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.
17 * NFSv4 reexport requires Lustre client on the re-exporting nodes to be
18 mounted with "32bitapi" mount option
20 Severity : enhancement
21 Description: Update RHEL5.6 kernel to 2.6.18-238.9.1.el5
24 Frequency : only with SLES10
26 Description: Use OFED "KMP" provided by Novell
27 Details : SLES10 SP3 ships with OFED in a separate "KMP" package.
28 Lustre is now built against this package. That means you need to
29 install the ofed-kmp package from Novell for the patchless client.
30 Note that the ofed-kmp that Novell ships may not exactly match the
31 kernel version but should still be compatible.
33 Severity : enhancement
35 Description: Update RHEL5.5 kernel to 2.6.18-194.17.1.el5
36 Update OEL5.5 kernel to 2.6.18-194.17.1.0.1.el5.
38 --------------------------------------------------------------------------------
40 2010-07-15 Oracle, Inc.
42 * Server support for kernels:
43 2.6.18-164.11.1.el5 (RHEL 5)
44 2.6.18-164.11.1.0.1.el5 (OEL 5)
45 * Client support for unpatched kernels:
46 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
47 2.6.18-164.11.1.el5 (RHEL 5),
48 2.6.18-164.11.1.0.1.el5 (OEL 5)
49 2.6.16.60-0.42.8 (SLES 10),
51 2.6.29.4-167.fc11 (FC11)
52 * Recommended e2fsprogs version: 1.41.10-sun2
54 Severity : enhancement
56 Description: Distribute lfs_migrate script with lustre
57 Details : The lfs_migrate mentioned in the manual is now included in the
58 lustre tree in lustre/scripts and will distributed in lustre RPMs.
60 Severity : enhancement
62 Description: Failover nid list added to import proc file.
64 Severity : enhancement
66 Description: "som_preview" mount option required to use SOM
67 Details : The clients will need to use som_preview mount option
71 Description: Lustre does not do 1MB IOs to HW RAID
72 Details : Bump MAX_PHYS/HW_SEGMENTS and SG_ALL to 256 in the RHEL5 kernel.
74 Severity : enhancement
76 Description: lfs quota output is not very convenient for awk/sed-parsing
77 Details : Some positions in lfs quota output table could be empty or
78 non-empty which made it hard to parse it with scripts, now
79 a dash is put instead of space where there is not supposed
84 Description: Update to OFED 1.5.1
85 Details : OFED 1.5.1 will be used in environments that don't use
90 Description: Remove lnet. prefix from lctl params display
91 Details : Also renamed lnet.memused to lnet_memused to
96 Description: Use conf_param -d to remove permanant settings.
100 Description: Add list_param -R
101 Details : lctl list_param now supports recursive mode.
104 Bugzilla : 21829, 22312
105 Description: Fix llobdstat.
106 Details : 1. Now updating the write_bytes counter
107 2. Port script fixes from 1.8.
111 Description: decrease the usage of memory on clients.
112 Details : 1. On clients, recycle dentries and inodes unused.
113 2. Delete the code related to ll_deathrow(att 6215 in bug 1443). It
118 Description: llobdstat fix and enhancement
119 Details : add a counter to set a limit to how many samples will be returned
120 fix a wildcard in the path to limit to obdfilter stats only
124 Description: lfs check only allowed for root
125 Details : Code cleanups and added a regresion test of non-root use of
130 Description: MDS Oopses in ptlrpc_dispatch_difficult_reply()/__list_add()
131 Details : Fix hr->hr_index overload.
135 Description: Fix FIEMAP functionality.
139 Description: FC11 patchless client support added
143 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and
144 OEL5.4 kernel to 2.6.18-164.11.1.0.1.el5.
148 Description: New acc-sm script and various test-framework enhancments.
152 Description: Previously landed patch was reverted.
156 Description: Update SLES10 kernel to 2.6.16.60-0.42.8 (patchless)
160 Description: Add ldiskfs maxdirsize mount option.
164 Description: Fix mkfs.lustre support for 16TB LUNs.
168 Description: Print mmp_check_interval
169 Details : print mmp_check_interval and make it possible to abort
170 mount operation in case it takes too long.
174 Description: Add cascading_rw to lustre mpi tests.
178 Description: Store SOM data in EA.
179 Details : Also several SOM related cleanups.
183 Description: Add refcount to md_open_data
187 Description: Use refcount on lov_request_set to prevent memory corruption.
191 Description: Statahead: use parent dir's i_mutex for synchronization.
195 Description: loadgen improvements.
199 Description: Enable peer health detection in ksocklnd and ko2iblnd.
203 Description: Control DCACHE_LUSTRE_INVALID flag with MDS_INODE_LOOKCUP lock
207 Description: Fix race on obd_devcie:md_stats
211 Description: Fix SLES 11 client builds.
212 Details : Resolved uninitalized pointer gcc warnings.
216 Description: Allow non-root access for "lfs check".
217 Details : Added a check in obd_class_ioctl() for OBD_IOC_PING_TARGET.
221 Description: Clear imp_force_reconnect correctly in ptlrpc_connect_interpret()
223 Severity : enhancement
225 Description: Update kernel to RHEL5.4 2.6.18-164.6.1.el5 and
226 OEL5 2.6.18-164.6.1.0.1.el5(Both in-kernel OFED enabled).
230 Description: Sanity test_27p fails.
231 Details : OBD_FAIL_OST_ENOSPC has been moved to filter_statfs().
232 Sanity tests 27 were cleaned up and improved.
236 Description: Unify req format on client/servers
237 Details : Use new req_capsule API [almost] everywhere instead of old PTLRPC
238 buffers and swabbers approach..
243 Description: drop join file stuff.
245 Severity : enhancement
247 Description: Add 2.6.27 and SLES11 (patchless client) support
251 Description: Fix in ptlrpc_expire_one_request() to print the signed time
254 Severity : enhancement
256 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
259 Severity : enhancement
261 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
263 Severity : enhancement
265 Description: correctly handle big reply message.
266 Details : send LNet event if reply is bigger then buffer and adjust this buffer
271 Description: Drop unnecessary __GFP_NOMEMALLOC flag from filter_get_page()
273 Severity : enhancement
275 Description: Add support for OEL5.
279 Description: Conf-sanity.sh 50g test - deactivated OST should not cause a panic.
281 Severity : enhancement
283 Description: "weak-modules" support
284 Details : Implement "weak-modules" support which enables kernel modules
285 to be used with any kernel that implements the same kABI. In
286 order to achieve this modules are now installed in
287 /lib/modules/$(uname -r)/updates/kernel on all distributions.
291 Description: client could not reconnect to OST because of an active request.
292 Details : The client now retries to reconnect to the same server, if a
293 connect request failed with EBUSY or -EAGAIN. Test case 35b
294 conf-sanity.sh has been created.
298 Description: Don't allow make backward step on assignin osc next id.
299 Details : race between allocation next id and ll_sync thread can be cause
300 of set wrong osc next id and can be kill valid ost objects.
304 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
308 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
312 Description: Error handling in osc_statfs_interpret() has been improved.
313 Details : Check in osc_statfs_interpret() for EBADR.
317 Description: Do not send statfs() requests to OSTs disabled by administrator.
318 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
320 Severity : enhancement
322 Description: Added sanity-quota test to for lfs(1) to work with numeric uid/gid.
324 Severity : enhancement
326 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
328 Severity : enhancement
330 Description: Enhancement for lfs(1) command to use numeric uid/gid.
334 Description: Improved cleanup procedure for conf-sanity test_46a.
338 Description: Do not start bulk transfer, if client tried to reconnect.
339 Details : Do not start bulk transfer, if client tried to reconnect,
340 but failed due to active rpcs. Check for exp_abort_active_req
341 has been implemented.
345 Description: Prevent inconsistences between linux and lustre mount structures.
346 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
347 Make the sleep interruptible.
349 Severity : enhancement
351 Description: provide server to client comms path
352 Details : server to client communications path via new LDLM_SET_INFO rpc
356 Description: truncate starts GFP_FS allocation under transaction causing deadlock
357 Details : ldiskfs_truncate calls grab_cache_page which may start page
358 allocation under an open transaction. This may lead to
359 calling prune_icache with consequent lustre reentrance.
363 Description: send correctly lsm on open replay
364 Details : MDS is trust to LSM size on replay open, but client can set wrong size
369 Description: enable client interface failover
370 Details : When a child reconnects from another NID, properly update export
371 nid hash position and ldlm reverse import.
373 Severity : enhancement
375 Description: Communicate OST degraded/readonly state via statfs to MDS
376 Details : Flags in the statfs returned from OSTs indicate whether the
377 OST is in a degraded RAID state, or if the filesystem has
378 turned read-only after a filesystem error is detected.
383 Description: don't panic if EPROTO was hit when reading symlink
384 Details : correctly handling request reference in error cases.
386 Severity : enhancement
388 Description: Add LustreNetLink, a kernel-userspace communcation path.
390 Severity : enhancement
392 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
397 Description: access to llog context before init.
398 Details : move handling CATALOGS file at osc layer and forbid access to llog
402 Frequency : rare, only when client is writing to OST on local node
404 Description: Avoid deadlock for local client writes
405 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
406 memory freeing context. This allows OST threads to set the
407 PF_MEMALLOC flag on task structures in order to allocate memory
408 from reserved pools and complete IO.
409 Use GFP_HIGHUSER for OST allocations for non-local client writes,
410 so that the OST threads generate memory pressure and allow
411 inactive pages to be reclaimed.
413 Severity : enhancement
415 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
420 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
421 Details : move ldlm namespace creation in setup phase to avoid grab
422 _lprocfs_lock with cli_sem held.
426 Description: Temporarily disable grant shrink.
427 Details : Disable the feature for debugging.
430 Frequency : only when formatting test filesystems
432 Description: Unable to run several mkfs.lustre on loop devices at the same time
433 Details : mkfs.lustre returns error 256 on the concurrent loop devices
434 formatting. The solution is to proper handle the error.
436 Severity : enhancement
438 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
440 Severity : enhancement
442 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
447 Description: resolve race between obd_disconnect and class_disconnect_exports
448 Details : if obd_disconnect will be called to already disconnected export he
449 forget release one reference and osc module can't unloaded.
451 Severity : enhancement
453 Description: Allow tuning service thread via /proc
454 Details : For each service a new
455 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
456 created that can be used to set min/max thread counts, and get the
457 current number of running threads.
462 Descriptoin: don't return error if have partially created objects for file.
463 Details : lov_update_create_set uses set->set_success as index for created
464 objects, so if some requests will be failed, they will have hole
465 at end of array and qos_shrink_lsm for allocate correct lsm.
467 Severity : enhancement
469 Description: Update OFED support to 1.4.1
474 Description: Reduce small size read RPC
475 Details : Set read-ahead limit for every file and only do read-ahead when
476 available read-ahead pages are bigger than 1M to avoid small size
479 Severity : enhancement
481 Description: move AT tunable parameters for more consistent usage
482 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
484 Severity : enhancement
486 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
487 Details : allow skip disconnected ost for send statfs request and hide error
493 Description: fix racy locking of mballoc block bitmaps causing BUG
494 Details : The locking of the mballoc buddy bitmap and the in-memory
495 block bitmap was using two different spin locks in some
496 cases. This made it possible to incorrectly access the
497 mballoc bitmap while another process was modifying it,
498 causing a sanity assertion to fail. While no on-disk corruption
499 was reported, there was some risk of this happening.
502 Frequency : rare, on llog test 6
504 Descriptoin: don't allow connect to already connected import
505 Details : allowing connect to already connected import is hide connecting problem.
510 Descriptoin: allow kill process which wait statahead result
511 Details : for some reasons 'ls' can stick in waiting result from statahead,
512 in this case need way for kill this process.
514 Severity : enhancement
516 Description: Add state history info file, enhance import info file
517 Details : Track import connection state changes in a new osc/mdc proc file;
518 add overview-type data to the osc/mdc import proc file.
520 Severity : enhancement
522 Description: MDS create should not wait for statfs RPC while holding DLM lock.
524 Severity : enhancement
526 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
531 Descriptoin: OST grant too much space to client even there are not enough space.
532 Details : Client will shrink its grant cache to OST if there are no write
533 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
534 this grant cache if there are already not enough avaible space
535 (left_space < total_clients * 32M).
538 Frequency : start MDS on uncleanly shutdowned MDS device
540 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
541 Details : stay in waiting mds<>ost recovery finished produce random bugs
542 due race between two ll_sync thread for one lov target. send
543 ACTIVATE event only if connect realy finished and import have
547 Frequency : rare, connect and disconnect target at same time
549 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
550 Details : don't call obd_disconnect under lov_lock. this long time
551 operation and can block ptlrpcd which answer to connect request.
556 Descriptoin: don't lose wakeup for imp_recovery_waitq
557 Details : recover_import_no_retry or invalidate_import and import_close can
558 both sleep on imp_recovery_waitq, but we was send only one wakeup
562 Frequency : always with long access acl
564 Descriptoin: mds can't pack reply with long acl.
565 Details : mds don't control size of acl but they limited by reint/getattr
568 Severity : enhancement
570 Description: Update to SLES10 kernel-2.6.16.60-0.33.
572 Severity : enhancement
574 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
577 Frequency : start MDS on uncleanly shutdowned MDS device
579 Descriptoin: aborting recovery hang on MDS
580 Details : don't throttle destroy RPCs for the MDT.
583 Frequency : on remount
585 Description: external journal device not working after the remount
586 Details : clear dev_rdonly flag for external journal devices in
592 Description: shutdown vs evict race
593 Details : client_disconnect_export vs connect request race.
594 if client will evicted at this time - we start invalidate
595 thread without referece to import and import can be freed
599 Frequency : rare, need acl's on inode.
601 Description: client can't handle ost additional correctly
602 Details : if ost was added after client connected to mds client can have
603 hit lnet_try_match_md ... to big messages to wide striped files.
604 in this case need teach client to handle config events about add
605 lov target and update client max ea size at that event.
607 Severity : enhancement
609 Description: Changelogs
610 Details : Changelogs are a lightweight mechanism to track filesystem
611 metadata and namespace changes. The changelog is recorded
612 permanently on the MDTs, and is periodically "consumed" / purged
613 when records are no longer needed.
615 Severity : enhancement
617 Description: compact fld format with extents
618 Details : Store range of seq rather than every seq in FLD. Seq
619 controller update FLD rather than clients. In Case of CMD, mdt0
620 has FLD, all other metadata server act as non persistent proxy
621 for FLD queries and cache fld entries in fld cache.
626 Description: don't skip ost target if they assigned to file
627 Details : Drop slow OSCs if we can, but not for requested start idx.
628 This means "if OSC is slow and it is not the requested
629 start OST, then it can be skipped, otherwise skip it only
630 if it is inactive/recovering/out-of-space.
634 Description: more cleanup in mds_lov
635 Details : not send LOV EA under replay, we can't know about they size at this
636 time. Don't allow client connect to mds before any ost connected,
637 for avoid problems with LOV EA size and returning EIO to client.
639 Severity : enhancement
641 Description: Interoperability at server side (Disk interoperability)
643 Severity : enhancement
645 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
647 Severity : enhancement
649 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
651 Severity : enhancement
653 Description: New client IO stack (CLIO).
655 Severity : enhancement
657 Description: Commit on sharing. Eliminate inter-client dependencies between
658 uncommitted transactions by doing transaction commits.
659 Thereby clients may recovery independently.
662 Frequency : Create a symlink file with a very long name
664 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
665 Details : If there is no extra space in the request for early cancels,
666 ldlm_req_handles_avail() returns 0 instead of a negative value.
668 Severity : enhancement
670 Description: Add /proc entry for import status
671 Details : The mdc, osc, and mgc import directories now have
672 an import directory that contains useful import data for debugging
675 Severity : enhancement
677 Description: Re-disable certain /proc logging
678 Details : Enable and disable client's offset_stats, extents_stats and
679 extents_stats_per_process stats logging on the fly.
682 Frequency : Only on FC kernels 2.6.22+
684 Description: oops in statahead
685 Details : Do not drop reference count for the dentry from VFS when lookup,
686 VFS will do that by itself.
688 Severity : enhancement
690 Description: Generic /proc file permissions
691 Details : Set /Proc file permissions in a more generic way to enable non-
692 root users operate on some /proc files.
696 Description: Hitting mdc_commit_close() ASSERTION
697 Details : Properly handle request reference release in
698 ll_release_openhandle().
702 Description: quota recovery deadlock during mds failover
703 Details : This patch includes att18982, att18236, att18237 in bz14840.
705 1. fix osts hang when mds does failover with quotaon
706 2. prevent watchdog storm when osts threads wait for the
711 Frequency : only patchless client
712 Description: add workaround for race between add/remove dentry from hash
714 Severity : enhancement
716 Description: Allow OST glimpses to return PW locks
720 Description: LBUG when llog conf file is full
721 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
725 Description: Prevent import from entering FULL state when server in recovery
729 Description: service mount cannot take device name with ":"
730 Details : Only when device name contains ":/" will mount treat it as
736 Description: replace ptlrpcd with the statahead thread to interpret the async
737 statahead RPC callback
741 Frequency : on recovery
742 Description: I/O failures after umount during fail back
743 Details : if client reconnected to restarted server we need join to recovery
744 instead of find server handler is changed and process self
745 eviction with cancel all locks.
747 Severity : enhancement
749 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
751 Severity : enhancement
753 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
755 Severity : enhancement
757 Description: Upcall on Lustre log has been dumped
758 Details : Allow for a user mode script to be called once a Lustre log has
759 been dumped. It passes the filename of the dumped log to the
760 script, the location of the script can be specified via
761 /proc/sys/lnet/debug_log_upcall.
766 Description: avoid idr_remove called for id which is not allocated.
767 Details : Move assigment s_dev for clustered nfs to end of initialization,
768 to avoid problem with error handling.
773 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
774 Details : When connection is reused this not moved from CONN_UNUSED_HASH
775 into CONN_USED_HASH and this prodice warning when put connection
776 again in unused hash.
779 Severity : enhancement
781 Description: File striping can now be set to use an arbitrary pool of OSTs.
783 Severity : enhancement
785 Description: Export bytes_read/bytes_write count on OSC/OST.
789 Description: Early reply size mismatch, MGC loses connection
790 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
791 the connect flags are properly negotiated.
796 Description: quotas are not honored with O_DIRECT
797 Details : all writes with the flag O_DIRECT will use grants which leads to
798 this problem. Now using OBD_BRW_SYNC to guard this.
802 Description: add quota statistics
803 Details : 1. sort out quota proc entries and proc code.
804 2. add quota statistics
806 Severity : enhancement
808 Description: enable quota support for HEAD.
812 Description: Properly propagate oinfo flags from lov to osc for statfs
813 Details : restore missing copy oi_flags to lov requests.
815 Severity : enhancement
817 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
818 : l_getgroups(8), lst(8), routerstat(8)
819 Details : included man pages for llobdstat(8), llstat(8),
820 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
822 Severity : enhancement
824 Description: configure's --enable-quota should check the
825 : kernel .config for CONFIG_QUOTA
826 Details : configure is terminated if --enable-quota is passed but
827 : no quota support is in kernel
831 Description: Remove portals compatibility
832 Details : Remove portals compatibility, not interoperable with releases
837 Description: Resolve device initialization race
838 Details : Prevent proc handler from accessing devices added to the
839 obd_devs array but yet be intialized.
841 Severity : enhancement
843 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
845 Severity : enhancement
847 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
852 Description: Using wrong pointer in osc_brw_prep_request
853 Details : Access to array[-1] can produce panic if kernel compiled with
854 CONFIG_PAGE_ALLOC enabled
858 Description: Client runs out of low memory
859 Details : Consider only lowmem when counting initial number of llap pages
863 Description: *optional* service tags registration
864 Details : if the "service tags" package is installed on a Lustre node
865 When the filesystem is mounted, a local-node service tag will
866 be created. See http://inventory.sun.com/ for more information
867 about the Service Tags asset management system.
871 Description: Kernel BUG tries to release flock
872 Details : Lustre does not destroy flock lock before last reference goes
873 away. So always drop flock locks when client is evicted and
874 perform unlock regardless of successfulness of speaking to MDS.
878 Description: add refcount protection for osc callbacks, avoid panic on shutdown
882 Description: sanity test 65a fails if stripecount of -1 is set
883 Details : handle -1 striping on filesystem in ll_dirstripe_verify
888 Description: ASSERTION(CheckWriteback(page,cmd)) failed
889 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
892 Severity : enhancement
894 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
898 Description: do not process already freed flock
899 Details : flock can possibly be freed by another thread before it reaches
900 to ldlm_flock_completion_ast.
904 Description: LBUG during stress test
905 Details : Need properly lock accesses the flock deadlock detection list.
909 Description: oops in page fault handler
910 Details : kernel page fault handler can return two special 'pages' in error
911 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
915 Description: timeout with invalidate import.
916 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
917 should be handled by ptlrpcd. This produce long age waiting and
918 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
920 Severity : enhancement
922 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
926 Description: enable MGS and MDT services start separately
927 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
928 a co-located MGS without starting the MGS, which is a complement
929 to 'nosvc' mount option.
933 Frequency : after recovery
934 Description: precreate to many object's after del orphan.
935 Details : del orphan st in oscc last_id == next_id and this triger growing
936 count of precreated objects. Set flag LOW to skip increase count
937 of precreated objects.
941 Frequency : rare, on clear nid stats
942 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
943 Details : when clean nid stats sometimes try destroy live entry,
944 and this produce panic in free.
948 Description: Stack overflow during MDS log replay
949 ease stack pressure by using a thread dealing llog_process.
953 Description: wait until IO finished before start new when do lock cancel.
954 Details : VM protocol want old IO finished before start new, in this case
955 need wait until PG_writeback is cleared until check dirty flag and
956 call writepages in lock cancel callback.
958 Severity : enhancement
960 Description: using special macro for print time and cleanup in includes.
964 Description: mds_mfd_close() ASSERTION(rc == 0)
965 Details : In mds_mfd_close(), we need protect inode's writecount change
966 within its orphan write semaphore to prevent possible races.
970 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
974 Frequency : rare, on shutdown ost
975 Description: don't hit live lock with umount ost.
976 Details : shrink_dcache_parent can be in long loop with destroy dentries,
977 use shrink_dcache_sb instead.
981 Description: don't panic with use echo client
982 Details : echo client pass NULL as client nid pointer and this produce null
987 Description: fix build on ppc32
988 Details : compile code with -m64 flag produce wrong object file for ppc32.
992 Description: add message levels for liblustreapi
996 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
997 Details : Move the 'good_osts' check before the 'total_bavail' check. This
998 will result in an -EAGAIN and in the exit call path we call
999 alloc_rr() which will with increasing aggressiveness attempt to
1000 aquire precreated objects on the minimum number of required OSCs.
1004 Description: Use old size assignment to avoid deadlock
1005 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
1006 the scheduling while holding a spinlock. We do not need locking
1007 for size in ll_update_inode() because size is only updated from
1008 the MDS for directories or files without objects, so there is no
1009 other place to do the update, and concurrent access to such inodes
1010 are protected by the inode lock.
1014 Description: resolve "_IOWR redefined" build error on SLES10
1018 Description: dump the memory debugging after all modules are unloaded to
1019 suppress false negative in conf_sanity test 39
1021 Severity : enhancement
1023 Description: build kernel-ib packages for OFED 1.3 in our release cycle
1028 Description: fix SLES kernel versioning
1029 Details : the kernel version for our SLES 10 kernel did not include a "-"
1030 before the "smp" at the end. while this was not a problem in
1031 general, it did mean that software trying to use the kernel
1032 version to try to detect a vendor specific kernel would fail.
