3 * Server support for kernels:
4 2.6.18-238.19.1.el5 (RHEL5)
5 2.6.32-131.6.1.el6 (RHEL6)
6 * Client support for unpatched kernels:
7 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
8 2.6.18-238.19.1.el5 (RHEL5)
9 2.6.32-131.6.1.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 RHEL6.1 kernel to 2.6.32-131.6.1.el6
23 Severity : enhancement
24 Description: Update RHEL5.6 kernel to 2.6.18-238.19.1.el5
26 Severity : enhancement
27 Description: Update RHEL5.6 kernel to 2.6.18-238.9.1.el5
30 Frequency : only with SLES10
32 Description: Use OFED "KMP" provided by Novell
33 Details : SLES10 SP3 ships with OFED in a separate "KMP" package.
34 Lustre is now built against this package. That means you need to
35 install the ofed-kmp package from Novell for the patchless client.
36 Note that the ofed-kmp that Novell ships may not exactly match the
37 kernel version but should still be compatible.
39 Severity : enhancement
41 Description: Update RHEL5.5 kernel to 2.6.18-194.17.1.el5
42 Update OEL5.5 kernel to 2.6.18-194.17.1.0.1.el5.
44 --------------------------------------------------------------------------------
46 2010-07-15 Oracle, Inc.
48 * Server support for kernels:
49 2.6.18-164.11.1.el5 (RHEL 5)
50 2.6.18-164.11.1.0.1.el5 (OEL 5)
51 * Client support for unpatched kernels:
52 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
53 2.6.18-164.11.1.el5 (RHEL 5),
54 2.6.18-164.11.1.0.1.el5 (OEL 5)
55 2.6.16.60-0.42.8 (SLES 10),
57 2.6.29.4-167.fc11 (FC11)
58 * Recommended e2fsprogs version: 1.41.10-sun2
60 Severity : enhancement
62 Description: Distribute lfs_migrate script with lustre
63 Details : The lfs_migrate mentioned in the manual is now included in the
64 lustre tree in lustre/scripts and will distributed in lustre RPMs.
66 Severity : enhancement
68 Description: Failover nid list added to import proc file.
70 Severity : enhancement
72 Description: "som_preview" mount option required to use SOM
73 Details : The clients will need to use som_preview mount option
77 Description: Lustre does not do 1MB IOs to HW RAID
78 Details : Bump MAX_PHYS/HW_SEGMENTS and SG_ALL to 256 in the RHEL5 kernel.
80 Severity : enhancement
82 Description: lfs quota output is not very convenient for awk/sed-parsing
83 Details : Some positions in lfs quota output table could be empty or
84 non-empty which made it hard to parse it with scripts, now
85 a dash is put instead of space where there is not supposed
90 Description: Update to OFED 1.5.1
91 Details : OFED 1.5.1 will be used in environments that don't use
96 Description: Remove lnet. prefix from lctl params display
97 Details : Also renamed lnet.memused to lnet_memused to
102 Description: Use conf_param -d to remove permanant settings.
106 Description: Add list_param -R
107 Details : lctl list_param now supports recursive mode.
110 Bugzilla : 21829, 22312
111 Description: Fix llobdstat.
112 Details : 1. Now updating the write_bytes counter
113 2. Port script fixes from 1.8.
117 Description: decrease the usage of memory on clients.
118 Details : 1. On clients, recycle dentries and inodes unused.
119 2. Delete the code related to ll_deathrow(att 6215 in bug 1443). It
124 Description: llobdstat fix and enhancement
125 Details : add a counter to set a limit to how many samples will be returned
126 fix a wildcard in the path to limit to obdfilter stats only
130 Description: lfs check only allowed for root
131 Details : Code cleanups and added a regresion test of non-root use of
136 Description: MDS Oopses in ptlrpc_dispatch_difficult_reply()/__list_add()
137 Details : Fix hr->hr_index overload.
141 Description: Fix FIEMAP functionality.
145 Description: FC11 patchless client support added
149 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and
150 OEL5.4 kernel to 2.6.18-164.11.1.0.1.el5.
154 Description: New acc-sm script and various test-framework enhancments.
158 Description: Previously landed patch was reverted.
162 Description: Update SLES10 kernel to 2.6.16.60-0.42.8 (patchless)
166 Description: Add ldiskfs maxdirsize mount option.
170 Description: Fix mkfs.lustre support for 16TB LUNs.
174 Description: Print mmp_check_interval
175 Details : print mmp_check_interval and make it possible to abort
176 mount operation in case it takes too long.
180 Description: Add cascading_rw to lustre mpi tests.
184 Description: Store SOM data in EA.
185 Details : Also several SOM related cleanups.
189 Description: Add refcount to md_open_data
193 Description: Use refcount on lov_request_set to prevent memory corruption.
197 Description: Statahead: use parent dir's i_mutex for synchronization.
201 Description: loadgen improvements.
205 Description: Enable peer health detection in ksocklnd and ko2iblnd.
209 Description: Control DCACHE_LUSTRE_INVALID flag with MDS_INODE_LOOKCUP lock
213 Description: Fix race on obd_devcie:md_stats
217 Description: Fix SLES 11 client builds.
218 Details : Resolved uninitalized pointer gcc warnings.
222 Description: Allow non-root access for "lfs check".
223 Details : Added a check in obd_class_ioctl() for OBD_IOC_PING_TARGET.
227 Description: Clear imp_force_reconnect correctly in ptlrpc_connect_interpret()
229 Severity : enhancement
231 Description: Update kernel to RHEL5.4 2.6.18-164.6.1.el5 and
232 OEL5 2.6.18-164.6.1.0.1.el5(Both in-kernel OFED enabled).
236 Description: Sanity test_27p fails.
237 Details : OBD_FAIL_OST_ENOSPC has been moved to filter_statfs().
238 Sanity tests 27 were cleaned up and improved.
242 Description: Unify req format on client/servers
243 Details : Use new req_capsule API [almost] everywhere instead of old PTLRPC
244 buffers and swabbers approach..
249 Description: drop join file stuff.
251 Severity : enhancement
253 Description: Add 2.6.27 and SLES11 (patchless client) support
257 Description: Fix in ptlrpc_expire_one_request() to print the signed time
260 Severity : enhancement
262 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
265 Severity : enhancement
267 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
269 Severity : enhancement
271 Description: correctly handle big reply message.
272 Details : send LNet event if reply is bigger then buffer and adjust this buffer
277 Description: Drop unnecessary __GFP_NOMEMALLOC flag from filter_get_page()
279 Severity : enhancement
281 Description: Add support for OEL5.
285 Description: Conf-sanity.sh 50g test - deactivated OST should not cause a panic.
287 Severity : enhancement
289 Description: "weak-modules" support
290 Details : Implement "weak-modules" support which enables kernel modules
291 to be used with any kernel that implements the same kABI. In
292 order to achieve this modules are now installed in
293 /lib/modules/$(uname -r)/updates/kernel on all distributions.
297 Description: client could not reconnect to OST because of an active request.
298 Details : The client now retries to reconnect to the same server, if a
299 connect request failed with EBUSY or -EAGAIN. Test case 35b
300 conf-sanity.sh has been created.
304 Description: Don't allow make backward step on assignin osc next id.
305 Details : race between allocation next id and ll_sync thread can be cause
306 of set wrong osc next id and can be kill valid ost objects.
310 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
314 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
318 Description: Error handling in osc_statfs_interpret() has been improved.
319 Details : Check in osc_statfs_interpret() for EBADR.
323 Description: Do not send statfs() requests to OSTs disabled by administrator.
324 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
326 Severity : enhancement
328 Description: Added sanity-quota test to for lfs(1) to work with numeric uid/gid.
330 Severity : enhancement
332 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
334 Severity : enhancement
336 Description: Enhancement for lfs(1) command to use numeric uid/gid.
340 Description: Improved cleanup procedure for conf-sanity test_46a.
344 Description: Do not start bulk transfer, if client tried to reconnect.
345 Details : Do not start bulk transfer, if client tried to reconnect,
346 but failed due to active rpcs. Check for exp_abort_active_req
347 has been implemented.
351 Description: Prevent inconsistences between linux and lustre mount structures.
352 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
353 Make the sleep interruptible.
355 Severity : enhancement
357 Description: provide server to client comms path
358 Details : server to client communications path via new LDLM_SET_INFO rpc
362 Description: truncate starts GFP_FS allocation under transaction causing deadlock
363 Details : ldiskfs_truncate calls grab_cache_page which may start page
364 allocation under an open transaction. This may lead to
365 calling prune_icache with consequent lustre reentrance.
369 Description: send correctly lsm on open replay
370 Details : MDS is trust to LSM size on replay open, but client can set wrong size
375 Description: enable client interface failover
376 Details : When a child reconnects from another NID, properly update export
377 nid hash position and ldlm reverse import.
379 Severity : enhancement
381 Description: Communicate OST degraded/readonly state via statfs to MDS
382 Details : Flags in the statfs returned from OSTs indicate whether the
383 OST is in a degraded RAID state, or if the filesystem has
384 turned read-only after a filesystem error is detected.
389 Description: don't panic if EPROTO was hit when reading symlink
390 Details : correctly handling request reference in error cases.
392 Severity : enhancement
394 Description: Add LustreNetLink, a kernel-userspace communcation path.
396 Severity : enhancement
398 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
403 Description: access to llog context before init.
404 Details : move handling CATALOGS file at osc layer and forbid access to llog
408 Frequency : rare, only when client is writing to OST on local node
410 Description: Avoid deadlock for local client writes
411 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
412 memory freeing context. This allows OST threads to set the
413 PF_MEMALLOC flag on task structures in order to allocate memory
414 from reserved pools and complete IO.
415 Use GFP_HIGHUSER for OST allocations for non-local client writes,
416 so that the OST threads generate memory pressure and allow
417 inactive pages to be reclaimed.
419 Severity : enhancement
421 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
426 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
427 Details : move ldlm namespace creation in setup phase to avoid grab
428 _lprocfs_lock with cli_sem held.
432 Description: Temporarily disable grant shrink.
433 Details : Disable the feature for debugging.
436 Frequency : only when formatting test filesystems
438 Description: Unable to run several mkfs.lustre on loop devices at the same time
439 Details : mkfs.lustre returns error 256 on the concurrent loop devices
440 formatting. The solution is to proper handle the error.
442 Severity : enhancement
444 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
446 Severity : enhancement
448 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
453 Description: resolve race between obd_disconnect and class_disconnect_exports
454 Details : if obd_disconnect will be called to already disconnected export he
455 forget release one reference and osc module can't unloaded.
457 Severity : enhancement
459 Description: Allow tuning service thread via /proc
460 Details : For each service a new
461 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
462 created that can be used to set min/max thread counts, and get the
463 current number of running threads.
468 Descriptoin: don't return error if have partially created objects for file.
469 Details : lov_update_create_set uses set->set_success as index for created
470 objects, so if some requests will be failed, they will have hole
471 at end of array and qos_shrink_lsm for allocate correct lsm.
473 Severity : enhancement
475 Description: Update OFED support to 1.4.1
480 Description: Reduce small size read RPC
481 Details : Set read-ahead limit for every file and only do read-ahead when
482 available read-ahead pages are bigger than 1M to avoid small size
485 Severity : enhancement
487 Description: move AT tunable parameters for more consistent usage
488 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
490 Severity : enhancement
492 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
493 Details : allow skip disconnected ost for send statfs request and hide error
499 Description: fix racy locking of mballoc block bitmaps causing BUG
500 Details : The locking of the mballoc buddy bitmap and the in-memory
501 block bitmap was using two different spin locks in some
502 cases. This made it possible to incorrectly access the
503 mballoc bitmap while another process was modifying it,
504 causing a sanity assertion to fail. While no on-disk corruption
505 was reported, there was some risk of this happening.
508 Frequency : rare, on llog test 6
510 Descriptoin: don't allow connect to already connected import
511 Details : allowing connect to already connected import is hide connecting problem.
516 Descriptoin: allow kill process which wait statahead result
517 Details : for some reasons 'ls' can stick in waiting result from statahead,
518 in this case need way for kill this process.
520 Severity : enhancement
522 Description: Add state history info file, enhance import info file
523 Details : Track import connection state changes in a new osc/mdc proc file;
524 add overview-type data to the osc/mdc import proc file.
526 Severity : enhancement
528 Description: MDS create should not wait for statfs RPC while holding DLM lock.
530 Severity : enhancement
532 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
537 Descriptoin: OST grant too much space to client even there are not enough space.
538 Details : Client will shrink its grant cache to OST if there are no write
539 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
540 this grant cache if there are already not enough avaible space
541 (left_space < total_clients * 32M).
544 Frequency : start MDS on uncleanly shutdowned MDS device
546 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
547 Details : stay in waiting mds<>ost recovery finished produce random bugs
548 due race between two ll_sync thread for one lov target. send
549 ACTIVATE event only if connect realy finished and import have
553 Frequency : rare, connect and disconnect target at same time
555 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
556 Details : don't call obd_disconnect under lov_lock. this long time
557 operation and can block ptlrpcd which answer to connect request.
562 Descriptoin: don't lose wakeup for imp_recovery_waitq
563 Details : recover_import_no_retry or invalidate_import and import_close can
564 both sleep on imp_recovery_waitq, but we was send only one wakeup
568 Frequency : always with long access acl
570 Descriptoin: mds can't pack reply with long acl.
571 Details : mds don't control size of acl but they limited by reint/getattr
574 Severity : enhancement
576 Description: Update to SLES10 kernel-2.6.16.60-0.33.
578 Severity : enhancement
580 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
583 Frequency : start MDS on uncleanly shutdowned MDS device
585 Descriptoin: aborting recovery hang on MDS
586 Details : don't throttle destroy RPCs for the MDT.
589 Frequency : on remount
591 Description: external journal device not working after the remount
592 Details : clear dev_rdonly flag for external journal devices in
598 Description: shutdown vs evict race
599 Details : client_disconnect_export vs connect request race.
600 if client will evicted at this time - we start invalidate
601 thread without referece to import and import can be freed
605 Frequency : rare, need acl's on inode.
607 Description: client can't handle ost additional correctly
608 Details : if ost was added after client connected to mds client can have
609 hit lnet_try_match_md ... to big messages to wide striped files.
610 in this case need teach client to handle config events about add
611 lov target and update client max ea size at that event.
613 Severity : enhancement
615 Description: Changelogs
616 Details : Changelogs are a lightweight mechanism to track filesystem
617 metadata and namespace changes. The changelog is recorded
618 permanently on the MDTs, and is periodically "consumed" / purged
619 when records are no longer needed.
621 Severity : enhancement
623 Description: compact fld format with extents
624 Details : Store range of seq rather than every seq in FLD. Seq
625 controller update FLD rather than clients. In Case of CMD, mdt0
626 has FLD, all other metadata server act as non persistent proxy
627 for FLD queries and cache fld entries in fld cache.
632 Description: don't skip ost target if they assigned to file
633 Details : Drop slow OSCs if we can, but not for requested start idx.
634 This means "if OSC is slow and it is not the requested
635 start OST, then it can be skipped, otherwise skip it only
636 if it is inactive/recovering/out-of-space.
640 Description: more cleanup in mds_lov
641 Details : not send LOV EA under replay, we can't know about they size at this
642 time. Don't allow client connect to mds before any ost connected,
643 for avoid problems with LOV EA size and returning EIO to client.
645 Severity : enhancement
647 Description: Interoperability at server side (Disk interoperability)
649 Severity : enhancement
651 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
653 Severity : enhancement
655 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
657 Severity : enhancement
659 Description: New client IO stack (CLIO).
661 Severity : enhancement
663 Description: Commit on sharing. Eliminate inter-client dependencies between
664 uncommitted transactions by doing transaction commits.
665 Thereby clients may recovery independently.
668 Frequency : Create a symlink file with a very long name
670 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
671 Details : If there is no extra space in the request for early cancels,
672 ldlm_req_handles_avail() returns 0 instead of a negative value.
674 Severity : enhancement
676 Description: Add /proc entry for import status
677 Details : The mdc, osc, and mgc import directories now have
678 an import directory that contains useful import data for debugging
681 Severity : enhancement
683 Description: Re-disable certain /proc logging
684 Details : Enable and disable client's offset_stats, extents_stats and
685 extents_stats_per_process stats logging on the fly.
688 Frequency : Only on FC kernels 2.6.22+
690 Description: oops in statahead
691 Details : Do not drop reference count for the dentry from VFS when lookup,
692 VFS will do that by itself.
694 Severity : enhancement
696 Description: Generic /proc file permissions
697 Details : Set /Proc file permissions in a more generic way to enable non-
698 root users operate on some /proc files.
702 Description: Hitting mdc_commit_close() ASSERTION
703 Details : Properly handle request reference release in
704 ll_release_openhandle().
708 Description: quota recovery deadlock during mds failover
709 Details : This patch includes att18982, att18236, att18237 in bz14840.
711 1. fix osts hang when mds does failover with quotaon
712 2. prevent watchdog storm when osts threads wait for the
717 Frequency : only patchless client
718 Description: add workaround for race between add/remove dentry from hash
720 Severity : enhancement
722 Description: Allow OST glimpses to return PW locks
726 Description: LBUG when llog conf file is full
727 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
731 Description: Prevent import from entering FULL state when server in recovery
735 Description: service mount cannot take device name with ":"
736 Details : Only when device name contains ":/" will mount treat it as
742 Description: replace ptlrpcd with the statahead thread to interpret the async
743 statahead RPC callback
747 Frequency : on recovery
748 Description: I/O failures after umount during fail back
749 Details : if client reconnected to restarted server we need join to recovery
750 instead of find server handler is changed and process self
751 eviction with cancel all locks.
753 Severity : enhancement
755 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
757 Severity : enhancement
759 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
761 Severity : enhancement
763 Description: Upcall on Lustre log has been dumped
764 Details : Allow for a user mode script to be called once a Lustre log has
765 been dumped. It passes the filename of the dumped log to the
766 script, the location of the script can be specified via
767 /proc/sys/lnet/debug_log_upcall.
772 Description: avoid idr_remove called for id which is not allocated.
773 Details : Move assigment s_dev for clustered nfs to end of initialization,
774 to avoid problem with error handling.
779 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
780 Details : When connection is reused this not moved from CONN_UNUSED_HASH
781 into CONN_USED_HASH and this prodice warning when put connection
782 again in unused hash.
785 Severity : enhancement
787 Description: File striping can now be set to use an arbitrary pool of OSTs.
789 Severity : enhancement
791 Description: Export bytes_read/bytes_write count on OSC/OST.
795 Description: Early reply size mismatch, MGC loses connection
796 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
797 the connect flags are properly negotiated.
802 Description: quotas are not honored with O_DIRECT
803 Details : all writes with the flag O_DIRECT will use grants which leads to
804 this problem. Now using OBD_BRW_SYNC to guard this.
808 Description: add quota statistics
809 Details : 1. sort out quota proc entries and proc code.
810 2. add quota statistics
812 Severity : enhancement
814 Description: enable quota support for HEAD.
818 Description: Properly propagate oinfo flags from lov to osc for statfs
819 Details : restore missing copy oi_flags to lov requests.
821 Severity : enhancement
823 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
824 : l_getgroups(8), lst(8), routerstat(8)
825 Details : included man pages for llobdstat(8), llstat(8),
826 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
828 Severity : enhancement
830 Description: configure's --enable-quota should check the
831 : kernel .config for CONFIG_QUOTA
832 Details : configure is terminated if --enable-quota is passed but
833 : no quota support is in kernel
837 Description: Remove portals compatibility
838 Details : Remove portals compatibility, not interoperable with releases
843 Description: Resolve device initialization race
844 Details : Prevent proc handler from accessing devices added to the
845 obd_devs array but yet be intialized.
847 Severity : enhancement
849 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
851 Severity : enhancement
853 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
858 Description: Using wrong pointer in osc_brw_prep_request
859 Details : Access to array[-1] can produce panic if kernel compiled with
860 CONFIG_PAGE_ALLOC enabled
864 Description: Client runs out of low memory
865 Details : Consider only lowmem when counting initial number of llap pages
869 Description: *optional* service tags registration
870 Details : if the "service tags" package is installed on a Lustre node
871 When the filesystem is mounted, a local-node service tag will
872 be created. See http://inventory.sun.com/ for more information
873 about the Service Tags asset management system.
877 Description: Kernel BUG tries to release flock
878 Details : Lustre does not destroy flock lock before last reference goes
879 away. So always drop flock locks when client is evicted and
880 perform unlock regardless of successfulness of speaking to MDS.
884 Description: add refcount protection for osc callbacks, avoid panic on shutdown
888 Description: sanity test 65a fails if stripecount of -1 is set
889 Details : handle -1 striping on filesystem in ll_dirstripe_verify
894 Description: ASSERTION(CheckWriteback(page,cmd)) failed
895 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
898 Severity : enhancement
900 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
904 Description: do not process already freed flock
905 Details : flock can possibly be freed by another thread before it reaches
906 to ldlm_flock_completion_ast.
910 Description: LBUG during stress test
911 Details : Need properly lock accesses the flock deadlock detection list.
915 Description: oops in page fault handler
916 Details : kernel page fault handler can return two special 'pages' in error
917 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
921 Description: timeout with invalidate import.
922 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
923 should be handled by ptlrpcd. This produce long age waiting and
924 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
926 Severity : enhancement
928 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
932 Description: enable MGS and MDT services start separately
933 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
934 a co-located MGS without starting the MGS, which is a complement
935 to 'nosvc' mount option.
939 Frequency : after recovery
940 Description: precreate to many object's after del orphan.
941 Details : del orphan st in oscc last_id == next_id and this triger growing
942 count of precreated objects. Set flag LOW to skip increase count
943 of precreated objects.
947 Frequency : rare, on clear nid stats
948 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
949 Details : when clean nid stats sometimes try destroy live entry,
950 and this produce panic in free.
954 Description: Stack overflow during MDS log replay
955 ease stack pressure by using a thread dealing llog_process.
959 Description: wait until IO finished before start new when do lock cancel.
960 Details : VM protocol want old IO finished before start new, in this case
961 need wait until PG_writeback is cleared until check dirty flag and
962 call writepages in lock cancel callback.
964 Severity : enhancement
966 Description: using special macro for print time and cleanup in includes.
970 Description: mds_mfd_close() ASSERTION(rc == 0)
971 Details : In mds_mfd_close(), we need protect inode's writecount change
972 within its orphan write semaphore to prevent possible races.
976 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
980 Frequency : rare, on shutdown ost
981 Description: don't hit live lock with umount ost.
982 Details : shrink_dcache_parent can be in long loop with destroy dentries,
983 use shrink_dcache_sb instead.
987 Description: don't panic with use echo client
988 Details : echo client pass NULL as client nid pointer and this produce null
993 Description: fix build on ppc32
994 Details : compile code with -m64 flag produce wrong object file for ppc32.
998 Description: add message levels for liblustreapi
1002 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
1003 Details : Move the 'good_osts' check before the 'total_bavail' check. This
1004 will result in an -EAGAIN and in the exit call path we call
1005 alloc_rr() which will with increasing aggressiveness attempt to
1006 aquire precreated objects on the minimum number of required OSCs.
1010 Description: Use old size assignment to avoid deadlock
1011 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
1012 the scheduling while holding a spinlock. We do not need locking
1013 for size in ll_update_inode() because size is only updated from
1014 the MDS for directories or files without objects, so there is no
1015 other place to do the update, and concurrent access to such inodes
1016 are protected by the inode lock.
