3 * Server support for kernels:
4 2.6.18-238.19.1.el5 (RHEL5)
5 2.6.32-131.2.1.el6 (RHEL6)
6 * Client support for unpatched kernels:
7 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
8 2.6.18-238.19.1.el5 (RHEL5)
9 2.6.32-131.2.1.el6 (RHEL6)
10 2.6.29.4-167.fc11 (FC11)
11 2.6.16.60-0.69.1 (SLES10)
13 * Recommended e2fsprogs version: 1.41.10-sun2
14 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
15 Lustre filesystem with 4K stack may cause a stack overflow. For more
16 information, please refer to bugzilla 17630.
17 * NFSv4 reexport requires Lustre client on the re-exporting nodes to be
18 mounted with "32bitapi" mount option
20 Severity : enhancement
21 Description: Update RHEL5.6 kernel to 2.6.18-238.19.1.el5
23 Severity : enhancement
24 Description: Update RHEL5.6 kernel to 2.6.18-238.9.1.el5
27 Frequency : only with SLES10
29 Description: Use OFED "KMP" provided by Novell
30 Details : SLES10 SP3 ships with OFED in a separate "KMP" package.
31 Lustre is now built against this package. That means you need to
32 install the ofed-kmp package from Novell for the patchless client.
33 Note that the ofed-kmp that Novell ships may not exactly match the
34 kernel version but should still be compatible.
36 Severity : enhancement
38 Description: Update RHEL5.5 kernel to 2.6.18-194.17.1.el5
39 Update OEL5.5 kernel to 2.6.18-194.17.1.0.1.el5.
41 --------------------------------------------------------------------------------
43 2010-07-15 Oracle, Inc.
45 * Server support for kernels:
46 2.6.18-164.11.1.el5 (RHEL 5)
47 2.6.18-164.11.1.0.1.el5 (OEL 5)
48 * Client support for unpatched kernels:
49 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
50 2.6.18-164.11.1.el5 (RHEL 5),
51 2.6.18-164.11.1.0.1.el5 (OEL 5)
52 2.6.16.60-0.42.8 (SLES 10),
54 2.6.29.4-167.fc11 (FC11)
55 * Recommended e2fsprogs version: 1.41.10-sun2
57 Severity : enhancement
59 Description: Distribute lfs_migrate script with lustre
60 Details : The lfs_migrate mentioned in the manual is now included in the
61 lustre tree in lustre/scripts and will distributed in lustre RPMs.
63 Severity : enhancement
65 Description: Failover nid list added to import proc file.
67 Severity : enhancement
69 Description: "som_preview" mount option required to use SOM
70 Details : The clients will need to use som_preview mount option
74 Description: Lustre does not do 1MB IOs to HW RAID
75 Details : Bump MAX_PHYS/HW_SEGMENTS and SG_ALL to 256 in the RHEL5 kernel.
77 Severity : enhancement
79 Description: lfs quota output is not very convenient for awk/sed-parsing
80 Details : Some positions in lfs quota output table could be empty or
81 non-empty which made it hard to parse it with scripts, now
82 a dash is put instead of space where there is not supposed
87 Description: Update to OFED 1.5.1
88 Details : OFED 1.5.1 will be used in environments that don't use
93 Description: Remove lnet. prefix from lctl params display
94 Details : Also renamed lnet.memused to lnet_memused to
99 Description: Use conf_param -d to remove permanant settings.
103 Description: Add list_param -R
104 Details : lctl list_param now supports recursive mode.
107 Bugzilla : 21829, 22312
108 Description: Fix llobdstat.
109 Details : 1. Now updating the write_bytes counter
110 2. Port script fixes from 1.8.
114 Description: decrease the usage of memory on clients.
115 Details : 1. On clients, recycle dentries and inodes unused.
116 2. Delete the code related to ll_deathrow(att 6215 in bug 1443). It
121 Description: llobdstat fix and enhancement
122 Details : add a counter to set a limit to how many samples will be returned
123 fix a wildcard in the path to limit to obdfilter stats only
127 Description: lfs check only allowed for root
128 Details : Code cleanups and added a regresion test of non-root use of
133 Description: MDS Oopses in ptlrpc_dispatch_difficult_reply()/__list_add()
134 Details : Fix hr->hr_index overload.
138 Description: Fix FIEMAP functionality.
142 Description: FC11 patchless client support added
146 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and
147 OEL5.4 kernel to 2.6.18-164.11.1.0.1.el5.
151 Description: New acc-sm script and various test-framework enhancments.
155 Description: Previously landed patch was reverted.
159 Description: Update SLES10 kernel to 2.6.16.60-0.42.8 (patchless)
163 Description: Add ldiskfs maxdirsize mount option.
167 Description: Fix mkfs.lustre support for 16TB LUNs.
171 Description: Print mmp_check_interval
172 Details : print mmp_check_interval and make it possible to abort
173 mount operation in case it takes too long.
177 Description: Add cascading_rw to lustre mpi tests.
181 Description: Store SOM data in EA.
182 Details : Also several SOM related cleanups.
186 Description: Add refcount to md_open_data
190 Description: Use refcount on lov_request_set to prevent memory corruption.
194 Description: Statahead: use parent dir's i_mutex for synchronization.
198 Description: loadgen improvements.
202 Description: Enable peer health detection in ksocklnd and ko2iblnd.
206 Description: Control DCACHE_LUSTRE_INVALID flag with MDS_INODE_LOOKCUP lock
210 Description: Fix race on obd_devcie:md_stats
214 Description: Fix SLES 11 client builds.
215 Details : Resolved uninitalized pointer gcc warnings.
219 Description: Allow non-root access for "lfs check".
220 Details : Added a check in obd_class_ioctl() for OBD_IOC_PING_TARGET.
224 Description: Clear imp_force_reconnect correctly in ptlrpc_connect_interpret()
226 Severity : enhancement
228 Description: Update kernel to RHEL5.4 2.6.18-164.6.1.el5 and
229 OEL5 2.6.18-164.6.1.0.1.el5(Both in-kernel OFED enabled).
233 Description: Sanity test_27p fails.
234 Details : OBD_FAIL_OST_ENOSPC has been moved to filter_statfs().
235 Sanity tests 27 were cleaned up and improved.
239 Description: Unify req format on client/servers
240 Details : Use new req_capsule API [almost] everywhere instead of old PTLRPC
241 buffers and swabbers approach..
246 Description: drop join file stuff.
248 Severity : enhancement
250 Description: Add 2.6.27 and SLES11 (patchless client) support
254 Description: Fix in ptlrpc_expire_one_request() to print the signed time
257 Severity : enhancement
259 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
262 Severity : enhancement
264 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
266 Severity : enhancement
268 Description: correctly handle big reply message.
269 Details : send LNet event if reply is bigger then buffer and adjust this buffer
274 Description: Drop unnecessary __GFP_NOMEMALLOC flag from filter_get_page()
276 Severity : enhancement
278 Description: Add support for OEL5.
282 Description: Conf-sanity.sh 50g test - deactivated OST should not cause a panic.
284 Severity : enhancement
286 Description: "weak-modules" support
287 Details : Implement "weak-modules" support which enables kernel modules
288 to be used with any kernel that implements the same kABI. In
289 order to achieve this modules are now installed in
290 /lib/modules/$(uname -r)/updates/kernel on all distributions.
294 Description: client could not reconnect to OST because of an active request.
295 Details : The client now retries to reconnect to the same server, if a
296 connect request failed with EBUSY or -EAGAIN. Test case 35b
297 conf-sanity.sh has been created.
301 Description: Don't allow make backward step on assignin osc next id.
302 Details : race between allocation next id and ll_sync thread can be cause
303 of set wrong osc next id and can be kill valid ost objects.
307 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
311 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
315 Description: Error handling in osc_statfs_interpret() has been improved.
316 Details : Check in osc_statfs_interpret() for EBADR.
320 Description: Do not send statfs() requests to OSTs disabled by administrator.
321 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
323 Severity : enhancement
325 Description: Added sanity-quota test to for lfs(1) to work with numeric uid/gid.
327 Severity : enhancement
329 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
331 Severity : enhancement
333 Description: Enhancement for lfs(1) command to use numeric uid/gid.
337 Description: Improved cleanup procedure for conf-sanity test_46a.
341 Description: Do not start bulk transfer, if client tried to reconnect.
342 Details : Do not start bulk transfer, if client tried to reconnect,
343 but failed due to active rpcs. Check for exp_abort_active_req
344 has been implemented.
348 Description: Prevent inconsistences between linux and lustre mount structures.
349 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
350 Make the sleep interruptible.
352 Severity : enhancement
354 Description: provide server to client comms path
355 Details : server to client communications path via new LDLM_SET_INFO rpc
359 Description: truncate starts GFP_FS allocation under transaction causing deadlock
360 Details : ldiskfs_truncate calls grab_cache_page which may start page
361 allocation under an open transaction. This may lead to
362 calling prune_icache with consequent lustre reentrance.
366 Description: send correctly lsm on open replay
367 Details : MDS is trust to LSM size on replay open, but client can set wrong size
372 Description: enable client interface failover
373 Details : When a child reconnects from another NID, properly update export
374 nid hash position and ldlm reverse import.
376 Severity : enhancement
378 Description: Communicate OST degraded/readonly state via statfs to MDS
379 Details : Flags in the statfs returned from OSTs indicate whether the
380 OST is in a degraded RAID state, or if the filesystem has
381 turned read-only after a filesystem error is detected.
386 Description: don't panic if EPROTO was hit when reading symlink
387 Details : correctly handling request reference in error cases.
389 Severity : enhancement
391 Description: Add LustreNetLink, a kernel-userspace communcation path.
393 Severity : enhancement
395 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
400 Description: access to llog context before init.
401 Details : move handling CATALOGS file at osc layer and forbid access to llog
405 Frequency : rare, only when client is writing to OST on local node
407 Description: Avoid deadlock for local client writes
408 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
409 memory freeing context. This allows OST threads to set the
410 PF_MEMALLOC flag on task structures in order to allocate memory
411 from reserved pools and complete IO.
412 Use GFP_HIGHUSER for OST allocations for non-local client writes,
413 so that the OST threads generate memory pressure and allow
414 inactive pages to be reclaimed.
416 Severity : enhancement
418 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
423 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
424 Details : move ldlm namespace creation in setup phase to avoid grab
425 _lprocfs_lock with cli_sem held.
429 Description: Temporarily disable grant shrink.
430 Details : Disable the feature for debugging.
433 Frequency : only when formatting test filesystems
435 Description: Unable to run several mkfs.lustre on loop devices at the same time
436 Details : mkfs.lustre returns error 256 on the concurrent loop devices
437 formatting. The solution is to proper handle the error.
439 Severity : enhancement
441 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
443 Severity : enhancement
445 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
450 Description: resolve race between obd_disconnect and class_disconnect_exports
451 Details : if obd_disconnect will be called to already disconnected export he
452 forget release one reference and osc module can't unloaded.
454 Severity : enhancement
456 Description: Allow tuning service thread via /proc
457 Details : For each service a new
458 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
459 created that can be used to set min/max thread counts, and get the
460 current number of running threads.
465 Descriptoin: don't return error if have partially created objects for file.
466 Details : lov_update_create_set uses set->set_success as index for created
467 objects, so if some requests will be failed, they will have hole
468 at end of array and qos_shrink_lsm for allocate correct lsm.
470 Severity : enhancement
472 Description: Update OFED support to 1.4.1
477 Description: Reduce small size read RPC
478 Details : Set read-ahead limit for every file and only do read-ahead when
479 available read-ahead pages are bigger than 1M to avoid small size
482 Severity : enhancement
484 Description: move AT tunable parameters for more consistent usage
485 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
487 Severity : enhancement
489 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
490 Details : allow skip disconnected ost for send statfs request and hide error
496 Description: fix racy locking of mballoc block bitmaps causing BUG
497 Details : The locking of the mballoc buddy bitmap and the in-memory
498 block bitmap was using two different spin locks in some
499 cases. This made it possible to incorrectly access the
500 mballoc bitmap while another process was modifying it,
501 causing a sanity assertion to fail. While no on-disk corruption
502 was reported, there was some risk of this happening.
505 Frequency : rare, on llog test 6
507 Descriptoin: don't allow connect to already connected import
508 Details : allowing connect to already connected import is hide connecting problem.
513 Descriptoin: allow kill process which wait statahead result
514 Details : for some reasons 'ls' can stick in waiting result from statahead,
515 in this case need way for kill this process.
517 Severity : enhancement
519 Description: Add state history info file, enhance import info file
520 Details : Track import connection state changes in a new osc/mdc proc file;
521 add overview-type data to the osc/mdc import proc file.
523 Severity : enhancement
525 Description: MDS create should not wait for statfs RPC while holding DLM lock.
527 Severity : enhancement
529 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
534 Descriptoin: OST grant too much space to client even there are not enough space.
535 Details : Client will shrink its grant cache to OST if there are no write
536 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
537 this grant cache if there are already not enough avaible space
538 (left_space < total_clients * 32M).
541 Frequency : start MDS on uncleanly shutdowned MDS device
543 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
544 Details : stay in waiting mds<>ost recovery finished produce random bugs
545 due race between two ll_sync thread for one lov target. send
546 ACTIVATE event only if connect realy finished and import have
550 Frequency : rare, connect and disconnect target at same time
552 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
553 Details : don't call obd_disconnect under lov_lock. this long time
554 operation and can block ptlrpcd which answer to connect request.
559 Descriptoin: don't lose wakeup for imp_recovery_waitq
560 Details : recover_import_no_retry or invalidate_import and import_close can
561 both sleep on imp_recovery_waitq, but we was send only one wakeup
565 Frequency : always with long access acl
567 Descriptoin: mds can't pack reply with long acl.
568 Details : mds don't control size of acl but they limited by reint/getattr
571 Severity : enhancement
573 Description: Update to SLES10 kernel-2.6.16.60-0.33.
575 Severity : enhancement
577 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
580 Frequency : start MDS on uncleanly shutdowned MDS device
582 Descriptoin: aborting recovery hang on MDS
583 Details : don't throttle destroy RPCs for the MDT.
586 Frequency : on remount
588 Description: external journal device not working after the remount
589 Details : clear dev_rdonly flag for external journal devices in
595 Description: shutdown vs evict race
596 Details : client_disconnect_export vs connect request race.
597 if client will evicted at this time - we start invalidate
598 thread without referece to import and import can be freed
602 Frequency : rare, need acl's on inode.
604 Description: client can't handle ost additional correctly
605 Details : if ost was added after client connected to mds client can have
606 hit lnet_try_match_md ... to big messages to wide striped files.
607 in this case need teach client to handle config events about add
608 lov target and update client max ea size at that event.
610 Severity : enhancement
612 Description: Changelogs
613 Details : Changelogs are a lightweight mechanism to track filesystem
614 metadata and namespace changes. The changelog is recorded
615 permanently on the MDTs, and is periodically "consumed" / purged
616 when records are no longer needed.
618 Severity : enhancement
620 Description: compact fld format with extents
621 Details : Store range of seq rather than every seq in FLD. Seq
622 controller update FLD rather than clients. In Case of CMD, mdt0
623 has FLD, all other metadata server act as non persistent proxy
624 for FLD queries and cache fld entries in fld cache.
629 Description: don't skip ost target if they assigned to file
630 Details : Drop slow OSCs if we can, but not for requested start idx.
631 This means "if OSC is slow and it is not the requested
632 start OST, then it can be skipped, otherwise skip it only
633 if it is inactive/recovering/out-of-space.
637 Description: more cleanup in mds_lov
638 Details : not send LOV EA under replay, we can't know about they size at this
639 time. Don't allow client connect to mds before any ost connected,
640 for avoid problems with LOV EA size and returning EIO to client.
642 Severity : enhancement
644 Description: Interoperability at server side (Disk interoperability)
646 Severity : enhancement
648 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
650 Severity : enhancement
652 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
654 Severity : enhancement
656 Description: New client IO stack (CLIO).
658 Severity : enhancement
660 Description: Commit on sharing. Eliminate inter-client dependencies between
661 uncommitted transactions by doing transaction commits.
662 Thereby clients may recovery independently.
665 Frequency : Create a symlink file with a very long name
667 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
668 Details : If there is no extra space in the request for early cancels,
669 ldlm_req_handles_avail() returns 0 instead of a negative value.
671 Severity : enhancement
673 Description: Add /proc entry for import status
674 Details : The mdc, osc, and mgc import directories now have
675 an import directory that contains useful import data for debugging
678 Severity : enhancement
680 Description: Re-disable certain /proc logging
681 Details : Enable and disable client's offset_stats, extents_stats and
682 extents_stats_per_process stats logging on the fly.
685 Frequency : Only on FC kernels 2.6.22+
687 Description: oops in statahead
688 Details : Do not drop reference count for the dentry from VFS when lookup,
689 VFS will do that by itself.
691 Severity : enhancement
693 Description: Generic /proc file permissions
694 Details : Set /Proc file permissions in a more generic way to enable non-
695 root users operate on some /proc files.
699 Description: Hitting mdc_commit_close() ASSERTION
700 Details : Properly handle request reference release in
701 ll_release_openhandle().
705 Description: quota recovery deadlock during mds failover
706 Details : This patch includes att18982, att18236, att18237 in bz14840.
708 1. fix osts hang when mds does failover with quotaon
709 2. prevent watchdog storm when osts threads wait for the
714 Frequency : only patchless client
715 Description: add workaround for race between add/remove dentry from hash
717 Severity : enhancement
719 Description: Allow OST glimpses to return PW locks
723 Description: LBUG when llog conf file is full
724 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
728 Description: Prevent import from entering FULL state when server in recovery
732 Description: service mount cannot take device name with ":"
733 Details : Only when device name contains ":/" will mount treat it as
739 Description: replace ptlrpcd with the statahead thread to interpret the async
740 statahead RPC callback
744 Frequency : on recovery
745 Description: I/O failures after umount during fail back
746 Details : if client reconnected to restarted server we need join to recovery
747 instead of find server handler is changed and process self
748 eviction with cancel all locks.
750 Severity : enhancement
752 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
754 Severity : enhancement
756 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
758 Severity : enhancement
760 Description: Upcall on Lustre log has been dumped
761 Details : Allow for a user mode script to be called once a Lustre log has
762 been dumped. It passes the filename of the dumped log to the
763 script, the location of the script can be specified via
764 /proc/sys/lnet/debug_log_upcall.
769 Description: avoid idr_remove called for id which is not allocated.
770 Details : Move assigment s_dev for clustered nfs to end of initialization,
771 to avoid problem with error handling.
776 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
777 Details : When connection is reused this not moved from CONN_UNUSED_HASH
778 into CONN_USED_HASH and this prodice warning when put connection
779 again in unused hash.
782 Severity : enhancement
784 Description: File striping can now be set to use an arbitrary pool of OSTs.
786 Severity : enhancement
788 Description: Export bytes_read/bytes_write count on OSC/OST.
792 Description: Early reply size mismatch, MGC loses connection
793 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
794 the connect flags are properly negotiated.
799 Description: quotas are not honored with O_DIRECT
800 Details : all writes with the flag O_DIRECT will use grants which leads to
801 this problem. Now using OBD_BRW_SYNC to guard this.
805 Description: add quota statistics
806 Details : 1. sort out quota proc entries and proc code.
807 2. add quota statistics
809 Severity : enhancement
811 Description: enable quota support for HEAD.
815 Description: Properly propagate oinfo flags from lov to osc for statfs
816 Details : restore missing copy oi_flags to lov requests.
818 Severity : enhancement
820 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
821 : l_getgroups(8), lst(8), routerstat(8)
822 Details : included man pages for llobdstat(8), llstat(8),
823 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
825 Severity : enhancement
827 Description: configure's --enable-quota should check the
828 : kernel .config for CONFIG_QUOTA
829 Details : configure is terminated if --enable-quota is passed but
830 : no quota support is in kernel
834 Description: Remove portals compatibility
835 Details : Remove portals compatibility, not interoperable with releases
840 Description: Resolve device initialization race
841 Details : Prevent proc handler from accessing devices added to the
842 obd_devs array but yet be intialized.
844 Severity : enhancement
846 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
848 Severity : enhancement
850 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
855 Description: Using wrong pointer in osc_brw_prep_request
856 Details : Access to array[-1] can produce panic if kernel compiled with
857 CONFIG_PAGE_ALLOC enabled
861 Description: Client runs out of low memory
862 Details : Consider only lowmem when counting initial number of llap pages
866 Description: *optional* service tags registration
867 Details : if the "service tags" package is installed on a Lustre node
868 When the filesystem is mounted, a local-node service tag will
869 be created. See http://inventory.sun.com/ for more information
870 about the Service Tags asset management system.
874 Description: Kernel BUG tries to release flock
875 Details : Lustre does not destroy flock lock before last reference goes
876 away. So always drop flock locks when client is evicted and
877 perform unlock regardless of successfulness of speaking to MDS.
881 Description: add refcount protection for osc callbacks, avoid panic on shutdown
885 Description: sanity test 65a fails if stripecount of -1 is set
886 Details : handle -1 striping on filesystem in ll_dirstripe_verify
891 Description: ASSERTION(CheckWriteback(page,cmd)) failed
892 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
895 Severity : enhancement
897 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
901 Description: do not process already freed flock
902 Details : flock can possibly be freed by another thread before it reaches
903 to ldlm_flock_completion_ast.
907 Description: LBUG during stress test
908 Details : Need properly lock accesses the flock deadlock detection list.
912 Description: oops in page fault handler
913 Details : kernel page fault handler can return two special 'pages' in error
914 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
918 Description: timeout with invalidate import.
919 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
920 should be handled by ptlrpcd. This produce long age waiting and
921 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
923 Severity : enhancement
925 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
929 Description: enable MGS and MDT services start separately
930 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
931 a co-located MGS without starting the MGS, which is a complement
932 to 'nosvc' mount option.
936 Frequency : after recovery
937 Description: precreate to many object's after del orphan.
938 Details : del orphan st in oscc last_id == next_id and this triger growing
939 count of precreated objects. Set flag LOW to skip increase count
940 of precreated objects.
944 Frequency : rare, on clear nid stats
945 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
946 Details : when clean nid stats sometimes try destroy live entry,
947 and this produce panic in free.
951 Description: Stack overflow during MDS log replay
952 ease stack pressure by using a thread dealing llog_process.
956 Description: wait until IO finished before start new when do lock cancel.
957 Details : VM protocol want old IO finished before start new, in this case
958 need wait until PG_writeback is cleared until check dirty flag and
959 call writepages in lock cancel callback.
961 Severity : enhancement
963 Description: using special macro for print time and cleanup in includes.
967 Description: mds_mfd_close() ASSERTION(rc == 0)
968 Details : In mds_mfd_close(), we need protect inode's writecount change
969 within its orphan write semaphore to prevent possible races.
973 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
977 Frequency : rare, on shutdown ost
978 Description: don't hit live lock with umount ost.
979 Details : shrink_dcache_parent can be in long loop with destroy dentries,
980 use shrink_dcache_sb instead.
984 Description: don't panic with use echo client
985 Details : echo client pass NULL as client nid pointer and this produce null
990 Description: fix build on ppc32
991 Details : compile code with -m64 flag produce wrong object file for ppc32.
995 Description: add message levels for liblustreapi
999 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
1000 Details : Move the 'good_osts' check before the 'total_bavail' check. This
1001 will result in an -EAGAIN and in the exit call path we call
1002 alloc_rr() which will with increasing aggressiveness attempt to
1003 aquire precreated objects on the minimum number of required OSCs.
