details : Introduce two new methods in lu_device_type_operations, that are
2298 invoked when first instance of a given type is created and last one
2299 is destroyed respectively. This is need by CLIO.
2303 Description: Add lu_ref support to struct lu_device.
2304 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
2305 track leaked references.
2309 Description: Introduce lu_kmem_descr.
2310 Details : lu_kmem_descr and its companion interface allow to create
2311 and destroy a number of kmem caches at once.
2315 Description: Fix lu_object finalization race.
2316 Details : Fix a race between lu_object_find() finding an object and its
2317 concurrent finalization. This race is (most likely) not possible
2318 on the server, but might happen on the client.
2322 Description: Introduce lu_ref interface.
2323 Details : lu_ref is a debugging module allowing to track references to
2324 a given object. It is quite cpu expensive, and has to be
2325 explicitly enabled with --enable-lu_ref. See usage description
2330 Description: Factor lu_site procfs stats into a separate function.
2331 Details : Separate lu_site stats printing code into a separate function
2332 that can be reused on a client.
2336 Description: Constify instances of struct {lu,dt,md}_device_operations.
2337 Details : Constify instances of struct {lu,dt,md}_device_operations.
2341 Description: Introduce struct md_site and move meta-data specific parts of
2342 struct lu_site here.
2343 Details : Move md-specific fields out of struct lu_site into special struct
2344 md_site, so that lu_site can be used on a client.
2348 Description: Kill mdd_lov_destroy().
2349 Details : Remove unused mdd code.
2353 Description: Add st_block checking to multistat.c.
2354 Details : Add st_block checking to multistat.c.
2358 Description: Add lu_ref support to struct obd_device.
2359 Details : Add lu_ref tracking to obd_device.
2363 Description: Kill obd_set_fail_loc().
2364 Details : Remove unused code.
2368 Description: Add special type for ptlrpc_request interpret functions.
2369 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
2370 there. Actual usage will be in CLIO.
2374 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
2375 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
2376 doesn't work with lockdep.
2380 Description: Add rwv.c test program.
2381 Details : New testing program exercising readv(2) and writev(2) (Qian).
2385 Description: Add sendfile.c test program.
2386 Details : New testing program exercising sendfile(2) (Jay).
2390 Description: Ratelimit a message that can be very frequent.
2391 Details : Ratelimit a memory allocation failure message that can
2396 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2397 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2401 Description: Liblustre build fixes.
2402 Details : Liblustre build fixes.
2406 Description: libcfs: add cfs_{need,cond}_resched() interface.
2407 Details : libcfs: add cfs_{need,cond}_resched() definition and
2408 implementations for Linux, NT, and liblustre.
2410 Severity : enhancement
2412 Description: More exported tunables for mballoc
2413 Details : Add support for tunable preallocation window and new tunables for
2414 large/small requests
2418 Description: Detect corruption of block bitmap and checking for preallocations
2419 Details : Checks validity of on-disk block bitmap. Also it does better
2420 checking of number of applied preallocations. When corruption is
2421 found, it turns filesystem readonly to prevent further corruptions.
2425 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
2426 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
2427 window to 0. If it does hit cache miss, and read-pattern is still
2428 stride-io mode, does not reset the stride window, but also does not
2429 increase the stride window length in this case.
2433 Frequency : only for big-endian servers
2434 Description: Check if system is big-endian while mounting fs with extents feature
2435 Details : Mounting a filesystem with extents feature will fail on big-endian
2436 systems since ext3-based ldiskfs is not supported on big-endian
2437 systems. This can be over-riden with "bigendian_extents" mount option.
2439 Severity : enhancement
2441 Description: The root squash functionality
2442 Details : A security feature, which is to prevent users from being able
2443 to mount lustre on their desktop, run as root, and delete
2444 all of the files in the filesystem. The goal is accomplished by
2445 remapping user id (UID) and group id (GID) of the root user to
2446 a UID and GID specified by the system administartor via Lustre
2447 configuration management server (MGS). The functionality also
2448 allows to specify sets of clients for which the remapping does
2453 Description: Excessive recovery window
2454 Details : With AT enabled, the recovery window can be excessively long (6000+
2455 seconds). To address this problem, we no longer use
2456 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
2457 timeout no longer depends on the service time, it is set to
2458 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
2459 time via pb_service_time.
2463 Description: Watchdog triggered on MDS failover
2464 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
2465 the OSTs know that the MDS "UUID" can be reused for the same export
2466 from a different NID, so we do not need to wait for the export to be
2470 Frequency : rare, only if using MMP with Linux RAID
2472 Description: MMP doesn't work with Linux RAID
2473 Details : While using HA for Lustre servers with Linux RAID, it is possible
2474 that MMP will not detect multiple mounts. To make this work we
2475 need to unplug the device queue in RAID when the MMP block is being
2476 written. Also while reading the MMP block, we should read it from
2477 disk and not the cached one.
2479 Severity : enhancement
2481 Description: open file using fid
2482 Details : A file can be opened using just its fid, like
2483 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
2486 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
2489 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
2490 Details : add proper locking for creation of jbd_4k slab cache
2494 Description: MMP check in ext3_remount() fails without displaying any error
2495 Details : When multiple mount protection fails during remount, proper error
2498 Severity : enhancement
2500 Description: Allow stripe size to be up to 4G-64k
2501 Details : Fix math logic to allow large stripe sizes.
2505 Description: add check for >8TB ldiskfs filesystems
2506 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
2507 Don't allow >8TB ldiskfs filesystems to be mounted without
2508 force_over_8tb mount option
2513 Descriptoin: OSS complains "too many blocks" and crashes
2514 Details : In filter_commitrw_write():
2515 1. lquota_chkquota();
2516 2. obdo_from_inode();
2517 3. lquota_pending_commit();
2518 "2." has changed uid and gid in this case, then caused
2519 lquota_chkquota() and lquota_pending_commit() to use
2520 different uid and gid. That is the root of the bug.
2524 Descriptoin: getstripe doesn't report stripe count = -1 sanely
2525 Details : Detect and print out the 'stripe count = -1' case properly.
2529 Description: Out or order replies might be lost on replay
2530 Details : In ptlrpc_retain_replayable_request if we cannot find retained
2531 request with tid smaller then one currently being added, add it
2532 to the start, not end of the list.
2534 --------------------------------------------------------------------------------
2536 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
2538 * Support for kernels:
2539 2.6.5-7.283 (SLES 9),
2540 2.6.9-55.EL (RHEL 4),
2541 2.6.16.46-0.14 (SLES 10),
2542 2.6.18.8 vanilla (kernel.org)
2543 * Client support for unpatched kernels:
2544 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2545 2.6.16 - 2.6.22 vanilla (kernel.org)
2546 * Due to recently discovered recovery problems, we do not recommend
2547 using patchless RHEL 4 clients with this or any earlier release.
2548 * Recommended e2fsprogs version: 1.39.cfs8
2549 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2550 * Starting with this release, the ldiskfs backing filesystem required
2551 by Lustre is now in its own package, lustre-ldiskfs. This package
2552 should be installed. It is versioned separately from Lustre and
2553 may be released separately in future.
2557 Description: block reactivating mgc import until all deactivates complete
2558 Details : Fix race when failing back MDT/MGS to itself (testing)
2561 Frequency : at statup only
2563 Description: mds_lov_synchronize race leads to various problems
2564 Details : simultaneous MDT->OST connections at startup can cause the
2565 sync to abort, leaving the OSC in a bad state.
2567 Severity : enhancement
2569 Description: add optional extra BUILD_VERSION info
2570 Details : add a new environment variable (namely LUSTRE_VERS) which allows
2571 to override the lustre version.
2574 Frequency : 2.6.18 servers only
2576 Description: ll_kern_mount() doesn't release the module reference
2577 Details : The ldiskfs module reference count never drops down to 0
2578 because ll_kern_mount() doesn't release the module reference.
2583 Description: server LBUG when using old ost_num_threads parameter
2584 Details : Accept the old ost_num_threads parameter but warn that it
2585 is deprecated, and fix an off-by-one error that caused an LBUG.
2590 Description: Transient SCSI error results in persistent IO issue
2591 Details : iobuf->dr_error is not reinitialized to 0 between two
2595 Frequency : sometimes when underlying device returns I/O errors
2597 Description: OSTs not going read-only during write failures
2598 Details : OSTs are not remounted read-only when the journal commit threads
2599 get I/O errors because fsfilt_ext3 calls journal_start/stop()
2600 instead of the ext3 wrappers.
2604 Description: poor connect scaling with increasing client count
2605 Details : Don't run filter_grant_sanity_check for more than 100 exports
2606 to improve scaling for large numbers of clients.
2609 Frequency : SLES10 only
2611 Description: sanity-quota.sh quotacheck failed: rc = -22
2612 Details : Quotas cannot be enabled on SLES10.
2615 Frequency : liblustre clients only
2617 Description: getdirentries does not give error when run on compute nodes
2618 Details : getdirentries does not fail when the size specified as an argument
2619 is too small to contain at least one entry
2621 Severity : enhancement
2623 Description: Add LNET router traceability for debug purposes
2624 Details : If a checksum failure occurs with a router as part of the
2625 IO path, the NID of the last router that forwarded the bulk data
2626 is printed so it can be identified.
2631 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
2632 Details : Due to a race condition, liblustre clients were occasionally
2633 evicted incorrectly.
2635 Severity : enhancement
2637 Description: lfs setstripe use optional parameters instead of postional
2640 Severity : enhancement
2642 Description: Nanosecond timestamp support for ldiskfs
2643 Details : The on-disk ldiskfs filesystem has added support for nanosecond
2644 resolution timestamps. There is not yet support for this at
2645 the Lustre filesystem level.
2648 Frequency : during server recovery
2650 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
2651 Details : request with rq_no_resend flag not awake l_wait_event if they get
2655 Frequency : nfs export on patchless client
2657 Description: connectathon hang when test nfs export over patchless client
2658 Details : Disconnected dentry cannot be found with lookup, so we do not need
2659 to unhash it or make it invalid
2662 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
2663 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
2664 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
2669 Description: open req refcounting wrong on reconnect
2670 Details : If reconnect happened between getting open reply from server and
2671 call to mdc_set_replay_data in ll_file_open, we will schedule
2672 replay for unreferenced request that we are about to free.
2673 Subsequent close will crash in variety of ways.
2674 Check that request is still eligible for replay in
2675 mdc_set_replay_data().
2680 Description: disable writes to filesystem when reading health_check file
2681 Details : the default for reading the health_check proc file has changed
2682 to NOT do a journal transaction and write to disk, because this
2683 can cause reads of the /proc file to hang and block HA state
2684 checking on a healthy but otherwise heavily loaded system. It
2685 is possible to return to the previous behaviour during configure
2686 with --enable-health-write.
2688 Severity : enhancement
2690 Description: 64-bit inode version
2691 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
2692 to the inode. This will be required for version-based recovery.
2697 Description: MDS fails to start if a duplicate client export is detected
2698 Details : in some rare cases it was possible for a client to connect to
2699 an MDS multiple times. Upon recovery the MDS would detect this
2700 and fail during startup. Handle this more gracefully.
2702 Severity : enhancement
2704 Description: Add -o localflock option to simulate old noflock
2706 Details : This will achieve local-only flock/fcntl locks
2712 Description: log_commit_thread vs filter_destroy race leads to crash
2713 Details : Take import reference before releasing llog record semaphore
2718 Description: Wrong request locking in request set processing
2719 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
2720 imp_delayed_list, in this place should be used imp_lock.
2723 Frequency : when reconnection
2725 Description: Grant Leak when osc reconnect to OST
2726 Details : When osc reconnect ost, OST(filter) should check whether it
2727 should grant more space to client by comparing fed_grant and
2728 cl_avail_grant, and return the granted space to client instead
2729 of "new granted" space, because client will call osc_init_grant
2730 to update the client grant space info.
2733 Frequency : when client reconnect to OST
2735 Description: Grant Leak when osc do resend and replay bulk write
2736 Details : When osc reconnect to OST, OST(filter)should clear grant info of
2737 bulk write request, because the grant info while be sync between
2738 OSC and OST when reconnect, and we should ignore the grant info
2739 these of resend/replay write req.
2744 Description: Grant space more than avaiable left space sometimes.
2745 Details : When then OST is about to be full, if two bulk writing from
2746 different clients came to OST. Accord the avaliable space of the
2747 OST, the first req should be permitted, and the second one
2748 should be denied by ENOSPC. But if the seconde arrived before
2749 the first one is commited. The OST might wrongly permit second
2750 writing, which will cause grant space > avaiable space.
2753 Frequency : when client is evicted
2755 Description: Grant might be wrongly erased when osc is evicted by OST
2756 Details : when the import is evicted by server, it will fork another
2757 thread ptlrpc_invalidate_import_thread to invalidate the
2758 import, where the grant will be set to 0. While the original
2759 thread will update the grant it got when connecting. So if
2760 the former happened latter, the grant will be wrongly errased
2761 because of this race.
2766 Description: Checking Stale with correct fid
2767 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
2768 to check whether it is stale, because sometimes, we want the
2769 enqueue happened anyway, and op_data.fid2 will not be initialized.
2771 Severity : enhancement
2773 Description: update patchless client
2774 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
2777 Frequency : only with 2.4 kernel
2779 Description: random memory corruption
2780 Details : size of struct ll_inode_info is to big for union inode.u and this
2781 can be cause of random memory corruption.
2786 Description: Memory leak in recovery
2787 Details : Lov_mds_md was not free in an error handler in mds_create_object.
2788 It should also check obd_fail before fsfilt_start, otherwise if
2789 fsfilt_start return -EROFS,(failover mds during mds recovery).
2790 then the req will return with repmsg->transno = 0 and rc = EROFS.
2791 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
2792 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
2793 be freed no matter whether fsfilt_commit success or not.
2796 Frequency : only with huge count clients
2798 Description: Prevents from taking the superblock lock in llap_from_page for
2800 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
2801 for llap_from_page call in ll_removepage prevents from taking the
2802 superblock lock for a soon died page.
2807 Description: Not check open intent error before release open handle
2808 Details : in some rare cases, the open intent error is not checked before
2809 release open handle, which may cause
2810 ASSERTION(open_req->rq_transno != 0), because it tries to release
2811 the failed open handle.
2816 Description: Set cat log bitmap only after create log success.
2817 Details : in some rare cases, the cat log bitmap is set too early. and it
2818 should be set only after create log success.
2822 Description: Accessing a block bevice can re-enable I/O when Lustre is
2823 tearing down a device.
2824 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
2829 Description: service threads may hog cpus when there are a lot of requests
2830 Details : Insert cond_resched to give other threads a chance to use some CPU
2835 Description: the cat log was not initialized in recovery
2836 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
2837 unlink log on mds will not be initialized until mds post
2838 recovery. And also in mds post recovery, the unlink log will
2839 initialization will be done asynchronausly, so there will be race
2840 between add unlink log and unlink log initialization.
2844 Description: brw_stats were being printed incorrectly
2845 Details : brw_stats were being printed as log2 but all of them were not
2846 recorded as log2. Also remove some code duplication arising from
2847 filter_tally_{read,write}.
2851 Frequency : rare, only in recovery.
2852 Description: ASSERTION(req->rq_type != LI_POISON) failed
2853 Details : imp_lock should be held while iterating over imp_sending_list for
2854 prevent destroy request after get timeout in ptlrpc_queue_wait.
2858 Description: replay-single.sh test 52 fails
2859 Details : A lock's skiplist need to be cleanup when it being unlinked
2860 from its resource list.
2864 Description: Short directio read returns full requested size rather than
2866 Details : Direct I/O operations should return actual amount of bytes
2867 transferred rather than requested size.
2869 Severity : enhancement
2871 Description: metadata RPC reduction (e.g. for rm performance)
2872 Details : decrease the amount of synchronous RPC between clients and servers
2873 by canceling conflicing lock before the operation on the client
2874 and packing thier handles into the main operation RPC to server.
2876 Severity : enhancement
2878 Description: Async OSC create to avoid the blocking unnecessarily.
2879 Details : If a OST has no remain object, system will block on the creating
2880 when need to create a new object on this OST. Now, ways use
2881 pre-created objects when available, instead of blocking on an
2882 empty osc while others are not empty. If we must block, we block
2883 for the shortest possible period of time.
2887 Description: improve handling recoverable errors
2888 Details : if request processig with error which can be recoverable on server
2889 request should be resend, otherwise page released from cache and
2892 Severity : enhancement
2894 Description: refine locking for avoid write wrong info into lov_objid file
2895 Details : fix possible races with add new target and write/update data in
2898 --------------------------------------------------------------------------------
2900 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
2905 Frequency : on some architectures
2907 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
2908 Details : Uninitialized flags sometimes cause configuration commands to
2912 Frequency : patchless clients only
2914 Description: missing __iget() symbol export
2915 Details : The __iget() symbol export is missing. To avoid the need for
2916 this on patchless clients the deathrow inode reaper is turned
2917 off, and we depend on the VM to clean up old inodes. This
2918 dependency was during via the fix for bug 12181.
