3 * Server support for kernels:
4 2.6.18-194.17.1.el5 (RHEL 5)
5 2.6.18-194.17.1.0.1.el5 (OEL 5)
6 * Client support for unpatched kernels:
7 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
8 2.6.18-194.17.1.el5 (RHEL 5),
9 2.6.18-194.17.1.0.1.el5 (OEL 5)
10 2.6.16.60-0.69.1 (SLES 10),
12 2.6.29.4-167.fc11 (FC11)
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.
19 Frequency : only with SLES10
21 Description: Use OFED "KMP" provided by Novell
22 Details : SLES10 SP3 ships with OFED in a separate "KMP" package.
23 Lustre is now built against this package. That means you need to
24 install the ofed-kmp package from Novell for the patchless client.
25 Note that the ofed-kmp that Novell ships may not exactly match the
26 kernel version but should still be compatible.
28 Severity : enhancement
30 Description: Update RHEL5.5 kernel to 2.6.18-194.17.1.el5
31 Update OEL5.5 kernel to 2.6.18-194.17.1.0.1.el5.
33 --------------------------------------------------------------------------------
35 2010-07-15 Oracle, Inc.
37 * Server support for kernels:
38 2.6.18-164.11.1.el5 (RHEL 5)
39 2.6.18-164.11.1.0.1.el5 (OEL 5)
40 * Client support for unpatched kernels:
41 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
42 2.6.18-164.11.1.el5 (RHEL 5),
43 2.6.18-164.11.1.0.1.el5 (OEL 5)
44 2.6.16.60-0.42.8 (SLES 10),
46 2.6.29.4-167.fc11 (FC11)
47 * Recommended e2fsprogs version: 1.41.10-sun2
49 Severity : enhancement
51 Description: Distribute lfs_migrate script with lustre
52 Details : The lfs_migrate mentioned in the manual is now included in the
53 lustre tree in lustre/scripts and will distributed in lustre RPMs.
55 Severity : enhancement
57 Description: Failover nid list added to import proc file.
59 Severity : enhancement
61 Description: "som_preview" mount option required to use SOM
62 Details : The clients will need to use som_preview mount option
66 Description: Lustre does not do 1MB IOs to HW RAID
67 Details : Bump MAX_PHYS/HW_SEGMENTS and SG_ALL to 256 in the RHEL5 kernel.
69 Severity : enhancement
71 Description: lfs quota output is not very convenient for awk/sed-parsing
72 Details : Some positions in lfs quota output table could be empty or
73 non-empty which made it hard to parse it with scripts, now
74 a dash is put instead of space where there is not supposed
79 Description: Update to OFED 1.5.1
80 Details : OFED 1.5.1 will be used in environments that don't use
85 Description: Remove lnet. prefix from lctl params display
86 Details : Also renamed lnet.memused to lnet_memused to
91 Description: Use conf_param -d to remove permanant settings.
95 Description: Add list_param -R
96 Details : lctl list_param now supports recursive mode.
99 Bugzilla : 21829, 22312
100 Description: Fix llobdstat.
101 Details : 1. Now updating the write_bytes counter
102 2. Port script fixes from 1.8.
106 Description: decrease the usage of memory on clients.
107 Details : 1. On clients, recycle dentries and inodes unused.
108 2. Delete the code related to ll_deathrow(att 6215 in bug 1443). It
113 Description: llobdstat fix and enhancement
114 Details : add a counter to set a limit to how many samples will be returned
115 fix a wildcard in the path to limit to obdfilter stats only
119 Description: lfs check only allowed for root
120 Details : Code cleanups and added a regresion test of non-root use of
125 Description: MDS Oopses in ptlrpc_dispatch_difficult_reply()/__list_add()
126 Details : Fix hr->hr_index overload.
130 Description: Fix FIEMAP functionality.
134 Description: FC11 patchless client support added
138 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and
139 OEL5.4 kernel to 2.6.18-164.11.1.0.1.el5.
143 Description: New acc-sm script and various test-framework enhancments.
147 Description: Previously landed patch was reverted.
151 Description: Update SLES10 kernel to 2.6.16.60-0.42.8 (patchless)
155 Description: Add ldiskfs maxdirsize mount option.
159 Description: Fix mkfs.lustre support for 16TB LUNs.
163 Description: Print mmp_check_interval
164 Details : print mmp_check_interval and make it possible to abort
165 mount operation in case it takes too long.
169 Description: Add cascading_rw to lustre mpi tests.
173 Description: Store SOM data in EA.
174 Details : Also several SOM related cleanups.
178 Description: Add refcount to md_open_data
182 Description: Use refcount on lov_request_set to prevent memory corruption.
186 Description: Statahead: use parent dir's i_mutex for synchronization.
190 Description: loadgen improvements.
194 Description: Enable peer health detection in ksocklnd and ko2iblnd.
198 Description: Control DCACHE_LUSTRE_INVALID flag with MDS_INODE_LOOKCUP lock
202 Description: Fix race on obd_devcie:md_stats
206 Description: Fix SLES 11 client builds.
207 Details : Resolved uninitalized pointer gcc warnings.
211 Description: Allow non-root access for "lfs check".
212 Details : Added a check in obd_class_ioctl() for OBD_IOC_PING_TARGET.
216 Description: Clear imp_force_reconnect correctly in ptlrpc_connect_interpret()
218 Severity : enhancement
220 Description: Update kernel to RHEL5.4 2.6.18-164.6.1.el5 and
221 OEL5 2.6.18-164.6.1.0.1.el5(Both in-kernel OFED enabled).
225 Description: Sanity test_27p fails.
226 Details : OBD_FAIL_OST_ENOSPC has been moved to filter_statfs().
227 Sanity tests 27 were cleaned up and improved.
231 Description: Unify req format on client/servers
232 Details : Use new req_capsule API [almost] everywhere instead of old PTLRPC
233 buffers and swabbers approach..
238 Description: drop join file stuff.
240 Severity : enhancement
242 Description: Add 2.6.27 and SLES11 (patchless client) support
246 Description: Fix in ptlrpc_expire_one_request() to print the signed time
249 Severity : enhancement
251 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
254 Severity : enhancement
256 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
258 Severity : enhancement
260 Description: correctly handle big reply message.
261 Details : send LNet event if reply is bigger then buffer and adjust this buffer
266 Description: Drop unnecessary __GFP_NOMEMALLOC flag from filter_get_page()
268 Severity : enhancement
270 Description: Add support for OEL5.
274 Description: Conf-sanity.sh 50g test - deactivated OST should not cause a panic.
276 Severity : enhancement
278 Description: "weak-modules" support
279 Details : Implement "weak-modules" support which enables kernel modules
280 to be used with any kernel that implements the same kABI. In
281 order to achieve this modules are now installed in
282 /lib/modules/$(uname -r)/updates/kernel on all distributions.
286 Description: client could not reconnect to OST because of an active request.
287 Details : The client now retries to reconnect to the same server, if a
288 connect request failed with EBUSY or -EAGAIN. Test case 35b
289 conf-sanity.sh has been created.
293 Description: Don't allow make backward step on assignin osc next id.
294 Details : race between allocation next id and ll_sync thread can be cause
295 of set wrong osc next id and can be kill valid ost objects.
299 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
303 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
307 Description: Error handling in osc_statfs_interpret() has been improved.
308 Details : Check in osc_statfs_interpret() for EBADR.
312 Description: Do not send statfs() requests to OSTs disabled by administrator.
313 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
315 Severity : enhancement
317 Description: Added sanity-quota test to for lfs(1) to work with numeric uid/gid.
319 Severity : enhancement
321 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
323 Severity : enhancement
325 Description: Enhancement for lfs(1) command to use numeric uid/gid.
329 Description: Improved cleanup procedure for conf-sanity test_46a.
333 Description: Do not start bulk transfer, if client tried to reconnect.
334 Details : Do not start bulk transfer, if client tried to reconnect,
335 but failed due to active rpcs. Check for exp_abort_active_req
336 has been implemented.
340 Description: Prevent inconsistences between linux and lustre mount structures.
341 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
342 Make the sleep interruptible.
344 Severity : enhancement
346 Description: provide server to client comms path
347 Details : server to client communications path via new LDLM_SET_INFO rpc
351 Description: truncate starts GFP_FS allocation under transaction causing deadlock
352 Details : ldiskfs_truncate calls grab_cache_page which may start page
353 allocation under an open transaction. This may lead to
354 calling prune_icache with consequent lustre reentrance.
358 Description: send correctly lsm on open replay
359 Details : MDS is trust to LSM size on replay open, but client can set wrong size
364 Description: enable client interface failover
365 Details : When a child reconnects from another NID, properly update export
366 nid hash position and ldlm reverse import.
368 Severity : enhancement
370 Description: Communicate OST degraded/readonly state via statfs to MDS
371 Details : Flags in the statfs returned from OSTs indicate whether the
372 OST is in a degraded RAID state, or if the filesystem has
373 turned read-only after a filesystem error is detected.
378 Description: don't panic if EPROTO was hit when reading symlink
379 Details : correctly handling request reference in error cases.
381 Severity : enhancement
383 Description: Add LustreNetLink, a kernel-userspace communcation path.
385 Severity : enhancement
387 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
392 Description: access to llog context before init.
393 Details : move handling CATALOGS file at osc layer and forbid access to llog
397 Frequency : rare, only when client is writing to OST on local node
399 Description: Avoid deadlock for local client writes
400 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
401 memory freeing context. This allows OST threads to set the
402 PF_MEMALLOC flag on task structures in order to allocate memory
403 from reserved pools and complete IO.
404 Use GFP_HIGHUSER for OST allocations for non-local client writes,
405 so that the OST threads generate memory pressure and allow
406 inactive pages to be reclaimed.
408 Severity : enhancement
410 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
415 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
416 Details : move ldlm namespace creation in setup phase to avoid grab
417 _lprocfs_lock with cli_sem held.
421 Description: Temporarily disable grant shrink.
422 Details : Disable the feature for debugging.
425 Frequency : only when formatting test filesystems
427 Description: Unable to run several mkfs.lustre on loop devices at the same time
428 Details : mkfs.lustre returns error 256 on the concurrent loop devices
429 formatting. The solution is to proper handle the error.
431 Severity : enhancement
433 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
435 Severity : enhancement
437 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
442 Description: resolve race between obd_disconnect and class_disconnect_exports
443 Details : if obd_disconnect will be called to already disconnected export he
444 forget release one reference and osc module can't unloaded.
446 Severity : enhancement
448 Description: Allow tuning service thread via /proc
449 Details : For each service a new
450 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
451 created that can be used to set min/max thread counts, and get the
452 current number of running threads.
457 Descriptoin: don't return error if have partially created objects for file.
458 Details : lov_update_create_set uses set->set_success as index for created
459 objects, so if some requests will be failed, they will have hole
460 at end of array and qos_shrink_lsm for allocate correct lsm.
462 Severity : enhancement
464 Description: Update OFED support to 1.4.1
469 Description: Reduce small size read RPC
470 Details : Set read-ahead limit for every file and only do read-ahead when
471 available read-ahead pages are bigger than 1M to avoid small size
474 Severity : enhancement
476 Description: move AT tunable parameters for more consistent usage
477 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
479 Severity : enhancement
481 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
482 Details : allow skip disconnected ost for send statfs request and hide error
488 Description: fix racy locking of mballoc block bitmaps causing BUG
489 Details : The locking of the mballoc buddy bitmap and the in-memory
490 block bitmap was using two different spin locks in some
491 cases. This made it possible to incorrectly access the
492 mballoc bitmap while another process was modifying it,
493 causing a sanity assertion to fail. While no on-disk corruption
494 was reported, there was some risk of this happening.
497 Frequency : rare, on llog test 6
499 Descriptoin: don't allow connect to already connected import
500 Details : allowing connect to already connected import is hide connecting problem.
505 Descriptoin: allow kill process which wait statahead result
506 Details : for some reasons 'ls' can stick in waiting result from statahead,
507 in this case need way for kill this process.
509 Severity : enhancement
511 Description: Add state history info file, enhance import info file
512 Details : Track import connection state changes in a new osc/mdc proc file;
513 add overview-type data to the osc/mdc import proc file.
515 Severity : enhancement
517 Description: MDS create should not wait for statfs RPC while holding DLM lock.
519 Severity : enhancement
521 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
526 Descriptoin: OST grant too much space to client even there are not enough space.
527 Details : Client will shrink its grant cache to OST if there are no write
528 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
529 this grant cache if there are already not enough avaible space
530 (left_space < total_clients * 32M).
533 Frequency : start MDS on uncleanly shutdowned MDS device
535 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
536 Details : stay in waiting mds<>ost recovery finished produce random bugs
537 due race between two ll_sync thread for one lov target. send
538 ACTIVATE event only if connect realy finished and import have
542 Frequency : rare, connect and disconnect target at same time
544 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
545 Details : don't call obd_disconnect under lov_lock. this long time
546 operation and can block ptlrpcd which answer to connect request.
551 Descriptoin: don't lose wakeup for imp_recovery_waitq
552 Details : recover_import_no_retry or invalidate_import and import_close can
553 both sleep on imp_recovery_waitq, but we was send only one wakeup
557 Frequency : always with long access acl
559 Descriptoin: mds can't pack reply with long acl.
560 Details : mds don't control size of acl but they limited by reint/getattr
563 Severity : enhancement
565 Description: Update to SLES10 kernel-2.6.16.60-0.33.
567 Severity : enhancement
569 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
572 Frequency : start MDS on uncleanly shutdowned MDS device
574 Descriptoin: aborting recovery hang on MDS
575 Details : don't throttle destroy RPCs for the MDT.
578 Frequency : on remount
580 Description: external journal device not working after the remount
581 Details : clear dev_rdonly flag for external journal devices in
587 Description: shutdown vs evict race
588 Details : client_disconnect_export vs connect request race.
589 if client will evicted at this time - we start invalidate
590 thread without referece to import and import can be freed
594 Frequency : rare, need acl's on inode.
596 Description: client can't handle ost additional correctly
597 Details : if ost was added after client connected to mds client can have
598 hit lnet_try_match_md ... to big messages to wide striped files.
599 in this case need teach client to handle config events about add
600 lov target and update client max ea size at that event.
602 Severity : enhancement
604 Description: Changelogs
605 Details : Changelogs are a lightweight mechanism to track filesystem
606 metadata and namespace changes. The changelog is recorded
607 permanently on the MDTs, and is periodically "consumed" / purged
608 when records are no longer needed.