1033 this was most evident by the OFED build scripts.
1037 Description: Don't update lov_desc members until making sure they are valid
1038 Details : When updating lov_desc members via proc fs, need fix their
1039 validities before doing the real update.
1043 Description: don't put request into delay list while invalidate in flight.
1044 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
1045 import in flight. this produce timeout for invalidate and sometimes
1046 can cause stale data.
1050 Frequency : on ppc only
1051 Description: not convert ost objects for directory because it's not exist.
1052 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
1056 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
1058 Severity : enhancement
1060 Description: Update to support 2.6.22.14 vanilla kernel.
1064 Frequency : rare, on recovery
1065 Description: read procfs can produce deadlock in some situation
1066 Details : Holding lprocfs lock which send rpc can produce block for destroy
1067 obd objects and this also block reconnect with -EALREADY. This
1068 isn't fix all lprocfs bugs - but make it rare.
1070 Severity : enhancement
1072 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
1075 Frequency : frequent on X2 node
1077 Description: mdc_set_open_replay_data LBUG
1078 Details : Set replay data for requests that are eligible for replay.
1082 Description: lustre_mgs: operation 101 on unconnected MGS
1083 Details : When MGC is disconnected from MGS long enough, MGS will evict the
1084 MGC, and late on MGC cannot successfully connect to MGS and a lot
1085 of the error messages complaining that MGS is not connected.
1089 Frequency : on network error
1090 Description: panic with double free request if network error
1091 Details : mdc_finish_enqueue is finish request if any network error ocuring,
1092 but it's true only for synchronus enqueue, for async enqueue
1093 (via ptlrpcd) this incorrect and ptlrpcd want finish request
1096 Severity : enhancement
1098 Description: client-side metadata stat-ahead during readdir(directory readahead)
1099 Details : perform client-side metadata stat-ahead when the client detects
1100 readdir and sequential stat of dir entries therein
1103 Frequency : on start mds
1105 Description: Implement get_info(last_id) in obdfilter.
1108 Frequency : occasional
1110 Description: Correctly check stale fid, not start epoch if ost not support SOM
1111 Details : open with flag O_CREATE need set old fid in op_fid3 because
1112 op_fid2 was overwritten with new generated fid, but mds can answer
1113 with one of these two fids and both is not stale. Setattr
1114 incorrectly started an epoch and assume will be called
1115 done_writing, but without SOM done_writing ever being called.
1118 Frequency : rare, depends on device drivers and load
1120 Description: MDS or OSS nodes crash due to stack overflow
1121 Details : Code changes in 1.8.0 increased the stack usage of some functions.
1122 In some cases, in conjunction with device drivers that use a lot
1123 of stack the MDS (or possibly OSS) service threads could overflow
1124 the stack. One change which was identified to consume additional
1125 stack has been reworked to avoid the extra stack usage.
1128 Frequency : occasional
1130 Description: Do not fail import if osc_interpret_create gets -EAGAIN
1131 Details : If osc_interpret_create got -EAGAIN it immediately exits and
1132 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
1133 oscc_has_objects and see OSC has no objests and call
1134 oscc_internal_create to resend create request.
1136 Severity : enhancement
1138 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
1140 Severity : enhancement
1142 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
1145 Frequency : very rare
1147 Description: Fix replay if there is an un-replied request and open
1148 Details : In some cases, older replay request will revert the
1149 mcd->mcd_last_xid on MDS which is used to record the client's
1150 latest sent request.
1152 Severity : enhancement
1154 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
1156 Severity : enhancement
1158 Description: Add rhel5 support to HEAD.
1160 Serverity : enhancement
1162 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
1167 Description: Don't allow skipping OSTs if index has been specified.
1168 Details : Don't allow skipping OSTs if index has been specified, make
1169 locking in internal create lots better.
1173 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
1174 Details : access to bitfield in structure is always rounded to long
1175 and this produce problem with not atomic change any bit.
1179 Description: Lustre make rpms failed.
1180 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
1181 builds Lustre rpms from tarball.
1184 Frequency : rare on shutdown ost
1186 Description: If llog cancel was not send before clean_exports phase, this can
1187 produce deadlock in llog code.
1188 Details : If llog thread has last reference to obd and call class_import_put
1189 this produce deadlock because llog_cleanup_commit_master wait when
1190 last llog_commit_thread exited, but this never success because was
1191 called from llog_commit_thread.
1195 Description: allow userland application know is lost one of stripes.
1196 Details : fill lvb_blocks with error code on ost and return it to
1197 application if error flag found.
1201 Description: NULL lov_tgts causing MDS oops
1202 Details : more safe checks for NULL lov_tgts for avoid oops.
1204 Severity : enhancement
1206 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
1210 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
1211 Details : If an OST is down the MDS will hang indefinitely in
1212 obd_statfs_rqset() waiting for the statfs data. While for
1213 MDS QOS usage of statfs, it should not stuck in waiting.
1215 Severity : enhancement
1217 Description: remote_acl support
1218 Details : Support ACL-based permission check for remote user.
1219 Support setfacl/getfacl for remote user with the utils
1220 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
1221 the system "{s,g}etfacl" utils.
1223 Severity : enhancement
1225 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
1227 Severity : enhancement
1229 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
1233 Description: make mgs_setparam() handle fsname containing dash
1234 Details : fsname containing a dash does not work with lctl conf_param
1236 Severity : enhancement
1238 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
1240 Severity : enhancement
1242 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
1246 Frequency : rare, at shutdown
1247 Description: access already free / zero obd_namespace.
1248 Details : if client_disconnect_export was called without force flag set,
1249 and exist connect request in flight, this can produce access to
1250 NULL pointer (or already free pointer) when connect_interpret
1251 store ocd flags in obd_namespace.
1255 Frequency : only at startup
1256 Description: not alloc memory with spinlock held.
1257 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
1258 if any spinlock held.
1260 Severity : enhancement
1262 Description: make lustre randomly fail allocating memory
1263 Details : Make lustre randomly failed allocating memory for testing purpose.
1265 Severity : enhancement
1267 Description: lost problems with lov objid file
1268 Details : Fixes some scability and access to not inited memory problems
1269 in work with lov objdid file.
1274 Description: lfs find does not continue on file error
1275 Details : Continue other files processing when a file/dir is absent.
1279 Description: Inconsistent usage of lustre_pack_reply()
1280 Details : Standardize the usage of lustre_pack_reply() such that it
1281 always generate a CERROR on failure.
1284 Frequency : occasional
1286 Description: MDS hang or stay in waiting lock
1287 Details : If client receive lock with CBPENDING flag ldlm need send lock
1288 cancel as separate rpc, to avoid situation when cancel request
1289 can't processed due all i/o threads stay in wait lock.
1293 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1297 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1301 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1305 Description: Update to RHEL4 latest kernel.
1307 Severity : enhancement
1309 Description: Build SLES10 patchless client fails
1310 Details : The configure was broken by run ./configure with
1311 --with-linux-obj=.... argument for patchless client. When the
1312 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1313 can't search header adequately. Use absolute path such as
1314 -I($LINUX)/include instead.
1318 Description: interrupt oig_wait produce painc on resend.
1319 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1320 private set, and this produce situation when rq_ptlrpcd_data not
1321 copyed to new allocated request and triggered LBUG on assert
1322 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1323 ptlrpcd set if request is changed and can be safety changed to use
1328 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1329 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1330 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1331 these macros and replace them with their difinitions instead.
1335 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1336 Details : Update which_patch & target file for SLES10 latest kernel.
1339 Description: Remove client patches from SLES 10 kernel.
1340 Details : This causes SLES 10 clients to behave as patchless clients
1341 even on a Lustre-patched (server) kernel.
1343 Severity : enhancement
1345 Description: self-adjustable client's lru lists
1346 Details : use adaptive algorithm for managing client cached locks lru
1347 lists according to current server load, other client's work
1348 pattern, memory activities, etc. Both, server and client
1349 side namespaces provide number of proc tunables for controlling
1352 Severity : enhancement
1354 Description: light-weight GSS support
1355 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
1356 and reduce performance overhead.
1358 Severity : enhancement
1360 Description: Linux keyring support
1361 Details : Support using service of Linux keyring for Lustre GSS internal
1362 context refresh/cache mechanism.
1366 Description: Fix errors in lfs documentation
1367 Details : Fixes man pages
1371 Description: don't use GFP_* in generic Lustre code.
1372 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
1376 Description: obdclass is limited by single OBD_ALLOC(idarray)
1377 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
1381 Description: warnings with build patchless client with vanila 2.6.19 and up
1382 Details : change old ctl_table style and replace ctl_table/ctl_table_header
1383 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
1387 Frequency : Only for SLES
1388 Description: sanity_quota fail test_1
1389 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
1390 file, which makes TSTID[2] inunique.
1394 Frequency : Only for SLES9
1395 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
1396 Details : Update target/ChangeLog/which_patch .
1400 Frequency : Only for test_10 in sanity-quota.sh
1401 Description: a bug in quota test code
1402 Details : checking a test flag in a wrong place
1406 Frequency : Only for test_13 in sanity-quota.sh
1407 Description: a bug in quota test script
1408 Details : shouldn't check overfull allocated quota
1412 Description: jbd statistics
1413 Details : Port older jbd statistics patch for sles10
1415 Severity : enhancement
1417 Description: Add printing inode info into message about error in writepage.
1422 Description: reply_lock_interpret crash due to race with it and lock cancel.
1423 Details : Do not replay locks that are being cancelled. Do not reference
1424 locks by their address during replay, just by their handle.
1428 Frequency : When flocks are used.
1429 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
1430 Details : Flock locks might destroy just granted lock if it could be merged
1431 with another existing flock, this is done in completion handler,
1432 so teach ldlm_cli_enquque_fini that this is a valid case for
1438 Description: Oops in read and write path when failing to allocate lock.
1439 Details : Check if lock allocation failed and return error back.
1443 Description: lstripe command fails for valid OST index
1444 Details : The stripe offset is compared to lov->desc.ld_tgt_count
1445 instead of lov->desc.ld_active_tgt_count.
1449 Description: sanity.sh failed test 103
1450 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
1455 Description: df doesn't work properly if diskfs blocksize != 4K
1456 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
1460 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
1461 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
1463 Severity : enhancement
1465 Description: Adaptive timeouts
1466 Details : RPC timeouts adapt to changing server load and network
1467 conditions to reduce resend attempts and improve recovery time.
1471 Description: llapi_file_create() does not allow some changes
1472 Details : add llapi_file_open() that allows specifying the mode and
1473 open flags, and also returns an open file handle.
1477 Description: merge and cleanup kernel patches.
1478 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
1482 Description: Add journal checksum support.(Kernel part)
1483 Details : The journal checksum feature adds two new flags i.e
1484 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
1485 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
1486 indicates that the commit block contains the checksum for
1487 the blocks described by the descriptor blocks. Now commit
1488 record can be sent to disk without waiting for descriptor
1489 blocks to be written to disk. This behavior is controlled
1490 using JBD2_FEATURE_ASYNC_COMMIT flag.
1494 Description: OSS needs mutliple precreate threads
1495 Details : Add ability to start more than one create thread per OSS.
1499 Description: Sanity.sh test_65k failed: setstripe should have succeeded
1500 Details : execute lfs setstripe on client
1504 Description: mds_obd_create error creating tmp object
1505 Details : When the user sets quota on root, llog will be affected and can't
1506 create files and write files.
1509 Frequency : Always on ia64 patchless client, and possibly others.
1511 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
1512 Details : This allows the patchless client to be loaded on architectures
1513 without this export.
1517 Description: RedHat Update kernel for RHEL5
1518 Details : Add the kernel config file for RHEL5.
1522 Description: RedHat Update kernel for RHEL5
1523 Details : Modify the kernel config file more closer RHEL5.
1527 Description: Build failure against Centos5 (RHEL5)
1528 Details : Define PAGE_SIZE when it isn't present.
1532 Description: test_7 on sanity-quota.sh can't be used on separate machines
1533 Details : add facet to handle it
1537 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
1538 Details : with NFS, the anon dentry's parent was set to itself in
1539 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
1540 corresponding dentry other than use rec->ur_name.
1542 Severity : enhancement
1544 Description: lfs setstripe enhancement
1545 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
1549 Description: enable data checksumming by default
1550 Details : enable checksum by default, allow --disable-checksum
1551 configure option and "-o nochecksum" mount option. Checksums
1552 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
1556 Description: lustre support for RHEL5
1557 Details : Add support for RHEL5.
1561 Description: Client eviction due to failover config
1562 Details : after a connection loss, the lustre client should attempt to
1563 reconnect to the last active server first before trying the
1564 other potential connections.
1568 Description: when mds and osts use different quota unit(32bit and 64bit),
1569 quota will be released repeatly.
1570 Details : void sending multiple quota reqs to mds, which will keep the
1571 status between the reqs.
1575 Description: osts not allocated evenly to files
1576 Details : change the condition to increase offset_idx
1579 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
1581 Description: Data corruption for OSTs that are formatted larger than 2TB
1583 Details : When generating the bio request for lustre file writes the
1584 sector number would overflow a temporary variable before being
1585 used for the IO. The data reads correctly from Lustre (which
1586 will overflow in a similar manner) but other file data or
1587 filesystem metadata may be corrupted in some cases.
1591 Description: Tune the kernel for good SCSI performance.
1592 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
1593 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
1598 Description: rewrite ext2-derived code in obdclass/uuid.c
1599 Details : rewrite inherited code (uuid parsing code from ext2 utils)
1600 from scratch preserving functionality.
1604 Description: "lfs find -obd UUID" prints directories
1605 Details : "lfs find -obd UUID" will return all directory names instead
1606 of just file names. It is incorrect because the directories
1607 do not reside on the OSTs.
1611 Description: use i_size_read and i_size_write in 2.6 port
1612 Details : replace inode->i_size access with i_size_read/write()
1615 Frequency : when removing large files
1617 Description: scheduling issue during removal of large Lustre files
1618 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
1619 It causes scheduling issues when removing large files (17TB in the
1623 Frequency : only with liblustre clients on XT3
1625 Description: evictions taking too long
1626 Details : allow llrd to evict clients directly on OSTs
1629 Frequency : only on ppc
1631 Description: /proc/fs/lustre/devices broken on ppc
1632 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
1633 We should make sure the type of 'index' is loff_t and then cast
1634 explicitly as needed below. Do not assign an explicitly cast
1638 Frequency : only for rhel5
1640 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
1641 Details : Modify the target file & which_kernel.
1643 Severity : enhancement
1645 Description: omit set fsid for export NFS
1646 Details : fix set/restore device id for avoid EMFILE error and mark lustre
1647 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
1651 Frequency : Always, for kernels after 2.6.16
1652 Description: Fix warning idr_remove called for id=.. which is not allocated.
1653 Details : Last kernels save old s_dev before kill super and not allow
1654 to restore from callback - restore it before call kill_anon_super.
1658 Description: buffer overruns could theoretically occur
1659 Details : llapi_semantic_traverse() modifies the "path" argument by
1660 appending values to the end of the origin string, and a
1661 overrun may occur. Adding buffer overrun check in liblustreapi.
1665 Description: Fix error on 'ls .' at the top of the Lustre mount.
1666 Details : Don't revalidate dentry if it is a root dentry.
1670 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
1671 Details : Modify vm-tunables-rhel4.patch.
1675 Description: Kernel config for 2.6.18-vanilla.
1676 Details : Modify targets/2.6-vanilla.target.in.
1677 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
1678 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
1679 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
1680 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
1684 Description: improve handling recoverable errors
1685 Details : if request processig with error which can be recoverable on server
1686 request should be resend, otherwise page released from cache and
1691 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
1692 Details : Modify target file & which_patch.
1693 A flaw was found in the IA32 system call emulation provided
1694 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
1695 value could be stored in the %RAX register, which could trigger an
1696 out-of-bounds system call table access. An untrusted local user
1697 could exploit this flaw to run code in the kernel
1698 (ie a root privilege escalation). (CVE-2007-4573).
1702 Description: O_DIRECT bypasses client statistics.
1703 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
1704 read_ahead_stats were updated. copied the stats section from
1705 osc_send_oap_rpc() into async_internal().
1709 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
1713 Description: change order of libsysio includes
1714 Details : '#include sysio.h' should always come before '#include xtio.h'
1718 Description: handle "serious error: objid * already exists" more gracefully
1719 Details : If LAST_ID value on disk is smaller than the objects existing in
1720 the O/0/d* directories, it indicates disk corruption and causes an
1721 LBUG(). If the object is 0-length, then we should use the existing
1722 object. This will help to avoid a full fsck in most cases.
1724 Severity : enhancement
1726 Description: adapt the lustre_config script to support the upgrade case
1727 Details : Add "-u" option for lustre_config script to support upgrading 1.4
1728 server targets to 1.6 in parallel.
1732 Description: To avoid grant space > avaible space when the disk is almost
1733 full. Without this patch you might see the error "grant XXXX >
1734 available" or some LBUG about grant, when the disk is almost
1736 Details : In filter_check_grant, for non_grant cache write, we should
1737 check the left space by if (*left > ungranted + bytes), instead
1738 of (*left > ungranted), because only we are sure the left space
1739 is enough for another "bytes", then the ungrant space should be
1740 increase. In client, we should update cl_avail_grant only there
1741 is OBD_MD_FLGRANT in the reply.
1745 Description: Update RHEL 4 kernel to fix local root privilege escalation.
1746 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
1747 described in CVE-2007-4573. This problem could allow untrusted
1748 local users to gain root access.
1751 Frequency : when using O_DIRECT and quotas
1753 Description: Incorrect file ownership on O_DIRECT output files
1754 Details : block usage reported by 'lfs quota' does not take into account
1755 files that have been written with O_DIRECT.
1760 Description: touch file failed when fs is not full
1761 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
1762 otherwise we can get ENOSP while fs is not full.
1766 Description: parallel lock callbacks
1767 Details : Instead of sending blocking and completion callbacks as separated
1768 requests, adding them to a set and sending in parallel.
1771 Frequency : only for Cray XT3
1772 Bugzilla : 12829/13455
1773 Description: Changing primary group doesn't change the group lustre assigns to
1775 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
1776 overridden with the primary group provided by the group upcall,
1777 whereas the supplied fsgid can be trusted if it is in the list of
1778 supplementary groups returned by the group upcall.
1780 Severity : enhancement
1782 Description: Allow masking D_WARNING, D_ERROR messages from console
1783 Details : Console messages can now be disabled via lnet.printk.
1787 Description: User code with malformed file open parameter crashes client node
1788 Details : Before packing join_file req, all the related reference should be
1789 checked carefully in case some malformed flags cause fake
1790 join_file req on client.
1794 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
1795 Details : ldlm_completion_ast() assumes that a lock is granted when the req
1796 mode is equal to the granted mode. However, it should also check
1797 that LDLM_FL_CP_REQD is not set.
1801 Description: Heavy nfs access might result in deadlocks
1802 Details : After ELC code landed, it is now improper to enqueue any mds
1803 locks under och_sem, because enqueue might want to decide to
1804 cancel open locks for same inode we are holding och_sem for.
1808 Description: Client eviction while running blogbench
1809 Details : A lot of unlink operations with concurrent I/O can lead to a
1810 deadlock causing evictions. To address the problem, the number of
1811 oustanding OST_DESTROY requests is now throttled to
1812 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
1813 callbacks are processed in priority.
1817 Description: enable ACLs on MDS by default
1818 Details : ACLs must be enabled on MDS by default.
1821 Frequency : PPC/PPC64 only
1823 Description: conflicts between asm-ppc64/types.h and lustre_types.h
1824 Details : fix duplicated definitions between asm-ppc64/types.h and
1825 lustre_types.h on PPC.
1828 Frequency : PPC/PPC64 only
1830 Description: asm-ppc/segment.h does not exist
1831 Details : fix compile issue on PPC.
1835 Description: better handle error messages in extents code
1838 Frequency : RHEL4 only
1840 Description: mkfs is very slow on IA64/RHEL4
1841 Details : A performance regression has been discovered in the MPT Fusion
1842 driver between versions 3.02.73rh and 3.02.99.00rh. As a
1843 consequence, we have downgraded the MPT Fusion driver in the RHEL4
1844 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
1846 Severity : enhancement
1848 Description: SNMP support enhancement
1849 Details : Adding total number of sampled request for an MDS node in snmp
1852 Severity : enhancement
1854 Description: Optimize ldlm waiting list processing for PR extent locks
1855 Details : When processing waiting list for read extent lock and meeting read
1856 lock that is same or wider to it that is not contended, skip
1857 processing rest of the list and immediatelly return current
1858 status of conflictness, since we are guaranteed there are no
1859 conflicting locks in the rest of the list.
1863 Description: Time out and refuse to reconnect
1864 Details : When the failover node is the primary node, it is possible
1865 to have two identical connections in imp_conn_list. We must
1866 compare not conn's pointers but NIDs, otherwise we can defeat
1867 connection throttling.
1871 Description: port llog fixes from b1_6 into HEAD
1872 Details : Port llog reference couting and some llog cleanups from b1_6
1873 (bug 10800) into HEAD, for protect from panic and access to already
1874 free llog structures.
1878 Description: Detect stride IO mode in read-ahead
1879 Details : When a client does stride read, read-ahead should detect that and
1880 read-ahead pages according to the detected stride pattern.
1884 Description: data checksumming impacts single node performance
1885 Details : add support for several checksum algorithm. Currently, only CRC32
1886 and Adler-32 are supported. The checksum type can be changed on
1887 the fly via /proc/fs/lustre/osc/*/checksum_type.
1891 Description: use adler32 for page checksums
1892 Details : when available, use the Adler-32 algorithm instead of CRC32 for
1897 Description: build for x2 fails
1898 Details : fix compile issue on Cray systems.
1902 Description: Properly match for duplicate locks
1903 Details : Due to different lock order from skiplists code, we need to
1904 traverse entire list for now
1907 Frequency : only on PPC/SLES10
1909 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
1910 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
1911 be defined. Add a hack in mkfs_lustre.c to work around this compile
1916 Description: LASSERT on MDS when client holding flock lock dies
1917 Details : ldlm pool logic depends on number of granted locks equal to
1918 number of released locks which is not true for flock locks, so
1919 just exclude such locks from consideration.
1923 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
1924 Details : Use fsfilt_sync() for both the whole filesystem sync and
1925 individual file sync to eliminate dangerous inode locking
1926 with I_LOCK that can lead to a deadlock.
1930 Description: performance in 1.6.3
1931 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
1932 sure that 1MB requests are not fragmented by the block layer.
1934 Severity : enhancement
1936 Description: organize the server-side client stats on per-nid basis
1937 Details : Change the structure of stats under obdfilter and mds to
1945 The "uuid"s file would list the uuids of _active_ exports.
1946 And the clear entry is to clear all stats and stale nids.
1948 Severity : enhancement
1950 Description: eliminate client locks in face of contention
1951 Details : file contention detection and lockless i/o implementation
1952 for contended files.
1956 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
1959 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
1961 Description: MDS may lose file striping (and hence file data) in some cases
1962 Details : If there are additional extended attributes stored on the MDS,
1963 in particular ACLs, SELinux, or user attributes (if user_xattr
1964 is specified for the client mount options) then there is a risk
1965 of attribute loss. Additionally, the Lustre file striping
1966 needs to be larger than default (e.g. striped over all OSTs),
1967 and an additional attribute must be stored initially in the
1968 inode and then increase in size enough to be moved to the
1969 external attribute block (e.g. ACL growing in size) for file
1974 Description: skiplist implementation simplification
1975 Details : skiplists are used to group compatible locks on granted list
1976 that was implemented as tracking first and last lock of each
1977 lock group the patch changes that to using doubly linked lists
1981 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
1982 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
1983 because LASSERT on both the data supplied by a client, and the
1984 data on disk is dangerous and incorrect.