1007 Description: Use old size assignment to avoid deadlock
1008 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
1009 the scheduling while holding a spinlock. We do not need locking
1010 for size in ll_update_inode() because size is only updated from
1011 the MDS for directories or files without objects, so there is no
1012 other place to do the update, and concurrent access to such inodes
1013 are protected by the inode lock.
1017 Description: resolve "_IOWR redefined" build error on SLES10
1021 Description: dump the memory debugging after all modules are unloaded to
1022 suppress false negative in conf_sanity test 39
1024 Severity : enhancement
1026 Description: build kernel-ib packages for OFED 1.3 in our release cycle
1031 Description: fix SLES kernel versioning
1032 Details : the kernel version for our SLES 10 kernel did not include a "-"
1033 before the "smp" at the end. while this was not a problem in
1034 general, it did mean that software trying to use the kernel
1035 version to try to detect a vendor specific kernel would fail.
1036 this was most evident by the OFED build scripts.
1040 Description: Don't update lov_desc members until making sure they are valid
1041 Details : When updating lov_desc members via proc fs, need fix their
1042 validities before doing the real update.
1046 Description: don't put request into delay list while invalidate in flight.
1047 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
1048 import in flight. this produce timeout for invalidate and sometimes
1049 can cause stale data.
1053 Frequency : on ppc only
1054 Description: not convert ost objects for directory because it's not exist.
1055 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
1059 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
1061 Severity : enhancement
1063 Description: Update to support 2.6.22.14 vanilla kernel.
1067 Frequency : rare, on recovery
1068 Description: read procfs can produce deadlock in some situation
1069 Details : Holding lprocfs lock which send rpc can produce block for destroy
1070 obd objects and this also block reconnect with -EALREADY. This
1071 isn't fix all lprocfs bugs - but make it rare.
1073 Severity : enhancement
1075 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
1078 Frequency : frequent on X2 node
1080 Description: mdc_set_open_replay_data LBUG
1081 Details : Set replay data for requests that are eligible for replay.
1085 Description: lustre_mgs: operation 101 on unconnected MGS
1086 Details : When MGC is disconnected from MGS long enough, MGS will evict the
1087 MGC, and late on MGC cannot successfully connect to MGS and a lot
1088 of the error messages complaining that MGS is not connected.
1092 Frequency : on network error
1093 Description: panic with double free request if network error
1094 Details : mdc_finish_enqueue is finish request if any network error ocuring,
1095 but it's true only for synchronus enqueue, for async enqueue
1096 (via ptlrpcd) this incorrect and ptlrpcd want finish request
1099 Severity : enhancement
1101 Description: client-side metadata stat-ahead during readdir(directory readahead)
1102 Details : perform client-side metadata stat-ahead when the client detects
1103 readdir and sequential stat of dir entries therein
1106 Frequency : on start mds
1108 Description: Implement get_info(last_id) in obdfilter.
1111 Frequency : occasional
1113 Description: Correctly check stale fid, not start epoch if ost not support SOM
1114 Details : open with flag O_CREATE need set old fid in op_fid3 because
1115 op_fid2 was overwritten with new generated fid, but mds can answer
1116 with one of these two fids and both is not stale. Setattr
1117 incorrectly started an epoch and assume will be called
1118 done_writing, but without SOM done_writing ever being called.
1121 Frequency : rare, depends on device drivers and load
1123 Description: MDS or OSS nodes crash due to stack overflow
1124 Details : Code changes in 1.8.0 increased the stack usage of some functions.
1125 In some cases, in conjunction with device drivers that use a lot
1126 of stack the MDS (or possibly OSS) service threads could overflow
1127 the stack. One change which was identified to consume additional
1128 stack has been reworked to avoid the extra stack usage.
1131 Frequency : occasional
1133 Description: Do not fail import if osc_interpret_create gets -EAGAIN
1134 Details : If osc_interpret_create got -EAGAIN it immediately exits and
1135 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
1136 oscc_has_objects and see OSC has no objests and call
1137 oscc_internal_create to resend create request.
1139 Severity : enhancement
1141 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
1143 Severity : enhancement
1145 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
1148 Frequency : very rare
1150 Description: Fix replay if there is an un-replied request and open
1151 Details : In some cases, older replay request will revert the
1152 mcd->mcd_last_xid on MDS which is used to record the client's
1153 latest sent request.
1155 Severity : enhancement
1157 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
1159 Severity : enhancement
1161 Description: Add rhel5 support to HEAD.
1163 Serverity : enhancement
1165 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
1170 Description: Don't allow skipping OSTs if index has been specified.
1171 Details : Don't allow skipping OSTs if index has been specified, make
1172 locking in internal create lots better.
1176 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
1177 Details : access to bitfield in structure is always rounded to long
1178 and this produce problem with not atomic change any bit.
1182 Description: Lustre make rpms failed.
1183 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
1184 builds Lustre rpms from tarball.
1187 Frequency : rare on shutdown ost
1189 Description: If llog cancel was not send before clean_exports phase, this can
1190 produce deadlock in llog code.
1191 Details : If llog thread has last reference to obd and call class_import_put
1192 this produce deadlock because llog_cleanup_commit_master wait when
1193 last llog_commit_thread exited, but this never success because was
1194 called from llog_commit_thread.
1198 Description: allow userland application know is lost one of stripes.
1199 Details : fill lvb_blocks with error code on ost and return it to
1200 application if error flag found.
1204 Description: NULL lov_tgts causing MDS oops
1205 Details : more safe checks for NULL lov_tgts for avoid oops.
1207 Severity : enhancement
1209 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
1213 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
1214 Details : If an OST is down the MDS will hang indefinitely in
1215 obd_statfs_rqset() waiting for the statfs data. While for
1216 MDS QOS usage of statfs, it should not stuck in waiting.
1218 Severity : enhancement
1220 Description: remote_acl support
1221 Details : Support ACL-based permission check for remote user.
1222 Support setfacl/getfacl for remote user with the utils
1223 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
1224 the system "{s,g}etfacl" utils.
1226 Severity : enhancement
1228 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
1230 Severity : enhancement
1232 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
1236 Description: make mgs_setparam() handle fsname containing dash
1237 Details : fsname containing a dash does not work with lctl conf_param
1239 Severity : enhancement
1241 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
1243 Severity : enhancement
1245 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
1249 Frequency : rare, at shutdown
1250 Description: access already free / zero obd_namespace.
1251 Details : if client_disconnect_export was called without force flag set,
1252 and exist connect request in flight, this can produce access to
1253 NULL pointer (or already free pointer) when connect_interpret
1254 store ocd flags in obd_namespace.
1258 Frequency : only at startup
1259 Description: not alloc memory with spinlock held.
1260 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
1261 if any spinlock held.
1263 Severity : enhancement
1265 Description: make lustre randomly fail allocating memory
1266 Details : Make lustre randomly failed allocating memory for testing purpose.
1268 Severity : enhancement
1270 Description: lost problems with lov objid file
1271 Details : Fixes some scability and access to not inited memory problems
1272 in work with lov objdid file.
1277 Description: lfs find does not continue on file error
1278 Details : Continue other files processing when a file/dir is absent.
1282 Description: Inconsistent usage of lustre_pack_reply()
1283 Details : Standardize the usage of lustre_pack_reply() such that it
1284 always generate a CERROR on failure.
1287 Frequency : occasional
1289 Description: MDS hang or stay in waiting lock
1290 Details : If client receive lock with CBPENDING flag ldlm need send lock
1291 cancel as separate rpc, to avoid situation when cancel request
1292 can't processed due all i/o threads stay in wait lock.
1296 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1300 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1304 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1308 Description: Update to RHEL4 latest kernel.
1310 Severity : enhancement
1312 Description: Build SLES10 patchless client fails
1313 Details : The configure was broken by run ./configure with
1314 --with-linux-obj=.... argument for patchless client. When the
1315 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1316 can't search header adequately. Use absolute path such as
1317 -I($LINUX)/include instead.
1321 Description: interrupt oig_wait produce painc on resend.
1322 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1323 private set, and this produce situation when rq_ptlrpcd_data not
1324 copyed to new allocated request and triggered LBUG on assert
1325 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1326 ptlrpcd set if request is changed and can be safety changed to use
1331 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1332 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1333 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1334 these macros and replace them with their difinitions instead.
1338 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1339 Details : Update which_patch & target file for SLES10 latest kernel.
1342 Description: Remove client patches from SLES 10 kernel.
1343 Details : This causes SLES 10 clients to behave as patchless clients
1344 even on a Lustre-patched (server) kernel.
1346 Severity : enhancement
1348 Description: self-adjustable client's lru lists
1349 Details : use adaptive algorithm for managing client cached locks lru
1350 lists according to current server load, other client's work
1351 pattern, memory activities, etc. Both, server and client
1352 side namespaces provide number of proc tunables for controlling
1355 Severity : enhancement
1357 Description: light-weight GSS support
1358 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
1359 and reduce performance overhead.
1361 Severity : enhancement
1363 Description: Linux keyring support
1364 Details : Support using service of Linux keyring for Lustre GSS internal
1365 context refresh/cache mechanism.
1369 Description: Fix errors in lfs documentation
1370 Details : Fixes man pages
1374 Description: don't use GFP_* in generic Lustre code.
1375 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
1379 Description: obdclass is limited by single OBD_ALLOC(idarray)
1380 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
1384 Description: warnings with build patchless client with vanila 2.6.19 and up
1385 Details : change old ctl_table style and replace ctl_table/ctl_table_header
1386 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
1390 Frequency : Only for SLES
1391 Description: sanity_quota fail test_1
1392 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
1393 file, which makes TSTID[2] inunique.
1397 Frequency : Only for SLES9
1398 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
1399 Details : Update target/ChangeLog/which_patch .
1403 Frequency : Only for test_10 in sanity-quota.sh
1404 Description: a bug in quota test code
1405 Details : checking a test flag in a wrong place
1409 Frequency : Only for test_13 in sanity-quota.sh
1410 Description: a bug in quota test script
1411 Details : shouldn't check overfull allocated quota
1415 Description: jbd statistics
1416 Details : Port older jbd statistics patch for sles10
1418 Severity : enhancement
1420 Description: Add printing inode info into message about error in writepage.
1425 Description: reply_lock_interpret crash due to race with it and lock cancel.
1426 Details : Do not replay locks that are being cancelled. Do not reference
1427 locks by their address during replay, just by their handle.
1431 Frequency : When flocks are used.
1432 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
1433 Details : Flock locks might destroy just granted lock if it could be merged
1434 with another existing flock, this is done in completion handler,
1435 so teach ldlm_cli_enquque_fini that this is a valid case for
1441 Description: Oops in read and write path when failing to allocate lock.
1442 Details : Check if lock allocation failed and return error back.
1446 Description: lstripe command fails for valid OST index
1447 Details : The stripe offset is compared to lov->desc.ld_tgt_count
1448 instead of lov->desc.ld_active_tgt_count.
1452 Description: sanity.sh failed test 103
1453 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
1458 Description: df doesn't work properly if diskfs blocksize != 4K
1459 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
1463 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
1464 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
1466 Severity : enhancement
1468 Description: Adaptive timeouts
1469 Details : RPC timeouts adapt to changing server load and network
1470 conditions to reduce resend attempts and improve recovery time.
1474 Description: llapi_file_create() does not allow some changes
1475 Details : add llapi_file_open() that allows specifying the mode and
1476 open flags, and also returns an open file handle.
1480 Description: merge and cleanup kernel patches.
1481 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
1485 Description: Add journal checksum support.(Kernel part)
1486 Details : The journal checksum feature adds two new flags i.e
1487 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
1488 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
1489 indicates that the commit block contains the checksum for
1490 the blocks described by the descriptor blocks. Now commit
1491 record can be sent to disk without waiting for descriptor
1492 blocks to be written to disk. This behavior is controlled
1493 using JBD2_FEATURE_ASYNC_COMMIT flag.
1497 Description: OSS needs mutliple precreate threads
1498 Details : Add ability to start more than one create thread per OSS.
1502 Description: Sanity.sh test_65k failed: setstripe should have succeeded
1503 Details : execute lfs setstripe on client
1507 Description: mds_obd_create error creating tmp object
1508 Details : When the user sets quota on root, llog will be affected and can't
1509 create files and write files.
1512 Frequency : Always on ia64 patchless client, and possibly others.
1514 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
1515 Details : This allows the patchless client to be loaded on architectures
1516 without this export.
1520 Description: RedHat Update kernel for RHEL5
1521 Details : Add the kernel config file for RHEL5.
1525 Description: RedHat Update kernel for RHEL5
1526 Details : Modify the kernel config file more closer RHEL5.
1530 Description: Build failure against Centos5 (RHEL5)
1531 Details : Define PAGE_SIZE when it isn't present.
1535 Description: test_7 on sanity-quota.sh can't be used on separate machines
1536 Details : add facet to handle it
1540 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
1541 Details : with NFS, the anon dentry's parent was set to itself in
1542 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
1543 corresponding dentry other than use rec->ur_name.
1545 Severity : enhancement
1547 Description: lfs setstripe enhancement
1548 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
1552 Description: enable data checksumming by default
1553 Details : enable checksum by default, allow --disable-checksum
1554 configure option and "-o nochecksum" mount option. Checksums
1555 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
1559 Description: lustre support for RHEL5
1560 Details : Add support for RHEL5.
1564 Description: Client eviction due to failover config
1565 Details : after a connection loss, the lustre client should attempt to
1566 reconnect to the last active server first before trying the
1567 other potential connections.
1571 Description: when mds and osts use different quota unit(32bit and 64bit),
1572 quota will be released repeatly.
1573 Details : void sending multiple quota reqs to mds, which will keep the
1574 status between the reqs.
1578 Description: osts not allocated evenly to files
1579 Details : change the condition to increase offset_idx
1582 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
1584 Description: Data corruption for OSTs that are formatted larger than 2TB
1586 Details : When generating the bio request for lustre file writes the
1587 sector number would overflow a temporary variable before being
1588 used for the IO. The data reads correctly from Lustre (which
1589 will overflow in a similar manner) but other file data or
1590 filesystem metadata may be corrupted in some cases.
1594 Description: Tune the kernel for good SCSI performance.
1595 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
1596 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
1601 Description: rewrite ext2-derived code in obdclass/uuid.c
1602 Details : rewrite inherited code (uuid parsing code from ext2 utils)
1603 from scratch preserving functionality.
1607 Description: "lfs find -obd UUID" prints directories
1608 Details : "lfs find -obd UUID" will return all directory names instead
1609 of just file names. It is incorrect because the directories
1610 do not reside on the OSTs.
1614 Description: use i_size_read and i_size_write in 2.6 port
1615 Details : replace inode->i_size access with i_size_read/write()
1618 Frequency : when removing large files
1620 Description: scheduling issue during removal of large Lustre files
1621 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
1622 It causes scheduling issues when removing large files (17TB in the
1626 Frequency : only with liblustre clients on XT3
1628 Description: evictions taking too long
1629 Details : allow llrd to evict clients directly on OSTs
1632 Frequency : only on ppc
1634 Description: /proc/fs/lustre/devices broken on ppc
1635 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
1636 We should make sure the type of 'index' is loff_t and then cast
1637 explicitly as needed below. Do not assign an explicitly cast
1641 Frequency : only for rhel5
1643 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
1644 Details : Modify the target file & which_kernel.
1646 Severity : enhancement
1648 Description: omit set fsid for export NFS
1649 Details : fix set/restore device id for avoid EMFILE error and mark lustre
1650 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
1654 Frequency : Always, for kernels after 2.6.16
1655 Description: Fix warning idr_remove called for id=.. which is not allocated.
1656 Details : Last kernels save old s_dev before kill super and not allow
1657 to restore from callback - restore it before call kill_anon_super.
1661 Description: buffer overruns could theoretically occur
1662 Details : llapi_semantic_traverse() modifies the "path" argument by
1663 appending values to the end of the origin string, and a
1664 overrun may occur. Adding buffer overrun check in liblustreapi.
1668 Description: Fix error on 'ls .' at the top of the Lustre mount.
1669 Details : Don't revalidate dentry if it is a root dentry.
1673 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
1674 Details : Modify vm-tunables-rhel4.patch.
1678 Description: Kernel config for 2.6.18-vanilla.
1679 Details : Modify targets/2.6-vanilla.target.in.
1680 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
1681 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
1682 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
1683 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
1687 Description: improve handling recoverable errors
1688 Details : if request processig with error which can be recoverable on server
1689 request should be resend, otherwise page released from cache and
1694 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
1695 Details : Modify target file & which_patch.
1696 A flaw was found in the IA32 system call emulation provided
1697 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
1698 value could be stored in the %RAX register, which could trigger an
1699 out-of-bounds system call table access. An untrusted local user
1700 could exploit this flaw to run code in the kernel
1701 (ie a root privilege escalation). (CVE-2007-4573).
1705 Description: O_DIRECT bypasses client statistics.
1706 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
1707 read_ahead_stats were updated. copied the stats section from
1708 osc_send_oap_rpc() into async_internal().
1712 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
1716 Description: change order of libsysio includes
1717 Details : '#include sysio.h' should always come before '#include xtio.h'
1721 Description: handle "serious error: objid * already exists" more gracefully
1722 Details : If LAST_ID value on disk is smaller than the objects existing in
1723 the O/0/d* directories, it indicates disk corruption and causes an
1724 LBUG(). If the object is 0-length, then we should use the existing
1725 object. This will help to avoid a full fsck in most cases.
1727 Severity : enhancement
1729 Description: adapt the lustre_config script to support the upgrade case
1730 Details : Add "-u" option for lustre_config script to support upgrading 1.4
1731 server targets to 1.6 in parallel.
1735 Description: To avoid grant space > avaible space when the disk is almost
1736 full. Without this patch you might see the error "grant XXXX >
1737 available" or some LBUG about grant, when the disk is almost
1739 Details : In filter_check_grant, for non_grant cache write, we should
1740 check the left space by if (*left > ungranted + bytes), instead
1741 of (*left > ungranted), because only we are sure the left space
1742 is enough for another "bytes", then the ungrant space should be
1743 increase. In client, we should update cl_avail_grant only there
1744 is OBD_MD_FLGRANT in the reply.
1748 Description: Update RHEL 4 kernel to fix local root privilege escalation.
1749 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
1750 described in CVE-2007-4573. This problem could allow untrusted
1751 local users to gain root access.
1754 Frequency : when using O_DIRECT and quotas
1756 Description: Incorrect file ownership on O_DIRECT output files
1757 Details : block usage reported by 'lfs quota' does not take into account
1758 files that have been written with O_DIRECT.
1763 Description: touch file failed when fs is not full
1764 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
1765 otherwise we can get ENOSP while fs is not full.
1769 Description: parallel lock callbacks
1770 Details : Instead of sending blocking and completion callbacks as separated
1771 requests, adding them to a set and sending in parallel.
1774 Frequency : only for Cray XT3
1775 Bugzilla : 12829/13455
1776 Description: Changing primary group doesn't change the group lustre assigns to
1778 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
1779 overridden with the primary group provided by the group upcall,
1780 whereas the supplied fsgid can be trusted if it is in the list of
1781 supplementary groups returned by the group upcall.
1783 Severity : enhancement
1785 Description: Allow masking D_WARNING, D_ERROR messages from console
1786 Details : Console messages can now be disabled via lnet.printk.
1790 Description: User code with malformed file open parameter crashes client node
1791 Details : Before packing join_file req, all the related reference should be
1792 checked carefully in case some malformed flags cause fake
1793 join_file req on client.
1797 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
1798 Details : ldlm_completion_ast() assumes that a lock is granted when the req
1799 mode is equal to the granted mode. However, it should also check
1800 that LDLM_FL_CP_REQD is not set.
1804 Description: Heavy nfs access might result in deadlocks
1805 Details : After ELC code landed, it is now improper to enqueue any mds
1806 locks under och_sem, because enqueue might want to decide to
1807 cancel open locks for same inode we are holding och_sem for.
1811 Description: Client eviction while running blogbench
1812 Details : A lot of unlink operations with concurrent I/O can lead to a
1813 deadlock causing evictions. To address the problem, the number of
1814 oustanding OST_DESTROY requests is now throttled to
1815 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
1816 callbacks are processed in priority.
1820 Description: enable ACLs on MDS by default
1821 Details : ACLs must be enabled on MDS by default.
1824 Frequency : PPC/PPC64 only
1826 Description: conflicts between asm-ppc64/types.h and lustre_types.h
1827 Details : fix duplicated definitions between asm-ppc64/types.h and
1828 lustre_types.h on PPC.
1831 Frequency : PPC/PPC64 only
1833 Description: asm-ppc/segment.h does not exist
1834 Details : fix compile issue on PPC.
1838 Description: better handle error messages in extents code
1841 Frequency : RHEL4 only
1843 Description: mkfs is very slow on IA64/RHEL4
1844 Details : A performance regression has been discovered in the MPT Fusion
1845 driver between versions 3.02.73rh and 3.02.99.00rh. As a
1846 consequence, we have downgraded the MPT Fusion driver in the RHEL4
1847 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
1849 Severity : enhancement
1851 Description: SNMP support enhancement
1852 Details : Adding total number of sampled request for an MDS node in snmp
1855 Severity : enhancement
1857 Description: Optimize ldlm waiting list processing for PR extent locks
1858 Details : When processing waiting list for read extent lock and meeting read
1859 lock that is same or wider to it that is not contended, skip
1860 processing rest of the list and immediatelly return current
1861 status of conflictness, since we are guaranteed there are no
1862 conflicting locks in the rest of the list.
1866 Description: Time out and refuse to reconnect
1867 Details : When the failover node is the primary node, it is possible
1868 to have two identical connections in imp_conn_list. We must
1869 compare not conn's pointers but NIDs, otherwise we can defeat
1870 connection throttling.
1874 Description: port llog fixes from b1_6 into HEAD
1875 Details : Port llog reference couting and some llog cleanups from b1_6
1876 (bug 10800) into HEAD, for protect from panic and access to already
1877 free llog structures.
1881 Description: Detect stride IO mode in read-ahead
1882 Details : When a client does stride read, read-ahead should detect that and
1883 read-ahead pages according to the detected stride pattern.
1887 Description: data checksumming impacts single node performance
1888 Details : add support for several checksum algorithm. Currently, only CRC32
1889 and Adler-32 are supported. The checksum type can be changed on
1890 the fly via /proc/fs/lustre/osc/*/checksum_type.
1894 Description: use adler32 for page checksums
1895 Details : when available, use the Adler-32 algorithm instead of CRC32 for
1900 Description: build for x2 fails
1901 Details : fix compile issue on Cray systems.
1905 Description: Properly match for duplicate locks
1906 Details : Due to different lock order from skiplists code, we need to
1907 traverse entire list for now
1910 Frequency : only on PPC/SLES10
1912 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
1913 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
1914 be defined. Add a hack in mkfs_lustre.c to work around this compile
1919 Description: LASSERT on MDS when client holding flock lock dies
1920 Details : ldlm pool logic depends on number of granted locks equal to
1921 number of released locks which is not true for flock locks, so
1922 just exclude such locks from consideration.
1926 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
1927 Details : Use fsfilt_sync() for both the whole filesystem sync and
1928 individual file sync to eliminate dangerous inode locking
1929 with I_LOCK that can lead to a deadlock.
1933 Description: performance in 1.6.3
1934 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
1935 sure that 1MB requests are not fragmented by the block layer.
1937 Severity : enhancement
1939 Description: organize the server-side client stats on per-nid basis
1940 Details : Change the structure of stats under obdfilter and mds to
1948 The "uuid"s file would list the uuids of _active_ exports.
1949 And the clear entry is to clear all stats and stale nids.
1951 Severity : enhancement
1953 Description: eliminate client locks in face of contention
1954 Details : file contention detection and lockless i/o implementation
1955 for contended files.