1020 Description: resolve "_IOWR redefined" build error on SLES10
1024 Description: dump the memory debugging after all modules are unloaded to
1025 suppress false negative in conf_sanity test 39
1027 Severity : enhancement
1029 Description: build kernel-ib packages for OFED 1.3 in our release cycle
1034 Description: fix SLES kernel versioning
1035 Details : the kernel version for our SLES 10 kernel did not include a "-"
1036 before the "smp" at the end. while this was not a problem in
1037 general, it did mean that software trying to use the kernel
1038 version to try to detect a vendor specific kernel would fail.
1039 this was most evident by the OFED build scripts.
1043 Description: Don't update lov_desc members until making sure they are valid
1044 Details : When updating lov_desc members via proc fs, need fix their
1045 validities before doing the real update.
1049 Description: don't put request into delay list while invalidate in flight.
1050 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
1051 import in flight. this produce timeout for invalidate and sometimes
1052 can cause stale data.
1056 Frequency : on ppc only
1057 Description: not convert ost objects for directory because it's not exist.
1058 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
1062 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
1064 Severity : enhancement
1066 Description: Update to support 2.6.22.14 vanilla kernel.
1070 Frequency : rare, on recovery
1071 Description: read procfs can produce deadlock in some situation
1072 Details : Holding lprocfs lock which send rpc can produce block for destroy
1073 obd objects and this also block reconnect with -EALREADY. This
1074 isn't fix all lprocfs bugs - but make it rare.
1076 Severity : enhancement
1078 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
1081 Frequency : frequent on X2 node
1083 Description: mdc_set_open_replay_data LBUG
1084 Details : Set replay data for requests that are eligible for replay.
1088 Description: lustre_mgs: operation 101 on unconnected MGS
1089 Details : When MGC is disconnected from MGS long enough, MGS will evict the
1090 MGC, and late on MGC cannot successfully connect to MGS and a lot
1091 of the error messages complaining that MGS is not connected.
1095 Frequency : on network error
1096 Description: panic with double free request if network error
1097 Details : mdc_finish_enqueue is finish request if any network error ocuring,
1098 but it's true only for synchronus enqueue, for async enqueue
1099 (via ptlrpcd) this incorrect and ptlrpcd want finish request
1102 Severity : enhancement
1104 Description: client-side metadata stat-ahead during readdir(directory readahead)
1105 Details : perform client-side metadata stat-ahead when the client detects
1106 readdir and sequential stat of dir entries therein
1109 Frequency : on start mds
1111 Description: Implement get_info(last_id) in obdfilter.
1114 Frequency : occasional
1116 Description: Correctly check stale fid, not start epoch if ost not support SOM
1117 Details : open with flag O_CREATE need set old fid in op_fid3 because
1118 op_fid2 was overwritten with new generated fid, but mds can answer
1119 with one of these two fids and both is not stale. Setattr
1120 incorrectly started an epoch and assume will be called
1121 done_writing, but without SOM done_writing ever being called.
1124 Frequency : rare, depends on device drivers and load
1126 Description: MDS or OSS nodes crash due to stack overflow
1127 Details : Code changes in 1.8.0 increased the stack usage of some functions.
1128 In some cases, in conjunction with device drivers that use a lot
1129 of stack the MDS (or possibly OSS) service threads could overflow
1130 the stack. One change which was identified to consume additional
1131 stack has been reworked to avoid the extra stack usage.
1134 Frequency : occasional
1136 Description: Do not fail import if osc_interpret_create gets -EAGAIN
1137 Details : If osc_interpret_create got -EAGAIN it immediately exits and
1138 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
1139 oscc_has_objects and see OSC has no objests and call
1140 oscc_internal_create to resend create request.
1142 Severity : enhancement
1144 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
1146 Severity : enhancement
1148 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
1151 Frequency : very rare
1153 Description: Fix replay if there is an un-replied request and open
1154 Details : In some cases, older replay request will revert the
1155 mcd->mcd_last_xid on MDS which is used to record the client's
1156 latest sent request.
1158 Severity : enhancement
1160 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
1162 Severity : enhancement
1164 Description: Add rhel5 support to HEAD.
1166 Serverity : enhancement
1168 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
1173 Description: Don't allow skipping OSTs if index has been specified.
1174 Details : Don't allow skipping OSTs if index has been specified, make
1175 locking in internal create lots better.
1179 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
1180 Details : access to bitfield in structure is always rounded to long
1181 and this produce problem with not atomic change any bit.
1185 Description: Lustre make rpms failed.
1186 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
1187 builds Lustre rpms from tarball.
1190 Frequency : rare on shutdown ost
1192 Description: If llog cancel was not send before clean_exports phase, this can
1193 produce deadlock in llog code.
1194 Details : If llog thread has last reference to obd and call class_import_put
1195 this produce deadlock because llog_cleanup_commit_master wait when
1196 last llog_commit_thread exited, but this never success because was
1197 called from llog_commit_thread.
1201 Description: allow userland application know is lost one of stripes.
1202 Details : fill lvb_blocks with error code on ost and return it to
1203 application if error flag found.
1207 Description: NULL lov_tgts causing MDS oops
1208 Details : more safe checks for NULL lov_tgts for avoid oops.
1210 Severity : enhancement
1212 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
1216 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
1217 Details : If an OST is down the MDS will hang indefinitely in
1218 obd_statfs_rqset() waiting for the statfs data. While for
1219 MDS QOS usage of statfs, it should not stuck in waiting.
1221 Severity : enhancement
1223 Description: remote_acl support
1224 Details : Support ACL-based permission check for remote user.
1225 Support setfacl/getfacl for remote user with the utils
1226 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
1227 the system "{s,g}etfacl" utils.
1229 Severity : enhancement
1231 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
1233 Severity : enhancement
1235 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
1239 Description: make mgs_setparam() handle fsname containing dash
1240 Details : fsname containing a dash does not work with lctl conf_param
1242 Severity : enhancement
1244 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
1246 Severity : enhancement
1248 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
1252 Frequency : rare, at shutdown
1253 Description: access already free / zero obd_namespace.
1254 Details : if client_disconnect_export was called without force flag set,
1255 and exist connect request in flight, this can produce access to
1256 NULL pointer (or already free pointer) when connect_interpret
1257 store ocd flags in obd_namespace.
1261 Frequency : only at startup
1262 Description: not alloc memory with spinlock held.
1263 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
1264 if any spinlock held.
1266 Severity : enhancement
1268 Description: make lustre randomly fail allocating memory
1269 Details : Make lustre randomly failed allocating memory for testing purpose.
1271 Severity : enhancement
1273 Description: lost problems with lov objid file
1274 Details : Fixes some scability and access to not inited memory problems
1275 in work with lov objdid file.
1280 Description: lfs find does not continue on file error
1281 Details : Continue other files processing when a file/dir is absent.
1285 Description: Inconsistent usage of lustre_pack_reply()
1286 Details : Standardize the usage of lustre_pack_reply() such that it
1287 always generate a CERROR on failure.
1290 Frequency : occasional
1292 Description: MDS hang or stay in waiting lock
1293 Details : If client receive lock with CBPENDING flag ldlm need send lock
1294 cancel as separate rpc, to avoid situation when cancel request
1295 can't processed due all i/o threads stay in wait lock.
1299 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1303 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1307 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1311 Description: Update to RHEL4 latest kernel.
1313 Severity : enhancement
1315 Description: Build SLES10 patchless client fails
1316 Details : The configure was broken by run ./configure with
1317 --with-linux-obj=.... argument for patchless client. When the
1318 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1319 can't search header adequately. Use absolute path such as
1320 -I($LINUX)/include instead.
1324 Description: interrupt oig_wait produce painc on resend.
1325 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1326 private set, and this produce situation when rq_ptlrpcd_data not
1327 copyed to new allocated request and triggered LBUG on assert
1328 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1329 ptlrpcd set if request is changed and can be safety changed to use
1334 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1335 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1336 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1337 these macros and replace them with their difinitions instead.
1341 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1342 Details : Update which_patch & target file for SLES10 latest kernel.
1345 Description: Remove client patches from SLES 10 kernel.
1346 Details : This causes SLES 10 clients to behave as patchless clients
1347 even on a Lustre-patched (server) kernel.
1349 Severity : enhancement
1351 Description: self-adjustable client's lru lists
1352 Details : use adaptive algorithm for managing client cached locks lru
1353 lists according to current server load, other client's work
1354 pattern, memory activities, etc. Both, server and client
1355 side namespaces provide number of proc tunables for controlling
1358 Severity : enhancement
1360 Description: light-weight GSS support
1361 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
1362 and reduce performance overhead.
1364 Severity : enhancement
1366 Description: Linux keyring support
1367 Details : Support using service of Linux keyring for Lustre GSS internal
1368 context refresh/cache mechanism.
1372 Description: Fix errors in lfs documentation
1373 Details : Fixes man pages
1377 Description: don't use GFP_* in generic Lustre code.
1378 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
1382 Description: obdclass is limited by single OBD_ALLOC(idarray)
1383 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
1387 Description: warnings with build patchless client with vanila 2.6.19 and up
1388 Details : change old ctl_table style and replace ctl_table/ctl_table_header
1389 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
1393 Frequency : Only for SLES
1394 Description: sanity_quota fail test_1
1395 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
1396 file, which makes TSTID[2] inunique.
1400 Frequency : Only for SLES9
1401 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
1402 Details : Update target/ChangeLog/which_patch .
1406 Frequency : Only for test_10 in sanity-quota.sh
1407 Description: a bug in quota test code
1408 Details : checking a test flag in a wrong place
1412 Frequency : Only for test_13 in sanity-quota.sh
1413 Description: a bug in quota test script
1414 Details : shouldn't check overfull allocated quota
1418 Description: jbd statistics
1419 Details : Port older jbd statistics patch for sles10
1421 Severity : enhancement
1423 Description: Add printing inode info into message about error in writepage.
1428 Description: reply_lock_interpret crash due to race with it and lock cancel.
1429 Details : Do not replay locks that are being cancelled. Do not reference
1430 locks by their address during replay, just by their handle.
1434 Frequency : When flocks are used.
1435 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
1436 Details : Flock locks might destroy just granted lock if it could be merged
1437 with another existing flock, this is done in completion handler,
1438 so teach ldlm_cli_enquque_fini that this is a valid case for
1444 Description: Oops in read and write path when failing to allocate lock.
1445 Details : Check if lock allocation failed and return error back.
1449 Description: lstripe command fails for valid OST index
1450 Details : The stripe offset is compared to lov->desc.ld_tgt_count
1451 instead of lov->desc.ld_active_tgt_count.
1455 Description: sanity.sh failed test 103
1456 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
1461 Description: df doesn't work properly if diskfs blocksize != 4K
1462 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
1466 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
1467 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
1469 Severity : enhancement
1471 Description: Adaptive timeouts
1472 Details : RPC timeouts adapt to changing server load and network
1473 conditions to reduce resend attempts and improve recovery time.
1477 Description: llapi_file_create() does not allow some changes
1478 Details : add llapi_file_open() that allows specifying the mode and
1479 open flags, and also returns an open file handle.
1483 Description: merge and cleanup kernel patches.
1484 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
1488 Description: Add journal checksum support.(Kernel part)
1489 Details : The journal checksum feature adds two new flags i.e
1490 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
1491 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
1492 indicates that the commit block contains the checksum for
1493 the blocks described by the descriptor blocks. Now commit
1494 record can be sent to disk without waiting for descriptor
1495 blocks to be written to disk. This behavior is controlled
1496 using JBD2_FEATURE_ASYNC_COMMIT flag.
1500 Description: OSS needs mutliple precreate threads
1501 Details : Add ability to start more than one create thread per OSS.
1505 Description: Sanity.sh test_65k failed: setstripe should have succeeded
1506 Details : execute lfs setstripe on client
1510 Description: mds_obd_create error creating tmp object
1511 Details : When the user sets quota on root, llog will be affected and can't
1512 create files and write files.
1515 Frequency : Always on ia64 patchless client, and possibly others.
1517 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
1518 Details : This allows the patchless client to be loaded on architectures
1519 without this export.
1523 Description: RedHat Update kernel for RHEL5
1524 Details : Add the kernel config file for RHEL5.
1528 Description: RedHat Update kernel for RHEL5
1529 Details : Modify the kernel config file more closer RHEL5.
1533 Description: Build failure against Centos5 (RHEL5)
1534 Details : Define PAGE_SIZE when it isn't present.
1538 Description: test_7 on sanity-quota.sh can't be used on separate machines
1539 Details : add facet to handle it
1543 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
1544 Details : with NFS, the anon dentry's parent was set to itself in
1545 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
1546 corresponding dentry other than use rec->ur_name.
1548 Severity : enhancement
1550 Description: lfs setstripe enhancement
1551 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
1555 Description: enable data checksumming by default
1556 Details : enable checksum by default, allow --disable-checksum
1557 configure option and "-o nochecksum" mount option. Checksums
1558 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
1562 Description: lustre support for RHEL5
1563 Details : Add support for RHEL5.
1567 Description: Client eviction due to failover config
1568 Details : after a connection loss, the lustre client should attempt to
1569 reconnect to the last active server first before trying the
1570 other potential connections.
1574 Description: when mds and osts use different quota unit(32bit and 64bit),
1575 quota will be released repeatly.
1576 Details : void sending multiple quota reqs to mds, which will keep the
1577 status between the reqs.
1581 Description: osts not allocated evenly to files
1582 Details : change the condition to increase offset_idx
1585 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
1587 Description: Data corruption for OSTs that are formatted larger than 2TB
1589 Details : When generating the bio request for lustre file writes the
1590 sector number would overflow a temporary variable before being
1591 used for the IO. The data reads correctly from Lustre (which
1592 will overflow in a similar manner) but other file data or
1593 filesystem metadata may be corrupted in some cases.
1597 Description: Tune the kernel for good SCSI performance.
1598 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
1599 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
1604 Description: rewrite ext2-derived code in obdclass/uuid.c
1605 Details : rewrite inherited code (uuid parsing code from ext2 utils)
1606 from scratch preserving functionality.
1610 Description: "lfs find -obd UUID" prints directories
1611 Details : "lfs find -obd UUID" will return all directory names instead
1612 of just file names. It is incorrect because the directories
1613 do not reside on the OSTs.
1617 Description: use i_size_read and i_size_write in 2.6 port
1618 Details : replace inode->i_size access with i_size_read/write()
1621 Frequency : when removing large files
1623 Description: scheduling issue during removal of large Lustre files
1624 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
1625 It causes scheduling issues when removing large files (17TB in the
1629 Frequency : only with liblustre clients on XT3
1631 Description: evictions taking too long
1632 Details : allow llrd to evict clients directly on OSTs
1635 Frequency : only on ppc
1637 Description: /proc/fs/lustre/devices broken on ppc
1638 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
1639 We should make sure the type of 'index' is loff_t and then cast
1640 explicitly as needed below. Do not assign an explicitly cast
1644 Frequency : only for rhel5
1646 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
1647 Details : Modify the target file & which_kernel.
1649 Severity : enhancement
1651 Description: omit set fsid for export NFS
1652 Details : fix set/restore device id for avoid EMFILE error and mark lustre
1653 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
1657 Frequency : Always, for kernels after 2.6.16
1658 Description: Fix warning idr_remove called for id=.. which is not allocated.
1659 Details : Last kernels save old s_dev before kill super and not allow
1660 to restore from callback - restore it before call kill_anon_super.
1664 Description: buffer overruns could theoretically occur
1665 Details : llapi_semantic_traverse() modifies the "path" argument by
1666 appending values to the end of the origin string, and a
1667 overrun may occur. Adding buffer overrun check in liblustreapi.
1671 Description: Fix error on 'ls .' at the top of the Lustre mount.
1672 Details : Don't revalidate dentry if it is a root dentry.
1676 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
1677 Details : Modify vm-tunables-rhel4.patch.
1681 Description: Kernel config for 2.6.18-vanilla.
1682 Details : Modify targets/2.6-vanilla.target.in.
1683 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
1684 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
1685 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
1686 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
1690 Description: improve handling recoverable errors
1691 Details : if request processig with error which can be recoverable on server
1692 request should be resend, otherwise page released from cache and
1697 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
1698 Details : Modify target file & which_patch.
1699 A flaw was found in the IA32 system call emulation provided
1700 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
1701 value could be stored in the %RAX register, which could trigger an
1702 out-of-bounds system call table access. An untrusted local user
1703 could exploit this flaw to run code in the kernel
1704 (ie a root privilege escalation). (CVE-2007-4573).
1708 Description: O_DIRECT bypasses client statistics.
1709 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
1710 read_ahead_stats were updated. copied the stats section from
1711 osc_send_oap_rpc() into async_internal().
1715 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
1719 Description: change order of libsysio includes
1720 Details : '#include sysio.h' should always come before '#include xtio.h'
1724 Description: handle "serious error: objid * already exists" more gracefully
1725 Details : If LAST_ID value on disk is smaller than the objects existing in
1726 the O/0/d* directories, it indicates disk corruption and causes an
1727 LBUG(). If the object is 0-length, then we should use the existing
1728 object. This will help to avoid a full fsck in most cases.
1730 Severity : enhancement
1732 Description: adapt the lustre_config script to support the upgrade case
1733 Details : Add "-u" option for lustre_config script to support upgrading 1.4
1734 server targets to 1.6 in parallel.
1738 Description: To avoid grant space > avaible space when the disk is almost
1739 full. Without this patch you might see the error "grant XXXX >
1740 available" or some LBUG about grant, when the disk is almost
1742 Details : In filter_check_grant, for non_grant cache write, we should
1743 check the left space by if (*left > ungranted + bytes), instead
1744 of (*left > ungranted), because only we are sure the left space
1745 is enough for another "bytes", then the ungrant space should be
1746 increase. In client, we should update cl_avail_grant only there
1747 is OBD_MD_FLGRANT in the reply.
1751 Description: Update RHEL 4 kernel to fix local root privilege escalation.
1752 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
1753 described in CVE-2007-4573. This problem could allow untrusted
1754 local users to gain root access.
1757 Frequency : when using O_DIRECT and quotas
1759 Description: Incorrect file ownership on O_DIRECT output files
1760 Details : block usage reported by 'lfs quota' does not take into account
1761 files that have been written with O_DIRECT.
1766 Description: touch file failed when fs is not full
1767 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
1768 otherwise we can get ENOSP while fs is not full.
1772 Description: parallel lock callbacks
1773 Details : Instead of sending blocking and completion callbacks as separated
1774 requests, adding them to a set and sending in parallel.
1777 Frequency : only for Cray XT3
1778 Bugzilla : 12829/13455
1779 Description: Changing primary group doesn't change the group lustre assigns to
1781 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
1782 overridden with the primary group provided by the group upcall,
1783 whereas the supplied fsgid can be trusted if it is in the list of
1784 supplementary groups returned by the group upcall.
1786 Severity : enhancement
1788 Description: Allow masking D_WARNING, D_ERROR messages from console
1789 Details : Console messages can now be disabled via lnet.printk.
1793 Description: User code with malformed file open parameter crashes client node
1794 Details : Before packing join_file req, all the related reference should be
1795 checked carefully in case some malformed flags cause fake
1796 join_file req on client.
1800 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
1801 Details : ldlm_completion_ast() assumes that a lock is granted when the req
1802 mode is equal to the granted mode. However, it should also check
1803 that LDLM_FL_CP_REQD is not set.
1807 Description: Heavy nfs access might result in deadlocks
1808 Details : After ELC code landed, it is now improper to enqueue any mds
1809 locks under och_sem, because enqueue might want to decide to
1810 cancel open locks for same inode we are holding och_sem for.
1814 Description: Client eviction while running blogbench
1815 Details : A lot of unlink operations with concurrent I/O can lead to a
1816 deadlock causing evictions. To address the problem, the number of
1817 oustanding OST_DESTROY requests is now throttled to
1818 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
1819 callbacks are processed in priority.
1823 Description: enable ACLs on MDS by default
1824 Details : ACLs must be enabled on MDS by default.
1827 Frequency : PPC/PPC64 only
1829 Description: conflicts between asm-ppc64/types.h and lustre_types.h
1830 Details : fix duplicated definitions between asm-ppc64/types.h and
1831 lustre_types.h on PPC.
1834 Frequency : PPC/PPC64 only
1836 Description: asm-ppc/segment.h does not exist
1837 Details : fix compile issue on PPC.
1841 Description: better handle error messages in extents code
1844 Frequency : RHEL4 only
1846 Description: mkfs is very slow on IA64/RHEL4
1847 Details : A performance regression has been discovered in the MPT Fusion
1848 driver between versions 3.02.73rh and 3.02.99.00rh. As a
1849 consequence, we have downgraded the MPT Fusion driver in the RHEL4
1850 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
1852 Severity : enhancement
1854 Description: SNMP support enhancement
1855 Details : Adding total number of sampled request for an MDS node in snmp
1858 Severity : enhancement
1860 Description: Optimize ldlm waiting list processing for PR extent locks
1861 Details : When processing waiting list for read extent lock and meeting read
1862 lock that is same or wider to it that is not contended, skip
1863 processing rest of the list and immediatelly return current
1864 status of conflictness, since we are guaranteed there are no
1865 conflicting locks in the rest of the list.
1869 Description: Time out and refuse to reconnect
1870 Details : When the failover node is the primary node, it is possible
1871 to have two identical connections in imp_conn_list. We must
1872 compare not conn's pointers but NIDs, otherwise we can defeat
1873 connection throttling.
1877 Description: port llog fixes from b1_6 into HEAD
1878 Details : Port llog reference couting and some llog cleanups from b1_6
1879 (bug 10800) into HEAD, for protect from panic and access to already
1880 free llog structures.
1884 Description: Detect stride IO mode in read-ahead
1885 Details : When a client does stride read, read-ahead should detect that and
1886 read-ahead pages according to the detected stride pattern.
1890 Description: data checksumming impacts single node performance
1891 Details : add support for several checksum algorithm. Currently, only CRC32
1892 and Adler-32 are supported. The checksum type can be changed on
1893 the fly via /proc/fs/lustre/osc/*/checksum_type.
1897 Description: use adler32 for page checksums
1898 Details : when available, use the Adler-32 algorithm instead of CRC32 for
1903 Description: build for x2 fails
1904 Details : fix compile issue on Cray systems.
1908 Description: Properly match for duplicate locks
1909 Details : Due to different lock order from skiplists code, we need to
1910 traverse entire list for now
1913 Frequency : only on PPC/SLES10
1915 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
1916 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
1917 be defined. Add a hack in mkfs_lustre.c to work around this compile
1922 Description: LASSERT on MDS when client holding flock lock dies
1923 Details : ldlm pool logic depends on number of granted locks equal to
1924 number of released locks which is not true for flock locks, so
1925 just exclude such locks from consideration.
1929 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
1930 Details : Use fsfilt_sync() for both the whole filesystem sync and
1931 individual file sync to eliminate dangerous inode locking
1932 with I_LOCK that can lead to a deadlock.
1936 Description: performance in 1.6.3
1937 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
1938 sure that 1MB requests are not fragmented by the block layer.
1940 Severity : enhancement
1942 Description: organize the server-side client stats on per-nid basis
1943 Details : Change the structure of stats under obdfilter and mds to
1951 The "uuid"s file would list the uuids of _active_ exports.
1952 And the clear entry is to clear all stats and stale nids.
1954 Severity : enhancement
1956 Description: eliminate client locks in face of contention
1957 Details : file contention detection and lockless i/o implementation
1958 for contended files.