610 Severity : enhancement
612 Description: compact fld format with extents
613 Details : Store range of seq rather than every seq in FLD. Seq
614 controller update FLD rather than clients. In Case of CMD, mdt0
615 has FLD, all other metadata server act as non persistent proxy
616 for FLD queries and cache fld entries in fld cache.
621 Description: don't skip ost target if they assigned to file
622 Details : Drop slow OSCs if we can, but not for requested start idx.
623 This means "if OSC is slow and it is not the requested
624 start OST, then it can be skipped, otherwise skip it only
625 if it is inactive/recovering/out-of-space.
629 Description: more cleanup in mds_lov
630 Details : not send LOV EA under replay, we can't know about they size at this
631 time. Don't allow client connect to mds before any ost connected,
632 for avoid problems with LOV EA size and returning EIO to client.
634 Severity : enhancement
636 Description: Interoperability at server side (Disk interoperability)
638 Severity : enhancement
640 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
642 Severity : enhancement
644 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
646 Severity : enhancement
648 Description: New client IO stack (CLIO).
650 Severity : enhancement
652 Description: Commit on sharing. Eliminate inter-client dependencies between
653 uncommitted transactions by doing transaction commits.
654 Thereby clients may recovery independently.
657 Frequency : Create a symlink file with a very long name
659 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
660 Details : If there is no extra space in the request for early cancels,
661 ldlm_req_handles_avail() returns 0 instead of a negative value.
663 Severity : enhancement
665 Description: Add /proc entry for import status
666 Details : The mdc, osc, and mgc import directories now have
667 an import directory that contains useful import data for debugging
670 Severity : enhancement
672 Description: Re-disable certain /proc logging
673 Details : Enable and disable client's offset_stats, extents_stats and
674 extents_stats_per_process stats logging on the fly.
677 Frequency : Only on FC kernels 2.6.22+
679 Description: oops in statahead
680 Details : Do not drop reference count for the dentry from VFS when lookup,
681 VFS will do that by itself.
683 Severity : enhancement
685 Description: Generic /proc file permissions
686 Details : Set /Proc file permissions in a more generic way to enable non-
687 root users operate on some /proc files.
691 Description: Hitting mdc_commit_close() ASSERTION
692 Details : Properly handle request reference release in
693 ll_release_openhandle().
697 Description: quota recovery deadlock during mds failover
698 Details : This patch includes att18982, att18236, att18237 in bz14840.
700 1. fix osts hang when mds does failover with quotaon
701 2. prevent watchdog storm when osts threads wait for the
706 Frequency : only patchless client
707 Description: add workaround for race between add/remove dentry from hash
709 Severity : enhancement
711 Description: Allow OST glimpses to return PW locks
715 Description: LBUG when llog conf file is full
716 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
720 Description: Prevent import from entering FULL state when server in recovery
724 Description: service mount cannot take device name with ":"
725 Details : Only when device name contains ":/" will mount treat it as
731 Description: replace ptlrpcd with the statahead thread to interpret the async
732 statahead RPC callback
736 Frequency : on recovery
737 Description: I/O failures after umount during fail back
738 Details : if client reconnected to restarted server we need join to recovery
739 instead of find server handler is changed and process self
740 eviction with cancel all locks.
742 Severity : enhancement
744 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
746 Severity : enhancement
748 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
750 Severity : enhancement
752 Description: Upcall on Lustre log has been dumped
753 Details : Allow for a user mode script to be called once a Lustre log has
754 been dumped. It passes the filename of the dumped log to the
755 script, the location of the script can be specified via
756 /proc/sys/lnet/debug_log_upcall.
761 Description: avoid idr_remove called for id which is not allocated.
762 Details : Move assigment s_dev for clustered nfs to end of initialization,
763 to avoid problem with error handling.
768 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
769 Details : When connection is reused this not moved from CONN_UNUSED_HASH
770 into CONN_USED_HASH and this prodice warning when put connection
771 again in unused hash.
774 Severity : enhancement
776 Description: File striping can now be set to use an arbitrary pool of OSTs.
778 Severity : enhancement
780 Description: Export bytes_read/bytes_write count on OSC/OST.
784 Description: Early reply size mismatch, MGC loses connection
785 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
786 the connect flags are properly negotiated.
791 Description: quotas are not honored with O_DIRECT
792 Details : all writes with the flag O_DIRECT will use grants which leads to
793 this problem. Now using OBD_BRW_SYNC to guard this.
797 Description: add quota statistics
798 Details : 1. sort out quota proc entries and proc code.
799 2. add quota statistics
801 Severity : enhancement
803 Description: enable quota support for HEAD.
807 Description: Properly propagate oinfo flags from lov to osc for statfs
808 Details : restore missing copy oi_flags to lov requests.
810 Severity : enhancement
812 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
813 : l_getgroups(8), lst(8), routerstat(8)
814 Details : included man pages for llobdstat(8), llstat(8),
815 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
817 Severity : enhancement
819 Description: configure's --enable-quota should check the
820 : kernel .config for CONFIG_QUOTA
821 Details : configure is terminated if --enable-quota is passed but
822 : no quota support is in kernel
826 Description: Remove portals compatibility
827 Details : Remove portals compatibility, not interoperable with releases
832 Description: Resolve device initialization race
833 Details : Prevent proc handler from accessing devices added to the
834 obd_devs array but yet be intialized.
836 Severity : enhancement
838 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
840 Severity : enhancement
842 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
847 Description: Using wrong pointer in osc_brw_prep_request
848 Details : Access to array[-1] can produce panic if kernel compiled with
849 CONFIG_PAGE_ALLOC enabled
853 Description: Client runs out of low memory
854 Details : Consider only lowmem when counting initial number of llap pages
858 Description: *optional* service tags registration
859 Details : if the "service tags" package is installed on a Lustre node
860 When the filesystem is mounted, a local-node service tag will
861 be created. See http://inventory.sun.com/ for more information
862 about the Service Tags asset management system.
866 Description: Kernel BUG tries to release flock
867 Details : Lustre does not destroy flock lock before last reference goes
868 away. So always drop flock locks when client is evicted and
869 perform unlock regardless of successfulness of speaking to MDS.
873 Description: add refcount protection for osc callbacks, avoid panic on shutdown
877 Description: sanity test 65a fails if stripecount of -1 is set
878 Details : handle -1 striping on filesystem in ll_dirstripe_verify
883 Description: ASSERTION(CheckWriteback(page,cmd)) failed
884 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
887 Severity : enhancement
889 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
893 Description: do not process already freed flock
894 Details : flock can possibly be freed by another thread before it reaches
895 to ldlm_flock_completion_ast.
899 Description: LBUG during stress test
900 Details : Need properly lock accesses the flock deadlock detection list.
904 Description: oops in page fault handler
905 Details : kernel page fault handler can return two special 'pages' in error
906 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
910 Description: timeout with invalidate import.
911 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
912 should be handled by ptlrpcd. This produce long age waiting and
913 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
915 Severity : enhancement
917 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
921 Description: enable MGS and MDT services start separately
922 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
923 a co-located MGS without starting the MGS, which is a complement
924 to 'nosvc' mount option.
928 Frequency : after recovery
929 Description: precreate to many object's after del orphan.
930 Details : del orphan st in oscc last_id == next_id and this triger growing
931 count of precreated objects. Set flag LOW to skip increase count
932 of precreated objects.
936 Frequency : rare, on clear nid stats
937 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
938 Details : when clean nid stats sometimes try destroy live entry,
939 and this produce panic in free.
943 Description: Stack overflow during MDS log replay
944 ease stack pressure by using a thread dealing llog_process.
948 Description: wait until IO finished before start new when do lock cancel.
949 Details : VM protocol want old IO finished before start new, in this case
950 need wait until PG_writeback is cleared until check dirty flag and
951 call writepages in lock cancel callback.
953 Severity : enhancement
955 Description: using special macro for print time and cleanup in includes.
959 Description: mds_mfd_close() ASSERTION(rc == 0)
960 Details : In mds_mfd_close(), we need protect inode's writecount change
961 within its orphan write semaphore to prevent possible races.
965 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
969 Frequency : rare, on shutdown ost
970 Description: don't hit live lock with umount ost.
971 Details : shrink_dcache_parent can be in long loop with destroy dentries,
972 use shrink_dcache_sb instead.
976 Description: don't panic with use echo client
977 Details : echo client pass NULL as client nid pointer and this produce null
982 Description: fix build on ppc32
983 Details : compile code with -m64 flag produce wrong object file for ppc32.
987 Description: add message levels for liblustreapi
991 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
992 Details : Move the 'good_osts' check before the 'total_bavail' check. This
993 will result in an -EAGAIN and in the exit call path we call
994 alloc_rr() which will with increasing aggressiveness attempt to
995 aquire precreated objects on the minimum number of required OSCs.
999 Description: Use old size assignment to avoid deadlock
1000 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
1001 the scheduling while holding a spinlock. We do not need locking
1002 for size in ll_update_inode() because size is only updated from
1003 the MDS for directories or files without objects, so there is no
1004 other place to do the update, and concurrent access to such inodes
1005 are protected by the inode lock.
1009 Description: resolve "_IOWR redefined" build error on SLES10
1013 Description: dump the memory debugging after all modules are unloaded to
1014 suppress false negative in conf_sanity test 39
1016 Severity : enhancement
1018 Description: build kernel-ib packages for OFED 1.3 in our release cycle
1023 Description: fix SLES kernel versioning
1024 Details : the kernel version for our SLES 10 kernel did not include a "-"
1025 before the "smp" at the end. while this was not a problem in
1026 general, it did mean that software trying to use the kernel
1027 version to try to detect a vendor specific kernel would fail.
1028 this was most evident by the OFED build scripts.
1032 Description: Don't update lov_desc members until making sure they are valid
1033 Details : When updating lov_desc members via proc fs, need fix their
1034 validities before doing the real update.
1038 Description: don't put request into delay list while invalidate in flight.
1039 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
1040 import in flight. this produce timeout for invalidate and sometimes
1041 can cause stale data.
1045 Frequency : on ppc only
1046 Description: not convert ost objects for directory because it's not exist.
1047 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
1051 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
1053 Severity : enhancement
1055 Description: Update to support 2.6.22.14 vanilla kernel.
1059 Frequency : rare, on recovery
1060 Description: read procfs can produce deadlock in some situation
1061 Details : Holding lprocfs lock which send rpc can produce block for destroy
1062 obd objects and this also block reconnect with -EALREADY. This
1063 isn't fix all lprocfs bugs - but make it rare.
1065 Severity : enhancement
1067 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
1070 Frequency : frequent on X2 node
1072 Description: mdc_set_open_replay_data LBUG
1073 Details : Set replay data for requests that are eligible for replay.
1077 Description: lustre_mgs: operation 101 on unconnected MGS
1078 Details : When MGC is disconnected from MGS long enough, MGS will evict the
1079 MGC, and late on MGC cannot successfully connect to MGS and a lot
1080 of the error messages complaining that MGS is not connected.
1084 Frequency : on network error
1085 Description: panic with double free request if network error
1086 Details : mdc_finish_enqueue is finish request if any network error ocuring,
1087 but it's true only for synchronus enqueue, for async enqueue
1088 (via ptlrpcd) this incorrect and ptlrpcd want finish request
1091 Severity : enhancement
1093 Description: client-side metadata stat-ahead during readdir(directory readahead)
1094 Details : perform client-side metadata stat-ahead when the client detects
1095 readdir and sequential stat of dir entries therein
1098 Frequency : on start mds
1100 Description: Implement get_info(last_id) in obdfilter.
1103 Frequency : occasional
1105 Description: Correctly check stale fid, not start epoch if ost not support SOM
1106 Details : open with flag O_CREATE need set old fid in op_fid3 because
1107 op_fid2 was overwritten with new generated fid, but mds can answer
1108 with one of these two fids and both is not stale. Setattr
1109 incorrectly started an epoch and assume will be called
1110 done_writing, but without SOM done_writing ever being called.
1113 Frequency : rare, depends on device drivers and load
1115 Description: MDS or OSS nodes crash due to stack overflow
1116 Details : Code changes in 1.8.0 increased the stack usage of some functions.
1117 In some cases, in conjunction with device drivers that use a lot
1118 of stack the MDS (or possibly OSS) service threads could overflow
1119 the stack. One change which was identified to consume additional
1120 stack has been reworked to avoid the extra stack usage.
1123 Frequency : occasional
1125 Description: Do not fail import if osc_interpret_create gets -EAGAIN
1126 Details : If osc_interpret_create got -EAGAIN it immediately exits and
1127 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
1128 oscc_has_objects and see OSC has no objests and call
1129 oscc_internal_create to resend create request.
1131 Severity : enhancement
1133 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
1135 Severity : enhancement
1137 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
1140 Frequency : very rare
1142 Description: Fix replay if there is an un-replied request and open
1143 Details : In some cases, older replay request will revert the
1144 mcd->mcd_last_xid on MDS which is used to record the client's
1145 latest sent request.
1147 Severity : enhancement
1149 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
1151 Severity : enhancement
1153 Description: Add rhel5 support to HEAD.
1155 Serverity : enhancement
1157 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
1162 Description: Don't allow skipping OSTs if index has been specified.
1163 Details : Don't allow skipping OSTs if index has been specified, make
1164 locking in internal create lots better.
1168 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
1169 Details : access to bitfield in structure is always rounded to long
1170 and this produce problem with not atomic change any bit.
1174 Description: Lustre make rpms failed.
1175 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
1176 builds Lustre rpms from tarball.
1179 Frequency : rare on shutdown ost
1181 Description: If llog cancel was not send before clean_exports phase, this can
1182 produce deadlock in llog code.
1183 Details : If llog thread has last reference to obd and call class_import_put
1184 this produce deadlock because llog_cleanup_commit_master wait when
1185 last llog_commit_thread exited, but this never success because was
1186 called from llog_commit_thread.
1190 Description: allow userland application know is lost one of stripes.
1191 Details : fill lvb_blocks with error code on ost and return it to
1192 application if error flag found.