1986 Severity : enhancement
1988 Description: Slow truncate/writes to huge files at high offsets.
1989 Details : Directly associate cached pages to lock that protect those pages,
1990 this allows us to quickly find what pages to write and remove
1991 once lock callback is received.
1995 Description: more ldlm soft lockups
1996 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
1997 starve other threads from the resource lock for a long time in
1998 case of long waiting queue, so change the debug level from
1999 D_OTHER to the less frequently used D_INFO.
2001 Severity : enhancement
2003 Description: add -gid, -group, -uid, -user options to lfs find
2007 Description: Hung threads in invalidate_inode_pages2_range
2008 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
2009 once one is completed. As a result, some RPCs are stuck in the
2010 queue and are never sent.
2014 Description: filter threads hungs on waiting journal commit
2015 Details : Cleanup filter group llog code, then only filter group llog will
2016 be only created in the MDS/OST syncing process.
2020 Description: Procfs and llog threads access destoryed import sometimes.
2021 Details : Sync the import destoryed process with procfs and llog threads by
2022 the import refcount and semaphore.
2024 Severity : enhancement
2026 Description: openlock cache of b1_6 port to HEAD
2031 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
2032 Details : If insertion of an extent fails, then discard the inode
2033 preallocation and free data blocks else it can lead to duplicate
2038 Description: don't always update ctime in ext3_xattr_set_handle()
2039 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
2040 In some cases the ctime should not be updated, for example for
2041 2.0->1.8 compatibility it is necessary to delete an xattr and it
2042 should not update the ctime.
2046 Bugzilla : 15713/16362
2047 Description: Assertion in iopen_connect_dentry in 1.6.3
2048 Details : looking up an inode via iopen with the wrong generation number can
2049 populate the dcache with a disconneced dentry while the inode
2050 number is in the process of being reallocated. This causes an
2051 assertion failure in iopen since the inode's dentry list contains
2052 both a connected and disconnected dentry.
2056 Description: assertion failure in ldlm_handle2lock()
2057 Details : fix a race between class_handle_unhash() and class_handle2object()
2058 introduced in lustre 1.6.5 by bug 13622.
2063 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
2064 Details : remove the limit of 256 scsi disks in the sd_iostat patch
2069 Description: oops in sd_iostats_seq_show()
2070 Details : unloading/reloading the scsi low level driver triggers a kernel
2071 bug when trying to access the sd iostat file.
2076 Description: Kernel panics during QLogic driver reload
2077 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
2078 patch, causing memory corruption.
2083 Description: journal_dev option does not work in b1_6
2084 Details : pass mount option during pre-mount.
2086 Severity : enhancement
2088 Description: Add a FIEMAP(FIle Extent MAP) ioctl
2089 Details : FIEMAP ioctl will allow an application to efficiently fetch the
2090 extent information of a file. It can be used to map logical blocks
2091 in a file to physical blocks in the block device.
2095 Description: LDLM soft lockups - improvement
2096 Details : It is be possible to send the lock handle along with each read
2097 or write request because the client is already doing a lock match
2098 itself so there isn't any reason the OST should have to re-do that
2102 Frequency : only on Cray X2
2104 Description: X2 build failures
2105 Details : fix build failures on Cray X2.
2109 Description: xid & resent requests
2110 Details : Initialize RPC XID from clock at startup (randomly if clock is
2113 Severity : enhancement
2115 Description: Add lustre_start utility to start or stop multiple Lustre servers
2120 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
2121 Details : In case of memory pressure, list_del() can be called twice on
2122 req->rq_history_list, causing a kernel oops.
2126 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
2127 Details : fix stack overflow in the distributed lock manager by defering
2128 export eviction after a failed AST to the elt thread instead of
2129 handling it in the dlm interpret routine.
2133 Description: Convert some comments to new format.
2134 Details : Update documenting comments to match doxygen conventions.
2138 Description: Grammar fixes.
2139 Details : A couple of trivial spelling fixes.
2143 Description: OSD_COUNTERS-mandatory
2144 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
2145 to simplify the code.
2149 Description: simplify cmm_device freeing logic.
2150 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
2154 Description: Add lockdep support to dt_object_operations locking interface.
2155 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
2156 parameter indicating lock ordering. Update mdd code to use new
2161 Description: Introduce failloc constants for lockless IO tests.
2162 Details : Add two new failloc constants to test lockless IO. Only one of
2163 them in implemented---another is checked in yet to be landed
2168 Description: Add lockdep support for inode mutex.
2169 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
2170 in the situations where Lustre has to lock more than one inode
2175 Description: Add optional invariants checking support.
2176 Details : Add new LINVRNT() macro, optional on new --enable-invariants
2177 configure switch. This macro is to be used for consistency and
2178 sanity checks that are too expensive to be left in `production'
2183 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
2184 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
2189 Description: Add lockdep support for ldlm_lock and ldlm_resource.
2190 Details : Use spin_lock_nested() in (the only) situation where more than
2191 one ldlm_lock is locked simultaneously. Also, fix possible
2192 dead-lock in ldlm_lock_change_resource() by enforcing particular
2197 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
2198 Details : Instead of specifying each ldlm_lock call-back through separate
2199 parameter, wrap them into struct ldlm_callback_suite.
2203 Description: Kill join_lru obd method and its callers.
2204 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
2205 in memory---a requirement implemented through ->o_join_lru() obd
2206 method in HEAD. Remove this method and its users.
2210 Description: Add asynchronous ldlm ENQUEUE completion handler.
2211 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
2212 case that stock ldlm_completion_ast() cannot handle as it waits
2213 until lock is granted. Introduce new ldlm_completion_ast_async()
2214 for this. Also comment ldlm_completion_ast().
2218 Description: ldlm_error <-> errno conversion.
2219 Details : Add functions to map (rather arbitrary) between LDLM error codes
2220 and standard errno values. CLIO needs this to prevent LDLM specific
2221 constants from escaping ldlm and osc.
2225 Description: Kill unused ldlm_handle2lock_ns() function.
2226 Details : Kill unused ldlm_handle2lock_ns() function.
2230 Description: Add lu_ref support to ldlm_lock
2231 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
2232 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
2233 and ldlm_resource. LDLM interface has to be changed, because code that
2234 releases a reference on a lock, has to "know" what reference this is.
2235 In the most frequent case
2237 lock = ldlm_handle2lock(handle);
2239 LDLM_LOCK_PUT(lock);
2241 no changes are required. When any other reference (received _not_ from
2242 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
2243 instead of LDLM_LOCK_PUT().
2245 Arguably, changes are pervasive, and interface requires some discipline
2246 for proper use. On the other hand, it was very instrumental in finding
2247 a few leaked lock references.
2251 Description: Add ldlm_lock_addref_try().
2252 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
2253 attempts to addref a lock that might be being canceled
2258 Description: Add ldlm_weigh_callback().
2259 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
2260 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
2261 hard-coded `number of pages' logic.
2265 Description: Add lockdep annotations to llog code.
2266 Details : Use appropriately tagged _nested() locking calls in the places
2267 where llog takes more than one ->lgh_lock lock.
2271 Description: Add loi_kms_set().
2272 Details : Wrap kms updates into a helper function.
2276 Description: Constify instances of struct lsm_operations.
2277 Details : Constify instances of struct lsm_operations.
2281 Description: lu_conf support.
2282 Details : On a server, a file system object is uniquely identified
2283 by a fid, which is sufficient to locate and load all object
2284 state (inode). On a client, on the other hand, more data are
2285 necessary instantiate an object. Change lu_object_find() and
2286 friends to take additional `lu_conf' argument describing object.
2287 Typically this includes layout information.
2291 Description: lu_context fixes.
2292 Details : Introduce new lu_context functions that are needed on the client
2293 side, where some system threads (ptlrpcd) are shared by multiple
2294 modules, and so cannot be stopped during module shutdown.
2298 Description: Add start and stop methods to lu_device_type_operations.
2299 Details : Introduce two new methods in lu_device_type_operations, that are
2300 invoked when first instance of a given type is created and last one
2301 is destroyed respectively. This is need by CLIO.
2305 Description: Add lu_ref support to struct lu_device.
2306 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
2307 track leaked references.
2311 Description: Introduce lu_kmem_descr.
2312 Details : lu_kmem_descr and its companion interface allow to create
2313 and destroy a number of kmem caches at once.
2317 Description: Fix lu_object finalization race.
2318 Details : Fix a race between lu_object_find() finding an object and its
2319 concurrent finalization. This race is (most likely) not possible
2320 on the server, but might happen on the client.
2324 Description: Introduce lu_ref interface.
2325 Details : lu_ref is a debugging module allowing to track references to
2326 a given object. It is quite cpu expensive, and has to be
2327 explicitly enabled with --enable-lu_ref. See usage description
2332 Description: Factor lu_site procfs stats into a separate function.
2333 Details : Separate lu_site stats printing code into a separate function
2334 that can be reused on a client.
2338 Description: Constify instances of struct {lu,dt,md}_device_operations.
2339 Details : Constify instances of struct {lu,dt,md}_device_operations.
2343 Description: Introduce struct md_site and move meta-data specific parts of
2344 struct lu_site here.
2345 Details : Move md-specific fields out of struct lu_site into special struct
2346 md_site, so that lu_site can be used on a client.
2350 Description: Kill mdd_lov_destroy().
2351 Details : Remove unused mdd code.
2355 Description: Add st_block checking to multistat.c.
2356 Details : Add st_block checking to multistat.c.
2360 Description: Add lu_ref support to struct obd_device.
2361 Details : Add lu_ref tracking to obd_device.
2365 Description: Kill obd_set_fail_loc().
2366 Details : Remove unused code.
2370 Description: Add special type for ptlrpc_request interpret functions.
2371 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
2372 there. Actual usage will be in CLIO.
2376 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
2377 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
2378 doesn't work with lockdep.
2382 Description: Add rwv.c test program.
2383 Details : New testing program exercising readv(2) and writev(2) (Qian).
2387 Description: Add sendfile.c test program.
2388 Details : New testing program exercising sendfile(2) (Jay).
2392 Description: Ratelimit a message that can be very frequent.
2393 Details : Ratelimit a memory allocation failure message that can
2398 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2399 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2403 Description: Liblustre build fixes.
2404 Details : Liblustre build fixes.
2408 Description: libcfs: add cfs_{need,cond}_resched() interface.
2409 Details : libcfs: add cfs_{need,cond}_resched() definition and
2410 implementations for Linux, NT, and liblustre.
2412 Severity : enhancement
2414 Description: More exported tunables for mballoc
2415 Details : Add support for tunable preallocation window and new tunables for
2416 large/small requests
2420 Description: Detect corruption of block bitmap and checking for preallocations
2421 Details : Checks validity of on-disk block bitmap. Also it does better
2422 checking of number of applied preallocations. When corruption is
2423 found, it turns filesystem readonly to prevent further corruptions.
2427 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
2428 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
2429 window to 0. If it does hit cache miss, and read-pattern is still
2430 stride-io mode, does not reset the stride window, but also does not
2431 increase the stride window length in this case.
2435 Frequency : only for big-endian servers
2436 Description: Check if system is big-endian while mounting fs with extents feature
2437 Details : Mounting a filesystem with extents feature will fail on big-endian
2438 systems since ext3-based ldiskfs is not supported on big-endian
2439 systems. This can be over-riden with "bigendian_extents" mount option.
2441 Severity : enhancement
2443 Description: The root squash functionality
2444 Details : A security feature, which is to prevent users from being able
2445 to mount lustre on their desktop, run as root, and delete
2446 all of the files in the filesystem. The goal is accomplished by
2447 remapping user id (UID) and group id (GID) of the root user to
2448 a UID and GID specified by the system administartor via Lustre
2449 configuration management server (MGS). The functionality also
2450 allows to specify sets of clients for which the remapping does
2455 Description: Excessive recovery window
2456 Details : With AT enabled, the recovery window can be excessively long (6000+
2457 seconds). To address this problem, we no longer use
2458 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
2459 timeout no longer depends on the service time, it is set to
2460 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
2461 time via pb_service_time.
2465 Description: Watchdog triggered on MDS failover
2466 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
2467 the OSTs know that the MDS "UUID" can be reused for the same export
2468 from a different NID, so we do not need to wait for the export to be
2472 Frequency : rare, only if using MMP with Linux RAID
2474 Description: MMP doesn't work with Linux RAID
2475 Details : While using HA for Lustre servers with Linux RAID, it is possible
2476 that MMP will not detect multiple mounts. To make this work we
2477 need to unplug the device queue in RAID when the MMP block is being
2478 written. Also while reading the MMP block, we should read it from
2479 disk and not the cached one.
2481 Severity : enhancement
2483 Description: open file using fid
2484 Details : A file can be opened using just its fid, like
2485 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
2488 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
2491 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
2492 Details : add proper locking for creation of jbd_4k slab cache
2496 Description: MMP check in ext3_remount() fails without displaying any error
2497 Details : When multiple mount protection fails during remount, proper error
2500 Severity : enhancement
2502 Description: Allow stripe size to be up to 4G-64k
2503 Details : Fix math logic to allow large stripe sizes.
2507 Description: add check for >8TB ldiskfs filesystems
2508 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
2509 Don't allow >8TB ldiskfs filesystems to be mounted without
2510 force_over_8tb mount option
2515 Descriptoin: OSS complains "too many blocks" and crashes
2516 Details : In filter_commitrw_write():
2517 1. lquota_chkquota();
2518 2. obdo_from_inode();
2519 3. lquota_pending_commit();
2520 "2." has changed uid and gid in this case, then caused
2521 lquota_chkquota() and lquota_pending_commit() to use
2522 different uid and gid. That is the root of the bug.
2526 Descriptoin: getstripe doesn't report stripe count = -1 sanely
2527 Details : Detect and print out the 'stripe count = -1' case properly.
2531 Description: Out or order replies might be lost on replay
2532 Details : In ptlrpc_retain_replayable_request if we cannot find retained
2533 request with tid smaller then one currently being added, add it
2534 to the start, not end of the list.
2536 --------------------------------------------------------------------------------
2538 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
2540 * Support for kernels:
2541 2.6.5-7.283 (SLES 9),
2542 2.6.9-55.EL (RHEL 4),
2543 2.6.16.46-0.14 (SLES 10),
2544 2.6.18.8 vanilla (kernel.org)
2545 * Client support for unpatched kernels:
2546 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2547 2.6.16 - 2.6.22 vanilla (kernel.org)
2548 * Due to recently discovered recovery problems, we do not recommend
2549 using patchless RHEL 4 clients with this or any earlier release.
2550 * Recommended e2fsprogs version: 1.39.cfs8
2551 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2552 * Starting with this release, the ldiskfs backing filesystem required
2553 by Lustre is now in its own package, lustre-ldiskfs. This package
2554 should be installed. It is versioned separately from Lustre and
2555 may be released separately in future.
2559 Description: block reactivating mgc import until all deactivates complete
2560 Details : Fix race when failing back MDT/MGS to itself (testing)
2563 Frequency : at statup only
2565 Description: mds_lov_synchronize race leads to various problems
2566 Details : simultaneous MDT->OST connections at startup can cause the
2567 sync to abort, leaving the OSC in a bad state.
2569 Severity : enhancement
2571 Description: add optional extra BUILD_VERSION info
2572 Details : add a new environment variable (namely LUSTRE_VERS) which allows
2573 to override the lustre version.
2576 Frequency : 2.6.18 servers only
2578 Description: ll_kern_mount() doesn't release the module reference
2579 Details : The ldiskfs module reference count never drops down to 0
2580 because ll_kern_mount() doesn't release the module reference.
2585 Description: server LBUG when using old ost_num_threads parameter
2586 Details : Accept the old ost_num_threads parameter but warn that it
2587 is deprecated, and fix an off-by-one error that caused an LBUG.
2592 Description: Transient SCSI error results in persistent IO issue
2593 Details : iobuf->dr_error is not reinitialized to 0 between two
2597 Frequency : sometimes when underlying device returns I/O errors
2599 Description: OSTs not going read-only during write failures
2600 Details : OSTs are not remounted read-only when the journal commit threads
2601 get I/O errors because fsfilt_ext3 calls journal_start/stop()
2602 instead of the ext3 wrappers.
2606 Description: poor connect scaling with increasing client count
2607 Details : Don't run filter_grant_sanity_check for more than 100 exports
2608 to improve scaling for large numbers of clients.
2611 Frequency : SLES10 only
2613 Description: sanity-quota.sh quotacheck failed: rc = -22
2614 Details : Quotas cannot be enabled on SLES10.
2617 Frequency : liblustre clients only
2619 Description: getdirentries does not give error when run on compute nodes
2620 Details : getdirentries does not fail when the size specified as an argument
2621 is too small to contain at least one entry
2623 Severity : enhancement
2625 Description: Add LNET router traceability for debug purposes
2626 Details : If a checksum failure occurs with a router as part of the
2627 IO path, the NID of the last router that forwarded the bulk data
2628 is printed so it can be identified.
2633 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
2634 Details : Due to a race condition, liblustre clients were occasionally
2635 evicted incorrectly.
2637 Severity : enhancement
2639 Description: lfs setstripe use optional parameters instead of postional
2642 Severity : enhancement
2644 Description: Nanosecond timestamp support for ldiskfs
2645 Details : The on-disk ldiskfs filesystem has added support for nanosecond
2646 resolution timestamps. There is not yet support for this at
2647 the Lustre filesystem level.
2650 Frequency : during server recovery
2652 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
2653 Details : request with rq_no_resend flag not awake l_wait_event if they get
2657 Frequency : nfs export on patchless client
2659 Description: connectathon hang when test nfs export over patchless client
2660 Details : Disconnected dentry cannot be found with lookup, so we do not need
2661 to unhash it or make it invalid
2664 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
2665 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
2666 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
2671 Description: open req refcounting wrong on reconnect
2672 Details : If reconnect happened between getting open reply from server and
2673 call to mdc_set_replay_data in ll_file_open, we will schedule
2674 replay for unreferenced request that we are about to free.
2675 Subsequent close will crash in variety of ways.
2676 Check that request is still eligible for replay in
2677 mdc_set_replay_data().
2682 Description: disable writes to filesystem when reading health_check file
2683 Details : the default for reading the health_check proc file has changed
2684 to NOT do a journal transaction and write to disk, because this
2685 can cause reads of the /proc file to hang and block HA state
2686 checking on a healthy but otherwise heavily loaded system. It
2687 is possible to return to the previous behaviour during configure
2688 with --enable-health-write.
2690 Severity : enhancement
2692 Description: 64-bit inode version
2693 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
2694 to the inode. This will be required for version-based recovery.
2699 Description: MDS fails to start if a duplicate client export is detected
2700 Details : in some rare cases it was possible for a client to connect to
2701 an MDS multiple times. Upon recovery the MDS would detect this
2702 and fail during startup. Handle this more gracefully.
2704 Severity : enhancement
2706 Description: Add -o localflock option to simulate old noflock
2708 Details : This will achieve local-only flock/fcntl locks
2714 Description: log_commit_thread vs filter_destroy race leads to crash
2715 Details : Take import reference before releasing llog record semaphore
2720 Description: Wrong request locking in request set processing
2721 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
2722 imp_delayed_list, in this place should be used imp_lock.
2725 Frequency : when reconnection
2727 Description: Grant Leak when osc reconnect to OST
2728 Details : When osc reconnect ost, OST(filter) should check whether it
2729 should grant more space to client by comparing fed_grant and
2730 cl_avail_grant, and return the granted space to client instead
2731 of "new granted" space, because client will call osc_init_grant
2732 to update the client grant space info.
2735 Frequency : when client reconnect to OST
2737 Description: Grant Leak when osc do resend and replay bulk write
2738 Details : When osc reconnect to OST, OST(filter)should clear grant info of
2739 bulk write request, because the grant info while be sync between
2740 OSC and OST when reconnect, and we should ignore the grant info
2741 these of resend/replay write req.
2746 Description: Grant space more than avaiable left space sometimes.
2747 Details : When then OST is about to be full, if two bulk writing from
2748 different clients came to OST. Accord the avaliable space of the
2749 OST, the first req should be permitted, and the second one
2750 should be denied by ENOSPC. But if the seconde arrived before
2751 the first one is commited. The OST might wrongly permit second
2752 writing, which will cause grant space > avaiable space.
2755 Frequency : when client is evicted
2757 Description: Grant might be wrongly erased when osc is evicted by OST
2758 Details : when the import is evicted by server, it will fork another
2759 thread ptlrpc_invalidate_import_thread to invalidate the
2760 import, where the grant will be set to 0. While the original
2761 thread will update the grant it got when connecting. So if
2762 the former happened latter, the grant will be wrongly errased
2763 because of this race.
2768 Description: Checking Stale with correct fid
2769 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
2770 to check whether it is stale, because sometimes, we want the
2771 enqueue happened anyway, and op_data.fid2 will not be initialized.
2773 Severity : enhancement
2775 Description: update patchless client
2776 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
2779 Frequency : only with 2.4 kernel
2781 Description: random memory corruption
2782 Details : size of struct ll_inode_info is to big for union inode.u and this
2783 can be cause of random memory corruption.
2788 Description: Memory leak in recovery
2789 Details : Lov_mds_md was not free in an error handler in mds_create_object.
2790 It should also check obd_fail before fsfilt_start, otherwise if
2791 fsfilt_start return -EROFS,(failover mds during mds recovery).
2792 then the req will return with repmsg->transno = 0 and rc = EROFS.
2793 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
2794 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
2795 be freed no matter whether fsfilt_commit success or not.
2798 Frequency : only with huge count clients
2800 Description: Prevents from taking the superblock lock in llap_from_page for
2802 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
2803 for llap_from_page call in ll_removepage prevents from taking the
2804 superblock lock for a soon died page.
2809 Description: Not check open intent error before release open handle
2810 Details : in some rare cases, the open intent error is not checked before
2811 release open handle, which may cause
2812 ASSERTION(open_req->rq_transno != 0), because it tries to release
2813 the failed open handle.
2818 Description: Set cat log bitmap only after create log success.
2819 Details : in some rare cases, the cat log bitmap is set too early. and it
2820 should be set only after create log success.
2824 Description: Accessing a block bevice can re-enable I/O when Lustre is
2825 tearing down a device.
2826 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
2831 Description: service threads may hog cpus when there are a lot of requests
2832 Details : Insert cond_resched to give other threads a chance to use some CPU
2837 Description: the cat log was not initialized in recovery
2838 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
2839 unlink log on mds will not be initialized until mds post
2840 recovery. And also in mds post recovery, the unlink log will
2841 initialization will be done asynchronausly, so there will be race
2842 between add unlink log and unlink log initialization.
2846 Description: brw_stats were being printed incorrectly
2847 Details : brw_stats were being printed as log2 but all of them were not
2848 recorded as log2. Also remove some code duplication arising from
2849 filter_tally_{read,write}.
2853 Frequency : rare, only in recovery.
2854 Description: ASSERTION(req->rq_type != LI_POISON) failed
2855 Details : imp_lock should be held while iterating over imp_sending_list for
2856 prevent destroy request after get timeout in ptlrpc_queue_wait.
2860 Description: replay-single.sh test 52 fails
2861 Details : A lock's skiplist need to be cleanup when it being unlinked
2862 from its resource list.