2920 --------------------------------------------------------------------------------
2922 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
2924 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
2925 INTEROPERATE with older versions automatically. In many cases a
2926 special upgrade step is needed. Please read the
2927 user documentation before upgrading any part of a live system.
2928 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
2929 version will not interoperate with 1.6 betas before beta5 (1.5.95).
2930 * WARNING: Lustre configuration and startup changes are required with
2931 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
2935 Severity : enhancement
2937 Description: MountConf
2938 Details : Lustre configuration is now managed via mkfs and mount
2939 commands instead of lmc and lconf. New obd types (MGS, MGC)
2940 are added for dynamic configuration management. See
2941 https://mail.clusterfs.com/wikis/lustre/MountConf for
2944 Severity : enhancement
2946 Description: dynamic OST addition
2947 Details : OSTs can now be added to a live filesystem
2949 Severity : enhancement
2951 Description: startup order invariance
2952 Details : MDTs and OSTs can be started in any order. Clients only
2953 require the MDT to complete startup.
2955 Severity : enhancement
2957 Description: parallel, asynchronous orphan cleanup
2958 Details : orphan cleanup is now performed in separate threads for each
2959 OST, allowing parallel non-blocking operation.
2961 Severity : enhancement
2963 Description: optimized stripe assignment
2964 Details : stripe assignments are now made based on ost space available,
2965 ost previous usage, and OSS previous usage, in order to try
2966 to optimize storage space and networking resources.
2968 Severity : enhancement
2970 Description: Permanently set tunables
2971 Details : All writable /proc/fs/lustre tunables can now be permanently
2972 set on a per-server basis, at mkfs time or on a live system.
2974 Severity : enhancement
2976 Description: Lustre message v2
2977 Details : Add lustre message format v2.
2979 Severity : enhancement
2981 Description: client OST exclusion list
2982 Details : Clients can be started with a list of OSTs that should be
2983 declared "inactive" for known non-responsive OSTs.
2987 Description: SPEC SFS validation failure on NFS v2 over lustre.
2988 Details : Changes the blocksize for regular files to be 2x RPC size,
2989 and not depend on stripe size.
2991 Severity : enhancement
2993 Description: Multiple MD RPCs in flight.
2994 Details : Further unserialise some read-only MDS RPCs - learn about intents.
2995 To avoid overly-overloading MDS, introduce a limit on number of
2996 MDS RPCs in flight for a single client and add /proc controls
2997 to adjust this limit.
2999 Severity : enhancement
3001 Description: client read/write statistics
3002 Details : Add client read/write call usage stats for performance
3003 analysis of user processes.
3004 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
3005 file access. extents_stats shows chunk size distribution.
3006 extents_stats_per_process show chunk size distribution per
3009 Severity : enhancement
3011 Description: mds statistics
3012 Details : Add detailed mds operations statistics in
3013 /proc/fs/lustre/mds/*/stats.
3017 Description: Failure of copying files with lustre special EAs.
3018 Details : Client side always return success for setxattr call for lustre
3019 special xattr (currently only "trusted.lov").
3024 Description: Refcount LNET uuids
3025 Details : The global LNET uuid list grew linearly with every startup;
3026 refcount repeated list entries instead of always adding to
3029 Severity : enhancement
3031 Description: Dynamic service threads
3032 Details : Within a small range, start extra service threads
3033 automatically when the request queue builds up.
3036 Frequency : mixed-endian client/server environments
3038 Description: mixed-endian crashes
3039 Details : The new msg_v2 system had some failures in mixed-endian
3042 Severity : enhancement
3044 Description: Easy OST removal
3045 Details : OSTs can be permanently deactivated with e.g. 'lctl
3046 conf_param lustre-OST0001.osc.active=0'
3048 Severity : enhancement
3050 Description: MGS proc entries
3051 Details : Added basic proc entries for the MGS showing what filesystems
3054 Severity : enhancement
3056 Description: provide MGS failover
3057 Details : Added config lock reacquisition after MGS server failover.
3059 Severity : enhancement
3061 Description: add Linux 2.4 support
3062 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
3066 Description: a large application tries to do I/O to the same resource and dies
3067 in the middle of it.
3068 Details : Check the req->rq_arrival time after the call to
3069 ost_brw_lock_get(), but before we do anything about
3070 processing it & sending the BULK transfer request. This
3071 should help move old stale pending locks off the queue as
3072 quickly as obd_timeout.
3075 Frequency : when an incorrect nid is specified during startup
3077 Description: ptlrpc connect to non-existant node causes kernel crash
3078 Details : LNET can't be re-entered from an event callback, which
3079 happened when we expire a message after the export has been
3080 cleaned up. Instead, hand the zombie cleanup off to another
3083 Severity : enhancement
3085 Description: plain/inodebits lock performance improvement
3086 Details : Grouping plain/inodebits in granted list by their request modes
3087 and bits policy, thus improving the performance of search through
3091 Frequency : only if OST filesystem is corrupted
3093 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3094 Details : for a short time RPCs with bulk IO are in the replay list,
3095 but replay of bulk IOs is unimplemented. If the OST filesystem
3096 is corrupted due to disk cache incoherency and then replay is
3097 started it is possible to trip an assertion. Avoid putting
3098 committed RPCs into the replay list at all to avoid this issue.
3101 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3103 Description: System hang on startup
3104 Details : This bug allowed the liblustre (e.g. catamount) client to
3105 return to the app before handling all startup RPCs. This
3106 could leave the node unresponsive to lustre network traffic
3107 and manifested as a server ptllnd timeout.
3109 Severity : enhancement
3111 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3112 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
3113 boolean to control whether to print peer debug info when a
3114 client's RPC times out.
3117 Frequency : only for kernels with patches from Lustre below 1.4.3
3119 Description: Remove old rdonly API
3120 Details : Remove old rdonly API which unused from at least lustre 1.4.3
3123 Frequency : only for devices with external journals
3125 Description: Set external device read-only also
3126 Details : During a commanded failover stop, we set the disk device
3127 read-only while the server shuts down. We now also set any
3128 external journal device read-only at the same time.
3131 Frequency : when upgrading from 1.4 while trying to change parameters
3133 Description: The wrong (new) MDC name was used when setting parameters for
3134 upgraded MDT's. Also allows changing of OSC (and MDC)
3135 parameters if --writeconf is specified at tunefs upgrade time.
3138 Frequency : when setting specific ost indicies
3140 Description: QOS code breaks on skipped indicies
3141 Details : Add checks for missing OST indicies in the QOS code, so OSTs
3142 created with --index need not be sequential.
3144 Severity : enhancement
3146 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3147 Details : The uninit_groups feature works in conjunction with the kernel
3148 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3149 up the pass1 processing of e2fsck. This is a read-only feature
3150 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3151 mount filesystems that have had this feature enabled.
3153 Severity : enhancement
3155 Description: Improve multi-block allocation algorithm to avoid fragmentation
3156 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3157 allocation locality and avoid filesystem fragmentation.
3159 ------------------------------------------------------------------------------
3161 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
3163 * Support for kernels:
3164 2.4.21-47.0.1.EL (RHEL 3)
3165 2.6.5-7.283 (SLES 9)
3166 2.6.9-42.0.10.EL (RHEL 4)
3167 2.6.12.6 vanilla (kernel.org)
3168 2.6.16.27-0.9 (SLES 10)
3169 * Recommended e2fsprogs version: 1.39.cfs5
3171 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3175 Frequency : occasional, depends on client load and configuration
3176 Bugzilla : 12181, 12203
3177 Description: data loss for recently-modified files
3179 Details : In some cases it is possible that recently written or created
3180 files may not be written to disk in a timely manner (this should
3181 normally be within 30s unless client IO load is very high).
3182 The problem appears as zero-length files or files that are a
3183 multiple of 1MB in size after a client crash or client eviction
3184 that are missing data at the end of the file.
3186 This problem is more likely to be hit on clients where files are
3187 repeatedly created and unlinked in the same directory, clients
3188 have a large amount of RAM, have many CPUs, the filesystem has
3189 many OSTs, the clients are rebooted frequently, and/or the files
3190 are not accessed by other nodes after being written.
3192 The presence of the problem can be detected by looking at
3193 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
3194 smaller than the second (nr_unused) then dirty files will not
3195 be flushed automatically to disk. "sync; sleep 10" should be
3196 run several times on the node before unmounting it to update
3197 Lustre (this is also safe to run on nodes without this problem).
3199 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
3200 that can cause this same problem, so customers using that kernel
3201 also need to update the kernel in addition to Lustre. In order
3202 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
3204 It is normal that files written just before a client crash (less
3205 than 30s) may not yet have been flushed to disk, even for local
3209 Frequency : frequent on thin XT3 nodes
3211 Description: UUID collision on thin XT3 Linux nodes
3212 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
3213 randomly, since we relied on an insufficiently-seeded PRNG.
3218 Description: OSS hangs after "All ost request buffers busy"
3219 Details : A deadlock between quota and journal operations caused OSS
3220 hangs after printing "All ost request buffers busy."
3223 Frequency : always on liblustre builds
3225 Description: Cleanup compiler warnings on liblustre
3228 Frequency : always on liblustre builds on XT3
3230 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
3234 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
3235 > 32000 subdirectories
3236 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
3237 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
3238 subdirectory count crosses 32000. This will aid e2fsck to
3239 correctly handle more than 32000 subdirectories.
3242 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3244 Description: System hang on startup
3245 Details : This bug allowed the liblustre (e.g. catamount) client to
3246 return to the app before handling all startup RPCs. This
3247 could leave the node unresponsive to lustre network traffic
3248 and manifested as a server ptllnd timeout.
3250 Severity : enhancement
3252 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3253 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
3254 boolean to control whether to print peer debug info when a
3255 client's RPC times out.
3260 Description: make O_SYNC working on 2.6 kernels
3261 Details : 2.6 kernels use different method for mark pages for write,
3262 so need add a code to lustre for O_SYNC work.
3267 Description: Failure to close file and release space on NFS
3268 Details : Put inode details into lock acquired in ll_intent_file_open.
3269 Use mdc_intent_lock in ll_intent_open to properly
3270 detect all kind of errors unhandled by mdc_enqueue.
3275 Description: proc file read during shutdown sometimes raced obd removal,
3277 Details : Add lock to prevent obd access after proc file removal.
3280 Frequency : Only for files larger than 4GB on 32-bit clients.
3282 Description: improperly doing page alignment of locks
3283 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
3284 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3289 Description: RPCs being resent when they shouldn't be
3290 Details : Some RPCs that should not be resent are being resent. This
3291 can cause inconsistencies in the RPC state machine. Do not
3292 resend such requests.
3295 Frequency : rare, only with NFS export
3297 Description: Crash on NFS re-export node
3298 Details : under very unusual load conditions an assertion is hit in
3299 ll_intent_file_open()
3302 Frequency : only if OST filesystem is corrupted
3304 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3305 Details : for a short time RPCs with bulk IO are in the replay list,
3306 but replay of bulk IOs is unimplemented. If the OST filesystem
3307 is corrupted due to disk cache incoherency and then replay is
3308 started it is possible to trip an assertion. Avoid putting
3309 committed RPCs into the replay list at all to avoid this issue.
3314 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3315 Details : Large single O_DIRECT read and write calls can fail to allocate
3316 a sufficiently large buffer to process the request. In case of
3317 allocation failure the allocation is retried with a smaller
3318 buffer and broken into smaller requests.
3320 Severity : enhancement
3322 Description: Add -o localflock option to simulate old noflock behaviour.
3323 Details : This will achieve local-only flock/fcntl locks coherentness.
3328 Description: versioning check is incomplete
3329 Details : Checking the version difference of client vs. server, report
3330 error if the gap is too big.
3335 Description: add support PG_writeback bit
3336 Details : add support for PG_writeback bit for Lustre, for more carefull
3337 work with page cache in 2.6 kernel. This also fix some deadlocks
3338 and remove hack for work O_SYNC with 2.6 kernel.
3340 Severity : enhancement
3342 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3343 Details : The uninit_groups feature works in conjunction with the kernel
3344 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3345 up the pass1 processing of e2fsck. This is a read-only feature
3346 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3347 mount filesystems that have had this feature enabled.
3349 Severity : enhancement
3351 Description: Improve multi-block allocation algorithm to avoid fragmentation
3352 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3353 allocation locality and avoid filesystem fragmentation.
3355 ------------------------------------------------------------------------------
3357 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3359 * Support for kernels:
3360 2.6.9-42.0.3.EL (RHEL 4)
3361 2.6.5-7.276 (SLES 9)
3362 2.4.21-47.0.1.EL (RHEL 3)
3363 2.6.12.6 vanilla (kernel.org)
3364 2.6.16.21-0.8 (SLES10)
3365 * Recommended e2fsprogs version: 1.39.cfs2-0
3367 * The backwards-compatible /proc/sys/portals symlink has been removed
3368 in this release. Before upgrading, please ensure that you change
3369 any configuration scripts or /etc/sysctl.conf files that access
3370 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3371 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3372 in advance of the upgrade on any system running Lustre 1.4.6 or
3373 newer, since /proc/sys/lnet was added in that version.
3374 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3378 Frequency : only when quota is used
3380 Description: avoid scanning export list for quota master
3381 Details : Change the algorithms to avoid scanning export list in order
3382 to improve the efficiency.
3385 Frequency : MDS failover only, very rarely
3387 Description: "went back in time" messages on mds failover
3388 Details : The greatest transno may be lost when the current operation
3389 finishes with an error (transno==0) and the client's last_rcvd
3390 record is over-written. Save the greatest transno in the
3391 mds_last_transno for this case.
3394 Frequency : always for specific kernels and striping counts
3396 Description: client may get "Matching packet too big" without ACL support
3397 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
3398 when trying to access files in certain configurations. The
3399 clients should in fact be denied when mounting because they do
3400 not understand ACLs.
3403 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
3405 Description: many clients connecting with IO in progress causes connect timeouts
3406 Details : Avoid synchronous journal commits to avoid delays caused by many
3407 clients connecting/disconnecting when bulk IO is in progress.
3408 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
3409 OST_IO_PORTAL to avoid delays behind potentially many pending
3413 Frequency : occasionally with multiple writers to a single file
3415 Description: shared writes to file may result in wrong size reported by stat()
3416 Details : Allow growing of kms when extent lock is cancelled
3419 Frequency : always with random mmap IO to multi-striped file
3421 Description: mmap write might be lost if we are writing to a 'hole' in stripe
3422 Details : Only if the hole is at the end of OST object so that kms is too
3423 small. Fix is to increase kms accordingly in ll_nopage.
3426 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
3428 Description: writes to a missing object would leak memory on the OST
3429 Details : If there is an inconsistency between the MDS and OST filesystems,
3430 such that the MDS references an object that doesn't exist, writes
3431 to that object will leak memory due to incorrect cleanup in the
3432 error handling path, eventually running out of memory on the OST.
3437 Description: Creating too long symlink causes lustre errors
3438 Details : Check symlink and name lengths before sending requests to MDS.
3441 Frequency : only if flock is enabled (not on by default)
3443 Description: posix locks not released on fd closure on 2.6.9+
3444 Details : We failed to add posix locks to list of inode locks on 2.6.9+
3445 kernels, this caused such locks not to be released on fd close and
3446 then assertions on fs unmount about still used locks.
3449 Frequency : MDS failover only, very rarely
3451 Description: clients may get ASSERTION(granted_lock != NULL)
3452 Details : When request was taking a long time, and a client was resending
3453 a getattr by name lock request. The were multiple lock requests
3454 with the same client lock handle and
3455 mds_getattr_name->fixup_handle_for_resent_request found one of the
3456 lock handles but later failed with ASSERTION(granted_lock != NULL).
3461 Description: handle->h_buffer_credits > 0, assertion failure
3462 Details : h_buffer_credits is zero after truncate, causing assertion
3463 failure. This patch extends the transaction or creates a new
3467 Frequency : NFS re-export or patchless client
3468 Bugzilla : 11179, 10796
3469 Description: Crash on NFS re-export node (__d_move)
3470 Details : We do not want to hash the dentry if we don't have a lock.
3471 But if this dentry is later used in d_move, we'd hit uninitialised
3472 list head d_hash, so we just do this to init d_hash field but
3473 leave dentry unhashed.
3476 Frequency : NFS re-export or patchless client
3478 Description: NFS exports has problem with symbolic link
3479 Details : lustre client didn't properly install dentry when re-exported
3480 to NFS or running patchless client.