1196 Description: NULL lov_tgts causing MDS oops
1197 Details : more safe checks for NULL lov_tgts for avoid oops.
1199 Severity : enhancement
1201 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
1205 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
1206 Details : If an OST is down the MDS will hang indefinitely in
1207 obd_statfs_rqset() waiting for the statfs data. While for
1208 MDS QOS usage of statfs, it should not stuck in waiting.
1210 Severity : enhancement
1212 Description: remote_acl support
1213 Details : Support ACL-based permission check for remote user.
1214 Support setfacl/getfacl for remote user with the utils
1215 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
1216 the system "{s,g}etfacl" utils.
1218 Severity : enhancement
1220 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
1222 Severity : enhancement
1224 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
1228 Description: make mgs_setparam() handle fsname containing dash
1229 Details : fsname containing a dash does not work with lctl conf_param
1231 Severity : enhancement
1233 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
1235 Severity : enhancement
1237 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
1241 Frequency : rare, at shutdown
1242 Description: access already free / zero obd_namespace.
1243 Details : if client_disconnect_export was called without force flag set,
1244 and exist connect request in flight, this can produce access to
1245 NULL pointer (or already free pointer) when connect_interpret
1246 store ocd flags in obd_namespace.
1250 Frequency : only at startup
1251 Description: not alloc memory with spinlock held.
1252 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
1253 if any spinlock held.
1255 Severity : enhancement
1257 Description: make lustre randomly fail allocating memory
1258 Details : Make lustre randomly failed allocating memory for testing purpose.
1260 Severity : enhancement
1262 Description: lost problems with lov objid file
1263 Details : Fixes some scability and access to not inited memory problems
1264 in work with lov objdid file.
1269 Description: lfs find does not continue on file error
1270 Details : Continue other files processing when a file/dir is absent.
1274 Description: Inconsistent usage of lustre_pack_reply()
1275 Details : Standardize the usage of lustre_pack_reply() such that it
1276 always generate a CERROR on failure.
1279 Frequency : occasional
1281 Description: MDS hang or stay in waiting lock
1282 Details : If client receive lock with CBPENDING flag ldlm need send lock
1283 cancel as separate rpc, to avoid situation when cancel request
1284 can't processed due all i/o threads stay in wait lock.
1288 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1292 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1296 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1300 Description: Update to RHEL4 latest kernel.
1302 Severity : enhancement
1304 Description: Build SLES10 patchless client fails
1305 Details : The configure was broken by run ./configure with
1306 --with-linux-obj=.... argument for patchless client. When the
1307 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1308 can't search header adequately. Use absolute path such as
1309 -I($LINUX)/include instead.
1313 Description: interrupt oig_wait produce painc on resend.
1314 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1315 private set, and this produce situation when rq_ptlrpcd_data not
1316 copyed to new allocated request and triggered LBUG on assert
1317 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1318 ptlrpcd set if request is changed and can be safety changed to use
1323 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1324 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1325 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1326 these macros and replace them with their difinitions instead.
1330 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1331 Details : Update which_patch & target file for SLES10 latest kernel.
1334 Description: Remove client patches from SLES 10 kernel.
1335 Details : This causes SLES 10 clients to behave as patchless clients
1336 even on a Lustre-patched (server) kernel.
1338 Severity : enhancement
1340 Description: self-adjustable client's lru lists
1341 Details : use adaptive algorithm for managing client cached locks lru
1342 lists according to current server load, other client's work
1343 pattern, memory activities, etc. Both, server and client
1344 side namespaces provide number of proc tunables for controlling
1347 Severity : enhancement
1349 Description: light-weight GSS support
1350 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
1351 and reduce performance overhead.
1353 Severity : enhancement
1355 Description: Linux keyring support
1356 Details : Support using service of Linux keyring for Lustre GSS internal
1357 context refresh/cache mechanism.
1361 Description: Fix errors in lfs documentation
1362 Details : Fixes man pages
1366 Description: don't use GFP_* in generic Lustre code.
1367 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
1371 Description: obdclass is limited by single OBD_ALLOC(idarray)
1372 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
1376 Description: warnings with build patchless client with vanila 2.6.19 and up
1377 Details : change old ctl_table style and replace ctl_table/ctl_table_header
1378 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
1382 Frequency : Only for SLES
1383 Description: sanity_quota fail test_1
1384 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
1385 file, which makes TSTID[2] inunique.
1389 Frequency : Only for SLES9
1390 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
1391 Details : Update target/ChangeLog/which_patch .
1395 Frequency : Only for test_10 in sanity-quota.sh
1396 Description: a bug in quota test code
1397 Details : checking a test flag in a wrong place
1401 Frequency : Only for test_13 in sanity-quota.sh
1402 Description: a bug in quota test script
1403 Details : shouldn't check overfull allocated quota
1407 Description: jbd statistics
1408 Details : Port older jbd statistics patch for sles10
1410 Severity : enhancement
1412 Description: Add printing inode info into message about error in writepage.
1417 Description: reply_lock_interpret crash due to race with it and lock cancel.
1418 Details : Do not replay locks that are being cancelled. Do not reference
1419 locks by their address during replay, just by their handle.
1423 Frequency : When flocks are used.
1424 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
1425 Details : Flock locks might destroy just granted lock if it could be merged
1426 with another existing flock, this is done in completion handler,
1427 so teach ldlm_cli_enquque_fini that this is a valid case for
1433 Description: Oops in read and write path when failing to allocate lock.
1434 Details : Check if lock allocation failed and return error back.
1438 Description: lstripe command fails for valid OST index
1439 Details : The stripe offset is compared to lov->desc.ld_tgt_count
1440 instead of lov->desc.ld_active_tgt_count.
1444 Description: sanity.sh failed test 103
1445 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
1450 Description: df doesn't work properly if diskfs blocksize != 4K
1451 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
1455 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
1456 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
1458 Severity : enhancement
1460 Description: Adaptive timeouts
1461 Details : RPC timeouts adapt to changing server load and network
1462 conditions to reduce resend attempts and improve recovery time.
1466 Description: llapi_file_create() does not allow some changes
1467 Details : add llapi_file_open() that allows specifying the mode and
1468 open flags, and also returns an open file handle.
1472 Description: merge and cleanup kernel patches.
1473 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
1477 Description: Add journal checksum support.(Kernel part)
1478 Details : The journal checksum feature adds two new flags i.e
1479 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
1480 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
1481 indicates that the commit block contains the checksum for
1482 the blocks described by the descriptor blocks. Now commit
1483 record can be sent to disk without waiting for descriptor
1484 blocks to be written to disk. This behavior is controlled
1485 using JBD2_FEATURE_ASYNC_COMMIT flag.
1489 Description: OSS needs mutliple precreate threads
1490 Details : Add ability to start more than one create thread per OSS.
1494 Description: Sanity.sh test_65k failed: setstripe should have succeeded
1495 Details : execute lfs setstripe on client
1499 Description: mds_obd_create error creating tmp object
1500 Details : When the user sets quota on root, llog will be affected and can't
1501 create files and write files.
1504 Frequency : Always on ia64 patchless client, and possibly others.
1506 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
1507 Details : This allows the patchless client to be loaded on architectures
1508 without this export.
1512 Description: RedHat Update kernel for RHEL5
1513 Details : Add the kernel config file for RHEL5.
1517 Description: RedHat Update kernel for RHEL5
1518 Details : Modify the kernel config file more closer RHEL5.
1522 Description: Build failure against Centos5 (RHEL5)
1523 Details : Define PAGE_SIZE when it isn't present.
1527 Description: test_7 on sanity-quota.sh can't be used on separate machines
1528 Details : add facet to handle it
1532 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
1533 Details : with NFS, the anon dentry's parent was set to itself in
1534 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
1535 corresponding dentry other than use rec->ur_name.
1537 Severity : enhancement
1539 Description: lfs setstripe enhancement
1540 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
1544 Description: enable data checksumming by default
1545 Details : enable checksum by default, allow --disable-checksum
1546 configure option and "-o nochecksum" mount option. Checksums
1547 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
1551 Description: lustre support for RHEL5
1552 Details : Add support for RHEL5.
1556 Description: Client eviction due to failover config
1557 Details : after a connection loss, the lustre client should attempt to
1558 reconnect to the last active server first before trying the
1559 other potential connections.
1563 Description: when mds and osts use different quota unit(32bit and 64bit),
1564 quota will be released repeatly.
1565 Details : void sending multiple quota reqs to mds, which will keep the
1566 status between the reqs.
1570 Description: osts not allocated evenly to files
1571 Details : change the condition to increase offset_idx
1574 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
1576 Description: Data corruption for OSTs that are formatted larger than 2TB
1578 Details : When generating the bio request for lustre file writes the
1579 sector number would overflow a temporary variable before being
1580 used for the IO. The data reads correctly from Lustre (which
1581 will overflow in a similar manner) but other file data or
1582 filesystem metadata may be corrupted in some cases.
1586 Description: Tune the kernel for good SCSI performance.
1587 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
1588 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
1593 Description: rewrite ext2-derived code in obdclass/uuid.c
1594 Details : rewrite inherited code (uuid parsing code from ext2 utils)
1595 from scratch preserving functionality.
1599 Description: "lfs find -obd UUID" prints directories
1600 Details : "lfs find -obd UUID" will return all directory names instead
1601 of just file names. It is incorrect because the directories
1602 do not reside on the OSTs.
1606 Description: use i_size_read and i_size_write in 2.6 port
1607 Details : replace inode->i_size access with i_size_read/write()
1610 Frequency : when removing large files
1612 Description: scheduling issue during removal of large Lustre files
1613 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
1614 It causes scheduling issues when removing large files (17TB in the
1618 Frequency : only with liblustre clients on XT3
1620 Description: evictions taking too long
1621 Details : allow llrd to evict clients directly on OSTs
1624 Frequency : only on ppc
1626 Description: /proc/fs/lustre/devices broken on ppc
1627 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
1628 We should make sure the type of 'index' is loff_t and then cast
1629 explicitly as needed below. Do not assign an explicitly cast
1633 Frequency : only for rhel5
1635 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
1636 Details : Modify the target file & which_kernel.
1638 Severity : enhancement
1640 Description: omit set fsid for export NFS
1641 Details : fix set/restore device id for avoid EMFILE error and mark lustre
1642 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
1646 Frequency : Always, for kernels after 2.6.16
1647 Description: Fix warning idr_remove called for id=.. which is not allocated.
1648 Details : Last kernels save old s_dev before kill super and not allow
1649 to restore from callback - restore it before call kill_anon_super.
1653 Description: buffer overruns could theoretically occur
1654 Details : llapi_semantic_traverse() modifies the "path" argument by
1655 appending values to the end of the origin string, and a
1656 overrun may occur. Adding buffer overrun check in liblustreapi.
1660 Description: Fix error on 'ls .' at the top of the Lustre mount.
1661 Details : Don't revalidate dentry if it is a root dentry.
1665 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
1666 Details : Modify vm-tunables-rhel4.patch.
1670 Description: Kernel config for 2.6.18-vanilla.
1671 Details : Modify targets/2.6-vanilla.target.in.
1672 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
1673 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
1674 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
1675 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
1679 Description: improve handling recoverable errors
1680 Details : if request processig with error which can be recoverable on server
1681 request should be resend, otherwise page released from cache and
1686 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
1687 Details : Modify target file & which_patch.
1688 A flaw was found in the IA32 system call emulation provided
1689 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
1690 value could be stored in the %RAX register, which could trigger an
1691 out-of-bounds system call table access. An untrusted local user
1692 could exploit this flaw to run code in the kernel
1693 (ie a root privilege escalation). (CVE-2007-4573).
1697 Description: O_DIRECT bypasses client statistics.
1698 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
1699 read_ahead_stats were updated. copied the stats section from
1700 osc_send_oap_rpc() into async_internal().
1704 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
1708 Description: change order of libsysio includes
1709 Details : '#include sysio.h' should always come before '#include xtio.h'
1713 Description: handle "serious error: objid * already exists" more gracefully
1714 Details : If LAST_ID value on disk is smaller than the objects existing in
1715 the O/0/d* directories, it indicates disk corruption and causes an
1716 LBUG(). If the object is 0-length, then we should use the existing
1717 object. This will help to avoid a full fsck in most cases.
1719 Severity : enhancement
1721 Description: adapt the lustre_config script to support the upgrade case
1722 Details : Add "-u" option for lustre_config script to support upgrading 1.4
1723 server targets to 1.6 in parallel.
1727 Description: To avoid grant space > avaible space when the disk is almost
1728 full. Without this patch you might see the error "grant XXXX >
1729 available" or some LBUG about grant, when the disk is almost
1731 Details : In filter_check_grant, for non_grant cache write, we should
1732 check the left space by if (*left > ungranted + bytes), instead
1733 of (*left > ungranted), because only we are sure the left space
1734 is enough for another "bytes", then the ungrant space should be
1735 increase. In client, we should update cl_avail_grant only there
1736 is OBD_MD_FLGRANT in the reply.
1740 Description: Update RHEL 4 kernel to fix local root privilege escalation.
1741 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
1742 described in CVE-2007-4573. This problem could allow untrusted
1743 local users to gain root access.
1746 Frequency : when using O_DIRECT and quotas
1748 Description: Incorrect file ownership on O_DIRECT output files
1749 Details : block usage reported by 'lfs quota' does not take into account
1750 files that have been written with O_DIRECT.
1755 Description: touch file failed when fs is not full
1756 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
1757 otherwise we can get ENOSP while fs is not full.
1761 Description: parallel lock callbacks
1762 Details : Instead of sending blocking and completion callbacks as separated
1763 requests, adding them to a set and sending in parallel.
1766 Frequency : only for Cray XT3
1767 Bugzilla : 12829/13455
1768 Description: Changing primary group doesn't change the group lustre assigns to
1770 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
1771 overridden with the primary group provided by the group upcall,
1772 whereas the supplied fsgid can be trusted if it is in the list of
1773 supplementary groups returned by the group upcall.
1775 Severity : enhancement
1777 Description: Allow masking D_WARNING, D_ERROR messages from console
1778 Details : Console messages can now be disabled via lnet.printk.
1782 Description: User code with malformed file open parameter crashes client node
1783 Details : Before packing join_file req, all the related reference should be
1784 checked carefully in case some malformed flags cause fake
1785 join_file req on client.
1789 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
1790 Details : ldlm_completion_ast() assumes that a lock is granted when the req
1791 mode is equal to the granted mode. However, it should also check
1792 that LDLM_FL_CP_REQD is not set.