1959 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
1962 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
1964 Description: MDS may lose file striping (and hence file data) in some cases
1965 Details : If there are additional extended attributes stored on the MDS,
1966 in particular ACLs, SELinux, or user attributes (if user_xattr
1967 is specified for the client mount options) then there is a risk
1968 of attribute loss. Additionally, the Lustre file striping
1969 needs to be larger than default (e.g. striped over all OSTs),
1970 and an additional attribute must be stored initially in the
1971 inode and then increase in size enough to be moved to the
1972 external attribute block (e.g. ACL growing in size) for file
1977 Description: skiplist implementation simplification
1978 Details : skiplists are used to group compatible locks on granted list
1979 that was implemented as tracking first and last lock of each
1980 lock group the patch changes that to using doubly linked lists
1984 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
1985 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
1986 because LASSERT on both the data supplied by a client, and the
1987 data on disk is dangerous and incorrect.
1989 Severity : enhancement
1991 Description: Slow truncate/writes to huge files at high offsets.
1992 Details : Directly associate cached pages to lock that protect those pages,
1993 this allows us to quickly find what pages to write and remove
1994 once lock callback is received.
1998 Description: more ldlm soft lockups
1999 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
2000 starve other threads from the resource lock for a long time in
2001 case of long waiting queue, so change the debug level from
2002 D_OTHER to the less frequently used D_INFO.
2004 Severity : enhancement
2006 Description: add -gid, -group, -uid, -user options to lfs find
2010 Description: Hung threads in invalidate_inode_pages2_range
2011 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
2012 once one is completed. As a result, some RPCs are stuck in the
2013 queue and are never sent.
2017 Description: filter threads hungs on waiting journal commit
2018 Details : Cleanup filter group llog code, then only filter group llog will
2019 be only created in the MDS/OST syncing process.
2023 Description: Procfs and llog threads access destoryed import sometimes.
2024 Details : Sync the import destoryed process with procfs and llog threads by
2025 the import refcount and semaphore.
2027 Severity : enhancement
2029 Description: openlock cache of b1_6 port to HEAD
2034 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
2035 Details : If insertion of an extent fails, then discard the inode
2036 preallocation and free data blocks else it can lead to duplicate
2041 Description: don't always update ctime in ext3_xattr_set_handle()
2042 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
2043 In some cases the ctime should not be updated, for example for
2044 2.0->1.8 compatibility it is necessary to delete an xattr and it
2045 should not update the ctime.
2049 Bugzilla : 15713/16362
2050 Description: Assertion in iopen_connect_dentry in 1.6.3
2051 Details : looking up an inode via iopen with the wrong generation number can
2052 populate the dcache with a disconneced dentry while the inode
2053 number is in the process of being reallocated. This causes an
2054 assertion failure in iopen since the inode's dentry list contains
2055 both a connected and disconnected dentry.
2059 Description: assertion failure in ldlm_handle2lock()
2060 Details : fix a race between class_handle_unhash() and class_handle2object()
2061 introduced in lustre 1.6.5 by bug 13622.
2066 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
2067 Details : remove the limit of 256 scsi disks in the sd_iostat patch
2072 Description: oops in sd_iostats_seq_show()
2073 Details : unloading/reloading the scsi low level driver triggers a kernel
2074 bug when trying to access the sd iostat file.
2079 Description: Kernel panics during QLogic driver reload
2080 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
2081 patch, causing memory corruption.
2086 Description: journal_dev option does not work in b1_6
2087 Details : pass mount option during pre-mount.
2089 Severity : enhancement
2091 Description: Add a FIEMAP(FIle Extent MAP) ioctl
2092 Details : FIEMAP ioctl will allow an application to efficiently fetch the
2093 extent information of a file. It can be used to map logical blocks
2094 in a file to physical blocks in the block device.
2098 Description: LDLM soft lockups - improvement
2099 Details : It is be possible to send the lock handle along with each read
2100 or write request because the client is already doing a lock match
2101 itself so there isn't any reason the OST should have to re-do that
2105 Frequency : only on Cray X2
2107 Description: X2 build failures
2108 Details : fix build failures on Cray X2.
2112 Description: xid & resent requests
2113 Details : Initialize RPC XID from clock at startup (randomly if clock is
2116 Severity : enhancement
2118 Description: Add lustre_start utility to start or stop multiple Lustre servers
2123 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
2124 Details : In case of memory pressure, list_del() can be called twice on
2125 req->rq_history_list, causing a kernel oops.
2129 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
2130 Details : fix stack overflow in the distributed lock manager by defering
2131 export eviction after a failed AST to the elt thread instead of
2132 handling it in the dlm interpret routine.
2136 Description: Convert some comments to new format.
2137 Details : Update documenting comments to match doxygen conventions.
2141 Description: Grammar fixes.
2142 Details : A couple of trivial spelling fixes.
2146 Description: OSD_COUNTERS-mandatory
2147 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
2148 to simplify the code.
2152 Description: simplify cmm_device freeing logic.
2153 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
2157 Description: Add lockdep support to dt_object_operations locking interface.
2158 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
2159 parameter indicating lock ordering. Update mdd code to use new
2164 Description: Introduce failloc constants for lockless IO tests.
2165 Details : Add two new failloc constants to test lockless IO. Only one of
2166 them in implemented---another is checked in yet to be landed
2171 Description: Add lockdep support for inode mutex.
2172 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
2173 in the situations where Lustre has to lock more than one inode
2178 Description: Add optional invariants checking support.
2179 Details : Add new LINVRNT() macro, optional on new --enable-invariants
2180 configure switch. This macro is to be used for consistency and
2181 sanity checks that are too expensive to be left in `production'
2186 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
2187 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
2192 Description: Add lockdep support for ldlm_lock and ldlm_resource.
2193 Details : Use spin_lock_nested() in (the only) situation where more than
2194 one ldlm_lock is locked simultaneously. Also, fix possible
2195 dead-lock in ldlm_lock_change_resource() by enforcing particular
2200 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
2201 Details : Instead of specifying each ldlm_lock call-back through separate
2202 parameter, wrap them into struct ldlm_callback_suite.
2206 Description: Kill join_lru obd method and its callers.
2207 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
2208 in memory---a requirement implemented through ->o_join_lru() obd
2209 method in HEAD. Remove this method and its users.
2213 Description: Add asynchronous ldlm ENQUEUE completion handler.
2214 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
2215 case that stock ldlm_completion_ast() cannot handle as it waits
2216 until lock is granted. Introduce new ldlm_completion_ast_async()
2217 for this. Also comment ldlm_completion_ast().
2221 Description: ldlm_error <-> errno conversion.
2222 Details : Add functions to map (rather arbitrary) between LDLM error codes
2223 and standard errno values. CLIO needs this to prevent LDLM specific
2224 constants from escaping ldlm and osc.
2228 Description: Kill unused ldlm_handle2lock_ns() function.
2229 Details : Kill unused ldlm_handle2lock_ns() function.
2233 Description: Add lu_ref support to ldlm_lock
2234 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
2235 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
2236 and ldlm_resource. LDLM interface has to be changed, because code that
2237 releases a reference on a lock, has to "know" what reference this is.
2238 In the most frequent case
2240 lock = ldlm_handle2lock(handle);
2242 LDLM_LOCK_PUT(lock);
2244 no changes are required. When any other reference (received _not_ from
2245 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
2246 instead of LDLM_LOCK_PUT().
2248 Arguably, changes are pervasive, and interface requires some discipline
2249 for proper use. On the other hand, it was very instrumental in finding
2250 a few leaked lock references.
2254 Description: Add ldlm_lock_addref_try().
2255 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
2256 attempts to addref a lock that might be being canceled
2261 Description: Add ldlm_weigh_callback().
2262 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
2263 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
2264 hard-coded `number of pages' logic.
2268 Description: Add lockdep annotations to llog code.
2269 Details : Use appropriately tagged _nested() locking calls in the places
2270 where llog takes more than one ->lgh_lock lock.
2274 Description: Add loi_kms_set().
2275 Details : Wrap kms updates into a helper function.
2279 Description: Constify instances of struct lsm_operations.
2280 Details : Constify instances of struct lsm_operations.
2284 Description: lu_conf support.
2285 Details : On a server, a file system object is uniquely identified
2286 by a fid, which is sufficient to locate and load all object
2287 state (inode). On a client, on the other hand, more data are
2288 necessary instantiate an object. Change lu_object_find() and
2289 friends to take additional `lu_conf' argument describing object.
2290 Typically this includes layout information.
2294 Description: lu_context fixes.
2295 Details : Introduce new lu_context functions that are needed on the client
2296 side, where some system threads (ptlrpcd) are shared by multiple
2297 modules, and so cannot be stopped during module shutdown.
2301 Description: Add start and stop methods to lu_device_type_operations.
2302 Details : Introduce two new methods in lu_device_type_operations, that are
2303 invoked when first instance of a given type is created and last one
2304 is destroyed respectively. This is need by CLIO.
2308 Description: Add lu_ref support to struct lu_device.
2309 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
2310 track leaked references.
2314 Description: Introduce lu_kmem_descr.
2315 Details : lu_kmem_descr and its companion interface allow to create
2316 and destroy a number of kmem caches at once.
2320 Description: Fix lu_object finalization race.
2321 Details : Fix a race between lu_object_find() finding an object and its
2322 concurrent finalization. This race is (most likely) not possible
2323 on the server, but might happen on the client.
2327 Description: Introduce lu_ref interface.
2328 Details : lu_ref is a debugging module allowing to track references to
2329 a given object. It is quite cpu expensive, and has to be
2330 explicitly enabled with --enable-lu_ref. See usage description
2335 Description: Factor lu_site procfs stats into a separate function.
2336 Details : Separate lu_site stats printing code into a separate function
2337 that can be reused on a client.
2341 Description: Constify instances of struct {lu,dt,md}_device_operations.
2342 Details : Constify instances of struct {lu,dt,md}_device_operations.
2346 Description: Introduce struct md_site and move meta-data specific parts of
2347 struct lu_site here.
2348 Details : Move md-specific fields out of struct lu_site into special struct
2349 md_site, so that lu_site can be used on a client.
2353 Description: Kill mdd_lov_destroy().
2354 Details : Remove unused mdd code.
2358 Description: Add st_block checking to multistat.c.
2359 Details : Add st_block checking to multistat.c.
2363 Description: Add lu_ref support to struct obd_device.
2364 Details : Add lu_ref tracking to obd_device.
2368 Description: Kill obd_set_fail_loc().
2369 Details : Remove unused code.
2373 Description: Add special type for ptlrpc_request interpret functions.
2374 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
2375 there. Actual usage will be in CLIO.
2379 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
2380 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
2381 doesn't work with lockdep.
2385 Description: Add rwv.c test program.
2386 Details : New testing program exercising readv(2) and writev(2) (Qian).
2390 Description: Add sendfile.c test program.
2391 Details : New testing program exercising sendfile(2) (Jay).
2395 Description: Ratelimit a message that can be very frequent.
2396 Details : Ratelimit a memory allocation failure message that can
2401 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2402 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2406 Description: Liblustre build fixes.
2407 Details : Liblustre build fixes.
2411 Description: libcfs: add cfs_{need,cond}_resched() interface.
2412 Details : libcfs: add cfs_{need,cond}_resched() definition and
2413 implementations for Linux, NT, and liblustre.
2415 Severity : enhancement
2417 Description: More exported tunables for mballoc
2418 Details : Add support for tunable preallocation window and new tunables for
2419 large/small requests
2423 Description: Detect corruption of block bitmap and checking for preallocations
2424 Details : Checks validity of on-disk block bitmap. Also it does better
2425 checking of number of applied preallocations. When corruption is
2426 found, it turns filesystem readonly to prevent further corruptions.
2430 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
2431 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
2432 window to 0. If it does hit cache miss, and read-pattern is still
2433 stride-io mode, does not reset the stride window, but also does not
2434 increase the stride window length in this case.
2438 Frequency : only for big-endian servers
2439 Description: Check if system is big-endian while mounting fs with extents feature
2440 Details : Mounting a filesystem with extents feature will fail on big-endian
2441 systems since ext3-based ldiskfs is not supported on big-endian
2442 systems. This can be over-riden with "bigendian_extents" mount option.
2444 Severity : enhancement
2446 Description: The root squash functionality
2447 Details : A security feature, which is to prevent users from being able
2448 to mount lustre on their desktop, run as root, and delete
2449 all of the files in the filesystem. The goal is accomplished by
2450 remapping user id (UID) and group id (GID) of the root user to
2451 a UID and GID specified by the system administartor via Lustre
2452 configuration management server (MGS). The functionality also
2453 allows to specify sets of clients for which the remapping does
2458 Description: Excessive recovery window
2459 Details : With AT enabled, the recovery window can be excessively long (6000+
2460 seconds). To address this problem, we no longer use
2461 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
2462 timeout no longer depends on the service time, it is set to
2463 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
2464 time via pb_service_time.
2468 Description: Watchdog triggered on MDS failover
2469 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
2470 the OSTs know that the MDS "UUID" can be reused for the same export
2471 from a different NID, so we do not need to wait for the export to be
2475 Frequency : rare, only if using MMP with Linux RAID
2477 Description: MMP doesn't work with Linux RAID
2478 Details : While using HA for Lustre servers with Linux RAID, it is possible
2479 that MMP will not detect multiple mounts. To make this work we
2480 need to unplug the device queue in RAID when the MMP block is being
2481 written. Also while reading the MMP block, we should read it from
2482 disk and not the cached one.
2484 Severity : enhancement
2486 Description: open file using fid
2487 Details : A file can be opened using just its fid, like
2488 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
2491 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
2494 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
2495 Details : add proper locking for creation of jbd_4k slab cache
2499 Description: MMP check in ext3_remount() fails without displaying any error
2500 Details : When multiple mount protection fails during remount, proper error
2503 Severity : enhancement
2505 Description: Allow stripe size to be up to 4G-64k
2506 Details : Fix math logic to allow large stripe sizes.
2510 Description: add check for >8TB ldiskfs filesystems
2511 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
2512 Don't allow >8TB ldiskfs filesystems to be mounted without
2513 force_over_8tb mount option
2518 Descriptoin: OSS complains "too many blocks" and crashes
2519 Details : In filter_commitrw_write():
2520 1. lquota_chkquota();
2521 2. obdo_from_inode();
2522 3. lquota_pending_commit();
2523 "2." has changed uid and gid in this case, then caused
2524 lquota_chkquota() and lquota_pending_commit() to use
2525 different uid and gid. That is the root of the bug.
2529 Descriptoin: getstripe doesn't report stripe count = -1 sanely
2530 Details : Detect and print out the 'stripe count = -1' case properly.
2534 Description: Out or order replies might be lost on replay
2535 Details : In ptlrpc_retain_replayable_request if we cannot find retained
2536 request with tid smaller then one currently being added, add it
2537 to the start, not end of the list.
2539 --------------------------------------------------------------------------------
2541 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
2543 * Support for kernels:
2544 2.6.5-7.283 (SLES 9),
2545 2.6.9-55.EL (RHEL 4),
2546 2.6.16.46-0.14 (SLES 10),
2547 2.6.18.8 vanilla (kernel.org)
2548 * Client support for unpatched kernels:
2549 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2550 2.6.16 - 2.6.22 vanilla (kernel.org)
2551 * Due to recently discovered recovery problems, we do not recommend
2552 using patchless RHEL 4 clients with this or any earlier release.
2553 * Recommended e2fsprogs version: 1.39.cfs8
2554 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2555 * Starting with this release, the ldiskfs backing filesystem required
2556 by Lustre is now in its own package, lustre-ldiskfs. This package
2557 should be installed. It is versioned separately from Lustre and
2558 may be released separately in future.
2562 Description: block reactivating mgc import until all deactivates complete
2563 Details : Fix race when failing back MDT/MGS to itself (testing)
2566 Frequency : at statup only
2568 Description: mds_lov_synchronize race leads to various problems
2569 Details : simultaneous MDT->OST connections at startup can cause the
2570 sync to abort, leaving the OSC in a bad state.
2572 Severity : enhancement
2574 Description: add optional extra BUILD_VERSION info
2575 Details : add a new environment variable (namely LUSTRE_VERS) which allows
2576 to override the lustre version.
2579 Frequency : 2.6.18 servers only
2581 Description: ll_kern_mount() doesn't release the module reference
2582 Details : The ldiskfs module reference count never drops down to 0
2583 because ll_kern_mount() doesn't release the module reference.
2588 Description: server LBUG when using old ost_num_threads parameter
2589 Details : Accept the old ost_num_threads parameter but warn that it
2590 is deprecated, and fix an off-by-one error that caused an LBUG.
2595 Description: Transient SCSI error results in persistent IO issue
2596 Details : iobuf->dr_error is not reinitialized to 0 between two
2600 Frequency : sometimes when underlying device returns I/O errors
2602 Description: OSTs not going read-only during write failures
2603 Details : OSTs are not remounted read-only when the journal commit threads
2604 get I/O errors because fsfilt_ext3 calls journal_start/stop()
2605 instead of the ext3 wrappers.
2609 Description: poor connect scaling with increasing client count
2610 Details : Don't run filter_grant_sanity_check for more than 100 exports
2611 to improve scaling for large numbers of clients.
2614 Frequency : SLES10 only
2616 Description: sanity-quota.sh quotacheck failed: rc = -22
2617 Details : Quotas cannot be enabled on SLES10.
2620 Frequency : liblustre clients only
2622 Description: getdirentries does not give error when run on compute nodes
2623 Details : getdirentries does not fail when the size specified as an argument
2624 is too small to contain at least one entry
2626 Severity : enhancement
2628 Description: Add LNET router traceability for debug purposes
2629 Details : If a checksum failure occurs with a router as part of the
2630 IO path, the NID of the last router that forwarded the bulk data
2631 is printed so it can be identified.
2636 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
2637 Details : Due to a race condition, liblustre clients were occasionally
2638 evicted incorrectly.
2640 Severity : enhancement
2642 Description: lfs setstripe use optional parameters instead of postional
2645 Severity : enhancement
2647 Description: Nanosecond timestamp support for ldiskfs
2648 Details : The on-disk ldiskfs filesystem has added support for nanosecond
2649 resolution timestamps. There is not yet support for this at
2650 the Lustre filesystem level.
2653 Frequency : during server recovery
2655 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
2656 Details : request with rq_no_resend flag not awake l_wait_event if they get
2660 Frequency : nfs export on patchless client
2662 Description: connectathon hang when test nfs export over patchless client
2663 Details : Disconnected dentry cannot be found with lookup, so we do not need
2664 to unhash it or make it invalid
2667 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
2668 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
2669 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
2674 Description: open req refcounting wrong on reconnect
2675 Details : If reconnect happened between getting open reply from server and
2676 call to mdc_set_replay_data in ll_file_open, we will schedule
2677 replay for unreferenced request that we are about to free.
2678 Subsequent close will crash in variety of ways.
2679 Check that request is still eligible for replay in
2680 mdc_set_replay_data().
2685 Description: disable writes to filesystem when reading health_check file
2686 Details : the default for reading the health_check proc file has changed
2687 to NOT do a journal transaction and write to disk, because this
2688 can cause reads of the /proc file to hang and block HA state
2689 checking on a healthy but otherwise heavily loaded system. It
2690 is possible to return to the previous behaviour during configure
2691 with --enable-health-write.
2693 Severity : enhancement
2695 Description: 64-bit inode version
2696 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
2697 to the inode. This will be required for version-based recovery.
2702 Description: MDS fails to start if a duplicate client export is detected
2703 Details : in some rare cases it was possible for a client to connect to
2704 an MDS multiple times. Upon recovery the MDS would detect this
2705 and fail during startup. Handle this more gracefully.
2707 Severity : enhancement
2709 Description: Add -o localflock option to simulate old noflock
2711 Details : This will achieve local-only flock/fcntl locks
2717 Description: log_commit_thread vs filter_destroy race leads to crash
2718 Details : Take import reference before releasing llog record semaphore
2723 Description: Wrong request locking in request set processing
2724 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
2725 imp_delayed_list, in this place should be used imp_lock.
2728 Frequency : when reconnection
2730 Description: Grant Leak when osc reconnect to OST
2731 Details : When osc reconnect ost, OST(filter) should check whether it
2732 should grant more space to client by comparing fed_grant and
2733 cl_avail_grant, and return the granted space to client instead
2734 of "new granted" space, because client will call osc_init_grant
2735 to update the client grant space info.
2738 Frequency : when client reconnect to OST
2740 Description: Grant Leak when osc do resend and replay bulk write
2741 Details : When osc reconnect to OST, OST(filter)should clear grant info of
2742 bulk write request, because the grant info while be sync between
2743 OSC and OST when reconnect, and we should ignore the grant info
2744 these of resend/replay write req.
2749 Description: Grant space more than avaiable left space sometimes.
2750 Details : When then OST is about to be full, if two bulk writing from
2751 different clients came to OST. Accord the avaliable space of the
2752 OST, the first req should be permitted, and the second one
2753 should be denied by ENOSPC. But if the seconde arrived before
2754 the first one is commited. The OST might wrongly permit second
2755 writing, which will cause grant space > avaiable space.
2758 Frequency : when client is evicted
2760 Description: Grant might be wrongly erased when osc is evicted by OST
2761 Details : when the import is evicted by server, it will fork another
2762 thread ptlrpc_invalidate_import_thread to invalidate the
2763 import, where the grant will be set to 0. While the original
2764 thread will update the grant it got when connecting. So if
2765 the former happened latter, the grant will be wrongly errased
2766 because of this race.
2771 Description: Checking Stale with correct fid
2772 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
2773 to check whether it is stale, because sometimes, we want the
2774 enqueue happened anyway, and op_data.fid2 will not be initialized.
2776 Severity : enhancement
2778 Description: update patchless client
2779 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
2782 Frequency : only with 2.4 kernel
2784 Description: random memory corruption
2785 Details : size of struct ll_inode_info is to big for union inode.u and this
2786 can be cause of random memory corruption.
2791 Description: Memory leak in recovery
2792 Details : Lov_mds_md was not free in an error handler in mds_create_object.
2793 It should also check obd_fail before fsfilt_start, otherwise if
2794 fsfilt_start return -EROFS,(failover mds during mds recovery).
2795 then the req will return with repmsg->transno = 0 and rc = EROFS.
2796 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
2797 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
2798 be freed no matter whether fsfilt_commit success or not.
2801 Frequency : only with huge count clients
2803 Description: Prevents from taking the superblock lock in llap_from_page for
2805 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
2806 for llap_from_page call in ll_removepage prevents from taking the
2807 superblock lock for a soon died page.
2812 Description: Not check open intent error before release open handle
2813 Details : in some rare cases, the open intent error is not checked before
2814 release open handle, which may cause
2815 ASSERTION(open_req->rq_transno != 0), because it tries to release
2816 the failed open handle.
2821 Description: Set cat log bitmap only after create log success.
2822 Details : in some rare cases, the cat log bitmap is set too early. and it
2823 should be set only after create log success.
2827 Description: Accessing a block bevice can re-enable I/O when Lustre is
2828 tearing down a device.
2829 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
2834 Description: service threads may hog cpus when there are a lot of requests
2835 Details : Insert cond_resched to give other threads a chance to use some CPU
2840 Description: the cat log was not initialized in recovery
2841 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
2842 unlink log on mds will not be initialized until mds post
2843 recovery. And also in mds post recovery, the unlink log will
2844 initialization will be done asynchronausly, so there will be race
2845 between add unlink log and unlink log initialization.
2849 Description: brw_stats were being printed incorrectly
2850 Details : brw_stats were being printed as log2 but all of them were not
2851 recorded as log2. Also remove some code duplication arising from
2852 filter_tally_{read,write}.