3483 Frequency : NFS re-export or patchless client
3485 Description: Various nfs/patchless fixes.
3486 Details : fixes reuse disconected alias for lookup process - this fixes
3487 warning "find_exported_dentry: npd != pd",
3488 fix permission error with open files at nfs.
3489 fix apply umask when do revalidate.
3492 Frequency : occasional
3494 Description: Crash on NFS re-export node
3495 Details : calling clear_page() on the wrong pointer triggered oops in
3496 generic_mapping_read().
3499 Frequency : rarely, using O_DIRECT IO
3501 Description: unaligned directio crashes client with LASSERT
3502 Details : check for unaligned buffers before trying any requests.
3505 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
3507 Description: stale data returned from RAID cache
3508 Details : If only a small amount of IO is done to the RAID device before
3509 reading it again it is possible to get stale data from the RAID
3510 cache instead of reading it from disk.
3513 Frequency : always for sles10 kernel
3515 Description: sles10 support
3516 Details : ll_follow_link: compile fixes and using of nd_set_link
3517 under newer kernels.
3520 Frequency : depends on arch, kernel and compiler version, always on sles10
3523 Description: recursive or deep enough symlinks cause stack overflow
3524 Details : getting rid of large stack-allocated variable in
3528 Frequency : depends on hardware
3530 Description: lustre write performance loss in the SLES10 kernel
3531 Details : the performance loss is caused by using of write barriers in the
3532 ext3 code. The SLES10 kernel turns barrier support on by
3533 default. The fix is to undo that change for ldiskfs.
3535 ------------------------------------------------------------------------------
3537 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
3539 * Support for kernels:
3540 2.6.9-42.0.3EL (RHEL 4)
3541 2.6.5-7.276 (SLES 9)
3542 2.4.21-47.0.1.EL (RHEL 3)
3543 2.6.12.6 vanilla (kernel.org)
3547 Frequency : quota enabled and large files being deleted
3549 Description: releasing more than 4GB of quota at once hangs OST
3550 Details : If a user deletes more than 4GB of files on a single OST it
3551 will cause the OST to spin in an infinite loop. Release
3552 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3557 Description: statfs data retrieved from /proc may be stale or zero
3558 Details : When reading per-device statfs data from /proc, in the
3559 {kbytes,files}_{total,free,avail} files, it may appear
3560 as zero or be out of date.
3563 Frequency : systems with MD RAID1 external journal devices
3565 Description: lconf's call to blkid is confused by RAID1 journal devices
3566 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3567 of returning all block devices that match the journal UUID.
3570 Frequency : always, for aggregate stripe size over 4GB
3572 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
3573 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
3574 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
3577 Frequency : always if "lfs find" used on a local file/directory
3579 Description: "lfs find" segfaults if used on a local file/directory
3580 Details : The case where a directory component was not specified wasn't
3581 handled correctly. Handle this properly.
3584 Frequency : always on ppc64
3586 Description: the write to an ext3 filesystem mounted with mballoc got stuck
3587 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3588 perform endianness conversion.
3591 Frequency : rarely (truncate to non-zero file size after write under load)
3592 Bugzilla : 10730, 10687
3593 Description: Files padded with zeros to next 4K multiple
3594 Details : With filesystems mounted using the "extents" option (2.6 kernels)
3595 it is possible that files that are truncated to a non-zero size
3596 immediately after being written are filled with zero bytes beyond
3597 the truncated size. No file data is lost.
3599 Severity : enhancement
3601 Description: Allow recovery/failover for liblustre clients.
3602 Details : liblustre clients were unaware of failover configurations until
3605 Severity : enhancement
3607 Description: user file locks should fail when not mounting with flock option
3608 Details : Set up an error-returning stub in ll_file_operations.lock field
3609 to prevent incorrect behaviour when client is mounted without
3610 flock option. Also, set up properly f_op->flock field for
3614 Frequency : always on ia64
3616 Description: "lfs df" loops on printing out MDS statfs information
3617 Details : The obd_ioctl_data was not initialized and in some systems
3618 this caused a failure during the ioctl that did not return
3619 an error. Initialize the struct and return an error on failure.
3622 Frequency : SLES 9 only
3624 Description: Error of copying files with lustre special EAs as root
3625 Details : Client side always return success for setxattr call for lustre
3626 special xattr (currently only "trusted.lov").
3629 Frequency : rarely on clusters with both ia64+i386 clients
3631 Description: ia64+i686 clients doing shared IO on the same file may LBUG
3632 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
3633 clients are doing concurrent writes to the same file it is
3634 possible that the ia64 clients may LASSERT because the OST
3635 extent locks are not PAGE_SIZE aligned. Ensure that grown
3636 locks are always aligned on the request boundary.
3639 Frequency : specific use, occasional
3641 Description: Overwriting in use executable truncates on-disk binary image
3642 Details : If one node attempts to overwrite an executable in use by
3643 another node, we now correctly return ETXTBSY instead of
3644 truncating the file.
3646 Severity : enhancement
3648 Description: Async OSC create to avoid the blocking unnecessarily.
3649 Details : If a OST has no remain object, system will block on the creating
3650 when need to create a new object on this OST. Now, ways use
3651 pre-created objects when available, instead of blocking on an
3652 empty osc while others are not empty. If we must block, we block
3653 for the shortest possible period of time.
3658 Description: chmod on Lustre root is propagated to other clients
3659 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
3660 invalid by the follow_mount time.
3663 Frequency : liblustre clients only
3665 Description: Race in 'instant cancel' lock handling could lead to such locks
3666 never to be granted in case of SMP MDS
3667 Details : Do not destroy not yet granted but cbpending locks in
3671 Frequency : replay/resend of open
3673 Description: non null lock assetion failure in mds_intent_policy
3674 Details : Trying to replay/resend lockless open requests resulted in
3675 mds_open() returning 0 with no lock. Now it sets a flag if
3676 a lock is going to be returned.
3678 Severity : enhancement
3680 Description: Checksum enhancements
3681 Details : New checksum enhancements allow for resending RPCs that failed
3684 Severity : enhancement
3686 Description: Tunables on number of dirty pages in cacche
3687 Details : Allow to set limit on number of dirty pages cached.
3692 Description: client crash on unmount - lock still has references
3693 Details : In some error handling cases it was possible to leak a lock
3694 reference on a client while accessing a file. This was not
3695 harmful to the client during operation, but would cause the
3696 client to crash when the filesystem is unmounted.
3699 Frequency : specific case, rare
3701 Description: ETXTBSY on mds though file not in use by client
3702 Details : ETXTBSY is no longer incorrectly returned when attempting to
3703 chmod or chown a directory that the user previously tried to
3704 execute or a currently-executing binary.
3707 Frequency : extremely rare except on liblustre-based clients
3709 Description: Lustre space not freed when files are deleted
3710 Details : Clean up open-unlinked files after client eviction. Previously
3711 the unlink was skipped and the files remained as orphans.
3716 Description: OST failure "would be an LBUG" in waiting_locks_callback()
3717 Details : In some cases it was possible to send a blocking callback to
3718 a client doing a glimpse, even though that client didn't get
3719 a lock granted. When the glimpse lock is cancelled on the OST
3720 the freed lock is left on the waiting list and corrupted the list.
3723 Frequency : all core dumps
3725 Description: Broke core dumps to lustre
3726 Details : Negative dentry may be unhashed if parent does not have UPDATE
3727 lock, but some callers, e.g. do_coredump, expect dentry to be
3728 hashed after successful create, hash it in ll_create_it.
3730 ------------------------------------------------------------------------------
3732 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
3734 * Support for kernels:
3735 2.6.9-42.0.2.EL (RHEL 4)
3736 2.6.5-7.276 (SLES 9)
3737 2.4.21-40.EL (RHEL 3)
3738 2.6.12.6 vanilla (kernel.org)
3742 Frequency : always on RHEL 3
3744 Description: Number of open files grows over time
3745 Details : The number of open files grows over time, whether or not
3746 Lustre is started. This was due to a filp leak introduced
3747 by one of our kernel patches.
3749 ------------------------------------------------------------------------------
3751 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
3753 * Support for kernels:
3754 2.6.9-42.EL (RHEL 4)
3755 2.6.5-7.276 (SLES 9)
3756 2.4.21-40.EL (RHEL 3)
3757 2.6.12.6 vanilla (kernel.org)
3762 Bugzilla : 5719, 9635, 9792, 9684
3763 Description: OST (or MDS) trips assertions in (re)connection under heavy load
3764 Details : If a server is under heavy load and cannot reply to new
3765 connection requests before the client resends the (re)connect,
3766 the connection handling code can behave badly if two service
3767 threads are concurrently handing separate (re)connections from
3768 the same client. Add better locking to the connection handling
3769 code, and ensure that only a single connection will be processed
3770 for a given client UUID, even if the lock is dropped.
3772 Severity : enhancement
3774 Description: add TCP zero-copy support to kernel
3775 Details : Add support to the kernel TCP stack to allow zero-copy bulk
3776 sends if the hardware supports scatter-gather and checksumming.
3777 This allows socklnd to do client-write and server-read more
3778 efficiently and reduce CPU utilization from skbuf copying.
3781 Frequency : only if NFS exporting from client
3783 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
3784 Details : If setting attributes on a file created under NFS that had
3785 never been opened it would be possible to oops the client
3786 if the file had no objects.
3789 Frequency : always for liblustre
3791 Description: liblustre client does MDS+OSTs setattr RPC for each write
3792 Details : When doing a write from a liblustre client, the client
3793 incorrectly issued an RPC to the MDS and each OST the file was
3794 striped over in order to update the timestamps. When writing
3795 with small chunks and many clients this could overwhelm the MDS
3796 with RPCs. In all cases it would slow down the write because
3797 these RPCs are unnecessary.
3799 Severity : enhancement
3801 Description: allow number of MDS service threads to be changed at module load
3802 Details : It is now possible to change the number of MDS service threads
3803 running. Adding "options mds mds_num_threads={N}" to the MDS's
3804 /etc/modprobe.conf will set the number of threads for the next
3805 time Lustre is restarted (assuming the "mds" module is also
3806 reloaded at that time). The default number of threads will
3807 stay the same, 32 for most systems.
3812 Description: OST crash if filesystem is unformatted or corrupt
3813 Details : If an OST is started on a device that has never been formatted
3814 or if the filesystem is corrupt and cannot even mount then the
3815 error handling cleanup routines would dereference a NULL pointer.
3820 Description: NULL pointer deref in llap_from_page.
3821 Details : get_cache_page_nowait can return a page with NULL (or otherwise
3822 incorrect) mapping if the page was truncated/reclaimed while it was
3823 searched for. Check for this condition and skip such pages when
3824 doing readahead. Introduce extra check to llap_from_page() to
3825 verify page->mapping->host is non-NULL (so page is not anonymous).
3828 Frequency : Sometimes when using sys_sendfile
3830 Description: "page not covered by a lock" warnings from ll_readpage
3831 Details : sendfile called ll_readpage without right page locks present.
3832 Now we introduced ll_file_sendfile that does necessary locking
3833 around call to generic_file_sendfile() much like we do in
3837 Frequency : with certain MDS communication failures at client mount time
3839 Description: NULL pointer deref after failed client mount
3840 Details : a client connection request may delayed by the network layer
3841 and not be sent until after the PTLRPC layer has timed out the
3842 request. If the client fails the mount immediately it will try
3843 to clean up before the network times out the request. Add a
3844 reference from the request import to the obd device and delay
3845 the cleanup until the network drops the request.
3848 Frequency : occasionally during client (re)connect
3850 Description: assertion failure during client (re)connect
3851 Details : processing a client connection request may be delayed by the
3852 client or server longer than the client connect timeout. This
3853 causes the client to resend the connection request. If the
3854 original connection request is replied in this interval, the
3855 client may trip an assertion failure in ptlrpc_connect_interpret()
3856 which thought it would be the only running connect process.
3859 Frequency : only with obd_echo servers and clients that are rebooted
3861 Description: kernel BUG accessing uninitialized data structure
3862 Details : When running an obd_echo server it did not start the ping_evictor
3863 thread, and when a client was evicted an uninitialized data
3864 structure was accessed. Start the ping_evictor in the RPC
3865 service startup instead of the OBD startup.
3867 Severity : enhancement
3868 Bugzilla : 10193 (patchless)
3869 Description: Remove dependency on various unexported kernel interfaces.
3870 Details : No longer need reparent_to_init, exit_mm, exit_files,
3871 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
3874 Frequency : rare (only users of deprecated and unsupported LDAP config)
3876 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
3877 Details : LDAP apparently contains 'lustreName' attributes instead of
3878 'name'. A simple remapping of the name is sufficient.
3881 Frequency : rare (only with non-default dump_on_timeout debug enabled)
3883 Description: waiting_locks_callback trips kernel BUG if client is evicted
3884 Details : Running with the dump_on_timeout debug flag turned on makes
3885 it possible that the waiting_locks_callback() can try to dump
3886 the Lustre kernel debug logs from an interrupt handler. Defer
3887 this log dumping to the expired_lock_main() thread.
3889 Severity : enhancement
3891 Description: Support NFS exporting on 2.6 kernels.
3892 Details : Implement non-rawops metadata methods for NFS server to use without
3893 changing NFS server code.
3896 Frequency : very rare (synthetic metadata workload only)
3898 Description: two racing renames might cause an MDS thread to deadlock
3899 Details : Running the "racer" program may cause one MDS thread to rename
3900 a file from being the source of a rename to being the target of
3901 a rename at exactly the same time that another thread is doing
3902 so, and the second thread has already enqueued these locks after
3903 doing a lookup of the target and is trying to relock them in
3904 order. Ensure that we don't try to re-lock the same resource.
3907 Frequency : only very large systems with liblustre clients
3909 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
3910 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
3911 all OSTs in parallel. This allows the eviction of stale liblustre
3912 clients to proceed much faster than if they were done in series,
3913 and also offers similar improvements for other set_info RPCs.
3918 Description: excessive CPU usage during initial read phase on client
3919 Details : During the initial read phase on a client, it would agressively
3920 retry readahead on the file, consuming too much CPU and impacting
3921 performance (since 1.4.5.8). Improve the readahead algorithm
3922 to avoid this, and also improve some other common cases (read
3923 of small files in particular, where "small" is files smaller than
3924 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
3929 Description: MDS crash when receiving packet with unknown intent.
3930 Details : Do not LBUG in unknown intent case, just return -EFAULT
3932 Severity : enhancement
3933 Bugzilla : 9293, 9385
3934 Description: MDS RPCs are serialised on client. This is unnecessary for some.
3935 Details : Do not serialize getattr (non-intent version) and statfs.
3938 Frequency : occasional, when OST network is overloaded/intermittent
3940 Description: client evicted by OST after bulk IO timeout
3941 Details : If a client sends a bulk IO request (read or write) the OST
3942 may evict the client if it is unresposive to its data GET/PUT
3943 request. This is incorrect if the network is overloaded (takes
3944 too long to transfer the RPC data) or dropped the OST GET/PUT
3945 request. There is no need to evict the client at all, since
3946 the pinger and/or lock callbacks will handle this, and the
3947 client can restart the bulk request.
3950 Frequency : Always when mmapping file with no objects
3952 Description: client crashes when mmapping file with no objects
3953 Details : Check that we actually have objects in a file before doing any
3954 operations on objects in ll_vm_open, ll_vm_close and
3960 Description: Request leak when working with deleted CWD
3961 Details : Introduce advanced request refcount tracking for requests
3962 referenced from lustre intent.
3964 Severity : Enhancement
3966 Description: Cache open file handles on client.
3967 Details : MDS now will return special lock along with openhandle, if
3968 requested and client is allowed to hold openhandle, even if unused,
3969 until such a lock is revoked. Helps NFS a lot, since NFS is opening
3970 closing files for every read/write openration.
3972 Severity : Enhancement
3974 Description: Cache open negative dentries on client when possible.
3975 Details : Guard negative dentries with UPDATE lock on parent dir, drop
3976 negative dentries on lock revocation.
3981 Description: Remounting a client read-only wasn't possible with a zconf mount
3982 Details : It wasn't possible to remount a client read-only with llmount.
3984 Severity : enhancement
3985 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
3986 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
3987 MPICH is included to add Lustre-specific ADIO interfaces.
3988 This is based closely on the UFS ADIO layer and only differs
3989 in file creation, in order to allow the OST striping to be set.
3990 This is user-contributed code and not supported by CFS.
3995 Description: extended inode attributes (immutable, append-only) work improperly
3996 when 2.4 and 2.6 kernels are used on client/server or vice versa
3997 Details : Introduce kernel-independent values for these flags.
3999 Severity : enhancement
4002 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
4003 Details : Many of the /proc/ tunables can only be tuned at a megabyte
4004 granularity. Now, Fractional MB granularity is be supported,
4005 this is very useful for low memory system.