1796 Description: Heavy nfs access might result in deadlocks
1797 Details : After ELC code landed, it is now improper to enqueue any mds
1798 locks under och_sem, because enqueue might want to decide to
1799 cancel open locks for same inode we are holding och_sem for.
1803 Description: Client eviction while running blogbench
1804 Details : A lot of unlink operations with concurrent I/O can lead to a
1805 deadlock causing evictions. To address the problem, the number of
1806 oustanding OST_DESTROY requests is now throttled to
1807 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
1808 callbacks are processed in priority.
1812 Description: enable ACLs on MDS by default
1813 Details : ACLs must be enabled on MDS by default.
1816 Frequency : PPC/PPC64 only
1818 Description: conflicts between asm-ppc64/types.h and lustre_types.h
1819 Details : fix duplicated definitions between asm-ppc64/types.h and
1820 lustre_types.h on PPC.
1823 Frequency : PPC/PPC64 only
1825 Description: asm-ppc/segment.h does not exist
1826 Details : fix compile issue on PPC.
1830 Description: better handle error messages in extents code
1833 Frequency : RHEL4 only
1835 Description: mkfs is very slow on IA64/RHEL4
1836 Details : A performance regression has been discovered in the MPT Fusion
1837 driver between versions 3.02.73rh and 3.02.99.00rh. As a
1838 consequence, we have downgraded the MPT Fusion driver in the RHEL4
1839 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
1841 Severity : enhancement
1843 Description: SNMP support enhancement
1844 Details : Adding total number of sampled request for an MDS node in snmp
1847 Severity : enhancement
1849 Description: Optimize ldlm waiting list processing for PR extent locks
1850 Details : When processing waiting list for read extent lock and meeting read
1851 lock that is same or wider to it that is not contended, skip
1852 processing rest of the list and immediatelly return current
1853 status of conflictness, since we are guaranteed there are no
1854 conflicting locks in the rest of the list.
1858 Description: Time out and refuse to reconnect
1859 Details : When the failover node is the primary node, it is possible
1860 to have two identical connections in imp_conn_list. We must
1861 compare not conn's pointers but NIDs, otherwise we can defeat
1862 connection throttling.
1866 Description: port llog fixes from b1_6 into HEAD
1867 Details : Port llog reference couting and some llog cleanups from b1_6
1868 (bug 10800) into HEAD, for protect from panic and access to already
1869 free llog structures.
1873 Description: Detect stride IO mode in read-ahead
1874 Details : When a client does stride read, read-ahead should detect that and
1875 read-ahead pages according to the detected stride pattern.
1879 Description: data checksumming impacts single node performance
1880 Details : add support for several checksum algorithm. Currently, only CRC32
1881 and Adler-32 are supported. The checksum type can be changed on
1882 the fly via /proc/fs/lustre/osc/*/checksum_type.
1886 Description: use adler32 for page checksums
1887 Details : when available, use the Adler-32 algorithm instead of CRC32 for
1892 Description: build for x2 fails
1893 Details : fix compile issue on Cray systems.
1897 Description: Properly match for duplicate locks
1898 Details : Due to different lock order from skiplists code, we need to
1899 traverse entire list for now
1902 Frequency : only on PPC/SLES10
1904 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
1905 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
1906 be defined. Add a hack in mkfs_lustre.c to work around this compile
1911 Description: LASSERT on MDS when client holding flock lock dies
1912 Details : ldlm pool logic depends on number of granted locks equal to
1913 number of released locks which is not true for flock locks, so
1914 just exclude such locks from consideration.
1918 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
1919 Details : Use fsfilt_sync() for both the whole filesystem sync and
1920 individual file sync to eliminate dangerous inode locking
1921 with I_LOCK that can lead to a deadlock.
1925 Description: performance in 1.6.3
1926 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
1927 sure that 1MB requests are not fragmented by the block layer.
1929 Severity : enhancement
1931 Description: organize the server-side client stats on per-nid basis
1932 Details : Change the structure of stats under obdfilter and mds to
1940 The "uuid"s file would list the uuids of _active_ exports.
1941 And the clear entry is to clear all stats and stale nids.
1943 Severity : enhancement
1945 Description: eliminate client locks in face of contention
1946 Details : file contention detection and lockless i/o implementation
1947 for contended files.
1951 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
1954 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
1956 Description: MDS may lose file striping (and hence file data) in some cases
1957 Details : If there are additional extended attributes stored on the MDS,
1958 in particular ACLs, SELinux, or user attributes (if user_xattr
1959 is specified for the client mount options) then there is a risk
1960 of attribute loss. Additionally, the Lustre file striping
1961 needs to be larger than default (e.g. striped over all OSTs),
1962 and an additional attribute must be stored initially in the
1963 inode and then increase in size enough to be moved to the
1964 external attribute block (e.g. ACL growing in size) for file
1969 Description: skiplist implementation simplification
1970 Details : skiplists are used to group compatible locks on granted list
1971 that was implemented as tracking first and last lock of each
1972 lock group the patch changes that to using doubly linked lists
1976 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
1977 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
1978 because LASSERT on both the data supplied by a client, and the
1979 data on disk is dangerous and incorrect.
1981 Severity : enhancement
1983 Description: Slow truncate/writes to huge files at high offsets.
1984 Details : Directly associate cached pages to lock that protect those pages,
1985 this allows us to quickly find what pages to write and remove
1986 once lock callback is received.
1990 Description: more ldlm soft lockups
1991 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
1992 starve other threads from the resource lock for a long time in
1993 case of long waiting queue, so change the debug level from
1994 D_OTHER to the less frequently used D_INFO.
1996 Severity : enhancement
1998 Description: add -gid, -group, -uid, -user options to lfs find
2002 Description: Hung threads in invalidate_inode_pages2_range
2003 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
2004 once one is completed. As a result, some RPCs are stuck in the
2005 queue and are never sent.
2009 Description: filter threads hungs on waiting journal commit
2010 Details : Cleanup filter group llog code, then only filter group llog will
2011 be only created in the MDS/OST syncing process.
2015 Description: Procfs and llog threads access destoryed import sometimes.
2016 Details : Sync the import destoryed process with procfs and llog threads by
2017 the import refcount and semaphore.
2019 Severity : enhancement
2021 Description: openlock cache of b1_6 port to HEAD
2026 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
2027 Details : If insertion of an extent fails, then discard the inode
2028 preallocation and free data blocks else it can lead to duplicate
2033 Description: don't always update ctime in ext3_xattr_set_handle()
2034 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
2035 In some cases the ctime should not be updated, for example for
2036 2.0->1.8 compatibility it is necessary to delete an xattr and it
2037 should not update the ctime.
2041 Bugzilla : 15713/16362
2042 Description: Assertion in iopen_connect_dentry in 1.6.3
2043 Details : looking up an inode via iopen with the wrong generation number can
2044 populate the dcache with a disconneced dentry while the inode
2045 number is in the process of being reallocated. This causes an
2046 assertion failure in iopen since the inode's dentry list contains
2047 both a connected and disconnected dentry.
2051 Description: assertion failure in ldlm_handle2lock()
2052 Details : fix a race between class_handle_unhash() and class_handle2object()
2053 introduced in lustre 1.6.5 by bug 13622.
2058 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
2059 Details : remove the limit of 256 scsi disks in the sd_iostat patch
2064 Description: oops in sd_iostats_seq_show()
2065 Details : unloading/reloading the scsi low level driver triggers a kernel
2066 bug when trying to access the sd iostat file.
2071 Description: Kernel panics during QLogic driver reload
2072 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
2073 patch, causing memory corruption.
2078 Description: journal_dev option does not work in b1_6
2079 Details : pass mount option during pre-mount.
2081 Severity : enhancement
2083 Description: Add a FIEMAP(FIle Extent MAP) ioctl
2084 Details : FIEMAP ioctl will allow an application to efficiently fetch the
2085 extent information of a file. It can be used to map logical blocks
2086 in a file to physical blocks in the block device.
2090 Description: LDLM soft lockups - improvement
2091 Details : It is be possible to send the lock handle along with each read
2092 or write request because the client is already doing a lock match
2093 itself so there isn't any reason the OST should have to re-do that
2097 Frequency : only on Cray X2
2099 Description: X2 build failures
2100 Details : fix build failures on Cray X2.
2104 Description: xid & resent requests
2105 Details : Initialize RPC XID from clock at startup (randomly if clock is
2108 Severity : enhancement
2110 Description: Add lustre_start utility to start or stop multiple Lustre servers
2115 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
2116 Details : In case of memory pressure, list_del() can be called twice on
2117 req->rq_history_list, causing a kernel oops.
2121 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
2122 Details : fix stack overflow in the distributed lock manager by defering
2123 export eviction after a failed AST to the elt thread instead of
2124 handling it in the dlm interpret routine.
2128 Description: Convert some comments to new format.
2129 Details : Update documenting comments to match doxygen conventions.
2133 Description: Grammar fixes.
2134 Details : A couple of trivial spelling fixes.
2138 Description: OSD_COUNTERS-mandatory
2139 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
2140 to simplify the code.
2144 Description: simplify cmm_device freeing logic.
2145 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
2149 Description: Add lockdep support to dt_object_operations locking interface.
2150 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
2151 parameter indicating lock ordering. Update mdd code to use new
2156 Description: Introduce failloc constants for lockless IO tests.
2157 Details : Add two new failloc constants to test lockless IO. Only one of
2158 them in implemented---another is checked in yet to be landed
2163 Description: Add lockdep support for inode mutex.
2164 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
2165 in the situations where Lustre has to lock more than one inode
2170 Description: Add optional invariants checking support.
2171 Details : Add new LINVRNT() macro, optional on new --enable-invariants
2172 configure switch. This macro is to be used for consistency and
2173 sanity checks that are too expensive to be left in `production'
2178 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
2179 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
2184 Description: Add lockdep support for ldlm_lock and ldlm_resource.
2185 Details : Use spin_lock_nested() in (the only) situation where more than
2186 one ldlm_lock is locked simultaneously. Also, fix possible
2187 dead-lock in ldlm_lock_change_resource() by enforcing particular
2192 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
2193 Details : Instead of specifying each ldlm_lock call-back through separate
2194 parameter, wrap them into struct ldlm_callback_suite.
2198 Description: Kill join_lru obd method and its callers.
2199 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
2200 in memory---a requirement implemented through ->o_join_lru() obd
2201 method in HEAD. Remove this method and its users.
2205 Description: Add asynchronous ldlm ENQUEUE completion handler.
2206 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
2207 case that stock ldlm_completion_ast() cannot handle as it waits
2208 until lock is granted. Introduce new ldlm_completion_ast_async()
2209 for this. Also comment ldlm_completion_ast().
2213 Description: ldlm_error <-> errno conversion.
2214 Details : Add functions to map (rather arbitrary) between LDLM error codes
2215 and standard errno values. CLIO needs this to prevent LDLM specific
2216 constants from escaping ldlm and osc.
2220 Description: Kill unused ldlm_handle2lock_ns() function.
2221 Details : Kill unused ldlm_handle2lock_ns() function.
2225 Description: Add lu_ref support to ldlm_lock
2226 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
2227 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
2228 and ldlm_resource. LDLM interface has to be changed, because code that
2229 releases a reference on a lock, has to "know" what reference this is.
2230 In the most frequent case
2232 lock = ldlm_handle2lock(handle);
2234 LDLM_LOCK_PUT(lock);
2236 no changes are required. When any other reference (received _not_ from
2237 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
2238 instead of LDLM_LOCK_PUT().
2240 Arguably, changes are pervasive, and interface requires some discipline
2241 for proper use. On the other hand, it was very instrumental in finding
2242 a few leaked lock references.
2246 Description: Add ldlm_lock_addref_try().
2247 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
2248 attempts to addref a lock that might be being canceled
2253 Description: Add ldlm_weigh_callback().
2254 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
2255 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
2256 hard-coded `number of pages' logic.
2260 Description: Add lockdep annotations to llog code.
2261 Details : Use appropriately tagged _nested() locking calls in the places
2262 where llog takes more than one ->lgh_lock lock.
2266 Description: Add loi_kms_set().
2267 Details : Wrap kms updates into a helper function.
2271 Description: Constify instances of struct lsm_operations.
2272 Details : Constify instances of struct lsm_operations.
2276 Description: lu_conf support.
2277 Details : On a server, a file system object is uniquely identified
2278 by a fid, which is sufficient to locate and load all object
2279 state (inode). On a client, on the other hand, more data are
2280 necessary instantiate an object. Change lu_object_find() and
2281 friends to take additional `lu_conf' argument describing object.
2282 Typically this includes layout information.
2286 Description: lu_context fixes.
2287 Details : Introduce new lu_context functions that are needed on the client
2288 side, where some system threads (ptlrpcd) are shared by multiple
2289 modules, and so cannot be stopped during module shutdown.
2293 Description: Add start and stop methods to lu_device_type_operations.
2294 Details : Introduce two new methods in lu_device_type_operations, that are
2295 invoked when first instance of a given type is created and last one
2296 is destroyed respectively. This is need by CLIO.
2300 Description: Add lu_ref support to struct lu_device.
2301 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
2302 track leaked references.
2306 Description: Introduce lu_kmem_descr.
2307 Details : lu_kmem_descr and its companion interface allow to create
2308 and destroy a number of kmem caches at once.
2312 Description: Fix lu_object finalization race.
2313 Details : Fix a race between lu_object_find() finding an object and its
2314 concurrent finalization. This race is (most likely) not possible
2315 on the server, but might happen on the client.
2319 Description: Introduce lu_ref interface.
2320 Details : lu_ref is a debugging module allowing to track references to
2321 a given object. It is quite cpu expensive, and has to be
2322 explicitly enabled with --enable-lu_ref. See usage description
2327 Description: Factor lu_site procfs stats into a separate function.
2328 Details : Separate lu_site stats printing code into a separate function
2329 that can be reused on a client.
2333 Description: Constify instances of struct {lu,dt,md}_device_operations.
2334 Details : Constify instances of struct {lu,dt,md}_device_operations.
2338 Description: Introduce struct md_site and move meta-data specific parts of
2339 struct lu_site here.
2340 Details : Move md-specific fields out of struct lu_site into special struct
2341 md_site, so that lu_site can be used on a client.