1962 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
1965 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
1967 Description: MDS may lose file striping (and hence file data) in some cases
1968 Details : If there are additional extended attributes stored on the MDS,
1969 in particular ACLs, SELinux, or user attributes (if user_xattr
1970 is specified for the client mount options) then there is a risk
1971 of attribute loss. Additionally, the Lustre file striping
1972 needs to be larger than default (e.g. striped over all OSTs),
1973 and an additional attribute must be stored initially in the
1974 inode and then increase in size enough to be moved to the
1975 external attribute block (e.g. ACL growing in size) for file
1980 Description: skiplist implementation simplification
1981 Details : skiplists are used to group compatible locks on granted list
1982 that was implemented as tracking first and last lock of each
1983 lock group the patch changes that to using doubly linked lists
1987 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
1988 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
1989 because LASSERT on both the data supplied by a client, and the
1990 data on disk is dangerous and incorrect.
1992 Severity : enhancement
1994 Description: Slow truncate/writes to huge files at high offsets.
1995 Details : Directly associate cached pages to lock that protect those pages,
1996 this allows us to quickly find what pages to write and remove
1997 once lock callback is received.
2001 Description: more ldlm soft lockups
2002 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
2003 starve other threads from the resource lock for a long time in
2004 case of long waiting queue, so change the debug level from
2005 D_OTHER to the less frequently used D_INFO.
2007 Severity : enhancement
2009 Description: add -gid, -group, -uid, -user options to lfs find
2013 Description: Hung threads in invalidate_inode_pages2_range
2014 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
2015 once one is completed. As a result, some RPCs are stuck in the
2016 queue and are never sent.
2020 Description: filter threads hungs on waiting journal commit
2021 Details : Cleanup filter group llog code, then only filter group llog will
2022 be only created in the MDS/OST syncing process.
2026 Description: Procfs and llog threads access destoryed import sometimes.
2027 Details : Sync the import destoryed process with procfs and llog threads by
2028 the import refcount and semaphore.
2030 Severity : enhancement
2032 Description: openlock cache of b1_6 port to HEAD
2037 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
2038 Details : If insertion of an extent fails, then discard the inode
2039 preallocation and free data blocks else it can lead to duplicate
2044 Description: don't always update ctime in ext3_xattr_set_handle()
2045 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
2046 In some cases the ctime should not be updated, for example for
2047 2.0->1.8 compatibility it is necessary to delete an xattr and it
2048 should not update the ctime.
2052 Bugzilla : 15713/16362
2053 Description: Assertion in iopen_connect_dentry in 1.6.3
2054 Details : looking up an inode via iopen with the wrong generation number can
2055 populate the dcache with a disconneced dentry while the inode
2056 number is in the process of being reallocated. This causes an
2057 assertion failure in iopen since the inode's dentry list contains
2058 both a connected and disconnected dentry.
2062 Description: assertion failure in ldlm_handle2lock()
2063 Details : fix a race between class_handle_unhash() and class_handle2object()
2064 introduced in lustre 1.6.5 by bug 13622.
2069 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
2070 Details : remove the limit of 256 scsi disks in the sd_iostat patch
2075 Description: oops in sd_iostats_seq_show()
2076 Details : unloading/reloading the scsi low level driver triggers a kernel
2077 bug when trying to access the sd iostat file.
2082 Description: Kernel panics during QLogic driver reload
2083 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
2084 patch, causing memory corruption.
2089 Description: journal_dev option does not work in b1_6
2090 Details : pass mount option during pre-mount.
2092 Severity : enhancement
2094 Description: Add a FIEMAP(FIle Extent MAP) ioctl
2095 Details : FIEMAP ioctl will allow an application to efficiently fetch the
2096 extent information of a file. It can be used to map logical blocks
2097 in a file to physical blocks in the block device.
2101 Description: LDLM soft lockups - improvement
2102 Details : It is be possible to send the lock handle along with each read
2103 or write request because the client is already doing a lock match
2104 itself so there isn't any reason the OST should have to re-do that
2108 Frequency : only on Cray X2
2110 Description: X2 build failures
2111 Details : fix build failures on Cray X2.
2115 Description: xid & resent requests
2116 Details : Initialize RPC XID from clock at startup (randomly if clock is
2119 Severity : enhancement
2121 Description: Add lustre_start utility to start or stop multiple Lustre servers
2126 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
2127 Details : In case of memory pressure, list_del() can be called twice on
2128 req->rq_history_list, causing a kernel oops.
2132 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
2133 Details : fix stack overflow in the distributed lock manager by defering
2134 export eviction after a failed AST to the elt thread instead of
2135 handling it in the dlm interpret routine.
2139 Description: Convert some comments to new format.
2140 Details : Update documenting comments to match doxygen conventions.
2144 Description: Grammar fixes.
2145 Details : A couple of trivial spelling fixes.
2149 Description: OSD_COUNTERS-mandatory
2150 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
2151 to simplify the code.
2155 Description: simplify cmm_device freeing logic.
2156 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
2160 Description: Add lockdep support to dt_object_operations locking interface.
2161 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
2162 parameter indicating lock ordering. Update mdd code to use new
2167 Description: Introduce failloc constants for lockless IO tests.
2168 Details : Add two new failloc constants to test lockless IO. Only one of
2169 them in implemented---another is checked in yet to be landed
2174 Description: Add lockdep support for inode mutex.
2175 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
2176 in the situations where Lustre has to lock more than one inode
2181 Description: Add optional invariants checking support.
2182 Details : Add new LINVRNT() macro, optional on new --enable-invariants
2183 configure switch. This macro is to be used for consistency and
2184 sanity checks that are too expensive to be left in `production'
2189 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
2190 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
2195 Description: Add lockdep support for ldlm_lock and ldlm_resource.
2196 Details : Use spin_lock_nested() in (the only) situation where more than
2197 one ldlm_lock is locked simultaneously. Also, fix possible
2198 dead-lock in ldlm_lock_change_resource() by enforcing particular
2203 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
2204 Details : Instead of specifying each ldlm_lock call-back through separate
2205 parameter, wrap them into struct ldlm_callback_suite.
2209 Description: Kill join_lru obd method and its callers.
2210 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
2211 in memory---a requirement implemented through ->o_join_lru() obd
2212 method in HEAD. Remove this method and its users.
2216 Description: Add asynchronous ldlm ENQUEUE completion handler.
2217 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
2218 case that stock ldlm_completion_ast() cannot handle as it waits
2219 until lock is granted. Introduce new ldlm_completion_ast_async()
2220 for this. Also comment ldlm_completion_ast().
2224 Description: ldlm_error <-> errno conversion.
2225 Details : Add functions to map (rather arbitrary) between LDLM error codes
2226 and standard errno values. CLIO needs this to prevent LDLM specific
2227 constants from escaping ldlm and osc.
2231 Description: Kill unused ldlm_handle2lock_ns() function.
2232 Details : Kill unused ldlm_handle2lock_ns() function.
2236 Description: Add lu_ref support to ldlm_lock
2237 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
2238 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
2239 and ldlm_resource. LDLM interface has to be changed, because code that
2240 releases a reference on a lock, has to "know" what reference this is.
2241 In the most frequent case
2243 lock = ldlm_handle2lock(handle);
2245 LDLM_LOCK_PUT(lock);
2247 no changes are required. When any other reference (received _not_ from
2248 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
2249 instead of LDLM_LOCK_PUT().
2251 Arguably, changes are pervasive, and interface requires some discipline
2252 for proper use. On the other hand, it was very instrumental in finding
2253 a few leaked lock references.
2257 Description: Add ldlm_lock_addref_try().
2258 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
2259 attempts to addref a lock that might be being canceled
2264 Description: Add ldlm_weigh_callback().
2265 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
2266 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
2267 hard-coded `number of pages' logic.
2271 Description: Add lockdep annotations to llog code.
2272 Details : Use appropriately tagged _nested() locking calls in the places
2273 where llog takes more than one ->lgh_lock lock.
2277 Description: Add loi_kms_set().
2278 Details : Wrap kms updates into a helper function.
2282 Description: Constify instances of struct lsm_operations.
2283 Details : Constify instances of struct lsm_operations.
2287 Description: lu_conf support.
2288 Details : On a server, a file system object is uniquely identified
2289 by a fid, which is sufficient to locate and load all object
2290 state (inode). On a client, on the other hand, more data are
2291 necessary instantiate an object. Change lu_object_find() and
2292 friends to take additional `lu_conf' argument describing object.
2293 Typically this includes layout information.
2297 Description: lu_context fixes.
2298 Details : Introduce new lu_context functions that are needed on the client
2299 side, where some system threads (ptlrpcd) are shared by multiple
2300 modules, and so cannot be stopped during module shutdown.
2304 Description: Add start and stop methods to lu_device_type_operations.
2305 Details : Introduce two new methods in lu_device_type_operations, that are
2306 invoked when first instance of a given type is created and last one
2307 is destroyed respectively. This is need by CLIO.
2311 Description: Add lu_ref support to struct lu_device.
2312 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
2313 track leaked references.
2317 Description: Introduce lu_kmem_descr.
2318 Details : lu_kmem_descr and its companion interface allow to create
2319 and destroy a number of kmem caches at once.
2323 Description: Fix lu_object finalization race.
2324 Details : Fix a race between lu_object_find() finding an object and its
2325 concurrent finalization. This race is (most likely) not possible
2326 on the server, but might happen on the client.
2330 Description: Introduce lu_ref interface.
2331 Details : lu_ref is a debugging module allowing to track references to
2332 a given object. It is quite cpu expensive, and has to be
2333 explicitly enabled with --enable-lu_ref. See usage description
2338 Description: Factor lu_site procfs stats into a separate function.
2339 Details : Separate lu_site stats printing code into a separate function
2340 that can be reused on a client.
2344 Description: Constify instances of struct {lu,dt,md}_device_operations.
2345 Details : Constify instances of struct {lu,dt,md}_device_operations.
2349 Description: Introduce struct md_site and move meta-data specific parts of
2350 struct lu_site here.
2351 Details : Move md-specific fields out of struct lu_site into special struct
2352 md_site, so that lu_site can be used on a client.
2356 Description: Kill mdd_lov_destroy().
2357 Details : Remove unused mdd code.
2361 Description: Add st_block checking to multistat.c.
2362 Details : Add st_block checking to multistat.c.
2366 Description: Add lu_ref support to struct obd_device.
2367 Details : Add lu_ref tracking to obd_device.
2371 Description: Kill obd_set_fail_loc().
2372 Details : Remove unused code.
2376 Description: Add special type for ptlrpc_request interpret functions.
2377 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
2378 there. Actual usage will be in CLIO.
2382 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
2383 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
2384 doesn't work with lockdep.
2388 Description: Add rwv.c test program.
2389 Details : New testing program exercising readv(2) and writev(2) (Qian).
2393 Description: Add sendfile.c test program.
2394 Details : New testing program exercising sendfile(2) (Jay).
2398 Description: Ratelimit a message that can be very frequent.
2399 Details : Ratelimit a memory allocation failure message that can
2404 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2405 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2409 Description: Liblustre build fixes.
2410 Details : Liblustre build fixes.
2414 Description: libcfs: add cfs_{need,cond}_resched() interface.
2415 Details : libcfs: add cfs_{need,cond}_resched() definition and
2416 implementations for Linux, NT, and liblustre.
2418 Severity : enhancement
2420 Description: More exported tunables for mballoc
2421 Details : Add support for tunable preallocation window and new tunables for
2422 large/small requests
2426 Description: Detect corruption of block bitmap and checking for preallocations
2427 Details : Checks validity of on-disk block bitmap. Also it does better
2428 checking of number of applied preallocations. When corruption is
2429 found, it turns filesystem readonly to prevent further corruptions.
2433 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
2434 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
2435 window to 0. If it does hit cache miss, and read-pattern is still
2436 stride-io mode, does not reset the stride window, but also does not
2437 increase the stride window length in this case.
2441 Frequency : only for big-endian servers
2442 Description: Check if system is big-endian while mounting fs with extents feature
2443 Details : Mounting a filesystem with extents feature will fail on big-endian
2444 systems since ext3-based ldiskfs is not supported on big-endian
2445 systems. This can be over-riden with "bigendian_extents" mount option.
2447 Severity : enhancement
2449 Description: The root squash functionality
2450 Details : A security feature, which is to prevent users from being able
2451 to mount lustre on their desktop, run as root, and delete
2452 all of the files in the filesystem. The goal is accomplished by
2453 remapping user id (UID) and group id (GID) of the root user to
2454 a UID and GID specified by the system administartor via Lustre
2455 configuration management server (MGS). The functionality also
2456 allows to specify sets of clients for which the remapping does
2461 Description: Excessive recovery window
2462 Details : With AT enabled, the recovery window can be excessively long (6000+
2463 seconds). To address this problem, we no longer use
2464 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
2465 timeout no longer depends on the service time, it is set to
2466 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
2467 time via pb_service_time.
2471 Description: Watchdog triggered on MDS failover
2472 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
2473 the OSTs know that the MDS "UUID" can be reused for the same export
2474 from a different NID, so we do not need to wait for the export to be
2478 Frequency : rare, only if using MMP with Linux RAID
2480 Description: MMP doesn't work with Linux RAID
2481 Details : While using HA for Lustre servers with Linux RAID, it is possible
2482 that MMP will not detect multiple mounts. To make this work we
2483 need to unplug the device queue in RAID when the MMP block is being
2484 written. Also while reading the MMP block, we should read it from
2485 disk and not the cached one.
2487 Severity : enhancement
2489 Description: open file using fid
2490 Details : A file can be opened using just its fid, like
2491 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
2494 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
2497 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
2498 Details : add proper locking for creation of jbd_4k slab cache
2502 Description: MMP check in ext3_remount() fails without displaying any error
2503 Details : When multiple mount protection fails during remount, proper error
2506 Severity : enhancement
2508 Description: Allow stripe size to be up to 4G-64k
2509 Details : Fix math logic to allow large stripe sizes.
2513 Description: add check for >8TB ldiskfs filesystems
2514 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
2515 Don't allow >8TB ldiskfs filesystems to be mounted without
2516 force_over_8tb mount option
2521 Descriptoin: OSS complains "too many blocks" and crashes
2522 Details : In filter_commitrw_write():
2523 1. lquota_chkquota();
2524 2. obdo_from_inode();
2525 3. lquota_pending_commit();
2526 "2." has changed uid and gid in this case, then caused
2527 lquota_chkquota() and lquota_pending_commit() to use
2528 different uid and gid. That is the root of the bug.
2532 Descriptoin: getstripe doesn't report stripe count = -1 sanely
2533 Details : Detect and print out the 'stripe count = -1' case properly.
2537 Description: Out or order replies might be lost on replay
2538 Details : In ptlrpc_retain_replayable_request if we cannot find retained
2539 request with tid smaller then one currently being added, add it
2540 to the start, not end of the list.
2542 --------------------------------------------------------------------------------
2544 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
2546 * Support for kernels:
2547 2.6.5-7.283 (SLES 9),
2548 2.6.9-55.EL (RHEL 4),
2549 2.6.16.46-0.14 (SLES 10),
2550 2.6.18.8 vanilla (kernel.org)
2551 * Client support for unpatched kernels:
2552 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2553 2.6.16 - 2.6.22 vanilla (kernel.org)
2554 * Due to recently discovered recovery problems, we do not recommend
2555 using patchless RHEL 4 clients with this or any earlier release.
2556 * Recommended e2fsprogs version: 1.39.cfs8
2557 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2558 * Starting with this release, the ldiskfs backing filesystem required
2559 by Lustre is now in its own package, lustre-ldiskfs. This package
2560 should be installed. It is versioned separately from Lustre and
2561 may be released separately in future.
2565 Description: block reactivating mgc import until all deactivates complete
2566 Details : Fix race when failing back MDT/MGS to itself (testing)
2569 Frequency : at statup only
2571 Description: mds_lov_synchronize race leads to various problems
2572 Details : simultaneous MDT->OST connections at startup can cause the
2573 sync to abort, leaving the OSC in a bad state.
2575 Severity : enhancement
2577 Description: add optional extra BUILD_VERSION info
2578 Details : add a new environment variable (namely LUSTRE_VERS) which allows
2579 to override the lustre version.
2582 Frequency : 2.6.18 servers only
2584 Description: ll_kern_mount() doesn't release the module reference
2585 Details : The ldiskfs module reference count never drops down to 0
2586 because ll_kern_mount() doesn't release the module reference.
2591 Description: server LBUG when using old ost_num_threads parameter
2592 Details : Accept the old ost_num_threads parameter but warn that it
2593 is deprecated, and fix an off-by-one error that caused an LBUG.
2598 Description: Transient SCSI error results in persistent IO issue
2599 Details : iobuf->dr_error is not reinitialized to 0 between two
2603 Frequency : sometimes when underlying device returns I/O errors
2605 Description: OSTs not going read-only during write failures
2606 Details : OSTs are not remounted read-only when the journal commit threads
2607 get I/O errors because fsfilt_ext3 calls journal_start/stop()
2608 instead of the ext3 wrappers.
2612 Description: poor connect scaling with increasing client count
2613 Details : Don't run filter_grant_sanity_check for more than 100 exports
2614 to improve scaling for large numbers of clients.
2617 Frequency : SLES10 only
2619 Description: sanity-quota.sh quotacheck failed: rc = -22
2620 Details : Quotas cannot be enabled on SLES10.
2623 Frequency : liblustre clients only
2625 Description: getdirentries does not give error when run on compute nodes
2626 Details : getdirentries does not fail when the size specified as an argument
2627 is too small to contain at least one entry
2629 Severity : enhancement
2631 Description: Add LNET router traceability for debug purposes
2632 Details : If a checksum failure occurs with a router as part of the
2633 IO path, the NID of the last router that forwarded the bulk data
2634 is printed so it can be identified.
2639 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
2640 Details : Due to a race condition, liblustre clients were occasionally
2641 evicted incorrectly.
2643 Severity : enhancement
2645 Description: lfs setstripe use optional parameters instead of postional
2648 Severity : enhancement
2650 Description: Nanosecond timestamp support for ldiskfs
2651 Details : The on-disk ldiskfs filesystem has added support for nanosecond
2652 resolution timestamps. There is not yet support for this at
2653 the Lustre filesystem level.
2656 Frequency : during server recovery
2658 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
2659 Details : request with rq_no_resend flag not awake l_wait_event if they get
2663 Frequency : nfs export on patchless client
2665 Description: connectathon hang when test nfs export over patchless client
2666 Details : Disconnected dentry cannot be found with lookup, so we do not need
2667 to unhash it or make it invalid
2670 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
2671 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
2672 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
2677 Description: open req refcounting wrong on reconnect
2678 Details : If reconnect happened between getting open reply from server and
2679 call to mdc_set_replay_data in ll_file_open, we will schedule
2680 replay for unreferenced request that we are about to free.
2681 Subsequent close will crash in variety of ways.
2682 Check that request is still eligible for replay in
2683 mdc_set_replay_data().
2688 Description: disable writes to filesystem when reading health_check file
2689 Details : the default for reading the health_check proc file has changed
2690 to NOT do a journal transaction and write to disk, because this
2691 can cause reads of the /proc file to hang and block HA state
2692 checking on a healthy but otherwise heavily loaded system. It
2693 is possible to return to the previous behaviour during configure
2694 with --enable-health-write.
2696 Severity : enhancement
2698 Description: 64-bit inode version
2699 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
2700 to the inode. This will be required for version-based recovery.
2705 Description: MDS fails to start if a duplicate client export is detected
2706 Details : in some rare cases it was possible for a client to connect to
2707 an MDS multiple times. Upon recovery the MDS would detect this
2708 and fail during startup. Handle this more gracefully.
2710 Severity : enhancement
2712 Description: Add -o localflock option to simulate old noflock
2714 Details : This will achieve local-only flock/fcntl locks
2720 Description: log_commit_thread vs filter_destroy race leads to crash
2721 Details : Take import reference before releasing llog record semaphore
2726 Description: Wrong request locking in request set processing
2727 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
2728 imp_delayed_list, in this place should be used imp_lock.
2731 Frequency : when reconnection
2733 Description: Grant Leak when osc reconnect to OST
2734 Details : When osc reconnect ost, OST(filter) should check whether it
2735 should grant more space to client by comparing fed_grant and
2736 cl_avail_grant, and return the granted space to client instead
2737 of "new granted" space, because client will call osc_init_grant
2738 to update the client grant space info.
2741 Frequency : when client reconnect to OST
2743 Description: Grant Leak when osc do resend and replay bulk write
2744 Details : When osc reconnect to OST, OST(filter)should clear grant info of
2745 bulk write request, because the grant info while be sync between
2746 OSC and OST when reconnect, and we should ignore the grant info
2747 these of resend/replay write req.
2752 Description: Grant space more than avaiable left space sometimes.
2753 Details : When then OST is about to be full, if two bulk writing from
2754 different clients came to OST. Accord the avaliable space of the
2755 OST, the first req should be permitted, and the second one
2756 should be denied by ENOSPC. But if the seconde arrived before
2757 the first one is commited. The OST might wrongly permit second
2758 writing, which will cause grant space > avaiable space.
2761 Frequency : when client is evicted
2763 Description: Grant might be wrongly erased when osc is evicted by OST
2764 Details : when the import is evicted by server, it will fork another
2765 thread ptlrpc_invalidate_import_thread to invalidate the
2766 import, where the grant will be set to 0. While the original
2767 thread will update the grant it got when connecting. So if
2768 the former happened latter, the grant will be wrongly errased
2769 because of this race.
2774 Description: Checking Stale with correct fid
2775 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
2776 to check whether it is stale, because sometimes, we want the
2777 enqueue happened anyway, and op_data.fid2 will not be initialized.
2779 Severity : enhancement
2781 Description: update patchless client
2782 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
2785 Frequency : only with 2.4 kernel
2787 Description: random memory corruption
2788 Details : size of struct ll_inode_info is to big for union inode.u and this
2789 can be cause of random memory corruption.
2794 Description: Memory leak in recovery
2795 Details : Lov_mds_md was not free in an error handler in mds_create_object.
2796 It should also check obd_fail before fsfilt_start, otherwise if
2797 fsfilt_start return -EROFS,(failover mds during mds recovery).
2798 then the req will return with repmsg->transno = 0 and rc = EROFS.
2799 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
2800 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
2801 be freed no matter whether fsfilt_commit success or not.
2804 Frequency : only with huge count clients
2806 Description: Prevents from taking the superblock lock in llap_from_page for
2808 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
2809 for llap_from_page call in ll_removepage prevents from taking the
2810 superblock lock for a soon died page.
2815 Description: Not check open intent error before release open handle
2816 Details : in some rare cases, the open intent error is not checked before
2817 release open handle, which may cause
2818 ASSERTION(open_req->rq_transno != 0), because it tries to release
2819 the failed open handle.
2824 Description: Set cat log bitmap only after create log success.
2825 Details : in some rare cases, the cat log bitmap is set too early. and it
2826 should be set only after create log success.
2830 Description: Accessing a block bevice can re-enable I/O when Lustre is
2831 tearing down a device.
2832 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
2837 Description: service threads may hog cpus when there are a lot of requests
2838 Details : Insert cond_resched to give other threads a chance to use some CPU
2843 Description: the cat log was not initialized in recovery
2844 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
2845 unlink log on mds will not be initialized until mds post
2846 recovery. And also in mds post recovery, the unlink log will
2847 initialization will be done asynchronausly, so there will be race
2848 between add unlink log and unlink log initialization.
2852 Description: brw_stats were being printed incorrectly
2853 Details : brw_stats were being printed as log2 but all of them were not
2854 recorded as log2. Also remove some code duplication arising from
2855 filter_tally_{read,write}.
2859 Frequency : rare, only in recovery.
2860 Description: ASSERTION(req->rq_type != LI_POISON) failed
2861 Details : imp_lock should be held while iterating over imp_sending_list for
2862 prevent destroy request after get timeout in ptlrpc_queue_wait.