4007 Severity : enhancement
4009 Description: Getattr by fid
4010 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
4011 locks, avoids extra getattr rpc requests to MDS, allows '/' to
4012 have locks and avoids getattr rpc requests for it on every stat.
4015 Frequency : Always, for filesystems larger than 2TB
4017 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4018 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
4019 filesystems larger than 2TB to be mounted with mballoc enabled.
4022 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
4024 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4025 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
4026 filesystems larger than 2TB is mounted then the kernel will
4027 silently corrupt the start of the filesystem. CONFIG_LBD is
4028 enabled for all CFS-supported kernels, but the possibility of
4029 this happening with a modified kernel config exists.
4031 Severity : enhancement
4033 Description: add client O_DIRECT support for 2.6 kernels
4034 Details : It is now possible to do O_DIRECT reads and writes to files
4035 in the Lustre client mountpoint on 2.6 kernel clients.
4037 Severity : enhancement
4039 Description: parallel glimpse, setattr, statfs, punch, destroy requests
4040 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
4041 parallel, not waiting for response from every OST before sending
4042 a rpc to the next OST.
4047 Description: setattr vs write race when updating file timestamps
4048 Details : Client processes that update a file timestamp into the past
4049 right after writing to the file (e.g. tar) it is possible that
4050 the updated file modification time can be reset to the current
4051 time due to a race between processing the setattr and write RPC.
4053 Severity : enhancement
4055 Description: Bring 'lfs find' closer in line with regular Linux find.
4056 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
4057 -print0 options and obtains all the needed info through the lustre
4060 Severity : enhancement
4062 Description: support up to 1024 configured devices on one node
4063 Details : change obd_dev array from statically allocated to dynamically
4064 allocated structs as they are first used to reduce memory usage
4069 Description: Flush dirty partially truncated pages during truncate
4070 Details : Immediatelly flush partially truncated pages in filter_setattr,
4071 this way we completely avoid having any pages in page cache on OST
4072 and can retire ugly workarounds during writes to flush such pages.
4077 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
4078 Details : protect inode from truncation within vfs_unlink() context
4079 just take a reference before calling vfs_unlink() and release it
4080 when parent's i_sem is free.
4085 Description: last_id value checked outside lock on OST caused LASSERT failure
4086 Details : If there were multiple MDS->OST object precreate requests in
4087 flight, it was possible that the OST's last object id was checked
4088 outside a lock and incorrectly tripped an assertion. Move checks
4089 inside locks, and discard old precreate requests.
4092 Frequency : always, if extents are used on OSTs
4094 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
4095 Details : OSTs using the extents format would not zero the high 16 bits of
4096 the index physical block number. This is not a problem for any
4097 OST filesystems smaller than 16TB, and no kernels support ext3
4098 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
4099 new/modified files) and can be fixed for existing filesystems
4100 with e2fsprogs-1.39-cfs1.
4105 Description: import connection selection may be incorrect if timer wraps
4106 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
4107 import connections to be ignored if the 32-bit jiffies counter
4108 wraps. Use a 64-bit jiffies counter.
4111 Frequency : very large clusters immediately after boot
4113 Description: LNET request buffers exhausted under heavy short-term load
4114 Details : If a large number of client requests are generated on a service
4115 that has previously never seen so many requests it is possible
4116 that the request buffer growth cannot keep up with the spike in
4117 demand. Instead of dropping incoming requests, they are held in
4118 the LND until the RPC service can accept more requests.
4121 Frequency : Sometimes during replay
4123 Description: Assertion failure in ll_local_open after replay.
4124 Details : If replay happened on an open request reply before we were able
4125 to set replay handler, reply will become not swabbed tripping the
4126 assertion in ll_local_open. Now we set the handler right after
4127 recognising of open request
4130 Frequency : very rare
4132 Description: kernel reports "badness in vsnprintf"
4133 Details : Reading from the "recovery_status" /proc file in small chunks
4134 may cause a negative length in lprocfs_obd_rd_recovery_status()
4135 call to vsnprintf() (which is otherwise harmless). Exit early
4136 if there is no more space in the output buffer.
4138 Severity : enhancement
4140 Description: clear OBD RPC statistics by writing to them
4141 Details : It is now possible to clear the OBD RPC statistics by writing
4142 to the "stats" file.
4147 Description: Client mtime is not the same on different clients after utimes
4148 Details : In some cases, the client was using the utimes() syscall on
4149 a file cached on another node. The clients now validate the
4150 ctime from the MDS + OSTs to determine which one is right.
4155 Description: Inability to activate failout mode
4156 Details : lconf script incorrectly assumed that in pythong string's numeric
4157 value is used in comparisons.
4160 Frequency : always with multiple stripes per file
4162 Description: Inefficient object allocation for mutli-stripe files
4163 Details : When selecting which OSTs to stripe files over, for files with
4164 a stripe count that divides evenly into the number of OSTs,
4165 the MDS is always picking the same starting OST for each file.
4166 Return the OST selection heuristic to the original design.
4171 Description: mount failures may take full timeout to return an error
4172 Details : Under some heavy load conditions it is possible that a
4173 failed mount can wait for the full obd_timeout interval,
4174 possibly several minutes, before reporting an error.
4175 Instead return an error as soon as the status is known.
4177 Frequency : quota enabled and large files being deleted
4179 Description: releasing more than 4GB of quota at once hangs OST
4180 Details : If a user deletes more than 4GB of files on a single OST it
4181 will cause the OST to spin in an infinite loop. Release
4182 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4187 Description: statfs data retrieved from /proc may be stale or zero
4188 Details : When reading per-device statfs data from /proc, in the
4189 {kbytes,files}_{total,free,avail} files, it may appear
4190 as zero or be out of date.
4193 Frequency : systems with MD RAID1 external journal devices
4195 Description: lconf's call to blkid is confused by RAID1 journal devices
4196 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4197 of returning all block devices that match the journal UUID.
4200 Frequency : always, for aggregate stripe size over 4GB
4202 Description: assertion fails when trying to use 4GB stripe size
4203 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
4204 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
4207 Frequency : always on ppc64
4209 Description: the first write on an ext3 filesystem with mballoc got stuck
4210 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4211 perform endianness conversion.
4213 ------------------------------------------------------------------------------
4215 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4217 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
4218 INTEROPERATE with older versions automatically. Please read the
4219 user documentation before upgrading any part of a live system.
4220 * WARNING: Lustre networking configuration changes are required with
4221 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
4224 * Support for kernels:
4225 2.6.9-22.0.2.EL (RHEL 4)
4226 2.6.5-7.244 (SLES 9)
4227 2.6.12.6 vanilla (kernel.org)
4230 Severity : enhancement
4231 Bugzilla : 7981/8208
4232 Description: Introduced Lustre Networking (LNET)
4233 Details : LNET is new networking infrastructure for Lustre, it includes
4234 a reorganized network configuration mode (see the user
4235 documentation for full details) as well as support for routing
4236 between different network fabrics. Lustre Networking Devices
4237 (LNDs) for the supported network fabrics have also been
4238 created for this new infrastructure.
4240 Severity : enhancement
4241 Description: Introduced Access control lists
4242 Details : clients can set ACLs on files and directories in order to have
4243 more fine-grained permissions than the standard Unix UGO+RWX.
4244 The MDS must be started with the "-o acl" mount option.
4246 Severity : enhancement
4247 Description: Introduced filesystem quotas
4248 Details : Administrators may now establish per-user quotas on the
4251 Severity : enhancement
4253 Description: Configuration change for the XT3
4254 The PTLLND is now used to run Lustre over Portals on the XT3
4255 The configure option(s) --with-cray-portals are no longer used.
4256 Rather --with-portals=<path-to-portals-includes> is used to
4257 enable building on the XT3. In addition to enable XT3 specific
4258 features the option --enable-cray-xt3 must be used.
4263 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
4264 Details : A race between allocating a new llcd and re-getting the llcd_lock
4265 allowed another thread to grab newly-allocated llcd.
4267 Severity : enhancement
4269 Description: 2.6 OST async journal commit and locking fix to improve performance
4270 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
4271 2.6 kernels are now async as they already were in 2.4 kernels so
4272 that they can commit concurrently with the network bulk transfer.
4273 For block-allocated files the filter allocation semaphore is held
4274 to avoid filesystem fragmentation during allocation. BKL lock
4275 removed for 2.6 xattr operations where it is no longer needed.
4280 Description: lconf incorrectly determined whether two IP networks could talk
4281 Details : In some more complicated routing and multiple-network
4282 configurations, lconf will avoid trying to make a network
4283 connection to a disjoint part of the IP space. It was doing the
4284 math incorrectly for one set of cases.
4289 Description: Fix for potential infinite loop processing records in an llog.
4290 Details : If an llog record is corrupted/zeroed, it is possible to loop
4291 forever in llog_process(). Validate the llog record length
4292 and skip the remainder of the block on error.
4295 Frequency : occasional (liblustre only)
4297 Description: liblustre could not open files whose last component is a symlink
4298 Details : sysio_path_walk() would incorrectly pass the open intent to
4299 intermediate path components.
4302 Frequency : rare (liblustre only with non-standard tuning)
4303 Bugzilla : 7201 (7350)
4304 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4305 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4306 after it has been released, only do this for referenced locks
4308 Severity : enhancement
4310 Description: specifying an (invalid) directory default stripe_size of -1
4311 would reset the directory default striping
4312 Details : stripe_size -1 was used internally to signal directory stripe
4313 removal, now use "all default" to signal dir stripe removal
4314 as a directory striping of "all default" is not useful
4317 Frequency : common for large clusters running liblustre clients
4319 Description: doing an ls when liblustre clients are running is slow
4320 Details : sending a glimpse AST to a liblustre client waits for every AST
4321 to time out, as liblustre clients will not respond. Since they
4322 cannot cache data we refresh the OST lock LVB from disk instead.
4324 Severity : enhancement
4326 Description: doing an ls at the same time as file IO can be slow
4327 Details : enqueue and other "small" requests can be blocked behind many
4328 large IO requests. Create a new OST IO portal for non-IO
4329 requests so they can be processed faster.
4332 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4334 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4335 created file with mode 000
4336 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4337 allow an ftruncate() as a non-root user to a file with mode 000.
4338 Lustre now respects this flag to disable mode checks when
4339 truncating a file owned by the user
4342 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4345 Description: Revoking locks from clients that went dead or catatonic might take
4347 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4348 cancellation of such locks instant on servers without waiting for
4349 any reply from clients. Clients drops these locks when cancel
4350 notification from server is received without replying.
4353 Frequency : liblustre-only, when liblustre client dies or becomes busy
4355 Description: Doing ls on Linux clients can take a long time with active
4357 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4358 granting such locks to it in the first place if possible. Locks
4359 are taken by proxy on the OST during the read or write and
4360 dropped immediately afterward. Add connect flags handling, do
4361 not grant locks to liblustre clients for glimpse ASTs.
4363 Severity : enhancement
4365 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4366 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4367 sequential reads and behaves badly with applications doing random
4368 reads. Improve it by reading ahead at least read region, and
4369 avoiding excessive large RPC for small reads.
4371 Severity : enhancement
4373 Description: Creating more than 1000 files for a single job may cause a load
4374 imbalance on the OSTs if there are also a large number of OSTs.
4375 Details : qos_prep_create() uses an OST index reseed value that is an
4376 even multiple of the number of available OSTs so that if the
4377 reseed happens in the middle of the object allocation it will
4378 still utilize the OSTs as uniformly as possible.
4383 Description: OST or MDS may oops in ping_evictor_main()
4384 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
4385 but doesn't restart at beginning of obd_exports_timed list
4388 Severity : enhancement
4390 Description: improve by-nid export eviction on the MDS and OST
4391 Details : allow multiple exports with the same NID to be evicted at one
4392 time without re-searching the exports list.
4395 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
4397 Description: MDS may oops in groups_free()
4398 Details : in rare race conditions a newly allocated group_info struct is
4399 freed again, and this can be NULL. The 2.4 compatibility code
4400 for groups_free() checked for a NULL pointer, but 2.6 did not.
4403 Frequency : common for liblustre clients doing little filesystem IO
4404 Bugzilla : 9352, 7313
4405 Description: server may evict liblustre clients accessing contended locks
4406 Details : if a client is granted a lock or receives a completion AST
4407 with a blocking AST already set it would not reply to the AST
4408 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
4411 Frequency : lfs setstripe, only systems with more than 160 OSTs
4413 Description: unable to set striping with a starting offset beyond OST 160
4414 Details : llapi_create_file() incorrectly limited the starting stripe
4415 index to the maximum single-file stripe count.
4418 Frequency : LDAP users only
4420 Description: lconf did not handle in-kernel recovery with LDAP properly
4421 Details : lconf/LustreDB get_refs() is searching the wrong namespace
4423 Severity : enhancement
4425 Description: bind OST threads to NUMA nodes to improve performance
4426 Details : all OST threads are uniformly bound to CPUs on a single NUMA
4427 node and do their allocations there to localize memory access
4429 Severity : enhancement
4431 Description: llmount can determine client NID directly from Myrinet (GM)
4432 Details : the client NID code from gmnalnid was moved directly into
4433 llmount, removing the need to use this or specifying the
4434 client NID explicitly when mounting GM clients with zeroconf
4437 Frequency : if client is started with down MDS
4439 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
4440 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
4441 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
4442 ptlrpc_import_delayed_req() because we didn't check if the
4443 request was interrupted, and we also didn't break out of the
4444 event loop if there was no timeout
4449 Description: data loss during non-page-aligned writes to a single file from
4450 both multiple nodes and multiple threads on one node at same time
4451 Details : updates to KMS and lsm weren't protected by common lock. Resulting
4452 inconsistency led to false short-reads, that were cached and later
4453 used by ->prepare_write() to fill in partially written page,
4454 leading to data loss.
4457 Frequency : always, if lconf --abort_recovery used
4459 Description: lconf --abort_recovery fails with 'Operation not supported'
4460 Details : lconf was attempting to abort recovery on the MDT device and not
4463 Severity : enhancement
4465 Description: remove cleanup logs
4466 Details : replace lconf-generated cleanup logs with lustre internal
4467 cleanup routines. Eliminates the need for client-cleanup and
4470 Severity : enhancement
4472 Description: add support for EAs (user and system) on lustre filesystems
4473 Details : it is now possible to store extended attributes in the Lustre
4474 client filesystem, and with the user_xattr mount option it
4475 is possible to allow users to store EAs on their files also
4477 Severity : enhancement
4479 Description: Add possibility (config option) to show minimal available OST free
4481 Details : When compiled with --enable-mindf configure option, statfs(2)
4482 (and so, df) will return least minimal free space available from
4483 all OSTs as amount of free space on FS, instead of summary of
4484 free spaces of all OSTs.
4486 Severity : enhancement
4488 Description: do not expand extent locks acquired on OST-side
4489 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
4490 by server: they are not cached anyway.
4493 Frequency : when mmap is used/binaries executed from Lustre
4495 Description: Unmmap pages before throwing them away from read cache.
4496 Details : llap_shrink cache now attempts to unmap pages before discarding
4497 them (if unmapping failed - do not discard). SLES9 kernel has
4498 extra checks that trigger if this unmapping is not done first.
4503 Description: lconf didn't resolve symlinks before checking to see whether a
4504 given mountpoint was already in use
4507 Frequency : when migrating failover services
4508 Bugzilla : 6395, 9514
4509 Description: When migrating a subset of services from a node (e.g. failback
4510 from a failover service node) the remaining services would
4511 time out and evict clients.
4512 Details : lconf --force (implied by --failover) sets the global obd_timeout
4513 to 5 seconds in order to quickly disconnect, but this caused
4514 other RPCs to time out too quickly. Do not change the global
4515 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
4517 Severity : enhancement
4518 Frequency : if MDS is started with down OST
4519 Bugzilla : 9439,5706
4520 Description: Allow startup/shutdown of an MDS without depending on the
4521 availability of the OSTs.
4522 Details : Asynchronously call mds_lov_synchronize during MDS startup.
4523 Add appropriate locking and lov-osc refcounts for safe
4524 cleaning. Add osc abort_inflight calls in case the
4525 synchronize never started.
4528 Frequency : occasional (Cray XT3 only)
4530 Description: root not authorized to access files in CRAY_PORTALS environment
4531 Details : The client process capabilities were not honoured on the MDS in
4532 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
4533 been accessed by an authorized user then root was able to access
4534 the file on the local client also. The root user capabilities
4535 are now allowed on the MDS, as this environment has secure UID.