2866 Description: Short directio read returns full requested size rather than
2868 Details : Direct I/O operations should return actual amount of bytes
2869 transferred rather than requested size.
2871 Severity : enhancement
2873 Description: metadata RPC reduction (e.g. for rm performance)
2874 Details : decrease the amount of synchronous RPC between clients and servers
2875 by canceling conflicing lock before the operation on the client
2876 and packing thier handles into the main operation RPC to server.
2878 Severity : enhancement
2880 Description: Async OSC create to avoid the blocking unnecessarily.
2881 Details : If a OST has no remain object, system will block on the creating
2882 when need to create a new object on this OST. Now, ways use
2883 pre-created objects when available, instead of blocking on an
2884 empty osc while others are not empty. If we must block, we block
2885 for the shortest possible period of time.
2889 Description: improve handling recoverable errors
2890 Details : if request processig with error which can be recoverable on server
2891 request should be resend, otherwise page released from cache and
2894 Severity : enhancement
2896 Description: refine locking for avoid write wrong info into lov_objid file
2897 Details : fix possible races with add new target and write/update data in
2900 --------------------------------------------------------------------------------
2902 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
2907 Frequency : on some architectures
2909 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
2910 Details : Uninitialized flags sometimes cause configuration commands to
2914 Frequency : patchless clients only
2916 Description: missing __iget() symbol export
2917 Details : The __iget() symbol export is missing. To avoid the need for
2918 this on patchless clients the deathrow inode reaper is turned
2919 off, and we depend on the VM to clean up old inodes. This
2920 dependency was during via the fix for bug 12181.
2922 --------------------------------------------------------------------------------
2924 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
2926 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
2927 INTEROPERATE with older versions automatically. In many cases a
2928 special upgrade step is needed. Please read the
2929 user documentation before upgrading any part of a live system.
2930 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
2931 version will not interoperate with 1.6 betas before beta5 (1.5.95).
2932 * WARNING: Lustre configuration and startup changes are required with
2933 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
2937 Severity : enhancement
2939 Description: MountConf
2940 Details : Lustre configuration is now managed via mkfs and mount
2941 commands instead of lmc and lconf. New obd types (MGS, MGC)
2942 are added for dynamic configuration management. See
2943 https://mail.clusterfs.com/wikis/lustre/MountConf for
2946 Severity : enhancement
2948 Description: dynamic OST addition
2949 Details : OSTs can now be added to a live filesystem
2951 Severity : enhancement
2953 Description: startup order invariance
2954 Details : MDTs and OSTs can be started in any order. Clients only
2955 require the MDT to complete startup.
2957 Severity : enhancement
2959 Description: parallel, asynchronous orphan cleanup
2960 Details : orphan cleanup is now performed in separate threads for each
2961 OST, allowing parallel non-blocking operation.
2963 Severity : enhancement
2965 Description: optimized stripe assignment
2966 Details : stripe assignments are now made based on ost space available,
2967 ost previous usage, and OSS previous usage, in order to try
2968 to optimize storage space and networking resources.
2970 Severity : enhancement
2972 Description: Permanently set tunables
2973 Details : All writable /proc/fs/lustre tunables can now be permanently
2974 set on a per-server basis, at mkfs time or on a live system.
2976 Severity : enhancement
2978 Description: Lustre message v2
2979 Details : Add lustre message format v2.
2981 Severity : enhancement
2983 Description: client OST exclusion list
2984 Details : Clients can be started with a list of OSTs that should be
2985 declared "inactive" for known non-responsive OSTs.
2989 Description: SPEC SFS validation failure on NFS v2 over lustre.
2990 Details : Changes the blocksize for regular files to be 2x RPC size,
2991 and not depend on stripe size.
2993 Severity : enhancement
2995 Description: Multiple MD RPCs in flight.
2996 Details : Further unserialise some read-only MDS RPCs - learn about intents.
2997 To avoid overly-overloading MDS, introduce a limit on number of
2998 MDS RPCs in flight for a single client and add /proc controls
2999 to adjust this limit.
3001 Severity : enhancement
3003 Description: client read/write statistics
3004 Details : Add client read/write call usage stats for performance
3005 analysis of user processes.
3006 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
3007 file access. extents_stats shows chunk size distribution.
3008 extents_stats_per_process show chunk size distribution per
3011 Severity : enhancement
3013 Description: mds statistics
3014 Details : Add detailed mds operations statistics in
3015 /proc/fs/lustre/mds/*/stats.
3019 Description: Failure of copying files with lustre special EAs.
3020 Details : Client side always return success for setxattr call for lustre
3021 special xattr (currently only "trusted.lov").
3026 Description: Refcount LNET uuids
3027 Details : The global LNET uuid list grew linearly with every startup;
3028 refcount repeated list entries instead of always adding to
3031 Severity : enhancement
3033 Description: Dynamic service threads
3034 Details : Within a small range, start extra service threads
3035 automatically when the request queue builds up.
3038 Frequency : mixed-endian client/server environments
3040 Description: mixed-endian crashes
3041 Details : The new msg_v2 system had some failures in mixed-endian
3044 Severity : enhancement
3046 Description: Easy OST removal
3047 Details : OSTs can be permanently deactivated with e.g. 'lctl
3048 conf_param lustre-OST0001.osc.active=0'
3050 Severity : enhancement
3052 Description: MGS proc entries
3053 Details : Added basic proc entries for the MGS showing what filesystems
3056 Severity : enhancement
3058 Description: provide MGS failover
3059 Details : Added config lock reacquisition after MGS server failover.
3061 Severity : enhancement
3063 Description: add Linux 2.4 support
3064 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
3068 Description: a large application tries to do I/O to the same resource and dies
3069 in the middle of it.
3070 Details : Check the req->rq_arrival time after the call to
3071 ost_brw_lock_get(), but before we do anything about
3072 processing it & sending the BULK transfer request. This
3073 should help move old stale pending locks off the queue as
3074 quickly as obd_timeout.
3077 Frequency : when an incorrect nid is specified during startup
3079 Description: ptlrpc connect to non-existant node causes kernel crash
3080 Details : LNET can't be re-entered from an event callback, which
3081 happened when we expire a message after the export has been
3082 cleaned up. Instead, hand the zombie cleanup off to another
3085 Severity : enhancement
3087 Description: plain/inodebits lock performance improvement
3088 Details : Grouping plain/inodebits in granted list by their request modes
3089 and bits policy, thus improving the performance of search through
3093 Frequency : only if OST filesystem is corrupted
3095 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3096 Details : for a short time RPCs with bulk IO are in the replay list,
3097 but replay of bulk IOs is unimplemented. If the OST filesystem
3098 is corrupted due to disk cache incoherency and then replay is
3099 started it is possible to trip an assertion. Avoid putting
3100 committed RPCs into the replay list at all to avoid this issue.
3103 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3105 Description: System hang on startup
3106 Details : This bug allowed the liblustre (e.g. catamount) client to
3107 return to the app before handling all startup RPCs. This
3108 could leave the node unresponsive to lustre network traffic
3109 and manifested as a server ptllnd timeout.
3111 Severity : enhancement
3113 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3114 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
3115 boolean to control whether to print peer debug info when a
3116 client's RPC times out.
3119 Frequency : only for kernels with patches from Lustre below 1.4.3
3121 Description: Remove old rdonly API
3122 Details : Remove old rdonly API which unused from at least lustre 1.4.3
3125 Frequency : only for devices with external journals
3127 Description: Set external device read-only also
3128 Details : During a commanded failover stop, we set the disk device
3129 read-only while the server shuts down. We now also set any
3130 external journal device read-only at the same time.
3133 Frequency : when upgrading from 1.4 while trying to change parameters
3135 Description: The wrong (new) MDC name was used when setting parameters for
3136 upgraded MDT's. Also allows changing of OSC (and MDC)
3137 parameters if --writeconf is specified at tunefs upgrade time.
3140 Frequency : when setting specific ost indicies
3142 Description: QOS code breaks on skipped indicies
3143 Details : Add checks for missing OST indicies in the QOS code, so OSTs
3144 created with --index need not be sequential.
3146 Severity : enhancement
3148 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3149 Details : The uninit_groups feature works in conjunction with the kernel
3150 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3151 up the pass1 processing of e2fsck. This is a read-only feature
3152 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3153 mount filesystems that have had this feature enabled.
3155 Severity : enhancement
3157 Description: Improve multi-block allocation algorithm to avoid fragmentation
3158 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3159 allocation locality and avoid filesystem fragmentation.
3161 ------------------------------------------------------------------------------
3163 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
3165 * Support for kernels:
3166 2.4.21-47.0.1.EL (RHEL 3)
3167 2.6.5-7.283 (SLES 9)
3168 2.6.9-42.0.10.EL (RHEL 4)
3169 2.6.12.6 vanilla (kernel.org)
3170 2.6.16.27-0.9 (SLES 10)
3171 * Recommended e2fsprogs version: 1.39.cfs5
3173 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3177 Frequency : occasional, depends on client load and configuration
3178 Bugzilla : 12181, 12203
3179 Description: data loss for recently-modified files
3181 Details : In some cases it is possible that recently written or created
3182 files may not be written to disk in a timely manner (this should
3183 normally be within 30s unless client IO load is very high).
3184 The problem appears as zero-length files or files that are a
3185 multiple of 1MB in size after a client crash or client eviction
3186 that are missing data at the end of the file.
3188 This problem is more likely to be hit on clients where files are
3189 repeatedly created and unlinked in the same directory, clients
3190 have a large amount of RAM, have many CPUs, the filesystem has
3191 many OSTs, the clients are rebooted frequently, and/or the files
3192 are not accessed by other nodes after being written.
3194 The presence of the problem can be detected by looking at
3195 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
3196 smaller than the second (nr_unused) then dirty files will not
3197 be flushed automatically to disk. "sync; sleep 10" should be
3198 run several times on the node before unmounting it to update
3199 Lustre (this is also safe to run on nodes without this problem).
3201 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
3202 that can cause this same problem, so customers using that kernel
3203 also need to update the kernel in addition to Lustre. In order
3204 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
3206 It is normal that files written just before a client crash (less
3207 than 30s) may not yet have been flushed to disk, even for local
3211 Frequency : frequent on thin XT3 nodes
3213 Description: UUID collision on thin XT3 Linux nodes
3214 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
3215 randomly, since we relied on an insufficiently-seeded PRNG.
3220 Description: OSS hangs after "All ost request buffers busy"
3221 Details : A deadlock between quota and journal operations caused OSS
3222 hangs after printing "All ost request buffers busy."
3225 Frequency : always on liblustre builds
3227 Description: Cleanup compiler warnings on liblustre
3230 Frequency : always on liblustre builds on XT3
3232 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
3236 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
3237 > 32000 subdirectories
3238 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
3239 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
3240 subdirectory count crosses 32000. This will aid e2fsck to
3241 correctly handle more than 32000 subdirectories.
3244 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3246 Description: System hang on startup
3247 Details : This bug allowed the liblustre (e.g. catamount) client to
3248 return to the app before handling all startup RPCs. This
3249 could leave the node unresponsive to lustre network traffic
3250 and manifested as a server ptllnd timeout.
3252 Severity : enhancement
3254 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3255 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
3256 boolean to control whether to print peer debug info when a
3257 client's RPC times out.
3262 Description: make O_SYNC working on 2.6 kernels
3263 Details : 2.6 kernels use different method for mark pages for write,
3264 so need add a code to lustre for O_SYNC work.
3269 Description: Failure to close file and release space on NFS
3270 Details : Put inode details into lock acquired in ll_intent_file_open.
3271 Use mdc_intent_lock in ll_intent_open to properly
3272 detect all kind of errors unhandled by mdc_enqueue.
3277 Description: proc file read during shutdown sometimes raced obd removal,
3279 Details : Add lock to prevent obd access after proc file removal.
3282 Frequency : Only for files larger than 4GB on 32-bit clients.
3284 Description: improperly doing page alignment of locks
3285 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
3286 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3291 Description: RPCs being resent when they shouldn't be
3292 Details : Some RPCs that should not be resent are being resent. This
3293 can cause inconsistencies in the RPC state machine. Do not
3294 resend such requests.
3297 Frequency : rare, only with NFS export
3299 Description: Crash on NFS re-export node
3300 Details : under very unusual load conditions an assertion is hit in
3301 ll_intent_file_open()
3304 Frequency : only if OST filesystem is corrupted
3306 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3307 Details : for a short time RPCs with bulk IO are in the replay list,
3308 but replay of bulk IOs is unimplemented. If the OST filesystem
3309 is corrupted due to disk cache incoherency and then replay is
3310 started it is possible to trip an assertion. Avoid putting
3311 committed RPCs into the replay list at all to avoid this issue.
3316 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3317 Details : Large single O_DIRECT read and write calls can fail to allocate
3318 a sufficiently large buffer to process the request. In case of
3319 allocation failure the allocation is retried with a smaller
3320 buffer and broken into smaller requests.
3322 Severity : enhancement
3324 Description: Add -o localflock option to simulate old noflock behaviour.
3325 Details : This will achieve local-only flock/fcntl locks coherentness.
3330 Description: versioning check is incomplete
3331 Details : Checking the version difference of client vs. server, report
3332 error if the gap is too big.
3337 Description: add support PG_writeback bit
3338 Details : add support for PG_writeback bit for Lustre, for more carefull
3339 work with page cache in 2.6 kernel. This also fix some deadlocks
3340 and remove hack for work O_SYNC with 2.6 kernel.
3342 Severity : enhancement
3344 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3345 Details : The uninit_groups feature works in conjunction with the kernel
3346 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3347 up the pass1 processing of e2fsck. This is a read-only feature
3348 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3349 mount filesystems that have had this feature enabled.
3351 Severity : enhancement
3353 Description: Improve multi-block allocation algorithm to avoid fragmentation
3354 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3355 allocation locality and avoid filesystem fragmentation.
3357 ------------------------------------------------------------------------------
3359 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3361 * Support for kernels:
3362 2.6.9-42.0.3.EL (RHEL 4)
3363 2.6.5-7.276 (SLES 9)
3364 2.4.21-47.0.1.EL (RHEL 3)
3365 2.6.12.6 vanilla (kernel.org)
3366 2.6.16.21-0.8 (SLES10)
3367 * Recommended e2fsprogs version: 1.39.cfs2-0
3369 * The backwards-compatible /proc/sys/portals symlink has been removed
3370 in this release. Before upgrading, please ensure that you change
3371 any configuration scripts or /etc/sysctl.conf files that access
3372 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3373 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3374 in advance of the upgrade on any system running Lustre 1.4.6 or
3375 newer, since /proc/sys/lnet was added in that version.
3376 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3380 Frequency : only when quota is used
3382 Description: avoid scanning export list for quota master
3383 Details : Change the algorithms to avoid scanning export list in order
3384 to improve the efficiency.
3387 Frequency : MDS failover only, very rarely
3389 Description: "went back in time" messages on mds failover
3390 Details : The greatest transno may be lost when the current operation
3391 finishes with an error (transno==0) and the client's last_rcvd
3392 record is over-written. Save the greatest transno in the
3393 mds_last_transno for this case.
3396 Frequency : always for specific kernels and striping counts
3398 Description: client may get "Matching packet too big" without ACL support
3399 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
3400 when trying to access files in certain configurations. The
3401 clients should in fact be denied when mounting because they do
3402 not understand ACLs.
3405 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
3407 Description: many clients connecting with IO in progress causes connect timeouts
3408 Details : Avoid synchronous journal commits to avoid delays caused by many
3409 clients connecting/disconnecting when bulk IO is in progress.
3410 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
3411 OST_IO_PORTAL to avoid delays behind potentially many pending
3415 Frequency : occasionally with multiple writers to a single file
3417 Description: shared writes to file may result in wrong size reported by stat()
3418 Details : Allow growing of kms when extent lock is cancelled
3421 Frequency : always with random mmap IO to multi-striped file
3423 Description: mmap write might be lost if we are writing to a 'hole' in stripe
3424 Details : Only if the hole is at the end of OST object so that kms is too
3425 small. Fix is to increase kms accordingly in ll_nopage.
3428 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
3430 Description: writes to a missing object would leak memory on the OST
3431 Details : If there is an inconsistency between the MDS and OST filesystems,
3432 such that the MDS references an object that doesn't exist, writes
3433 to that object will leak memory due to incorrect cleanup in the
3434 error handling path, eventually running out of memory on the OST.
3439 Description: Creating too long symlink causes lustre errors
3440 Details : Check symlink and name lengths before sending requests to MDS.
3443 Frequency : only if flock is enabled (not on by default)
3445 Description: posix locks not released on fd closure on 2.6.9+
3446 Details : We failed to add posix locks to list of inode locks on 2.6.9+
3447 kernels, this caused such locks not to be released on fd close and
3448 then assertions on fs unmount about still used locks.
3451 Frequency : MDS failover only, very rarely
3453 Description: clients may get ASSERTION(granted_lock != NULL)
3454 Details : When request was taking a long time, and a client was resending
3455 a getattr by name lock request. The were multiple lock requests
3456 with the same client lock handle and
3457 mds_getattr_name->fixup_handle_for_resent_request found one of the
3458 lock handles but later failed with ASSERTION(granted_lock != NULL).
3463 Description: handle->h_buffer_credits > 0, assertion failure
3464 Details : h_buffer_credits is zero after truncate, causing assertion
3465 failure. This patch extends the transaction or creates a new
3469 Frequency : NFS re-export or patchless client
3470 Bugzilla : 11179, 10796
3471 Description: Crash on NFS re-export node (__d_move)
3472 Details : We do not want to hash the dentry if we don't have a lock.
3473 But if this dentry is later used in d_move, we'd hit uninitialised
3474 list head d_hash, so we just do this to init d_hash field but
3475 leave dentry unhashed.
3478 Frequency : NFS re-export or patchless client
3480 Description: NFS exports has problem with symbolic link
3481 Details : lustre client didn't properly install dentry when re-exported
3482 to NFS or running patchless client.
3485 Frequency : NFS re-export or patchless client
3487 Description: Various nfs/patchless fixes.
3488 Details : fixes reuse disconected alias for lookup process - this fixes
3489 warning "find_exported_dentry: npd != pd",
3490 fix permission error with open files at nfs.
3491 fix apply umask when do revalidate.
3494 Frequency : occasional
3496 Description: Crash on NFS re-export node
3497 Details : calling clear_page() on the wrong pointer triggered oops in
3498 generic_mapping_read().
3501 Frequency : rarely, using O_DIRECT IO
3503 Description: unaligned directio crashes client with LASSERT
3504 Details : check for unaligned buffers before trying any requests.
3507 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
3509 Description: stale data returned from RAID cache
3510 Details : If only a small amount of IO is done to the RAID device before
3511 reading it again it is possible to get stale data from the RAID
3512 cache instead of reading it from disk.
3515 Frequency : always for sles10 kernel
3517 Description: sles10 support
3518 Details : ll_follow_link: compile fixes and using of nd_set_link
3519 under newer kernels.
3522 Frequency : depends on arch, kernel and compiler version, always on sles10
3525 Description: recursive or deep enough symlinks cause stack overflow
3526 Details : getting rid of large stack-allocated variable in
3530 Frequency : depends on hardware
3532 Description: lustre write performance loss in the SLES10 kernel
3533 Details : the performance loss is caused by using of write barriers in the
3534 ext3 code. The SLES10 kernel turns barrier support on by
3535 default. The fix is to undo that change for ldiskfs.
3537 ------------------------------------------------------------------------------
3539 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
3541 * Support for kernels:
3542 2.6.9-42.0.3EL (RHEL 4)
3543 2.6.5-7.276 (SLES 9)
3544 2.4.21-47.0.1.EL (RHEL 3)
3545 2.6.12.6 vanilla (kernel.org)
3549 Frequency : quota enabled and large files being deleted
3551 Description: releasing more than 4GB of quota at once hangs OST
3552 Details : If a user deletes more than 4GB of files on a single OST it
3553 will cause the OST to spin in an infinite loop. Release
3554 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3559 Description: statfs data retrieved from /proc may be stale or zero
3560 Details : When reading per-device statfs data from /proc, in the
3561 {kbytes,files}_{total,free,avail} files, it may appear
3562 as zero or be out of date.
3565 Frequency : systems with MD RAID1 external journal devices
3567 Description: lconf's call to blkid is confused by RAID1 journal devices
3568 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3569 of returning all block devices that match the journal UUID.
3572 Frequency : always, for aggregate stripe size over 4GB
3574 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
3575 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
3576 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
3579 Frequency : always if "lfs find" used on a local file/directory
3581 Description: "lfs find" segfaults if used on a local file/directory
3582 Details : The case where a directory component was not specified wasn't
3583 handled correctly. Handle this properly.
3586 Frequency : always on ppc64
3588 Description: the write to an ext3 filesystem mounted with mballoc got stuck
3589 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3590 perform endianness conversion.
3593 Frequency : rarely (truncate to non-zero file size after write under load)
3594 Bugzilla : 10730, 10687
3595 Description: Files padded with zeros to next 4K multiple
3596 Details : With filesystems mounted using the "extents" option (2.6 kernels)
3597 it is possible that files that are truncated to a non-zero size
3598 immediately after being written are filled with zero bytes beyond
3599 the truncated size. No file data is lost.
3601 Severity : enhancement
3603 Description: Allow recovery/failover for liblustre clients.
3604 Details : liblustre clients were unaware of failover configurations until
3607 Severity : enhancement
3609 Description: user file locks should fail when not mounting with flock option
3610 Details : Set up an error-returning stub in ll_file_operations.lock field
3611 to prevent incorrect behaviour when client is mounted without
3612 flock option. Also, set up properly f_op->flock field for
3616 Frequency : always on ia64
3618 Description: "lfs df" loops on printing out MDS statfs information
3619 Details : The obd_ioctl_data was not initialized and in some systems
3620 this caused a failure during the ioctl that did not return
3621 an error. Initialize the struct and return an error on failure.
3624 Frequency : SLES 9 only
3626 Description: Error of copying files with lustre special EAs as root
3627 Details : Client side always return success for setxattr call for lustre
3628 special xattr (currently only "trusted.lov").
3631 Frequency : rarely on clusters with both ia64+i386 clients
3633 Description: ia64+i686 clients doing shared IO on the same file may LBUG
3634 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
3635 clients are doing concurrent writes to the same file it is
3636 possible that the ia64 clients may LASSERT because the OST
3637 extent locks are not PAGE_SIZE aligned. Ensure that grown
3638 locks are always aligned on the request boundary.
3641 Frequency : specific use, occasional
3643 Description: Overwriting in use executable truncates on-disk binary image
3644 Details : If one node attempts to overwrite an executable in use by
3645 another node, we now correctly return ETXTBSY instead of
3646 truncating the file.
3648 Severity : enhancement
3650 Description: Async OSC create to avoid the blocking unnecessarily.
3651 Details : If a OST has no remain object, system will block on the creating
3652 when need to create a new object on this OST. Now, ways use
3653 pre-created objects when available, instead of blocking on an
3654 empty osc while others are not empty. If we must block, we block
3655 for the shortest possible period of time.
3660 Description: chmod on Lustre root is propagated to other clients
3661 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
3662 invalid by the follow_mount time.
3665 Frequency : liblustre clients only
3667 Description: Race in 'instant cancel' lock handling could lead to such locks
3668 never to be granted in case of SMP MDS
3669 Details : Do not destroy not yet granted but cbpending locks in
3673 Frequency : replay/resend of open
3675 Description: non null lock assetion failure in mds_intent_policy
3676 Details : Trying to replay/resend lockless open requests resulted in
3677 mds_open() returning 0 with no lock. Now it sets a flag if
3678 a lock is going to be returned.