2345 Description: Kill mdd_lov_destroy().
2346 Details : Remove unused mdd code.
2350 Description: Add st_block checking to multistat.c.
2351 Details : Add st_block checking to multistat.c.
2355 Description: Add lu_ref support to struct obd_device.
2356 Details : Add lu_ref tracking to obd_device.
2360 Description: Kill obd_set_fail_loc().
2361 Details : Remove unused code.
2365 Description: Add special type for ptlrpc_request interpret functions.
2366 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
2367 there. Actual usage will be in CLIO.
2371 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
2372 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
2373 doesn't work with lockdep.
2377 Description: Add rwv.c test program.
2378 Details : New testing program exercising readv(2) and writev(2) (Qian).
2382 Description: Add sendfile.c test program.
2383 Details : New testing program exercising sendfile(2) (Jay).
2387 Description: Ratelimit a message that can be very frequent.
2388 Details : Ratelimit a memory allocation failure message that can
2393 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2394 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2398 Description: Liblustre build fixes.
2399 Details : Liblustre build fixes.
2403 Description: libcfs: add cfs_{need,cond}_resched() interface.
2404 Details : libcfs: add cfs_{need,cond}_resched() definition and
2405 implementations for Linux, NT, and liblustre.
2407 Severity : enhancement
2409 Description: More exported tunables for mballoc
2410 Details : Add support for tunable preallocation window and new tunables for
2411 large/small requests
2415 Description: Detect corruption of block bitmap and checking for preallocations
2416 Details : Checks validity of on-disk block bitmap. Also it does better
2417 checking of number of applied preallocations. When corruption is
2418 found, it turns filesystem readonly to prevent further corruptions.
2422 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
2423 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
2424 window to 0. If it does hit cache miss, and read-pattern is still
2425 stride-io mode, does not reset the stride window, but also does not
2426 increase the stride window length in this case.
2430 Frequency : only for big-endian servers
2431 Description: Check if system is big-endian while mounting fs with extents feature
2432 Details : Mounting a filesystem with extents feature will fail on big-endian
2433 systems since ext3-based ldiskfs is not supported on big-endian
2434 systems. This can be over-riden with "bigendian_extents" mount option.
2436 Severity : enhancement
2438 Description: The root squash functionality
2439 Details : A security feature, which is to prevent users from being able
2440 to mount lustre on their desktop, run as root, and delete
2441 all of the files in the filesystem. The goal is accomplished by
2442 remapping user id (UID) and group id (GID) of the root user to
2443 a UID and GID specified by the system administartor via Lustre
2444 configuration management server (MGS). The functionality also
2445 allows to specify sets of clients for which the remapping does
2450 Description: Excessive recovery window
2451 Details : With AT enabled, the recovery window can be excessively long (6000+
2452 seconds). To address this problem, we no longer use
2453 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
2454 timeout no longer depends on the service time, it is set to
2455 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
2456 time via pb_service_time.
2460 Description: Watchdog triggered on MDS failover
2461 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
2462 the OSTs know that the MDS "UUID" can be reused for the same export
2463 from a different NID, so we do not need to wait for the export to be
2467 Frequency : rare, only if using MMP with Linux RAID
2469 Description: MMP doesn't work with Linux RAID
2470 Details : While using HA for Lustre servers with Linux RAID, it is possible
2471 that MMP will not detect multiple mounts. To make this work we
2472 need to unplug the device queue in RAID when the MMP block is being
2473 written. Also while reading the MMP block, we should read it from
2474 disk and not the cached one.
2476 Severity : enhancement
2478 Description: open file using fid
2479 Details : A file can be opened using just its fid, like
2480 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
2483 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
2486 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
2487 Details : add proper locking for creation of jbd_4k slab cache
2491 Description: MMP check in ext3_remount() fails without displaying any error
2492 Details : When multiple mount protection fails during remount, proper error
2495 Severity : enhancement
2497 Description: Allow stripe size to be up to 4G-64k
2498 Details : Fix math logic to allow large stripe sizes.
2502 Description: add check for >8TB ldiskfs filesystems
2503 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
2504 Don't allow >8TB ldiskfs filesystems to be mounted without
2505 force_over_8tb mount option
2510 Descriptoin: OSS complains "too many blocks" and crashes
2511 Details : In filter_commitrw_write():
2512 1. lquota_chkquota();
2513 2. obdo_from_inode();
2514 3. lquota_pending_commit();
2515 "2." has changed uid and gid in this case, then caused
2516 lquota_chkquota() and lquota_pending_commit() to use
2517 different uid and gid. That is the root of the bug.
2521 Descriptoin: getstripe doesn't report stripe count = -1 sanely
2522 Details : Detect and print out the 'stripe count = -1' case properly.
2526 Description: Out or order replies might be lost on replay
2527 Details : In ptlrpc_retain_replayable_request if we cannot find retained
2528 request with tid smaller then one currently being added, add it
2529 to the start, not end of the list.
2531 --------------------------------------------------------------------------------
2533 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
2535 * Support for kernels:
2536 2.6.5-7.283 (SLES 9),
2537 2.6.9-55.EL (RHEL 4),
2538 2.6.16.46-0.14 (SLES 10),
2539 2.6.18.8 vanilla (kernel.org)
2540 * Client support for unpatched kernels:
2541 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2542 2.6.16 - 2.6.22 vanilla (kernel.org)
2543 * Due to recently discovered recovery problems, we do not recommend
2544 using patchless RHEL 4 clients with this or any earlier release.
2545 * Recommended e2fsprogs version: 1.39.cfs8
2546 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2547 * Starting with this release, the ldiskfs backing filesystem required
2548 by Lustre is now in its own package, lustre-ldiskfs. This package
2549 should be installed. It is versioned separately from Lustre and
2550 may be released separately in future.
2554 Description: block reactivating mgc import until all deactivates complete
2555 Details : Fix race when failing back MDT/MGS to itself (testing)
2558 Frequency : at statup only
2560 Description: mds_lov_synchronize race leads to various problems
2561 Details : simultaneous MDT->OST connections at startup can cause the
2562 sync to abort, leaving the OSC in a bad state.
2564 Severity : enhancement
2566 Description: add optional extra BUILD_VERSION info
2567 Details : add a new environment variable (namely LUSTRE_VERS) which allows
2568 to override the lustre version.
2571 Frequency : 2.6.18 servers only
2573 Description: ll_kern_mount() doesn't release the module reference
2574 Details : The ldiskfs module reference count never drops down to 0
2575 because ll_kern_mount() doesn't release the module reference.
2580 Description: server LBUG when using old ost_num_threads parameter
2581 Details : Accept the old ost_num_threads parameter but warn that it
2582 is deprecated, and fix an off-by-one error that caused an LBUG.
2587 Description: Transient SCSI error results in persistent IO issue
2588 Details : iobuf->dr_error is not reinitialized to 0 between two
2592 Frequency : sometimes when underlying device returns I/O errors
2594 Description: OSTs not going read-only during write failures
2595 Details : OSTs are not remounted read-only when the journal commit threads
2596 get I/O errors because fsfilt_ext3 calls journal_start/stop()
2597 instead of the ext3 wrappers.
2601 Description: poor connect scaling with increasing client count
2602 Details : Don't run filter_grant_sanity_check for more than 100 exports
2603 to improve scaling for large numbers of clients.
2606 Frequency : SLES10 only
2608 Description: sanity-quota.sh quotacheck failed: rc = -22
2609 Details : Quotas cannot be enabled on SLES10.
2612 Frequency : liblustre clients only
2614 Description: getdirentries does not give error when run on compute nodes
2615 Details : getdirentries does not fail when the size specified as an argument
2616 is too small to contain at least one entry
2618 Severity : enhancement
2620 Description: Add LNET router traceability for debug purposes
2621 Details : If a checksum failure occurs with a router as part of the
2622 IO path, the NID of the last router that forwarded the bulk data
2623 is printed so it can be identified.
2628 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
2629 Details : Due to a race condition, liblustre clients were occasionally
2630 evicted incorrectly.
2632 Severity : enhancement
2634 Description: lfs setstripe use optional parameters instead of postional
2637 Severity : enhancement
2639 Description: Nanosecond timestamp support for ldiskfs
2640 Details : The on-disk ldiskfs filesystem has added support for nanosecond
2641 resolution timestamps. There is not yet support for this at
2642 the Lustre filesystem level.
2645 Frequency : during server recovery
2647 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
2648 Details : request with rq_no_resend flag not awake l_wait_event if they get
2652 Frequency : nfs export on patchless client
2654 Description: connectathon hang when test nfs export over patchless client
2655 Details : Disconnected dentry cannot be found with lookup, so we do not need
2656 to unhash it or make it invalid
2659 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
2660 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
2661 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
2666 Description: open req refcounting wrong on reconnect
2667 Details : If reconnect happened between getting open reply from server and
2668 call to mdc_set_replay_data in ll_file_open, we will schedule
2669 replay for unreferenced request that we are about to free.
2670 Subsequent close will crash in variety of ways.
2671 Check that request is still eligible for replay in
2672 mdc_set_replay_data().
2677 Description: disable writes to filesystem when reading health_check file
2678 Details : the default for reading the health_check proc file has changed
2679 to NOT do a journal transaction and write to disk, because this
2680 can cause reads of the /proc file to hang and block HA state
2681 checking on a healthy but otherwise heavily loaded system. It
2682 is possible to return to the previous behaviour during configure
2683 with --enable-health-write.
2685 Severity : enhancement
2687 Description: 64-bit inode version
2688 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
2689 to the inode. This will be required for version-based recovery.
2694 Description: MDS fails to start if a duplicate client export is detected
2695 Details : in some rare cases it was possible for a client to connect to
2696 an MDS multiple times. Upon recovery the MDS would detect this
2697 and fail during startup. Handle this more gracefully.
2699 Severity : enhancement
2701 Description: Add -o localflock option to simulate old noflock
2703 Details : This will achieve local-only flock/fcntl locks
2709 Description: log_commit_thread vs filter_destroy race leads to crash
2710 Details : Take import reference before releasing llog record semaphore
2715 Description: Wrong request locking in request set processing
2716 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
2717 imp_delayed_list, in this place should be used imp_lock.
2720 Frequency : when reconnection
2722 Description: Grant Leak when osc reconnect to OST
2723 Details : When osc reconnect ost, OST(filter) should check whether it
2724 should grant more space to client by comparing fed_grant and
2725 cl_avail_grant, and return the granted space to client instead
2726 of "new granted" space, because client will call osc_init_grant
2727 to update the client grant space info.
2730 Frequency : when client reconnect to OST
2732 Description: Grant Leak when osc do resend and replay bulk write
2733 Details : When osc reconnect to OST, OST(filter)should clear grant info of
2734 bulk write request, because the grant info while be sync between
2735 OSC and OST when reconnect, and we should ignore the grant info
2736 these of resend/replay write req.
2741 Description: Grant space more than avaiable left space sometimes.
2742 Details : When then OST is about to be full, if two bulk writing from
2743 different clients came to OST. Accord the avaliable space of the
2744 OST, the first req should be permitted, and the second one
2745 should be denied by ENOSPC. But if the seconde arrived before
2746 the first one is commited. The OST might wrongly permit second
2747 writing, which will cause grant space > avaiable space.
2750 Frequency : when client is evicted
2752 Description: Grant might be wrongly erased when osc is evicted by OST
2753 Details : when the import is evicted by server, it will fork another
2754 thread ptlrpc_invalidate_import_thread to invalidate the
2755 import, where the grant will be set to 0. While the original
2756 thread will update the grant it got when connecting. So if
2757 the former happened latter, the grant will be wrongly errased
2758 because of this race.
2763 Description: Checking Stale with correct fid
2764 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
2765 to check whether it is stale, because sometimes, we want the
2766 enqueue happened anyway, and op_data.fid2 will not be initialized.
2768 Severity : enhancement
2770 Description: update patchless client
2771 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
2774 Frequency : only with 2.4 kernel
2776 Description: random memory corruption
2777 Details : size of struct ll_inode_info is to big for union inode.u and this
2778 can be cause of random memory corruption.
2783 Description: Memory leak in recovery
2784 Details : Lov_mds_md was not free in an error handler in mds_create_object.
2785 It should also check obd_fail before fsfilt_start, otherwise if
2786 fsfilt_start return -EROFS,(failover mds during mds recovery).
2787 then the req will return with repmsg->transno = 0 and rc = EROFS.
2788 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
2789 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
2790 be freed no matter whether fsfilt_commit success or not.
2793 Frequency : only with huge count clients
2795 Description: Prevents from taking the superblock lock in llap_from_page for
2797 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
2798 for llap_from_page call in ll_removepage prevents from taking the
2799 superblock lock for a soon died page.
2804 Description: Not check open intent error before release open handle
2805 Details : in some rare cases, the open intent error is not checked before
2806 release open handle, which may cause
2807 ASSERTION(open_req->rq_transno != 0), because it tries to release
2808 the failed open handle.
2813 Description: Set cat log bitmap only after create log success.
2814 Details : in some rare cases, the cat log bitmap is set too early. and it
2815 should be set only after create log success.
2819 Description: Accessing a block bevice can re-enable I/O when Lustre is
2820 tearing down a device.
2821 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
2826 Description: service threads may hog cpus when there are a lot of requests
2827 Details : Insert cond_resched to give other threads a chance to use some CPU
2832 Description: the cat log was not initialized in recovery
2833 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
2834 unlink log on mds will not be initialized until mds post
2835 recovery. And also in mds post recovery, the unlink log will
2836 initialization will be done asynchronausly, so there will be race
2837 between add unlink log and unlink log initialization.
2841 Description: brw_stats were being printed incorrectly
2842 Details : brw_stats were being printed as log2 but all of them were not
2843 recorded as log2. Also remove some code duplication arising from
2844 filter_tally_{read,write}.
2848 Frequency : rare, only in recovery.
2849 Description: ASSERTION(req->rq_type != LI_POISON) failed
2850 Details : imp_lock should be held while iterating over imp_sending_list for
2851 prevent destroy request after get timeout in ptlrpc_queue_wait.
2855 Description: replay-single.sh test 52 fails
2856 Details : A lock's skiplist need to be cleanup when it being unlinked
2857 from its resource list.
2861 Description: Short directio read returns full requested size rather than
2863 Details : Direct I/O operations should return actual amount of bytes
2864 transferred rather than requested size.
2866 Severity : enhancement
2868 Description: metadata RPC reduction (e.g. for rm performance)
2869 Details : decrease the amount of synchronous RPC between clients and servers
2870 by canceling conflicing lock before the operation on the client
2871 and packing thier handles into the main operation RPC to server.