4538 Frequency : occasional
4540 Description: ldiskfs "too long searching" message happens too often
4541 Details : A debugging message (otherwise harmless) prints too often on
4542 the OST console. This has been reduced to only happen when
4543 there are fragmentation problems on the filesystem.
4548 Description: Division by zero in statfs when all OSCs are inactive
4549 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
4550 lov_statfs divides by value returned by lov_get_stripecnt().
4554 Bugzilla : 9489, 3273
4555 Description: First write from each client to each OST was only 4kB in size,
4556 to initialize client writeback cache, which caused sub-optimal
4557 RPCs and poor layout on disk for the first writen file.
4558 Details : Clients now request an initial cache grant at (re)connect time
4559 and so that they can start streaming writes to the cache right
4560 away and always do full-sized RPCs if there is enough data.
4561 If the OST is rebooted the client also re-establishes its grant
4562 so that client cached writes will be honoured under the grant.
4567 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
4568 Details : Now I/O RPCs go to different portal number and (presumably) fast
4569 lock requests (and glimses) and other RPCs get their own service
4570 threads pool that should be able to service those RPCs
4573 Severity : enhancement
4575 Description: Ability to exchange lustre version between client and servers and
4576 issue warnings at client side if client is too old. Also for
4577 liblustre clients there is ability to refuse connection of too old
4579 Details : New 'version' field is added to connect data structure that is
4580 filled with version info. That info is later checked by server and
4584 Frequency : rare, liblustre only.
4585 Bugzilla : 9296, 9581
4586 Description: Two simultaneous writes from liblustre at offset within same page
4587 might proceed at the same time overwriting eachother with stale
4589 Details : I/O lock withing llu_file_prwv was released too early, before data
4590 actually was hitting the wire. Extended lock-holding time until
4591 server acknowledges receiving data.
4594 Frequency : extremely rare. Never observed in practice.
4596 Description: avoid generating lustre_handle cookie of 0.
4597 Details : class_handle_hash() generates handle cookies by incrementing
4598 global counter, and can hit 0 occasionaly (this is unlikely, but
4599 not impossible, because initial value of cookie counter is
4600 selected randonly). Value of 0 is used as a sentinel meaning
4601 "unassigned handle" --- avoid it. Also coalesce two critical
4602 sections in this function into one.
4604 Severity : enhancement
4606 Description: allow liblustre clients to delegate truncate locking to OST
4607 Details : To avoid overhead of locking, liblustre client instructs OST to
4608 take extent lock in ost_punch() on client's behalf. New connection
4609 flag is added to handle backward compatibility.
4611 Severity : enhancement
4612 Bugzilla : 4928, 7341, 9758
4613 Description: allow number of OST service threads to be specified
4614 Details : a module parameter allows the number of OST service threads
4615 to be specified via "options ost ost_num_threads={N}" in the
4616 OSS's /etc/modules.conf or /etc/modprobe.conf.
4620 Bugzilla : 6146, 9635, 9895
4621 Description: servers crash with bad pointer in target_handle_connect()
4622 Details : In rare cases when a client is reconnecting it was possible that
4623 the connection request was the last reference for that export.
4624 We would temporarily drop the export reference and get a new
4625 one, but this may have been the last reference and the export
4626 was just destroyed. Get new reference before dropping old one.
4628 Severity : enhancement
4629 Frequency : if client is started with failover MDS
4631 Description: Allow multiple MDS hostnames in the mount command
4632 Details : Try to read the configuration from all specified MDS
4633 hostnames during a client mount in case the "primary"
4636 Severity : enhancement
4638 Description: Stop sending data to evicted clients as soon as possible.
4639 Details : Check if the client we are about to send or are sending data to
4640 was evicted already. (Check is done every second of waiting,
4641 for which l_wait_event interface was extended to allow checking
4642 of exit condition at specified intervals).
4645 Frequency : rare, normally only when NFS exporting is done from client
4647 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
4649 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
4652 Frequency : always, if srand() is called before liblustre initialization
4654 Description: Liblustre uses system PRNG disturbing its usage by user application
4655 Details : Introduce internal to lustre fast and high-quality PRNG for
4656 lustre usage and make liblustre and some other places in generic
4657 lustre code to use it.
4659 Severity : enhancement
4660 Bugzilla : 9477, 9557, 9870
4661 Description: Verify that the MDS configuration logs are updated when xml is
4662 Details : Check if the .xml configuration logs are newer than the config
4663 logs stored on the MDS and report an error if this is the case.
4664 Request --write-conf, or allow starting with --old_conf.
4666 Severity : enhancement
4668 Description: Handle symlinks in the path when checking if Lustre is mounted.
4669 Details : Resolve intermediate symlinks when checking if a client has
4670 mounted a filesystem to avoid duplicate client mounts.
4675 Description: lconf can hit an error exception but still return success.
4676 Details : The lconf command catches the Command error exception at the top
4677 level script context and will exit with the associated exit
4678 status, but doesn't ensure that this exit status is non-zero.
4683 Description: failure of ptlrpc thread startup can cause oops
4684 Details : Starting a ptlrpc service thread can fail if there are a large
4685 number of threads or the server memory is very fragmented.
4686 Handle this without oopsing.
4689 Frequency : always, only if liblustre and non-default acceptor port was used
4691 Description: liblustre cannot connect to servers with non-default acceptor port
4692 Details : tcpnal_set_default_params() was not called and was therefore
4693 ignoring the environment varaible TCPNAL_PORT, as well as other
4694 TCPNAL_ environment variables
4699 Description: two objects could be created on the same OST for a single file
4700 Details : If an OST is down, in some cases it was possible to create two
4701 objects on a single OST for a single file. No problems other
4702 than potential performance impact and spurious error messages.
4706 Bugzilla : 5681, 9562
4707 Description: Client may oops in ll_unhash_aliases
4708 Details : Client dcache may become inconsistent in race condition.
4709 In some cases "getcwd" can fail if the current directory is
4715 Description: Inode refcounting problems in NFS export code
4716 Details : link_raw functions used to call d_instantiate without obtaining
4717 extra inode reference first.
4721 Bugzilla : 9942, 9903
4722 Description: Referencing freed requests leading to crash, memleaks with NFS.
4723 Details : We used to require that call to ll_revalidate_it was always
4724 followed by ll_lookup_it. Also with revalidate_special() it is
4725 possible to call ll_revalidate_it() twice for the same dentry
4726 even if first occurence returned success. This fix changes semantic
4727 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
4728 reference on a request referred from the intent.
4729 ll_intent_release() then releases such a request.
4732 Frequency : rare, normally benchmark loads only
4734 Description: unlinked inodes were kept in memory on the client
4735 Details : If a client is repeatedly creating and unlinking files it
4736 can accumulate a lot of stale inodes in the inode slab cache.
4737 If there is no other client load running this can cause the
4738 client node to run out of memory. Instead flush old inodes
4739 from client cache that have the same inode number as a new inode.
4742 Frequency : SLES9 2.6.5 kernel and long filenames only
4743 Bugzilla : 9969, 10379
4744 Description: utime reports stale NFS file handle
4745 Details : SLES9 uses out-of-dentry names in some cases, which confused
4746 the lustre dentry revalidation. Change it to always use the
4750 Frequency : rare, unless heavy write-truncate concurrency is continuous
4751 Bugzilla : 4180, 6984, 7171, 9963, 9331
4752 Description: OST becomes very slow and/or deadlocked during object unlink
4753 Details : filter_destroy() was holding onto the parent directory lock
4754 while truncating+unlinking objects. For very large objects this
4755 may block other threads for a long time and slow overall OST
4756 responsiveness. It may also be possible to get a lock ordering
4757 deadlock in this case, or run out of journal credits because of
4758 the combined truncate+unlink. Solution is to do object truncate
4759 first in one transaction without parent lock, and then do the
4760 final unlink in a new transaction with the parent lock. This
4761 reduces the lock hold time dramatically.
4764 Frequency : rare, 2.4 kernels only
4766 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
4767 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
4768 before unmounting their filesystem in order to avoid deadlock.
4769 The kernel_locked() function in 2.4 kernels only checks whether
4770 the kernel lock is held, not whether it is this process that is
4771 holding it as 2.6 kernels do.
4776 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
4777 Details : The client ptlrpc code may be trying to reconnect to a down
4778 server before a previous connection attempt has timed out.
4779 Increase the reconnect interval to be longer than the connection
4780 timeout interval to avoid sending duplicate connections to
4784 Frequency : echo_client brw_test command
4786 Description: fix echo_client to work with OST preallocated code
4787 Details : OST preallocation code (5137) didn't take echo_client IO path
4788 into account: echo_client calls filter methods outside of any
4789 OST thread and, hence, there is no per-thread preallocated
4790 pages and buffers to use. Solution: hijack pga pages for IO. As
4791 a byproduct, this avoids unnecessary data copying.
4795 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
4796 Description: Client can oops in mdc_commit_close() after open replay
4797 Details : It was possible for the MDS to return an open request with no
4798 transaction number in mds_finish_transno() if the client was
4799 evicted, but without actually returning an error. Clients
4800 would later try to replay that open and may trip an assertion
4801 Simplify the client close codepath, and always return an error
4802 from the MDS in case the open is not successful.
4805 Frequency : rare, 2.6 OSTs only
4807 Description: OST may deadlock under high load on fragmented files
4808 Details : If there was a heavy load and highly-fragmented OST filesystems
4809 it was possible to have all the OST threads deadlock waiting on
4810 allocation of biovecs, because the biovecs were not released
4811 until the entire RPC IO was completed. Instead, release biovecs
4812 as soon as they are complete to ensure forward IO progress.
4814 Severity : enhancement
4816 Description: Support for specifying external journal device at mount
4817 Details : If an OST or MDS device is formatted with an external journal
4818 device, this device major/minor is stored in the ext3 superblock
4819 and may not be valid for failover. Allow detecting and
4820 specifying the external journal at mount time.
4825 Description: Mounting an MDS with pending unlinked files may cause oops
4826 Details : target_finish_recovery() calls mds_postrecov() which returned
4827 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
4828 considers this an error and immediately begins cleaning up the
4829 lov, just after starting the mds_lov process
4831 Severity : enhancement
4833 Description: Implement 'lfs df' to report actual free space on per-OST basis
4834 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
4835 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
4836 usage of objects; '-h' to report in human readable format.
4838 ------------------------------------------------------------------------------
4840 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
4847 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
4848 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
4849 at startup during orphan unlinking. Instead, skip buddy update
4850 before setup, as it will be regenerated after recovery is complete.
4855 Description: If an OST is inactive, its locks might reference stale inodes.
4856 Details : lov_change_cbdata() must iterate over all namespaces, even if
4857 they are inactive to clear inode references from the lock.
4859 Severity : enhancement
4860 Frequency : occasional, if non-standard max_dirty_mb used
4862 Description: Client will block write RPCs if not enough grant
4863 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
4864 then the client will block writes while waiting for another RPC
4865 to complete instead of consuming its dirty limit. With change
4866 we get improved performance when max_dirty_mb is small.
4868 Severity : enhancement
4869 Bugzilla : 3389, 6253
4870 Description: Add support for supplementary groups on the MDS.
4871 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
4872 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
4873 lookups for user supplementary groups into a cache.
4877 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
4878 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
4879 for liblustre O_EXCL is enforced by MDS.
4883 Description: readdir never returns NULL in liblustre.
4884 Details : Corrected llu_iop_getdirentries logic, to return offset of next
4885 dentry in struct dirent.
4889 Frequency : liblustre only, depends on application IO pattern
4890 Description: liblustre clients evicted if not contacting servers
4891 Details : Don't put liblustre clients into the ping_evictor list, so
4892 they will not be evicted by the pinger ever.
4894 Severity : enhancement
4896 Description: Add ability to evict clients by NID from MDS.
4897 Details : By echoing "nid:$NID" string into
4898 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
4899 $NID would be instantly evicted from this MDS and from all active
4900 OSTs connected to it.
4904 Description: Do not query file size twice, somewhat slowing stat(2) calls.
4905 Details : lookup_it_finish() used to query file size from OSTs that was not
4910 Description: service threads change working directory to that of init
4911 Details : Starting lustre service threads may pin the working directory
4912 of the parent thread, making that filesystem busy. Threads
4913 now change to the working directory of init to avoid this.
4917 Frequency : during shutdown only
4918 Description: shutdown with a failed MDS or OST can cause unmount to hang
4919 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
4924 Frequency : 2.6 only
4925 Description: chmod/chown may include an extra supplementary group
4926 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
4927 supplementary group and if none is specified this is used.
4930 Bugzilla : 5479 (6816)
4932 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
4933 Details : If lookup is in progress on a file that is unlinked we might try
4934 to revalidate the inode and fail in revalidate after lookup is
4935 complete and ll_file_open() enqueues the open again but
4936 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
4939 Frequency : always, if lconf --abort_recovery used
4941 Description: lconf --abort_recovery fails with 'Operation not supported'
4942 Details : lconf was attempting to abort recovery on the MDT device and not
4945 ------------------------------------------------------------------------------
4947 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
4952 Frequency : rare (only unsupported configurations with a node running as an
4954 Bugzilla : 6514, 5137
4955 Description: Mounting a Lustre file system on a node running as an OST could
4957 Details : OSTs now preallocates memory needed to write out data at
4958 startup, instead of when needed, to avoid having to
4959 allocate memory in possibly low memory situations.
4960 Specifically, if the file system is mounted on on OST,
4961 memory pressure could force it to try to write out data,
4962 which it needed to allocate memory to do. Due to the low
4963 memory, it would be unable to do so and the node would
4964 become unresponsive.
4966 Severity : enhancement
4968 Description: Addition of lconf --service command line option
4969 Details : lconf now accepts a '--service <arg>' option, which is
4970 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
4972 Severity : enhancement
4974 Description: Failover mode is now the default for OSTs.
4975 Details : By default, OSTs will now run in failover mode. To return to
4976 the old behaviour, add '--failout' to the lmc line for OSTs.
4978 Severity : enhancement
4980 Description: Health checks are now provided for MDS and OSTs
4981 Details : Additional detailed health check information on MSD and OSTs
4982 is now provided through the procfs health_check value.
4985 Frequency : occasional, depends on IO load
4987 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
4988 after numerous create/delete cycles
4989 Details : The ext3 inode allocation policy would not allocate new inodes
4990 very well on the OSTs because there are no new directories
4991 being created. Instead we look for groups with free space if
4992 the parent directories are nearly full.
4997 Description: Network or server problems during mount may cause partially
4998 mounted clients instead of returning an error.
4999 Details : The config llog parsing code may overwrite the error return
5000 code during mount error handling, returning success instead
5006 Description: MDS can fail to allocate large reply buffers
5007 Details : After long uptimes the MDS can fail to allocate large reply
5008 buffers (e.g. zconf client mount config records) due to memory
5009 fragmentation or consumption by the buffer cache. Preallocate
5010 some large reply buffers so that these replies can be sent even
5011 under memory pressure.
5015 Frequency : rare (liblustre)
5016 Description: fsx running with liblustre complained that using truncate() to
5017 extend the file doesn't work. This patch corrects that issue.
5018 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
5019 ATTR_SIZE and lsm use in llu_setattr_raw.
5023 Frequency : rare, only 2.6 kernels
5024 Description: Unusual file access patterns on the MDS may result in inode
5025 data being lost in very rare circumstances.
5026 Details : Bad interaction between the ea-in-inode patch and the "no-read"
5027 code in the 2.6 kernel caused the inode and/or EA data not to
5028 be read from disk, causing single-file corruption.
5032 Frequency : rare, only 2.6 filesystems using extents
5033 Description: Heavy concurrent write and delete load may cause data corruption.
5034 Details : It was possible under high-load situations to have an extent
5035 metadata block in the block device cache from a just-unlinked
5036 file overwrite a newly-allocated data block. We now unmap any
5037 metadata buffers that alias just-allocated data blocks.
5041 Frequency : filesystems with default stripe_count larger than 77
5042 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
5043 Details : lconf specifies an inode size of 4096 bytes when the default
5044 stripe_count is larger than 77. This conflicts with the default
5045 inode density of 1 per 4096 bytes. Allocate smaller inodes in
5046 this case to avoid pinning too much memory for large EAs.
5048 ------------------------------------------------------------------------------
5050 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
5055 Frequency : rare (extremely heavy IO load with hundreds of clients)
5057 Description: Client is evicted, gets IO error writing to file
5058 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
5059 caused clients to be evicted for AST timeouts. The fixes in
5060 bug 5192 mean we no longer need to have such short AST timeouts
5061 so ldlm_timeout has been increased.