2866 Description: replay-single.sh test 52 fails
2867 Details : A lock's skiplist need to be cleanup when it being unlinked
2868 from its resource list.
2872 Description: Short directio read returns full requested size rather than
2874 Details : Direct I/O operations should return actual amount of bytes
2875 transferred rather than requested size.
2877 Severity : enhancement
2879 Description: metadata RPC reduction (e.g. for rm performance)
2880 Details : decrease the amount of synchronous RPC between clients and servers
2881 by canceling conflicing lock before the operation on the client
2882 and packing thier handles into the main operation RPC to server.
2884 Severity : enhancement
2886 Description: Async OSC create to avoid the blocking unnecessarily.
2887 Details : If a OST has no remain object, system will block on the creating
2888 when need to create a new object on this OST. Now, ways use
2889 pre-created objects when available, instead of blocking on an
2890 empty osc while others are not empty. If we must block, we block
2891 for the shortest possible period of time.
2895 Description: improve handling recoverable errors
2896 Details : if request processig with error which can be recoverable on server
2897 request should be resend, otherwise page released from cache and
2900 Severity : enhancement
2902 Description: refine locking for avoid write wrong info into lov_objid file
2903 Details : fix possible races with add new target and write/update data in
2906 --------------------------------------------------------------------------------
2908 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
2913 Frequency : on some architectures
2915 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
2916 Details : Uninitialized flags sometimes cause configuration commands to
2920 Frequency : patchless clients only
2922 Description: missing __iget() symbol export
2923 Details : The __iget() symbol export is missing. To avoid the need for
2924 this on patchless clients the deathrow inode reaper is turned
2925 off, and we depend on the VM to clean up old inodes. This
2926 dependency was during via the fix for bug 12181.
2928 --------------------------------------------------------------------------------
2930 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
2932 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
2933 INTEROPERATE with older versions automatically. In many cases a
2934 special upgrade step is needed. Please read the
2935 user documentation before upgrading any part of a live system.
2936 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
2937 version will not interoperate with 1.6 betas before beta5 (1.5.95).
2938 * WARNING: Lustre configuration and startup changes are required with
2939 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
2943 Severity : enhancement
2945 Description: MountConf
2946 Details : Lustre configuration is now managed via mkfs and mount
2947 commands instead of lmc and lconf. New obd types (MGS, MGC)
2948 are added for dynamic configuration management. See
2949 https://mail.clusterfs.com/wikis/lustre/MountConf for
2952 Severity : enhancement
2954 Description: dynamic OST addition
2955 Details : OSTs can now be added to a live filesystem
2957 Severity : enhancement
2959 Description: startup order invariance
2960 Details : MDTs and OSTs can be started in any order. Clients only
2961 require the MDT to complete startup.
2963 Severity : enhancement
2965 Description: parallel, asynchronous orphan cleanup
2966 Details : orphan cleanup is now performed in separate threads for each
2967 OST, allowing parallel non-blocking operation.
2969 Severity : enhancement
2971 Description: optimized stripe assignment
2972 Details : stripe assignments are now made based on ost space available,
2973 ost previous usage, and OSS previous usage, in order to try
2974 to optimize storage space and networking resources.
2976 Severity : enhancement
2978 Description: Permanently set tunables
2979 Details : All writable /proc/fs/lustre tunables can now be permanently
2980 set on a per-server basis, at mkfs time or on a live system.
2982 Severity : enhancement
2984 Description: Lustre message v2
2985 Details : Add lustre message format v2.
2987 Severity : enhancement
2989 Description: client OST exclusion list
2990 Details : Clients can be started with a list of OSTs that should be
2991 declared "inactive" for known non-responsive OSTs.
2995 Description: SPEC SFS validation failure on NFS v2 over lustre.
2996 Details : Changes the blocksize for regular files to be 2x RPC size,
2997 and not depend on stripe size.
2999 Severity : enhancement
3001 Description: Multiple MD RPCs in flight.
3002 Details : Further unserialise some read-only MDS RPCs - learn about intents.
3003 To avoid overly-overloading MDS, introduce a limit on number of
3004 MDS RPCs in flight for a single client and add /proc controls
3005 to adjust this limit.
3007 Severity : enhancement
3009 Description: client read/write statistics
3010 Details : Add client read/write call usage stats for performance
3011 analysis of user processes.
3012 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
3013 file access. extents_stats shows chunk size distribution.
3014 extents_stats_per_process show chunk size distribution per
3017 Severity : enhancement
3019 Description: mds statistics
3020 Details : Add detailed mds operations statistics in
3021 /proc/fs/lustre/mds/*/stats.
3025 Description: Failure of copying files with lustre special EAs.
3026 Details : Client side always return success for setxattr call for lustre
3027 special xattr (currently only "trusted.lov").
3032 Description: Refcount LNET uuids
3033 Details : The global LNET uuid list grew linearly with every startup;
3034 refcount repeated list entries instead of always adding to
3037 Severity : enhancement
3039 Description: Dynamic service threads
3040 Details : Within a small range, start extra service threads
3041 automatically when the request queue builds up.
3044 Frequency : mixed-endian client/server environments
3046 Description: mixed-endian crashes
3047 Details : The new msg_v2 system had some failures in mixed-endian
3050 Severity : enhancement
3052 Description: Easy OST removal
3053 Details : OSTs can be permanently deactivated with e.g. 'lctl
3054 conf_param lustre-OST0001.osc.active=0'
3056 Severity : enhancement
3058 Description: MGS proc entries
3059 Details : Added basic proc entries for the MGS showing what filesystems
3062 Severity : enhancement
3064 Description: provide MGS failover
3065 Details : Added config lock reacquisition after MGS server failover.
3067 Severity : enhancement
3069 Description: add Linux 2.4 support
3070 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
3074 Description: a large application tries to do I/O to the same resource and dies
3075 in the middle of it.
3076 Details : Check the req->rq_arrival time after the call to
3077 ost_brw_lock_get(), but before we do anything about
3078 processing it & sending the BULK transfer request. This
3079 should help move old stale pending locks off the queue as
3080 quickly as obd_timeout.
3083 Frequency : when an incorrect nid is specified during startup
3085 Description: ptlrpc connect to non-existant node causes kernel crash
3086 Details : LNET can't be re-entered from an event callback, which
3087 happened when we expire a message after the export has been
3088 cleaned up. Instead, hand the zombie cleanup off to another
3091 Severity : enhancement
3093 Description: plain/inodebits lock performance improvement
3094 Details : Grouping plain/inodebits in granted list by their request modes
3095 and bits policy, thus improving the performance of search through
3099 Frequency : only if OST filesystem is corrupted
3101 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3102 Details : for a short time RPCs with bulk IO are in the replay list,
3103 but replay of bulk IOs is unimplemented. If the OST filesystem
3104 is corrupted due to disk cache incoherency and then replay is
3105 started it is possible to trip an assertion. Avoid putting
3106 committed RPCs into the replay list at all to avoid this issue.
3109 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3111 Description: System hang on startup
3112 Details : This bug allowed the liblustre (e.g. catamount) client to
3113 return to the app before handling all startup RPCs. This
3114 could leave the node unresponsive to lustre network traffic
3115 and manifested as a server ptllnd timeout.
3117 Severity : enhancement
3119 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3120 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
3121 boolean to control whether to print peer debug info when a
3122 client's RPC times out.
3125 Frequency : only for kernels with patches from Lustre below 1.4.3
3127 Description: Remove old rdonly API
3128 Details : Remove old rdonly API which unused from at least lustre 1.4.3
3131 Frequency : only for devices with external journals
3133 Description: Set external device read-only also
3134 Details : During a commanded failover stop, we set the disk device
3135 read-only while the server shuts down. We now also set any
3136 external journal device read-only at the same time.
3139 Frequency : when upgrading from 1.4 while trying to change parameters
3141 Description: The wrong (new) MDC name was used when setting parameters for
3142 upgraded MDT's. Also allows changing of OSC (and MDC)
3143 parameters if --writeconf is specified at tunefs upgrade time.
3146 Frequency : when setting specific ost indicies
3148 Description: QOS code breaks on skipped indicies
3149 Details : Add checks for missing OST indicies in the QOS code, so OSTs
3150 created with --index need not be sequential.
3152 Severity : enhancement
3154 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3155 Details : The uninit_groups feature works in conjunction with the kernel
3156 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3157 up the pass1 processing of e2fsck. This is a read-only feature
3158 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3159 mount filesystems that have had this feature enabled.
3161 Severity : enhancement
3163 Description: Improve multi-block allocation algorithm to avoid fragmentation
3164 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3165 allocation locality and avoid filesystem fragmentation.
3167 ------------------------------------------------------------------------------
3169 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
3171 * Support for kernels:
3172 2.4.21-47.0.1.EL (RHEL 3)
3173 2.6.5-7.283 (SLES 9)
3174 2.6.9-42.0.10.EL (RHEL 4)
3175 2.6.12.6 vanilla (kernel.org)
3176 2.6.16.27-0.9 (SLES 10)
3177 * Recommended e2fsprogs version: 1.39.cfs5
3179 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3183 Frequency : occasional, depends on client load and configuration
3184 Bugzilla : 12181, 12203
3185 Description: data loss for recently-modified files
3187 Details : In some cases it is possible that recently written or created
3188 files may not be written to disk in a timely manner (this should
3189 normally be within 30s unless client IO load is very high).
3190 The problem appears as zero-length files or files that are a
3191 multiple of 1MB in size after a client crash or client eviction
3192 that are missing data at the end of the file.
3194 This problem is more likely to be hit on clients where files are
3195 repeatedly created and unlinked in the same directory, clients
3196 have a large amount of RAM, have many CPUs, the filesystem has
3197 many OSTs, the clients are rebooted frequently, and/or the files
3198 are not accessed by other nodes after being written.
3200 The presence of the problem can be detected by looking at
3201 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
3202 smaller than the second (nr_unused) then dirty files will not
3203 be flushed automatically to disk. "sync; sleep 10" should be
3204 run several times on the node before unmounting it to update
3205 Lustre (this is also safe to run on nodes without this problem).
3207 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
3208 that can cause this same problem, so customers using that kernel
3209 also need to update the kernel in addition to Lustre. In order
3210 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
3212 It is normal that files written just before a client crash (less
3213 than 30s) may not yet have been flushed to disk, even for local
3217 Frequency : frequent on thin XT3 nodes
3219 Description: UUID collision on thin XT3 Linux nodes
3220 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
3221 randomly, since we relied on an insufficiently-seeded PRNG.
3226 Description: OSS hangs after "All ost request buffers busy"
3227 Details : A deadlock between quota and journal operations caused OSS
3228 hangs after printing "All ost request buffers busy."
3231 Frequency : always on liblustre builds
3233 Description: Cleanup compiler warnings on liblustre
3236 Frequency : always on liblustre builds on XT3
3238 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
3242 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
3243 > 32000 subdirectories
3244 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
3245 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
3246 subdirectory count crosses 32000. This will aid e2fsck to
3247 correctly handle more than 32000 subdirectories.
3250 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3252 Description: System hang on startup
3253 Details : This bug allowed the liblustre (e.g. catamount) client to
3254 return to the app before handling all startup RPCs. This
3255 could leave the node unresponsive to lustre network traffic
3256 and manifested as a server ptllnd timeout.
3258 Severity : enhancement
3260 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3261 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
3262 boolean to control whether to print peer debug info when a
3263 client's RPC times out.
3268 Description: make O_SYNC working on 2.6 kernels
3269 Details : 2.6 kernels use different method for mark pages for write,
3270 so need add a code to lustre for O_SYNC work.
3275 Description: Failure to close file and release space on NFS
3276 Details : Put inode details into lock acquired in ll_intent_file_open.
3277 Use mdc_intent_lock in ll_intent_open to properly
3278 detect all kind of errors unhandled by mdc_enqueue.
3283 Description: proc file read during shutdown sometimes raced obd removal,
3285 Details : Add lock to prevent obd access after proc file removal.
3288 Frequency : Only for files larger than 4GB on 32-bit clients.
3290 Description: improperly doing page alignment of locks
3291 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
3292 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3297 Description: RPCs being resent when they shouldn't be
3298 Details : Some RPCs that should not be resent are being resent. This
3299 can cause inconsistencies in the RPC state machine. Do not
3300 resend such requests.
3303 Frequency : rare, only with NFS export
3305 Description: Crash on NFS re-export node
3306 Details : under very unusual load conditions an assertion is hit in
3307 ll_intent_file_open()
3310 Frequency : only if OST filesystem is corrupted
3312 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3313 Details : for a short time RPCs with bulk IO are in the replay list,
3314 but replay of bulk IOs is unimplemented. If the OST filesystem
3315 is corrupted due to disk cache incoherency and then replay is
3316 started it is possible to trip an assertion. Avoid putting
3317 committed RPCs into the replay list at all to avoid this issue.
3322 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3323 Details : Large single O_DIRECT read and write calls can fail to allocate
3324 a sufficiently large buffer to process the request. In case of
3325 allocation failure the allocation is retried with a smaller
3326 buffer and broken into smaller requests.
3328 Severity : enhancement
3330 Description: Add -o localflock option to simulate old noflock behaviour.
3331 Details : This will achieve local-only flock/fcntl locks coherentness.
3336 Description: versioning check is incomplete
3337 Details : Checking the version difference of client vs. server, report
3338 error if the gap is too big.
3343 Description: add support PG_writeback bit
3344 Details : add support for PG_writeback bit for Lustre, for more carefull
3345 work with page cache in 2.6 kernel. This also fix some deadlocks
3346 and remove hack for work O_SYNC with 2.6 kernel.
3348 Severity : enhancement
3350 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3351 Details : The uninit_groups feature works in conjunction with the kernel
3352 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3353 up the pass1 processing of e2fsck. This is a read-only feature
3354 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3355 mount filesystems that have had this feature enabled.
3357 Severity : enhancement
3359 Description: Improve multi-block allocation algorithm to avoid fragmentation
3360 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3361 allocation locality and avoid filesystem fragmentation.
3363 ------------------------------------------------------------------------------
3365 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3367 * Support for kernels:
3368 2.6.9-42.0.3.EL (RHEL 4)
3369 2.6.5-7.276 (SLES 9)
3370 2.4.21-47.0.1.EL (RHEL 3)
3371 2.6.12.6 vanilla (kernel.org)
3372 2.6.16.21-0.8 (SLES10)
3373 * Recommended e2fsprogs version: 1.39.cfs2-0
3375 * The backwards-compatible /proc/sys/portals symlink has been removed
3376 in this release. Before upgrading, please ensure that you change
3377 any configuration scripts or /etc/sysctl.conf files that access
3378 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3379 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3380 in advance of the upgrade on any system running Lustre 1.4.6 or
3381 newer, since /proc/sys/lnet was added in that version.
3382 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3386 Frequency : only when quota is used
3388 Description: avoid scanning export list for quota master
3389 Details : Change the algorithms to avoid scanning export list in order
3390 to improve the efficiency.
3393 Frequency : MDS failover only, very rarely
3395 Description: "went back in time" messages on mds failover
3396 Details : The greatest transno may be lost when the current operation
3397 finishes with an error (transno==0) and the client's last_rcvd
3398 record is over-written. Save the greatest transno in the
3399 mds_last_transno for this case.
3402 Frequency : always for specific kernels and striping counts
3404 Description: client may get "Matching packet too big" without ACL support
3405 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
3406 when trying to access files in certain configurations. The
3407 clients should in fact be denied when mounting because they do
3408 not understand ACLs.
3411 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
3413 Description: many clients connecting with IO in progress causes connect timeouts
3414 Details : Avoid synchronous journal commits to avoid delays caused by many
3415 clients connecting/disconnecting when bulk IO is in progress.
3416 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
3417 OST_IO_PORTAL to avoid delays behind potentially many pending
3421 Frequency : occasionally with multiple writers to a single file
3423 Description: shared writes to file may result in wrong size reported by stat()
3424 Details : Allow growing of kms when extent lock is cancelled
3427 Frequency : always with random mmap IO to multi-striped file
3429 Description: mmap write might be lost if we are writing to a 'hole' in stripe
3430 Details : Only if the hole is at the end of OST object so that kms is too
3431 small. Fix is to increase kms accordingly in ll_nopage.
3434 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
3436 Description: writes to a missing object would leak memory on the OST
3437 Details : If there is an inconsistency between the MDS and OST filesystems,
3438 such that the MDS references an object that doesn't exist, writes
3439 to that object will leak memory due to incorrect cleanup in the
3440 error handling path, eventually running out of memory on the OST.
3445 Description: Creating too long symlink causes lustre errors
3446 Details : Check symlink and name lengths before sending requests to MDS.
3449 Frequency : only if flock is enabled (not on by default)
3451 Description: posix locks not released on fd closure on 2.6.9+
3452 Details : We failed to add posix locks to list of inode locks on 2.6.9+
3453 kernels, this caused such locks not to be released on fd close and
3454 then assertions on fs unmount about still used locks.
3457 Frequency : MDS failover only, very rarely
3459 Description: clients may get ASSERTION(granted_lock != NULL)
3460 Details : When request was taking a long time, and a client was resending
3461 a getattr by name lock request. The were multiple lock requests
3462 with the same client lock handle and
3463 mds_getattr_name->fixup_handle_for_resent_request found one of the
3464 lock handles but later failed with ASSERTION(granted_lock != NULL).
3469 Description: handle->h_buffer_credits > 0, assertion failure
3470 Details : h_buffer_credits is zero after truncate, causing assertion
3471 failure. This patch extends the transaction or creates a new
3475 Frequency : NFS re-export or patchless client
3476 Bugzilla : 11179, 10796
3477 Description: Crash on NFS re-export node (__d_move)
3478 Details : We do not want to hash the dentry if we don't have a lock.
3479 But if this dentry is later used in d_move, we'd hit uninitialised
3480 list head d_hash, so we just do this to init d_hash field but
3481 leave dentry unhashed.
3484 Frequency : NFS re-export or patchless client
3486 Description: NFS exports has problem with symbolic link
3487 Details : lustre client didn't properly install dentry when re-exported
3488 to NFS or running patchless client.
3491 Frequency : NFS re-export or patchless client
3493 Description: Various nfs/patchless fixes.
3494 Details : fixes reuse disconected alias for lookup process - this fixes
3495 warning "find_exported_dentry: npd != pd",
3496 fix permission error with open files at nfs.
3497 fix apply umask when do revalidate.
3500 Frequency : occasional
3502 Description: Crash on NFS re-export node
3503 Details : calling clear_page() on the wrong pointer triggered oops in
3504 generic_mapping_read().
3507 Frequency : rarely, using O_DIRECT IO
3509 Description: unaligned directio crashes client with LASSERT
3510 Details : check for unaligned buffers before trying any requests.
3513 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
3515 Description: stale data returned from RAID cache
3516 Details : If only a small amount of IO is done to the RAID device before
3517 reading it again it is possible to get stale data from the RAID
3518 cache instead of reading it from disk.
3521 Frequency : always for sles10 kernel
3523 Description: sles10 support
3524 Details : ll_follow_link: compile fixes and using of nd_set_link
3525 under newer kernels.
3528 Frequency : depends on arch, kernel and compiler version, always on sles10
3531 Description: recursive or deep enough symlinks cause stack overflow
3532 Details : getting rid of large stack-allocated variable in
3536 Frequency : depends on hardware
3538 Description: lustre write performance loss in the SLES10 kernel
3539 Details : the performance loss is caused by using of write barriers in the
3540 ext3 code. The SLES10 kernel turns barrier support on by
3541 default. The fix is to undo that change for ldiskfs.
3543 ------------------------------------------------------------------------------
3545 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
3547 * Support for kernels:
3548 2.6.9-42.0.3EL (RHEL 4)
3549 2.6.5-7.276 (SLES 9)
3550 2.4.21-47.0.1.EL (RHEL 3)
3551 2.6.12.6 vanilla (kernel.org)
3555 Frequency : quota enabled and large files being deleted
3557 Description: releasing more than 4GB of quota at once hangs OST
3558 Details : If a user deletes more than 4GB of files on a single OST it
3559 will cause the OST to spin in an infinite loop. Release
3560 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3565 Description: statfs data retrieved from /proc may be stale or zero
3566 Details : When reading per-device statfs data from /proc, in the
3567 {kbytes,files}_{total,free,avail} files, it may appear
3568 as zero or be out of date.
3571 Frequency : systems with MD RAID1 external journal devices
3573 Description: lconf's call to blkid is confused by RAID1 journal devices
3574 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3575 of returning all block devices that match the journal UUID.
3578 Frequency : always, for aggregate stripe size over 4GB
3580 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
3581 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
3582 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
3585 Frequency : always if "lfs find" used on a local file/directory
3587 Description: "lfs find" segfaults if used on a local file/directory
3588 Details : The case where a directory component was not specified wasn't
3589 handled correctly. Handle this properly.
3592 Frequency : always on ppc64
3594 Description: the write to an ext3 filesystem mounted with mballoc got stuck
3595 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3596 perform endianness conversion.
3599 Frequency : rarely (truncate to non-zero file size after write under load)
3600 Bugzilla : 10730, 10687
3601 Description: Files padded with zeros to next 4K multiple
3602 Details : With filesystems mounted using the "extents" option (2.6 kernels)
3603 it is possible that files that are truncated to a non-zero size
3604 immediately after being written are filled with zero bytes beyond
3605 the truncated size. No file data is lost.
3607 Severity : enhancement
3609 Description: Allow recovery/failover for liblustre clients.
3610 Details : liblustre clients were unaware of failover configurations until
3613 Severity : enhancement
3615 Description: user file locks should fail when not mounting with flock option
3616 Details : Set up an error-returning stub in ll_file_operations.lock field
3617 to prevent incorrect behaviour when client is mounted without
3618 flock option. Also, set up properly f_op->flock field for
3622 Frequency : always on ia64
3624 Description: "lfs df" loops on printing out MDS statfs information
3625 Details : The obd_ioctl_data was not initialized and in some systems
3626 this caused a failure during the ioctl that did not return
3627 an error. Initialize the struct and return an error on failure.
3630 Frequency : SLES 9 only
3632 Description: Error of copying files with lustre special EAs as root
3633 Details : Client side always return success for setxattr call for lustre
3634 special xattr (currently only "trusted.lov").
3637 Frequency : rarely on clusters with both ia64+i386 clients
3639 Description: ia64+i686 clients doing shared IO on the same file may LBUG
3640 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
3641 clients are doing concurrent writes to the same file it is
3642 possible that the ia64 clients may LASSERT because the OST
3643 extent locks are not PAGE_SIZE aligned. Ensure that grown
3644 locks are always aligned on the request boundary.
3647 Frequency : specific use, occasional
3649 Description: Overwriting in use executable truncates on-disk binary image
3650 Details : If one node attempts to overwrite an executable in use by
3651 another node, we now correctly return ETXTBSY instead of
3652 truncating the file.
3654 Severity : enhancement
3656 Description: Async OSC create to avoid the blocking unnecessarily.
3657 Details : If a OST has no remain object, system will block on the creating
3658 when need to create a new object on this OST. Now, ways use
3659 pre-created objects when available, instead of blocking on an
3660 empty osc while others are not empty. If we must block, we block
3661 for the shortest possible period of time.
3666 Description: chmod on Lustre root is propagated to other clients
3667 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
3668 invalid by the follow_mount time.