3680 Severity : enhancement
3682 Description: Checksum enhancements
3683 Details : New checksum enhancements allow for resending RPCs that failed
3686 Severity : enhancement
3688 Description: Tunables on number of dirty pages in cacche
3689 Details : Allow to set limit on number of dirty pages cached.
3694 Description: client crash on unmount - lock still has references
3695 Details : In some error handling cases it was possible to leak a lock
3696 reference on a client while accessing a file. This was not
3697 harmful to the client during operation, but would cause the
3698 client to crash when the filesystem is unmounted.
3701 Frequency : specific case, rare
3703 Description: ETXTBSY on mds though file not in use by client
3704 Details : ETXTBSY is no longer incorrectly returned when attempting to
3705 chmod or chown a directory that the user previously tried to
3706 execute or a currently-executing binary.
3709 Frequency : extremely rare except on liblustre-based clients
3711 Description: Lustre space not freed when files are deleted
3712 Details : Clean up open-unlinked files after client eviction. Previously
3713 the unlink was skipped and the files remained as orphans.
3718 Description: OST failure "would be an LBUG" in waiting_locks_callback()
3719 Details : In some cases it was possible to send a blocking callback to
3720 a client doing a glimpse, even though that client didn't get
3721 a lock granted. When the glimpse lock is cancelled on the OST
3722 the freed lock is left on the waiting list and corrupted the list.
3725 Frequency : all core dumps
3727 Description: Broke core dumps to lustre
3728 Details : Negative dentry may be unhashed if parent does not have UPDATE
3729 lock, but some callers, e.g. do_coredump, expect dentry to be
3730 hashed after successful create, hash it in ll_create_it.
3732 ------------------------------------------------------------------------------
3734 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
3736 * Support for kernels:
3737 2.6.9-42.0.2.EL (RHEL 4)
3738 2.6.5-7.276 (SLES 9)
3739 2.4.21-40.EL (RHEL 3)
3740 2.6.12.6 vanilla (kernel.org)
3744 Frequency : always on RHEL 3
3746 Description: Number of open files grows over time
3747 Details : The number of open files grows over time, whether or not
3748 Lustre is started. This was due to a filp leak introduced
3749 by one of our kernel patches.
3751 ------------------------------------------------------------------------------
3753 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
3755 * Support for kernels:
3756 2.6.9-42.EL (RHEL 4)
3757 2.6.5-7.276 (SLES 9)
3758 2.4.21-40.EL (RHEL 3)
3759 2.6.12.6 vanilla (kernel.org)
3764 Bugzilla : 5719, 9635, 9792, 9684
3765 Description: OST (or MDS) trips assertions in (re)connection under heavy load
3766 Details : If a server is under heavy load and cannot reply to new
3767 connection requests before the client resends the (re)connect,
3768 the connection handling code can behave badly if two service
3769 threads are concurrently handing separate (re)connections from
3770 the same client. Add better locking to the connection handling
3771 code, and ensure that only a single connection will be processed
3772 for a given client UUID, even if the lock is dropped.
3774 Severity : enhancement
3776 Description: add TCP zero-copy support to kernel
3777 Details : Add support to the kernel TCP stack to allow zero-copy bulk
3778 sends if the hardware supports scatter-gather and checksumming.
3779 This allows socklnd to do client-write and server-read more
3780 efficiently and reduce CPU utilization from skbuf copying.
3783 Frequency : only if NFS exporting from client
3785 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
3786 Details : If setting attributes on a file created under NFS that had
3787 never been opened it would be possible to oops the client
3788 if the file had no objects.
3791 Frequency : always for liblustre
3793 Description: liblustre client does MDS+OSTs setattr RPC for each write
3794 Details : When doing a write from a liblustre client, the client
3795 incorrectly issued an RPC to the MDS and each OST the file was
3796 striped over in order to update the timestamps. When writing
3797 with small chunks and many clients this could overwhelm the MDS
3798 with RPCs. In all cases it would slow down the write because
3799 these RPCs are unnecessary.
3801 Severity : enhancement
3803 Description: allow number of MDS service threads to be changed at module load
3804 Details : It is now possible to change the number of MDS service threads
3805 running. Adding "options mds mds_num_threads={N}" to the MDS's
3806 /etc/modprobe.conf will set the number of threads for the next
3807 time Lustre is restarted (assuming the "mds" module is also
3808 reloaded at that time). The default number of threads will
3809 stay the same, 32 for most systems.
3814 Description: OST crash if filesystem is unformatted or corrupt
3815 Details : If an OST is started on a device that has never been formatted
3816 or if the filesystem is corrupt and cannot even mount then the
3817 error handling cleanup routines would dereference a NULL pointer.
3822 Description: NULL pointer deref in llap_from_page.
3823 Details : get_cache_page_nowait can return a page with NULL (or otherwise
3824 incorrect) mapping if the page was truncated/reclaimed while it was
3825 searched for. Check for this condition and skip such pages when
3826 doing readahead. Introduce extra check to llap_from_page() to
3827 verify page->mapping->host is non-NULL (so page is not anonymous).
3830 Frequency : Sometimes when using sys_sendfile
3832 Description: "page not covered by a lock" warnings from ll_readpage
3833 Details : sendfile called ll_readpage without right page locks present.
3834 Now we introduced ll_file_sendfile that does necessary locking
3835 around call to generic_file_sendfile() much like we do in
3839 Frequency : with certain MDS communication failures at client mount time
3841 Description: NULL pointer deref after failed client mount
3842 Details : a client connection request may delayed by the network layer
3843 and not be sent until after the PTLRPC layer has timed out the
3844 request. If the client fails the mount immediately it will try
3845 to clean up before the network times out the request. Add a
3846 reference from the request import to the obd device and delay
3847 the cleanup until the network drops the request.
3850 Frequency : occasionally during client (re)connect
3852 Description: assertion failure during client (re)connect
3853 Details : processing a client connection request may be delayed by the
3854 client or server longer than the client connect timeout. This
3855 causes the client to resend the connection request. If the
3856 original connection request is replied in this interval, the
3857 client may trip an assertion failure in ptlrpc_connect_interpret()
3858 which thought it would be the only running connect process.
3861 Frequency : only with obd_echo servers and clients that are rebooted
3863 Description: kernel BUG accessing uninitialized data structure
3864 Details : When running an obd_echo server it did not start the ping_evictor
3865 thread, and when a client was evicted an uninitialized data
3866 structure was accessed. Start the ping_evictor in the RPC
3867 service startup instead of the OBD startup.
3869 Severity : enhancement
3870 Bugzilla : 10193 (patchless)
3871 Description: Remove dependency on various unexported kernel interfaces.
3872 Details : No longer need reparent_to_init, exit_mm, exit_files,
3873 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
3876 Frequency : rare (only users of deprecated and unsupported LDAP config)
3878 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
3879 Details : LDAP apparently contains 'lustreName' attributes instead of
3880 'name'. A simple remapping of the name is sufficient.
3883 Frequency : rare (only with non-default dump_on_timeout debug enabled)
3885 Description: waiting_locks_callback trips kernel BUG if client is evicted
3886 Details : Running with the dump_on_timeout debug flag turned on makes
3887 it possible that the waiting_locks_callback() can try to dump
3888 the Lustre kernel debug logs from an interrupt handler. Defer
3889 this log dumping to the expired_lock_main() thread.
3891 Severity : enhancement
3893 Description: Support NFS exporting on 2.6 kernels.
3894 Details : Implement non-rawops metadata methods for NFS server to use without
3895 changing NFS server code.
3898 Frequency : very rare (synthetic metadata workload only)
3900 Description: two racing renames might cause an MDS thread to deadlock
3901 Details : Running the "racer" program may cause one MDS thread to rename
3902 a file from being the source of a rename to being the target of
3903 a rename at exactly the same time that another thread is doing
3904 so, and the second thread has already enqueued these locks after
3905 doing a lookup of the target and is trying to relock them in
3906 order. Ensure that we don't try to re-lock the same resource.
3909 Frequency : only very large systems with liblustre clients
3911 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
3912 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
3913 all OSTs in parallel. This allows the eviction of stale liblustre
3914 clients to proceed much faster than if they were done in series,
3915 and also offers similar improvements for other set_info RPCs.
3920 Description: excessive CPU usage during initial read phase on client
3921 Details : During the initial read phase on a client, it would agressively
3922 retry readahead on the file, consuming too much CPU and impacting
3923 performance (since 1.4.5.8). Improve the readahead algorithm
3924 to avoid this, and also improve some other common cases (read
3925 of small files in particular, where "small" is files smaller than
3926 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
3931 Description: MDS crash when receiving packet with unknown intent.
3932 Details : Do not LBUG in unknown intent case, just return -EFAULT
3934 Severity : enhancement
3935 Bugzilla : 9293, 9385
3936 Description: MDS RPCs are serialised on client. This is unnecessary for some.
3937 Details : Do not serialize getattr (non-intent version) and statfs.
3940 Frequency : occasional, when OST network is overloaded/intermittent
3942 Description: client evicted by OST after bulk IO timeout
3943 Details : If a client sends a bulk IO request (read or write) the OST
3944 may evict the client if it is unresposive to its data GET/PUT
3945 request. This is incorrect if the network is overloaded (takes
3946 too long to transfer the RPC data) or dropped the OST GET/PUT
3947 request. There is no need to evict the client at all, since
3948 the pinger and/or lock callbacks will handle this, and the
3949 client can restart the bulk request.
3952 Frequency : Always when mmapping file with no objects
3954 Description: client crashes when mmapping file with no objects
3955 Details : Check that we actually have objects in a file before doing any
3956 operations on objects in ll_vm_open, ll_vm_close and
3962 Description: Request leak when working with deleted CWD
3963 Details : Introduce advanced request refcount tracking for requests
3964 referenced from lustre intent.
3966 Severity : Enhancement
3968 Description: Cache open file handles on client.
3969 Details : MDS now will return special lock along with openhandle, if
3970 requested and client is allowed to hold openhandle, even if unused,
3971 until such a lock is revoked. Helps NFS a lot, since NFS is opening
3972 closing files for every read/write openration.
3974 Severity : Enhancement
3976 Description: Cache open negative dentries on client when possible.
3977 Details : Guard negative dentries with UPDATE lock on parent dir, drop
3978 negative dentries on lock revocation.
3983 Description: Remounting a client read-only wasn't possible with a zconf mount
3984 Details : It wasn't possible to remount a client read-only with llmount.
3986 Severity : enhancement
3987 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
3988 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
3989 MPICH is included to add Lustre-specific ADIO interfaces.
3990 This is based closely on the UFS ADIO layer and only differs
3991 in file creation, in order to allow the OST striping to be set.
3992 This is user-contributed code and not supported by CFS.
3997 Description: extended inode attributes (immutable, append-only) work improperly
3998 when 2.4 and 2.6 kernels are used on client/server or vice versa
3999 Details : Introduce kernel-independent values for these flags.
4001 Severity : enhancement
4004 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
4005 Details : Many of the /proc/ tunables can only be tuned at a megabyte
4006 granularity. Now, Fractional MB granularity is be supported,
4007 this is very useful for low memory system.
4009 Severity : enhancement
4011 Description: Getattr by fid
4012 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
4013 locks, avoids extra getattr rpc requests to MDS, allows '/' to
4014 have locks and avoids getattr rpc requests for it on every stat.
4017 Frequency : Always, for filesystems larger than 2TB
4019 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4020 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
4021 filesystems larger than 2TB to be mounted with mballoc enabled.
4024 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
4026 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4027 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
4028 filesystems larger than 2TB is mounted then the kernel will
4029 silently corrupt the start of the filesystem. CONFIG_LBD is
4030 enabled for all CFS-supported kernels, but the possibility of
4031 this happening with a modified kernel config exists.
4033 Severity : enhancement
4035 Description: add client O_DIRECT support for 2.6 kernels
4036 Details : It is now possible to do O_DIRECT reads and writes to files
4037 in the Lustre client mountpoint on 2.6 kernel clients.
4039 Severity : enhancement
4041 Description: parallel glimpse, setattr, statfs, punch, destroy requests
4042 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
4043 parallel, not waiting for response from every OST before sending
4044 a rpc to the next OST.
4049 Description: setattr vs write race when updating file timestamps
4050 Details : Client processes that update a file timestamp into the past
4051 right after writing to the file (e.g. tar) it is possible that
4052 the updated file modification time can be reset to the current
4053 time due to a race between processing the setattr and write RPC.
4055 Severity : enhancement
4057 Description: Bring 'lfs find' closer in line with regular Linux find.
4058 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
4059 -print0 options and obtains all the needed info through the lustre
4062 Severity : enhancement
4064 Description: support up to 1024 configured devices on one node
4065 Details : change obd_dev array from statically allocated to dynamically
4066 allocated structs as they are first used to reduce memory usage
4071 Description: Flush dirty partially truncated pages during truncate
4072 Details : Immediatelly flush partially truncated pages in filter_setattr,
4073 this way we completely avoid having any pages in page cache on OST
4074 and can retire ugly workarounds during writes to flush such pages.
4079 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
4080 Details : protect inode from truncation within vfs_unlink() context
4081 just take a reference before calling vfs_unlink() and release it
4082 when parent's i_sem is free.
4087 Description: last_id value checked outside lock on OST caused LASSERT failure
4088 Details : If there were multiple MDS->OST object precreate requests in
4089 flight, it was possible that the OST's last object id was checked
4090 outside a lock and incorrectly tripped an assertion. Move checks
4091 inside locks, and discard old precreate requests.
4094 Frequency : always, if extents are used on OSTs
4096 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
4097 Details : OSTs using the extents format would not zero the high 16 bits of
4098 the index physical block number. This is not a problem for any
4099 OST filesystems smaller than 16TB, and no kernels support ext3
4100 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
4101 new/modified files) and can be fixed for existing filesystems
4102 with e2fsprogs-1.39-cfs1.
4107 Description: import connection selection may be incorrect if timer wraps
4108 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
4109 import connections to be ignored if the 32-bit jiffies counter
4110 wraps. Use a 64-bit jiffies counter.
4113 Frequency : very large clusters immediately after boot
4115 Description: LNET request buffers exhausted under heavy short-term load
4116 Details : If a large number of client requests are generated on a service
4117 that has previously never seen so many requests it is possible
4118 that the request buffer growth cannot keep up with the spike in
4119 demand. Instead of dropping incoming requests, they are held in
4120 the LND until the RPC service can accept more requests.
4123 Frequency : Sometimes during replay
4125 Description: Assertion failure in ll_local_open after replay.
4126 Details : If replay happened on an open request reply before we were able
4127 to set replay handler, reply will become not swabbed tripping the
4128 assertion in ll_local_open. Now we set the handler right after
4129 recognising of open request
4132 Frequency : very rare
4134 Description: kernel reports "badness in vsnprintf"
4135 Details : Reading from the "recovery_status" /proc file in small chunks
4136 may cause a negative length in lprocfs_obd_rd_recovery_status()
4137 call to vsnprintf() (which is otherwise harmless). Exit early
4138 if there is no more space in the output buffer.
4140 Severity : enhancement
4142 Description: clear OBD RPC statistics by writing to them
4143 Details : It is now possible to clear the OBD RPC statistics by writing
4144 to the "stats" file.
4149 Description: Client mtime is not the same on different clients after utimes
4150 Details : In some cases, the client was using the utimes() syscall on
4151 a file cached on another node. The clients now validate the
4152 ctime from the MDS + OSTs to determine which one is right.
4157 Description: Inability to activate failout mode
4158 Details : lconf script incorrectly assumed that in pythong string's numeric
4159 value is used in comparisons.
4162 Frequency : always with multiple stripes per file
4164 Description: Inefficient object allocation for mutli-stripe files
4165 Details : When selecting which OSTs to stripe files over, for files with
4166 a stripe count that divides evenly into the number of OSTs,
4167 the MDS is always picking the same starting OST for each file.
4168 Return the OST selection heuristic to the original design.
4173 Description: mount failures may take full timeout to return an error
4174 Details : Under some heavy load conditions it is possible that a
4175 failed mount can wait for the full obd_timeout interval,
4176 possibly several minutes, before reporting an error.
4177 Instead return an error as soon as the status is known.
4179 Frequency : quota enabled and large files being deleted
4181 Description: releasing more than 4GB of quota at once hangs OST
4182 Details : If a user deletes more than 4GB of files on a single OST it
4183 will cause the OST to spin in an infinite loop. Release
4184 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4189 Description: statfs data retrieved from /proc may be stale or zero
4190 Details : When reading per-device statfs data from /proc, in the
4191 {kbytes,files}_{total,free,avail} files, it may appear
4192 as zero or be out of date.
4195 Frequency : systems with MD RAID1 external journal devices
4197 Description: lconf's call to blkid is confused by RAID1 journal devices
4198 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4199 of returning all block devices that match the journal UUID.
4202 Frequency : always, for aggregate stripe size over 4GB
4204 Description: assertion fails when trying to use 4GB stripe size
4205 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
4206 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
4209 Frequency : always on ppc64
4211 Description: the first write on an ext3 filesystem with mballoc got stuck
4212 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4213 perform endianness conversion.
4215 ------------------------------------------------------------------------------
4217 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4219 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
4220 INTEROPERATE with older versions automatically. Please read the
4221 user documentation before upgrading any part of a live system.
4222 * WARNING: Lustre networking configuration changes are required with
4223 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
4226 * Support for kernels:
4227 2.6.9-22.0.2.EL (RHEL 4)
4228 2.6.5-7.244 (SLES 9)
4229 2.6.12.6 vanilla (kernel.org)
4232 Severity : enhancement
4233 Bugzilla : 7981/8208
4234 Description: Introduced Lustre Networking (LNET)
4235 Details : LNET is new networking infrastructure for Lustre, it includes
4236 a reorganized network configuration mode (see the user
4237 documentation for full details) as well as support for routing
4238 between different network fabrics. Lustre Networking Devices
4239 (LNDs) for the supported network fabrics have also been
4240 created for this new infrastructure.
4242 Severity : enhancement
4243 Description: Introduced Access control lists
4244 Details : clients can set ACLs on files and directories in order to have
4245 more fine-grained permissions than the standard Unix UGO+RWX.
4246 The MDS must be started with the "-o acl" mount option.
4248 Severity : enhancement
4249 Description: Introduced filesystem quotas
4250 Details : Administrators may now establish per-user quotas on the
4253 Severity : enhancement
4255 Description: Configuration change for the XT3
4256 The PTLLND is now used to run Lustre over Portals on the XT3
4257 The configure option(s) --with-cray-portals are no longer used.
4258 Rather --with-portals=<path-to-portals-includes> is used to
4259 enable building on the XT3. In addition to enable XT3 specific
4260 features the option --enable-cray-xt3 must be used.
4265 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
4266 Details : A race between allocating a new llcd and re-getting the llcd_lock
4267 allowed another thread to grab newly-allocated llcd.
4269 Severity : enhancement
4271 Description: 2.6 OST async journal commit and locking fix to improve performance
4272 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
4273 2.6 kernels are now async as they already were in 2.4 kernels so
4274 that they can commit concurrently with the network bulk transfer.
4275 For block-allocated files the filter allocation semaphore is held
4276 to avoid filesystem fragmentation during allocation. BKL lock
4277 removed for 2.6 xattr operations where it is no longer needed.
4282 Description: lconf incorrectly determined whether two IP networks could talk
4283 Details : In some more complicated routing and multiple-network
4284 configurations, lconf will avoid trying to make a network
4285 connection to a disjoint part of the IP space. It was doing the
4286 math incorrectly for one set of cases.
4291 Description: Fix for potential infinite loop processing records in an llog.
4292 Details : If an llog record is corrupted/zeroed, it is possible to loop
4293 forever in llog_process(). Validate the llog record length
4294 and skip the remainder of the block on error.
4297 Frequency : occasional (liblustre only)
4299 Description: liblustre could not open files whose last component is a symlink
4300 Details : sysio_path_walk() would incorrectly pass the open intent to
4301 intermediate path components.
4304 Frequency : rare (liblustre only with non-standard tuning)
4305 Bugzilla : 7201 (7350)
4306 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4307 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4308 after it has been released, only do this for referenced locks
4310 Severity : enhancement
4312 Description: specifying an (invalid) directory default stripe_size of -1
4313 would reset the directory default striping
4314 Details : stripe_size -1 was used internally to signal directory stripe
4315 removal, now use "all default" to signal dir stripe removal
4316 as a directory striping of "all default" is not useful
4319 Frequency : common for large clusters running liblustre clients
4321 Description: doing an ls when liblustre clients are running is slow
4322 Details : sending a glimpse AST to a liblustre client waits for every AST
4323 to time out, as liblustre clients will not respond. Since they
4324 cannot cache data we refresh the OST lock LVB from disk instead.
4326 Severity : enhancement
4328 Description: doing an ls at the same time as file IO can be slow
4329 Details : enqueue and other "small" requests can be blocked behind many
4330 large IO requests. Create a new OST IO portal for non-IO
4331 requests so they can be processed faster.
4334 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4336 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4337 created file with mode 000
4338 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4339 allow an ftruncate() as a non-root user to a file with mode 000.
4340 Lustre now respects this flag to disable mode checks when
4341 truncating a file owned by the user
4344 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4347 Description: Revoking locks from clients that went dead or catatonic might take
4349 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4350 cancellation of such locks instant on servers without waiting for
4351 any reply from clients. Clients drops these locks when cancel
4352 notification from server is received without replying.
4355 Frequency : liblustre-only, when liblustre client dies or becomes busy
4357 Description: Doing ls on Linux clients can take a long time with active
4359 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4360 granting such locks to it in the first place if possible. Locks
4361 are taken by proxy on the OST during the read or write and
4362 dropped immediately afterward. Add connect flags handling, do
4363 not grant locks to liblustre clients for glimpse ASTs.
4365 Severity : enhancement
4367 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4368 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4369 sequential reads and behaves badly with applications doing random
4370 reads. Improve it by reading ahead at least read region, and
4371 avoiding excessive large RPC for small reads.
4373 Severity : enhancement
4375 Description: Creating more than 1000 files for a single job may cause a load
4376 imbalance on the OSTs if there are also a large number of OSTs.
4377 Details : qos_prep_create() uses an OST index reseed value that is an
4378 even multiple of the number of available OSTs so that if the
4379 reseed happens in the middle of the object allocation it will
4380 still utilize the OSTs as uniformly as possible.
4385 Description: OST or MDS may oops in ping_evictor_main()
4386 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
4387 but doesn't restart at beginning of obd_exports_timed list
4390 Severity : enhancement
4392 Description: improve by-nid export eviction on the MDS and OST
4393 Details : allow multiple exports with the same NID to be evicted at one
4394 time without re-searching the exports list.
4397 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
4399 Description: MDS may oops in groups_free()
4400 Details : in rare race conditions a newly allocated group_info struct is
4401 freed again, and this can be NULL. The 2.4 compatibility code
4402 for groups_free() checked for a NULL pointer, but 2.6 did not.
4405 Frequency : common for liblustre clients doing little filesystem IO
4406 Bugzilla : 9352, 7313
4407 Description: server may evict liblustre clients accessing contended locks
4408 Details : if a client is granted a lock or receives a completion AST
4409 with a blocking AST already set it would not reply to the AST
4410 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
4413 Frequency : lfs setstripe, only systems with more than 160 OSTs
4415 Description: unable to set striping with a starting offset beyond OST 160
4416 Details : llapi_create_file() incorrectly limited the starting stripe
4417 index to the maximum single-file stripe count.