5064 Frequency : occasional during --force or --failover shutdown under load
5065 Bugzilla : 5949, 4834
5066 Description: Server oops/LBUG if stopped with --force or --failover under load
5067 Details : a collection of import/export refcount and cleanup ordering
5068 issues fixed for safer force cleanup
5071 Frequency : only filesystems larger than 120 OSTs
5072 Bugzilla : 5990, 6223
5073 Description: lfs getstripe would oops on a very large filesystem
5074 Details : lov_getconfig used kfree on vmalloc'd memory
5077 Frequency : only filesystems exporting via NFS to Solaris 10 clients
5078 Bugzilla : 6242, 6243
5079 Description: reading from files that had been truncated to a non-zero size
5080 but never opened returned no data
5081 Details : ll_file_read() reads zeros from no-object files to EOF
5086 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
5087 Details : There's little harm in aborting MDS/OSS recovery and letting it
5088 try again, so I removed the LASSERT and return an error instead.
5090 Severity : enhancement
5092 Description: New debugging infrastructure for tracking down data corruption
5093 Details : The I/O checksum code was replaced to: (a) control it at runtime,
5094 (b) cover more of the client-side code path, and (c) try to narrow
5095 down where problems occurred
5099 Bugzilla : 3819, 4364, 4397, 6313
5100 Description: Racing close and eviction MDS could cause assertion in mds_close
5101 Details : It was possible to get multiple mfd references during close and
5102 client eviction, leading to one thread referencing a freed mfd.
5104 Severity: : enhancement
5105 Bugzilla : 3262, 6359
5106 Description: Attempts to reconnect to servers are now more aggressive.
5107 Details : This builds on the enhanced upcall-less recovery that was added
5108 in 1.4.2. When trying to reconnect to servers, clients will
5109 now try each server in the failover group every 10 seconds. By
5110 default, clients would previously try one server every 25 seconds.
5115 Description: After recovery, certain operations trigger a failed
5116 assertion on a client.
5117 Details : Failing over an mds, using lconf -d --failover, while a
5118 client was doing a readdir() call would cause the client to
5119 LBUG after recovery completed and the readdir() was resent.
5121 Severity : enhancement
5123 Description: Default groups are now added by lconf
5124 Details : You can now run lconf --group <servicename> without having to
5125 manually add groups with lmc.
5128 Frequency : occasional
5130 Description: Nodes with an elan id of 0 trigger a failed assertion
5133 Frequency : always when accessing e.g. tty/console device nodes
5135 Description: tty and some other devices nodes cannot be used on lustre
5136 Details : file's private_data field is used by device data and lustre
5137 values in there got lost. New field was added to struct file to
5138 store fs-specific private data.
5141 Frequency : when exporting Lustre via NFS
5143 Description: NFSD failed occasionally when looking up a path component
5144 Details : NFSD is looking up ".." which was broken in ext3 directories
5145 that had grown large enough to become hashed.
5148 Frequency : Clusters with multiple interfaces not on the same subnet
5150 Description: Nodes will repeatedly try to reconnect to an interface which it
5151 cannot reach and report an error to the log.
5152 Details : Extra peer list entries will be created by lconf with some peers
5153 unreachable. lconf now validates the peer before adding it.
5156 Frequency : Only if a default stripe is set on the filesystem root.
5158 Description: Setting a default stripe on the filesystem root prevented the
5159 filesystem from being remounted.
5160 Details : The client was sending extra request flags in the root getattr
5161 request and did not allocate a reply buffer for the dir EA.
5164 Frequency : occasional, higher if lots of files are accessed by one client
5165 Bugzilla : 6159, 6097
5166 Description: Client trips assertion regarding lsm mismatch/magic
5167 Details : While revalidating inodes the VFS looks up inodes with ifind()
5168 and in rare cases can find an inode that is being freed.
5169 The ll_test_inode() code will free the lsm during ifind()
5170 when it finds an existing inode and then the VFS later attaches
5171 this free lsm to a new inode.
5175 Bugzilla : 6422, 7030
5176 Description: MDS deadlock between mkdir and client eviction
5177 Details : Creating a new file via mkdir or mknod (starting a transaction
5178 and getting the ns lock) can deadlock with client eviction
5179 (gets ns lock and trying to finish a synchronous transaction).
5182 Frequency : occasional
5183 Description: While starting a server, the fsfilt_ext3 module could not be
5185 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
5186 kernels. Previously, lconf would still use the ext3 name
5187 when trying to load modules. Now, it will correctly use
5188 ext3 on 2.4 and ldiskfs on 2.6.
5190 Severity : enhancement
5191 Description: The default stripe count has been changed to 1
5192 Details : The interpretation of the default stripe count (0, to lfs
5193 or lmc) has been changed to mean striping across a single
5194 OST, rather than all available. For general usage we have
5195 found a stripe count of 1 or 2 works best.
5197 Severity : enhancement
5198 Description: Add support for compiling against Cray portals.
5199 Details : Conditional compiling for some areas that are different
5203 Frequency : occasional
5204 Bugzilla : 6409, 6834
5205 Description: Creating files with an explicit stripe count may lead to
5206 a failed assertion on the MDS
5207 Details : If some OSTs are full or unavailable, creating files may
5208 trigger a failed assertion on the MDS. Now, Lustre will
5209 try to use other servers or return an error to the
5213 Frequency : occasional
5215 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
5216 caused lock timeout during readahead (since 1.4.2).
5217 Details : Processes doing readahead might match a lock that hasn't been
5218 granted yet if there are overlapping and conflicting lock
5219 requests. The readahead process waits on ungranted lock
5220 (original lock is CBPENDING), while OST waits for that process
5221 to cancel CBPENDING read lock and eventually evicts client.
5223 Severity : enhancement
5225 Description: Initial enabling of flock support for clients
5226 Details : Implements fcntl advisory locking and file status functions.
5227 This feature is provided as an optional mount flag (default
5228 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
5229 locking are implemented yet, and those that are are not guaranteed
5230 to be completely correct in production environments.
5231 mount -t lustre -o [flock|noflock] ...
5234 Frequency : occasional
5236 Description: OSTs running 2.4 kernels but with extents enabled might trip an
5237 assertion in the ext3 JBD (journaling) layer.
5238 Details : The b_committed_data struct is protected by the big kernel lock
5239 in 2.4 kernels, serializing journal_commit_transaction() and
5240 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
5241 access to this struct. In 2.6 kernels there is finer grained
5242 locking to improve SMP performance of the JBD layer.
5246 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
5251 Description: merge and cleanup kernel patches.
5254 -----------------------------------------------------------------------------
5256 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
5258 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
5259 versions of Lustre. Please update all servers and clients to
5260 version 1.4.2 or later at the same time. You must also run
5261 "lconf --write-conf {config}.xml" on the MDS while it is stopped
5262 to update the configuration logs.
5264 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
5265 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
5266 - swab llog records of type '0' so we get proper header size/idx (5861)
5267 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
5268 - fix rename of one directory over another leaking an inode (5953)
5269 - avoid SetPageDirty on 2.6 (5981)
5270 - don't re-add just-being-destroyed locks to the waiting list (5653)
5271 - when creating new directories, inherit the parent's custom
5272 striping settings if present parent (3048)
5273 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
5274 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
5275 - don't hold client locks on temporary worklist from l_lru (5666)
5276 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
5277 - automatically evict dead clients (5921)
5278 - Update file size properly in create+truncate+fstat case (6196)
5279 - Do not unhash mountpoint dentries, do not allow removal of
5281 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
5282 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
5283 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
5284 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
5285 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
5286 - Correctly handle reads of files with no objects (6243)
5287 - lctl recover will also mark a device active if deactivate used (5933)
5289 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
5290 - allow --write-conf on an MDS with different nettype than client (5619)
5291 - don't write config llogs to MDS for mounts not from that MDS (5617)
5292 - lconf should create multiple TCP connections from a client (5201)
5293 - init scripts are now turned off by default; run chkconfig --on
5294 lustre and chkconfig --on lustrefs to use them
5295 - upcalls are no longer needed for clients to recover to failover
5297 - add --abort-recovery option to lconf to abort recovery on device
5299 - add support for an arbitrary number of OSTs (3026)
5300 - Quota support protocol changes.
5301 - forward compatibility changes to wire structs (6007)
5302 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5303 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5304 - improved "lustre status" script
5305 - initialize blocksize for non-regular files (6062)
5306 - added --disable-server and --disable-client configure options (5782)
5307 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5308 - Vanilla 2.4.29 support
5309 - increase maximum number of obd devices to 520 (6242)
5310 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5311 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5312 SLES 9 2.6.5 kernels
5313 - limit stripes per file to 160 (the maximum EA size) (6093)
5315 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5318 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5319 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5320 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5321 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5322 - small ext3 extents cleanups and fixes (5733)
5323 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5324 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5325 - avoid needless client->OST connect, fix handle mismatch (5317)
5326 - fix DLM error path that led to out-of-sync client, long delays (5779)
5327 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5328 - fix several locking issues related to i_size (5492,5624,5654,5672)
5329 - don't move pending lock onto export if it is already evicted (5683)
5330 - fix kernel oops when creating .foo in unlinked directory (5548)
5331 - fix deadlock in obdfilter statistics vs. object create (5811)
5332 - use time_{before,after} to avoid timer jiffies wrap (5882)
5333 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5334 - Do not leak request if server was not able to process it (5154)
5335 - If mds_open unable to find parent dir, make that negative lookup(5154)
5336 - don't create new directories with extent-mapping (5909, 5936)
5338 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5339 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5340 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5341 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5342 - support for RHEL4 kernel on i686 (5773)
5343 - provide error messages when incompatible logs are encountered (5898)
5345 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5346 * version 1.4.0.10 (1.4.1 release candidate 1)
5348 - don't keep a lock reference when lock is not granted (4238)
5349 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5350 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5351 - drop import inflight refcount on signal_completed_replay error (5255)
5352 - unlock page after async write error during send (3677)
5353 - handle missing objects in filter_preprw_read properly (5265)
5354 - no transno return for symlink open, don't save no-trasno open (3440)
5355 - don't try to complete elan receive that already failed (4012)
5356 - free RPC server reply state on error (5406)
5357 - clean up thread from ptlrpc_start_thread() on error (5160)
5358 - readahead could read extra page into cache that wasn't ejected (5388)
5359 - prevent races in class_attach/setup/cleanup/detach (5260)
5360 - don't dereference de->d_inode after l_dput of de (5458)
5361 - use "int" for stripe value returned from lock_to_stripe (5544)
5362 - mballoc allocation and error-checking fixes in 2.6 (5504)
5363 - block device patches to fix I/O request sizes in 2.6 (5482)
5364 - look up hostnames for IB nals (5602)
5365 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5366 - don't start multiple acceptors for the same port (5277)
5367 - fix incorrect LASSERT in mds_getattr_name (5635)
5368 - export a proc file for general "ping" checking (5628)
5369 - fix "lfs check" to not block when the MDS is down (5628)
5371 - service request history (4965)
5372 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5373 - create an "evict_client" /proc entry on OSTs, like the MDS has
5374 - fix mount usage message, return errors per mount(8) (5168)
5375 - change grep [] to grep "[]" in tests so they work in more UMLs
5376 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5377 - remove ancient LOV_MAGIC_V0 EA support (5047)
5378 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5379 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5380 - allow readahead allocations to fail when low on memory (5383)
5381 - mmap locking landed again, after considerable improvement (2828)
5382 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
5384 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
5387 - send OST transaction number in read/write reply to free req (4966)
5388 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5389 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5390 - the watchdog thread now runs as interruptible (5246)
5391 - flock/lockf fixes (but it's still disabled, pending 5135)
5392 - don't use EXT3 constants in llite code (5094)
5393 - memory shortage at startup could cause assertion (5176)
5395 - reorganization of lov code
5396 - single portals codebase
5398 - add extents/mballoc support (5025)
5399 - direct I/O reads in the obdfilter (4048)
5400 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
5402 tbd Cluster File Systems, Inc. <info@clusterfs.com>
5405 - send OST transaction number in read/write reply to free req (4966)
5406 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5407 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5408 - the watchdog thread now runs as interruptible (5246)
5409 - handle missing objects in filter_preprw_read properly (5265)
5410 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5411 - drop import inflight refcount on signal_completed_replay error (5255)
5412 - unlock page after async write error during send (3677)
5413 - return original error code on reconstructed replies (3761)
5414 - no transno return for symlink open, don't save no-trasno open (3440)
5416 - add pid to ldlm debugging output (4922)
5417 - bump the watchdog timeouts -- we can't handle 30sec yet
5418 - extra debugging for orphan dentry/inode bug (5259)
5420 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
5423 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
5424 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
5425 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
5426 - fix NULL dereference in /proc/sys/portals/routes (4827)
5427 - allow failed mdc_close() operations to be interrupted (4561)
5428 - stop precreate on OST before MDS would time out on it (4778)
5429 - don't send partial-page writes before EOF from client (4410)
5430 - discard client grant for sub-page writes on large-page clients (4520)
5431 - don't free dentries not owned by NFS code, check generation (4806)
5432 - fix lsm leak if mds_create_objects() fails (4801)
5433 - limit debug_daemon file size, always print CERROR messages (4789)
5434 - use transno after validating reply (3892)
5435 - process timed out requests if import state changes (3754)
5436 - update mtime on OST during writes, return in glimpse (4829)
5437 - add mkfsoptions to LDAP (4679)
5438 - use ->max_readahead method instead of zapping global ra (5039)
5439 - don't interrupt __l_wait_event() during strace
5441 - add software watchdogs to catch hung threads quickly (4941)
5442 - make lustrefs init script start after nfs is mounted
5443 - fix CWARN/ERROR duplication (4930)
5444 - return async write errors to application if possible (2248)
5445 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
5446 - print NAL number in %x format (4645)
5447 - update barely-supported suse-2.4.21-171 series (4842)
5448 - support for sles 9 %post scripts
5449 - support for building 2.6 kernel-source packages
5450 - support for sles km_* packages
5452 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
5455 - ignore -ENOENT errors in osc_destroy (3639)
5456 - notify osc create thread that OSC is being cleaned up (4600)
5457 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
5458 - reconstruct ost_handle() like mds_handle() (4657)
5459 - create a new thread to do import eviction to avoid deadlock (3969)
5460 - let lconf resolve symlinked-to devices (4629)
5461 - don't unlink "objects" from directory with default EA (4554)
5462 - hold socknal file ref over connect in case target is down (4394)
5463 - allow more than 32000 subdirectories in a single directory (3244)
5464 - fix blocks count for O_DIRECT writes (3751)
5465 - OST returns ENOSPC from object create when no space left (4539)
5466 - don't send truncate RPC if file size isn't changing (4410)
5467 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
5468 - bind to privileged port in socknal and tcpnal (3689)
5470 - rate limit CERROR/CWARN console message to avoid overload (4519)
5471 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
5472 - basic mmap support (3918)
5473 - kernel patch series update from b1_4 (4711)
5475 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
5478 - avoid crash during MDS cleanup with OST shut down (2775)
5479 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
5480 - don't use bad inodes on the MDS (3744)
5481 - dynamic object preallocation to improve recovery speed (4236)
5482 - don't hold spinlock over lock dumping or change debug flags (4401)
5483 - don't zero obd_dev when it is force cleaned (3651)
5484 - print grants to console if they go negative (4431)
5485 - "lctl deactivate" will stop automatic recovery attempts (3406)
5486 - look for existing locks in ldlm_handle_enqueue() (3764)
5487 - don't resolve lock handle twice in recovery avoiding race (4401)
5488 - revalidate should check working dir is a directory (4134)
5490 - don't always mark "slow" obdfilter messages as errors (4418)
5492 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
5495 - don't close LustreDB during write_conf until it is done (3860)
5496 - fix typo in lconf for_each_profile (3821)
5497 - allow dumping logs from multiple threads at one time (3820)
5498 - don't allow multiple threads in OSC recovery (3812)
5499 - fix debug_size parameters (3864)
5500 - fix mds_postrecov to initialize import for llog ctxt (3121)
5501 - replace config semaphore with spinlock (3306)
5502 - be sure to send a reply for a CANCEL rpc with bad export (3863)
5503 - don't allow enqueue to complete on a destroyed export (3822)
5504 - down write_lock before checking llog header bitmap (3825)
5505 - recover from lock replay timeout (3764)
5506 - up llog sem before sending rpc (3652)
5507 - reduce ns lock hold times when setting kms (3267)
5508 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
5509 - fix NULL deref and obd_dev leak on setup error (3312)
5510 - replace some LBUG about llog ops with error handling (3841)
5511 - don't match INVALID dentries from d_lookup and spin (3784)
5512 - hold dcache_lock while marking dentries INVALID and hashing (4255)
5513 - fix invalid assertion in ptlrpc_set_wait (3880)
5515 - add libwrap support for the TCP acceptor (3996)
5516 - add /proc/sys/portals/routes for non-root route listing (3994)
5517 - allow setting MDS UUID in .xml (2580)
5518 - print the stack of a process that LBUGs (4228)
5520 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
5523 - don't cleanup request in ll_file_open() on failed MDS open (3430)
5524 - make sure to unset replay flag from failed open requests (3440)
5525 - if default stripe count is 0, use OST count for inode size (3636)
5526 - update parent mtime/ctime on client for create/unlink (2611)
5527 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
5528 - free recovery state on server during a forced cleanup (3571)
5529 - unregister_reply for resent reqs (3063)
5530 - loop back devices mounting and status check on 2.6 (3563)
5531 - fix resource-creation race that can provoke i_size == 0 (3513)
5532 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
5533 - more debugging for page-accounting assertion (3746)
5534 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
5535 - avoid deadlock after precreation failure (3758)
5536 - fix race and lock order deadlock in orphan handling (3450, 3750)
5537 - add validity checks when grabbing inodes from l_ast_data (3599)
5539 - add /proc/.../recovery_status to obdfilter (3428)
5540 - lightweight CDEBUG infrastructure, debug daemon (3668)
5541 - change default OSC RPC parameters to be better on small clusters
5542 - turn off OST read cache for files smaller than 32MB
5543 - install man pages and include them in rpms (3100)
5544 - add new init script for (un)mounting lustre filesystems (2593)
5545 - run chkconfig in %post for init scripts (3701)
5546 - drop scimac NAL (unmaintained)
5548 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
5551 - clean kiobufs before and after use (3485)
5552 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
5553 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
5554 - change init script to fail more clearly if not run as root (1528)