3671 Frequency : liblustre clients only
3673 Description: Race in 'instant cancel' lock handling could lead to such locks
3674 never to be granted in case of SMP MDS
3675 Details : Do not destroy not yet granted but cbpending locks in
3679 Frequency : replay/resend of open
3681 Description: non null lock assetion failure in mds_intent_policy
3682 Details : Trying to replay/resend lockless open requests resulted in
3683 mds_open() returning 0 with no lock. Now it sets a flag if
3684 a lock is going to be returned.
3686 Severity : enhancement
3688 Description: Checksum enhancements
3689 Details : New checksum enhancements allow for resending RPCs that failed
3692 Severity : enhancement
3694 Description: Tunables on number of dirty pages in cacche
3695 Details : Allow to set limit on number of dirty pages cached.
3700 Description: client crash on unmount - lock still has references
3701 Details : In some error handling cases it was possible to leak a lock
3702 reference on a client while accessing a file. This was not
3703 harmful to the client during operation, but would cause the
3704 client to crash when the filesystem is unmounted.
3707 Frequency : specific case, rare
3709 Description: ETXTBSY on mds though file not in use by client
3710 Details : ETXTBSY is no longer incorrectly returned when attempting to
3711 chmod or chown a directory that the user previously tried to
3712 execute or a currently-executing binary.
3715 Frequency : extremely rare except on liblustre-based clients
3717 Description: Lustre space not freed when files are deleted
3718 Details : Clean up open-unlinked files after client eviction. Previously
3719 the unlink was skipped and the files remained as orphans.
3724 Description: OST failure "would be an LBUG" in waiting_locks_callback()
3725 Details : In some cases it was possible to send a blocking callback to
3726 a client doing a glimpse, even though that client didn't get
3727 a lock granted. When the glimpse lock is cancelled on the OST
3728 the freed lock is left on the waiting list and corrupted the list.
3731 Frequency : all core dumps
3733 Description: Broke core dumps to lustre
3734 Details : Negative dentry may be unhashed if parent does not have UPDATE
3735 lock, but some callers, e.g. do_coredump, expect dentry to be
3736 hashed after successful create, hash it in ll_create_it.
3738 ------------------------------------------------------------------------------
3740 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
3742 * Support for kernels:
3743 2.6.9-42.0.2.EL (RHEL 4)
3744 2.6.5-7.276 (SLES 9)
3745 2.4.21-40.EL (RHEL 3)
3746 2.6.12.6 vanilla (kernel.org)
3750 Frequency : always on RHEL 3
3752 Description: Number of open files grows over time
3753 Details : The number of open files grows over time, whether or not
3754 Lustre is started. This was due to a filp leak introduced
3755 by one of our kernel patches.
3757 ------------------------------------------------------------------------------
3759 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
3761 * Support for kernels:
3762 2.6.9-42.EL (RHEL 4)
3763 2.6.5-7.276 (SLES 9)
3764 2.4.21-40.EL (RHEL 3)
3765 2.6.12.6 vanilla (kernel.org)
3770 Bugzilla : 5719, 9635, 9792, 9684
3771 Description: OST (or MDS) trips assertions in (re)connection under heavy load
3772 Details : If a server is under heavy load and cannot reply to new
3773 connection requests before the client resends the (re)connect,
3774 the connection handling code can behave badly if two service
3775 threads are concurrently handing separate (re)connections from
3776 the same client. Add better locking to the connection handling
3777 code, and ensure that only a single connection will be processed
3778 for a given client UUID, even if the lock is dropped.
3780 Severity : enhancement
3782 Description: add TCP zero-copy support to kernel
3783 Details : Add support to the kernel TCP stack to allow zero-copy bulk
3784 sends if the hardware supports scatter-gather and checksumming.
3785 This allows socklnd to do client-write and server-read more
3786 efficiently and reduce CPU utilization from skbuf copying.
3789 Frequency : only if NFS exporting from client
3791 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
3792 Details : If setting attributes on a file created under NFS that had
3793 never been opened it would be possible to oops the client
3794 if the file had no objects.
3797 Frequency : always for liblustre
3799 Description: liblustre client does MDS+OSTs setattr RPC for each write
3800 Details : When doing a write from a liblustre client, the client
3801 incorrectly issued an RPC to the MDS and each OST the file was
3802 striped over in order to update the timestamps. When writing
3803 with small chunks and many clients this could overwhelm the MDS
3804 with RPCs. In all cases it would slow down the write because
3805 these RPCs are unnecessary.
3807 Severity : enhancement
3809 Description: allow number of MDS service threads to be changed at module load
3810 Details : It is now possible to change the number of MDS service threads
3811 running. Adding "options mds mds_num_threads={N}" to the MDS's
3812 /etc/modprobe.conf will set the number of threads for the next
3813 time Lustre is restarted (assuming the "mds" module is also
3814 reloaded at that time). The default number of threads will
3815 stay the same, 32 for most systems.
3820 Description: OST crash if filesystem is unformatted or corrupt
3821 Details : If an OST is started on a device that has never been formatted
3822 or if the filesystem is corrupt and cannot even mount then the
3823 error handling cleanup routines would dereference a NULL pointer.
3828 Description: NULL pointer deref in llap_from_page.
3829 Details : get_cache_page_nowait can return a page with NULL (or otherwise
3830 incorrect) mapping if the page was truncated/reclaimed while it was
3831 searched for. Check for this condition and skip such pages when
3832 doing readahead. Introduce extra check to llap_from_page() to
3833 verify page->mapping->host is non-NULL (so page is not anonymous).
3836 Frequency : Sometimes when using sys_sendfile
3838 Description: "page not covered by a lock" warnings from ll_readpage
3839 Details : sendfile called ll_readpage without right page locks present.
3840 Now we introduced ll_file_sendfile that does necessary locking
3841 around call to generic_file_sendfile() much like we do in
3845 Frequency : with certain MDS communication failures at client mount time
3847 Description: NULL pointer deref after failed client mount
3848 Details : a client connection request may delayed by the network layer
3849 and not be sent until after the PTLRPC layer has timed out the
3850 request. If the client fails the mount immediately it will try
3851 to clean up before the network times out the request. Add a
3852 reference from the request import to the obd device and delay
3853 the cleanup until the network drops the request.
3856 Frequency : occasionally during client (re)connect
3858 Description: assertion failure during client (re)connect
3859 Details : processing a client connection request may be delayed by the
3860 client or server longer than the client connect timeout. This
3861 causes the client to resend the connection request. If the
3862 original connection request is replied in this interval, the
3863 client may trip an assertion failure in ptlrpc_connect_interpret()
3864 which thought it would be the only running connect process.
3867 Frequency : only with obd_echo servers and clients that are rebooted
3869 Description: kernel BUG accessing uninitialized data structure
3870 Details : When running an obd_echo server it did not start the ping_evictor
3871 thread, and when a client was evicted an uninitialized data
3872 structure was accessed. Start the ping_evictor in the RPC
3873 service startup instead of the OBD startup.
3875 Severity : enhancement
3876 Bugzilla : 10193 (patchless)
3877 Description: Remove dependency on various unexported kernel interfaces.
3878 Details : No longer need reparent_to_init, exit_mm, exit_files,
3879 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
3882 Frequency : rare (only users of deprecated and unsupported LDAP config)
3884 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
3885 Details : LDAP apparently contains 'lustreName' attributes instead of
3886 'name'. A simple remapping of the name is sufficient.
3889 Frequency : rare (only with non-default dump_on_timeout debug enabled)
3891 Description: waiting_locks_callback trips kernel BUG if client is evicted
3892 Details : Running with the dump_on_timeout debug flag turned on makes
3893 it possible that the waiting_locks_callback() can try to dump
3894 the Lustre kernel debug logs from an interrupt handler. Defer
3895 this log dumping to the expired_lock_main() thread.
3897 Severity : enhancement
3899 Description: Support NFS exporting on 2.6 kernels.
3900 Details : Implement non-rawops metadata methods for NFS server to use without
3901 changing NFS server code.
3904 Frequency : very rare (synthetic metadata workload only)
3906 Description: two racing renames might cause an MDS thread to deadlock
3907 Details : Running the "racer" program may cause one MDS thread to rename
3908 a file from being the source of a rename to being the target of
3909 a rename at exactly the same time that another thread is doing
3910 so, and the second thread has already enqueued these locks after
3911 doing a lookup of the target and is trying to relock them in
3912 order. Ensure that we don't try to re-lock the same resource.
3915 Frequency : only very large systems with liblustre clients
3917 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
3918 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
3919 all OSTs in parallel. This allows the eviction of stale liblustre
3920 clients to proceed much faster than if they were done in series,
3921 and also offers similar improvements for other set_info RPCs.
3926 Description: excessive CPU usage during initial read phase on client
3927 Details : During the initial read phase on a client, it would agressively
3928 retry readahead on the file, consuming too much CPU and impacting
3929 performance (since 1.4.5.8). Improve the readahead algorithm
3930 to avoid this, and also improve some other common cases (read
3931 of small files in particular, where "small" is files smaller than
3932 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
3937 Description: MDS crash when receiving packet with unknown intent.
3938 Details : Do not LBUG in unknown intent case, just return -EFAULT
3940 Severity : enhancement
3941 Bugzilla : 9293, 9385
3942 Description: MDS RPCs are serialised on client. This is unnecessary for some.
3943 Details : Do not serialize getattr (non-intent version) and statfs.
3946 Frequency : occasional, when OST network is overloaded/intermittent
3948 Description: client evicted by OST after bulk IO timeout
3949 Details : If a client sends a bulk IO request (read or write) the OST
3950 may evict the client if it is unresposive to its data GET/PUT
3951 request. This is incorrect if the network is overloaded (takes
3952 too long to transfer the RPC data) or dropped the OST GET/PUT
3953 request. There is no need to evict the client at all, since
3954 the pinger and/or lock callbacks will handle this, and the
3955 client can restart the bulk request.
3958 Frequency : Always when mmapping file with no objects
3960 Description: client crashes when mmapping file with no objects
3961 Details : Check that we actually have objects in a file before doing any
3962 operations on objects in ll_vm_open, ll_vm_close and
3968 Description: Request leak when working with deleted CWD
3969 Details : Introduce advanced request refcount tracking for requests
3970 referenced from lustre intent.
3972 Severity : Enhancement
3974 Description: Cache open file handles on client.
3975 Details : MDS now will return special lock along with openhandle, if
3976 requested and client is allowed to hold openhandle, even if unused,
3977 until such a lock is revoked. Helps NFS a lot, since NFS is opening
3978 closing files for every read/write openration.
3980 Severity : Enhancement
3982 Description: Cache open negative dentries on client when possible.
3983 Details : Guard negative dentries with UPDATE lock on parent dir, drop
3984 negative dentries on lock revocation.
3989 Description: Remounting a client read-only wasn't possible with a zconf mount
3990 Details : It wasn't possible to remount a client read-only with llmount.
3992 Severity : enhancement
3993 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
3994 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
3995 MPICH is included to add Lustre-specific ADIO interfaces.
3996 This is based closely on the UFS ADIO layer and only differs
3997 in file creation, in order to allow the OST striping to be set.
3998 This is user-contributed code and not supported by CFS.
4003 Description: extended inode attributes (immutable, append-only) work improperly
4004 when 2.4 and 2.6 kernels are used on client/server or vice versa
4005 Details : Introduce kernel-independent values for these flags.
4007 Severity : enhancement
4010 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
4011 Details : Many of the /proc/ tunables can only be tuned at a megabyte
4012 granularity. Now, Fractional MB granularity is be supported,
4013 this is very useful for low memory system.
4015 Severity : enhancement
4017 Description: Getattr by fid
4018 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
4019 locks, avoids extra getattr rpc requests to MDS, allows '/' to
4020 have locks and avoids getattr rpc requests for it on every stat.
4023 Frequency : Always, for filesystems larger than 2TB
4025 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4026 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
4027 filesystems larger than 2TB to be mounted with mballoc enabled.
4030 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
4032 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4033 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
4034 filesystems larger than 2TB is mounted then the kernel will
4035 silently corrupt the start of the filesystem. CONFIG_LBD is
4036 enabled for all CFS-supported kernels, but the possibility of
4037 this happening with a modified kernel config exists.
4039 Severity : enhancement
4041 Description: add client O_DIRECT support for 2.6 kernels
4042 Details : It is now possible to do O_DIRECT reads and writes to files
4043 in the Lustre client mountpoint on 2.6 kernel clients.
4045 Severity : enhancement
4047 Description: parallel glimpse, setattr, statfs, punch, destroy requests
4048 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
4049 parallel, not waiting for response from every OST before sending
4050 a rpc to the next OST.
4055 Description: setattr vs write race when updating file timestamps
4056 Details : Client processes that update a file timestamp into the past
4057 right after writing to the file (e.g. tar) it is possible that
4058 the updated file modification time can be reset to the current
4059 time due to a race between processing the setattr and write RPC.
4061 Severity : enhancement
4063 Description: Bring 'lfs find' closer in line with regular Linux find.
4064 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
4065 -print0 options and obtains all the needed info through the lustre
4068 Severity : enhancement
4070 Description: support up to 1024 configured devices on one node
4071 Details : change obd_dev array from statically allocated to dynamically
4072 allocated structs as they are first used to reduce memory usage
4077 Description: Flush dirty partially truncated pages during truncate
4078 Details : Immediatelly flush partially truncated pages in filter_setattr,
4079 this way we completely avoid having any pages in page cache on OST
4080 and can retire ugly workarounds during writes to flush such pages.
4085 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
4086 Details : protect inode from truncation within vfs_unlink() context
4087 just take a reference before calling vfs_unlink() and release it
4088 when parent's i_sem is free.
4093 Description: last_id value checked outside lock on OST caused LASSERT failure
4094 Details : If there were multiple MDS->OST object precreate requests in
4095 flight, it was possible that the OST's last object id was checked
4096 outside a lock and incorrectly tripped an assertion. Move checks
4097 inside locks, and discard old precreate requests.
4100 Frequency : always, if extents are used on OSTs
4102 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
4103 Details : OSTs using the extents format would not zero the high 16 bits of
4104 the index physical block number. This is not a problem for any
4105 OST filesystems smaller than 16TB, and no kernels support ext3
4106 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
4107 new/modified files) and can be fixed for existing filesystems
4108 with e2fsprogs-1.39-cfs1.
4113 Description: import connection selection may be incorrect if timer wraps
4114 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
4115 import connections to be ignored if the 32-bit jiffies counter
4116 wraps. Use a 64-bit jiffies counter.
4119 Frequency : very large clusters immediately after boot
4121 Description: LNET request buffers exhausted under heavy short-term load
4122 Details : If a large number of client requests are generated on a service
4123 that has previously never seen so many requests it is possible
4124 that the request buffer growth cannot keep up with the spike in
4125 demand. Instead of dropping incoming requests, they are held in
4126 the LND until the RPC service can accept more requests.
4129 Frequency : Sometimes during replay
4131 Description: Assertion failure in ll_local_open after replay.
4132 Details : If replay happened on an open request reply before we were able
4133 to set replay handler, reply will become not swabbed tripping the
4134 assertion in ll_local_open. Now we set the handler right after
4135 recognising of open request
4138 Frequency : very rare
4140 Description: kernel reports "badness in vsnprintf"
4141 Details : Reading from the "recovery_status" /proc file in small chunks
4142 may cause a negative length in lprocfs_obd_rd_recovery_status()
4143 call to vsnprintf() (which is otherwise harmless). Exit early
4144 if there is no more space in the output buffer.
4146 Severity : enhancement
4148 Description: clear OBD RPC statistics by writing to them
4149 Details : It is now possible to clear the OBD RPC statistics by writing
4150 to the "stats" file.
4155 Description: Client mtime is not the same on different clients after utimes
4156 Details : In some cases, the client was using the utimes() syscall on
4157 a file cached on another node. The clients now validate the
4158 ctime from the MDS + OSTs to determine which one is right.
4163 Description: Inability to activate failout mode
4164 Details : lconf script incorrectly assumed that in pythong string's numeric
4165 value is used in comparisons.
4168 Frequency : always with multiple stripes per file
4170 Description: Inefficient object allocation for mutli-stripe files
4171 Details : When selecting which OSTs to stripe files over, for files with
4172 a stripe count that divides evenly into the number of OSTs,
4173 the MDS is always picking the same starting OST for each file.
4174 Return the OST selection heuristic to the original design.
4179 Description: mount failures may take full timeout to return an error
4180 Details : Under some heavy load conditions it is possible that a
4181 failed mount can wait for the full obd_timeout interval,
4182 possibly several minutes, before reporting an error.
4183 Instead return an error as soon as the status is known.
4185 Frequency : quota enabled and large files being deleted
4187 Description: releasing more than 4GB of quota at once hangs OST
4188 Details : If a user deletes more than 4GB of files on a single OST it
4189 will cause the OST to spin in an infinite loop. Release
4190 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4195 Description: statfs data retrieved from /proc may be stale or zero
4196 Details : When reading per-device statfs data from /proc, in the
4197 {kbytes,files}_{total,free,avail} files, it may appear
4198 as zero or be out of date.
4201 Frequency : systems with MD RAID1 external journal devices
4203 Description: lconf's call to blkid is confused by RAID1 journal devices
4204 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4205 of returning all block devices that match the journal UUID.
4208 Frequency : always, for aggregate stripe size over 4GB
4210 Description: assertion fails when trying to use 4GB stripe size
4211 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
4212 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
4215 Frequency : always on ppc64
4217 Description: the first write on an ext3 filesystem with mballoc got stuck
4218 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4219 perform endianness conversion.
4221 ------------------------------------------------------------------------------
4223 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4225 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
4226 INTEROPERATE with older versions automatically. Please read the
4227 user documentation before upgrading any part of a live system.
4228 * WARNING: Lustre networking configuration changes are required with
4229 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
4232 * Support for kernels:
4233 2.6.9-22.0.2.EL (RHEL 4)
4234 2.6.5-7.244 (SLES 9)
4235 2.6.12.6 vanilla (kernel.org)
4238 Severity : enhancement
4239 Bugzilla : 7981/8208
4240 Description: Introduced Lustre Networking (LNET)
4241 Details : LNET is new networking infrastructure for Lustre, it includes
4242 a reorganized network configuration mode (see the user
4243 documentation for full details) as well as support for routing
4244 between different network fabrics. Lustre Networking Devices
4245 (LNDs) for the supported network fabrics have also been
4246 created for this new infrastructure.
4248 Severity : enhancement
4249 Description: Introduced Access control lists
4250 Details : clients can set ACLs on files and directories in order to have
4251 more fine-grained permissions than the standard Unix UGO+RWX.
4252 The MDS must be started with the "-o acl" mount option.
4254 Severity : enhancement
4255 Description: Introduced filesystem quotas
4256 Details : Administrators may now establish per-user quotas on the
4259 Severity : enhancement
4261 Description: Configuration change for the XT3
4262 The PTLLND is now used to run Lustre over Portals on the XT3
4263 The configure option(s) --with-cray-portals are no longer used.
4264 Rather --with-portals=<path-to-portals-includes> is used to
4265 enable building on the XT3. In addition to enable XT3 specific
4266 features the option --enable-cray-xt3 must be used.
4271 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
4272 Details : A race between allocating a new llcd and re-getting the llcd_lock
4273 allowed another thread to grab newly-allocated llcd.
4275 Severity : enhancement
4277 Description: 2.6 OST async journal commit and locking fix to improve performance
4278 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
4279 2.6 kernels are now async as they already were in 2.4 kernels so
4280 that they can commit concurrently with the network bulk transfer.
4281 For block-allocated files the filter allocation semaphore is held
4282 to avoid filesystem fragmentation during allocation. BKL lock
4283 removed for 2.6 xattr operations where it is no longer needed.
4288 Description: lconf incorrectly determined whether two IP networks could talk
4289 Details : In some more complicated routing and multiple-network
4290 configurations, lconf will avoid trying to make a network
4291 connection to a disjoint part of the IP space. It was doing the
4292 math incorrectly for one set of cases.
4297 Description: Fix for potential infinite loop processing records in an llog.
4298 Details : If an llog record is corrupted/zeroed, it is possible to loop
4299 forever in llog_process(). Validate the llog record length
4300 and skip the remainder of the block on error.
4303 Frequency : occasional (liblustre only)
4305 Description: liblustre could not open files whose last component is a symlink
4306 Details : sysio_path_walk() would incorrectly pass the open intent to
4307 intermediate path components.
4310 Frequency : rare (liblustre only with non-standard tuning)
4311 Bugzilla : 7201 (7350)
4312 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4313 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4314 after it has been released, only do this for referenced locks
4316 Severity : enhancement
4318 Description: specifying an (invalid) directory default stripe_size of -1
4319 would reset the directory default striping
4320 Details : stripe_size -1 was used internally to signal directory stripe
4321 removal, now use "all default" to signal dir stripe removal
4322 as a directory striping of "all default" is not useful
4325 Frequency : common for large clusters running liblustre clients
4327 Description: doing an ls when liblustre clients are running is slow
4328 Details : sending a glimpse AST to a liblustre client waits for every AST
4329 to time out, as liblustre clients will not respond. Since they
4330 cannot cache data we refresh the OST lock LVB from disk instead.
4332 Severity : enhancement
4334 Description: doing an ls at the same time as file IO can be slow
4335 Details : enqueue and other "small" requests can be blocked behind many
4336 large IO requests. Create a new OST IO portal for non-IO
4337 requests so they can be processed faster.
4340 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4342 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4343 created file with mode 000
4344 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4345 allow an ftruncate() as a non-root user to a file with mode 000.
4346 Lustre now respects this flag to disable mode checks when
4347 truncating a file owned by the user
4350 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4353 Description: Revoking locks from clients that went dead or catatonic might take
4355 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4356 cancellation of such locks instant on servers without waiting for
4357 any reply from clients. Clients drops these locks when cancel
4358 notification from server is received without replying.
4361 Frequency : liblustre-only, when liblustre client dies or becomes busy
4363 Description: Doing ls on Linux clients can take a long time with active
4365 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4366 granting such locks to it in the first place if possible. Locks
4367 are taken by proxy on the OST during the read or write and
4368 dropped immediately afterward. Add connect flags handling, do
4369 not grant locks to liblustre clients for glimpse ASTs.
4371 Severity : enhancement
4373 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4374 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4375 sequential reads and behaves badly with applications doing random
4376 reads. Improve it by reading ahead at least read region, and
4377 avoiding excessive large RPC for small reads.
4379 Severity : enhancement
4381 Description: Creating more than 1000 files for a single job may cause a load
4382 imbalance on the OSTs if there are also a large number of OSTs.
4383 Details : qos_prep_create() uses an OST index reseed value that is an
4384 even multiple of the number of available OSTs so that if the
4385 reseed happens in the middle of the object allocation it will
4386 still utilize the OSTs as uniformly as possible.
4391 Description: OST or MDS may oops in ping_evictor_main()
4392 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
4393 but doesn't restart at beginning of obd_exports_timed list
4396 Severity : enhancement
4398 Description: improve by-nid export eviction on the MDS and OST
4399 Details : allow multiple exports with the same NID to be evicted at one
4400 time without re-searching the exports list.
4403 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
4405 Description: MDS may oops in groups_free()
4406 Details : in rare race conditions a newly allocated group_info struct is
4407 freed again, and this can be NULL. The 2.4 compatibility code
4408 for groups_free() checked for a NULL pointer, but 2.6 did not.
4411 Frequency : common for liblustre clients doing little filesystem IO
4412 Bugzilla : 9352, 7313
4413 Description: server may evict liblustre clients accessing contended locks
4414 Details : if a client is granted a lock or receives a completion AST
4415 with a blocking AST already set it would not reply to the AST
4416 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
4419 Frequency : lfs setstripe, only systems with more than 160 OSTs
4421 Description: unable to set striping with a starting offset beyond OST 160
4422 Details : llapi_create_file() incorrectly limited the starting stripe
4423 index to the maximum single-file stripe count.