2873 Severity : enhancement
2875 Description: Async OSC create to avoid the blocking unnecessarily.
2876 Details : If a OST has no remain object, system will block on the creating
2877 when need to create a new object on this OST. Now, ways use
2878 pre-created objects when available, instead of blocking on an
2879 empty osc while others are not empty. If we must block, we block
2880 for the shortest possible period of time.
2884 Description: improve handling recoverable errors
2885 Details : if request processig with error which can be recoverable on server
2886 request should be resend, otherwise page released from cache and
2889 Severity : enhancement
2891 Description: refine locking for avoid write wrong info into lov_objid file
2892 Details : fix possible races with add new target and write/update data in
2895 --------------------------------------------------------------------------------
2897 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
2902 Frequency : on some architectures
2904 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
2905 Details : Uninitialized flags sometimes cause configuration commands to
2909 Frequency : patchless clients only
2911 Description: missing __iget() symbol export
2912 Details : The __iget() symbol export is missing. To avoid the need for
2913 this on patchless clients the deathrow inode reaper is turned
2914 off, and we depend on the VM to clean up old inodes. This
2915 dependency was during via the fix for bug 12181.
2917 --------------------------------------------------------------------------------
2919 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
2921 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
2922 INTEROPERATE with older versions automatically. In many cases a
2923 special upgrade step is needed. Please read the
2924 user documentation before upgrading any part of a live system.
2925 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
2926 version will not interoperate with 1.6 betas before beta5 (1.5.95).
2927 * WARNING: Lustre configuration and startup changes are required with
2928 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
2932 Severity : enhancement
2934 Description: MountConf
2935 Details : Lustre configuration is now managed via mkfs and mount
2936 commands instead of lmc and lconf. New obd types (MGS, MGC)
2937 are added for dynamic configuration management. See
2938 https://mail.clusterfs.com/wikis/lustre/MountConf for
2941 Severity : enhancement
2943 Description: dynamic OST addition
2944 Details : OSTs can now be added to a live filesystem
2946 Severity : enhancement
2948 Description: startup order invariance
2949 Details : MDTs and OSTs can be started in any order. Clients only
2950 require the MDT to complete startup.
2952 Severity : enhancement
2954 Description: parallel, asynchronous orphan cleanup
2955 Details : orphan cleanup is now performed in separate threads for each
2956 OST, allowing parallel non-blocking operation.
2958 Severity : enhancement
2960 Description: optimized stripe assignment
2961 Details : stripe assignments are now made based on ost space available,
2962 ost previous usage, and OSS previous usage, in order to try
2963 to optimize storage space and networking resources.
2965 Severity : enhancement
2967 Description: Permanently set tunables
2968 Details : All writable /proc/fs/lustre tunables can now be permanently
2969 set on a per-server basis, at mkfs time or on a live system.
2971 Severity : enhancement
2973 Description: Lustre message v2
2974 Details : Add lustre message format v2.
2976 Severity : enhancement
2978 Description: client OST exclusion list
2979 Details : Clients can be started with a list of OSTs that should be
2980 declared "inactive" for known non-responsive OSTs.
2984 Description: SPEC SFS validation failure on NFS v2 over lustre.
2985 Details : Changes the blocksize for regular files to be 2x RPC size,
2986 and not depend on stripe size.
2988 Severity : enhancement
2990 Description: Multiple MD RPCs in flight.
2991 Details : Further unserialise some read-only MDS RPCs - learn about intents.
2992 To avoid overly-overloading MDS, introduce a limit on number of
2993 MDS RPCs in flight for a single client and add /proc controls
2994 to adjust this limit.
2996 Severity : enhancement
2998 Description: client read/write statistics
2999 Details : Add client read/write call usage stats for performance
3000 analysis of user processes.
3001 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
3002 file access. extents_stats shows chunk size distribution.
3003 extents_stats_per_process show chunk size distribution per
3006 Severity : enhancement
3008 Description: mds statistics
3009 Details : Add detailed mds operations statistics in
3010 /proc/fs/lustre/mds/*/stats.
3014 Description: Failure of copying files with lustre special EAs.
3015 Details : Client side always return success for setxattr call for lustre
3016 special xattr (currently only "trusted.lov").
3021 Description: Refcount LNET uuids
3022 Details : The global LNET uuid list grew linearly with every startup;
3023 refcount repeated list entries instead of always adding to
3026 Severity : enhancement
3028 Description: Dynamic service threads
3029 Details : Within a small range, start extra service threads
3030 automatically when the request queue builds up.
3033 Frequency : mixed-endian client/server environments
3035 Description: mixed-endian crashes
3036 Details : The new msg_v2 system had some failures in mixed-endian
3039 Severity : enhancement
3041 Description: Easy OST removal
3042 Details : OSTs can be permanently deactivated with e.g. 'lctl
3043 conf_param lustre-OST0001.osc.active=0'
3045 Severity : enhancement
3047 Description: MGS proc entries
3048 Details : Added basic proc entries for the MGS showing what filesystems
3051 Severity : enhancement
3053 Description: provide MGS failover
3054 Details : Added config lock reacquisition after MGS server failover.
3056 Severity : enhancement
3058 Description: add Linux 2.4 support
3059 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
3063 Description: a large application tries to do I/O to the same resource and dies
3064 in the middle of it.
3065 Details : Check the req->rq_arrival time after the call to
3066 ost_brw_lock_get(), but before we do anything about
3067 processing it & sending the BULK transfer request. This
3068 should help move old stale pending locks off the queue as
3069 quickly as obd_timeout.
3072 Frequency : when an incorrect nid is specified during startup
3074 Description: ptlrpc connect to non-existant node causes kernel crash
3075 Details : LNET can't be re-entered from an event callback, which
3076 happened when we expire a message after the export has been
3077 cleaned up. Instead, hand the zombie cleanup off to another
3080 Severity : enhancement
3082 Description: plain/inodebits lock performance improvement
3083 Details : Grouping plain/inodebits in granted list by their request modes
3084 and bits policy, thus improving the performance of search through
3088 Frequency : only if OST filesystem is corrupted
3090 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3091 Details : for a short time RPCs with bulk IO are in the replay list,
3092 but replay of bulk IOs is unimplemented. If the OST filesystem
3093 is corrupted due to disk cache incoherency and then replay is
3094 started it is possible to trip an assertion. Avoid putting
3095 committed RPCs into the replay list at all to avoid this issue.
3098 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3100 Description: System hang on startup
3101 Details : This bug allowed the liblustre (e.g. catamount) client to
3102 return to the app before handling all startup RPCs. This
3103 could leave the node unresponsive to lustre network traffic
3104 and manifested as a server ptllnd timeout.
3106 Severity : enhancement
3108 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3109 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
3110 boolean to control whether to print peer debug info when a
3111 client's RPC times out.
3114 Frequency : only for kernels with patches from Lustre below 1.4.3
3116 Description: Remove old rdonly API
3117 Details : Remove old rdonly API which unused from at least lustre 1.4.3
3120 Frequency : only for devices with external journals
3122 Description: Set external device read-only also
3123 Details : During a commanded failover stop, we set the disk device
3124 read-only while the server shuts down. We now also set any
3125 external journal device read-only at the same time.
3128 Frequency : when upgrading from 1.4 while trying to change parameters
3130 Description: The wrong (new) MDC name was used when setting parameters for
3131 upgraded MDT's. Also allows changing of OSC (and MDC)
3132 parameters if --writeconf is specified at tunefs upgrade time.
3135 Frequency : when setting specific ost indicies
3137 Description: QOS code breaks on skipped indicies
3138 Details : Add checks for missing OST indicies in the QOS code, so OSTs
3139 created with --index need not be sequential.
3141 Severity : enhancement
3143 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3144 Details : The uninit_groups feature works in conjunction with the kernel
3145 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3146 up the pass1 processing of e2fsck. This is a read-only feature
3147 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3148 mount filesystems that have had this feature enabled.
3150 Severity : enhancement
3152 Description: Improve multi-block allocation algorithm to avoid fragmentation
3153 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3154 allocation locality and avoid filesystem fragmentation.
3156 ------------------------------------------------------------------------------
3158 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
3160 * Support for kernels:
3161 2.4.21-47.0.1.EL (RHEL 3)
3162 2.6.5-7.283 (SLES 9)
3163 2.6.9-42.0.10.EL (RHEL 4)
3164 2.6.12.6 vanilla (kernel.org)
3165 2.6.16.27-0.9 (SLES 10)
3166 * Recommended e2fsprogs version: 1.39.cfs5
3168 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3172 Frequency : occasional, depends on client load and configuration
3173 Bugzilla : 12181, 12203
3174 Description: data loss for recently-modified files
3176 Details : In some cases it is possible that recently written or created
3177 files may not be written to disk in a timely manner (this should
3178 normally be within 30s unless client IO load is very high).
3179 The problem appears as zero-length files or files that are a
3180 multiple of 1MB in size after a client crash or client eviction
3181 that are missing data at the end of the file.
3183 This problem is more likely to be hit on clients where files are
3184 repeatedly created and unlinked in the same directory, clients
3185 have a large amount of RAM, have many CPUs, the filesystem has
3186 many OSTs, the clients are rebooted frequently, and/or the files
3187 are not accessed by other nodes after being written.
3189 The presence of the problem can be detected by looking at
3190 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
3191 smaller than the second (nr_unused) then dirty files will not
3192 be flushed automatically to disk. "sync; sleep 10" should be
3193 run several times on the node before unmounting it to update
3194 Lustre (this is also safe to run on nodes without this problem).
3196 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
3197 that can cause this same problem, so customers using that kernel
3198 also need to update the kernel in addition to Lustre. In order
3199 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
3201 It is normal that files written just before a client crash (less
3202 than 30s) may not yet have been flushed to disk, even for local
3206 Frequency : frequent on thin XT3 nodes
3208 Description: UUID collision on thin XT3 Linux nodes
3209 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
3210 randomly, since we relied on an insufficiently-seeded PRNG.
3215 Description: OSS hangs after "All ost request buffers busy"
3216 Details : A deadlock between quota and journal operations caused OSS
3217 hangs after printing "All ost request buffers busy."
3220 Frequency : always on liblustre builds
3222 Description: Cleanup compiler warnings on liblustre
3225 Frequency : always on liblustre builds on XT3
3227 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
3231 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
3232 > 32000 subdirectories
3233 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
3234 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
3235 subdirectory count crosses 32000. This will aid e2fsck to
3236 correctly handle more than 32000 subdirectories.
3239 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3241 Description: System hang on startup
3242 Details : This bug allowed the liblustre (e.g. catamount) client to
3243 return to the app before handling all startup RPCs. This
3244 could leave the node unresponsive to lustre network traffic
3245 and manifested as a server ptllnd timeout.
3247 Severity : enhancement
3249 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3250 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
3251 boolean to control whether to print peer debug info when a
3252 client's RPC times out.
3257 Description: make O_SYNC working on 2.6 kernels
3258 Details : 2.6 kernels use different method for mark pages for write,
3259 so need add a code to lustre for O_SYNC work.
3264 Description: Failure to close file and release space on NFS
3265 Details : Put inode details into lock acquired in ll_intent_file_open.
3266 Use mdc_intent_lock in ll_intent_open to properly
3267 detect all kind of errors unhandled by mdc_enqueue.
3272 Description: proc file read during shutdown sometimes raced obd removal,
3274 Details : Add lock to prevent obd access after proc file removal.
3277 Frequency : Only for files larger than 4GB on 32-bit clients.
3279 Description: improperly doing page alignment of locks
3280 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
3281 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3286 Description: RPCs being resent when they shouldn't be
3287 Details : Some RPCs that should not be resent are being resent. This
3288 can cause inconsistencies in the RPC state machine. Do not
3289 resend such requests.
3292 Frequency : rare, only with NFS export
3294 Description: Crash on NFS re-export node
3295 Details : under very unusual load conditions an assertion is hit in
3296 ll_intent_file_open()
3299 Frequency : only if OST filesystem is corrupted
3301 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3302 Details : for a short time RPCs with bulk IO are in the replay list,
3303 but replay of bulk IOs is unimplemented. If the OST filesystem
3304 is corrupted due to disk cache incoherency and then replay is
3305 started it is possible to trip an assertion. Avoid putting
3306 committed RPCs into the replay list at all to avoid this issue.
3311 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3312 Details : Large single O_DIRECT read and write calls can fail to allocate
3313 a sufficiently large buffer to process the request. In case of
3314 allocation failure the allocation is retried with a smaller
3315 buffer and broken into smaller requests.
3317 Severity : enhancement
3319 Description: Add -o localflock option to simulate old noflock behaviour.
3320 Details : This will achieve local-only flock/fcntl locks coherentness.
3325 Description: versioning check is incomplete
3326 Details : Checking the version difference of client vs. server, report
3327 error if the gap is too big.
3332 Description: add support PG_writeback bit
3333 Details : add support for PG_writeback bit for Lustre, for more carefull
3334 work with page cache in 2.6 kernel. This also fix some deadlocks
3335 and remove hack for work O_SYNC with 2.6 kernel.
3337 Severity : enhancement
3339 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3340 Details : The uninit_groups feature works in conjunction with the kernel
3341 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3342 up the pass1 processing of e2fsck. This is a read-only feature
3343 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3344 mount filesystems that have had this feature enabled.
3346 Severity : enhancement
3348 Description: Improve multi-block allocation algorithm to avoid fragmentation
3349 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3350 allocation locality and avoid filesystem fragmentation.
3352 ------------------------------------------------------------------------------
3354 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3356 * Support for kernels:
3357 2.6.9-42.0.3.EL (RHEL 4)
3358 2.6.5-7.276 (SLES 9)
3359 2.4.21-47.0.1.EL (RHEL 3)
3360 2.6.12.6 vanilla (kernel.org)
3361 2.6.16.21-0.8 (SLES10)
3362 * Recommended e2fsprogs version: 1.39.cfs2-0
3364 * The backwards-compatible /proc/sys/portals symlink has been removed
3365 in this release. Before upgrading, please ensure that you change
3366 any configuration scripts or /etc/sysctl.conf files that access
3367 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3368 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3369 in advance of the upgrade on any system running Lustre 1.4.6 or
3370 newer, since /proc/sys/lnet was added in that version.