4420 Frequency : LDAP users only
4422 Description: lconf did not handle in-kernel recovery with LDAP properly
4423 Details : lconf/LustreDB get_refs() is searching the wrong namespace
4425 Severity : enhancement
4427 Description: bind OST threads to NUMA nodes to improve performance
4428 Details : all OST threads are uniformly bound to CPUs on a single NUMA
4429 node and do their allocations there to localize memory access
4431 Severity : enhancement
4433 Description: llmount can determine client NID directly from Myrinet (GM)
4434 Details : the client NID code from gmnalnid was moved directly into
4435 llmount, removing the need to use this or specifying the
4436 client NID explicitly when mounting GM clients with zeroconf
4439 Frequency : if client is started with down MDS
4441 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
4442 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
4443 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
4444 ptlrpc_import_delayed_req() because we didn't check if the
4445 request was interrupted, and we also didn't break out of the
4446 event loop if there was no timeout
4451 Description: data loss during non-page-aligned writes to a single file from
4452 both multiple nodes and multiple threads on one node at same time
4453 Details : updates to KMS and lsm weren't protected by common lock. Resulting
4454 inconsistency led to false short-reads, that were cached and later
4455 used by ->prepare_write() to fill in partially written page,
4456 leading to data loss.
4459 Frequency : always, if lconf --abort_recovery used
4461 Description: lconf --abort_recovery fails with 'Operation not supported'
4462 Details : lconf was attempting to abort recovery on the MDT device and not
4465 Severity : enhancement
4467 Description: remove cleanup logs
4468 Details : replace lconf-generated cleanup logs with lustre internal
4469 cleanup routines. Eliminates the need for client-cleanup and
4472 Severity : enhancement
4474 Description: add support for EAs (user and system) on lustre filesystems
4475 Details : it is now possible to store extended attributes in the Lustre
4476 client filesystem, and with the user_xattr mount option it
4477 is possible to allow users to store EAs on their files also
4479 Severity : enhancement
4481 Description: Add possibility (config option) to show minimal available OST free
4483 Details : When compiled with --enable-mindf configure option, statfs(2)
4484 (and so, df) will return least minimal free space available from
4485 all OSTs as amount of free space on FS, instead of summary of
4486 free spaces of all OSTs.
4488 Severity : enhancement
4490 Description: do not expand extent locks acquired on OST-side
4491 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
4492 by server: they are not cached anyway.
4495 Frequency : when mmap is used/binaries executed from Lustre
4497 Description: Unmmap pages before throwing them away from read cache.
4498 Details : llap_shrink cache now attempts to unmap pages before discarding
4499 them (if unmapping failed - do not discard). SLES9 kernel has
4500 extra checks that trigger if this unmapping is not done first.
4505 Description: lconf didn't resolve symlinks before checking to see whether a
4506 given mountpoint was already in use
4509 Frequency : when migrating failover services
4510 Bugzilla : 6395, 9514
4511 Description: When migrating a subset of services from a node (e.g. failback
4512 from a failover service node) the remaining services would
4513 time out and evict clients.
4514 Details : lconf --force (implied by --failover) sets the global obd_timeout
4515 to 5 seconds in order to quickly disconnect, but this caused
4516 other RPCs to time out too quickly. Do not change the global
4517 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
4519 Severity : enhancement
4520 Frequency : if MDS is started with down OST
4521 Bugzilla : 9439,5706
4522 Description: Allow startup/shutdown of an MDS without depending on the
4523 availability of the OSTs.
4524 Details : Asynchronously call mds_lov_synchronize during MDS startup.
4525 Add appropriate locking and lov-osc refcounts for safe
4526 cleaning. Add osc abort_inflight calls in case the
4527 synchronize never started.
4530 Frequency : occasional (Cray XT3 only)
4532 Description: root not authorized to access files in CRAY_PORTALS environment
4533 Details : The client process capabilities were not honoured on the MDS in
4534 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
4535 been accessed by an authorized user then root was able to access
4536 the file on the local client also. The root user capabilities
4537 are now allowed on the MDS, as this environment has secure UID.
4540 Frequency : occasional
4542 Description: ldiskfs "too long searching" message happens too often
4543 Details : A debugging message (otherwise harmless) prints too often on
4544 the OST console. This has been reduced to only happen when
4545 there are fragmentation problems on the filesystem.
4550 Description: Division by zero in statfs when all OSCs are inactive
4551 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
4552 lov_statfs divides by value returned by lov_get_stripecnt().
4556 Bugzilla : 9489, 3273
4557 Description: First write from each client to each OST was only 4kB in size,
4558 to initialize client writeback cache, which caused sub-optimal
4559 RPCs and poor layout on disk for the first writen file.
4560 Details : Clients now request an initial cache grant at (re)connect time
4561 and so that they can start streaming writes to the cache right
4562 away and always do full-sized RPCs if there is enough data.
4563 If the OST is rebooted the client also re-establishes its grant
4564 so that client cached writes will be honoured under the grant.
4569 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
4570 Details : Now I/O RPCs go to different portal number and (presumably) fast
4571 lock requests (and glimses) and other RPCs get their own service
4572 threads pool that should be able to service those RPCs
4575 Severity : enhancement
4577 Description: Ability to exchange lustre version between client and servers and
4578 issue warnings at client side if client is too old. Also for
4579 liblustre clients there is ability to refuse connection of too old
4581 Details : New 'version' field is added to connect data structure that is
4582 filled with version info. That info is later checked by server and
4586 Frequency : rare, liblustre only.
4587 Bugzilla : 9296, 9581
4588 Description: Two simultaneous writes from liblustre at offset within same page
4589 might proceed at the same time overwriting eachother with stale
4591 Details : I/O lock withing llu_file_prwv was released too early, before data
4592 actually was hitting the wire. Extended lock-holding time until
4593 server acknowledges receiving data.
4596 Frequency : extremely rare. Never observed in practice.
4598 Description: avoid generating lustre_handle cookie of 0.
4599 Details : class_handle_hash() generates handle cookies by incrementing
4600 global counter, and can hit 0 occasionaly (this is unlikely, but
4601 not impossible, because initial value of cookie counter is
4602 selected randonly). Value of 0 is used as a sentinel meaning
4603 "unassigned handle" --- avoid it. Also coalesce two critical
4604 sections in this function into one.
4606 Severity : enhancement
4608 Description: allow liblustre clients to delegate truncate locking to OST
4609 Details : To avoid overhead of locking, liblustre client instructs OST to
4610 take extent lock in ost_punch() on client's behalf. New connection
4611 flag is added to handle backward compatibility.
4613 Severity : enhancement
4614 Bugzilla : 4928, 7341, 9758
4615 Description: allow number of OST service threads to be specified
4616 Details : a module parameter allows the number of OST service threads
4617 to be specified via "options ost ost_num_threads={N}" in the
4618 OSS's /etc/modules.conf or /etc/modprobe.conf.
4622 Bugzilla : 6146, 9635, 9895
4623 Description: servers crash with bad pointer in target_handle_connect()
4624 Details : In rare cases when a client is reconnecting it was possible that
4625 the connection request was the last reference for that export.
4626 We would temporarily drop the export reference and get a new
4627 one, but this may have been the last reference and the export
4628 was just destroyed. Get new reference before dropping old one.
4630 Severity : enhancement
4631 Frequency : if client is started with failover MDS
4633 Description: Allow multiple MDS hostnames in the mount command
4634 Details : Try to read the configuration from all specified MDS
4635 hostnames during a client mount in case the "primary"
4638 Severity : enhancement
4640 Description: Stop sending data to evicted clients as soon as possible.
4641 Details : Check if the client we are about to send or are sending data to
4642 was evicted already. (Check is done every second of waiting,
4643 for which l_wait_event interface was extended to allow checking
4644 of exit condition at specified intervals).
4647 Frequency : rare, normally only when NFS exporting is done from client
4649 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
4651 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
4654 Frequency : always, if srand() is called before liblustre initialization
4656 Description: Liblustre uses system PRNG disturbing its usage by user application
4657 Details : Introduce internal to lustre fast and high-quality PRNG for
4658 lustre usage and make liblustre and some other places in generic
4659 lustre code to use it.
4661 Severity : enhancement
4662 Bugzilla : 9477, 9557, 9870
4663 Description: Verify that the MDS configuration logs are updated when xml is
4664 Details : Check if the .xml configuration logs are newer than the config
4665 logs stored on the MDS and report an error if this is the case.
4666 Request --write-conf, or allow starting with --old_conf.
4668 Severity : enhancement
4670 Description: Handle symlinks in the path when checking if Lustre is mounted.
4671 Details : Resolve intermediate symlinks when checking if a client has
4672 mounted a filesystem to avoid duplicate client mounts.
4677 Description: lconf can hit an error exception but still return success.
4678 Details : The lconf command catches the Command error exception at the top
4679 level script context and will exit with the associated exit
4680 status, but doesn't ensure that this exit status is non-zero.
4685 Description: failure of ptlrpc thread startup can cause oops
4686 Details : Starting a ptlrpc service thread can fail if there are a large
4687 number of threads or the server memory is very fragmented.
4688 Handle this without oopsing.
4691 Frequency : always, only if liblustre and non-default acceptor port was used
4693 Description: liblustre cannot connect to servers with non-default acceptor port
4694 Details : tcpnal_set_default_params() was not called and was therefore
4695 ignoring the environment varaible TCPNAL_PORT, as well as other
4696 TCPNAL_ environment variables
4701 Description: two objects could be created on the same OST for a single file
4702 Details : If an OST is down, in some cases it was possible to create two
4703 objects on a single OST for a single file. No problems other
4704 than potential performance impact and spurious error messages.
4708 Bugzilla : 5681, 9562
4709 Description: Client may oops in ll_unhash_aliases
4710 Details : Client dcache may become inconsistent in race condition.
4711 In some cases "getcwd" can fail if the current directory is
4717 Description: Inode refcounting problems in NFS export code
4718 Details : link_raw functions used to call d_instantiate without obtaining
4719 extra inode reference first.
4723 Bugzilla : 9942, 9903
4724 Description: Referencing freed requests leading to crash, memleaks with NFS.
4725 Details : We used to require that call to ll_revalidate_it was always
4726 followed by ll_lookup_it. Also with revalidate_special() it is
4727 possible to call ll_revalidate_it() twice for the same dentry
4728 even if first occurence returned success. This fix changes semantic
4729 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
4730 reference on a request referred from the intent.
4731 ll_intent_release() then releases such a request.
4734 Frequency : rare, normally benchmark loads only
4736 Description: unlinked inodes were kept in memory on the client
4737 Details : If a client is repeatedly creating and unlinking files it
4738 can accumulate a lot of stale inodes in the inode slab cache.
4739 If there is no other client load running this can cause the
4740 client node to run out of memory. Instead flush old inodes
4741 from client cache that have the same inode number as a new inode.
4744 Frequency : SLES9 2.6.5 kernel and long filenames only
4745 Bugzilla : 9969, 10379
4746 Description: utime reports stale NFS file handle
4747 Details : SLES9 uses out-of-dentry names in some cases, which confused
4748 the lustre dentry revalidation. Change it to always use the
4752 Frequency : rare, unless heavy write-truncate concurrency is continuous
4753 Bugzilla : 4180, 6984, 7171, 9963, 9331
4754 Description: OST becomes very slow and/or deadlocked during object unlink
4755 Details : filter_destroy() was holding onto the parent directory lock
4756 while truncating+unlinking objects. For very large objects this
4757 may block other threads for a long time and slow overall OST
4758 responsiveness. It may also be possible to get a lock ordering
4759 deadlock in this case, or run out of journal credits because of
4760 the combined truncate+unlink. Solution is to do object truncate
4761 first in one transaction without parent lock, and then do the
4762 final unlink in a new transaction with the parent lock. This
4763 reduces the lock hold time dramatically.
4766 Frequency : rare, 2.4 kernels only
4768 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
4769 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
4770 before unmounting their filesystem in order to avoid deadlock.
4771 The kernel_locked() function in 2.4 kernels only checks whether
4772 the kernel lock is held, not whether it is this process that is
4773 holding it as 2.6 kernels do.
4778 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
4779 Details : The client ptlrpc code may be trying to reconnect to a down
4780 server before a previous connection attempt has timed out.
4781 Increase the reconnect interval to be longer than the connection
4782 timeout interval to avoid sending duplicate connections to
4786 Frequency : echo_client brw_test command
4788 Description: fix echo_client to work with OST preallocated code
4789 Details : OST preallocation code (5137) didn't take echo_client IO path
4790 into account: echo_client calls filter methods outside of any
4791 OST thread and, hence, there is no per-thread preallocated
4792 pages and buffers to use. Solution: hijack pga pages for IO. As
4793 a byproduct, this avoids unnecessary data copying.
4797 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
4798 Description: Client can oops in mdc_commit_close() after open replay
4799 Details : It was possible for the MDS to return an open request with no
4800 transaction number in mds_finish_transno() if the client was
4801 evicted, but without actually returning an error. Clients
4802 would later try to replay that open and may trip an assertion
4803 Simplify the client close codepath, and always return an error
4804 from the MDS in case the open is not successful.
4807 Frequency : rare, 2.6 OSTs only
4809 Description: OST may deadlock under high load on fragmented files
4810 Details : If there was a heavy load and highly-fragmented OST filesystems
4811 it was possible to have all the OST threads deadlock waiting on
4812 allocation of biovecs, because the biovecs were not released
4813 until the entire RPC IO was completed. Instead, release biovecs
4814 as soon as they are complete to ensure forward IO progress.
4816 Severity : enhancement
4818 Description: Support for specifying external journal device at mount
4819 Details : If an OST or MDS device is formatted with an external journal
4820 device, this device major/minor is stored in the ext3 superblock
4821 and may not be valid for failover. Allow detecting and
4822 specifying the external journal at mount time.
4827 Description: Mounting an MDS with pending unlinked files may cause oops
4828 Details : target_finish_recovery() calls mds_postrecov() which returned
4829 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
4830 considers this an error and immediately begins cleaning up the
4831 lov, just after starting the mds_lov process
4833 Severity : enhancement
4835 Description: Implement 'lfs df' to report actual free space on per-OST basis
4836 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
4837 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
4838 usage of objects; '-h' to report in human readable format.
4840 ------------------------------------------------------------------------------
4842 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
4849 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
4850 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
4851 at startup during orphan unlinking. Instead, skip buddy update
4852 before setup, as it will be regenerated after recovery is complete.
4857 Description: If an OST is inactive, its locks might reference stale inodes.
4858 Details : lov_change_cbdata() must iterate over all namespaces, even if
4859 they are inactive to clear inode references from the lock.
4861 Severity : enhancement
4862 Frequency : occasional, if non-standard max_dirty_mb used
4864 Description: Client will block write RPCs if not enough grant
4865 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
4866 then the client will block writes while waiting for another RPC
4867 to complete instead of consuming its dirty limit. With change
4868 we get improved performance when max_dirty_mb is small.
4870 Severity : enhancement
4871 Bugzilla : 3389, 6253
4872 Description: Add support for supplementary groups on the MDS.
4873 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
4874 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
4875 lookups for user supplementary groups into a cache.
4879 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
4880 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
4881 for liblustre O_EXCL is enforced by MDS.
4885 Description: readdir never returns NULL in liblustre.
4886 Details : Corrected llu_iop_getdirentries logic, to return offset of next
4887 dentry in struct dirent.
4891 Frequency : liblustre only, depends on application IO pattern
4892 Description: liblustre clients evicted if not contacting servers
4893 Details : Don't put liblustre clients into the ping_evictor list, so
4894 they will not be evicted by the pinger ever.
4896 Severity : enhancement
4898 Description: Add ability to evict clients by NID from MDS.
4899 Details : By echoing "nid:$NID" string into
4900 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
4901 $NID would be instantly evicted from this MDS and from all active
4902 OSTs connected to it.
4906 Description: Do not query file size twice, somewhat slowing stat(2) calls.
4907 Details : lookup_it_finish() used to query file size from OSTs that was not
4912 Description: service threads change working directory to that of init
4913 Details : Starting lustre service threads may pin the working directory
4914 of the parent thread, making that filesystem busy. Threads
4915 now change to the working directory of init to avoid this.
4919 Frequency : during shutdown only
4920 Description: shutdown with a failed MDS or OST can cause unmount to hang
4921 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
4926 Frequency : 2.6 only
4927 Description: chmod/chown may include an extra supplementary group
4928 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
4929 supplementary group and if none is specified this is used.
4932 Bugzilla : 5479 (6816)
4934 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
4935 Details : If lookup is in progress on a file that is unlinked we might try
4936 to revalidate the inode and fail in revalidate after lookup is
4937 complete and ll_file_open() enqueues the open again but
4938 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
4941 Frequency : always, if lconf --abort_recovery used
4943 Description: lconf --abort_recovery fails with 'Operation not supported'
4944 Details : lconf was attempting to abort recovery on the MDT device and not
4947 ------------------------------------------------------------------------------
4949 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
4954 Frequency : rare (only unsupported configurations with a node running as an
4956 Bugzilla : 6514, 5137
4957 Description: Mounting a Lustre file system on a node running as an OST could
4959 Details : OSTs now preallocates memory needed to write out data at
4960 startup, instead of when needed, to avoid having to
4961 allocate memory in possibly low memory situations.
4962 Specifically, if the file system is mounted on on OST,
4963 memory pressure could force it to try to write out data,
4964 which it needed to allocate memory to do. Due to the low
4965 memory, it would be unable to do so and the node would
4966 become unresponsive.
4968 Severity : enhancement
4970 Description: Addition of lconf --service command line option
4971 Details : lconf now accepts a '--service <arg>' option, which is
4972 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
4974 Severity : enhancement
4976 Description: Failover mode is now the default for OSTs.
4977 Details : By default, OSTs will now run in failover mode. To return to
4978 the old behaviour, add '--failout' to the lmc line for OSTs.
4980 Severity : enhancement
4982 Description: Health checks are now provided for MDS and OSTs
4983 Details : Additional detailed health check information on MSD and OSTs
4984 is now provided through the procfs health_check value.
4987 Frequency : occasional, depends on IO load
4989 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
4990 after numerous create/delete cycles
4991 Details : The ext3 inode allocation policy would not allocate new inodes
4992 very well on the OSTs because there are no new directories
4993 being created. Instead we look for groups with free space if
4994 the parent directories are nearly full.
4999 Description: Network or server problems during mount may cause partially
5000 mounted clients instead of returning an error.
5001 Details : The config llog parsing code may overwrite the error return
5002 code during mount error handling, returning success instead
5008 Description: MDS can fail to allocate large reply buffers
5009 Details : After long uptimes the MDS can fail to allocate large reply
5010 buffers (e.g. zconf client mount config records) due to memory
5011 fragmentation or consumption by the buffer cache. Preallocate
5012 some large reply buffers so that these replies can be sent even
5013 under memory pressure.
5017 Frequency : rare (liblustre)
5018 Description: fsx running with liblustre complained that using truncate() to
5019 extend the file doesn't work. This patch corrects that issue.
5020 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
5021 ATTR_SIZE and lsm use in llu_setattr_raw.
5025 Frequency : rare, only 2.6 kernels
5026 Description: Unusual file access patterns on the MDS may result in inode
5027 data being lost in very rare circumstances.
5028 Details : Bad interaction between the ea-in-inode patch and the "no-read"
5029 code in the 2.6 kernel caused the inode and/or EA data not to
5030 be read from disk, causing single-file corruption.
5034 Frequency : rare, only 2.6 filesystems using extents
5035 Description: Heavy concurrent write and delete load may cause data corruption.
5036 Details : It was possible under high-load situations to have an extent
5037 metadata block in the block device cache from a just-unlinked
5038 file overwrite a newly-allocated data block. We now unmap any
5039 metadata buffers that alias just-allocated data blocks.
5043 Frequency : filesystems with default stripe_count larger than 77
5044 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
5045 Details : lconf specifies an inode size of 4096 bytes when the default
5046 stripe_count is larger than 77. This conflicts with the default
5047 inode density of 1 per 4096 bytes. Allocate smaller inodes in
5048 this case to avoid pinning too much memory for large EAs.
5050 ------------------------------------------------------------------------------
5052 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
5057 Frequency : rare (extremely heavy IO load with hundreds of clients)
5059 Description: Client is evicted, gets IO error writing to file
5060 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
5061 caused clients to be evicted for AST timeouts. The fixes in
5062 bug 5192 mean we no longer need to have such short AST timeouts
5063 so ldlm_timeout has been increased.
5066 Frequency : occasional during --force or --failover shutdown under load
5067 Bugzilla : 5949, 4834
5068 Description: Server oops/LBUG if stopped with --force or --failover under load
5069 Details : a collection of import/export refcount and cleanup ordering
5070 issues fixed for safer force cleanup
5073 Frequency : only filesystems larger than 120 OSTs
5074 Bugzilla : 5990, 6223
5075 Description: lfs getstripe would oops on a very large filesystem
5076 Details : lov_getconfig used kfree on vmalloc'd memory
5079 Frequency : only filesystems exporting via NFS to Solaris 10 clients
5080 Bugzilla : 6242, 6243
5081 Description: reading from files that had been truncated to a non-zero size
5082 but never opened returned no data
5083 Details : ll_file_read() reads zeros from no-object files to EOF
5088 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
5089 Details : There's little harm in aborting MDS/OSS recovery and letting it
5090 try again, so I removed the LASSERT and return an error instead.
5092 Severity : enhancement
5094 Description: New debugging infrastructure for tracking down data corruption
5095 Details : The I/O checksum code was replaced to: (a) control it at runtime,
5096 (b) cover more of the client-side code path, and (c) try to narrow
5097 down where problems occurred
5101 Bugzilla : 3819, 4364, 4397, 6313
5102 Description: Racing close and eviction MDS could cause assertion in mds_close
5103 Details : It was possible to get multiple mfd references during close and
5104 client eviction, leading to one thread referencing a freed mfd.
5106 Severity: : enhancement
5107 Bugzilla : 3262, 6359
5108 Description: Attempts to reconnect to servers are now more aggressive.
5109 Details : This builds on the enhanced upcall-less recovery that was added
5110 in 1.4.2. When trying to reconnect to servers, clients will
5111 now try each server in the failover group every 10 seconds. By
5112 default, clients would previously try one server every 25 seconds.
5117 Description: After recovery, certain operations trigger a failed
5118 assertion on a client.
5119 Details : Failing over an mds, using lconf -d --failover, while a
5120 client was doing a readdir() call would cause the client to
5121 LBUG after recovery completed and the readdir() was resent.
5123 Severity : enhancement
5125 Description: Default groups are now added by lconf
5126 Details : You can now run lconf --group <servicename> without having to
5127 manually add groups with lmc.
5130 Frequency : occasional
5132 Description: Nodes with an elan id of 0 trigger a failed assertion
5135 Frequency : always when accessing e.g. tty/console device nodes
5137 Description: tty and some other devices nodes cannot be used on lustre
5138 Details : file's private_data field is used by device data and lustre
5139 values in there got lost. New field was added to struct file to
5140 store fs-specific private data.
5143 Frequency : when exporting Lustre via NFS
5145 Description: NFSD failed occasionally when looking up a path component
5146 Details : NFSD is looking up ".." which was broken in ext3 directories
5147 that had grown large enough to become hashed.
5150 Frequency : Clusters with multiple interfaces not on the same subnet
5152 Description: Nodes will repeatedly try to reconnect to an interface which it
5153 cannot reach and report an error to the log.
5154 Details : Extra peer list entries will be created by lconf with some peers
5155 unreachable. lconf now validates the peer before adding it.