4426 Frequency : LDAP users only
4428 Description: lconf did not handle in-kernel recovery with LDAP properly
4429 Details : lconf/LustreDB get_refs() is searching the wrong namespace
4431 Severity : enhancement
4433 Description: bind OST threads to NUMA nodes to improve performance
4434 Details : all OST threads are uniformly bound to CPUs on a single NUMA
4435 node and do their allocations there to localize memory access
4437 Severity : enhancement
4439 Description: llmount can determine client NID directly from Myrinet (GM)
4440 Details : the client NID code from gmnalnid was moved directly into
4441 llmount, removing the need to use this or specifying the
4442 client NID explicitly when mounting GM clients with zeroconf
4445 Frequency : if client is started with down MDS
4447 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
4448 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
4449 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
4450 ptlrpc_import_delayed_req() because we didn't check if the
4451 request was interrupted, and we also didn't break out of the
4452 event loop if there was no timeout
4457 Description: data loss during non-page-aligned writes to a single file from
4458 both multiple nodes and multiple threads on one node at same time
4459 Details : updates to KMS and lsm weren't protected by common lock. Resulting
4460 inconsistency led to false short-reads, that were cached and later
4461 used by ->prepare_write() to fill in partially written page,
4462 leading to data loss.
4465 Frequency : always, if lconf --abort_recovery used
4467 Description: lconf --abort_recovery fails with 'Operation not supported'
4468 Details : lconf was attempting to abort recovery on the MDT device and not
4471 Severity : enhancement
4473 Description: remove cleanup logs
4474 Details : replace lconf-generated cleanup logs with lustre internal
4475 cleanup routines. Eliminates the need for client-cleanup and
4478 Severity : enhancement
4480 Description: add support for EAs (user and system) on lustre filesystems
4481 Details : it is now possible to store extended attributes in the Lustre
4482 client filesystem, and with the user_xattr mount option it
4483 is possible to allow users to store EAs on their files also
4485 Severity : enhancement
4487 Description: Add possibility (config option) to show minimal available OST free
4489 Details : When compiled with --enable-mindf configure option, statfs(2)
4490 (and so, df) will return least minimal free space available from
4491 all OSTs as amount of free space on FS, instead of summary of
4492 free spaces of all OSTs.
4494 Severity : enhancement
4496 Description: do not expand extent locks acquired on OST-side
4497 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
4498 by server: they are not cached anyway.
4501 Frequency : when mmap is used/binaries executed from Lustre
4503 Description: Unmmap pages before throwing them away from read cache.
4504 Details : llap_shrink cache now attempts to unmap pages before discarding
4505 them (if unmapping failed - do not discard). SLES9 kernel has
4506 extra checks that trigger if this unmapping is not done first.
4511 Description: lconf didn't resolve symlinks before checking to see whether a
4512 given mountpoint was already in use
4515 Frequency : when migrating failover services
4516 Bugzilla : 6395, 9514
4517 Description: When migrating a subset of services from a node (e.g. failback
4518 from a failover service node) the remaining services would
4519 time out and evict clients.
4520 Details : lconf --force (implied by --failover) sets the global obd_timeout
4521 to 5 seconds in order to quickly disconnect, but this caused
4522 other RPCs to time out too quickly. Do not change the global
4523 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
4525 Severity : enhancement
4526 Frequency : if MDS is started with down OST
4527 Bugzilla : 9439,5706
4528 Description: Allow startup/shutdown of an MDS without depending on the
4529 availability of the OSTs.
4530 Details : Asynchronously call mds_lov_synchronize during MDS startup.
4531 Add appropriate locking and lov-osc refcounts for safe
4532 cleaning. Add osc abort_inflight calls in case the
4533 synchronize never started.
4536 Frequency : occasional (Cray XT3 only)
4538 Description: root not authorized to access files in CRAY_PORTALS environment
4539 Details : The client process capabilities were not honoured on the MDS in
4540 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
4541 been accessed by an authorized user then root was able to access
4542 the file on the local client also. The root user capabilities
4543 are now allowed on the MDS, as this environment has secure UID.
4546 Frequency : occasional
4548 Description: ldiskfs "too long searching" message happens too often
4549 Details : A debugging message (otherwise harmless) prints too often on
4550 the OST console. This has been reduced to only happen when
4551 there are fragmentation problems on the filesystem.
4556 Description: Division by zero in statfs when all OSCs are inactive
4557 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
4558 lov_statfs divides by value returned by lov_get_stripecnt().
4562 Bugzilla : 9489, 3273
4563 Description: First write from each client to each OST was only 4kB in size,
4564 to initialize client writeback cache, which caused sub-optimal
4565 RPCs and poor layout on disk for the first writen file.
4566 Details : Clients now request an initial cache grant at (re)connect time
4567 and so that they can start streaming writes to the cache right
4568 away and always do full-sized RPCs if there is enough data.
4569 If the OST is rebooted the client also re-establishes its grant
4570 so that client cached writes will be honoured under the grant.
4575 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
4576 Details : Now I/O RPCs go to different portal number and (presumably) fast
4577 lock requests (and glimses) and other RPCs get their own service
4578 threads pool that should be able to service those RPCs
4581 Severity : enhancement
4583 Description: Ability to exchange lustre version between client and servers and
4584 issue warnings at client side if client is too old. Also for
4585 liblustre clients there is ability to refuse connection of too old
4587 Details : New 'version' field is added to connect data structure that is
4588 filled with version info. That info is later checked by server and
4592 Frequency : rare, liblustre only.
4593 Bugzilla : 9296, 9581
4594 Description: Two simultaneous writes from liblustre at offset within same page
4595 might proceed at the same time overwriting eachother with stale
4597 Details : I/O lock withing llu_file_prwv was released too early, before data
4598 actually was hitting the wire. Extended lock-holding time until
4599 server acknowledges receiving data.
4602 Frequency : extremely rare. Never observed in practice.
4604 Description: avoid generating lustre_handle cookie of 0.
4605 Details : class_handle_hash() generates handle cookies by incrementing
4606 global counter, and can hit 0 occasionaly (this is unlikely, but
4607 not impossible, because initial value of cookie counter is
4608 selected randonly). Value of 0 is used as a sentinel meaning
4609 "unassigned handle" --- avoid it. Also coalesce two critical
4610 sections in this function into one.
4612 Severity : enhancement
4614 Description: allow liblustre clients to delegate truncate locking to OST
4615 Details : To avoid overhead of locking, liblustre client instructs OST to
4616 take extent lock in ost_punch() on client's behalf. New connection
4617 flag is added to handle backward compatibility.
4619 Severity : enhancement
4620 Bugzilla : 4928, 7341, 9758
4621 Description: allow number of OST service threads to be specified
4622 Details : a module parameter allows the number of OST service threads
4623 to be specified via "options ost ost_num_threads={N}" in the
4624 OSS's /etc/modules.conf or /etc/modprobe.conf.
4628 Bugzilla : 6146, 9635, 9895
4629 Description: servers crash with bad pointer in target_handle_connect()
4630 Details : In rare cases when a client is reconnecting it was possible that
4631 the connection request was the last reference for that export.
4632 We would temporarily drop the export reference and get a new
4633 one, but this may have been the last reference and the export
4634 was just destroyed. Get new reference before dropping old one.
4636 Severity : enhancement
4637 Frequency : if client is started with failover MDS
4639 Description: Allow multiple MDS hostnames in the mount command
4640 Details : Try to read the configuration from all specified MDS
4641 hostnames during a client mount in case the "primary"
4644 Severity : enhancement
4646 Description: Stop sending data to evicted clients as soon as possible.
4647 Details : Check if the client we are about to send or are sending data to
4648 was evicted already. (Check is done every second of waiting,
4649 for which l_wait_event interface was extended to allow checking
4650 of exit condition at specified intervals).
4653 Frequency : rare, normally only when NFS exporting is done from client
4655 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
4657 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
4660 Frequency : always, if srand() is called before liblustre initialization
4662 Description: Liblustre uses system PRNG disturbing its usage by user application
4663 Details : Introduce internal to lustre fast and high-quality PRNG for
4664 lustre usage and make liblustre and some other places in generic
4665 lustre code to use it.
4667 Severity : enhancement
4668 Bugzilla : 9477, 9557, 9870
4669 Description: Verify that the MDS configuration logs are updated when xml is
4670 Details : Check if the .xml configuration logs are newer than the config
4671 logs stored on the MDS and report an error if this is the case.
4672 Request --write-conf, or allow starting with --old_conf.
4674 Severity : enhancement
4676 Description: Handle symlinks in the path when checking if Lustre is mounted.
4677 Details : Resolve intermediate symlinks when checking if a client has
4678 mounted a filesystem to avoid duplicate client mounts.
4683 Description: lconf can hit an error exception but still return success.
4684 Details : The lconf command catches the Command error exception at the top
4685 level script context and will exit with the associated exit
4686 status, but doesn't ensure that this exit status is non-zero.
4691 Description: failure of ptlrpc thread startup can cause oops
4692 Details : Starting a ptlrpc service thread can fail if there are a large
4693 number of threads or the server memory is very fragmented.
4694 Handle this without oopsing.
4697 Frequency : always, only if liblustre and non-default acceptor port was used
4699 Description: liblustre cannot connect to servers with non-default acceptor port
4700 Details : tcpnal_set_default_params() was not called and was therefore
4701 ignoring the environment varaible TCPNAL_PORT, as well as other
4702 TCPNAL_ environment variables
4707 Description: two objects could be created on the same OST for a single file
4708 Details : If an OST is down, in some cases it was possible to create two
4709 objects on a single OST for a single file. No problems other
4710 than potential performance impact and spurious error messages.
4714 Bugzilla : 5681, 9562
4715 Description: Client may oops in ll_unhash_aliases
4716 Details : Client dcache may become inconsistent in race condition.
4717 In some cases "getcwd" can fail if the current directory is
4723 Description: Inode refcounting problems in NFS export code
4724 Details : link_raw functions used to call d_instantiate without obtaining
4725 extra inode reference first.
4729 Bugzilla : 9942, 9903
4730 Description: Referencing freed requests leading to crash, memleaks with NFS.
4731 Details : We used to require that call to ll_revalidate_it was always
4732 followed by ll_lookup_it. Also with revalidate_special() it is
4733 possible to call ll_revalidate_it() twice for the same dentry
4734 even if first occurence returned success. This fix changes semantic
4735 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
4736 reference on a request referred from the intent.
4737 ll_intent_release() then releases such a request.
4740 Frequency : rare, normally benchmark loads only
4742 Description: unlinked inodes were kept in memory on the client
4743 Details : If a client is repeatedly creating and unlinking files it
4744 can accumulate a lot of stale inodes in the inode slab cache.
4745 If there is no other client load running this can cause the
4746 client node to run out of memory. Instead flush old inodes
4747 from client cache that have the same inode number as a new inode.
4750 Frequency : SLES9 2.6.5 kernel and long filenames only
4751 Bugzilla : 9969, 10379
4752 Description: utime reports stale NFS file handle
4753 Details : SLES9 uses out-of-dentry names in some cases, which confused
4754 the lustre dentry revalidation. Change it to always use the
4758 Frequency : rare, unless heavy write-truncate concurrency is continuous
4759 Bugzilla : 4180, 6984, 7171, 9963, 9331
4760 Description: OST becomes very slow and/or deadlocked during object unlink
4761 Details : filter_destroy() was holding onto the parent directory lock
4762 while truncating+unlinking objects. For very large objects this
4763 may block other threads for a long time and slow overall OST
4764 responsiveness. It may also be possible to get a lock ordering
4765 deadlock in this case, or run out of journal credits because of
4766 the combined truncate+unlink. Solution is to do object truncate
4767 first in one transaction without parent lock, and then do the
4768 final unlink in a new transaction with the parent lock. This
4769 reduces the lock hold time dramatically.
4772 Frequency : rare, 2.4 kernels only
4774 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
4775 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
4776 before unmounting their filesystem in order to avoid deadlock.
4777 The kernel_locked() function in 2.4 kernels only checks whether
4778 the kernel lock is held, not whether it is this process that is
4779 holding it as 2.6 kernels do.
4784 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
4785 Details : The client ptlrpc code may be trying to reconnect to a down
4786 server before a previous connection attempt has timed out.
4787 Increase the reconnect interval to be longer than the connection
4788 timeout interval to avoid sending duplicate connections to
4792 Frequency : echo_client brw_test command
4794 Description: fix echo_client to work with OST preallocated code
4795 Details : OST preallocation code (5137) didn't take echo_client IO path
4796 into account: echo_client calls filter methods outside of any
4797 OST thread and, hence, there is no per-thread preallocated
4798 pages and buffers to use. Solution: hijack pga pages for IO. As
4799 a byproduct, this avoids unnecessary data copying.
4803 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
4804 Description: Client can oops in mdc_commit_close() after open replay
4805 Details : It was possible for the MDS to return an open request with no
4806 transaction number in mds_finish_transno() if the client was
4807 evicted, but without actually returning an error. Clients
4808 would later try to replay that open and may trip an assertion
4809 Simplify the client close codepath, and always return an error
4810 from the MDS in case the open is not successful.
4813 Frequency : rare, 2.6 OSTs only
4815 Description: OST may deadlock under high load on fragmented files
4816 Details : If there was a heavy load and highly-fragmented OST filesystems
4817 it was possible to have all the OST threads deadlock waiting on
4818 allocation of biovecs, because the biovecs were not released
4819 until the entire RPC IO was completed. Instead, release biovecs
4820 as soon as they are complete to ensure forward IO progress.
4822 Severity : enhancement
4824 Description: Support for specifying external journal device at mount
4825 Details : If an OST or MDS device is formatted with an external journal
4826 device, this device major/minor is stored in the ext3 superblock
4827 and may not be valid for failover. Allow detecting and
4828 specifying the external journal at mount time.
4833 Description: Mounting an MDS with pending unlinked files may cause oops
4834 Details : target_finish_recovery() calls mds_postrecov() which returned
4835 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
4836 considers this an error and immediately begins cleaning up the
4837 lov, just after starting the mds_lov process
4839 Severity : enhancement
4841 Description: Implement 'lfs df' to report actual free space on per-OST basis
4842 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
4843 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
4844 usage of objects; '-h' to report in human readable format.
4846 ------------------------------------------------------------------------------
4848 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
4855 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
4856 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
4857 at startup during orphan unlinking. Instead, skip buddy update
4858 before setup, as it will be regenerated after recovery is complete.
4863 Description: If an OST is inactive, its locks might reference stale inodes.
4864 Details : lov_change_cbdata() must iterate over all namespaces, even if
4865 they are inactive to clear inode references from the lock.
4867 Severity : enhancement
4868 Frequency : occasional, if non-standard max_dirty_mb used
4870 Description: Client will block write RPCs if not enough grant
4871 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
4872 then the client will block writes while waiting for another RPC
4873 to complete instead of consuming its dirty limit. With change
4874 we get improved performance when max_dirty_mb is small.
4876 Severity : enhancement
4877 Bugzilla : 3389, 6253
4878 Description: Add support for supplementary groups on the MDS.
4879 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
4880 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
4881 lookups for user supplementary groups into a cache.
4885 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
4886 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
4887 for liblustre O_EXCL is enforced by MDS.
4891 Description: readdir never returns NULL in liblustre.
4892 Details : Corrected llu_iop_getdirentries logic, to return offset of next
4893 dentry in struct dirent.
4897 Frequency : liblustre only, depends on application IO pattern
4898 Description: liblustre clients evicted if not contacting servers
4899 Details : Don't put liblustre clients into the ping_evictor list, so
4900 they will not be evicted by the pinger ever.
4902 Severity : enhancement
4904 Description: Add ability to evict clients by NID from MDS.
4905 Details : By echoing "nid:$NID" string into
4906 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
4907 $NID would be instantly evicted from this MDS and from all active
4908 OSTs connected to it.
4912 Description: Do not query file size twice, somewhat slowing stat(2) calls.
4913 Details : lookup_it_finish() used to query file size from OSTs that was not
4918 Description: service threads change working directory to that of init
4919 Details : Starting lustre service threads may pin the working directory
4920 of the parent thread, making that filesystem busy. Threads
4921 now change to the working directory of init to avoid this.
4925 Frequency : during shutdown only
4926 Description: shutdown with a failed MDS or OST can cause unmount to hang
4927 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
4932 Frequency : 2.6 only
4933 Description: chmod/chown may include an extra supplementary group
4934 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
4935 supplementary group and if none is specified this is used.
4938 Bugzilla : 5479 (6816)
4940 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
4941 Details : If lookup is in progress on a file that is unlinked we might try
4942 to revalidate the inode and fail in revalidate after lookup is
4943 complete and ll_file_open() enqueues the open again but
4944 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
4947 Frequency : always, if lconf --abort_recovery used
4949 Description: lconf --abort_recovery fails with 'Operation not supported'
4950 Details : lconf was attempting to abort recovery on the MDT device and not
4953 ------------------------------------------------------------------------------
4955 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
4960 Frequency : rare (only unsupported configurations with a node running as an
4962 Bugzilla : 6514, 5137
4963 Description: Mounting a Lustre file system on a node running as an OST could
4965 Details : OSTs now preallocates memory needed to write out data at
4966 startup, instead of when needed, to avoid having to
4967 allocate memory in possibly low memory situations.
4968 Specifically, if the file system is mounted on on OST,
4969 memory pressure could force it to try to write out data,
4970 which it needed to allocate memory to do. Due to the low
4971 memory, it would be unable to do so and the node would
4972 become unresponsive.
4974 Severity : enhancement
4976 Description: Addition of lconf --service command line option
4977 Details : lconf now accepts a '--service <arg>' option, which is
4978 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
4980 Severity : enhancement
4982 Description: Failover mode is now the default for OSTs.
4983 Details : By default, OSTs will now run in failover mode. To return to
4984 the old behaviour, add '--failout' to the lmc line for OSTs.
4986 Severity : enhancement
4988 Description: Health checks are now provided for MDS and OSTs
4989 Details : Additional detailed health check information on MSD and OSTs
4990 is now provided through the procfs health_check value.
4993 Frequency : occasional, depends on IO load
4995 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
4996 after numerous create/delete cycles
4997 Details : The ext3 inode allocation policy would not allocate new inodes
4998 very well on the OSTs because there are no new directories
4999 being created. Instead we look for groups with free space if
5000 the parent directories are nearly full.
5005 Description: Network or server problems during mount may cause partially
5006 mounted clients instead of returning an error.
5007 Details : The config llog parsing code may overwrite the error return
5008 code during mount error handling, returning success instead
5014 Description: MDS can fail to allocate large reply buffers
5015 Details : After long uptimes the MDS can fail to allocate large reply
5016 buffers (e.g. zconf client mount config records) due to memory
5017 fragmentation or consumption by the buffer cache. Preallocate
5018 some large reply buffers so that these replies can be sent even
5019 under memory pressure.
5023 Frequency : rare (liblustre)
5024 Description: fsx running with liblustre complained that using truncate() to
5025 extend the file doesn't work. This patch corrects that issue.
5026 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
5027 ATTR_SIZE and lsm use in llu_setattr_raw.
5031 Frequency : rare, only 2.6 kernels
5032 Description: Unusual file access patterns on the MDS may result in inode
5033 data being lost in very rare circumstances.
5034 Details : Bad interaction between the ea-in-inode patch and the "no-read"
5035 code in the 2.6 kernel caused the inode and/or EA data not to
5036 be read from disk, causing single-file corruption.
5040 Frequency : rare, only 2.6 filesystems using extents
5041 Description: Heavy concurrent write and delete load may cause data corruption.
5042 Details : It was possible under high-load situations to have an extent
5043 metadata block in the block device cache from a just-unlinked
5044 file overwrite a newly-allocated data block. We now unmap any
5045 metadata buffers that alias just-allocated data blocks.
5049 Frequency : filesystems with default stripe_count larger than 77
5050 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
5051 Details : lconf specifies an inode size of 4096 bytes when the default
5052 stripe_count is larger than 77. This conflicts with the default
5053 inode density of 1 per 4096 bytes. Allocate smaller inodes in
5054 this case to avoid pinning too much memory for large EAs.
5056 ------------------------------------------------------------------------------
5058 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
5063 Frequency : rare (extremely heavy IO load with hundreds of clients)
5065 Description: Client is evicted, gets IO error writing to file
5066 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
5067 caused clients to be evicted for AST timeouts. The fixes in
5068 bug 5192 mean we no longer need to have such short AST timeouts
5069 so ldlm_timeout has been increased.
5072 Frequency : occasional during --force or --failover shutdown under load
5073 Bugzilla : 5949, 4834
5074 Description: Server oops/LBUG if stopped with --force or --failover under load
5075 Details : a collection of import/export refcount and cleanup ordering
5076 issues fixed for safer force cleanup
5079 Frequency : only filesystems larger than 120 OSTs
5080 Bugzilla : 5990, 6223
5081 Description: lfs getstripe would oops on a very large filesystem
5082 Details : lov_getconfig used kfree on vmalloc'd memory
5085 Frequency : only filesystems exporting via NFS to Solaris 10 clients
5086 Bugzilla : 6242, 6243
5087 Description: reading from files that had been truncated to a non-zero size
5088 but never opened returned no data
5089 Details : ll_file_read() reads zeros from no-object files to EOF
5094 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
5095 Details : There's little harm in aborting MDS/OSS recovery and letting it
5096 try again, so I removed the LASSERT and return an error instead.
5098 Severity : enhancement
5100 Description: New debugging infrastructure for tracking down data corruption
5101 Details : The I/O checksum code was replaced to: (a) control it at runtime,
5102 (b) cover more of the client-side code path, and (c) try to narrow
5103 down where problems occurred
5107 Bugzilla : 3819, 4364, 4397, 6313
5108 Description: Racing close and eviction MDS could cause assertion in mds_close
5109 Details : It was possible to get multiple mfd references during close and
5110 client eviction, leading to one thread referencing a freed mfd.
5112 Severity: : enhancement
5113 Bugzilla : 3262, 6359
5114 Description: Attempts to reconnect to servers are now more aggressive.
5115 Details : This builds on the enhanced upcall-less recovery that was added
5116 in 1.4.2. When trying to reconnect to servers, clients will
5117 now try each server in the failover group every 10 seconds. By
5118 default, clients would previously try one server every 25 seconds.
5123 Description: After recovery, certain operations trigger a failed
5124 assertion on a client.
5125 Details : Failing over an mds, using lconf -d --failover, while a
5126 client was doing a readdir() call would cause the client to
5127 LBUG after recovery completed and the readdir() was resent.
5129 Severity : enhancement
5131 Description: Default groups are now added by lconf
5132 Details : You can now run lconf --group <servicename> without having to
5133 manually add groups with lmc.
5136 Frequency : occasional
5138 Description: Nodes with an elan id of 0 trigger a failed assertion
5141 Frequency : always when accessing e.g. tty/console device nodes
5143 Description: tty and some other devices nodes cannot be used on lustre
5144 Details : file's private_data field is used by device data and lustre
5145 values in there got lost. New field was added to struct file to
5146 store fs-specific private data.
5149 Frequency : when exporting Lustre via NFS
5151 Description: NFSD failed occasionally when looking up a path component
5152 Details : NFSD is looking up ".." which was broken in ext3 directories
5153 that had grown large enough to become hashed.
5156 Frequency : Clusters with multiple interfaces not on the same subnet
5158 Description: Nodes will repeatedly try to reconnect to an interface which it
5159 cannot reach and report an error to the log.