3371 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3375 Frequency : only when quota is used
3377 Description: avoid scanning export list for quota master
3378 Details : Change the algorithms to avoid scanning export list in order
3379 to improve the efficiency.
3382 Frequency : MDS failover only, very rarely
3384 Description: "went back in time" messages on mds failover
3385 Details : The greatest transno may be lost when the current operation
3386 finishes with an error (transno==0) and the client's last_rcvd
3387 record is over-written. Save the greatest transno in the
3388 mds_last_transno for this case.
3391 Frequency : always for specific kernels and striping counts
3393 Description: client may get "Matching packet too big" without ACL support
3394 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
3395 when trying to access files in certain configurations. The
3396 clients should in fact be denied when mounting because they do
3397 not understand ACLs.
3400 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
3402 Description: many clients connecting with IO in progress causes connect timeouts
3403 Details : Avoid synchronous journal commits to avoid delays caused by many
3404 clients connecting/disconnecting when bulk IO is in progress.
3405 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
3406 OST_IO_PORTAL to avoid delays behind potentially many pending
3410 Frequency : occasionally with multiple writers to a single file
3412 Description: shared writes to file may result in wrong size reported by stat()
3413 Details : Allow growing of kms when extent lock is cancelled
3416 Frequency : always with random mmap IO to multi-striped file
3418 Description: mmap write might be lost if we are writing to a 'hole' in stripe
3419 Details : Only if the hole is at the end of OST object so that kms is too
3420 small. Fix is to increase kms accordingly in ll_nopage.
3423 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
3425 Description: writes to a missing object would leak memory on the OST
3426 Details : If there is an inconsistency between the MDS and OST filesystems,
3427 such that the MDS references an object that doesn't exist, writes
3428 to that object will leak memory due to incorrect cleanup in the
3429 error handling path, eventually running out of memory on the OST.
3434 Description: Creating too long symlink causes lustre errors
3435 Details : Check symlink and name lengths before sending requests to MDS.
3438 Frequency : only if flock is enabled (not on by default)
3440 Description: posix locks not released on fd closure on 2.6.9+
3441 Details : We failed to add posix locks to list of inode locks on 2.6.9+
3442 kernels, this caused such locks not to be released on fd close and
3443 then assertions on fs unmount about still used locks.
3446 Frequency : MDS failover only, very rarely
3448 Description: clients may get ASSERTION(granted_lock != NULL)
3449 Details : When request was taking a long time, and a client was resending
3450 a getattr by name lock request. The were multiple lock requests
3451 with the same client lock handle and
3452 mds_getattr_name->fixup_handle_for_resent_request found one of the
3453 lock handles but later failed with ASSERTION(granted_lock != NULL).
3458 Description: handle->h_buffer_credits > 0, assertion failure
3459 Details : h_buffer_credits is zero after truncate, causing assertion
3460 failure. This patch extends the transaction or creates a new
3464 Frequency : NFS re-export or patchless client
3465 Bugzilla : 11179, 10796
3466 Description: Crash on NFS re-export node (__d_move)
3467 Details : We do not want to hash the dentry if we don't have a lock.
3468 But if this dentry is later used in d_move, we'd hit uninitialised
3469 list head d_hash, so we just do this to init d_hash field but
3470 leave dentry unhashed.
3473 Frequency : NFS re-export or patchless client
3475 Description: NFS exports has problem with symbolic link
3476 Details : lustre client didn't properly install dentry when re-exported
3477 to NFS or running patchless client.
3480 Frequency : NFS re-export or patchless client
3482 Description: Various nfs/patchless fixes.
3483 Details : fixes reuse disconected alias for lookup process - this fixes
3484 warning "find_exported_dentry: npd != pd",
3485 fix permission error with open files at nfs.
3486 fix apply umask when do revalidate.
3489 Frequency : occasional
3491 Description: Crash on NFS re-export node
3492 Details : calling clear_page() on the wrong pointer triggered oops in
3493 generic_mapping_read().
3496 Frequency : rarely, using O_DIRECT IO
3498 Description: unaligned directio crashes client with LASSERT
3499 Details : check for unaligned buffers before trying any requests.
3502 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
3504 Description: stale data returned from RAID cache
3505 Details : If only a small amount of IO is done to the RAID device before
3506 reading it again it is possible to get stale data from the RAID
3507 cache instead of reading it from disk.
3510 Frequency : always for sles10 kernel
3512 Description: sles10 support
3513 Details : ll_follow_link: compile fixes and using of nd_set_link
3514 under newer kernels.
3517 Frequency : depends on arch, kernel and compiler version, always on sles10
3520 Description: recursive or deep enough symlinks cause stack overflow
3521 Details : getting rid of large stack-allocated variable in
3525 Frequency : depends on hardware
3527 Description: lustre write performance loss in the SLES10 kernel
3528 Details : the performance loss is caused by using of write barriers in the
3529 ext3 code. The SLES10 kernel turns barrier support on by
3530 default. The fix is to undo that change for ldiskfs.
3532 ------------------------------------------------------------------------------
3534 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
3536 * Support for kernels:
3537 2.6.9-42.0.3EL (RHEL 4)
3538 2.6.5-7.276 (SLES 9)
3539 2.4.21-47.0.1.EL (RHEL 3)
3540 2.6.12.6 vanilla (kernel.org)
3544 Frequency : quota enabled and large files being deleted
3546 Description: releasing more than 4GB of quota at once hangs OST
3547 Details : If a user deletes more than 4GB of files on a single OST it
3548 will cause the OST to spin in an infinite loop. Release
3549 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3554 Description: statfs data retrieved from /proc may be stale or zero
3555 Details : When reading per-device statfs data from /proc, in the
3556 {kbytes,files}_{total,free,avail} files, it may appear
3557 as zero or be out of date.
3560 Frequency : systems with MD RAID1 external journal devices
3562 Description: lconf's call to blkid is confused by RAID1 journal devices
3563 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3564 of returning all block devices that match the journal UUID.
3567 Frequency : always, for aggregate stripe size over 4GB
3569 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
3570 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
3571 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
3574 Frequency : always if "lfs find" used on a local file/directory
3576 Description: "lfs find" segfaults if used on a local file/directory
3577 Details : The case where a directory component was not specified wasn't
3578 handled correctly. Handle this properly.
3581 Frequency : always on ppc64
3583 Description: the write to an ext3 filesystem mounted with mballoc got stuck
3584 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3585 perform endianness conversion.
3588 Frequency : rarely (truncate to non-zero file size after write under load)
3589 Bugzilla : 10730, 10687
3590 Description: Files padded with zeros to next 4K multiple
3591 Details : With filesystems mounted using the "extents" option (2.6 kernels)
3592 it is possible that files that are truncated to a non-zero size
3593 immediately after being written are filled with zero bytes beyond
3594 the truncated size. No file data is lost.
3596 Severity : enhancement
3598 Description: Allow recovery/failover for liblustre clients.
3599 Details : liblustre clients were unaware of failover configurations until
3602 Severity : enhancement
3604 Description: user file locks should fail when not mounting with flock option
3605 Details : Set up an error-returning stub in ll_file_operations.lock field
3606 to prevent incorrect behaviour when client is mounted without
3607 flock option. Also, set up properly f_op->flock field for
3611 Frequency : always on ia64
3613 Description: "lfs df" loops on printing out MDS statfs information
3614 Details : The obd_ioctl_data was not initialized and in some systems
3615 this caused a failure during the ioctl that did not return
3616 an error. Initialize the struct and return an error on failure.
3619 Frequency : SLES 9 only
3621 Description: Error of copying files with lustre special EAs as root
3622 Details : Client side always return success for setxattr call for lustre
3623 special xattr (currently only "trusted.lov").
3626 Frequency : rarely on clusters with both ia64+i386 clients
3628 Description: ia64+i686 clients doing shared IO on the same file may LBUG
3629 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
3630 clients are doing concurrent writes to the same file it is
3631 possible that the ia64 clients may LASSERT because the OST
3632 extent locks are not PAGE_SIZE aligned. Ensure that grown
3633 locks are always aligned on the request boundary.
3636 Frequency : specific use, occasional
3638 Description: Overwriting in use executable truncates on-disk binary image
3639 Details : If one node attempts to overwrite an executable in use by
3640 another node, we now correctly return ETXTBSY instead of
3641 truncating the file.
3643 Severity : enhancement
3645 Description: Async OSC create to avoid the blocking unnecessarily.
3646 Details : If a OST has no remain object, system will block on the creating
3647 when need to create a new object on this OST. Now, ways use
3648 pre-created objects when available, instead of blocking on an
3649 empty osc while others are not empty. If we must block, we block
3650 for the shortest possible period of time.
3655 Description: chmod on Lustre root is propagated to other clients
3656 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
3657 invalid by the follow_mount time.
3660 Frequency : liblustre clients only
3662 Description: Race in 'instant cancel' lock handling could lead to such locks
3663 never to be granted in case of SMP MDS
3664 Details : Do not destroy not yet granted but cbpending locks in
3668 Frequency : replay/resend of open
3670 Description: non null lock assetion failure in mds_intent_policy
3671 Details : Trying to replay/resend lockless open requests resulted in
3672 mds_open() returning 0 with no lock. Now it sets a flag if
3673 a lock is going to be returned.
3675 Severity : enhancement
3677 Description: Checksum enhancements
3678 Details : New checksum enhancements allow for resending RPCs that failed
3681 Severity : enhancement
3683 Description: Tunables on number of dirty pages in cacche
3684 Details : Allow to set limit on number of dirty pages cached.
3689 Description: client crash on unmount - lock still has references
3690 Details : In some error handling cases it was possible to leak a lock
3691 reference on a client while accessing a file. This was not
3692 harmful to the client during operation, but would cause the
3693 client to crash when the filesystem is unmounted.
3696 Frequency : specific case, rare
3698 Description: ETXTBSY on mds though file not in use by client
3699 Details : ETXTBSY is no longer incorrectly returned when attempting to
3700 chmod or chown a directory that the user previously tried to
3701 execute or a currently-executing binary.
3704 Frequency : extremely rare except on liblustre-based clients
3706 Description: Lustre space not freed when files are deleted
3707 Details : Clean up open-unlinked files after client eviction. Previously
3708 the unlink was skipped and the files remained as orphans.
3713 Description: OST failure "would be an LBUG" in waiting_locks_callback()
3714 Details : In some cases it was possible to send a blocking callback to
3715 a client doing a glimpse, even though that client didn't get
3716 a lock granted. When the glimpse lock is cancelled on the OST
3717 the freed lock is left on the waiting list and corrupted the list.
3720 Frequency : all core dumps
3722 Description: Broke core dumps to lustre
3723 Details : Negative dentry may be unhashed if parent does not have UPDATE
3724 lock, but some callers, e.g. do_coredump, expect dentry to be
3725 hashed after successful create, hash it in ll_create_it.
3727 ------------------------------------------------------------------------------
3729 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
3731 * Support for kernels:
3732 2.6.9-42.0.2.EL (RHEL 4)
3733 2.6.5-7.276 (SLES 9)
3734 2.4.21-40.EL (RHEL 3)
3735 2.6.12.6 vanilla (kernel.org)
3739 Frequency : always on RHEL 3
3741 Description: Number of open files grows over time
3742 Details : The number of open files grows over time, whether or not
3743 Lustre is started. This was due to a filp leak introduced
3744 by one of our kernel patches.
3746 ------------------------------------------------------------------------------
3748 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
3750 * Support for kernels:
3751 2.6.9-42.EL (RHEL 4)
3752 2.6.5-7.276 (SLES 9)
3753 2.4.21-40.EL (RHEL 3)
3754 2.6.12.6 vanilla (kernel.org)
3759 Bugzilla : 5719, 9635, 9792, 9684
3760 Description: OST (or MDS) trips assertions in (re)connection under heavy load
3761 Details : If a server is under heavy load and cannot reply to new
3762 connection requests before the client resends the (re)connect,
3763 the connection handling code can behave badly if two service
3764 threads are concurrently handing separate (re)connections from
3765 the same client. Add better locking to the connection handling
3766 code, and ensure that only a single connection will be processed
3767 for a given client UUID, even if the lock is dropped.
3769 Severity : enhancement
3771 Description: add TCP zero-copy support to kernel
3772 Details : Add support to the kernel TCP stack to allow zero-copy bulk
3773 sends if the hardware supports scatter-gather and checksumming.
3774 This allows socklnd to do client-write and server-read more
3775 efficiently and reduce CPU utilization from skbuf copying.
3778 Frequency : only if NFS exporting from client
3780 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
3781 Details : If setting attributes on a file created under NFS that had
3782 never been opened it would be possible to oops the client
3783 if the file had no objects.
3786 Frequency : always for liblustre
3788 Description: liblustre client does MDS+OSTs setattr RPC for each write
3789 Details : When doing a write from a liblustre client, the client
3790 incorrectly issued an RPC to the MDS and each OST the file was
3791 striped over in order to update the timestamps. When writing
3792 with small chunks and many clients this could overwhelm the MDS
3793 with RPCs. In all cases it would slow down the write because
3794 these RPCs are unnecessary.
3796 Severity : enhancement
3798 Description: allow number of MDS service threads to be changed at module load
3799 Details : It is now possible to change the number of MDS service threads
3800 running. Adding "options mds mds_num_threads={N}" to the MDS's
3801 /etc/modprobe.conf will set the number of threads for the next
3802 time Lustre is restarted (assuming the "mds" module is also
3803 reloaded at that time). The default number of threads will
3804 stay the same, 32 for most systems.
3809 Description: OST crash if filesystem is unformatted or corrupt
3810 Details : If an OST is started on a device that has never been formatted
3811 or if the filesystem is corrupt and cannot even mount then the
3812 error handling cleanup routines would dereference a NULL pointer.
3817 Description: NULL pointer deref in llap_from_page.
3818 Details : get_cache_page_nowait can return a page with NULL (or otherwise
3819 incorrect) mapping if the page was truncated/reclaimed while it was
3820 searched for. Check for this condition and skip such pages when
3821 doing readahead. Introduce extra check to llap_from_page() to
3822 verify page->mapping->host is non-NULL (so page is not anonymous).