2856 Frequency : rare, only in recovery.
2857 Description: ASSERTION(req->rq_type != LI_POISON) failed
2858 Details : imp_lock should be held while iterating over imp_sending_list for
2859 prevent destroy request after get timeout in ptlrpc_queue_wait.
2863 Description: replay-single.sh test 52 fails
2864 Details : A lock's skiplist need to be cleanup when it being unlinked
2865 from its resource list.
2869 Description: Short directio read returns full requested size rather than
2871 Details : Direct I/O operations should return actual amount of bytes
2872 transferred rather than requested size.
2874 Severity : enhancement
2876 Description: metadata RPC reduction (e.g. for rm performance)
2877 Details : decrease the amount of synchronous RPC between clients and servers
2878 by canceling conflicing lock before the operation on the client
2879 and packing thier handles into the main operation RPC to server.
2881 Severity : enhancement
2883 Description: Async OSC create to avoid the blocking unnecessarily.
2884 Details : If a OST has no remain object, system will block on the creating
2885 when need to create a new object on this OST. Now, ways use
2886 pre-created objects when available, instead of blocking on an
2887 empty osc while others are not empty. If we must block, we block
2888 for the shortest possible period of time.
2892 Description: improve handling recoverable errors
2893 Details : if request processig with error which can be recoverable on server
2894 request should be resend, otherwise page released from cache and
2897 Severity : enhancement
2899 Description: refine locking for avoid write wrong info into lov_objid file
2900 Details : fix possible races with add new target and write/update data in
2903 --------------------------------------------------------------------------------
2905 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
2910 Frequency : on some architectures
2912 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
2913 Details : Uninitialized flags sometimes cause configuration commands to
2917 Frequency : patchless clients only
2919 Description: missing __iget() symbol export
2920 Details : The __iget() symbol export is missing. To avoid the need for
2921 this on patchless clients the deathrow inode reaper is turned
2922 off, and we depend on the VM to clean up old inodes. This
2923 dependency was during via the fix for bug 12181.
2925 --------------------------------------------------------------------------------
2927 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
2929 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
2930 INTEROPERATE with older versions automatically. In many cases a
2931 special upgrade step is needed. Please read the
2932 user documentation before upgrading any part of a live system.
2933 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
2934 version will not interoperate with 1.6 betas before beta5 (1.5.95).
2935 * WARNING: Lustre configuration and startup changes are required with
2936 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
2940 Severity : enhancement
2942 Description: MountConf
2943 Details : Lustre configuration is now managed via mkfs and mount
2944 commands instead of lmc and lconf. New obd types (MGS, MGC)
2945 are added for dynamic configuration management. See
2946 https://mail.clusterfs.com/wikis/lustre/MountConf for
2949 Severity : enhancement
2951 Description: dynamic OST addition
2952 Details : OSTs can now be added to a live filesystem
2954 Severity : enhancement
2956 Description: startup order invariance
2957 Details : MDTs and OSTs can be started in any order. Clients only
2958 require the MDT to complete startup.
2960 Severity : enhancement
2962 Description: parallel, asynchronous orphan cleanup
2963 Details : orphan cleanup is now performed in separate threads for each
2964 OST, allowing parallel non-blocking operation.
2966 Severity : enhancement
2968 Description: optimized stripe assignment
2969 Details : stripe assignments are now made based on ost space available,
2970 ost previous usage, and OSS previous usage, in order to try
2971 to optimize storage space and networking resources.
2973 Severity : enhancement
2975 Description: Permanently set tunables
2976 Details : All writable /proc/fs/lustre tunables can now be permanently
2977 set on a per-server basis, at mkfs time or on a live system.
2979 Severity : enhancement
2981 Description: Lustre message v2
2982 Details : Add lustre message format v2.
2984 Severity : enhancement
2986 Description: client OST exclusion list
2987 Details : Clients can be started with a list of OSTs that should be
2988 declared "inactive" for known non-responsive OSTs.
2992 Description: SPEC SFS validation failure on NFS v2 over lustre.
2993 Details : Changes the blocksize for regular files to be 2x RPC size,
2994 and not depend on stripe size.
2996 Severity : enhancement
2998 Description: Multiple MD RPCs in flight.
2999 Details : Further unserialise some read-only MDS RPCs - learn about intents.
3000 To avoid overly-overloading MDS, introduce a limit on number of
3001 MDS RPCs in flight for a single client and add /proc controls
3002 to adjust this limit.
3004 Severity : enhancement
3006 Description: client read/write statistics
3007 Details : Add client read/write call usage stats for performance
3008 analysis of user processes.
3009 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
3010 file access. extents_stats shows chunk size distribution.
3011 extents_stats_per_process show chunk size distribution per
3014 Severity : enhancement
3016 Description: mds statistics
3017 Details : Add detailed mds operations statistics in
3018 /proc/fs/lustre/mds/*/stats.
3022 Description: Failure of copying files with lustre special EAs.
3023 Details : Client side always return success for setxattr call for lustre
3024 special xattr (currently only "trusted.lov").
3029 Description: Refcount LNET uuids
3030 Details : The global LNET uuid list grew linearly with every startup;
3031 refcount repeated list entries instead of always adding to
3034 Severity : enhancement
3036 Description: Dynamic service threads
3037 Details : Within a small range, start extra service threads
3038 automatically when the request queue builds up.
3041 Frequency : mixed-endian client/server environments
3043 Description: mixed-endian crashes
3044 Details : The new msg_v2 system had some failures in mixed-endian
3047 Severity : enhancement
3049 Description: Easy OST removal
3050 Details : OSTs can be permanently deactivated with e.g. 'lctl
3051 conf_param lustre-OST0001.osc.active=0'
3053 Severity : enhancement
3055 Description: MGS proc entries
3056 Details : Added basic proc entries for the MGS showing what filesystems
3059 Severity : enhancement
3061 Description: provide MGS failover
3062 Details : Added config lock reacquisition after MGS server failover.
3064 Severity : enhancement
3066 Description: add Linux 2.4 support
3067 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
3071 Description: a large application tries to do I/O to the same resource and dies
3072 in the middle of it.
3073 Details : Check the req->rq_arrival time after the call to
3074 ost_brw_lock_get(), but before we do anything about
3075 processing it & sending the BULK transfer request. This
3076 should help move old stale pending locks off the queue as
3077 quickly as obd_timeout.
3080 Frequency : when an incorrect nid is specified during startup
3082 Description: ptlrpc connect to non-existant node causes kernel crash
3083 Details : LNET can't be re-entered from an event callback, which
3084 happened when we expire a message after the export has been
3085 cleaned up. Instead, hand the zombie cleanup off to another
3088 Severity : enhancement
3090 Description: plain/inodebits lock performance improvement
3091 Details : Grouping plain/inodebits in granted list by their request modes
3092 and bits policy, thus improving the performance of search through
3096 Frequency : only if OST filesystem is corrupted
3098 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3099 Details : for a short time RPCs with bulk IO are in the replay list,
3100 but replay of bulk IOs is unimplemented. If the OST filesystem
3101 is corrupted due to disk cache incoherency and then replay is
3102 started it is possible to trip an assertion. Avoid putting
3103 committed RPCs into the replay list at all to avoid this issue.
3106 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3108 Description: System hang on startup
3109 Details : This bug allowed the liblustre (e.g. catamount) client to
3110 return to the app before handling all startup RPCs. This
3111 could leave the node unresponsive to lustre network traffic
3112 and manifested as a server ptllnd timeout.
3114 Severity : enhancement
3116 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3117 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
3118 boolean to control whether to print peer debug info when a
3119 client's RPC times out.
3122 Frequency : only for kernels with patches from Lustre below 1.4.3
3124 Description: Remove old rdonly API
3125 Details : Remove old rdonly API which unused from at least lustre 1.4.3
3128 Frequency : only for devices with external journals
3130 Description: Set external device read-only also
3131 Details : During a commanded failover stop, we set the disk device
3132 read-only while the server shuts down. We now also set any
3133 external journal device read-only at the same time.
3136 Frequency : when upgrading from 1.4 while trying to change parameters
3138 Description: The wrong (new) MDC name was used when setting parameters for
3139 upgraded MDT's. Also allows changing of OSC (and MDC)
3140 parameters if --writeconf is specified at tunefs upgrade time.
3143 Frequency : when setting specific ost indicies
3145 Description: QOS code breaks on skipped indicies
3146 Details : Add checks for missing OST indicies in the QOS code, so OSTs
3147 created with --index need not be sequential.
3149 Severity : enhancement
3151 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3152 Details : The uninit_groups feature works in conjunction with the kernel
3153 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3154 up the pass1 processing of e2fsck. This is a read-only feature
3155 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3156 mount filesystems that have had this feature enabled.
3158 Severity : enhancement
3160 Description: Improve multi-block allocation algorithm to avoid fragmentation
3161 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3162 allocation locality and avoid filesystem fragmentation.
3164 ------------------------------------------------------------------------------
3166 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
3168 * Support for kernels:
3169 2.4.21-47.0.1.EL (RHEL 3)
3170 2.6.5-7.283 (SLES 9)
3171 2.6.9-42.0.10.EL (RHEL 4)
3172 2.6.12.6 vanilla (kernel.org)
3173 2.6.16.27-0.9 (SLES 10)
3174 * Recommended e2fsprogs version: 1.39.cfs5
3176 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3180 Frequency : occasional, depends on client load and configuration
3181 Bugzilla : 12181, 12203
3182 Description: data loss for recently-modified files
3184 Details : In some cases it is possible that recently written or created
3185 files may not be written to disk in a timely manner (this should
3186 normally be within 30s unless client IO load is very high).
3187 The problem appears as zero-length files or files that are a
3188 multiple of 1MB in size after a client crash or client eviction
3189 that are missing data at the end of the file.
3191 This problem is more likely to be hit on clients where files are
3192 repeatedly created and unlinked in the same directory, clients
3193 have a large amount of RAM, have many CPUs, the filesystem has
3194 many OSTs, the clients are rebooted frequently, and/or the files
3195 are not accessed by other nodes after being written.
3197 The presence of the problem can be detected by looking at
3198 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
3199 smaller than the second (nr_unused) then dirty files will not
3200 be flushed automatically to disk. "sync; sleep 10" should be
3201 run several times on the node before unmounting it to update
3202 Lustre (this is also safe to run on nodes without this problem).
3204 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
3205 that can cause this same problem, so customers using that kernel
3206 also need to update the kernel in addition to Lustre. In order
3207 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
3209 It is normal that files written just before a client crash (less
3210 than 30s) may not yet have been flushed to disk, even for local
3214 Frequency : frequent on thin XT3 nodes
3216 Description: UUID collision on thin XT3 Linux nodes
3217 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
3218 randomly, since we relied on an insufficiently-seeded PRNG.
3223 Description: OSS hangs after "All ost request buffers busy"
3224 Details : A deadlock between quota and journal operations caused OSS
3225 hangs after printing "All ost request buffers busy."
3228 Frequency : always on liblustre builds
3230 Description: Cleanup compiler warnings on liblustre
3233 Frequency : always on liblustre builds on XT3
3235 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
3239 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
3240 > 32000 subdirectories
3241 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
3242 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
3243 subdirectory count crosses 32000. This will aid e2fsck to
3244 correctly handle more than 32000 subdirectories.
3247 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3249 Description: System hang on startup
3250 Details : This bug allowed the liblustre (e.g. catamount) client to
3251 return to the app before handling all startup RPCs. This
3252 could leave the node unresponsive to lustre network traffic
3253 and manifested as a server ptllnd timeout.
3255 Severity : enhancement
3257 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3258 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
3259 boolean to control whether to print peer debug info when a
3260 client's RPC times out.
3265 Description: make O_SYNC working on 2.6 kernels
3266 Details : 2.6 kernels use different method for mark pages for write,
3267 so need add a code to lustre for O_SYNC work.
3272 Description: Failure to close file and release space on NFS
3273 Details : Put inode details into lock acquired in ll_intent_file_open.
3274 Use mdc_intent_lock in ll_intent_open to properly
3275 detect all kind of errors unhandled by mdc_enqueue.
3280 Description: proc file read during shutdown sometimes raced obd removal,
3282 Details : Add lock to prevent obd access after proc file removal.
3285 Frequency : Only for files larger than 4GB on 32-bit clients.
3287 Description: improperly doing page alignment of locks
3288 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
3289 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3294 Description: RPCs being resent when they shouldn't be
3295 Details : Some RPCs that should not be resent are being resent. This
3296 can cause inconsistencies in the RPC state machine. Do not
3297 resend such requests.
3300 Frequency : rare, only with NFS export
3302 Description: Crash on NFS re-export node
3303 Details : under very unusual load conditions an assertion is hit in
3304 ll_intent_file_open()
3307 Frequency : only if OST filesystem is corrupted
3309 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3310 Details : for a short time RPCs with bulk IO are in the replay list,
3311 but replay of bulk IOs is unimplemented. If the OST filesystem
3312 is corrupted due to disk cache incoherency and then replay is
3313 started it is possible to trip an assertion. Avoid putting
3314 committed RPCs into the replay list at all to avoid this issue.
3319 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3320 Details : Large single O_DIRECT read and write calls can fail to allocate
3321 a sufficiently large buffer to process the request. In case of
3322 allocation failure the allocation is retried with a smaller
3323 buffer and broken into smaller requests.
3325 Severity : enhancement
3327 Description: Add -o localflock option to simulate old noflock behaviour.
3328 Details : This will achieve local-only flock/fcntl locks coherentness.
3333 Description: versioning check is incomplete
3334 Details : Checking the version difference of client vs. server, report
3335 error if the gap is too big.
3340 Description: add support PG_writeback bit
3341 Details : add support for PG_writeback bit for Lustre, for more carefull
3342 work with page cache in 2.6 kernel. This also fix some deadlocks
3343 and remove hack for work O_SYNC with 2.6 kernel.
3345 Severity : enhancement
3347 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3348 Details : The uninit_groups feature works in conjunction with the kernel
3349 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3350 up the pass1 processing of e2fsck. This is a read-only feature
3351 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3352 mount filesystems that have had this feature enabled.
3354 Severity : enhancement
3356 Description: Improve multi-block allocation algorithm to avoid fragmentation
3357 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3358 allocation locality and avoid filesystem fragmentation.
3360 ------------------------------------------------------------------------------
3362 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3364 * Support for kernels:
3365 2.6.9-42.0.3.EL (RHEL 4)
3366 2.6.5-7.276 (SLES 9)
3367 2.4.21-47.0.1.EL (RHEL 3)
3368 2.6.12.6 vanilla (kernel.org)
3369 2.6.16.21-0.8 (SLES10)
3370 * Recommended e2fsprogs version: 1.39.cfs2-0
3372 * The backwards-compatible /proc/sys/portals symlink has been removed
3373 in this release. Before upgrading, please ensure that you change
3374 any configuration scripts or /etc/sysctl.conf files that access
3375 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3376 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3377 in advance of the upgrade on any system running Lustre 1.4.6 or
3378 newer, since /proc/sys/lnet was added in that version.
3379 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3383 Frequency : only when quota is used
3385 Description: avoid scanning export list for quota master
3386 Details : Change the algorithms to avoid scanning export list in order
3387 to improve the efficiency.
3390 Frequency : MDS failover only, very rarely
3392 Description: "went back in time" messages on mds failover
3393 Details : The greatest transno may be lost when the current operation
3394 finishes with an error (transno==0) and the client's last_rcvd
3395 record is over-written. Save the greatest transno in the
3396 mds_last_transno for this case.
3399 Frequency : always for specific kernels and striping counts
3401 Description: client may get "Matching packet too big" without ACL support
3402 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
3403 when trying to access files in certain configurations. The
3404 clients should in fact be denied when mounting because they do
3405 not understand ACLs.
3408 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
3410 Description: many clients connecting with IO in progress causes connect timeouts
3411 Details : Avoid synchronous journal commits to avoid delays caused by many
3412 clients connecting/disconnecting when bulk IO is in progress.
3413 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
3414 OST_IO_PORTAL to avoid delays behind potentially many pending
3418 Frequency : occasionally with multiple writers to a single file
3420 Description: shared writes to file may result in wrong size reported by stat()
3421 Details : Allow growing of kms when extent lock is cancelled
3424 Frequency : always with random mmap IO to multi-striped file
3426 Description: mmap write might be lost if we are writing to a 'hole' in stripe
3427 Details : Only if the hole is at the end of OST object so that kms is too
3428 small. Fix is to increase kms accordingly in ll_nopage.
3431 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
3433 Description: writes to a missing object would leak memory on the OST
3434 Details : If there is an inconsistency between the MDS and OST filesystems,
3435 such that the MDS references an object that doesn't exist, writes
3436 to that object will leak memory due to incorrect cleanup in the
3437 error handling path, eventually running out of memory on the OST.
3442 Description: Creating too long symlink causes lustre errors
3443 Details : Check symlink and name lengths before sending requests to MDS.
3446 Frequency : only if flock is enabled (not on by default)
3448 Description: posix locks not released on fd closure on 2.6.9+
3449 Details : We failed to add posix locks to list of inode locks on 2.6.9+
3450 kernels, this caused such locks not to be released on fd close and
3451 then assertions on fs unmount about still used locks.
3454 Frequency : MDS failover only, very rarely
3456 Description: clients may get ASSERTION(granted_lock != NULL)
3457 Details : When request was taking a long time, and a client was resending
3458 a getattr by name lock request. The were multiple lock requests
3459 with the same client lock handle and
3460 mds_getattr_name->fixup_handle_for_resent_request found one of the
3461 lock handles but later failed with ASSERTION(granted_lock != NULL).
3466 Description: handle->h_buffer_credits > 0, assertion failure
3467 Details : h_buffer_credits is zero after truncate, causing assertion
3468 failure. This patch extends the transaction or creates a new
3472 Frequency : NFS re-export or patchless client
3473 Bugzilla : 11179, 10796
3474 Description: Crash on NFS re-export node (__d_move)
3475 Details : We do not want to hash the dentry if we don't have a lock.
3476 But if this dentry is later used in d_move, we'd hit uninitialised
3477 list head d_hash, so we just do this to init d_hash field but
3478 leave dentry unhashed.
3481 Frequency : NFS re-export or patchless client
3483 Description: NFS exports has problem with symbolic link
3484 Details : lustre client didn't properly install dentry when re-exported
3485 to NFS or running patchless client.
3488 Frequency : NFS re-export or patchless client
3490 Description: Various nfs/patchless fixes.
3491 Details : fixes reuse disconected alias for lookup process - this fixes
3492 warning "find_exported_dentry: npd != pd",
3493 fix permission error with open files at nfs.
3494 fix apply umask when do revalidate.
3497 Frequency : occasional
3499 Description: Crash on NFS re-export node
3500 Details : calling clear_page() on the wrong pointer triggered oops in
3501 generic_mapping_read().
3504 Frequency : rarely, using O_DIRECT IO
3506 Description: unaligned directio crashes client with LASSERT
3507 Details : check for unaligned buffers before trying any requests.
3510 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
3512 Description: stale data returned from RAID cache
3513 Details : If only a small amount of IO is done to the RAID device before
3514 reading it again it is possible to get stale data from the RAID
3515 cache instead of reading it from disk.
3518 Frequency : always for sles10 kernel
3520 Description: sles10 support
3521 Details : ll_follow_link: compile fixes and using of nd_set_link
3522 under newer kernels.
3525 Frequency : depends on arch, kernel and compiler version, always on sles10
3528 Description: recursive or deep enough symlinks cause stack overflow
3529 Details : getting rid of large stack-allocated variable in
3533 Frequency : depends on hardware
3535 Description: lustre write performance loss in the SLES10 kernel
3536 Details : the performance loss is caused by using of write barriers in the
3537 ext3 code. The SLES10 kernel turns barrier support on by
3538 default. The fix is to undo that change for ldiskfs.
3540 ------------------------------------------------------------------------------
3542 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
3544 * Support for kernels:
3545 2.6.9-42.0.3EL (RHEL 4)
3546 2.6.5-7.276 (SLES 9)
3547 2.4.21-47.0.1.EL (RHEL 3)
3548 2.6.12.6 vanilla (kernel.org)
3552 Frequency : quota enabled and large files being deleted
3554 Description: releasing more than 4GB of quota at once hangs OST
3555 Details : If a user deletes more than 4GB of files on a single OST it
3556 will cause the OST to spin in an infinite loop. Release
3557 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3562 Description: statfs data retrieved from /proc may be stale or zero
3563 Details : When reading per-device statfs data from /proc, in the
3564 {kbytes,files}_{total,free,avail} files, it may appear
3565 as zero or be out of date.
3568 Frequency : systems with MD RAID1 external journal devices
3570 Description: lconf's call to blkid is confused by RAID1 journal devices
3571 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3572 of returning all block devices that match the journal UUID.
3575 Frequency : always, for aggregate stripe size over 4GB
3577 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
3578 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
3579 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
3582 Frequency : always if "lfs find" used on a local file/directory
3584 Description: "lfs find" segfaults if used on a local file/directory
3585 Details : The case where a directory component was not specified wasn't
3586 handled correctly. Handle this properly.
3589 Frequency : always on ppc64
3591 Description: the write to an ext3 filesystem mounted with mballoc got stuck
3592 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3593 perform endianness conversion.
3596 Frequency : rarely (truncate to non-zero file size after write under load)
3597 Bugzilla : 10730, 10687
3598 Description: Files padded with zeros to next 4K multiple
3599 Details : With filesystems mounted using the "extents" option (2.6 kernels)
3600 it is possible that files that are truncated to a non-zero size
3601 immediately after being written are filled with zero bytes beyond
3602 the truncated size. No file data is lost.
3604 Severity : enhancement
3606 Description: Allow recovery/failover for liblustre clients.
3607 Details : liblustre clients were unaware of failover configurations until
3610 Severity : enhancement
3612 Description: user file locks should fail when not mounting with flock option
3613 Details : Set up an error-returning stub in ll_file_operations.lock field
3614 to prevent incorrect behaviour when client is mounted without
3615 flock option. Also, set up properly f_op->flock field for
3619 Frequency : always on ia64
3621 Description: "lfs df" loops on printing out MDS statfs information
3622 Details : The obd_ioctl_data was not initialized and in some systems
3623 this caused a failure during the ioctl that did not return
3624 an error. Initialize the struct and return an error on failure.
3627 Frequency : SLES 9 only
3629 Description: Error of copying files with lustre special EAs as root
3630 Details : Client side always return success for setxattr call for lustre
3631 special xattr (currently only "trusted.lov").
3634 Frequency : rarely on clusters with both ia64+i386 clients
3636 Description: ia64+i686 clients doing shared IO on the same file may LBUG
3637 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
3638 clients are doing concurrent writes to the same file it is
3639 possible that the ia64 clients may LASSERT because the OST
3640 extent locks are not PAGE_SIZE aligned. Ensure that grown
3641 locks are always aligned on the request boundary.
3644 Frequency : specific use, occasional
3646 Description: Overwriting in use executable truncates on-disk binary image
3647 Details : If one node attempts to overwrite an executable in use by
3648 another node, we now correctly return ETXTBSY instead of
3649 truncating the file.
3651 Severity : enhancement
3653 Description: Async OSC create to avoid the blocking unnecessarily.
3654 Details : If a OST has no remain object, system will block on the creating
3655 when need to create a new object on this OST. Now, ways use
3656 pre-created objects when available, instead of blocking on an
3657 empty osc while others are not empty. If we must block, we block
3658 for the shortest possible period of time.
3663 Description: chmod on Lustre root is propagated to other clients
3664 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
3665 invalid by the follow_mount time.