5158 Frequency : Only if a default stripe is set on the filesystem root.
5160 Description: Setting a default stripe on the filesystem root prevented the
5161 filesystem from being remounted.
5162 Details : The client was sending extra request flags in the root getattr
5163 request and did not allocate a reply buffer for the dir EA.
5166 Frequency : occasional, higher if lots of files are accessed by one client
5167 Bugzilla : 6159, 6097
5168 Description: Client trips assertion regarding lsm mismatch/magic
5169 Details : While revalidating inodes the VFS looks up inodes with ifind()
5170 and in rare cases can find an inode that is being freed.
5171 The ll_test_inode() code will free the lsm during ifind()
5172 when it finds an existing inode and then the VFS later attaches
5173 this free lsm to a new inode.
5177 Bugzilla : 6422, 7030
5178 Description: MDS deadlock between mkdir and client eviction
5179 Details : Creating a new file via mkdir or mknod (starting a transaction
5180 and getting the ns lock) can deadlock with client eviction
5181 (gets ns lock and trying to finish a synchronous transaction).
5184 Frequency : occasional
5185 Description: While starting a server, the fsfilt_ext3 module could not be
5187 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
5188 kernels. Previously, lconf would still use the ext3 name
5189 when trying to load modules. Now, it will correctly use
5190 ext3 on 2.4 and ldiskfs on 2.6.
5192 Severity : enhancement
5193 Description: The default stripe count has been changed to 1
5194 Details : The interpretation of the default stripe count (0, to lfs
5195 or lmc) has been changed to mean striping across a single
5196 OST, rather than all available. For general usage we have
5197 found a stripe count of 1 or 2 works best.
5199 Severity : enhancement
5200 Description: Add support for compiling against Cray portals.
5201 Details : Conditional compiling for some areas that are different
5205 Frequency : occasional
5206 Bugzilla : 6409, 6834
5207 Description: Creating files with an explicit stripe count may lead to
5208 a failed assertion on the MDS
5209 Details : If some OSTs are full or unavailable, creating files may
5210 trigger a failed assertion on the MDS. Now, Lustre will
5211 try to use other servers or return an error to the
5215 Frequency : occasional
5217 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
5218 caused lock timeout during readahead (since 1.4.2).
5219 Details : Processes doing readahead might match a lock that hasn't been
5220 granted yet if there are overlapping and conflicting lock
5221 requests. The readahead process waits on ungranted lock
5222 (original lock is CBPENDING), while OST waits for that process
5223 to cancel CBPENDING read lock and eventually evicts client.
5225 Severity : enhancement
5227 Description: Initial enabling of flock support for clients
5228 Details : Implements fcntl advisory locking and file status functions.
5229 This feature is provided as an optional mount flag (default
5230 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
5231 locking are implemented yet, and those that are are not guaranteed
5232 to be completely correct in production environments.
5233 mount -t lustre -o [flock|noflock] ...
5236 Frequency : occasional
5238 Description: OSTs running 2.4 kernels but with extents enabled might trip an
5239 assertion in the ext3 JBD (journaling) layer.
5240 Details : The b_committed_data struct is protected by the big kernel lock
5241 in 2.4 kernels, serializing journal_commit_transaction() and
5242 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
5243 access to this struct. In 2.6 kernels there is finer grained
5244 locking to improve SMP performance of the JBD layer.
5248 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
5253 Description: merge and cleanup kernel patches.
5256 -----------------------------------------------------------------------------
5258 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
5260 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
5261 versions of Lustre. Please update all servers and clients to
5262 version 1.4.2 or later at the same time. You must also run
5263 "lconf --write-conf {config}.xml" on the MDS while it is stopped
5264 to update the configuration logs.
5266 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
5267 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
5268 - swab llog records of type '0' so we get proper header size/idx (5861)
5269 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
5270 - fix rename of one directory over another leaking an inode (5953)
5271 - avoid SetPageDirty on 2.6 (5981)
5272 - don't re-add just-being-destroyed locks to the waiting list (5653)
5273 - when creating new directories, inherit the parent's custom
5274 striping settings if present parent (3048)
5275 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
5276 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
5277 - don't hold client locks on temporary worklist from l_lru (5666)
5278 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
5279 - automatically evict dead clients (5921)
5280 - Update file size properly in create+truncate+fstat case (6196)
5281 - Do not unhash mountpoint dentries, do not allow removal of
5283 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
5284 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
5285 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
5286 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
5287 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
5288 - Correctly handle reads of files with no objects (6243)
5289 - lctl recover will also mark a device active if deactivate used (5933)
5291 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
5292 - allow --write-conf on an MDS with different nettype than client (5619)
5293 - don't write config llogs to MDS for mounts not from that MDS (5617)
5294 - lconf should create multiple TCP connections from a client (5201)
5295 - init scripts are now turned off by default; run chkconfig --on
5296 lustre and chkconfig --on lustrefs to use them
5297 - upcalls are no longer needed for clients to recover to failover
5299 - add --abort-recovery option to lconf to abort recovery on device
5301 - add support for an arbitrary number of OSTs (3026)
5302 - Quota support protocol changes.
5303 - forward compatibility changes to wire structs (6007)
5304 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5305 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5306 - improved "lustre status" script
5307 - initialize blocksize for non-regular files (6062)
5308 - added --disable-server and --disable-client configure options (5782)
5309 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5310 - Vanilla 2.4.29 support
5311 - increase maximum number of obd devices to 520 (6242)
5312 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5313 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5314 SLES 9 2.6.5 kernels
5315 - limit stripes per file to 160 (the maximum EA size) (6093)
5317 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5320 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5321 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5322 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5323 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5324 - small ext3 extents cleanups and fixes (5733)
5325 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5326 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5327 - avoid needless client->OST connect, fix handle mismatch (5317)
5328 - fix DLM error path that led to out-of-sync client, long delays (5779)
5329 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5330 - fix several locking issues related to i_size (5492,5624,5654,5672)
5331 - don't move pending lock onto export if it is already evicted (5683)
5332 - fix kernel oops when creating .foo in unlinked directory (5548)
5333 - fix deadlock in obdfilter statistics vs. object create (5811)
5334 - use time_{before,after} to avoid timer jiffies wrap (5882)
5335 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5336 - Do not leak request if server was not able to process it (5154)
5337 - If mds_open unable to find parent dir, make that negative lookup(5154)
5338 - don't create new directories with extent-mapping (5909, 5936)
5340 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5341 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5342 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5343 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5344 - support for RHEL4 kernel on i686 (5773)
5345 - provide error messages when incompatible logs are encountered (5898)
5347 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5348 * version 1.4.0.10 (1.4.1 release candidate 1)
5350 - don't keep a lock reference when lock is not granted (4238)
5351 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5352 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5353 - drop import inflight refcount on signal_completed_replay error (5255)
5354 - unlock page after async write error during send (3677)
5355 - handle missing objects in filter_preprw_read properly (5265)
5356 - no transno return for symlink open, don't save no-trasno open (3440)
5357 - don't try to complete elan receive that already failed (4012)
5358 - free RPC server reply state on error (5406)
5359 - clean up thread from ptlrpc_start_thread() on error (5160)
5360 - readahead could read extra page into cache that wasn't ejected (5388)
5361 - prevent races in class_attach/setup/cleanup/detach (5260)
5362 - don't dereference de->d_inode after l_dput of de (5458)
5363 - use "int" for stripe value returned from lock_to_stripe (5544)
5364 - mballoc allocation and error-checking fixes in 2.6 (5504)
5365 - block device patches to fix I/O request sizes in 2.6 (5482)
5366 - look up hostnames for IB nals (5602)
5367 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5368 - don't start multiple acceptors for the same port (5277)
5369 - fix incorrect LASSERT in mds_getattr_name (5635)
5370 - export a proc file for general "ping" checking (5628)
5371 - fix "lfs check" to not block when the MDS is down (5628)
5373 - service request history (4965)
5374 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5375 - create an "evict_client" /proc entry on OSTs, like the MDS has
5376 - fix mount usage message, return errors per mount(8) (5168)
5377 - change grep [] to grep "[]" in tests so they work in more UMLs
5378 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5379 - remove ancient LOV_MAGIC_V0 EA support (5047)
5380 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5381 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5382 - allow readahead allocations to fail when low on memory (5383)
5383 - mmap locking landed again, after considerable improvement (2828)
5384 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
5386 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
5389 - send OST transaction number in read/write reply to free req (4966)
5390 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5391 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5392 - the watchdog thread now runs as interruptible (5246)
5393 - flock/lockf fixes (but it's still disabled, pending 5135)
5394 - don't use EXT3 constants in llite code (5094)
5395 - memory shortage at startup could cause assertion (5176)
5397 - reorganization of lov code
5398 - single portals codebase
5400 - add extents/mballoc support (5025)
5401 - direct I/O reads in the obdfilter (4048)
5402 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
5404 tbd Cluster File Systems, Inc. <info@clusterfs.com>
5407 - send OST transaction number in read/write reply to free req (4966)
5408 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5409 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5410 - the watchdog thread now runs as interruptible (5246)
5411 - handle missing objects in filter_preprw_read properly (5265)
5412 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5413 - drop import inflight refcount on signal_completed_replay error (5255)
5414 - unlock page after async write error during send (3677)
5415 - return original error code on reconstructed replies (3761)
5416 - no transno return for symlink open, don't save no-trasno open (3440)
5418 - add pid to ldlm debugging output (4922)
5419 - bump the watchdog timeouts -- we can't handle 30sec yet
5420 - extra debugging for orphan dentry/inode bug (5259)
5422 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
5425 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
5426 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
5427 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
5428 - fix NULL dereference in /proc/sys/portals/routes (4827)
5429 - allow failed mdc_close() operations to be interrupted (4561)
5430 - stop precreate on OST before MDS would time out on it (4778)
5431 - don't send partial-page writes before EOF from client (4410)
5432 - discard client grant for sub-page writes on large-page clients (4520)
5433 - don't free dentries not owned by NFS code, check generation (4806)
5434 - fix lsm leak if mds_create_objects() fails (4801)
5435 - limit debug_daemon file size, always print CERROR messages (4789)
5436 - use transno after validating reply (3892)
5437 - process timed out requests if import state changes (3754)
5438 - update mtime on OST during writes, return in glimpse (4829)
5439 - add mkfsoptions to LDAP (4679)
5440 - use ->max_readahead method instead of zapping global ra (5039)
5441 - don't interrupt __l_wait_event() during strace
5443 - add software watchdogs to catch hung threads quickly (4941)
5444 - make lustrefs init script start after nfs is mounted
5445 - fix CWARN/ERROR duplication (4930)
5446 - return async write errors to application if possible (2248)
5447 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
5448 - print NAL number in %x format (4645)
5449 - update barely-supported suse-2.4.21-171 series (4842)
5450 - support for sles 9 %post scripts
5451 - support for building 2.6 kernel-source packages
5452 - support for sles km_* packages
5454 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
5457 - ignore -ENOENT errors in osc_destroy (3639)
5458 - notify osc create thread that OSC is being cleaned up (4600)
5459 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
5460 - reconstruct ost_handle() like mds_handle() (4657)
5461 - create a new thread to do import eviction to avoid deadlock (3969)
5462 - let lconf resolve symlinked-to devices (4629)
5463 - don't unlink "objects" from directory with default EA (4554)
5464 - hold socknal file ref over connect in case target is down (4394)
5465 - allow more than 32000 subdirectories in a single directory (3244)
5466 - fix blocks count for O_DIRECT writes (3751)
5467 - OST returns ENOSPC from object create when no space left (4539)
5468 - don't send truncate RPC if file size isn't changing (4410)
5469 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
5470 - bind to privileged port in socknal and tcpnal (3689)
5472 - rate limit CERROR/CWARN console message to avoid overload (4519)
5473 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
5474 - basic mmap support (3918)
5475 - kernel patch series update from b1_4 (4711)
5477 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
5480 - avoid crash during MDS cleanup with OST shut down (2775)
5481 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
5482 - don't use bad inodes on the MDS (3744)
5483 - dynamic object preallocation to improve recovery speed (4236)
5484 - don't hold spinlock over lock dumping or change debug flags (4401)
5485 - don't zero obd_dev when it is force cleaned (3651)
5486 - print grants to console if they go negative (4431)
5487 - "lctl deactivate" will stop automatic recovery attempts (3406)
5488 - look for existing locks in ldlm_handle_enqueue() (3764)
5489 - don't resolve lock handle twice in recovery avoiding race (4401)
5490 - revalidate should check working dir is a directory (4134)
5492 - don't always mark "slow" obdfilter messages as errors (4418)
5494 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
5497 - don't close LustreDB during write_conf until it is done (3860)
5498 - fix typo in lconf for_each_profile (3821)
5499 - allow dumping logs from multiple threads at one time (3820)
5500 - don't allow multiple threads in OSC recovery (3812)
5501 - fix debug_size parameters (3864)
5502 - fix mds_postrecov to initialize import for llog ctxt (3121)
5503 - replace config semaphore with spinlock (3306)
5504 - be sure to send a reply for a CANCEL rpc with bad export (3863)
5505 - don't allow enqueue to complete on a destroyed export (3822)
5506 - down write_lock before checking llog header bitmap (3825)
5507 - recover from lock replay timeout (3764)
5508 - up llog sem before sending rpc (3652)
5509 - reduce ns lock hold times when setting kms (3267)
5510 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
5511 - fix NULL deref and obd_dev leak on setup error (3312)
5512 - replace some LBUG about llog ops with error handling (3841)
5513 - don't match INVALID dentries from d_lookup and spin (3784)
5514 - hold dcache_lock while marking dentries INVALID and hashing (4255)
5515 - fix invalid assertion in ptlrpc_set_wait (3880)
5517 - add libwrap support for the TCP acceptor (3996)
5518 - add /proc/sys/portals/routes for non-root route listing (3994)
5519 - allow setting MDS UUID in .xml (2580)
5520 - print the stack of a process that LBUGs (4228)
5522 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
5525 - don't cleanup request in ll_file_open() on failed MDS open (3430)
5526 - make sure to unset replay flag from failed open requests (3440)
5527 - if default stripe count is 0, use OST count for inode size (3636)
5528 - update parent mtime/ctime on client for create/unlink (2611)
5529 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
5530 - free recovery state on server during a forced cleanup (3571)
5531 - unregister_reply for resent reqs (3063)
5532 - loop back devices mounting and status check on 2.6 (3563)
5533 - fix resource-creation race that can provoke i_size == 0 (3513)
5534 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
5535 - more debugging for page-accounting assertion (3746)
5536 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
5537 - avoid deadlock after precreation failure (3758)
5538 - fix race and lock order deadlock in orphan handling (3450, 3750)
5539 - add validity checks when grabbing inodes from l_ast_data (3599)
5541 - add /proc/.../recovery_status to obdfilter (3428)
5542 - lightweight CDEBUG infrastructure, debug daemon (3668)
5543 - change default OSC RPC parameters to be better on small clusters
5544 - turn off OST read cache for files smaller than 32MB
5545 - install man pages and include them in rpms (3100)
5546 - add new init script for (un)mounting lustre filesystems (2593)
5547 - run chkconfig in %post for init scripts (3701)
5548 - drop scimac NAL (unmaintained)
5550 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
5553 - clean kiobufs before and after use (3485)
5554 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
5555 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
5556 - change init script to fail more clearly if not run as root (1528)
5557 - allow clients to reconnect during replay (1742)
5558 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
5559 - don't do DNS lookups on NIDs too small for IP addresses (3442)
5560 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
5561 - fix cond_resched (3554)
5562 - only evict unfinished clients after recovery (3515)
5563 - allow bulk resend, prevent data loss (3570)
5564 - dynamic ptlrpc request buffer allocation (2102)
5565 - don't allow unlinking open directory if it isn't empty (2904)
5566 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
5567 - remove extraneous obd dereference causing LASSERT failure (3334)
5568 - don't use get_cycles() when creating temp. files on the mds (3156)
5569 - hold i_sem when setting i_size in ll_extent_lock() (3564)
5570 - handle EEXIST for set-stripe, set proper directory name (3336)
5572 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
5573 - fix ksocknal_fmb_callback() error messages (2918)
5575 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
5578 - don't copy lvb into (possibly NULL) reply on error (2983)
5579 - don't deref dentry after dput, don't free lvb on error (2922)
5580 - use the kms to determine writeback rpc length (2947)
5581 - increment oti_logcookies when osc is inactive (2948)
5582 - update client's i_blocks count via lvb messages (2543)
5583 - handle intent open/close of special files properly (1557)
5584 - mount MDS with errors=remount-ro, like obdfilter (2009)
5585 - initialize lock handle to avoid ASSERT on error cleanup (3057)
5586 - don't use cancelling-locks' kms values (2947)
5587 - use highest lock extent for kms, not last one (2925)
5588 - don't dereference ERR_PTR() dentry in error handling path (3107)
5589 - fix thread race in portals_debug_dumplog() (3122)
5590 - create lprocfs device entries at setup instead of at attach (1519)
5591 - common AST error handler, don't evict client on completion race (3145)
5592 - zero nameidata in detach_mnt in 2.6 (3118)
5593 - verify d_inode after revalidate_special is valid in 2.6 (3116)
5594 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
5595 - initialize RPC timeout timer earlier for 2.6 (3219)
5596 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
5597 - print nal/nid for unknown nid (3258)
5598 - additional checks for oscc recovery before doing precreate (3284)
5599 - fix ll_extent_lock() error return code for 64-bit systems (3043)
5600 - don't crash in mdc_close for bad permissions on open (3285)
5601 - zero i_rdev for non-device files (3147)
5602 - clear page->private before handing to FS, better assertion (3119)
5603 - tune the read pipeline (3236)
5604 - fix incorrect decref of invalidated dentry (2350)
5605 - provide read-ahead stats and refine rpc in flight stats (3328)
5606 - don't hold journal transaction open across create RPC (3313)
5607 - update atime on MDS at close time (3265)
5608 - close LDAP connection when recovering to avoid server load (3315)
5609 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
5610 - don't leak open file on MDS after open resend (3325)
5611 - serialize filter_precreate and filter_destroy_precreated (3329)
5612 - loop device shouldn't call sync_dev() for nul device (3092)
5613 - clear page cache after eviction (2766)
5614 - resynchronize MDS->OST in background (2824)
5615 - refuse to mount the same filesystem twice on same mountpoint (3394)
5616 - allow llmount to create routes for mounting behind routers (3320)
5617 - push lock cancellation to blocking thread for glimpse ASTs (3409)
5618 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
5619 - fix rare problem with rename on htree directories (3417)
5621 - allow default OST striping configuration per directory (1414)
5622 - fix compilation for qswnal for 2.6 kernels (3125)
5623 - increase maximum number of MDS request buffers for large systems
5624 - change liblustreapi to be useful for external progs like lfsck (3098)
5625 - increase local configuration timeout for slow disks (3353)
5626 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
5628 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5631 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
5632 - don't overwrite extent policy data in reply if lock was blocked (2901)
5633 - drop filter export grants atomically with removal from device (2663)
5634 - del obd_self_export from work_list in class_disconnect_exports (2908)
5635 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
5636 - swab reply message in mdc_close, other PPC fixes (2464)
5637 - fix destroying of named logs (2325)
5638 - overwrite old logs when running lconf --write_conf (2264)
5639 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
5640 - fix race in target_handle_connect (2898)
5641 - mds_reint_create() should take same inode create lock (2926)
5642 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
5643 - don't close files for self_export to avoid uninitialized obd (2936)
5644 - allow MDS with the same name as client node (2939)
5645 - hold dentry reference for closed log files for unlink (2325)
5646 - reserve space for all logs during transactions (2059)
5647 - don't evict page beyond end of stripe extent (2925)
5648 - don't oops on a deleted current working directory (2399)
5649 - handle hard links to targets without a parent properly (2517)
5650 - don't dereference NULL lock when racing during eviction (2867)
5651 - don't grow lock extents when lots of conflicting locks (2919)
5653 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
5656 - account for cache space usage on clients to avoid data loss (974)
5657 - lfsck support in lustre kernel code (2349)
5658 - reduce journal credits needed for BRW writes (2370)
5659 - orphan handling to avoid losing space on client/server crashes
5660 - ptlrpcd can be blocked, stopping ALL progress (2477)
5661 - use lock value blocks to assist in proper KMS, faster stat (1021)
5662 - takes i_sem instead of DLM locks internally on obdfilter (2720)
5663 - recovery for initial connections (2355)
5664 - fixes for mds_cleanup_orphans (1934)
5665 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
5666 - block all file creations until orphan recovery completes (1901)
5667 - client remove rq_connection from request struct (2423)
5668 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
5669 - recovery timer race (2670)
5670 - mdc_close recovey bug (2532)
5671 - ptlrpc cleanup bug (2710)
5672 - mds timeout on local locks (2588)
5673 - namespace lock held during RPCs (2431)
5674 - handle interrupted sync write properly (2503)
5675 - don't try to handle a message that hasn't been replied to (2699)
5676 - client assert failure during cleanup after abort recovery (2701)
5677 - leak mdc device after failed mount (2712)
5678 - ptlrpc_check_set allows timedout requests to complete (2714)
5679 - wait for inflight reqs when ptlrpcd finishes (2710)
5680 - make sure unregistered services are removed from the srv_list
5681 - reset bulk XID's when resending them (caught by 1138 test)
5682 - unregister_bulk after timeout
5683 - fix lconf error (2694)
5684 - handle write after unfinished setstripe, stripe-only getstripe (2388)
5685 - readahead locks pages, leaves pending causing memory pressure (2673)
5686 - increase OST request buffers to 4096 on large machines (2729)
5687 - fix up permission of existing directories in simple_mkdir (2661)
5688 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
5689 - don't assign transno to errored transactions (2742)
5690 - don't delete objects on OST if given a bogus objid from MDS (2751)
5691 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
5692 - if rq_no_resend, then timeout request after recovery (2432)
5693 - fix MDS llog_logid record size, 64-bit array alignment (2733)
5694 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
5695 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
5696 - MDS recovery shouldn't delete objects that it has given out (2730)
5697 - if enqueue arrives after completion, don't clobber LVB (2819)
5698 - don't unlock pages twice when trigger_group_io returns error (2814)
5699 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
5700 - don't write pages to disk if there was an error (1450)
5701 - don't ping imports that have recovery disabled (2676)
5702 - take buffered bytes into account when balancing socknal conn (2817)
5703 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
5704 - reconnect unlink llog connection after MDS reconnects to OST (2816)
5705 - remove little-endian swabbing of llog records (1987)
5706 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