5160 Details : Extra peer list entries will be created by lconf with some peers
5161 unreachable. lconf now validates the peer before adding it.
5164 Frequency : Only if a default stripe is set on the filesystem root.
5166 Description: Setting a default stripe on the filesystem root prevented the
5167 filesystem from being remounted.
5168 Details : The client was sending extra request flags in the root getattr
5169 request and did not allocate a reply buffer for the dir EA.
5172 Frequency : occasional, higher if lots of files are accessed by one client
5173 Bugzilla : 6159, 6097
5174 Description: Client trips assertion regarding lsm mismatch/magic
5175 Details : While revalidating inodes the VFS looks up inodes with ifind()
5176 and in rare cases can find an inode that is being freed.
5177 The ll_test_inode() code will free the lsm during ifind()
5178 when it finds an existing inode and then the VFS later attaches
5179 this free lsm to a new inode.
5183 Bugzilla : 6422, 7030
5184 Description: MDS deadlock between mkdir and client eviction
5185 Details : Creating a new file via mkdir or mknod (starting a transaction
5186 and getting the ns lock) can deadlock with client eviction
5187 (gets ns lock and trying to finish a synchronous transaction).
5190 Frequency : occasional
5191 Description: While starting a server, the fsfilt_ext3 module could not be
5193 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
5194 kernels. Previously, lconf would still use the ext3 name
5195 when trying to load modules. Now, it will correctly use
5196 ext3 on 2.4 and ldiskfs on 2.6.
5198 Severity : enhancement
5199 Description: The default stripe count has been changed to 1
5200 Details : The interpretation of the default stripe count (0, to lfs
5201 or lmc) has been changed to mean striping across a single
5202 OST, rather than all available. For general usage we have
5203 found a stripe count of 1 or 2 works best.
5205 Severity : enhancement
5206 Description: Add support for compiling against Cray portals.
5207 Details : Conditional compiling for some areas that are different
5211 Frequency : occasional
5212 Bugzilla : 6409, 6834
5213 Description: Creating files with an explicit stripe count may lead to
5214 a failed assertion on the MDS
5215 Details : If some OSTs are full or unavailable, creating files may
5216 trigger a failed assertion on the MDS. Now, Lustre will
5217 try to use other servers or return an error to the
5221 Frequency : occasional
5223 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
5224 caused lock timeout during readahead (since 1.4.2).
5225 Details : Processes doing readahead might match a lock that hasn't been
5226 granted yet if there are overlapping and conflicting lock
5227 requests. The readahead process waits on ungranted lock
5228 (original lock is CBPENDING), while OST waits for that process
5229 to cancel CBPENDING read lock and eventually evicts client.
5231 Severity : enhancement
5233 Description: Initial enabling of flock support for clients
5234 Details : Implements fcntl advisory locking and file status functions.
5235 This feature is provided as an optional mount flag (default
5236 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
5237 locking are implemented yet, and those that are are not guaranteed
5238 to be completely correct in production environments.
5239 mount -t lustre -o [flock|noflock] ...
5242 Frequency : occasional
5244 Description: OSTs running 2.4 kernels but with extents enabled might trip an
5245 assertion in the ext3 JBD (journaling) layer.
5246 Details : The b_committed_data struct is protected by the big kernel lock
5247 in 2.4 kernels, serializing journal_commit_transaction() and
5248 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
5249 access to this struct. In 2.6 kernels there is finer grained
5250 locking to improve SMP performance of the JBD layer.
5254 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
5259 Description: merge and cleanup kernel patches.
5262 -----------------------------------------------------------------------------
5264 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
5266 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
5267 versions of Lustre. Please update all servers and clients to
5268 version 1.4.2 or later at the same time. You must also run
5269 "lconf --write-conf {config}.xml" on the MDS while it is stopped
5270 to update the configuration logs.
5272 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
5273 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
5274 - swab llog records of type '0' so we get proper header size/idx (5861)
5275 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
5276 - fix rename of one directory over another leaking an inode (5953)
5277 - avoid SetPageDirty on 2.6 (5981)
5278 - don't re-add just-being-destroyed locks to the waiting list (5653)
5279 - when creating new directories, inherit the parent's custom
5280 striping settings if present parent (3048)
5281 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
5282 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
5283 - don't hold client locks on temporary worklist from l_lru (5666)
5284 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
5285 - automatically evict dead clients (5921)
5286 - Update file size properly in create+truncate+fstat case (6196)
5287 - Do not unhash mountpoint dentries, do not allow removal of
5289 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
5290 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
5291 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
5292 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
5293 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
5294 - Correctly handle reads of files with no objects (6243)
5295 - lctl recover will also mark a device active if deactivate used (5933)
5297 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
5298 - allow --write-conf on an MDS with different nettype than client (5619)
5299 - don't write config llogs to MDS for mounts not from that MDS (5617)
5300 - lconf should create multiple TCP connections from a client (5201)
5301 - init scripts are now turned off by default; run chkconfig --on
5302 lustre and chkconfig --on lustrefs to use them
5303 - upcalls are no longer needed for clients to recover to failover
5305 - add --abort-recovery option to lconf to abort recovery on device
5307 - add support for an arbitrary number of OSTs (3026)
5308 - Quota support protocol changes.
5309 - forward compatibility changes to wire structs (6007)
5310 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5311 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5312 - improved "lustre status" script
5313 - initialize blocksize for non-regular files (6062)
5314 - added --disable-server and --disable-client configure options (5782)
5315 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5316 - Vanilla 2.4.29 support
5317 - increase maximum number of obd devices to 520 (6242)
5318 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5319 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5320 SLES 9 2.6.5 kernels
5321 - limit stripes per file to 160 (the maximum EA size) (6093)
5323 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5326 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5327 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5328 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5329 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5330 - small ext3 extents cleanups and fixes (5733)
5331 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5332 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5333 - avoid needless client->OST connect, fix handle mismatch (5317)
5334 - fix DLM error path that led to out-of-sync client, long delays (5779)
5335 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5336 - fix several locking issues related to i_size (5492,5624,5654,5672)
5337 - don't move pending lock onto export if it is already evicted (5683)
5338 - fix kernel oops when creating .foo in unlinked directory (5548)
5339 - fix deadlock in obdfilter statistics vs. object create (5811)
5340 - use time_{before,after} to avoid timer jiffies wrap (5882)
5341 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5342 - Do not leak request if server was not able to process it (5154)
5343 - If mds_open unable to find parent dir, make that negative lookup(5154)
5344 - don't create new directories with extent-mapping (5909, 5936)
5346 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5347 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5348 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5349 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5350 - support for RHEL4 kernel on i686 (5773)
5351 - provide error messages when incompatible logs are encountered (5898)
5353 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5354 * version 1.4.0.10 (1.4.1 release candidate 1)
5356 - don't keep a lock reference when lock is not granted (4238)
5357 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5358 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5359 - drop import inflight refcount on signal_completed_replay error (5255)
5360 - unlock page after async write error during send (3677)
5361 - handle missing objects in filter_preprw_read properly (5265)
5362 - no transno return for symlink open, don't save no-trasno open (3440)
5363 - don't try to complete elan receive that already failed (4012)
5364 - free RPC server reply state on error (5406)
5365 - clean up thread from ptlrpc_start_thread() on error (5160)
5366 - readahead could read extra page into cache that wasn't ejected (5388)
5367 - prevent races in class_attach/setup/cleanup/detach (5260)
5368 - don't dereference de->d_inode after l_dput of de (5458)
5369 - use "int" for stripe value returned from lock_to_stripe (5544)
5370 - mballoc allocation and error-checking fixes in 2.6 (5504)
5371 - block device patches to fix I/O request sizes in 2.6 (5482)
5372 - look up hostnames for IB nals (5602)
5373 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5374 - don't start multiple acceptors for the same port (5277)
5375 - fix incorrect LASSERT in mds_getattr_name (5635)
5376 - export a proc file for general "ping" checking (5628)
5377 - fix "lfs check" to not block when the MDS is down (5628)
5379 - service request history (4965)
5380 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5381 - create an "evict_client" /proc entry on OSTs, like the MDS has
5382 - fix mount usage message, return errors per mount(8) (5168)
5383 - change grep [] to grep "[]" in tests so they work in more UMLs
5384 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5385 - remove ancient LOV_MAGIC_V0 EA support (5047)
5386 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5387 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5388 - allow readahead allocations to fail when low on memory (5383)
5389 - mmap locking landed again, after considerable improvement (2828)
5390 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
5392 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
5395 - send OST transaction number in read/write reply to free req (4966)
5396 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5397 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5398 - the watchdog thread now runs as interruptible (5246)
5399 - flock/lockf fixes (but it's still disabled, pending 5135)
5400 - don't use EXT3 constants in llite code (5094)
5401 - memory shortage at startup could cause assertion (5176)
5403 - reorganization of lov code
5404 - single portals codebase
5406 - add extents/mballoc support (5025)
5407 - direct I/O reads in the obdfilter (4048)
5408 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
5410 tbd Cluster File Systems, Inc. <info@clusterfs.com>
5413 - send OST transaction number in read/write reply to free req (4966)
5414 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5415 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5416 - the watchdog thread now runs as interruptible (5246)
5417 - handle missing objects in filter_preprw_read properly (5265)
5418 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5419 - drop import inflight refcount on signal_completed_replay error (5255)
5420 - unlock page after async write error during send (3677)
5421 - return original error code on reconstructed replies (3761)
5422 - no transno return for symlink open, don't save no-trasno open (3440)
5424 - add pid to ldlm debugging output (4922)
5425 - bump the watchdog timeouts -- we can't handle 30sec yet
5426 - extra debugging for orphan dentry/inode bug (5259)
5428 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
5431 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
5432 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
5433 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
5434 - fix NULL dereference in /proc/sys/portals/routes (4827)
5435 - allow failed mdc_close() operations to be interrupted (4561)
5436 - stop precreate on OST before MDS would time out on it (4778)
5437 - don't send partial-page writes before EOF from client (4410)
5438 - discard client grant for sub-page writes on large-page clients (4520)
5439 - don't free dentries not owned by NFS code, check generation (4806)
5440 - fix lsm leak if mds_create_objects() fails (4801)
5441 - limit debug_daemon file size, always print CERROR messages (4789)
5442 - use transno after validating reply (3892)
5443 - process timed out requests if import state changes (3754)
5444 - update mtime on OST during writes, return in glimpse (4829)
5445 - add mkfsoptions to LDAP (4679)
5446 - use ->max_readahead method instead of zapping global ra (5039)
5447 - don't interrupt __l_wait_event() during strace
5449 - add software watchdogs to catch hung threads quickly (4941)
5450 - make lustrefs init script start after nfs is mounted
5451 - fix CWARN/ERROR duplication (4930)
5452 - return async write errors to application if possible (2248)
5453 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
5454 - print NAL number in %x format (4645)
5455 - update barely-supported suse-2.4.21-171 series (4842)
5456 - support for sles 9 %post scripts
5457 - support for building 2.6 kernel-source packages
5458 - support for sles km_* packages
5460 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
5463 - ignore -ENOENT errors in osc_destroy (3639)
5464 - notify osc create thread that OSC is being cleaned up (4600)
5465 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
5466 - reconstruct ost_handle() like mds_handle() (4657)
5467 - create a new thread to do import eviction to avoid deadlock (3969)
5468 - let lconf resolve symlinked-to devices (4629)
5469 - don't unlink "objects" from directory with default EA (4554)
5470 - hold socknal file ref over connect in case target is down (4394)
5471 - allow more than 32000 subdirectories in a single directory (3244)
5472 - fix blocks count for O_DIRECT writes (3751)
5473 - OST returns ENOSPC from object create when no space left (4539)
5474 - don't send truncate RPC if file size isn't changing (4410)
5475 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
5476 - bind to privileged port in socknal and tcpnal (3689)
5478 - rate limit CERROR/CWARN console message to avoid overload (4519)
5479 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
5480 - basic mmap support (3918)
5481 - kernel patch series update from b1_4 (4711)
5483 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
5486 - avoid crash during MDS cleanup with OST shut down (2775)
5487 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
5488 - don't use bad inodes on the MDS (3744)
5489 - dynamic object preallocation to improve recovery speed (4236)
5490 - don't hold spinlock over lock dumping or change debug flags (4401)
5491 - don't zero obd_dev when it is force cleaned (3651)
5492 - print grants to console if they go negative (4431)
5493 - "lctl deactivate" will stop automatic recovery attempts (3406)
5494 - look for existing locks in ldlm_handle_enqueue() (3764)
5495 - don't resolve lock handle twice in recovery avoiding race (4401)
5496 - revalidate should check working dir is a directory (4134)
5498 - don't always mark "slow" obdfilter messages as errors (4418)
5500 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
5503 - don't close LustreDB during write_conf until it is done (3860)
5504 - fix typo in lconf for_each_profile (3821)
5505 - allow dumping logs from multiple threads at one time (3820)
5506 - don't allow multiple threads in OSC recovery (3812)
5507 - fix debug_size parameters (3864)
5508 - fix mds_postrecov to initialize import for llog ctxt (3121)
5509 - replace config semaphore with spinlock (3306)
5510 - be sure to send a reply for a CANCEL rpc with bad export (3863)
5511 - don't allow enqueue to complete on a destroyed export (3822)
5512 - down write_lock before checking llog header bitmap (3825)
5513 - recover from lock replay timeout (3764)
5514 - up llog sem before sending rpc (3652)
5515 - reduce ns lock hold times when setting kms (3267)
5516 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
5517 - fix NULL deref and obd_dev leak on setup error (3312)
5518 - replace some LBUG about llog ops with error handling (3841)
5519 - don't match INVALID dentries from d_lookup and spin (3784)
5520 - hold dcache_lock while marking dentries INVALID and hashing (4255)
5521 - fix invalid assertion in ptlrpc_set_wait (3880)
5523 - add libwrap support for the TCP acceptor (3996)
5524 - add /proc/sys/portals/routes for non-root route listing (3994)
5525 - allow setting MDS UUID in .xml (2580)
5526 - print the stack of a process that LBUGs (4228)
5528 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
5531 - don't cleanup request in ll_file_open() on failed MDS open (3430)
5532 - make sure to unset replay flag from failed open requests (3440)
5533 - if default stripe count is 0, use OST count for inode size (3636)
5534 - update parent mtime/ctime on client for create/unlink (2611)
5535 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
5536 - free recovery state on server during a forced cleanup (3571)
5537 - unregister_reply for resent reqs (3063)
5538 - loop back devices mounting and status check on 2.6 (3563)
5539 - fix resource-creation race that can provoke i_size == 0 (3513)
5540 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
5541 - more debugging for page-accounting assertion (3746)
5542 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
5543 - avoid deadlock after precreation failure (3758)
5544 - fix race and lock order deadlock in orphan handling (3450, 3750)
5545 - add validity checks when grabbing inodes from l_ast_data (3599)
5547 - add /proc/.../recovery_status to obdfilter (3428)
5548 - lightweight CDEBUG infrastructure, debug daemon (3668)
5549 - change default OSC RPC parameters to be better on small clusters
5550 - turn off OST read cache for files smaller than 32MB
5551 - install man pages and include them in rpms (3100)
5552 - add new init script for (un)mounting lustre filesystems (2593)
5553 - run chkconfig in %post for init scripts (3701)
5554 - drop scimac NAL (unmaintained)
5556 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
5559 - clean kiobufs before and after use (3485)
5560 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
5561 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
5562 - change init script to fail more clearly if not run as root (1528)
5563 - allow clients to reconnect during replay (1742)
5564 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
5565 - don't do DNS lookups on NIDs too small for IP addresses (3442)
5566 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
5567 - fix cond_resched (3554)
5568 - only evict unfinished clients after recovery (3515)
5569 - allow bulk resend, prevent data loss (3570)
5570 - dynamic ptlrpc request buffer allocation (2102)
5571 - don't allow unlinking open directory if it isn't empty (2904)
5572 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
5573 - remove extraneous obd dereference causing LASSERT failure (3334)
5574 - don't use get_cycles() when creating temp. files on the mds (3156)
5575 - hold i_sem when setting i_size in ll_extent_lock() (3564)
5576 - handle EEXIST for set-stripe, set proper directory name (3336)
5578 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
5579 - fix ksocknal_fmb_callback() error messages (2918)
5581 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
5584 - don't copy lvb into (possibly NULL) reply on error (2983)
5585 - don't deref dentry after dput, don't free lvb on error (2922)
5586 - use the kms to determine writeback rpc length (2947)
5587 - increment oti_logcookies when osc is inactive (2948)
5588 - update client's i_blocks count via lvb messages (2543)
5589 - handle intent open/close of special files properly (1557)
5590 - mount MDS with errors=remount-ro, like obdfilter (2009)
5591 - initialize lock handle to avoid ASSERT on error cleanup (3057)
5592 - don't use cancelling-locks' kms values (2947)
5593 - use highest lock extent for kms, not last one (2925)
5594 - don't dereference ERR_PTR() dentry in error handling path (3107)
5595 - fix thread race in portals_debug_dumplog() (3122)
5596 - create lprocfs device entries at setup instead of at attach (1519)
5597 - common AST error handler, don't evict client on completion race (3145)
5598 - zero nameidata in detach_mnt in 2.6 (3118)
5599 - verify d_inode after revalidate_special is valid in 2.6 (3116)
5600 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
5601 - initialize RPC timeout timer earlier for 2.6 (3219)
5602 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
5603 - print nal/nid for unknown nid (3258)
5604 - additional checks for oscc recovery before doing precreate (3284)
5605 - fix ll_extent_lock() error return code for 64-bit systems (3043)
5606 - don't crash in mdc_close for bad permissions on open (3285)
5607 - zero i_rdev for non-device files (3147)
5608 - clear page->private before handing to FS, better assertion (3119)
5609 - tune the read pipeline (3236)
5610 - fix incorrect decref of invalidated dentry (2350)
5611 - provide read-ahead stats and refine rpc in flight stats (3328)
5612 - don't hold journal transaction open across create RPC (3313)
5613 - update atime on MDS at close time (3265)
5614 - close LDAP connection when recovering to avoid server load (3315)
5615 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
5616 - don't leak open file on MDS after open resend (3325)
5617 - serialize filter_precreate and filter_destroy_precreated (3329)
5618 - loop device shouldn't call sync_dev() for nul device (3092)
5619 - clear page cache after eviction (2766)
5620 - resynchronize MDS->OST in background (2824)
5621 - refuse to mount the same filesystem twice on same mountpoint (3394)
5622 - allow llmount to create routes for mounting behind routers (3320)
5623 - push lock cancellation to blocking thread for glimpse ASTs (3409)
5624 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
5625 - fix rare problem with rename on htree directories (3417)
5627 - allow default OST striping configuration per directory (1414)
5628 - fix compilation for qswnal for 2.6 kernels (3125)
5629 - increase maximum number of MDS request buffers for large systems
5630 - change liblustreapi to be useful for external progs like lfsck (3098)
5631 - increase local configuration timeout for slow disks (3353)
5632 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
5634 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5637 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
5638 - don't overwrite extent policy data in reply if lock was blocked (2901)
5639 - drop filter export grants atomically with removal from device (2663)
5640 - del obd_self_export from work_list in class_disconnect_exports (2908)
5641 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
5642 - swab reply message in mdc_close, other PPC fixes (2464)
5643 - fix destroying of named logs (2325)
5644 - overwrite old logs when running lconf --write_conf (2264)
5645 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
5646 - fix race in target_handle_connect (2898)
5647 - mds_reint_create() should take same inode create lock (2926)
5648 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
5649 - don't close files for self_export to avoid uninitialized obd (2936)
5650 - allow MDS with the same name as client node (2939)
5651 - hold dentry reference for closed log files for unlink (2325)
5652 - reserve space for all logs during transactions (2059)
5653 - don't evict page beyond end of stripe extent (2925)
5654 - don't oops on a deleted current working directory (2399)
5655 - handle hard links to targets without a parent properly (2517)
5656 - don't dereference NULL lock when racing during eviction (2867)
5657 - don't grow lock extents when lots of conflicting locks (2919)
5659 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
5662 - account for cache space usage on clients to avoid data loss (974)
5663 - lfsck support in lustre kernel code (2349)
5664 - reduce journal credits needed for BRW writes (2370)
5665 - orphan handling to avoid losing space on client/server crashes
5666 - ptlrpcd can be blocked, stopping ALL progress (2477)
5667 - use lock value blocks to assist in proper KMS, faster stat (1021)
5668 - takes i_sem instead of DLM locks internally on obdfilter (2720)
5669 - recovery for initial connections (2355)
5670 - fixes for mds_cleanup_orphans (1934)
5671 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
5672 - block all file creations until orphan recovery completes (1901)
5673 - client remove rq_connection from request struct (2423)
5674 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
5675 - recovery timer race (2670)
5676 - mdc_close recovey bug (2532)
5677 - ptlrpc cleanup bug (2710)
5678 - mds timeout on local locks (2588)
5679 - namespace lock held during RPCs (2431)
5680 - handle interrupted sync write properly (2503)
5681 - don't try to handle a message that hasn't been replied to (2699)
5682 - client assert failure during cleanup after abort recovery (2701)
5683 - leak mdc device after failed mount (2712)
5684 - ptlrpc_check_set allows timedout requests to complete (2714)
5685 - wait for inflight reqs when ptlrpcd finishes (2710)
5686 - make sure unregistered services are removed from the srv_list
5687 - reset bulk XID's when resending them (caught by 1138 test)
5688 - unregister_bulk after timeout
5689 - fix lconf error (2694)
5690 - handle write after unfinished setstripe, stripe-only getstripe (2388)
5691 - readahead locks pages, leaves pending causing memory pressure (2673)
5692 - increase OST request buffers to 4096 on large machines (2729)
5693 - fix up permission of existing directories in simple_mkdir (2661)
5694 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
5695 - don't assign transno to errored transactions (2742)
5696 - don't delete objects on OST if given a bogus objid from MDS (2751)
5697 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
5698 - if rq_no_resend, then timeout request after recovery (2432)
5699 - fix MDS llog_logid record size, 64-bit array alignment (2733)
5700 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
5701 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
5702 - MDS recovery shouldn't delete objects that it has given out (2730)
5703 - if enqueue arrives after completion, don't clobber LVB (2819)
5704 - don't unlock pages twice when trigger_group_io returns error (2814)
5705 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
5706 - don't write pages to disk if there was an error (1450)
5707 - don't ping imports that have recovery disabled (2676)
5708 - take buffered bytes into account when balancing socknal conn (2817)
5709 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
5710 - reconnect unlink llog connection after MDS reconnects to OST (2816)
5711 - remove little-endian swabbing of llog records (1987)