3668 Frequency : liblustre clients only
3670 Description: Race in 'instant cancel' lock handling could lead to such locks
3671 never to be granted in case of SMP MDS
3672 Details : Do not destroy not yet granted but cbpending locks in
3676 Frequency : replay/resend of open
3678 Description: non null lock assetion failure in mds_intent_policy
3679 Details : Trying to replay/resend lockless open requests resulted in
3680 mds_open() returning 0 with no lock. Now it sets a flag if
3681 a lock is going to be returned.
3683 Severity : enhancement
3685 Description: Checksum enhancements
3686 Details : New checksum enhancements allow for resending RPCs that failed
3689 Severity : enhancement
3691 Description: Tunables on number of dirty pages in cacche
3692 Details : Allow to set limit on number of dirty pages cached.
3697 Description: client crash on unmount - lock still has references
3698 Details : In some error handling cases it was possible to leak a lock
3699 reference on a client while accessing a file. This was not
3700 harmful to the client during operation, but would cause the
3701 client to crash when the filesystem is unmounted.
3704 Frequency : specific case, rare
3706 Description: ETXTBSY on mds though file not in use by client
3707 Details : ETXTBSY is no longer incorrectly returned when attempting to
3708 chmod or chown a directory that the user previously tried to
3709 execute or a currently-executing binary.
3712 Frequency : extremely rare except on liblustre-based clients
3714 Description: Lustre space not freed when files are deleted
3715 Details : Clean up open-unlinked files after client eviction. Previously
3716 the unlink was skipped and the files remained as orphans.
3721 Description: OST failure "would be an LBUG" in waiting_locks_callback()
3722 Details : In some cases it was possible to send a blocking callback to
3723 a client doing a glimpse, even though that client didn't get
3724 a lock granted. When the glimpse lock is cancelled on the OST
3725 the freed lock is left on the waiting list and corrupted the list.
3728 Frequency : all core dumps
3730 Description: Broke core dumps to lustre
3731 Details : Negative dentry may be unhashed if parent does not have UPDATE
3732 lock, but some callers, e.g. do_coredump, expect dentry to be
3733 hashed after successful create, hash it in ll_create_it.
3735 ------------------------------------------------------------------------------
3737 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
3739 * Support for kernels:
3740 2.6.9-42.0.2.EL (RHEL 4)
3741 2.6.5-7.276 (SLES 9)
3742 2.4.21-40.EL (RHEL 3)
3743 2.6.12.6 vanilla (kernel.org)
3747 Frequency : always on RHEL 3
3749 Description: Number of open files grows over time
3750 Details : The number of open files grows over time, whether or not
3751 Lustre is started. This was due to a filp leak introduced
3752 by one of our kernel patches.
3754 ------------------------------------------------------------------------------
3756 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
3758 * Support for kernels:
3759 2.6.9-42.EL (RHEL 4)
3760 2.6.5-7.276 (SLES 9)
3761 2.4.21-40.EL (RHEL 3)
3762 2.6.12.6 vanilla (kernel.org)
3767 Bugzilla : 5719, 9635, 9792, 9684
3768 Description: OST (or MDS) trips assertions in (re)connection under heavy load
3769 Details : If a server is under heavy load and cannot reply to new
3770 connection requests before the client resends the (re)connect,
3771 the connection handling code can behave badly if two service
3772 threads are concurrently handing separate (re)connections from
3773 the same client. Add better locking to the connection handling
3774 code, and ensure that only a single connection will be processed
3775 for a given client UUID, even if the lock is dropped.
3777 Severity : enhancement
3779 Description: add TCP zero-copy support to kernel
3780 Details : Add support to the kernel TCP stack to allow zero-copy bulk
3781 sends if the hardware supports scatter-gather and checksumming.
3782 This allows socklnd to do client-write and server-read more
3783 efficiently and reduce CPU utilization from skbuf copying.
3786 Frequency : only if NFS exporting from client
3788 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
3789 Details : If setting attributes on a file created under NFS that had
3790 never been opened it would be possible to oops the client
3791 if the file had no objects.
3794 Frequency : always for liblustre
3796 Description: liblustre client does MDS+OSTs setattr RPC for each write
3797 Details : When doing a write from a liblustre client, the client
3798 incorrectly issued an RPC to the MDS and each OST the file was
3799 striped over in order to update the timestamps. When writing
3800 with small chunks and many clients this could overwhelm the MDS
3801 with RPCs. In all cases it would slow down the write because
3802 these RPCs are unnecessary.
3804 Severity : enhancement
3806 Description: allow number of MDS service threads to be changed at module load
3807 Details : It is now possible to change the number of MDS service threads
3808 running. Adding "options mds mds_num_threads={N}" to the MDS's
3809 /etc/modprobe.conf will set the number of threads for the next
3810 time Lustre is restarted (assuming the "mds" module is also
3811 reloaded at that time). The default number of threads will
3812 stay the same, 32 for most systems.
3817 Description: OST crash if filesystem is unformatted or corrupt
3818 Details : If an OST is started on a device that has never been formatted
3819 or if the filesystem is corrupt and cannot even mount then the
3820 error handling cleanup routines would dereference a NULL pointer.
3825 Description: NULL pointer deref in llap_from_page.
3826 Details : get_cache_page_nowait can return a page with NULL (or otherwise
3827 incorrect) mapping if the page was truncated/reclaimed while it was
3828 searched for. Check for this condition and skip such pages when
3829 doing readahead. Introduce extra check to llap_from_page() to
3830 verify page->mapping->host is non-NULL (so page is not anonymous).
3833 Frequency : Sometimes when using sys_sendfile
3835 Description: "page not covered by a lock" warnings from ll_readpage
3836 Details : sendfile called ll_readpage without right page locks present.
3837 Now we introduced ll_file_sendfile that does necessary locking
3838 around call to generic_file_sendfile() much like we do in
3842 Frequency : with certain MDS communication failures at client mount time
3844 Description: NULL pointer deref after failed client mount
3845 Details : a client connection request may delayed by the network layer
3846 and not be sent until after the PTLRPC layer has timed out the
3847 request. If the client fails the mount immediately it will try
3848 to clean up before the network times out the request. Add a
3849 reference from the request import to the obd device and delay
3850 the cleanup until the network drops the request.
3853 Frequency : occasionally during client (re)connect
3855 Description: assertion failure during client (re)connect
3856 Details : processing a client connection request may be delayed by the
3857 client or server longer than the client connect timeout. This
3858 causes the client to resend the connection request. If the
3859 original connection request is replied in this interval, the
3860 client may trip an assertion failure in ptlrpc_connect_interpret()
3861 which thought it would be the only running connect process.
3864 Frequency : only with obd_echo servers and clients that are rebooted
3866 Description: kernel BUG accessing uninitialized data structure
3867 Details : When running an obd_echo server it did not start the ping_evictor
3868 thread, and when a client was evicted an uninitialized data
3869 structure was accessed. Start the ping_evictor in the RPC
3870 service startup instead of the OBD startup.
3872 Severity : enhancement
3873 Bugzilla : 10193 (patchless)
3874 Description: Remove dependency on various unexported kernel interfaces.
3875 Details : No longer need reparent_to_init, exit_mm, exit_files,
3876 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
3879 Frequency : rare (only users of deprecated and unsupported LDAP config)
3881 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
3882 Details : LDAP apparently contains 'lustreName' attributes instead of
3883 'name'. A simple remapping of the name is sufficient.
3886 Frequency : rare (only with non-default dump_on_timeout debug enabled)
3888 Description: waiting_locks_callback trips kernel BUG if client is evicted
3889 Details : Running with the dump_on_timeout debug flag turned on makes
3890 it possible that the waiting_locks_callback() can try to dump
3891 the Lustre kernel debug logs from an interrupt handler. Defer
3892 this log dumping to the expired_lock_main() thread.
3894 Severity : enhancement
3896 Description: Support NFS exporting on 2.6 kernels.
3897 Details : Implement non-rawops metadata methods for NFS server to use without
3898 changing NFS server code.
3901 Frequency : very rare (synthetic metadata workload only)
3903 Description: two racing renames might cause an MDS thread to deadlock
3904 Details : Running the "racer" program may cause one MDS thread to rename
3905 a file from being the source of a rename to being the target of
3906 a rename at exactly the same time that another thread is doing
3907 so, and the second thread has already enqueued these locks after
3908 doing a lookup of the target and is trying to relock them in
3909 order. Ensure that we don't try to re-lock the same resource.
3912 Frequency : only very large systems with liblustre clients
3914 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
3915 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
3916 all OSTs in parallel. This allows the eviction of stale liblustre
3917 clients to proceed much faster than if they were done in series,
3918 and also offers similar improvements for other set_info RPCs.
3923 Description: excessive CPU usage during initial read phase on client
3924 Details : During the initial read phase on a client, it would agressively
3925 retry readahead on the file, consuming too much CPU and impacting
3926 performance (since 1.4.5.8). Improve the readahead algorithm
3927 to avoid this, and also improve some other common cases (read
3928 of small files in particular, where "small" is files smaller than
3929 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
3934 Description: MDS crash when receiving packet with unknown intent.
3935 Details : Do not LBUG in unknown intent case, just return -EFAULT
3937 Severity : enhancement
3938 Bugzilla : 9293, 9385
3939 Description: MDS RPCs are serialised on client. This is unnecessary for some.
3940 Details : Do not serialize getattr (non-intent version) and statfs.
3943 Frequency : occasional, when OST network is overloaded/intermittent
3945 Description: client evicted by OST after bulk IO timeout
3946 Details : If a client sends a bulk IO request (read or write) the OST
3947 may evict the client if it is unresposive to its data GET/PUT
3948 request. This is incorrect if the network is overloaded (takes
3949 too long to transfer the RPC data) or dropped the OST GET/PUT
3950 request. There is no need to evict the client at all, since
3951 the pinger and/or lock callbacks will handle this, and the
3952 client can restart the bulk request.
3955 Frequency : Always when mmapping file with no objects
3957 Description: client crashes when mmapping file with no objects
3958 Details : Check that we actually have objects in a file before doing any
3959 operations on objects in ll_vm_open, ll_vm_close and
3965 Description: Request leak when working with deleted CWD
3966 Details : Introduce advanced request refcount tracking for requests
3967 referenced from lustre intent.
3969 Severity : Enhancement
3971 Description: Cache open file handles on client.
3972 Details : MDS now will return special lock along with openhandle, if
3973 requested and client is allowed to hold openhandle, even if unused,
3974 until such a lock is revoked. Helps NFS a lot, since NFS is opening
3975 closing files for every read/write openration.
3977 Severity : Enhancement
3979 Description: Cache open negative dentries on client when possible.
3980 Details : Guard negative dentries with UPDATE lock on parent dir, drop
3981 negative dentries on lock revocation.
3986 Description: Remounting a client read-only wasn't possible with a zconf mount
3987 Details : It wasn't possible to remount a client read-only with llmount.
3989 Severity : enhancement
3990 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
3991 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
3992 MPICH is included to add Lustre-specific ADIO interfaces.
3993 This is based closely on the UFS ADIO layer and only differs
3994 in file creation, in order to allow the OST striping to be set.
3995 This is user-contributed code and not supported by CFS.
4000 Description: extended inode attributes (immutable, append-only) work improperly
4001 when 2.4 and 2.6 kernels are used on client/server or vice versa
4002 Details : Introduce kernel-independent values for these flags.
4004 Severity : enhancement
4007 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
4008 Details : Many of the /proc/ tunables can only be tuned at a megabyte
4009 granularity. Now, Fractional MB granularity is be supported,
4010 this is very useful for low memory system.
4012 Severity : enhancement
4014 Description: Getattr by fid
4015 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
4016 locks, avoids extra getattr rpc requests to MDS, allows '/' to
4017 have locks and avoids getattr rpc requests for it on every stat.
4020 Frequency : Always, for filesystems larger than 2TB
4022 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4023 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
4024 filesystems larger than 2TB to be mounted with mballoc enabled.
4027 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
4029 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4030 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
4031 filesystems larger than 2TB is mounted then the kernel will
4032 silently corrupt the start of the filesystem. CONFIG_LBD is
4033 enabled for all CFS-supported kernels, but the possibility of
4034 this happening with a modified kernel config exists.
4036 Severity : enhancement
4038 Description: add client O_DIRECT support for 2.6 kernels
4039 Details : It is now possible to do O_DIRECT reads and writes to files
4040 in the Lustre client mountpoint on 2.6 kernel clients.
4042 Severity : enhancement
4044 Description: parallel glimpse, setattr, statfs, punch, destroy requests
4045 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
4046 parallel, not waiting for response from every OST before sending
4047 a rpc to the next OST.
4052 Description: setattr vs write race when updating file timestamps
4053 Details : Client processes that update a file timestamp into the past
4054 right after writing to the file (e.g. tar) it is possible that
4055 the updated file modification time can be reset to the current
4056 time due to a race between processing the setattr and write RPC.
4058 Severity : enhancement
4060 Description: Bring 'lfs find' closer in line with regular Linux find.
4061 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
4062 -print0 options and obtains all the needed info through the lustre
4065 Severity : enhancement
4067 Description: support up to 1024 configured devices on one node
4068 Details : change obd_dev array from statically allocated to dynamically
4069 allocated structs as they are first used to reduce memory usage
4074 Description: Flush dirty partially truncated pages during truncate
4075 Details : Immediatelly flush partially truncated pages in filter_setattr,
4076 this way we completely avoid having any pages in page cache on OST
4077 and can retire ugly workarounds during writes to flush such pages.
4082 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
4083 Details : protect inode from truncation within vfs_unlink() context
4084 just take a reference before calling vfs_unlink() and release it
4085 when parent's i_sem is free.
4090 Description: last_id value checked outside lock on OST caused LASSERT failure
4091 Details : If there were multiple MDS->OST object precreate requests in
4092 flight, it was possible that the OST's last object id was checked
4093 outside a lock and incorrectly tripped an assertion. Move checks
4094 inside locks, and discard old precreate requests.
4097 Frequency : always, if extents are used on OSTs
4099 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
4100 Details : OSTs using the extents format would not zero the high 16 bits of
4101 the index physical block number. This is not a problem for any
4102 OST filesystems smaller than 16TB, and no kernels support ext3
4103 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
4104 new/modified files) and can be fixed for existing filesystems
4105 with e2fsprogs-1.39-cfs1.
4110 Description: import connection selection may be incorrect if timer wraps
4111 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
4112 import connections to be ignored if the 32-bit jiffies counter
4113 wraps. Use a 64-bit jiffies counter.
4116 Frequency : very large clusters immediately after boot
4118 Description: LNET request buffers exhausted under heavy short-term load
4119 Details : If a large number of client requests are generated on a service
4120 that has previously never seen so many requests it is possible
4121 that the request buffer growth cannot keep up with the spike in
4122 demand. Instead of dropping incoming requests, they are held in
4123 the LND until the RPC service can accept more requests.
4126 Frequency : Sometimes during replay
4128 Description: Assertion failure in ll_local_open after replay.
4129 Details : If replay happened on an open request reply before we were able
4130 to set replay handler, reply will become not swabbed tripping the
4131 assertion in ll_local_open. Now we set the handler right after
4132 recognising of open request
4135 Frequency : very rare
4137 Description: kernel reports "badness in vsnprintf"
4138 Details : Reading from the "recovery_status" /proc file in small chunks
4139 may cause a negative length in lprocfs_obd_rd_recovery_status()
4140 call to vsnprintf() (which is otherwise harmless). Exit early
4141 if there is no more space in the output buffer.
4143 Severity : enhancement
4145 Description: clear OBD RPC statistics by writing to them
4146 Details : It is now possible to clear the OBD RPC statistics by writing
4147 to the "stats" file.
4152 Description: Client mtime is not the same on different clients after utimes
4153 Details : In some cases, the client was using the utimes() syscall on
4154 a file cached on another node. The clients now validate the
4155 ctime from the MDS + OSTs to determine which one is right.
4160 Description: Inability to activate failout mode
4161 Details : lconf script incorrectly assumed that in pythong string's numeric
4162 value is used in comparisons.
4165 Frequency : always with multiple stripes per file
4167 Description: Inefficient object allocation for mutli-stripe files
4168 Details : When selecting which OSTs to stripe files over, for files with
4169 a stripe count that divides evenly into the number of OSTs,
4170 the MDS is always picking the same starting OST for each file.
4171 Return the OST selection heuristic to the original design.
4176 Description: mount failures may take full timeout to return an error
4177 Details : Under some heavy load conditions it is possible that a
4178 failed mount can wait for the full obd_timeout interval,
4179 possibly several minutes, before reporting an error.
4180 Instead return an error as soon as the status is known.
4182 Frequency : quota enabled and large files being deleted
4184 Description: releasing more than 4GB of quota at once hangs OST
4185 Details : If a user deletes more than 4GB of files on a single OST it
4186 will cause the OST to spin in an infinite loop. Release
4187 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4192 Description: statfs data retrieved from /proc may be stale or zero
4193 Details : When reading per-device statfs data from /proc, in the
4194 {kbytes,files}_{total,free,avail} files, it may appear
4195 as zero or be out of date.
4198 Frequency : systems with MD RAID1 external journal devices
4200 Description: lconf's call to blkid is confused by RAID1 journal devices
4201 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4202 of returning all block devices that match the journal UUID.
4205 Frequency : always, for aggregate stripe size over 4GB
4207 Description: assertion fails when trying to use 4GB stripe size
4208 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
4209 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
4212 Frequency : always on ppc64
4214 Description: the first write on an ext3 filesystem with mballoc got stuck
4215 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4216 perform endianness conversion.
4218 ------------------------------------------------------------------------------
4220 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4222 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
4223 INTEROPERATE with older versions automatically. Please read the
4224 user documentation before upgrading any part of a live system.
4225 * WARNING: Lustre networking configuration changes are required with
4226 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
4229 * Support for kernels:
4230 2.6.9-22.0.2.EL (RHEL 4)
4231 2.6.5-7.244 (SLES 9)
4232 2.6.12.6 vanilla (kernel.org)
4235 Severity : enhancement
4236 Bugzilla : 7981/8208
4237 Description: Introduced Lustre Networking (LNET)
4238 Details : LNET is new networking infrastructure for Lustre, it includes
4239 a reorganized network configuration mode (see the user
4240 documentation for full details) as well as support for routing
4241 between different network fabrics. Lustre Networking Devices
4242 (LNDs) for the supported network fabrics have also been
4243 created for this new infrastructure.
4245 Severity : enhancement
4246 Description: Introduced Access control lists
4247 Details : clients can set ACLs on files and directories in order to have
4248 more fine-grained permissions than the standard Unix UGO+RWX.
4249 The MDS must be started with the "-o acl" mount option.
4251 Severity : enhancement
4252 Description: Introduced filesystem quotas
4253 Details : Administrators may now establish per-user quotas on the
4256 Severity : enhancement
4258 Description: Configuration change for the XT3
4259 The PTLLND is now used to run Lustre over Portals on the XT3
4260 The configure option(s) --with-cray-portals are no longer used.
4261 Rather --with-portals=<path-to-portals-includes> is used to
4262 enable building on the XT3. In addition to enable XT3 specific
4263 features the option --enable-cray-xt3 must be used.
4268 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
4269 Details : A race between allocating a new llcd and re-getting the llcd_lock
4270 allowed another thread to grab newly-allocated llcd.
4272 Severity : enhancement
4274 Description: 2.6 OST async journal commit and locking fix to improve performance
4275 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
4276 2.6 kernels are now async as they already were in 2.4 kernels so
4277 that they can commit concurrently with the network bulk transfer.
4278 For block-allocated files the filter allocation semaphore is held
4279 to avoid filesystem fragmentation during allocation. BKL lock
4280 removed for 2.6 xattr operations where it is no longer needed.
4285 Description: lconf incorrectly determined whether two IP networks could talk
4286 Details : In some more complicated routing and multiple-network
4287 configurations, lconf will avoid trying to make a network
4288 connection to a disjoint part of the IP space. It was doing the
4289 math incorrectly for one set of cases.
4294 Description: Fix for potential infinite loop processing records in an llog.
4295 Details : If an llog record is corrupted/zeroed, it is possible to loop
4296 forever in llog_process(). Validate the llog record length
4297 and skip the remainder of the block on error.
4300 Frequency : occasional (liblustre only)
4302 Description: liblustre could not open files whose last component is a symlink
4303 Details : sysio_path_walk() would incorrectly pass the open intent to
4304 intermediate path components.
4307 Frequency : rare (liblustre only with non-standard tuning)
4308 Bugzilla : 7201 (7350)
4309 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4310 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4311 after it has been released, only do this for referenced locks
4313 Severity : enhancement
4315 Description: specifying an (invalid) directory default stripe_size of -1
4316 would reset the directory default striping
4317 Details : stripe_size -1 was used internally to signal directory stripe
4318 removal, now use "all default" to signal dir stripe removal
4319 as a directory striping of "all default" is not useful
4322 Frequency : common for large clusters running liblustre clients
4324 Description: doing an ls when liblustre clients are running is slow
4325 Details : sending a glimpse AST to a liblustre client waits for every AST
4326 to time out, as liblustre clients will not respond. Since they
4327 cannot cache data we refresh the OST lock LVB from disk instead.
4329 Severity : enhancement
4331 Description: doing an ls at the same time as file IO can be slow
4332 Details : enqueue and other "small" requests can be blocked behind many
4333 large IO requests. Create a new OST IO portal for non-IO
4334 requests so they can be processed faster.
4337 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4339 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4340 created file with mode 000
4341 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4342 allow an ftruncate() as a non-root user to a file with mode 000.
4343 Lustre now respects this flag to disable mode checks when
4344 truncating a file owned by the user
4347 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4350 Description: Revoking locks from clients that went dead or catatonic might take
4352 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4353 cancellation of such locks instant on servers without waiting for
4354 any reply from clients. Clients drops these locks when cancel
4355 notification from server is received without replying.
4358 Frequency : liblustre-only, when liblustre client dies or becomes busy
4360 Description: Doing ls on Linux clients can take a long time with active
4362 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4363 granting such locks to it in the first place if possible. Locks
4364 are taken by proxy on the OST during the read or write and
4365 dropped immediately afterward. Add connect flags handling, do
4366 not grant locks to liblustre clients for glimpse ASTs.
4368 Severity : enhancement
4370 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4371 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4372 sequential reads and behaves badly with applications doing random
4373 reads. Improve it by reading ahead at least read region, and
4374 avoiding excessive large RPC for small reads.
4376 Severity : enhancement
4378 Description: Creating more than 1000 files for a single job may cause a load
4379 imbalance on the OSTs if there are also a large number of OSTs.
4380 Details : qos_prep_create() uses an OST index reseed value that is an
4381 even multiple of the number of available OSTs so that if the
4382 reseed happens in the middle of the object allocation it will
4383 still utilize the OSTs as uniformly as possible.
4388 Description: OST or MDS may oops in ping_evictor_main()
4389 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
4390 but doesn't restart at beginning of obd_exports_timed list
4393 Severity : enhancement
4395 Description: improve by-nid export eviction on the MDS and OST
4396 Details : allow multiple exports with the same NID to be evicted at one
4397 time without re-searching the exports list.
4400 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
4402 Description: MDS may oops in groups_free()
4403 Details : in rare race conditions a newly allocated group_info struct is
4404 freed again, and this can be NULL. The 2.4 compatibility code
4405 for groups_free() checked for a NULL pointer, but 2.6 did not.