5707 - retry reposting request buffers if they fail (1191)
5708 - grow extent at grant time to avoid granting a revoked lock (2809)
5709 - lock revoke doesn't evict page if covered by a second lock (2765)
5710 - disable VM readahead to avoid reading outside lock extents (2805)
5712 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
5713 - updated kernel patches for hp-2.4.20 kernel (2681)
5715 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
5718 - fix truncated write corruption (2366)
5719 - fix for failed assertion in iopen_connect_dentry (1792,2517)
5721 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
5722 - ensure len(uuid) < 37 in lmc (1171)
5723 - fix ia64 OOPS in llog_test (2255)
5724 - zero end of page at obdfilter for partial page writes (2648)
5725 - don't leave stale dentries around after renames (bug 2428)
5726 - fix timeouts when evicting a client with a single lock held (2642)
5727 - set deadline for the initial HELLO message to drain (2634)
5728 - print out dotted-quad IP addresses in the socknal (2302)
5730 - additional debugging for MDS client eviction problem (2443)
5731 - fix mkfsoptions support for osts (2603, 2604)
5733 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
5736 - add series for the vanilla 2.6.0 kernel
5737 - add series for the vanilla 2.4.24 kernel
5738 - add series for a cray x86/64 UL kernel drop
5739 - fix xattr patches for the vanilla 2.4.19 series
5741 - generate true UUIDs in lmc (1171)
5742 - have portals stack dumping break in UML (2466)
5743 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
5744 - allocate with _NOFS in ldlm to avoid deadlock (1933)
5745 - wake callback waiting threads on client eviction (2460)
5746 - Add --ptldebug and --subsystem to lmc (1719)
5747 - update assertion to allow safe interrupt allocation
5748 - set rq_no_resend for cancel requests (2432)
5749 - recalculate ptlrpcd timeout after resend (2494)
5750 - call vfs_rmdir when removing pending directories (2368)
5751 - fix renaming a file to itself (2429)
5752 - lmc creates a default one-stripe lov (2454)
5753 - expand procfs space to handle large clusters (2326)
5754 - increase UML stack to avoid overflow
5755 - update lconf's list of debug and subsystem masks
5756 - fix lfs find --obd (2510)
5757 - /proc tunable for disabling filter read caching (2591)
5758 - stop rpm packages from altering slapd.conf (2301)
5759 - disable nagle in the socknal under 0conf (2578)
5760 - choose mds inode size based on stripe count (2572)
5761 - fix kernel-source rpm problems (2516)
5763 - add --disable-doc to avoid pdf generation (2421)
5764 - update documentation, tests, type-os, comments
5765 - avoid format warnings on ia64
5766 - remove the TOE NAL
5767 - tiny code cleanups by removing unused fields
5769 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
5772 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
5773 - remove the most blatant lies from BUILDING file (2371)
5774 - change default debug level to reasonable production setting
5775 - reduce client side cache size to reduce cache flush time
5776 - reduce max RPCs in flight to avoid unnecessary file fragmentation
5777 - make TCP zerocopy and pinger support enabled by default (2476)
5778 - sync writes completed after process exits caused crashes (2319)
5779 - maintain correct mount count on the MDS (2356)
5780 - backout 1557, because 2316 wasn't really fixed
5781 - better file I/O statistics gathering in /proc
5782 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
5783 - another kernel patch to fix zero-copy TCP function export
5784 - don't take duplicate lock when processing re-sent getattr (2420)
5785 - lctl uses obd_self_export instead of creating new conn (2353)
5786 - MDS/OST recovery case which requires object creation asserted (2425)
5787 - move lfs from /usr/sbin to /usr/bin in packages
5788 - fix race between mds_client_add and mds_client_free (2417)
5789 - use kmalloc instead of slabs in portals (2430)
5790 - don't create duplicate records when a failover MDS is present (2442)
5791 - remove unnecessary mount age check (2332)
5792 - don't remove directory inodes from locks prematurely (2451)
5793 - don't break if MDS service name is the same as hostname (2103)
5794 - fix races in client write RPC generation when cache full (2482)
5796 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
5799 - remove now-unused request->rq_obd (278)
5800 - if an allocation fails, print out how much memory we've used (1933)
5801 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
5802 - add the "configurable stack size" patch to most series files (1256)
5803 - ability to write large log records, for 100+ OST configs (2306)
5804 - fix NULL deref when filter_prep fails (2314)
5805 - fix operator precedence error in filter_sync
5806 - dynamic allocation of socknal TX descriptors (2315)
5807 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
5808 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
5809 - ensure that CWARN messages in Portals always get to the syslog
5810 - __init/__exit are not for prototype decls (ldlm_init/exit)
5811 - x86-64 compile warning fixes
5812 - fix gateway LMC keyword conflict (2318)
5813 - fix MDS lock inversions in getattr/reint paths (1844)
5814 - fix a rare lock re-ordering bug, which caused deadlock (2322)
5815 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
5816 - DLM race condition prevented some lock evictions (2328)
5817 - ENOMEM detection and retry on socknal sends (2230)
5818 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
5819 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
5820 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
5821 - better error messages when a client is rejected during recovery (1505)
5822 - avoid cancelling locks which were never granted, after failure (2330)
5823 - fix i_sem/journal inversion in mds_client_add (2333)
5824 - fix truncate/getattr lock cycle deadlock (2334)
5825 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
5826 - fix two rare exit paths which could leak an l_lock() ref (2321)
5827 - fencepost error in MDS/OST orphan recovery (2226)
5828 - make log record alignment 8 bytes (1988)
5829 - lstripe now fails when requested offset > ost_count (2237)
5830 - ensure that all kernel series have a complete list.h (1607)
5831 - fix crashes in special-file operations (2316)
5832 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
5834 - allow configurable automake binary, for testing new versions
5835 - small update to the lfs documentation
5837 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
5839 * fix negative export reference count in fsfilt_sync (2312)
5841 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
5842 * release candidate 0.9.1
5844 - orphans are moved into the PENDING directory for possible recovery
5845 - replayed opens now open by fid for orphan/rename safety (1042)
5846 - last close of an orphan inode generates a transno (683)
5847 - chdir() and mount() now pin the directory entry (1020)
5848 - avoid CERROR in normal ll_setattr_raw() error case (1500)
5849 - discard very old requests without processing them (1502)
5850 - remove some common, well-understood CERRORs (1505)
5851 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
5852 - clear "grant" flags in OST replies until OST grant code lands (1644)
5853 - fix read performance by not clobbering i_blksize on client (1598)
5854 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
5855 - make LRU size a /proc tunable, clears locks when reduced (707)
5856 - fix some lprocfs rot that prevented ptlbd from loading (1732)
5857 - server locks take references on exports now (1558)
5858 - build fixes for 2.4.20-rh trees (1663)
5859 - return an error from lov_create if all OSCs are inactive (1751)
5860 - fix import levels when a reconnect happens without a timeout (1597)
5861 - exit early from mds_open if we get a lookup error (1749)
5862 - partial page read at EOF wouldn't wait for disk before sending (1642)
5863 - avoid NULL deref in obdfilter when reading page past EOF (1592)
5864 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
5865 - fix LBUG in ll_it_open_error with rc = -2 (1861)
5866 - write/truncate lock inversion (1639)
5867 - Don't auto-load obdclass, portals modules during cleanup (1495)
5868 - fix timestamps from jumping to "now" (1763)
5869 - extra journal assertions (1648)
5870 - add an extra multiunlink test (1771)
5871 - fix read_record/write_record API (1776)
5872 - fix leak of offset_extent, possible incorrect i_size later (1772)
5873 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
5874 - Debugging for the kqswnal_get_idle_tx problems (1820)
5875 - Allow recovery to be attempted multiple times (1536)
5876 - Write out MDS last_rcvd file after it is first created (1600)
5877 - Fix tx_descriptor leak in failed transmit situations (1827)
5878 - ext3 journaling fixes for assertion failure after IO error (1871)
5879 - class_export_put() on freed export after completion AST error (1896)
5880 - Fix revalidate looping in VFS (1322)
5881 - Don't access a freed export during MDS_REINT timeout (1521)
5882 - Add open-unlink recovery support on the MDS (1673,1764)
5883 - Return an error if no MDS data was read from last_rcvd (1946)
5884 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
5885 - Don't setup a disk device that doesn't match exported UUID (317)
5886 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
5887 - avoid committing NULL handle in force close
5888 - local.sh is now a one-stripe LOV configuration
5889 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
5890 - fix ext3 htree duplicate directory entry corruption (1516)
5891 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
5892 - update to new LOV EA format (2097)
5893 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
5895 2003-06-15 Phil Schwan <phil@clusterfs.com>
5898 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
5899 - per-import/export recovery handling (958, 931, 959)
5900 - multiple last-rcvd slots, for serving multiple FSes (949)
5901 - connections are again shared between multiple imp/exports (963, 964)
5902 - "umount -f" would hang if any requests needed to be sent (393, 978)
5903 - avoid pinning large req buffer by copying for queued messages (989)
5904 - add "uuid" to "lctl device" command to help upcalls (991)
5905 - "open" RPCs with transnos would confuse recovery counters (1037)
5906 - do proper endian conversion of all wire messages (288, 340, 891)
5907 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
5908 - call waiting locks callback from LDLM recovery thread (1127, 1151)
5909 - fix ptlrpc_connection leak in target_handle_connect (1174)
5910 - fix import refcounting bug in OST and MDS cleanup (1134)
5911 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
5912 - fix very unlikely obd_types race condition (501)
5913 - remove osc_open hack for echo_client (1187)
5914 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
5915 - a failure in read_inode2 leads to deadlock (1139)
5916 - cancel ack-locks as soon as transaction is committed (1072)
5917 - fix major leaks and crashes in the bulk I/O path (937, 1057)
5918 - make sure to commitrw after any preprw to avoid deadlock (1162)
5919 - failing to execute a file in a lustre FS would lock inode (1203)
5920 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
5921 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
5922 - obd_uuid2tgt would walk off the end of the list (1255)
5923 - on IA64 the osc would give portals incorrect bulk size (1258)
5924 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
5925 - fix lock inversion caused by new llite matching code (1282)
5926 - limit the number of dirty pages on a client to 10MB (1286)
5927 - timed out locks were not being corrected cancelled (1289)
5928 - fix O_DIRECT above 4GB on IA-32 (1292)
5929 * major user-visible changes
5930 - fail out/fail over policy now controlled by the upcall (993)
5932 - add OBD_PING to check server availability and failure (954)
5933 - lustre messages are now sent in sending host order (288, 340, 891)
5934 - add eadatalen to MDS getattr reply (340)
5935 - OST read replies may contain second buffer, with per-page status (593)
5937 2003-03-11 Phil Schwan <phil@clusterfs.com>
5940 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
5941 - failed open()s could cause deadlock; fixed (867, 869)
5942 - stop cancelling OST locks when files are closed (481)
5943 - overlapping XID spaces caused network corruption (851, 853)
5944 - fix unsafe fsfilt counter arithmetic; change to atomic_t
5945 - setattr_raw added, to do single-RPC, server-side setattrs
5946 - lmc/lconf syntax change for OST UUIDs
5947 - fix crashy race condition between ptlrpc_free_req and osc_close
5948 - don't use request in mdc_enqueue if we hit a timeout (889)
5949 - don't set the inode i_size for regular files from the MDS (896)
5950 - handle out of order completion AST (842)
5951 - don't LBUG if a lock request times out after receiving AST (913)
5952 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
5953 - if a bad lock AST arrives, send an error instead of dropping entirely
5954 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
5955 - fix leak in bulk IO when only partially completed (899, 900, 926)
5956 - fix O_DIRECT for ia64 (55)
5957 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
5958 - C-z after timeout could hang a process forever; fixed (977)
5960 - client-side I/O cache (678, 924, 929, 941, 970)
5962 - READPAGE and SETATTRs which don't take server-side locks get
5965 2003-02-11 Phil Schwan <phil@clusterfs.com>
5968 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
5969 - Fix fsfilt_extN_readpage() to read a full page of directory
5970 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
5971 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
5972 - Fix imbalanced LOV object allocation and out-of-bound access (469)
5973 - Most intent operations were removed, in favour of a new RPC mode
5974 that does a single RPC to the server and bypasses most of the VFS
5975 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
5976 - Aggressively cancel local locks on DLM servers
5977 - mds_reint_unlink sends EA to the client if it's the last nlink.
5978 client uses that EA to unlink OST objects.
5979 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
5980 - recursive symlinks were fixed (439)
5981 - fixed NULL deref in DEBUG_REQ
5982 - filter_update_lastobjid no longer calls sync, which annoyed extN
5983 - fixed multi-client small-writes to a single file problem (445)
5984 - fixed mtime updates during file writes (607)
5985 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
5986 - fixed bug in obd_brw_read/write() (under guise of testing 367)
5987 - fixed Linux OST size reporting problem (444, 656)
5988 - OST now updates object mtime with writes or setattr (607, 619)
5989 - client verifies file size before zeroing page past EOF (445)
5990 - OST now writes last allocated objid to disk with allocation (108)
5991 - LOV on echo now works (409)
5993 - mds_reint_unlink sends a new buffer, with the EA included. this
5994 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
5995 set if a regular file was being unlinked, and it was the last link
5996 - use PtlGet from the target for bulk writes (315)
5997 - OST now updates object mtime with writes or setattr (607, 619)
5998 - LDLM now has a grant-time callback to revalidate locked items, if
6000 - Many MDS operations were reorganized to combat race conditions
6002 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
6003 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
6005 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
6008 - Fully reactivate OST imports after reconnection (512, others)
6009 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
6010 - More graceful error handling for truncating on dead OST (515)
6011 - Don't error out unless we're actually accessing dead stripes (474)
6012 - Fix garbage sizes when stripes are missing (410)
6013 - LRU counters were broken, causing constant lock purge (433, 432)
6014 - garbage on read from stripes with failed OSTs (441)
6015 - mark OSCs as active before reconnecting during recovery (438)
6016 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
6017 - lfind did not preserve OST order in output (443)
6018 - symlinks cause hung clients, incorrect data (439)
6019 - stop dereferencing request after dropping refcount (457)
6020 - don't LASSERT(spin_is_locked) on non-SMP (455)
6021 - fixes for many rename() bugs
6022 - fstat didn't correctly synchronize attributes (399)
6023 - server must handle lock cancellation during blocking AST prep (487)
6024 - bulk descriptors were free()d too soon (511)
6025 - fix paths in lconf, which would load incorrect modules (451, 507)
6026 - fix confusing lconf 'host not found' error message (386)
6027 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
6028 - fix race condition in mdc_blocking_ast() for inode access (526)
6029 - fix lov_unpackmd() unpacking wrong number of stripes (537)
6030 - fix lov_set_osc_active() marking wrong OSC inactive (440)
6031 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
6032 - fix multiple writes of stripe MD to MDS (358, maybe 519)
6033 - fix lstripe in several ways (kernel side) (527)
6034 - fix request leak in ldlm_cli_enqueue (262)
6035 - incorrect OSC was marked inactive after OST failure
6036 - call mds_fs_cleanup before unmounting filesystem (524)
6037 - fix races between taking ns_lock and ldlm_lock_change_resource
6038 - fix races updating LOV export open file list
6039 - fix lov_enqueue error path, avoid decref-ing bad lock handle
6040 - fix recovery NULL deref in ldlm_cli_cancel_unused
6041 - fix some DLM races by using new hash table for lock handles (419)
6042 - permit the client to specify desired inodes, at replay
6043 - duplicate requests when we queue them for replay reintegration
6044 - fix last_rcvd offset calculation
6045 - sync after each recovered transaction, so we always make progress
6046 - never, not always, ERESTART requests without transnos
6047 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
6049 - skip replay if the MDS says that the client is already connected
6050 - don't check for a recovery-enabled export to match lctl's UUID
6051 - don't INC_USE_COUNT for phantom exports
6052 - don't crash when cleaning up phantom exports (567)
6053 - don't double-finish or set replay data for errored mdc_open requests
6054 - abort requests when they time out, so we don't get old replies
6055 - send/receive replies for AST messages again
6056 - if the client says that it doesn't have the lock, cancel it on the
6058 - if we timeout during I/O, don't try to cancel an in-use lock; instead
6059 mark it as destroyed, it will all work out when decref is called
6060 - fix module use counts (22, 581)
6062 - ASTs now expect a reply (server cancels lock on error reply)
6064 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
6067 - fix many simultaneous client startup (392)
6068 - fix dentry->d_it clobbering
6069 - credentials weren't being shipped for readdir/getattr operations
6070 - remove invalid assertions triggered during some concurrent MD
6072 - proper Lustre versions added (336, 389)
6073 - fix memory leak for create error case (398)
6074 - fix LOV locking bug that would get cli/srv out of sync
6075 - fix echo client over LOV (409)
6076 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
6077 - fix double-O_EXCL intent crash (424)
6078 - avoid sending multiple lock CANCELs (352)
6080 - MDS can do multi-client recovery (modulo bugs in new code)
6082 - many updates, edits, cleanups
6084 2002-11-18 Phil Schwan <phil@clusterfs.com>
6087 - fix null d_it dereference (346)
6088 - fix full OST/dbench hang (333)
6089 - fix permission problem with file removal (286)
6090 - fix removal of OSCs from LOV when they fail
6091 - fix NULL deref during bulk timeout (214)
6092 - fix problems related to multiple filesystems on one MDS (241)
6093 - fixed serious subtle metadata locking bugs
6094 - free locks on clients when inodes are removed due to memory
6096 - fix inode pointer in lock data (285)
6097 - partial support for multiple MDS on a single host (241)
6098 - data locks weren't cancelled at clear_inode time (290, 311)
6099 - intent locks could lead to unbounded lock growth (205)
6100 - added a maximum lock count, an LRU list, and a flusher
6101 - fix multiple rename (365)
6102 - properly abstracted the echo client
6103 - OSC locked 1 byte too many; fixed
6104 - rewrote brw callback code:
6105 - fixed recovery bugs related to LOVs (306)
6106 - fixed too-many-pages-in-one-write crash (191)
6107 - fixed (again) crash in sync_io_timeout (214)
6108 - probably fixed callback-related race (385)
6110 - Add capability to MDS protocol
6111 - LDLM cancellations and callbacks on different portals
6113 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
6116 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
6117 - defer open object destruction to close time (601981, 138)
6118 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
6119 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
6120 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
6121 - don't call obd_setattr in ll_file_release for destroyed objects
6122 * protocol change to lustre_msg: move |version| and add |flags|
6123 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
6124 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
6125 * added replay of create, unlink, link and rename operations during
6126 MDS failover; recovery should be much more robust now
6127 * remove failed OSCs from LOVs (only lov_create uses this so far)
6128 * the lustre-HOWTO was brought (more) up to date (582544)
6130 2002-10-23 Phil Schwan <phil@clusterfs.com>
6133 - in-use dentries weren't being reused properly (617851)
6134 - prevent multiple LDLM setup (599178)
6135 - fix LOV size calculations for truncate (617853)
6136 - fix client handling of MDS intent errors (POSIX)
6137 - fix permission bug in lovstripe.c test (624321)
6138 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
6139 - truncate past end of file could corrupt data
6140 - proper cleanup after timeouts, crashes, etc (592524, 550815)
6141 - a race in recovery could return ETIMEDOUT to apps (623947)
6142 - building outside the source directory was fixed
6143 * the lustre-HOWTO was brought (more) up to date (582544)
6144 * major progress was made on recovery functionality
6146 2002-10-10 Phil Schwan <phil@clusterfs.com>
6149 - recovery deadlock fix
6150 - rm -rf causes LBUG fix (617817)
6151 - file open by multiple tasks fix (618962)
6152 - directory permissions bugs (602707 and 620007)
6153 - journal_stop fixed with locking (611313)
6154 - O_APPEND failures resolved (618273, perhaps 614459)
6155 - lconf PATH fix (619770)
6156 - IA64 build fix (621450)
6157 - RPC buffer sizes scale with amount of memory
6159 2002-10-01 Phil Schwan <phil@clusterfs.com>
6162 - locks would be cancelled without throwing away data pages,
6163 resulting in inconsistent data (605627)
6164 - inode attributes were not always being refreshed (605627, 612449)
6165 - lconf now continues to cleanup after lctl reports an error
6166 - MDS now enforces user permissions (602707)
6167 - lprocfs cleanup fixed, but not yet enabled (614157)
6168 - fixed infinite server hang, should a client not respond to an AST
6169 - avoid going into recovery if user calls readlink() with a buffer
6170 that's too small (613941)
6171 - AST RPCs no longer require replies (614867) -- this may be changed
6172 - don't crash server if client sends an IOV that's too big (611336)
6173 - fixed lock conversion deadlock (611892)
6174 - fixed the following of symlinks (614622)
6175 * recovery: the server can remove locks from a client that dies, other
6176 clients can make progress
6177 * more extN patch fixes
6178 * compile-time configurable ptlrpc buffer allocations
6180 - collaborative read cache document
6181 - Lustre Lite Performance CDR document-in-progress
6183 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6186 - fix typo in patch-2.4.18
6188 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6191 - clear ptlrpc request each time in handle_incoming_request()
6192 - unlink of files now destroys the object on the OST
6194 2002-09-19 Peter Braam <braam@clusterfs.com>
6196 * add hard link support
6197 * change obdfile creation method
6198 * kernel patch changed
6200 2002-09-19 Peter Braam <braam@clusterfs.com>
6203 - stack overflow bug in extN fixed
6205 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
6207 * documentation updates
6208 - add man pages for config tools
6209 - update tests/README to describe testing with new config tools
6210 - finish metadata API descriptions
6211 * bug fixes and cleanups
6212 - statfs workaround for 16TB limit
6213 - LOV stripe allocation improved, can stripe on subset of OSTs
6214 - LOV file size/IO offset was wrong for files > 4GB in size
6215 - object EA data was being dropped, caused files to be unreadable
6216 - memory overflow with non-LOV OST caused memory corruption
6217 - fixed regression tests to work with new config tools, obdfilter
6218 - fixed bug when directory size became larger than 1 block
6219 - fixed bug (for single client case) when PWD was deleted
6220 - invalidate local directory pages when doing intent-based ops
6221 - avoid LDLM oops when lock callback contained bad data
6223 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
6225 * documentation updates
6226 * bug fixes and cleanups
6227 - configuration tools
6230 - 64-bit compile warnings
6231 - 64-bit internal statfs data
6233 * test_brw on persistent OST devices
6235 * lprocfs (disabled)
6237 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
6239 * documentation updates
6240 * bug fixes and cleanups
6241 * configuration tools
6243 2002-08-30 Peter J. Braam <braam@clusterfs.com>
6246 * many small fixes to 0_5_4
6247 * io/network handling
6248 * thinkos in MDS operations
6250 2002-08-24 Peter J. Braam <braam@clusterfs.com>
6253 * crucial basic fixes to 0.5.3
6254 * IOR, Iozone work over Elan
6257 2002-08-07 Phil Schwan <phil@clusterfs.com>
6258 * version 0_5_3, our first alpha
6259 * we use the new Portals iovs
6260 * documentation updates
6261 * bug fixes and cleanups
6262 * small changes in the DLM wire protocol
6264 2002-07-25 Peter J. Braam <braam@clusterfs.com>
6265 * version 0_5_1 with some initial stability,
6266 * locking on MD and file I/O.
6267 * documentation updates
6268 * several bug fixes since 0.5.0
6269 * small changes in wire protocol
6271 2002-07-18 Phil Schwan <phil@clusterfs.com>
6273 * delivered as Lustre Light Alpha
6274 * fixed a crash after handling invalid MDS requests
6275 * fixed directory pages for architectures with non-4k pages sizes
6277 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
6278 * release version v0_4_4
6279 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
6281 * Fixes a number of interruption problems with OST operations.
6282 * Update documentation for portals header changes
6283 * Move all wire protocol structs/defines to lustre_idl.h
6284 * Fixes symlink length bug.
6285 * Add tcpdump to repository.
6287 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
6288 * release version v0_4_3
6289 * Fixes statfs for inodes on extN.
6290 * Fixes bug in runtests which would delete /etc/hosts.
6291 * Use 64-bit object IDs wherever possible (not into VFS though)
6292 Remove ost_get_info, which is unused by lustre, and out of date.
6294 2002-07-03 Peter Braam <braam@clusterfs.com>
6295 * release version v0_4_2 Fixes a lookup error (type not passed)
6296 * move forward to head of Portals
6297 * move forward to latest Lustre kernel
6299 2002-06-25 Peter Braam <braam@clusterfs.com>
6300 * release version v0_4_1. Hopefully stable on single node use.