5555 - allow clients to reconnect during replay (1742)
5556 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
5557 - don't do DNS lookups on NIDs too small for IP addresses (3442)
5558 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
5559 - fix cond_resched (3554)
5560 - only evict unfinished clients after recovery (3515)
5561 - allow bulk resend, prevent data loss (3570)
5562 - dynamic ptlrpc request buffer allocation (2102)
5563 - don't allow unlinking open directory if it isn't empty (2904)
5564 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
5565 - remove extraneous obd dereference causing LASSERT failure (3334)
5566 - don't use get_cycles() when creating temp. files on the mds (3156)
5567 - hold i_sem when setting i_size in ll_extent_lock() (3564)
5568 - handle EEXIST for set-stripe, set proper directory name (3336)
5570 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
5571 - fix ksocknal_fmb_callback() error messages (2918)
5573 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
5576 - don't copy lvb into (possibly NULL) reply on error (2983)
5577 - don't deref dentry after dput, don't free lvb on error (2922)
5578 - use the kms to determine writeback rpc length (2947)
5579 - increment oti_logcookies when osc is inactive (2948)
5580 - update client's i_blocks count via lvb messages (2543)
5581 - handle intent open/close of special files properly (1557)
5582 - mount MDS with errors=remount-ro, like obdfilter (2009)
5583 - initialize lock handle to avoid ASSERT on error cleanup (3057)
5584 - don't use cancelling-locks' kms values (2947)
5585 - use highest lock extent for kms, not last one (2925)
5586 - don't dereference ERR_PTR() dentry in error handling path (3107)
5587 - fix thread race in portals_debug_dumplog() (3122)
5588 - create lprocfs device entries at setup instead of at attach (1519)
5589 - common AST error handler, don't evict client on completion race (3145)
5590 - zero nameidata in detach_mnt in 2.6 (3118)
5591 - verify d_inode after revalidate_special is valid in 2.6 (3116)
5592 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
5593 - initialize RPC timeout timer earlier for 2.6 (3219)
5594 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
5595 - print nal/nid for unknown nid (3258)
5596 - additional checks for oscc recovery before doing precreate (3284)
5597 - fix ll_extent_lock() error return code for 64-bit systems (3043)
5598 - don't crash in mdc_close for bad permissions on open (3285)
5599 - zero i_rdev for non-device files (3147)
5600 - clear page->private before handing to FS, better assertion (3119)
5601 - tune the read pipeline (3236)
5602 - fix incorrect decref of invalidated dentry (2350)
5603 - provide read-ahead stats and refine rpc in flight stats (3328)
5604 - don't hold journal transaction open across create RPC (3313)
5605 - update atime on MDS at close time (3265)
5606 - close LDAP connection when recovering to avoid server load (3315)
5607 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
5608 - don't leak open file on MDS after open resend (3325)
5609 - serialize filter_precreate and filter_destroy_precreated (3329)
5610 - loop device shouldn't call sync_dev() for nul device (3092)
5611 - clear page cache after eviction (2766)
5612 - resynchronize MDS->OST in background (2824)
5613 - refuse to mount the same filesystem twice on same mountpoint (3394)
5614 - allow llmount to create routes for mounting behind routers (3320)
5615 - push lock cancellation to blocking thread for glimpse ASTs (3409)
5616 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
5617 - fix rare problem with rename on htree directories (3417)
5619 - allow default OST striping configuration per directory (1414)
5620 - fix compilation for qswnal for 2.6 kernels (3125)
5621 - increase maximum number of MDS request buffers for large systems
5622 - change liblustreapi to be useful for external progs like lfsck (3098)
5623 - increase local configuration timeout for slow disks (3353)
5624 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
5626 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5629 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
5630 - don't overwrite extent policy data in reply if lock was blocked (2901)
5631 - drop filter export grants atomically with removal from device (2663)
5632 - del obd_self_export from work_list in class_disconnect_exports (2908)
5633 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
5634 - swab reply message in mdc_close, other PPC fixes (2464)
5635 - fix destroying of named logs (2325)
5636 - overwrite old logs when running lconf --write_conf (2264)
5637 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
5638 - fix race in target_handle_connect (2898)
5639 - mds_reint_create() should take same inode create lock (2926)
5640 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
5641 - don't close files for self_export to avoid uninitialized obd (2936)
5642 - allow MDS with the same name as client node (2939)
5643 - hold dentry reference for closed log files for unlink (2325)
5644 - reserve space for all logs during transactions (2059)
5645 - don't evict page beyond end of stripe extent (2925)
5646 - don't oops on a deleted current working directory (2399)
5647 - handle hard links to targets without a parent properly (2517)
5648 - don't dereference NULL lock when racing during eviction (2867)
5649 - don't grow lock extents when lots of conflicting locks (2919)
5651 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
5654 - account for cache space usage on clients to avoid data loss (974)
5655 - lfsck support in lustre kernel code (2349)
5656 - reduce journal credits needed for BRW writes (2370)
5657 - orphan handling to avoid losing space on client/server crashes
5658 - ptlrpcd can be blocked, stopping ALL progress (2477)
5659 - use lock value blocks to assist in proper KMS, faster stat (1021)
5660 - takes i_sem instead of DLM locks internally on obdfilter (2720)
5661 - recovery for initial connections (2355)
5662 - fixes for mds_cleanup_orphans (1934)
5663 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
5664 - block all file creations until orphan recovery completes (1901)
5665 - client remove rq_connection from request struct (2423)
5666 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
5667 - recovery timer race (2670)
5668 - mdc_close recovey bug (2532)
5669 - ptlrpc cleanup bug (2710)
5670 - mds timeout on local locks (2588)
5671 - namespace lock held during RPCs (2431)
5672 - handle interrupted sync write properly (2503)
5673 - don't try to handle a message that hasn't been replied to (2699)
5674 - client assert failure during cleanup after abort recovery (2701)
5675 - leak mdc device after failed mount (2712)
5676 - ptlrpc_check_set allows timedout requests to complete (2714)
5677 - wait for inflight reqs when ptlrpcd finishes (2710)
5678 - make sure unregistered services are removed from the srv_list
5679 - reset bulk XID's when resending them (caught by 1138 test)
5680 - unregister_bulk after timeout
5681 - fix lconf error (2694)
5682 - handle write after unfinished setstripe, stripe-only getstripe (2388)
5683 - readahead locks pages, leaves pending causing memory pressure (2673)
5684 - increase OST request buffers to 4096 on large machines (2729)
5685 - fix up permission of existing directories in simple_mkdir (2661)
5686 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
5687 - don't assign transno to errored transactions (2742)
5688 - don't delete objects on OST if given a bogus objid from MDS (2751)
5689 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
5690 - if rq_no_resend, then timeout request after recovery (2432)
5691 - fix MDS llog_logid record size, 64-bit array alignment (2733)
5692 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
5693 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
5694 - MDS recovery shouldn't delete objects that it has given out (2730)
5695 - if enqueue arrives after completion, don't clobber LVB (2819)
5696 - don't unlock pages twice when trigger_group_io returns error (2814)
5697 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
5698 - don't write pages to disk if there was an error (1450)
5699 - don't ping imports that have recovery disabled (2676)
5700 - take buffered bytes into account when balancing socknal conn (2817)
5701 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
5702 - reconnect unlink llog connection after MDS reconnects to OST (2816)
5703 - remove little-endian swabbing of llog records (1987)
5704 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
5705 - retry reposting request buffers if they fail (1191)
5706 - grow extent at grant time to avoid granting a revoked lock (2809)
5707 - lock revoke doesn't evict page if covered by a second lock (2765)
5708 - disable VM readahead to avoid reading outside lock extents (2805)
5710 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
5711 - updated kernel patches for hp-2.4.20 kernel (2681)
5713 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
5716 - fix truncated write corruption (2366)
5717 - fix for failed assertion in iopen_connect_dentry (1792,2517)
5719 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
5720 - ensure len(uuid) < 37 in lmc (1171)
5721 - fix ia64 OOPS in llog_test (2255)
5722 - zero end of page at obdfilter for partial page writes (2648)
5723 - don't leave stale dentries around after renames (bug 2428)
5724 - fix timeouts when evicting a client with a single lock held (2642)
5725 - set deadline for the initial HELLO message to drain (2634)
5726 - print out dotted-quad IP addresses in the socknal (2302)
5728 - additional debugging for MDS client eviction problem (2443)
5729 - fix mkfsoptions support for osts (2603, 2604)
5731 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
5734 - add series for the vanilla 2.6.0 kernel
5735 - add series for the vanilla 2.4.24 kernel
5736 - add series for a cray x86/64 UL kernel drop
5737 - fix xattr patches for the vanilla 2.4.19 series
5739 - generate true UUIDs in lmc (1171)
5740 - have portals stack dumping break in UML (2466)
5741 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
5742 - allocate with _NOFS in ldlm to avoid deadlock (1933)
5743 - wake callback waiting threads on client eviction (2460)
5744 - Add --ptldebug and --subsystem to lmc (1719)
5745 - update assertion to allow safe interrupt allocation
5746 - set rq_no_resend for cancel requests (2432)
5747 - recalculate ptlrpcd timeout after resend (2494)
5748 - call vfs_rmdir when removing pending directories (2368)
5749 - fix renaming a file to itself (2429)
5750 - lmc creates a default one-stripe lov (2454)
5751 - expand procfs space to handle large clusters (2326)
5752 - increase UML stack to avoid overflow
5753 - update lconf's list of debug and subsystem masks
5754 - fix lfs find --obd (2510)
5755 - /proc tunable for disabling filter read caching (2591)
5756 - stop rpm packages from altering slapd.conf (2301)
5757 - disable nagle in the socknal under 0conf (2578)
5758 - choose mds inode size based on stripe count (2572)
5759 - fix kernel-source rpm problems (2516)
5761 - add --disable-doc to avoid pdf generation (2421)
5762 - update documentation, tests, type-os, comments
5763 - avoid format warnings on ia64
5764 - remove the TOE NAL
5765 - tiny code cleanups by removing unused fields
5767 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
5770 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
5771 - remove the most blatant lies from BUILDING file (2371)
5772 - change default debug level to reasonable production setting
5773 - reduce client side cache size to reduce cache flush time
5774 - reduce max RPCs in flight to avoid unnecessary file fragmentation
5775 - make TCP zerocopy and pinger support enabled by default (2476)
5776 - sync writes completed after process exits caused crashes (2319)
5777 - maintain correct mount count on the MDS (2356)
5778 - backout 1557, because 2316 wasn't really fixed
5779 - better file I/O statistics gathering in /proc
5780 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
5781 - another kernel patch to fix zero-copy TCP function export
5782 - don't take duplicate lock when processing re-sent getattr (2420)
5783 - lctl uses obd_self_export instead of creating new conn (2353)
5784 - MDS/OST recovery case which requires object creation asserted (2425)
5785 - move lfs from /usr/sbin to /usr/bin in packages
5786 - fix race between mds_client_add and mds_client_free (2417)
5787 - use kmalloc instead of slabs in portals (2430)
5788 - don't create duplicate records when a failover MDS is present (2442)
5789 - remove unnecessary mount age check (2332)
5790 - don't remove directory inodes from locks prematurely (2451)
5791 - don't break if MDS service name is the same as hostname (2103)
5792 - fix races in client write RPC generation when cache full (2482)
5794 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
5797 - remove now-unused request->rq_obd (278)
5798 - if an allocation fails, print out how much memory we've used (1933)
5799 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
5800 - add the "configurable stack size" patch to most series files (1256)
5801 - ability to write large log records, for 100+ OST configs (2306)
5802 - fix NULL deref when filter_prep fails (2314)
5803 - fix operator precedence error in filter_sync
5804 - dynamic allocation of socknal TX descriptors (2315)
5805 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
5806 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
5807 - ensure that CWARN messages in Portals always get to the syslog
5808 - __init/__exit are not for prototype decls (ldlm_init/exit)
5809 - x86-64 compile warning fixes
5810 - fix gateway LMC keyword conflict (2318)
5811 - fix MDS lock inversions in getattr/reint paths (1844)
5812 - fix a rare lock re-ordering bug, which caused deadlock (2322)
5813 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
5814 - DLM race condition prevented some lock evictions (2328)
5815 - ENOMEM detection and retry on socknal sends (2230)
5816 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
5817 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
5818 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
5819 - better error messages when a client is rejected during recovery (1505)
5820 - avoid cancelling locks which were never granted, after failure (2330)
5821 - fix i_sem/journal inversion in mds_client_add (2333)
5822 - fix truncate/getattr lock cycle deadlock (2334)
5823 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
5824 - fix two rare exit paths which could leak an l_lock() ref (2321)
5825 - fencepost error in MDS/OST orphan recovery (2226)
5826 - make log record alignment 8 bytes (1988)
5827 - lstripe now fails when requested offset > ost_count (2237)
5828 - ensure that all kernel series have a complete list.h (1607)
5829 - fix crashes in special-file operations (2316)
5830 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
5832 - allow configurable automake binary, for testing new versions
5833 - small update to the lfs documentation
5835 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
5837 * fix negative export reference count in fsfilt_sync (2312)
5839 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
5840 * release candidate 0.9.1
5842 - orphans are moved into the PENDING directory for possible recovery
5843 - replayed opens now open by fid for orphan/rename safety (1042)
5844 - last close of an orphan inode generates a transno (683)
5845 - chdir() and mount() now pin the directory entry (1020)
5846 - avoid CERROR in normal ll_setattr_raw() error case (1500)
5847 - discard very old requests without processing them (1502)
5848 - remove some common, well-understood CERRORs (1505)
5849 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
5850 - clear "grant" flags in OST replies until OST grant code lands (1644)
5851 - fix read performance by not clobbering i_blksize on client (1598)
5852 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
5853 - make LRU size a /proc tunable, clears locks when reduced (707)
5854 - fix some lprocfs rot that prevented ptlbd from loading (1732)
5855 - server locks take references on exports now (1558)
5856 - build fixes for 2.4.20-rh trees (1663)
5857 - return an error from lov_create if all OSCs are inactive (1751)
5858 - fix import levels when a reconnect happens without a timeout (1597)
5859 - exit early from mds_open if we get a lookup error (1749)
5860 - partial page read at EOF wouldn't wait for disk before sending (1642)
5861 - avoid NULL deref in obdfilter when reading page past EOF (1592)
5862 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
5863 - fix LBUG in ll_it_open_error with rc = -2 (1861)
5864 - write/truncate lock inversion (1639)
5865 - Don't auto-load obdclass, portals modules during cleanup (1495)
5866 - fix timestamps from jumping to "now" (1763)
5867 - extra journal assertions (1648)
5868 - add an extra multiunlink test (1771)
5869 - fix read_record/write_record API (1776)
5870 - fix leak of offset_extent, possible incorrect i_size later (1772)
5871 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
5872 - Debugging for the kqswnal_get_idle_tx problems (1820)
5873 - Allow recovery to be attempted multiple times (1536)
5874 - Write out MDS last_rcvd file after it is first created (1600)
5875 - Fix tx_descriptor leak in failed transmit situations (1827)
5876 - ext3 journaling fixes for assertion failure after IO error (1871)
5877 - class_export_put() on freed export after completion AST error (1896)
5878 - Fix revalidate looping in VFS (1322)
5879 - Don't access a freed export during MDS_REINT timeout (1521)
5880 - Add open-unlink recovery support on the MDS (1673,1764)
5881 - Return an error if no MDS data was read from last_rcvd (1946)
5882 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
5883 - Don't setup a disk device that doesn't match exported UUID (317)
5884 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
5885 - avoid committing NULL handle in force close
5886 - local.sh is now a one-stripe LOV configuration
5887 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
5888 - fix ext3 htree duplicate directory entry corruption (1516)
5889 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
5890 - update to new LOV EA format (2097)
5891 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
5893 2003-06-15 Phil Schwan <phil@clusterfs.com>
5896 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
5897 - per-import/export recovery handling (958, 931, 959)
5898 - multiple last-rcvd slots, for serving multiple FSes (949)
5899 - connections are again shared between multiple imp/exports (963, 964)
5900 - "umount -f" would hang if any requests needed to be sent (393, 978)
5901 - avoid pinning large req buffer by copying for queued messages (989)
5902 - add "uuid" to "lctl device" command to help upcalls (991)
5903 - "open" RPCs with transnos would confuse recovery counters (1037)
5904 - do proper endian conversion of all wire messages (288, 340, 891)
5905 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
5906 - call waiting locks callback from LDLM recovery thread (1127, 1151)
5907 - fix ptlrpc_connection leak in target_handle_connect (1174)
5908 - fix import refcounting bug in OST and MDS cleanup (1134)
5909 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
5910 - fix very unlikely obd_types race condition (501)
5911 - remove osc_open hack for echo_client (1187)
5912 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
5913 - a failure in read_inode2 leads to deadlock (1139)
5914 - cancel ack-locks as soon as transaction is committed (1072)
5915 - fix major leaks and crashes in the bulk I/O path (937, 1057)
5916 - make sure to commitrw after any preprw to avoid deadlock (1162)
5917 - failing to execute a file in a lustre FS would lock inode (1203)
5918 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
5919 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
5920 - obd_uuid2tgt would walk off the end of the list (1255)
5921 - on IA64 the osc would give portals incorrect bulk size (1258)
5922 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
5923 - fix lock inversion caused by new llite matching code (1282)
5924 - limit the number of dirty pages on a client to 10MB (1286)
5925 - timed out locks were not being corrected cancelled (1289)
5926 - fix O_DIRECT above 4GB on IA-32 (1292)
5927 * major user-visible changes
5928 - fail out/fail over policy now controlled by the upcall (993)
5930 - add OBD_PING to check server availability and failure (954)
5931 - lustre messages are now sent in sending host order (288, 340, 891)
5932 - add eadatalen to MDS getattr reply (340)
5933 - OST read replies may contain second buffer, with per-page status (593)
5935 2003-03-11 Phil Schwan <phil@clusterfs.com>
5938 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
5939 - failed open()s could cause deadlock; fixed (867, 869)
5940 - stop cancelling OST locks when files are closed (481)
5941 - overlapping XID spaces caused network corruption (851, 853)
5942 - fix unsafe fsfilt counter arithmetic; change to atomic_t
5943 - setattr_raw added, to do single-RPC, server-side setattrs
5944 - lmc/lconf syntax change for OST UUIDs
5945 - fix crashy race condition between ptlrpc_free_req and osc_close
5946 - don't use request in mdc_enqueue if we hit a timeout (889)
5947 - don't set the inode i_size for regular files from the MDS (896)
5948 - handle out of order completion AST (842)
5949 - don't LBUG if a lock request times out after receiving AST (913)
5950 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
5951 - if a bad lock AST arrives, send an error instead of dropping entirely
5952 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
5953 - fix leak in bulk IO when only partially completed (899, 900, 926)
5954 - fix O_DIRECT for ia64 (55)
5955 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
5956 - C-z after timeout could hang a process forever; fixed (977)
5958 - client-side I/O cache (678, 924, 929, 941, 970)
5960 - READPAGE and SETATTRs which don't take server-side locks get
5963 2003-02-11 Phil Schwan <phil@clusterfs.com>
5966 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
5967 - Fix fsfilt_extN_readpage() to read a full page of directory
5968 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
5969 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
5970 - Fix imbalanced LOV object allocation and out-of-bound access (469)
5971 - Most intent operations were removed, in favour of a new RPC mode
5972 that does a single RPC to the server and bypasses most of the VFS
5973 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
5974 - Aggressively cancel local locks on DLM servers
5975 - mds_reint_unlink sends EA to the client if it's the last nlink.