4408 Frequency : common for liblustre clients doing little filesystem IO
4409 Bugzilla : 9352, 7313
4410 Description: server may evict liblustre clients accessing contended locks
4411 Details : if a client is granted a lock or receives a completion AST
4412 with a blocking AST already set it would not reply to the AST
4413 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
4416 Frequency : lfs setstripe, only systems with more than 160 OSTs
4418 Description: unable to set striping with a starting offset beyond OST 160
4419 Details : llapi_create_file() incorrectly limited the starting stripe
4420 index to the maximum single-file stripe count.
4423 Frequency : LDAP users only
4425 Description: lconf did not handle in-kernel recovery with LDAP properly
4426 Details : lconf/LustreDB get_refs() is searching the wrong namespace
4428 Severity : enhancement
4430 Description: bind OST threads to NUMA nodes to improve performance
4431 Details : all OST threads are uniformly bound to CPUs on a single NUMA
4432 node and do their allocations there to localize memory access
4434 Severity : enhancement
4436 Description: llmount can determine client NID directly from Myrinet (GM)
4437 Details : the client NID code from gmnalnid was moved directly into
4438 llmount, removing the need to use this or specifying the
4439 client NID explicitly when mounting GM clients with zeroconf
4442 Frequency : if client is started with down MDS
4444 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
4445 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
4446 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
4447 ptlrpc_import_delayed_req() because we didn't check if the
4448 request was interrupted, and we also didn't break out of the
4449 event loop if there was no timeout
4454 Description: data loss during non-page-aligned writes to a single file from
4455 both multiple nodes and multiple threads on one node at same time
4456 Details : updates to KMS and lsm weren't protected by common lock. Resulting
4457 inconsistency led to false short-reads, that were cached and later
4458 used by ->prepare_write() to fill in partially written page,
4459 leading to data loss.
4462 Frequency : always, if lconf --abort_recovery used
4464 Description: lconf --abort_recovery fails with 'Operation not supported'
4465 Details : lconf was attempting to abort recovery on the MDT device and not
4468 Severity : enhancement
4470 Description: remove cleanup logs
4471 Details : replace lconf-generated cleanup logs with lustre internal
4472 cleanup routines. Eliminates the need for client-cleanup and
4475 Severity : enhancement
4477 Description: add support for EAs (user and system) on lustre filesystems
4478 Details : it is now possible to store extended attributes in the Lustre
4479 client filesystem, and with the user_xattr mount option it
4480 is possible to allow users to store EAs on their files also
4482 Severity : enhancement
4484 Description: Add possibility (config option) to show minimal available OST free
4486 Details : When compiled with --enable-mindf configure option, statfs(2)
4487 (and so, df) will return least minimal free space available from
4488 all OSTs as amount of free space on FS, instead of summary of
4489 free spaces of all OSTs.
4491 Severity : enhancement
4493 Description: do not expand extent locks acquired on OST-side
4494 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
4495 by server: they are not cached anyway.
4498 Frequency : when mmap is used/binaries executed from Lustre
4500 Description: Unmmap pages before throwing them away from read cache.
4501 Details : llap_shrink cache now attempts to unmap pages before discarding
4502 them (if unmapping failed - do not discard). SLES9 kernel has
4503 extra checks that trigger if this unmapping is not done first.
4508 Description: lconf didn't resolve symlinks before checking to see whether a
4509 given mountpoint was already in use
4512 Frequency : when migrating failover services
4513 Bugzilla : 6395, 9514
4514 Description: When migrating a subset of services from a node (e.g. failback
4515 from a failover service node) the remaining services would
4516 time out and evict clients.
4517 Details : lconf --force (implied by --failover) sets the global obd_timeout
4518 to 5 seconds in order to quickly disconnect, but this caused
4519 other RPCs to time out too quickly. Do not change the global
4520 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
4522 Severity : enhancement
4523 Frequency : if MDS is started with down OST
4524 Bugzilla : 9439,5706
4525 Description: Allow startup/shutdown of an MDS without depending on the
4526 availability of the OSTs.
4527 Details : Asynchronously call mds_lov_synchronize during MDS startup.
4528 Add appropriate locking and lov-osc refcounts for safe
4529 cleaning. Add osc abort_inflight calls in case the
4530 synchronize never started.
4533 Frequency : occasional (Cray XT3 only)
4535 Description: root not authorized to access files in CRAY_PORTALS environment
4536 Details : The client process capabilities were not honoured on the MDS in
4537 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
4538 been accessed by an authorized user then root was able to access
4539 the file on the local client also. The root user capabilities
4540 are now allowed on the MDS, as this environment has secure UID.
4543 Frequency : occasional
4545 Description: ldiskfs "too long searching" message happens too often
4546 Details : A debugging message (otherwise harmless) prints too often on
4547 the OST console. This has been reduced to only happen when
4548 there are fragmentation problems on the filesystem.
4553 Description: Division by zero in statfs when all OSCs are inactive
4554 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
4555 lov_statfs divides by value returned by lov_get_stripecnt().
4559 Bugzilla : 9489, 3273
4560 Description: First write from each client to each OST was only 4kB in size,
4561 to initialize client writeback cache, which caused sub-optimal
4562 RPCs and poor layout on disk for the first writen file.
4563 Details : Clients now request an initial cache grant at (re)connect time
4564 and so that they can start streaming writes to the cache right
4565 away and always do full-sized RPCs if there is enough data.
4566 If the OST is rebooted the client also re-establishes its grant
4567 so that client cached writes will be honoured under the grant.
4572 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
4573 Details : Now I/O RPCs go to different portal number and (presumably) fast
4574 lock requests (and glimses) and other RPCs get their own service
4575 threads pool that should be able to service those RPCs
4578 Severity : enhancement
4580 Description: Ability to exchange lustre version between client and servers and
4581 issue warnings at client side if client is too old. Also for
4582 liblustre clients there is ability to refuse connection of too old
4584 Details : New 'version' field is added to connect data structure that is
4585 filled with version info. That info is later checked by server and
4589 Frequency : rare, liblustre only.
4590 Bugzilla : 9296, 9581
4591 Description: Two simultaneous writes from liblustre at offset within same page
4592 might proceed at the same time overwriting eachother with stale
4594 Details : I/O lock withing llu_file_prwv was released too early, before data
4595 actually was hitting the wire. Extended lock-holding time until
4596 server acknowledges receiving data.
4599 Frequency : extremely rare. Never observed in practice.
4601 Description: avoid generating lustre_handle cookie of 0.
4602 Details : class_handle_hash() generates handle cookies by incrementing
4603 global counter, and can hit 0 occasionaly (this is unlikely, but
4604 not impossible, because initial value of cookie counter is
4605 selected randonly). Value of 0 is used as a sentinel meaning
4606 "unassigned handle" --- avoid it. Also coalesce two critical
4607 sections in this function into one.
4609 Severity : enhancement
4611 Description: allow liblustre clients to delegate truncate locking to OST
4612 Details : To avoid overhead of locking, liblustre client instructs OST to
4613 take extent lock in ost_punch() on client's behalf. New connection
4614 flag is added to handle backward compatibility.
4616 Severity : enhancement
4617 Bugzilla : 4928, 7341, 9758
4618 Description: allow number of OST service threads to be specified
4619 Details : a module parameter allows the number of OST service threads
4620 to be specified via "options ost ost_num_threads={N}" in the
4621 OSS's /etc/modules.conf or /etc/modprobe.conf.
4625 Bugzilla : 6146, 9635, 9895
4626 Description: servers crash with bad pointer in target_handle_connect()
4627 Details : In rare cases when a client is reconnecting it was possible that
4628 the connection request was the last reference for that export.
4629 We would temporarily drop the export reference and get a new
4630 one, but this may have been the last reference and the export
4631 was just destroyed. Get new reference before dropping old one.
4633 Severity : enhancement
4634 Frequency : if client is started with failover MDS
4636 Description: Allow multiple MDS hostnames in the mount command
4637 Details : Try to read the configuration from all specified MDS
4638 hostnames during a client mount in case the "primary"
4641 Severity : enhancement
4643 Description: Stop sending data to evicted clients as soon as possible.
4644 Details : Check if the client we are about to send or are sending data to
4645 was evicted already. (Check is done every second of waiting,
4646 for which l_wait_event interface was extended to allow checking
4647 of exit condition at specified intervals).
4650 Frequency : rare, normally only when NFS exporting is done from client
4652 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
4654 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
4657 Frequency : always, if srand() is called before liblustre initialization
4659 Description: Liblustre uses system PRNG disturbing its usage by user application
4660 Details : Introduce internal to lustre fast and high-quality PRNG for
4661 lustre usage and make liblustre and some other places in generic
4662 lustre code to use it.
4664 Severity : enhancement
4665 Bugzilla : 9477, 9557, 9870
4666 Description: Verify that the MDS configuration logs are updated when xml is
4667 Details : Check if the .xml configuration logs are newer than the config
4668 logs stored on the MDS and report an error if this is the case.
4669 Request --write-conf, or allow starting with --old_conf.
4671 Severity : enhancement
4673 Description: Handle symlinks in the path when checking if Lustre is mounted.
4674 Details : Resolve intermediate symlinks when checking if a client has
4675 mounted a filesystem to avoid duplicate client mounts.
4680 Description: lconf can hit an error exception but still return success.
4681 Details : The lconf command catches the Command error exception at the top
4682 level script context and will exit with the associated exit
4683 status, but doesn't ensure that this exit status is non-zero.
4688 Description: failure of ptlrpc thread startup can cause oops
4689 Details : Starting a ptlrpc service thread can fail if there are a large
4690 number of threads or the server memory is very fragmented.
4691 Handle this without oopsing.
4694 Frequency : always, only if liblustre and non-default acceptor port was used
4696 Description: liblustre cannot connect to servers with non-default acceptor port
4697 Details : tcpnal_set_default_params() was not called and was therefore
4698 ignoring the environment varaible TCPNAL_PORT, as well as other
4699 TCPNAL_ environment variables
4704 Description: two objects could be created on the same OST for a single file
4705 Details : If an OST is down, in some cases it was possible to create two
4706 objects on a single OST for a single file. No problems other
4707 than potential performance impact and spurious error messages.
4711 Bugzilla : 5681, 9562
4712 Description: Client may oops in ll_unhash_aliases
4713 Details : Client dcache may become inconsistent in race condition.
4714 In some cases "getcwd" can fail if the current directory is
4720 Description: Inode refcounting problems in NFS export code
4721 Details : link_raw functions used to call d_instantiate without obtaining
4722 extra inode reference first.
4726 Bugzilla : 9942, 9903
4727 Description: Referencing freed requests leading to crash, memleaks with NFS.
4728 Details : We used to require that call to ll_revalidate_it was always
4729 followed by ll_lookup_it. Also with revalidate_special() it is
4730 possible to call ll_revalidate_it() twice for the same dentry
4731 even if first occurence returned success. This fix changes semantic
4732 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
4733 reference on a request referred from the intent.
4734 ll_intent_release() then releases such a request.
4737 Frequency : rare, normally benchmark loads only
4739 Description: unlinked inodes were kept in memory on the client
4740 Details : If a client is repeatedly creating and unlinking files it
4741 can accumulate a lot of stale inodes in the inode slab cache.
4742 If there is no other client load running this can cause the
4743 client node to run out of memory. Instead flush old inodes
4744 from client cache that have the same inode number as a new inode.
4747 Frequency : SLES9 2.6.5 kernel and long filenames only
4748 Bugzilla : 9969, 10379
4749 Description: utime reports stale NFS file handle
4750 Details : SLES9 uses out-of-dentry names in some cases, which confused
4751 the lustre dentry revalidation. Change it to always use the
4755 Frequency : rare, unless heavy write-truncate concurrency is continuous
4756 Bugzilla : 4180, 6984, 7171, 9963, 9331
4757 Description: OST becomes very slow and/or deadlocked during object unlink
4758 Details : filter_destroy() was holding onto the parent directory lock
4759 while truncating+unlinking objects. For very large objects this
4760 may block other threads for a long time and slow overall OST
4761 responsiveness. It may also be possible to get a lock ordering
4762 deadlock in this case, or run out of journal credits because of
4763 the combined truncate+unlink. Solution is to do object truncate
4764 first in one transaction without parent lock, and then do the
4765 final unlink in a new transaction with the parent lock. This
4766 reduces the lock hold time dramatically.
4769 Frequency : rare, 2.4 kernels only
4771 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
4772 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
4773 before unmounting their filesystem in order to avoid deadlock.
4774 The kernel_locked() function in 2.4 kernels only checks whether
4775 the kernel lock is held, not whether it is this process that is
4776 holding it as 2.6 kernels do.
4781 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
4782 Details : The client ptlrpc code may be trying to reconnect to a down
4783 server before a previous connection attempt has timed out.
4784 Increase the reconnect interval to be longer than the connection
4785 timeout interval to avoid sending duplicate connections to
4789 Frequency : echo_client brw_test command
4791 Description: fix echo_client to work with OST preallocated code
4792 Details : OST preallocation code (5137) didn't take echo_client IO path
4793 into account: echo_client calls filter methods outside of any
4794 OST thread and, hence, there is no per-thread preallocated
4795 pages and buffers to use. Solution: hijack pga pages for IO. As
4796 a byproduct, this avoids unnecessary data copying.
4800 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
4801 Description: Client can oops in mdc_commit_close() after open replay
4802 Details : It was possible for the MDS to return an open request with no
4803 transaction number in mds_finish_transno() if the client was
4804 evicted, but without actually returning an error. Clients
4805 would later try to replay that open and may trip an assertion
4806 Simplify the client close codepath, and always return an error
4807 from the MDS in case the open is not successful.
4810 Frequency : rare, 2.6 OSTs only
4812 Description: OST may deadlock under high load on fragmented files
4813 Details : If there was a heavy load and highly-fragmented OST filesystems
4814 it was possible to have all the OST threads deadlock waiting on
4815 allocation of biovecs, because the biovecs were not released
4816 until the entire RPC IO was completed. Instead, release biovecs
4817 as soon as they are complete to ensure forward IO progress.
4819 Severity : enhancement
4821 Description: Support for specifying external journal device at mount
4822 Details : If an OST or MDS device is formatted with an external journal
4823 device, this device major/minor is stored in the ext3 superblock
4824 and may not be valid for failover. Allow detecting and
4825 specifying the external journal at mount time.
4830 Description: Mounting an MDS with pending unlinked files may cause oops
4831 Details : target_finish_recovery() calls mds_postrecov() which returned
4832 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
4833 considers this an error and immediately begins cleaning up the
4834 lov, just after starting the mds_lov process
4836 Severity : enhancement
4838 Description: Implement 'lfs df' to report actual free space on per-OST basis
4839 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
4840 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
4841 usage of objects; '-h' to report in human readable format.
4843 ------------------------------------------------------------------------------
4845 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
4852 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
4853 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
4854 at startup during orphan unlinking. Instead, skip buddy update
4855 before setup, as it will be regenerated after recovery is complete.
4860 Description: If an OST is inactive, its locks might reference stale inodes.
4861 Details : lov_change_cbdata() must iterate over all namespaces, even if
4862 they are inactive to clear inode references from the lock.
4864 Severity : enhancement
4865 Frequency : occasional, if non-standard max_dirty_mb used
4867 Description: Client will block write RPCs if not enough grant
4868 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
4869 then the client will block writes while waiting for another RPC
4870 to complete instead of consuming its dirty limit. With change
4871 we get improved performance when max_dirty_mb is small.
4873 Severity : enhancement
4874 Bugzilla : 3389, 6253
4875 Description: Add support for supplementary groups on the MDS.
4876 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
4877 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
4878 lookups for user supplementary groups into a cache.
4882 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
4883 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
4884 for liblustre O_EXCL is enforced by MDS.
4888 Description: readdir never returns NULL in liblustre.
4889 Details : Corrected llu_iop_getdirentries logic, to return offset of next
4890 dentry in struct dirent.
4894 Frequency : liblustre only, depends on application IO pattern
4895 Description: liblustre clients evicted if not contacting servers
4896 Details : Don't put liblustre clients into the ping_evictor list, so
4897 they will not be evicted by the pinger ever.
4899 Severity : enhancement
4901 Description: Add ability to evict clients by NID from MDS.
4902 Details : By echoing "nid:$NID" string into
4903 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
4904 $NID would be instantly evicted from this MDS and from all active
4905 OSTs connected to it.
4909 Description: Do not query file size twice, somewhat slowing stat(2) calls.
4910 Details : lookup_it_finish() used to query file size from OSTs that was not
4915 Description: service threads change working directory to that of init
4916 Details : Starting lustre service threads may pin the working directory
4917 of the parent thread, making that filesystem busy. Threads
4918 now change to the working directory of init to avoid this.
4922 Frequency : during shutdown only
4923 Description: shutdown with a failed MDS or OST can cause unmount to hang
4924 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
4929 Frequency : 2.6 only
4930 Description: chmod/chown may include an extra supplementary group
4931 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
4932 supplementary group and if none is specified this is used.
4935 Bugzilla : 5479 (6816)
4937 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
4938 Details : If lookup is in progress on a file that is unlinked we might try
4939 to revalidate the inode and fail in revalidate after lookup is
4940 complete and ll_file_open() enqueues the open again but
4941 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
4944 Frequency : always, if lconf --abort_recovery used
4946 Description: lconf --abort_recovery fails with 'Operation not supported'
4947 Details : lconf was attempting to abort recovery on the MDT device and not
4950 ------------------------------------------------------------------------------
4952 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
4957 Frequency : rare (only unsupported configurations with a node running as an
4959 Bugzilla : 6514, 5137
4960 Description: Mounting a Lustre file system on a node running as an OST could
4962 Details : OSTs now preallocates memory needed to write out data at
4963 startup, instead of when needed, to avoid having to
4964 allocate memory in possibly low memory situations.
4965 Specifically, if the file system is mounted on on OST,
4966 memory pressure could force it to try to write out data,
4967 which it needed to allocate memory to do. Due to the low
4968 memory, it would be unable to do so and the node would
4969 become unresponsive.
4971 Severity : enhancement
4973 Description: Addition of lconf --service command line option
4974 Details : lconf now accepts a '--service <arg>' option, which is
4975 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
4977 Severity : enhancement
4979 Description: Failover mode is now the default for OSTs.
4980 Details : By default, OSTs will now run in failover mode. To return to
4981 the old behaviour, add '--failout' to the lmc line for OSTs.
4983 Severity : enhancement
4985 Description: Health checks are now provided for MDS and OSTs
4986 Details : Additional detailed health check information on MSD and OSTs
4987 is now provided through the procfs health_check value.
4990 Frequency : occasional, depends on IO load
4992 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
4993 after numerous create/delete cycles
4994 Details : The ext3 inode allocation policy would not allocate new inodes
4995 very well on the OSTs because there are no new directories
4996 being created. Instead we look for groups with free space if
4997 the parent directories are nearly full.
5002 Description: Network or server problems during mount may cause partially
5003 mounted clients instead of returning an error.
5004 Details : The config llog parsing code may overwrite the error return
5005 code during mount error handling, returning success instead
5011 Description: MDS can fail to allocate large reply buffers
5012 Details : After long uptimes the MDS can fail to allocate large reply
5013 buffers (e.g. zconf client mount config records) due to memory
5014 fragmentation or consumption by the buffer cache. Preallocate
5015 some large reply buffers so that these replies can be sent even
5016 under memory pressure.
5020 Frequency : rare (liblustre)
5021 Description: fsx running with liblustre complained that using truncate() to
5022 extend the file doesn't work. This patch corrects that issue.
5023 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
5024 ATTR_SIZE and lsm use in llu_setattr_raw.
5028 Frequency : rare, only 2.6 kernels
5029 Description: Unusual file access patterns on the MDS may result in inode
5030 data being lost in very rare circumstances.
5031 Details : Bad interaction between the ea-in-inode patch and the "no-read"
5032 code in the 2.6 kernel caused the inode and/or EA data not to
5033 be read from disk, causing single-file corruption.
5037 Frequency : rare, only 2.6 filesystems using extents
5038 Description: Heavy concurrent write and delete load may cause data corruption.
5039 Details : It was possible under high-load situations to have an extent
5040 metadata block in the block device cache from a just-unlinked
5041 file overwrite a newly-allocated data block. We now unmap any
5042 metadata buffers that alias just-allocated data blocks.
5046 Frequency : filesystems with default stripe_count larger than 77
5047 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
5048 Details : lconf specifies an inode size of 4096 bytes when the default
5049 stripe_count is larger than 77. This conflicts with the default
5050 inode density of 1 per 4096 bytes. Allocate smaller inodes in
5051 this case to avoid pinning too much memory for large EAs.
5053 ------------------------------------------------------------------------------
5055 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
5060 Frequency : rare (extremely heavy IO load with hundreds of clients)
5062 Description: Client is evicted, gets IO error writing to file
5063 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
5064 caused clients to be evicted for AST timeouts. The fixes in
5065 bug 5192 mean we no longer need to have such short AST timeouts
5066 so ldlm_timeout has been increased.
5069 Frequency : occasional during --force or --failover shutdown under load
5070 Bugzilla : 5949, 4834
5071 Description: Server oops/LBUG if stopped with --force or --failover under load
5072 Details : a collection of import/export refcount and cleanup ordering
5073 issues fixed for safer force cleanup
5076 Frequency : only filesystems larger than 120 OSTs
5077 Bugzilla : 5990, 6223
5078 Description: lfs getstripe would oops on a very large filesystem
5079 Details : lov_getconfig used kfree on vmalloc'd memory
5082 Frequency : only filesystems exporting via NFS to Solaris 10 clients
5083 Bugzilla : 6242, 6243
5084 Description: reading from files that had been truncated to a non-zero size
5085 but never opened returned no data
5086 Details : ll_file_read() reads zeros from no-object files to EOF
5091 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
5092 Details : There's little harm in aborting MDS/OSS recovery and letting it
5093 try again, so I removed the LASSERT and return an error instead.
5095 Severity : enhancement
5097 Description: New debugging infrastructure for tracking down data corruption
5098 Details : The I/O checksum code was replaced to: (a) control it at runtime,
5099 (b) cover more of the client-side code path, and (c) try to narrow
5100 down where problems occurred
5104 Bugzilla : 3819, 4364, 4397, 6313
5105 Description: Racing close and eviction MDS could cause assertion in mds_close
5106 Details : It was possible to get multiple mfd references during close and
5107 client eviction, leading to one thread referencing a freed mfd.
5109 Severity: : enhancement
5110 Bugzilla : 3262, 6359
5111 Description: Attempts to reconnect to servers are now more aggressive.
5112 Details : This builds on the enhanced upcall-less recovery that was added
5113 in 1.4.2. When trying to reconnect to servers, clients will
5114 now try each server in the failover group every 10 seconds. By
5115 default, clients would previously try one server every 25 seconds.
5120 Description: After recovery, certain operations trigger a failed
5121 assertion on a client.
5122 Details : Failing over an mds, using lconf -d --failover, while a
5123 client was doing a readdir() call would cause the client to
5124 LBUG after recovery completed and the readdir() was resent.
5126 Severity : enhancement
5128 Description: Default groups are now added by lconf
5129 Details : You can now run lconf --group <servicename> without having to
5130 manually add groups with lmc.
5133 Frequency : occasional
5135 Description: Nodes with an elan id of 0 trigger a failed assertion
5138 Frequency : always when accessing e.g. tty/console device nodes
5140 Description: tty and some other devices nodes cannot be used on lustre
5141 Details : file's private_data field is used by device data and lustre
5142 values in there got lost. New field was added to struct file to
5143 store fs-specific private data.
5146 Frequency : when exporting Lustre via NFS
5148 Description: NFSD failed occasionally when looking up a path component
5149 Details : NFSD is looking up ".." which was broken in ext3 directories
5150 that had grown large enough to become hashed.
5153 Frequency : Clusters with multiple interfaces not on the same subnet
5155 Description: Nodes will repeatedly try to reconnect to an interface which it
5156 cannot reach and report an error to the log.