5712 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
5713 - retry reposting request buffers if they fail (1191)
5714 - grow extent at grant time to avoid granting a revoked lock (2809)
5715 - lock revoke doesn't evict page if covered by a second lock (2765)
5716 - disable VM readahead to avoid reading outside lock extents (2805)
5718 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
5719 - updated kernel patches for hp-2.4.20 kernel (2681)
5721 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
5724 - fix truncated write corruption (2366)
5725 - fix for failed assertion in iopen_connect_dentry (1792,2517)
5727 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
5728 - ensure len(uuid) < 37 in lmc (1171)
5729 - fix ia64 OOPS in llog_test (2255)
5730 - zero end of page at obdfilter for partial page writes (2648)
5731 - don't leave stale dentries around after renames (bug 2428)
5732 - fix timeouts when evicting a client with a single lock held (2642)
5733 - set deadline for the initial HELLO message to drain (2634)
5734 - print out dotted-quad IP addresses in the socknal (2302)
5736 - additional debugging for MDS client eviction problem (2443)
5737 - fix mkfsoptions support for osts (2603, 2604)
5739 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
5742 - add series for the vanilla 2.6.0 kernel
5743 - add series for the vanilla 2.4.24 kernel
5744 - add series for a cray x86/64 UL kernel drop
5745 - fix xattr patches for the vanilla 2.4.19 series
5747 - generate true UUIDs in lmc (1171)
5748 - have portals stack dumping break in UML (2466)
5749 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
5750 - allocate with _NOFS in ldlm to avoid deadlock (1933)
5751 - wake callback waiting threads on client eviction (2460)
5752 - Add --ptldebug and --subsystem to lmc (1719)
5753 - update assertion to allow safe interrupt allocation
5754 - set rq_no_resend for cancel requests (2432)
5755 - recalculate ptlrpcd timeout after resend (2494)
5756 - call vfs_rmdir when removing pending directories (2368)
5757 - fix renaming a file to itself (2429)
5758 - lmc creates a default one-stripe lov (2454)
5759 - expand procfs space to handle large clusters (2326)
5760 - increase UML stack to avoid overflow
5761 - update lconf's list of debug and subsystem masks
5762 - fix lfs find --obd (2510)
5763 - /proc tunable for disabling filter read caching (2591)
5764 - stop rpm packages from altering slapd.conf (2301)
5765 - disable nagle in the socknal under 0conf (2578)
5766 - choose mds inode size based on stripe count (2572)
5767 - fix kernel-source rpm problems (2516)
5769 - add --disable-doc to avoid pdf generation (2421)
5770 - update documentation, tests, type-os, comments
5771 - avoid format warnings on ia64
5772 - remove the TOE NAL
5773 - tiny code cleanups by removing unused fields
5775 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
5778 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
5779 - remove the most blatant lies from BUILDING file (2371)
5780 - change default debug level to reasonable production setting
5781 - reduce client side cache size to reduce cache flush time
5782 - reduce max RPCs in flight to avoid unnecessary file fragmentation
5783 - make TCP zerocopy and pinger support enabled by default (2476)
5784 - sync writes completed after process exits caused crashes (2319)
5785 - maintain correct mount count on the MDS (2356)
5786 - backout 1557, because 2316 wasn't really fixed
5787 - better file I/O statistics gathering in /proc
5788 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
5789 - another kernel patch to fix zero-copy TCP function export
5790 - don't take duplicate lock when processing re-sent getattr (2420)
5791 - lctl uses obd_self_export instead of creating new conn (2353)
5792 - MDS/OST recovery case which requires object creation asserted (2425)
5793 - move lfs from /usr/sbin to /usr/bin in packages
5794 - fix race between mds_client_add and mds_client_free (2417)
5795 - use kmalloc instead of slabs in portals (2430)
5796 - don't create duplicate records when a failover MDS is present (2442)
5797 - remove unnecessary mount age check (2332)
5798 - don't remove directory inodes from locks prematurely (2451)
5799 - don't break if MDS service name is the same as hostname (2103)
5800 - fix races in client write RPC generation when cache full (2482)
5802 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
5805 - remove now-unused request->rq_obd (278)
5806 - if an allocation fails, print out how much memory we've used (1933)
5807 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
5808 - add the "configurable stack size" patch to most series files (1256)
5809 - ability to write large log records, for 100+ OST configs (2306)
5810 - fix NULL deref when filter_prep fails (2314)
5811 - fix operator precedence error in filter_sync
5812 - dynamic allocation of socknal TX descriptors (2315)
5813 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
5814 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
5815 - ensure that CWARN messages in Portals always get to the syslog
5816 - __init/__exit are not for prototype decls (ldlm_init/exit)
5817 - x86-64 compile warning fixes
5818 - fix gateway LMC keyword conflict (2318)
5819 - fix MDS lock inversions in getattr/reint paths (1844)
5820 - fix a rare lock re-ordering bug, which caused deadlock (2322)
5821 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
5822 - DLM race condition prevented some lock evictions (2328)
5823 - ENOMEM detection and retry on socknal sends (2230)
5824 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
5825 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
5826 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
5827 - better error messages when a client is rejected during recovery (1505)
5828 - avoid cancelling locks which were never granted, after failure (2330)
5829 - fix i_sem/journal inversion in mds_client_add (2333)
5830 - fix truncate/getattr lock cycle deadlock (2334)
5831 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
5832 - fix two rare exit paths which could leak an l_lock() ref (2321)
5833 - fencepost error in MDS/OST orphan recovery (2226)
5834 - make log record alignment 8 bytes (1988)
5835 - lstripe now fails when requested offset > ost_count (2237)
5836 - ensure that all kernel series have a complete list.h (1607)
5837 - fix crashes in special-file operations (2316)
5838 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
5840 - allow configurable automake binary, for testing new versions
5841 - small update to the lfs documentation
5843 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
5845 * fix negative export reference count in fsfilt_sync (2312)
5847 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
5848 * release candidate 0.9.1
5850 - orphans are moved into the PENDING directory for possible recovery
5851 - replayed opens now open by fid for orphan/rename safety (1042)
5852 - last close of an orphan inode generates a transno (683)
5853 - chdir() and mount() now pin the directory entry (1020)
5854 - avoid CERROR in normal ll_setattr_raw() error case (1500)
5855 - discard very old requests without processing them (1502)
5856 - remove some common, well-understood CERRORs (1505)
5857 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
5858 - clear "grant" flags in OST replies until OST grant code lands (1644)
5859 - fix read performance by not clobbering i_blksize on client (1598)
5860 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
5861 - make LRU size a /proc tunable, clears locks when reduced (707)
5862 - fix some lprocfs rot that prevented ptlbd from loading (1732)
5863 - server locks take references on exports now (1558)
5864 - build fixes for 2.4.20-rh trees (1663)
5865 - return an error from lov_create if all OSCs are inactive (1751)
5866 - fix import levels when a reconnect happens without a timeout (1597)
5867 - exit early from mds_open if we get a lookup error (1749)
5868 - partial page read at EOF wouldn't wait for disk before sending (1642)
5869 - avoid NULL deref in obdfilter when reading page past EOF (1592)
5870 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
5871 - fix LBUG in ll_it_open_error with rc = -2 (1861)
5872 - write/truncate lock inversion (1639)
5873 - Don't auto-load obdclass, portals modules during cleanup (1495)
5874 - fix timestamps from jumping to "now" (1763)
5875 - extra journal assertions (1648)
5876 - add an extra multiunlink test (1771)
5877 - fix read_record/write_record API (1776)
5878 - fix leak of offset_extent, possible incorrect i_size later (1772)
5879 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
5880 - Debugging for the kqswnal_get_idle_tx problems (1820)
5881 - Allow recovery to be attempted multiple times (1536)
5882 - Write out MDS last_rcvd file after it is first created (1600)
5883 - Fix tx_descriptor leak in failed transmit situations (1827)
5884 - ext3 journaling fixes for assertion failure after IO error (1871)
5885 - class_export_put() on freed export after completion AST error (1896)
5886 - Fix revalidate looping in VFS (1322)
5887 - Don't access a freed export during MDS_REINT timeout (1521)
5888 - Add open-unlink recovery support on the MDS (1673,1764)
5889 - Return an error if no MDS data was read from last_rcvd (1946)
5890 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
5891 - Don't setup a disk device that doesn't match exported UUID (317)
5892 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
5893 - avoid committing NULL handle in force close
5894 - local.sh is now a one-stripe LOV configuration
5895 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
5896 - fix ext3 htree duplicate directory entry corruption (1516)
5897 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
5898 - update to new LOV EA format (2097)
5899 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
5901 2003-06-15 Phil Schwan <phil@clusterfs.com>
5904 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
5905 - per-import/export recovery handling (958, 931, 959)
5906 - multiple last-rcvd slots, for serving multiple FSes (949)
5907 - connections are again shared between multiple imp/exports (963, 964)
5908 - "umount -f" would hang if any requests needed to be sent (393, 978)
5909 - avoid pinning large req buffer by copying for queued messages (989)
5910 - add "uuid" to "lctl device" command to help upcalls (991)
5911 - "open" RPCs with transnos would confuse recovery counters (1037)
5912 - do proper endian conversion of all wire messages (288, 340, 891)
5913 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
5914 - call waiting locks callback from LDLM recovery thread (1127, 1151)
5915 - fix ptlrpc_connection leak in target_handle_connect (1174)
5916 - fix import refcounting bug in OST and MDS cleanup (1134)
5917 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
5918 - fix very unlikely obd_types race condition (501)
5919 - remove osc_open hack for echo_client (1187)
5920 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
5921 - a failure in read_inode2 leads to deadlock (1139)
5922 - cancel ack-locks as soon as transaction is committed (1072)
5923 - fix major leaks and crashes in the bulk I/O path (937, 1057)
5924 - make sure to commitrw after any preprw to avoid deadlock (1162)
5925 - failing to execute a file in a lustre FS would lock inode (1203)
5926 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
5927 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
5928 - obd_uuid2tgt would walk off the end of the list (1255)
5929 - on IA64 the osc would give portals incorrect bulk size (1258)
5930 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
5931 - fix lock inversion caused by new llite matching code (1282)
5932 - limit the number of dirty pages on a client to 10MB (1286)
5933 - timed out locks were not being corrected cancelled (1289)
5934 - fix O_DIRECT above 4GB on IA-32 (1292)
5935 * major user-visible changes
5936 - fail out/fail over policy now controlled by the upcall (993)
5938 - add OBD_PING to check server availability and failure (954)
5939 - lustre messages are now sent in sending host order (288, 340, 891)
5940 - add eadatalen to MDS getattr reply (340)
5941 - OST read replies may contain second buffer, with per-page status (593)
5943 2003-03-11 Phil Schwan <phil@clusterfs.com>
5946 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
5947 - failed open()s could cause deadlock; fixed (867, 869)
5948 - stop cancelling OST locks when files are closed (481)
5949 - overlapping XID spaces caused network corruption (851, 853)
5950 - fix unsafe fsfilt counter arithmetic; change to atomic_t
5951 - setattr_raw added, to do single-RPC, server-side setattrs
5952 - lmc/lconf syntax change for OST UUIDs
5953 - fix crashy race condition between ptlrpc_free_req and osc_close
5954 - don't use request in mdc_enqueue if we hit a timeout (889)
5955 - don't set the inode i_size for regular files from the MDS (896)
5956 - handle out of order completion AST (842)
5957 - don't LBUG if a lock request times out after receiving AST (913)
5958 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
5959 - if a bad lock AST arrives, send an error instead of dropping entirely
5960 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
5961 - fix leak in bulk IO when only partially completed (899, 900, 926)
5962 - fix O_DIRECT for ia64 (55)
5963 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
5964 - C-z after timeout could hang a process forever; fixed (977)
5966 - client-side I/O cache (678, 924, 929, 941, 970)
5968 - READPAGE and SETATTRs which don't take server-side locks get
5971 2003-02-11 Phil Schwan <phil@clusterfs.com>
5974 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
5975 - Fix fsfilt_extN_readpage() to read a full page of directory
5976 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
5977 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
5978 - Fix imbalanced LOV object allocation and out-of-bound access (469)
5979 - Most intent operations were removed, in favour of a new RPC mode
5980 that does a single RPC to the server and bypasses most of the VFS
5981 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
5982 - Aggressively cancel local locks on DLM servers
5983 - mds_reint_unlink sends EA to the client if it's the last nlink.
5984 client uses that EA to unlink OST objects.
5985 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
5986 - recursive symlinks were fixed (439)
5987 - fixed NULL deref in DEBUG_REQ
5988 - filter_update_lastobjid no longer calls sync, which annoyed extN
5989 - fixed multi-client small-writes to a single file problem (445)
5990 - fixed mtime updates during file writes (607)
5991 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
5992 - fixed bug in obd_brw_read/write() (under guise of testing 367)
5993 - fixed Linux OST size reporting problem (444, 656)
5994 - OST now updates object mtime with writes or setattr (607, 619)
5995 - client verifies file size before zeroing page past EOF (445)
5996 - OST now writes last allocated objid to disk with allocation (108)
5997 - LOV on echo now works (409)
5999 - mds_reint_unlink sends a new buffer, with the EA included. this
6000 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
6001 set if a regular file was being unlinked, and it was the last link
6002 - use PtlGet from the target for bulk writes (315)
6003 - OST now updates object mtime with writes or setattr (607, 619)
6004 - LDLM now has a grant-time callback to revalidate locked items, if
6006 - Many MDS operations were reorganized to combat race conditions
6008 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
6009 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
6011 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
6014 - Fully reactivate OST imports after reconnection (512, others)
6015 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
6016 - More graceful error handling for truncating on dead OST (515)
6017 - Don't error out unless we're actually accessing dead stripes (474)
6018 - Fix garbage sizes when stripes are missing (410)
6019 - LRU counters were broken, causing constant lock purge (433, 432)
6020 - garbage on read from stripes with failed OSTs (441)
6021 - mark OSCs as active before reconnecting during recovery (438)
6022 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
6023 - lfind did not preserve OST order in output (443)
6024 - symlinks cause hung clients, incorrect data (439)
6025 - stop dereferencing request after dropping refcount (457)
6026 - don't LASSERT(spin_is_locked) on non-SMP (455)
6027 - fixes for many rename() bugs
6028 - fstat didn't correctly synchronize attributes (399)
6029 - server must handle lock cancellation during blocking AST prep (487)
6030 - bulk descriptors were free()d too soon (511)
6031 - fix paths in lconf, which would load incorrect modules (451, 507)
6032 - fix confusing lconf 'host not found' error message (386)
6033 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
6034 - fix race condition in mdc_blocking_ast() for inode access (526)
6035 - fix lov_unpackmd() unpacking wrong number of stripes (537)
6036 - fix lov_set_osc_active() marking wrong OSC inactive (440)
6037 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
6038 - fix multiple writes of stripe MD to MDS (358, maybe 519)
6039 - fix lstripe in several ways (kernel side) (527)
6040 - fix request leak in ldlm_cli_enqueue (262)
6041 - incorrect OSC was marked inactive after OST failure
6042 - call mds_fs_cleanup before unmounting filesystem (524)
6043 - fix races between taking ns_lock and ldlm_lock_change_resource
6044 - fix races updating LOV export open file list
6045 - fix lov_enqueue error path, avoid decref-ing bad lock handle
6046 - fix recovery NULL deref in ldlm_cli_cancel_unused
6047 - fix some DLM races by using new hash table for lock handles (419)
6048 - permit the client to specify desired inodes, at replay
6049 - duplicate requests when we queue them for replay reintegration
6050 - fix last_rcvd offset calculation
6051 - sync after each recovered transaction, so we always make progress
6052 - never, not always, ERESTART requests without transnos
6053 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
6055 - skip replay if the MDS says that the client is already connected
6056 - don't check for a recovery-enabled export to match lctl's UUID
6057 - don't INC_USE_COUNT for phantom exports
6058 - don't crash when cleaning up phantom exports (567)
6059 - don't double-finish or set replay data for errored mdc_open requests
6060 - abort requests when they time out, so we don't get old replies
6061 - send/receive replies for AST messages again
6062 - if the client says that it doesn't have the lock, cancel it on the
6064 - if we timeout during I/O, don't try to cancel an in-use lock; instead
6065 mark it as destroyed, it will all work out when decref is called
6066 - fix module use counts (22, 581)
6068 - ASTs now expect a reply (server cancels lock on error reply)
6070 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
6073 - fix many simultaneous client startup (392)
6074 - fix dentry->d_it clobbering
6075 - credentials weren't being shipped for readdir/getattr operations
6076 - remove invalid assertions triggered during some concurrent MD
6078 - proper Lustre versions added (336, 389)
6079 - fix memory leak for create error case (398)
6080 - fix LOV locking bug that would get cli/srv out of sync
6081 - fix echo client over LOV (409)
6082 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
6083 - fix double-O_EXCL intent crash (424)
6084 - avoid sending multiple lock CANCELs (352)
6086 - MDS can do multi-client recovery (modulo bugs in new code)
6088 - many updates, edits, cleanups
6090 2002-11-18 Phil Schwan <phil@clusterfs.com>
6093 - fix null d_it dereference (346)
6094 - fix full OST/dbench hang (333)
6095 - fix permission problem with file removal (286)
6096 - fix removal of OSCs from LOV when they fail
6097 - fix NULL deref during bulk timeout (214)
6098 - fix problems related to multiple filesystems on one MDS (241)
6099 - fixed serious subtle metadata locking bugs
6100 - free locks on clients when inodes are removed due to memory
6102 - fix inode pointer in lock data (285)
6103 - partial support for multiple MDS on a single host (241)
6104 - data locks weren't cancelled at clear_inode time (290, 311)
6105 - intent locks could lead to unbounded lock growth (205)
6106 - added a maximum lock count, an LRU list, and a flusher
6107 - fix multiple rename (365)
6108 - properly abstracted the echo client
6109 - OSC locked 1 byte too many; fixed
6110 - rewrote brw callback code:
6111 - fixed recovery bugs related to LOVs (306)
6112 - fixed too-many-pages-in-one-write crash (191)
6113 - fixed (again) crash in sync_io_timeout (214)
6114 - probably fixed callback-related race (385)
6116 - Add capability to MDS protocol
6117 - LDLM cancellations and callbacks on different portals
6119 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
6122 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
6123 - defer open object destruction to close time (601981, 138)
6124 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
6125 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
6126 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
6127 - don't call obd_setattr in ll_file_release for destroyed objects
6128 * protocol change to lustre_msg: move |version| and add |flags|
6129 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
6130 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
6131 * added replay of create, unlink, link and rename operations during
6132 MDS failover; recovery should be much more robust now
6133 * remove failed OSCs from LOVs (only lov_create uses this so far)
6134 * the lustre-HOWTO was brought (more) up to date (582544)
6136 2002-10-23 Phil Schwan <phil@clusterfs.com>
6139 - in-use dentries weren't being reused properly (617851)
6140 - prevent multiple LDLM setup (599178)
6141 - fix LOV size calculations for truncate (617853)
6142 - fix client handling of MDS intent errors (POSIX)
6143 - fix permission bug in lovstripe.c test (624321)
6144 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
6145 - truncate past end of file could corrupt data
6146 - proper cleanup after timeouts, crashes, etc (592524, 550815)
6147 - a race in recovery could return ETIMEDOUT to apps (623947)
6148 - building outside the source directory was fixed
6149 * the lustre-HOWTO was brought (more) up to date (582544)
6150 * major progress was made on recovery functionality
6152 2002-10-10 Phil Schwan <phil@clusterfs.com>
6155 - recovery deadlock fix
6156 - rm -rf causes LBUG fix (617817)
6157 - file open by multiple tasks fix (618962)
6158 - directory permissions bugs (602707 and 620007)
6159 - journal_stop fixed with locking (611313)
6160 - O_APPEND failures resolved (618273, perhaps 614459)
6161 - lconf PATH fix (619770)
6162 - IA64 build fix (621450)
6163 - RPC buffer sizes scale with amount of memory
6165 2002-10-01 Phil Schwan <phil@clusterfs.com>
6168 - locks would be cancelled without throwing away data pages,
6169 resulting in inconsistent data (605627)
6170 - inode attributes were not always being refreshed (605627, 612449)
6171 - lconf now continues to cleanup after lctl reports an error
6172 - MDS now enforces user permissions (602707)
6173 - lprocfs cleanup fixed, but not yet enabled (614157)
6174 - fixed infinite server hang, should a client not respond to an AST
6175 - avoid going into recovery if user calls readlink() with a buffer
6176 that's too small (613941)
6177 - AST RPCs no longer require replies (614867) -- this may be changed
6178 - don't crash server if client sends an IOV that's too big (611336)
6179 - fixed lock conversion deadlock (611892)
6180 - fixed the following of symlinks (614622)
6181 * recovery: the server can remove locks from a client that dies, other
6182 clients can make progress
6183 * more extN patch fixes
6184 * compile-time configurable ptlrpc buffer allocations
6186 - collaborative read cache document
6187 - Lustre Lite Performance CDR document-in-progress
6189 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6192 - fix typo in patch-2.4.18
6194 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6197 - clear ptlrpc request each time in handle_incoming_request()
6198 - unlink of files now destroys the object on the OST
6200 2002-09-19 Peter Braam <braam@clusterfs.com>
6202 * add hard link support
6203 * change obdfile creation method
6204 * kernel patch changed
6206 2002-09-19 Peter Braam <braam@clusterfs.com>
6209 - stack overflow bug in extN fixed
6211 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
6213 * documentation updates
6214 - add man pages for config tools
6215 - update tests/README to describe testing with new config tools
6216 - finish metadata API descriptions
6217 * bug fixes and cleanups
6218 - statfs workaround for 16TB limit
6219 - LOV stripe allocation improved, can stripe on subset of OSTs
6220 - LOV file size/IO offset was wrong for files > 4GB in size
6221 - object EA data was being dropped, caused files to be unreadable
6222 - memory overflow with non-LOV OST caused memory corruption
6223 - fixed regression tests to work with new config tools, obdfilter
6224 - fixed bug when directory size became larger than 1 block
6225 - fixed bug (for single client case) when PWD was deleted
6226 - invalidate local directory pages when doing intent-based ops
6227 - avoid LDLM oops when lock callback contained bad data
6229 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
6231 * documentation updates
6232 * bug fixes and cleanups
6233 - configuration tools
6236 - 64-bit compile warnings
6237 - 64-bit internal statfs data
6239 * test_brw on persistent OST devices
6241 * lprocfs (disabled)
6243 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
6245 * documentation updates
6246 * bug fixes and cleanups
6247 * configuration tools
6249 2002-08-30 Peter J. Braam <braam@clusterfs.com>
6252 * many small fixes to 0_5_4
6253 * io/network handling
6254 * thinkos in MDS operations
6256 2002-08-24 Peter J. Braam <braam@clusterfs.com>
6259 * crucial basic fixes to 0.5.3
6260 * IOR, Iozone work over Elan
6263 2002-08-07 Phil Schwan <phil@clusterfs.com>
6264 * version 0_5_3, our first alpha
6265 * we use the new Portals iovs
6266 * documentation updates
6267 * bug fixes and cleanups
6268 * small changes in the DLM wire protocol
6270 2002-07-25 Peter J. Braam <braam@clusterfs.com>
6271 * version 0_5_1 with some initial stability,
6272 * locking on MD and file I/O.
6273 * documentation updates
6274 * several bug fixes since 0.5.0
6275 * small changes in wire protocol
6277 2002-07-18 Phil Schwan <phil@clusterfs.com>
6279 * delivered as Lustre Light Alpha
6280 * fixed a crash after handling invalid MDS requests
6281 * fixed directory pages for architectures with non-4k pages sizes
6283 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
6284 * release version v0_4_4
6285 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
6287 * Fixes a number of interruption problems with OST operations.
6288 * Update documentation for portals header changes
6289 * Move all wire protocol structs/defines to lustre_idl.h
6290 * Fixes symlink length bug.
6291 * Add tcpdump to repository.
6293 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
6294 * release version v0_4_3
6295 * Fixes statfs for inodes on extN.
6296 * Fixes bug in runtests which would delete /etc/hosts.
6297 * Use 64-bit object IDs wherever possible (not into VFS though)
6298 Remove ost_get_info, which is unused by lustre, and out of date.
6300 2002-07-03 Peter Braam <braam@clusterfs.com>
6301 * release version v0_4_2 Fixes a lookup error (type not passed)
6302 * move forward to head of Portals
6303 * move forward to latest Lustre kernel
6305 2002-06-25 Peter Braam <braam@clusterfs.com>
6306 * release version v0_4_1. Hopefully stable on single node use.