5976 client uses that EA to unlink OST objects.
5977 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
5978 - recursive symlinks were fixed (439)
5979 - fixed NULL deref in DEBUG_REQ
5980 - filter_update_lastobjid no longer calls sync, which annoyed extN
5981 - fixed multi-client small-writes to a single file problem (445)
5982 - fixed mtime updates during file writes (607)
5983 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
5984 - fixed bug in obd_brw_read/write() (under guise of testing 367)
5985 - fixed Linux OST size reporting problem (444, 656)
5986 - OST now updates object mtime with writes or setattr (607, 619)
5987 - client verifies file size before zeroing page past EOF (445)
5988 - OST now writes last allocated objid to disk with allocation (108)
5989 - LOV on echo now works (409)
5991 - mds_reint_unlink sends a new buffer, with the EA included. this
5992 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
5993 set if a regular file was being unlinked, and it was the last link
5994 - use PtlGet from the target for bulk writes (315)
5995 - OST now updates object mtime with writes or setattr (607, 619)
5996 - LDLM now has a grant-time callback to revalidate locked items, if
5998 - Many MDS operations were reorganized to combat race conditions
6000 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
6001 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
6003 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
6006 - Fully reactivate OST imports after reconnection (512, others)
6007 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
6008 - More graceful error handling for truncating on dead OST (515)
6009 - Don't error out unless we're actually accessing dead stripes (474)
6010 - Fix garbage sizes when stripes are missing (410)
6011 - LRU counters were broken, causing constant lock purge (433, 432)
6012 - garbage on read from stripes with failed OSTs (441)
6013 - mark OSCs as active before reconnecting during recovery (438)
6014 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
6015 - lfind did not preserve OST order in output (443)
6016 - symlinks cause hung clients, incorrect data (439)
6017 - stop dereferencing request after dropping refcount (457)
6018 - don't LASSERT(spin_is_locked) on non-SMP (455)
6019 - fixes for many rename() bugs
6020 - fstat didn't correctly synchronize attributes (399)
6021 - server must handle lock cancellation during blocking AST prep (487)
6022 - bulk descriptors were free()d too soon (511)
6023 - fix paths in lconf, which would load incorrect modules (451, 507)
6024 - fix confusing lconf 'host not found' error message (386)
6025 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
6026 - fix race condition in mdc_blocking_ast() for inode access (526)
6027 - fix lov_unpackmd() unpacking wrong number of stripes (537)
6028 - fix lov_set_osc_active() marking wrong OSC inactive (440)
6029 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
6030 - fix multiple writes of stripe MD to MDS (358, maybe 519)
6031 - fix lstripe in several ways (kernel side) (527)
6032 - fix request leak in ldlm_cli_enqueue (262)
6033 - incorrect OSC was marked inactive after OST failure
6034 - call mds_fs_cleanup before unmounting filesystem (524)
6035 - fix races between taking ns_lock and ldlm_lock_change_resource
6036 - fix races updating LOV export open file list
6037 - fix lov_enqueue error path, avoid decref-ing bad lock handle
6038 - fix recovery NULL deref in ldlm_cli_cancel_unused
6039 - fix some DLM races by using new hash table for lock handles (419)
6040 - permit the client to specify desired inodes, at replay
6041 - duplicate requests when we queue them for replay reintegration
6042 - fix last_rcvd offset calculation
6043 - sync after each recovered transaction, so we always make progress
6044 - never, not always, ERESTART requests without transnos
6045 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
6047 - skip replay if the MDS says that the client is already connected
6048 - don't check for a recovery-enabled export to match lctl's UUID
6049 - don't INC_USE_COUNT for phantom exports
6050 - don't crash when cleaning up phantom exports (567)
6051 - don't double-finish or set replay data for errored mdc_open requests
6052 - abort requests when they time out, so we don't get old replies
6053 - send/receive replies for AST messages again
6054 - if the client says that it doesn't have the lock, cancel it on the
6056 - if we timeout during I/O, don't try to cancel an in-use lock; instead
6057 mark it as destroyed, it will all work out when decref is called
6058 - fix module use counts (22, 581)
6060 - ASTs now expect a reply (server cancels lock on error reply)
6062 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
6065 - fix many simultaneous client startup (392)
6066 - fix dentry->d_it clobbering
6067 - credentials weren't being shipped for readdir/getattr operations
6068 - remove invalid assertions triggered during some concurrent MD
6070 - proper Lustre versions added (336, 389)
6071 - fix memory leak for create error case (398)
6072 - fix LOV locking bug that would get cli/srv out of sync
6073 - fix echo client over LOV (409)
6074 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
6075 - fix double-O_EXCL intent crash (424)
6076 - avoid sending multiple lock CANCELs (352)
6078 - MDS can do multi-client recovery (modulo bugs in new code)
6080 - many updates, edits, cleanups
6082 2002-11-18 Phil Schwan <phil@clusterfs.com>
6085 - fix null d_it dereference (346)
6086 - fix full OST/dbench hang (333)
6087 - fix permission problem with file removal (286)
6088 - fix removal of OSCs from LOV when they fail
6089 - fix NULL deref during bulk timeout (214)
6090 - fix problems related to multiple filesystems on one MDS (241)
6091 - fixed serious subtle metadata locking bugs
6092 - free locks on clients when inodes are removed due to memory
6094 - fix inode pointer in lock data (285)
6095 - partial support for multiple MDS on a single host (241)
6096 - data locks weren't cancelled at clear_inode time (290, 311)
6097 - intent locks could lead to unbounded lock growth (205)
6098 - added a maximum lock count, an LRU list, and a flusher
6099 - fix multiple rename (365)
6100 - properly abstracted the echo client
6101 - OSC locked 1 byte too many; fixed
6102 - rewrote brw callback code:
6103 - fixed recovery bugs related to LOVs (306)
6104 - fixed too-many-pages-in-one-write crash (191)
6105 - fixed (again) crash in sync_io_timeout (214)
6106 - probably fixed callback-related race (385)
6108 - Add capability to MDS protocol
6109 - LDLM cancellations and callbacks on different portals
6111 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
6114 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
6115 - defer open object destruction to close time (601981, 138)
6116 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
6117 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
6118 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
6119 - don't call obd_setattr in ll_file_release for destroyed objects
6120 * protocol change to lustre_msg: move |version| and add |flags|
6121 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
6122 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
6123 * added replay of create, unlink, link and rename operations during
6124 MDS failover; recovery should be much more robust now
6125 * remove failed OSCs from LOVs (only lov_create uses this so far)
6126 * the lustre-HOWTO was brought (more) up to date (582544)
6128 2002-10-23 Phil Schwan <phil@clusterfs.com>
6131 - in-use dentries weren't being reused properly (617851)
6132 - prevent multiple LDLM setup (599178)
6133 - fix LOV size calculations for truncate (617853)
6134 - fix client handling of MDS intent errors (POSIX)
6135 - fix permission bug in lovstripe.c test (624321)
6136 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
6137 - truncate past end of file could corrupt data
6138 - proper cleanup after timeouts, crashes, etc (592524, 550815)
6139 - a race in recovery could return ETIMEDOUT to apps (623947)
6140 - building outside the source directory was fixed
6141 * the lustre-HOWTO was brought (more) up to date (582544)
6142 * major progress was made on recovery functionality
6144 2002-10-10 Phil Schwan <phil@clusterfs.com>
6147 - recovery deadlock fix
6148 - rm -rf causes LBUG fix (617817)
6149 - file open by multiple tasks fix (618962)
6150 - directory permissions bugs (602707 and 620007)
6151 - journal_stop fixed with locking (611313)
6152 - O_APPEND failures resolved (618273, perhaps 614459)
6153 - lconf PATH fix (619770)
6154 - IA64 build fix (621450)
6155 - RPC buffer sizes scale with amount of memory
6157 2002-10-01 Phil Schwan <phil@clusterfs.com>
6160 - locks would be cancelled without throwing away data pages,
6161 resulting in inconsistent data (605627)
6162 - inode attributes were not always being refreshed (605627, 612449)
6163 - lconf now continues to cleanup after lctl reports an error
6164 - MDS now enforces user permissions (602707)
6165 - lprocfs cleanup fixed, but not yet enabled (614157)
6166 - fixed infinite server hang, should a client not respond to an AST
6167 - avoid going into recovery if user calls readlink() with a buffer
6168 that's too small (613941)
6169 - AST RPCs no longer require replies (614867) -- this may be changed
6170 - don't crash server if client sends an IOV that's too big (611336)
6171 - fixed lock conversion deadlock (611892)
6172 - fixed the following of symlinks (614622)
6173 * recovery: the server can remove locks from a client that dies, other
6174 clients can make progress
6175 * more extN patch fixes
6176 * compile-time configurable ptlrpc buffer allocations
6178 - collaborative read cache document
6179 - Lustre Lite Performance CDR document-in-progress
6181 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6184 - fix typo in patch-2.4.18
6186 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6189 - clear ptlrpc request each time in handle_incoming_request()
6190 - unlink of files now destroys the object on the OST
6192 2002-09-19 Peter Braam <braam@clusterfs.com>
6194 * add hard link support
6195 * change obdfile creation method
6196 * kernel patch changed
6198 2002-09-19 Peter Braam <braam@clusterfs.com>
6201 - stack overflow bug in extN fixed
6203 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
6205 * documentation updates
6206 - add man pages for config tools
6207 - update tests/README to describe testing with new config tools
6208 - finish metadata API descriptions
6209 * bug fixes and cleanups
6210 - statfs workaround for 16TB limit
6211 - LOV stripe allocation improved, can stripe on subset of OSTs
6212 - LOV file size/IO offset was wrong for files > 4GB in size
6213 - object EA data was being dropped, caused files to be unreadable
6214 - memory overflow with non-LOV OST caused memory corruption
6215 - fixed regression tests to work with new config tools, obdfilter
6216 - fixed bug when directory size became larger than 1 block
6217 - fixed bug (for single client case) when PWD was deleted
6218 - invalidate local directory pages when doing intent-based ops
6219 - avoid LDLM oops when lock callback contained bad data
6221 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
6223 * documentation updates
6224 * bug fixes and cleanups
6225 - configuration tools
6228 - 64-bit compile warnings
6229 - 64-bit internal statfs data
6231 * test_brw on persistent OST devices
6233 * lprocfs (disabled)
6235 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
6237 * documentation updates
6238 * bug fixes and cleanups
6239 * configuration tools
6241 2002-08-30 Peter J. Braam <braam@clusterfs.com>
6244 * many small fixes to 0_5_4
6245 * io/network handling
6246 * thinkos in MDS operations
6248 2002-08-24 Peter J. Braam <braam@clusterfs.com>
6251 * crucial basic fixes to 0.5.3
6252 * IOR, Iozone work over Elan
6255 2002-08-07 Phil Schwan <phil@clusterfs.com>
6256 * version 0_5_3, our first alpha
6257 * we use the new Portals iovs
6258 * documentation updates
6259 * bug fixes and cleanups
6260 * small changes in the DLM wire protocol
6262 2002-07-25 Peter J. Braam <braam@clusterfs.com>
6263 * version 0_5_1 with some initial stability,
6264 * locking on MD and file I/O.
6265 * documentation updates
6266 * several bug fixes since 0.5.0
6267 * small changes in wire protocol
6269 2002-07-18 Phil Schwan <phil@clusterfs.com>
6271 * delivered as Lustre Light Alpha
6272 * fixed a crash after handling invalid MDS requests
6273 * fixed directory pages for architectures with non-4k pages sizes
6275 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
6276 * release version v0_4_4
6277 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
6279 * Fixes a number of interruption problems with OST operations.
6280 * Update documentation for portals header changes
6281 * Move all wire protocol structs/defines to lustre_idl.h
6282 * Fixes symlink length bug.
6283 * Add tcpdump to repository.
6285 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
6286 * release version v0_4_3
6287 * Fixes statfs for inodes on extN.
6288 * Fixes bug in runtests which would delete /etc/hosts.
6289 * Use 64-bit object IDs wherever possible (not into VFS though)
6290 Remove ost_get_info, which is unused by lustre, and out of date.
6292 2002-07-03 Peter Braam <braam@clusterfs.com>
6293 * release version v0_4_2 Fixes a lookup error (type not passed)
6294 * move forward to head of Portals
6295 * move forward to latest Lustre kernel
6297 2002-06-25 Peter Braam <braam@clusterfs.com>
6298 * release version v0_4_1. Hopefully stable on single node use.