5157 Details : Extra peer list entries will be created by lconf with some peers
5158 unreachable. lconf now validates the peer before adding it.
5161 Frequency : Only if a default stripe is set on the filesystem root.
5163 Description: Setting a default stripe on the filesystem root prevented the
5164 filesystem from being remounted.
5165 Details : The client was sending extra request flags in the root getattr
5166 request and did not allocate a reply buffer for the dir EA.
5169 Frequency : occasional, higher if lots of files are accessed by one client
5170 Bugzilla : 6159, 6097
5171 Description: Client trips assertion regarding lsm mismatch/magic
5172 Details : While revalidating inodes the VFS looks up inodes with ifind()
5173 and in rare cases can find an inode that is being freed.
5174 The ll_test_inode() code will free the lsm during ifind()
5175 when it finds an existing inode and then the VFS later attaches
5176 this free lsm to a new inode.
5180 Bugzilla : 6422, 7030
5181 Description: MDS deadlock between mkdir and client eviction
5182 Details : Creating a new file via mkdir or mknod (starting a transaction
5183 and getting the ns lock) can deadlock with client eviction
5184 (gets ns lock and trying to finish a synchronous transaction).
5187 Frequency : occasional
5188 Description: While starting a server, the fsfilt_ext3 module could not be
5190 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
5191 kernels. Previously, lconf would still use the ext3 name
5192 when trying to load modules. Now, it will correctly use
5193 ext3 on 2.4 and ldiskfs on 2.6.
5195 Severity : enhancement
5196 Description: The default stripe count has been changed to 1
5197 Details : The interpretation of the default stripe count (0, to lfs
5198 or lmc) has been changed to mean striping across a single
5199 OST, rather than all available. For general usage we have
5200 found a stripe count of 1 or 2 works best.
5202 Severity : enhancement
5203 Description: Add support for compiling against Cray portals.
5204 Details : Conditional compiling for some areas that are different
5208 Frequency : occasional
5209 Bugzilla : 6409, 6834
5210 Description: Creating files with an explicit stripe count may lead to
5211 a failed assertion on the MDS
5212 Details : If some OSTs are full or unavailable, creating files may
5213 trigger a failed assertion on the MDS. Now, Lustre will
5214 try to use other servers or return an error to the
5218 Frequency : occasional
5220 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
5221 caused lock timeout during readahead (since 1.4.2).
5222 Details : Processes doing readahead might match a lock that hasn't been
5223 granted yet if there are overlapping and conflicting lock
5224 requests. The readahead process waits on ungranted lock
5225 (original lock is CBPENDING), while OST waits for that process
5226 to cancel CBPENDING read lock and eventually evicts client.
5228 Severity : enhancement
5230 Description: Initial enabling of flock support for clients
5231 Details : Implements fcntl advisory locking and file status functions.
5232 This feature is provided as an optional mount flag (default
5233 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
5234 locking are implemented yet, and those that are are not guaranteed
5235 to be completely correct in production environments.
5236 mount -t lustre -o [flock|noflock] ...
5239 Frequency : occasional
5241 Description: OSTs running 2.4 kernels but with extents enabled might trip an
5242 assertion in the ext3 JBD (journaling) layer.
5243 Details : The b_committed_data struct is protected by the big kernel lock
5244 in 2.4 kernels, serializing journal_commit_transaction() and
5245 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
5246 access to this struct. In 2.6 kernels there is finer grained
5247 locking to improve SMP performance of the JBD layer.
5251 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
5256 Description: merge and cleanup kernel patches.
5259 -----------------------------------------------------------------------------
5261 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
5263 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
5264 versions of Lustre. Please update all servers and clients to
5265 version 1.4.2 or later at the same time. You must also run
5266 "lconf --write-conf {config}.xml" on the MDS while it is stopped
5267 to update the configuration logs.
5269 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
5270 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
5271 - swab llog records of type '0' so we get proper header size/idx (5861)
5272 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
5273 - fix rename of one directory over another leaking an inode (5953)
5274 - avoid SetPageDirty on 2.6 (5981)
5275 - don't re-add just-being-destroyed locks to the waiting list (5653)
5276 - when creating new directories, inherit the parent's custom
5277 striping settings if present parent (3048)
5278 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
5279 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
5280 - don't hold client locks on temporary worklist from l_lru (5666)
5281 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
5282 - automatically evict dead clients (5921)
5283 - Update file size properly in create+truncate+fstat case (6196)
5284 - Do not unhash mountpoint dentries, do not allow removal of
5286 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
5287 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
5288 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
5289 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
5290 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
5291 - Correctly handle reads of files with no objects (6243)
5292 - lctl recover will also mark a device active if deactivate used (5933)
5294 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
5295 - allow --write-conf on an MDS with different nettype than client (5619)
5296 - don't write config llogs to MDS for mounts not from that MDS (5617)
5297 - lconf should create multiple TCP connections from a client (5201)
5298 - init scripts are now turned off by default; run chkconfig --on
5299 lustre and chkconfig --on lustrefs to use them
5300 - upcalls are no longer needed for clients to recover to failover
5302 - add --abort-recovery option to lconf to abort recovery on device
5304 - add support for an arbitrary number of OSTs (3026)
5305 - Quota support protocol changes.
5306 - forward compatibility changes to wire structs (6007)
5307 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5308 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5309 - improved "lustre status" script
5310 - initialize blocksize for non-regular files (6062)
5311 - added --disable-server and --disable-client configure options (5782)
5312 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5313 - Vanilla 2.4.29 support
5314 - increase maximum number of obd devices to 520 (6242)
5315 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5316 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5317 SLES 9 2.6.5 kernels
5318 - limit stripes per file to 160 (the maximum EA size) (6093)
5320 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5323 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5324 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5325 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5326 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5327 - small ext3 extents cleanups and fixes (5733)
5328 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5329 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5330 - avoid needless client->OST connect, fix handle mismatch (5317)
5331 - fix DLM error path that led to out-of-sync client, long delays (5779)
5332 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5333 - fix several locking issues related to i_size (5492,5624,5654,5672)
5334 - don't move pending lock onto export if it is already evicted (5683)
5335 - fix kernel oops when creating .foo in unlinked directory (5548)
5336 - fix deadlock in obdfilter statistics vs. object create (5811)
5337 - use time_{before,after} to avoid timer jiffies wrap (5882)
5338 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5339 - Do not leak request if server was not able to process it (5154)
5340 - If mds_open unable to find parent dir, make that negative lookup(5154)
5341 - don't create new directories with extent-mapping (5909, 5936)
5343 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5344 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5345 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5346 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5347 - support for RHEL4 kernel on i686 (5773)
5348 - provide error messages when incompatible logs are encountered (5898)
5350 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5351 * version 1.4.0.10 (1.4.1 release candidate 1)
5353 - don't keep a lock reference when lock is not granted (4238)
5354 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5355 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5356 - drop import inflight refcount on signal_completed_replay error (5255)
5357 - unlock page after async write error during send (3677)
5358 - handle missing objects in filter_preprw_read properly (5265)
5359 - no transno return for symlink open, don't save no-trasno open (3440)
5360 - don't try to complete elan receive that already failed (4012)
5361 - free RPC server reply state on error (5406)
5362 - clean up thread from ptlrpc_start_thread() on error (5160)
5363 - readahead could read extra page into cache that wasn't ejected (5388)
5364 - prevent races in class_attach/setup/cleanup/detach (5260)
5365 - don't dereference de->d_inode after l_dput of de (5458)
5366 - use "int" for stripe value returned from lock_to_stripe (5544)
5367 - mballoc allocation and error-checking fixes in 2.6 (5504)
5368 - block device patches to fix I/O request sizes in 2.6 (5482)
5369 - look up hostnames for IB nals (5602)
5370 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5371 - don't start multiple acceptors for the same port (5277)
5372 - fix incorrect LASSERT in mds_getattr_name (5635)
5373 - export a proc file for general "ping" checking (5628)
5374 - fix "lfs check" to not block when the MDS is down (5628)
5376 - service request history (4965)
5377 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5378 - create an "evict_client" /proc entry on OSTs, like the MDS has
5379 - fix mount usage message, return errors per mount(8) (5168)
5380 - change grep [] to grep "[]" in tests so they work in more UMLs
5381 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5382 - remove ancient LOV_MAGIC_V0 EA support (5047)
5383 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5384 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5385 - allow readahead allocations to fail when low on memory (5383)
5386 - mmap locking landed again, after considerable improvement (2828)
5387 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
5389 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
5392 - send OST transaction number in read/write reply to free req (4966)
5393 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5394 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5395 - the watchdog thread now runs as interruptible (5246)
5396 - flock/lockf fixes (but it's still disabled, pending 5135)
5397 - don't use EXT3 constants in llite code (5094)
5398 - memory shortage at startup could cause assertion (5176)
5400 - reorganization of lov code
5401 - single portals codebase
5403 - add extents/mballoc support (5025)
5404 - direct I/O reads in the obdfilter (4048)
5405 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
5407 tbd Cluster File Systems, Inc. <info@clusterfs.com>
5410 - send OST transaction number in read/write reply to free req (4966)
5411 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5412 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5413 - the watchdog thread now runs as interruptible (5246)
5414 - handle missing objects in filter_preprw_read properly (5265)
5415 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5416 - drop import inflight refcount on signal_completed_replay error (5255)
5417 - unlock page after async write error during send (3677)
5418 - return original error code on reconstructed replies (3761)
5419 - no transno return for symlink open, don't save no-trasno open (3440)
5421 - add pid to ldlm debugging output (4922)
5422 - bump the watchdog timeouts -- we can't handle 30sec yet
5423 - extra debugging for orphan dentry/inode bug (5259)
5425 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
5428 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
5429 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
5430 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
5431 - fix NULL dereference in /proc/sys/portals/routes (4827)
5432 - allow failed mdc_close() operations to be interrupted (4561)
5433 - stop precreate on OST before MDS would time out on it (4778)
5434 - don't send partial-page writes before EOF from client (4410)
5435 - discard client grant for sub-page writes on large-page clients (4520)
5436 - don't free dentries not owned by NFS code, check generation (4806)
5437 - fix lsm leak if mds_create_objects() fails (4801)
5438 - limit debug_daemon file size, always print CERROR messages (4789)
5439 - use transno after validating reply (3892)
5440 - process timed out requests if import state changes (3754)
5441 - update mtime on OST during writes, return in glimpse (4829)
5442 - add mkfsoptions to LDAP (4679)
5443 - use ->max_readahead method instead of zapping global ra (5039)
5444 - don't interrupt __l_wait_event() during strace
5446 - add software watchdogs to catch hung threads quickly (4941)
5447 - make lustrefs init script start after nfs is mounted
5448 - fix CWARN/ERROR duplication (4930)
5449 - return async write errors to application if possible (2248)
5450 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
5451 - print NAL number in %x format (4645)
5452 - update barely-supported suse-2.4.21-171 series (4842)
5453 - support for sles 9 %post scripts
5454 - support for building 2.6 kernel-source packages
5455 - support for sles km_* packages
5457 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
5460 - ignore -ENOENT errors in osc_destroy (3639)
5461 - notify osc create thread that OSC is being cleaned up (4600)
5462 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
5463 - reconstruct ost_handle() like mds_handle() (4657)
5464 - create a new thread to do import eviction to avoid deadlock (3969)
5465 - let lconf resolve symlinked-to devices (4629)
5466 - don't unlink "objects" from directory with default EA (4554)
5467 - hold socknal file ref over connect in case target is down (4394)
5468 - allow more than 32000 subdirectories in a single directory (3244)
5469 - fix blocks count for O_DIRECT writes (3751)
5470 - OST returns ENOSPC from object create when no space left (4539)
5471 - don't send truncate RPC if file size isn't changing (4410)
5472 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
5473 - bind to privileged port in socknal and tcpnal (3689)
5475 - rate limit CERROR/CWARN console message to avoid overload (4519)
5476 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
5477 - basic mmap support (3918)
5478 - kernel patch series update from b1_4 (4711)
5480 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
5483 - avoid crash during MDS cleanup with OST shut down (2775)
5484 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
5485 - don't use bad inodes on the MDS (3744)
5486 - dynamic object preallocation to improve recovery speed (4236)
5487 - don't hold spinlock over lock dumping or change debug flags (4401)
5488 - don't zero obd_dev when it is force cleaned (3651)
5489 - print grants to console if they go negative (4431)
5490 - "lctl deactivate" will stop automatic recovery attempts (3406)
5491 - look for existing locks in ldlm_handle_enqueue() (3764)
5492 - don't resolve lock handle twice in recovery avoiding race (4401)
5493 - revalidate should check working dir is a directory (4134)
5495 - don't always mark "slow" obdfilter messages as errors (4418)
5497 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
5500 - don't close LustreDB during write_conf until it is done (3860)
5501 - fix typo in lconf for_each_profile (3821)
5502 - allow dumping logs from multiple threads at one time (3820)
5503 - don't allow multiple threads in OSC recovery (3812)
5504 - fix debug_size parameters (3864)
5505 - fix mds_postrecov to initialize import for llog ctxt (3121)
5506 - replace config semaphore with spinlock (3306)
5507 - be sure to send a reply for a CANCEL rpc with bad export (3863)
5508 - don't allow enqueue to complete on a destroyed export (3822)
5509 - down write_lock before checking llog header bitmap (3825)
5510 - recover from lock replay timeout (3764)
5511 - up llog sem before sending rpc (3652)
5512 - reduce ns lock hold times when setting kms (3267)
5513 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
5514 - fix NULL deref and obd_dev leak on setup error (3312)
5515 - replace some LBUG about llog ops with error handling (3841)
5516 - don't match INVALID dentries from d_lookup and spin (3784)
5517 - hold dcache_lock while marking dentries INVALID and hashing (4255)
5518 - fix invalid assertion in ptlrpc_set_wait (3880)
5520 - add libwrap support for the TCP acceptor (3996)
5521 - add /proc/sys/portals/routes for non-root route listing (3994)
5522 - allow setting MDS UUID in .xml (2580)
5523 - print the stack of a process that LBUGs (4228)
5525 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
5528 - don't cleanup request in ll_file_open() on failed MDS open (3430)
5529 - make sure to unset replay flag from failed open requests (3440)
5530 - if default stripe count is 0, use OST count for inode size (3636)
5531 - update parent mtime/ctime on client for create/unlink (2611)
5532 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
5533 - free recovery state on server during a forced cleanup (3571)
5534 - unregister_reply for resent reqs (3063)
5535 - loop back devices mounting and status check on 2.6 (3563)
5536 - fix resource-creation race that can provoke i_size == 0 (3513)
5537 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
5538 - more debugging for page-accounting assertion (3746)
5539 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
5540 - avoid deadlock after precreation failure (3758)
5541 - fix race and lock order deadlock in orphan handling (3450, 3750)
5542 - add validity checks when grabbing inodes from l_ast_data (3599)
5544 - add /proc/.../recovery_status to obdfilter (3428)
5545 - lightweight CDEBUG infrastructure, debug daemon (3668)
5546 - change default OSC RPC parameters to be better on small clusters
5547 - turn off OST read cache for files smaller than 32MB
5548 - install man pages and include them in rpms (3100)
5549 - add new init script for (un)mounting lustre filesystems (2593)
5550 - run chkconfig in %post for init scripts (3701)
5551 - drop scimac NAL (unmaintained)
5553 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
5556 - clean kiobufs before and after use (3485)
5557 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
5558 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
5559 - change init script to fail more clearly if not run as root (1528)
5560 - allow clients to reconnect during replay (1742)
5561 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
5562 - don't do DNS lookups on NIDs too small for IP addresses (3442)
5563 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
5564 - fix cond_resched (3554)
5565 - only evict unfinished clients after recovery (3515)
5566 - allow bulk resend, prevent data loss (3570)
5567 - dynamic ptlrpc request buffer allocation (2102)
5568 - don't allow unlinking open directory if it isn't empty (2904)
5569 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
5570 - remove extraneous obd dereference causing LASSERT failure (3334)
5571 - don't use get_cycles() when creating temp. files on the mds (3156)
5572 - hold i_sem when setting i_size in ll_extent_lock() (3564)
5573 - handle EEXIST for set-stripe, set proper directory name (3336)
5575 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
5576 - fix ksocknal_fmb_callback() error messages (2918)
5578 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
5581 - don't copy lvb into (possibly NULL) reply on error (2983)
5582 - don't deref dentry after dput, don't free lvb on error (2922)
5583 - use the kms to determine writeback rpc length (2947)
5584 - increment oti_logcookies when osc is inactive (2948)
5585 - update client's i_blocks count via lvb messages (2543)
5586 - handle intent open/close of special files properly (1557)
5587 - mount MDS with errors=remount-ro, like obdfilter (2009)
5588 - initialize lock handle to avoid ASSERT on error cleanup (3057)
5589 - don't use cancelling-locks' kms values (2947)
5590 - use highest lock extent for kms, not last one (2925)
5591 - don't dereference ERR_PTR() dentry in error handling path (3107)
5592 - fix thread race in portals_debug_dumplog() (3122)
5593 - create lprocfs device entries at setup instead of at attach (1519)
5594 - common AST error handler, don't evict client on completion race (3145)
5595 - zero nameidata in detach_mnt in 2.6 (3118)
5596 - verify d_inode after revalidate_special is valid in 2.6 (3116)
5597 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
5598 - initialize RPC timeout timer earlier for 2.6 (3219)
5599 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
5600 - print nal/nid for unknown nid (3258)
5601 - additional checks for oscc recovery before doing precreate (3284)
5602 - fix ll_extent_lock() error return code for 64-bit systems (3043)
5603 - don't crash in mdc_close for bad permissions on open (3285)
5604 - zero i_rdev for non-device files (3147)
5605 - clear page->private before handing to FS, better assertion (3119)
5606 - tune the read pipeline (3236)
5607 - fix incorrect decref of invalidated dentry (2350)
5608 - provide read-ahead stats and refine rpc in flight stats (3328)
5609 - don't hold journal transaction open across create RPC (3313)
5610 - update atime on MDS at close time (3265)
5611 - close LDAP connection when recovering to avoid server load (3315)
5612 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
5613 - don't leak open file on MDS after open resend (3325)
5614 - serialize filter_precreate and filter_destroy_precreated (3329)
5615 - loop device shouldn't call sync_dev() for nul device (3092)
5616 - clear page cache after eviction (2766)
5617 - resynchronize MDS->OST in background (2824)
5618 - refuse to mount the same filesystem twice on same mountpoint (3394)
5619 - allow llmount to create routes for mounting behind routers (3320)
5620 - push lock cancellation to blocking thread for glimpse ASTs (3409)
5621 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
5622 - fix rare problem with rename on htree directories (3417)
5624 - allow default OST striping configuration per directory (1414)
5625 - fix compilation for qswnal for 2.6 kernels (3125)
5626 - increase maximum number of MDS request buffers for large systems
5627 - change liblustreapi to be useful for external progs like lfsck (3098)
5628 - increase local configuration timeout for slow disks (3353)
5629 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
5631 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5634 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
5635 - don't overwrite extent policy data in reply if lock was blocked (2901)
5636 - drop filter export grants atomically with removal from device (2663)
5637 - del obd_self_export from work_list in class_disconnect_exports (2908)
5638 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
5639 - swab reply message in mdc_close, other PPC fixes (2464)
5640 - fix destroying of named logs (2325)
5641 - overwrite old logs when running lconf --write_conf (2264)
5642 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
5643 - fix race in target_handle_connect (2898)
5644 - mds_reint_create() should take same inode create lock (2926)
5645 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
5646 - don't close files for self_export to avoid uninitialized obd (2936)
5647 - allow MDS with the same name as client node (2939)
5648 - hold dentry reference for closed log files for unlink (2325)
5649 - reserve space for all logs during transactions (2059)
5650 - don't evict page beyond end of stripe extent (2925)
5651 - don't oops on a deleted current working directory (2399)
5652 - handle hard links to targets without a parent properly (2517)
5653 - don't dereference NULL lock when racing during eviction (2867)
5654 - don't grow lock extents when lots of conflicting locks (2919)
5656 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
5659 - account for cache space usage on clients to avoid data loss (974)
5660 - lfsck support in lustre kernel code (2349)
5661 - reduce journal credits needed for BRW writes (2370)
5662 - orphan handling to avoid losing space on client/server crashes
5663 - ptlrpcd can be blocked, stopping ALL progress (2477)
5664 - use lock value blocks to assist in proper KMS, faster stat (1021)
5665 - takes i_sem instead of DLM locks internally on obdfilter (2720)
5666 - recovery for initial connections (2355)
5667 - fixes for mds_cleanup_orphans (1934)
5668 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
5669 - block all file creations until orphan recovery completes (1901)
5670 - client remove rq_connection from request struct (2423)
5671 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
5672 - recovery timer race (2670)
5673 - mdc_close recovey bug (2532)
5674 - ptlrpc cleanup bug (2710)
5675 - mds timeout on local locks (2588)
5676 - namespace lock held during RPCs (2431)
5677 - handle interrupted sync write properly (2503)
5678 - don't try to handle a message that hasn't been replied to (2699)
5679 - client assert failure during cleanup after abort recovery (2701)
5680 - leak mdc device after failed mount (2712)
5681 - ptlrpc_check_set allows timedout requests to complete (2714)
5682 - wait for inflight reqs when ptlrpcd finishes (2710)
5683 - make sure unregistered services are removed from the srv_list
5684 - reset bulk XID's when resending them (caught by 1138 test)
5685 - unregister_bulk after timeout
5686 - fix lconf error (2694)
5687 - handle write after unfinished setstripe, stripe-only getstripe (2388)
5688 - readahead locks pages, leaves pending causing memory pressure (2673)
5689 - increase OST request buffers to 4096 on large machines (2729)
5690 - fix up permission of existing directories in simple_mkdir (2661)
5691 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
5692 - don't assign transno to errored transactions (2742)
5693 - don't delete objects on OST if given a bogus objid from MDS (2751)
5694 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
5695 - if rq_no_resend, then timeout request after recovery (2432)
5696 - fix MDS llog_logid record size, 64-bit array alignment (2733)
5697 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
5698 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
5699 - MDS recovery shouldn't delete objects that it has given out (2730)
5700 - if enqueue arrives after completion, don't clobber LVB (2819)
5701 - don't unlock pages twice when trigger_group_io returns error (2814)
5702 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
5703 - don't write pages to disk if there was an error (1450)
5704 - don't ping imports that have recovery disabled (2676)
5705 - take buffered bytes into account when balancing socknal conn (2817)
5706 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
5707 - reconnect unlink llog connection after MDS reconnects to OST (2816)
5708 - remove little-endian swabbing of llog records (1987)
5709 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
5710 - retry reposting request buffers if they fail (1191)
5711 - grow extent at grant time to avoid granting a revoked lock (2809)
5712 - lock revoke doesn't evict page if covered by a second lock (2765)
5713 - disable VM readahead to avoid reading outside lock extents (2805)
5715 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
5716 - updated kernel patches for hp-2.4.20 kernel (2681)
5718 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
5721 - fix truncated write corruption (2366)
5722 - fix for failed assertion in iopen_connect_dentry (1792,2517)
5724 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
5725 - ensure len(uuid) < 37 in lmc (1171)
5726 - fix ia64 OOPS in llog_test (2255)
5727 - zero end of page at obdfilter for partial page writes (2648)
5728 - don't leave stale dentries around after renames (bug 2428)
5729 - fix timeouts when evicting a client with a single lock held (2642)
5730 - set deadline for the initial HELLO message to drain (2634)
5731 - print out dotted-quad IP addresses in the socknal (2302)
5733 - additional debugging for MDS client eviction problem (2443)
5734 - fix mkfsoptions support for osts (2603, 2604)
5736 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
5739 - add series for the vanilla 2.6.0 kernel
5740 - add series for the vanilla 2.4.24 kernel
5741 - add series for a cray x86/64 UL kernel drop
5742 - fix xattr patches for the vanilla 2.4.19 series
5744 - generate true UUIDs in lmc (1171)
5745 - have portals stack dumping break in UML (2466)
5746 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
5747 - allocate with _NOFS in ldlm to avoid deadlock (1933)
5748 - wake callback waiting threads on client eviction (2460)
5749 - Add --ptldebug and --subsystem to lmc (1719)
5750 - update assertion to allow safe interrupt allocation
5751 - set rq_no_resend for cancel requests (2432)
5752 - recalculate ptlrpcd timeout after resend (2494)
5753 - call vfs_rmdir when removing pending directories (2368)
5754 - fix renaming a file to itself (2429)
5755 - lmc creates a default one-stripe lov (2454)
5756 - expand procfs space to handle large clusters (2326)
5757 - increase UML stack to avoid overflow
5758 - update lconf's list of debug and subsystem masks
5759 - fix lfs find --obd (2510)
5760 - /proc tunable for disabling filter read caching (2591)
5761 - stop rpm packages from altering slapd.conf (2301)
5762 - disable nagle in the socknal under 0conf (2578)
5763 - choose mds inode size based on stripe count (2572)
5764 - fix kernel-source rpm problems (2516)
5766 - add --disable-doc to avoid pdf generation (2421)
5767 - update documentation, tests, type-os, comments
5768 - avoid format warnings on ia64
5769 - remove the TOE NAL
5770 - tiny code cleanups by removing unused fields
5772 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
5775 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
5776 - remove the most blatant lies from BUILDING file (2371)
5777 - change default debug level to reasonable production setting
5778 - reduce client side cache size to reduce cache flush time
5779 - reduce max RPCs in flight to avoid unnecessary file fragmentation
5780 - make TCP zerocopy and pinger support enabled by default (2476)
5781 - sync writes completed after process exits caused crashes (2319)
5782 - maintain correct mount count on the MDS (2356)
5783 - backout 1557, because 2316 wasn't really fixed
5784 - better file I/O statistics gathering in /proc
5785 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
5786 - another kernel patch to fix zero-copy TCP function export
5787 - don't take duplicate lock when processing re-sent getattr (2420)
5788 - lctl uses obd_self_export instead of creating new conn (2353)
5789 - MDS/OST recovery case which requires object creation asserted (2425)
5790 - move lfs from /usr/sbin to /usr/bin in packages
5791 - fix race between mds_client_add and mds_client_free (2417)
5792 - use kmalloc instead of slabs in portals (2430)
5793 - don't create duplicate records when a failover MDS is present (2442)
5794 - remove unnecessary mount age check (2332)
5795 - don't remove directory inodes from locks prematurely (2451)
5796 - don't break if MDS service name is the same as hostname (2103)
5797 - fix races in client write RPC generation when cache full (2482)
5799 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
5802 - remove now-unused request->rq_obd (278)
5803 - if an allocation fails, print out how much memory we've used (1933)
5804 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
5805 - add the "configurable stack size" patch to most series files (1256)
5806 - ability to write large log records, for 100+ OST configs (2306)
5807 - fix NULL deref when filter_prep fails (2314)
5808 - fix operator precedence error in filter_sync
5809 - dynamic allocation of socknal TX descriptors (2315)
5810 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
5811 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
5812 - ensure that CWARN messages in Portals always get to the syslog
5813 - __init/__exit are not for prototype decls (ldlm_init/exit)
5814 - x86-64 compile warning fixes
5815 - fix gateway LMC keyword conflict (2318)
5816 - fix MDS lock inversions in getattr/reint paths (1844)
5817 - fix a rare lock re-ordering bug, which caused deadlock (2322)
5818 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
5819 - DLM race condition prevented some lock evictions (2328)
5820 - ENOMEM detection and retry on socknal sends (2230)
5821 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
5822 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
5823 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
5824 - better error messages when a client is rejected during recovery (1505)
5825 - avoid cancelling locks which were never granted, after failure (2330)
5826 - fix i_sem/journal inversion in mds_client_add (2333)
5827 - fix truncate/getattr lock cycle deadlock (2334)
5828 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
5829 - fix two rare exit paths which could leak an l_lock() ref (2321)
5830 - fencepost error in MDS/OST orphan recovery (2226)
5831 - make log record alignment 8 bytes (1988)
5832 - lstripe now fails when requested offset > ost_count (2237)
5833 - ensure that all kernel series have a complete list.h (1607)
5834 - fix crashes in special-file operations (2316)
5835 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
5837 - allow configurable automake binary, for testing new versions
5838 - small update to the lfs documentation
5840 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
5842 * fix negative export reference count in fsfilt_sync (2312)
5844 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
5845 * release candidate 0.9.1
5847 - orphans are moved into the PENDING directory for possible recovery
5848 - replayed opens now open by fid for orphan/rename safety (1042)
5849 - last close of an orphan inode generates a transno (683)
5850 - chdir() and mount() now pin the directory entry (1020)
5851 - avoid CERROR in normal ll_setattr_raw() error case (1500)
5852 - discard very old requests without processing them (1502)
5853 - remove some common, well-understood CERRORs (1505)
5854 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
5855 - clear "grant" flags in OST replies until OST grant code lands (1644)
5856 - fix read performance by not clobbering i_blksize on client (1598)
5857 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
5858 - make LRU size a /proc tunable, clears locks when reduced (707)
5859 - fix some lprocfs rot that prevented ptlbd from loading (1732)
5860 - server locks take references on exports now (1558)
5861 - build fixes for 2.4.20-rh trees (1663)
5862 - return an error from lov_create if all OSCs are inactive (1751)
5863 - fix import levels when a reconnect happens without a timeout (1597)
5864 - exit early from mds_open if we get a lookup error (1749)
5865 - partial page read at EOF wouldn't wait for disk before sending (1642)
5866 - avoid NULL deref in obdfilter when reading page past EOF (1592)
5867 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
5868 - fix LBUG in ll_it_open_error with rc = -2 (1861)
5869 - write/truncate lock inversion (1639)
5870 - Don't auto-load obdclass, portals modules during cleanup (1495)
5871 - fix timestamps from jumping to "now" (1763)
5872 - extra journal assertions (1648)
5873 - add an extra multiunlink test (1771)
5874 - fix read_record/write_record API (1776)
5875 - fix leak of offset_extent, possible incorrect i_size later (1772)
5876 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
5877 - Debugging for the kqswnal_get_idle_tx problems (1820)
5878 - Allow recovery to be attempted multiple times (1536)
5879 - Write out MDS last_rcvd file after it is first created (1600)
5880 - Fix tx_descriptor leak in failed transmit situations (1827)
5881 - ext3 journaling fixes for assertion failure after IO error (1871)
5882 - class_export_put() on freed export after completion AST error (1896)
5883 - Fix revalidate looping in VFS (1322)
5884 - Don't access a freed export during MDS_REINT timeout (1521)
5885 - Add open-unlink recovery support on the MDS (1673,1764)
5886 - Return an error if no MDS data was read from last_rcvd (1946)
5887 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
5888 - Don't setup a disk device that doesn't match exported UUID (317)
5889 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
5890 - avoid committing NULL handle in force close
5891 - local.sh is now a one-stripe LOV configuration
5892 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
5893 - fix ext3 htree duplicate directory entry corruption (1516)
5894 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
5895 - update to new LOV EA format (2097)
5896 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
5898 2003-06-15 Phil Schwan <phil@clusterfs.com>
5901 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
5902 - per-import/export recovery handling (958, 931, 959)
5903 - multiple last-rcvd slots, for serving multiple FSes (949)
5904 - connections are again shared between multiple imp/exports (963, 964)
5905 - "umount -f" would hang if any requests needed to be sent (393, 978)
5906 - avoid pinning large req buffer by copying for queued messages (989)
5907 - add "uuid" to "lctl device" command to help upcalls (991)
5908 - "open" RPCs with transnos would confuse recovery counters (1037)
5909 - do proper endian conversion of all wire messages (288, 340, 891)
5910 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
5911 - call waiting locks callback from LDLM recovery thread (1127, 1151)
5912 - fix ptlrpc_connection leak in target_handle_connect (1174)
5913 - fix import refcounting bug in OST and MDS cleanup (1134)
5914 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
5915 - fix very unlikely obd_types race condition (501)
5916 - remove osc_open hack for echo_client (1187)
5917 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
5918 - a failure in read_inode2 leads to deadlock (1139)
5919 - cancel ack-locks as soon as transaction is committed (1072)
5920 - fix major leaks and crashes in the bulk I/O path (937, 1057)
5921 - make sure to commitrw after any preprw to avoid deadlock (1162)
5922 - failing to execute a file in a lustre FS would lock inode (1203)
5923 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
5924 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
5925 - obd_uuid2tgt would walk off the end of the list (1255)
5926 - on IA64 the osc would give portals incorrect bulk size (1258)
5927 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
5928 - fix lock inversion caused by new llite matching code (1282)
5929 - limit the number of dirty pages on a client to 10MB (1286)
5930 - timed out locks were not being corrected cancelled (1289)
5931 - fix O_DIRECT above 4GB on IA-32 (1292)
5932 * major user-visible changes
5933 - fail out/fail over policy now controlled by the upcall (993)
5935 - add OBD_PING to check server availability and failure (954)
5936 - lustre messages are now sent in sending host order (288, 340, 891)
5937 - add eadatalen to MDS getattr reply (340)
5938 - OST read replies may contain second buffer, with per-page status (593)
5940 2003-03-11 Phil Schwan <phil@clusterfs.com>
5943 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
5944 - failed open()s could cause deadlock; fixed (867, 869)
5945 - stop cancelling OST locks when files are closed (481)
5946 - overlapping XID spaces caused network corruption (851, 853)
5947 - fix unsafe fsfilt counter arithmetic; change to atomic_t
5948 - setattr_raw added, to do single-RPC, server-side setattrs
5949 - lmc/lconf syntax change for OST UUIDs
5950 - fix crashy race condition between ptlrpc_free_req and osc_close
5951 - don't use request in mdc_enqueue if we hit a timeout (889)
5952 - don't set the inode i_size for regular files from the MDS (896)
5953 - handle out of order completion AST (842)
5954 - don't LBUG if a lock request times out after receiving AST (913)
5955 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
5956 - if a bad lock AST arrives, send an error instead of dropping entirely
5957 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
5958 - fix leak in bulk IO when only partially completed (899, 900, 926)
5959 - fix O_DIRECT for ia64 (55)
5960 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
5961 - C-z after timeout could hang a process forever; fixed (977)
5963 - client-side I/O cache (678, 924, 929, 941, 970)
5965 - READPAGE and SETATTRs which don't take server-side locks get
5968 2003-02-11 Phil Schwan <phil@clusterfs.com>
5971 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
5972 - Fix fsfilt_extN_readpage() to read a full page of directory
5973 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
5974 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
5975 - Fix imbalanced LOV object allocation and out-of-bound access (469)
5976 - Most intent operations were removed, in favour of a new RPC mode
5977 that does a single RPC to the server and bypasses most of the VFS
5978 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
5979 - Aggressively cancel local locks on DLM servers
5980 - mds_reint_unlink sends EA to the client if it's the last nlink.
5981 client uses that EA to unlink OST objects.
5982 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
5983 - recursive symlinks were fixed (439)
5984 - fixed NULL deref in DEBUG_REQ
5985 - filter_update_lastobjid no longer calls sync, which annoyed extN
5986 - fixed multi-client small-writes to a single file problem (445)
5987 - fixed mtime updates during file writes (607)
5988 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
5989 - fixed bug in obd_brw_read/write() (under guise of testing 367)
5990 - fixed Linux OST size reporting problem (444, 656)
5991 - OST now updates object mtime with writes or setattr (607, 619)
5992 - client verifies file size before zeroing page past EOF (445)
5993 - OST now writes last allocated objid to disk with allocation (108)
5994 - LOV on echo now works (409)
5996 - mds_reint_unlink sends a new buffer, with the EA included. this
5997 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
5998 set if a regular file was being unlinked, and it was the last link
5999 - use PtlGet from the target for bulk writes (315)
6000 - OST now updates object mtime with writes or setattr (607, 619)
6001 - LDLM now has a grant-time callback to revalidate locked items, if
6003 - Many MDS operations were reorganized to combat race conditions
6005 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
6006 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
6008 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
6011 - Fully reactivate OST imports after reconnection (512, others)
6012 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
6013 - More graceful error handling for truncating on dead OST (515)
6014 - Don't error out unless we're actually accessing dead stripes (474)
6015 - Fix garbage sizes when stripes are missing (410)
6016 - LRU counters were broken, causing constant lock purge (433, 432)
6017 - garbage on read from stripes with failed OSTs (441)
6018 - mark OSCs as active before reconnecting during recovery (438)
6019 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
6020 - lfind did not preserve OST order in output (443)
6021 - symlinks cause hung clients, incorrect data (439)
6022 - stop dereferencing request after dropping refcount (457)
6023 - don't LASSERT(spin_is_locked) on non-SMP (455)
6024 - fixes for many rename() bugs
6025 - fstat didn't correctly synchronize attributes (399)
6026 - server must handle lock cancellation during blocking AST prep (487)
6027 - bulk descriptors were free()d too soon (511)
6028 - fix paths in lconf, which would load incorrect modules (451, 507)
6029 - fix confusing lconf 'host not found' error message (386)
6030 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
6031 - fix race condition in mdc_blocking_ast() for inode access (526)
6032 - fix lov_unpackmd() unpacking wrong number of stripes (537)
6033 - fix lov_set_osc_active() marking wrong OSC inactive (440)
6034 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
6035 - fix multiple writes of stripe MD to MDS (358, maybe 519)
6036 - fix lstripe in several ways (kernel side) (527)
6037 - fix request leak in ldlm_cli_enqueue (262)
6038 - incorrect OSC was marked inactive after OST failure
6039 - call mds_fs_cleanup before unmounting filesystem (524)
6040 - fix races between taking ns_lock and ldlm_lock_change_resource
6041 - fix races updating LOV export open file list
6042 - fix lov_enqueue error path, avoid decref-ing bad lock handle
6043 - fix recovery NULL deref in ldlm_cli_cancel_unused
6044 - fix some DLM races by using new hash table for lock handles (419)
6045 - permit the client to specify desired inodes, at replay
6046 - duplicate requests when we queue them for replay reintegration
6047 - fix last_rcvd offset calculation
6048 - sync after each recovered transaction, so we always make progress
6049 - never, not always, ERESTART requests without transnos
6050 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
6052 - skip replay if the MDS says that the client is already connected
6053 - don't check for a recovery-enabled export to match lctl's UUID
6054 - don't INC_USE_COUNT for phantom exports
6055 - don't crash when cleaning up phantom exports (567)
6056 - don't double-finish or set replay data for errored mdc_open requests
6057 - abort requests when they time out, so we don't get old replies
6058 - send/receive replies for AST messages again
6059 - if the client says that it doesn't have the lock, cancel it on the
6061 - if we timeout during I/O, don't try to cancel an in-use lock; instead
6062 mark it as destroyed, it will all work out when decref is called
6063 - fix module use counts (22, 581)
6065 - ASTs now expect a reply (server cancels lock on error reply)
6067 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
6070 - fix many simultaneous client startup (392)
6071 - fix dentry->d_it clobbering
6072 - credentials weren't being shipped for readdir/getattr operations
6073 - remove invalid assertions triggered during some concurrent MD
6075 - proper Lustre versions added (336, 389)
6076 - fix memory leak for create error case (398)
6077 - fix LOV locking bug that would get cli/srv out of sync
6078 - fix echo client over LOV (409)
6079 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
6080 - fix double-O_EXCL intent crash (424)
6081 - avoid sending multiple lock CANCELs (352)
6083 - MDS can do multi-client recovery (modulo bugs in new code)
6085 - many updates, edits, cleanups
6087 2002-11-18 Phil Schwan <phil@clusterfs.com>
6090 - fix null d_it dereference (346)
6091 - fix full OST/dbench hang (333)
6092 - fix permission problem with file removal (286)
6093 - fix removal of OSCs from LOV when they fail
6094 - fix NULL deref during bulk timeout (214)
6095 - fix problems related to multiple filesystems on one MDS (241)
6096 - fixed serious subtle metadata locking bugs
6097 - free locks on clients when inodes are removed due to memory
6099 - fix inode pointer in lock data (285)
6100 - partial support for multiple MDS on a single host (241)
6101 - data locks weren't cancelled at clear_inode time (290, 311)
6102 - intent locks could lead to unbounded lock growth (205)
6103 - added a maximum lock count, an LRU list, and a flusher
6104 - fix multiple rename (365)
6105 - properly abstracted the echo client
6106 - OSC locked 1 byte too many; fixed
6107 - rewrote brw callback code:
6108 - fixed recovery bugs related to LOVs (306)
6109 - fixed too-many-pages-in-one-write crash (191)
6110 - fixed (again) crash in sync_io_timeout (214)
6111 - probably fixed callback-related race (385)
6113 - Add capability to MDS protocol
6114 - LDLM cancellations and callbacks on different portals
6116 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
6119 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
6120 - defer open object destruction to close time (601981, 138)
6121 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
6122 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
6123 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
6124 - don't call obd_setattr in ll_file_release for destroyed objects
6125 * protocol change to lustre_msg: move |version| and add |flags|
6126 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
6127 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
6128 * added replay of create, unlink, link and rename operations during
6129 MDS failover; recovery should be much more robust now
6130 * remove failed OSCs from LOVs (only lov_create uses this so far)
6131 * the lustre-HOWTO was brought (more) up to date (582544)
6133 2002-10-23 Phil Schwan <phil@clusterfs.com>
6136 - in-use dentries weren't being reused properly (617851)
6137 - prevent multiple LDLM setup (599178)
6138 - fix LOV size calculations for truncate (617853)
6139 - fix client handling of MDS intent errors (POSIX)
6140 - fix permission bug in lovstripe.c test (624321)
6141 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
6142 - truncate past end of file could corrupt data
6143 - proper cleanup after timeouts, crashes, etc (592524, 550815)
6144 - a race in recovery could return ETIMEDOUT to apps (623947)
6145 - building outside the source directory was fixed
6146 * the lustre-HOWTO was brought (more) up to date (582544)
6147 * major progress was made on recovery functionality
6149 2002-10-10 Phil Schwan <phil@clusterfs.com>
6152 - recovery deadlock fix
6153 - rm -rf causes LBUG fix (617817)
6154 - file open by multiple tasks fix (618962)
6155 - directory permissions bugs (602707 and 620007)
6156 - journal_stop fixed with locking (611313)
6157 - O_APPEND failures resolved (618273, perhaps 614459)
6158 - lconf PATH fix (619770)
6159 - IA64 build fix (621450)
6160 - RPC buffer sizes scale with amount of memory
6162 2002-10-01 Phil Schwan <phil@clusterfs.com>
6165 - locks would be cancelled without throwing away data pages,
6166 resulting in inconsistent data (605627)
6167 - inode attributes were not always being refreshed (605627, 612449)
6168 - lconf now continues to cleanup after lctl reports an error
6169 - MDS now enforces user permissions (602707)
6170 - lprocfs cleanup fixed, but not yet enabled (614157)
6171 - fixed infinite server hang, should a client not respond to an AST
6172 - avoid going into recovery if user calls readlink() with a buffer
6173 that's too small (613941)
6174 - AST RPCs no longer require replies (614867) -- this may be changed
6175 - don't crash server if client sends an IOV that's too big (611336)
6176 - fixed lock conversion deadlock (611892)
6177 - fixed the following of symlinks (614622)
6178 * recovery: the server can remove locks from a client that dies, other
6179 clients can make progress
6180 * more extN patch fixes
6181 * compile-time configurable ptlrpc buffer allocations
6183 - collaborative read cache document
6184 - Lustre Lite Performance CDR document-in-progress
6186 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6189 - fix typo in patch-2.4.18
6191 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6194 - clear ptlrpc request each time in handle_incoming_request()
6195 - unlink of files now destroys the object on the OST
6197 2002-09-19 Peter Braam <braam@clusterfs.com>
6199 * add hard link support
6200 * change obdfile creation method
6201 * kernel patch changed
6203 2002-09-19 Peter Braam <braam@clusterfs.com>
6206 - stack overflow bug in extN fixed
6208 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
6210 * documentation updates
6211 - add man pages for config tools
6212 - update tests/README to describe testing with new config tools
6213 - finish metadata API descriptions
6214 * bug fixes and cleanups
6215 - statfs workaround for 16TB limit
6216 - LOV stripe allocation improved, can stripe on subset of OSTs
6217 - LOV file size/IO offset was wrong for files > 4GB in size
6218 - object EA data was being dropped, caused files to be unreadable
6219 - memory overflow with non-LOV OST caused memory corruption
6220 - fixed regression tests to work with new config tools, obdfilter
6221 - fixed bug when directory size became larger than 1 block
6222 - fixed bug (for single client case) when PWD was deleted
6223 - invalidate local directory pages when doing intent-based ops
6224 - avoid LDLM oops when lock callback contained bad data
6226 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
6228 * documentation updates
6229 * bug fixes and cleanups
6230 - configuration tools
6233 - 64-bit compile warnings
6234 - 64-bit internal statfs data
6236 * test_brw on persistent OST devices
6238 * lprocfs (disabled)
6240 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
6242 * documentation updates
6243 * bug fixes and cleanups
6244 * configuration tools
6246 2002-08-30 Peter J. Braam <braam@clusterfs.com>
6249 * many small fixes to 0_5_4
6250 * io/network handling
6251 * thinkos in MDS operations
6253 2002-08-24 Peter J. Braam <braam@clusterfs.com>
6256 * crucial basic fixes to 0.5.3
6257 * IOR, Iozone work over Elan
6260 2002-08-07 Phil Schwan <phil@clusterfs.com>
6261 * version 0_5_3, our first alpha
6262 * we use the new Portals iovs
6263 * documentation updates
6264 * bug fixes and cleanups
6265 * small changes in the DLM wire protocol
6267 2002-07-25 Peter J. Braam <braam@clusterfs.com>
6268 * version 0_5_1 with some initial stability,
6269 * locking on MD and file I/O.
6270 * documentation updates
6271 * several bug fixes since 0.5.0
6272 * small changes in wire protocol
6274 2002-07-18 Phil Schwan <phil@clusterfs.com>
6276 * delivered as Lustre Light Alpha
6277 * fixed a crash after handling invalid MDS requests
6278 * fixed directory pages for architectures with non-4k pages sizes
6280 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
6281 * release version v0_4_4
6282 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
6284 * Fixes a number of interruption problems with OST operations.
6285 * Update documentation for portals header changes
6286 * Move all wire protocol structs/defines to lustre_idl.h
6287 * Fixes symlink length bug.
6288 * Add tcpdump to repository.
6290 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
6291 * release version v0_4_3
6292 * Fixes statfs for inodes on extN.
6293 * Fixes bug in runtests which would delete /etc/hosts.
6294 * Use 64-bit object IDs wherever possible (not into VFS though)
6295 Remove ost_get_info, which is unused by lustre, and out of date.
6297 2002-07-03 Peter Braam <braam@clusterfs.com>
6298 * release version v0_4_2 Fixes a lookup error (type not passed)
6299 * move forward to head of Portals
6300 * move forward to latest Lustre kernel
6302 2002-06-25 Peter Braam <braam@clusterfs.com>
6303 * release version v0_4_1. Hopefully stable on single node use.