3825 Frequency : Sometimes when using sys_sendfile
3827 Description: "page not covered by a lock" warnings from ll_readpage
3828 Details : sendfile called ll_readpage without right page locks present.
3829 Now we introduced ll_file_sendfile that does necessary locking
3830 around call to generic_file_sendfile() much like we do in
3834 Frequency : with certain MDS communication failures at client mount time
3836 Description: NULL pointer deref after failed client mount
3837 Details : a client connection request may delayed by the network layer
3838 and not be sent until after the PTLRPC layer has timed out the
3839 request. If the client fails the mount immediately it will try
3840 to clean up before the network times out the request. Add a
3841 reference from the request import to the obd device and delay
3842 the cleanup until the network drops the request.
3845 Frequency : occasionally during client (re)connect
3847 Description: assertion failure during client (re)connect
3848 Details : processing a client connection request may be delayed by the
3849 client or server longer than the client connect timeout. This
3850 causes the client to resend the connection request. If the
3851 original connection request is replied in this interval, the
3852 client may trip an assertion failure in ptlrpc_connect_interpret()
3853 which thought it would be the only running connect process.
3856 Frequency : only with obd_echo servers and clients that are rebooted
3858 Description: kernel BUG accessing uninitialized data structure
3859 Details : When running an obd_echo server it did not start the ping_evictor
3860 thread, and when a client was evicted an uninitialized data
3861 structure was accessed. Start the ping_evictor in the RPC
3862 service startup instead of the OBD startup.
3864 Severity : enhancement
3865 Bugzilla : 10193 (patchless)
3866 Description: Remove dependency on various unexported kernel interfaces.
3867 Details : No longer need reparent_to_init, exit_mm, exit_files,
3868 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
3871 Frequency : rare (only users of deprecated and unsupported LDAP config)
3873 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
3874 Details : LDAP apparently contains 'lustreName' attributes instead of
3875 'name'. A simple remapping of the name is sufficient.
3878 Frequency : rare (only with non-default dump_on_timeout debug enabled)
3880 Description: waiting_locks_callback trips kernel BUG if client is evicted
3881 Details : Running with the dump_on_timeout debug flag turned on makes
3882 it possible that the waiting_locks_callback() can try to dump
3883 the Lustre kernel debug logs from an interrupt handler. Defer
3884 this log dumping to the expired_lock_main() thread.
3886 Severity : enhancement
3888 Description: Support NFS exporting on 2.6 kernels.
3889 Details : Implement non-rawops metadata methods for NFS server to use without
3890 changing NFS server code.
3893 Frequency : very rare (synthetic metadata workload only)
3895 Description: two racing renames might cause an MDS thread to deadlock
3896 Details : Running the "racer" program may cause one MDS thread to rename
3897 a file from being the source of a rename to being the target of
3898 a rename at exactly the same time that another thread is doing
3899 so, and the second thread has already enqueued these locks after
3900 doing a lookup of the target and is trying to relock them in
3901 order. Ensure that we don't try to re-lock the same resource.
3904 Frequency : only very large systems with liblustre clients
3906 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
3907 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
3908 all OSTs in parallel. This allows the eviction of stale liblustre
3909 clients to proceed much faster than if they were done in series,
3910 and also offers similar improvements for other set_info RPCs.
3915 Description: excessive CPU usage during initial read phase on client
3916 Details : During the initial read phase on a client, it would agressively
3917 retry readahead on the file, consuming too much CPU and impacting
3918 performance (since 1.4.5.8). Improve the readahead algorithm
3919 to avoid this, and also improve some other common cases (read
3920 of small files in particular, where "small" is files smaller than
3921 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
3926 Description: MDS crash when receiving packet with unknown intent.
3927 Details : Do not LBUG in unknown intent case, just return -EFAULT
3929 Severity : enhancement
3930 Bugzilla : 9293, 9385
3931 Description: MDS RPCs are serialised on client. This is unnecessary for some.
3932 Details : Do not serialize getattr (non-intent version) and statfs.
3935 Frequency : occasional, when OST network is overloaded/intermittent
3937 Description: client evicted by OST after bulk IO timeout
3938 Details : If a client sends a bulk IO request (read or write) the OST
3939 may evict the client if it is unresposive to its data GET/PUT
3940 request. This is incorrect if the network is overloaded (takes
3941 too long to transfer the RPC data) or dropped the OST GET/PUT
3942 request. There is no need to evict the client at all, since
3943 the pinger and/or lock callbacks will handle this, and the
3944 client can restart the bulk request.
3947 Frequency : Always when mmapping file with no objects
3949 Description: client crashes when mmapping file with no objects
3950 Details : Check that we actually have objects in a file before doing any
3951 operations on objects in ll_vm_open, ll_vm_close and
3957 Description: Request leak when working with deleted CWD
3958 Details : Introduce advanced request refcount tracking for requests
3959 referenced from lustre intent.
3961 Severity : Enhancement
3963 Description: Cache open file handles on client.
3964 Details : MDS now will return special lock along with openhandle, if
3965 requested and client is allowed to hold openhandle, even if unused,
3966 until such a lock is revoked. Helps NFS a lot, since NFS is opening
3967 closing files for every read/write openration.
3969 Severity : Enhancement
3971 Description: Cache open negative dentries on client when possible.
3972 Details : Guard negative dentries with UPDATE lock on parent dir, drop
3973 negative dentries on lock revocation.
3978 Description: Remounting a client read-only wasn't possible with a zconf mount
3979 Details : It wasn't possible to remount a client read-only with llmount.
3981 Severity : enhancement
3982 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
3983 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
3984 MPICH is included to add Lustre-specific ADIO interfaces.
3985 This is based closely on the UFS ADIO layer and only differs
3986 in file creation, in order to allow the OST striping to be set.
3987 This is user-contributed code and not supported by CFS.
3992 Description: extended inode attributes (immutable, append-only) work improperly
3993 when 2.4 and 2.6 kernels are used on client/server or vice versa
3994 Details : Introduce kernel-independent values for these flags.
3996 Severity : enhancement
3999 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
4000 Details : Many of the /proc/ tunables can only be tuned at a megabyte
4001 granularity. Now, Fractional MB granularity is be supported,
4002 this is very useful for low memory system.
4004 Severity : enhancement
4006 Description: Getattr by fid
4007 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
4008 locks, avoids extra getattr rpc requests to MDS, allows '/' to
4009 have locks and avoids getattr rpc requests for it on every stat.
4012 Frequency : Always, for filesystems larger than 2TB
4014 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4015 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
4016 filesystems larger than 2TB to be mounted with mballoc enabled.
4019 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
4021 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4022 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
4023 filesystems larger than 2TB is mounted then the kernel will
4024 silently corrupt the start of the filesystem. CONFIG_LBD is
4025 enabled for all CFS-supported kernels, but the possibility of
4026 this happening with a modified kernel config exists.
4028 Severity : enhancement
4030 Description: add client O_DIRECT support for 2.6 kernels
4031 Details : It is now possible to do O_DIRECT reads and writes to files
4032 in the Lustre client mountpoint on 2.6 kernel clients.
4034 Severity : enhancement
4036 Description: parallel glimpse, setattr, statfs, punch, destroy requests
4037 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
4038 parallel, not waiting for response from every OST before sending
4039 a rpc to the next OST.
4044 Description: setattr vs write race when updating file timestamps
4045 Details : Client processes that update a file timestamp into the past
4046 right after writing to the file (e.g. tar) it is possible that
4047 the updated file modification time can be reset to the current
4048 time due to a race between processing the setattr and write RPC.
4050 Severity : enhancement
4052 Description: Bring 'lfs find' closer in line with regular Linux find.
4053 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
4054 -print0 options and obtains all the needed info through the lustre
4057 Severity : enhancement
4059 Description: support up to 1024 configured devices on one node
4060 Details : change obd_dev array from statically allocated to dynamically
4061 allocated structs as they are first used to reduce memory usage
4066 Description: Flush dirty partially truncated pages during truncate
4067 Details : Immediatelly flush partially truncated pages in filter_setattr,
4068 this way we completely avoid having any pages in page cache on OST
4069 and can retire ugly workarounds during writes to flush such pages.
4074 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
4075 Details : protect inode from truncation within vfs_unlink() context
4076 just take a reference before calling vfs_unlink() and release it
4077 when parent's i_sem is free.
4082 Description: last_id value checked outside lock on OST caused LASSERT failure
4083 Details : If there were multiple MDS->OST object precreate requests in
4084 flight, it was possible that the OST's last object id was checked
4085 outside a lock and incorrectly tripped an assertion. Move checks
4086 inside locks, and discard old precreate requests.
4089 Frequency : always, if extents are used on OSTs
4091 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
4092 Details : OSTs using the extents format would not zero the high 16 bits of
4093 the index physical block number. This is not a problem for any
4094 OST filesystems smaller than 16TB, and no kernels support ext3
4095 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
4096 new/modified files) and can be fixed for existing filesystems
4097 with e2fsprogs-1.39-cfs1.
4102 Description: import connection selection may be incorrect if timer wraps
4103 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
4104 import connections to be ignored if the 32-bit jiffies counter
4105 wraps. Use a 64-bit jiffies counter.
4108 Frequency : very large clusters immediately after boot
4110 Description: LNET request buffers exhausted under heavy short-term load
4111 Details : If a large number of client requests are generated on a service
4112 that has previously never seen so many requests it is possible
4113 that the request buffer growth cannot keep up with the spike in
4114 demand. Instead of dropping incoming requests, they are held in
4115 the LND until the RPC service can accept more requests.
4118 Frequency : Sometimes during replay
4120 Description: Assertion failure in ll_local_open after replay.
4121 Details : If replay happened on an open request reply before we were able
4122 to set replay handler, reply will become not swabbed tripping the
4123 assertion in ll_local_open. Now we set the handler right after
4124 recognising of open request
4127 Frequency : very rare
4129 Description: kernel reports "badness in vsnprintf"
4130 Details : Reading from the "recovery_status" /proc file in small chunks
4131 may cause a negative length in lprocfs_obd_rd_recovery_status()
4132 call to vsnprintf() (which is otherwise harmless). Exit early
4133 if there is no more space in the output buffer.
4135 Severity : enhancement
4137 Description: clear OBD RPC statistics by writing to them
4138 Details : It is now possible to clear the OBD RPC statistics by writing
4139 to the "stats" file.
4144 Description: Client mtime is not the same on different clients after utimes
4145 Details : In some cases, the client was using the utimes() syscall on
4146 a file cached on another node. The clients now validate the
4147 ctime from the MDS + OSTs to determine which one is right.
4152 Description: Inability to activate failout mode
4153 Details : lconf script incorrectly assumed that in pythong string's numeric
4154 value is used in comparisons.
4157 Frequency : always with multiple stripes per file
4159 Description: Inefficient object allocation for mutli-stripe files
4160 Details : When selecting which OSTs to stripe files over, for files with
4161 a stripe count that divides evenly into the number of OSTs,
4162 the MDS is always picking the same starting OST for each file.
4163 Return the OST selection heuristic to the original design.
4168 Description: mount failures may take full timeout to return an error
4169 Details : Under some heavy load conditions it is possible that a
4170 failed mount can wait for the full obd_timeout interval,
4171 possibly several minutes, before reporting an error.
4172 Instead return an error as soon as the status is known.
4174 Frequency : quota enabled and large files being deleted
4176 Description: releasing more than 4GB of quota at once hangs OST
4177 Details : If a user deletes more than 4GB of files on a single OST it
4178 will cause the OST to spin in an infinite loop. Release
4179 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4184 Description: statfs data retrieved from /proc may be stale or zero
4185 Details : When reading per-device statfs data from /proc, in the
4186 {kbytes,files}_{total,free,avail} files, it may appear
4187 as zero or be out of date.
4190 Frequency : systems with MD RAID1 external journal devices
4192 Description: lconf's call to blkid is confused by RAID1 journal devices
4193 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4194 of returning all block devices that match the journal UUID.
4197 Frequency : always, for aggregate stripe size over 4GB
4199 Description: assertion fails when trying to use 4GB stripe size
4200 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
4201 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
4204 Frequency : always on ppc64
4206 Description: the first write on an ext3 filesystem with mballoc got stuck
4207 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4208 perform endianness conversion.
4210 ------------------------------------------------------------------------------
4212 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4214 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
4215 INTEROPERATE with older versions automatically. Please read the
4216 user documentation before upgrading any part of a live system.
4217 * WARNING: Lustre networking configuration changes are required with
4218 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
4221 * Support for kernels:
4222 2.6.9-22.0.2.EL (RHEL 4)
4223 2.6.5-7.244 (SLES 9)
4224 2.6.12.6 vanilla (kernel.org)
4227 Severity : enhancement
4228 Bugzilla : 7981/8208
4229 Description: Introduced Lustre Networking (LNET)
4230 Details : LNET is new networking infrastructure for Lustre, it includes
4231 a reorganized network configuration mode (see the user
4232 documentation for full details) as well as support for routing
4233 between different network fabrics. Lustre Networking Devices
4234 (LNDs) for the supported network fabrics have also been
4235 created for this new infrastructure.
4237 Severity : enhancement
4238 Description: Introduced Access control lists
4239 Details : clients can set ACLs on files and directories in order to have
4240 more fine-grained permissions than the standard Unix UGO+RWX.
4241 The MDS must be started with the "-o acl" mount option.
4243 Severity : enhancement
4244 Description: Introduced filesystem quotas
4245 Details : Administrators may now establish per-user quotas on the
4248 Severity : enhancement
4250 Description: Configuration change for the XT3
4251 The PTLLND is now used to run Lustre over Portals on the XT3
4252 The configure option(s) --with-cray-portals are no longer used.
4253 Rather --with-portals=<path-to-portals-includes> is used to
4254 enable building on the XT3. In addition to enable XT3 specific
4255 features the option --enable-cray-xt3 must be used.
4260 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
4261 Details : A race between allocating a new llcd and re-getting the llcd_lock
4262 allowed another thread to grab newly-allocated llcd.
4264 Severity : enhancement
4266 Description: 2.6 OST async journal commit and locking fix to improve performance
4267 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
4268 2.6 kernels are now async as they already were in 2.4 kernels so
4269 that they can commit concurrently with the network bulk transfer.
4270 For block-allocated files the filter allocation semaphore is held
4271 to avoid filesystem fragmentation during allocation. BKL lock
4272 removed for 2.6 xattr operations where it is no longer needed.
4277 Description: lconf incorrectly determined whether two IP networks could talk
4278 Details : In some more complicated routing and multiple-network
4279 configurations, lconf will avoid trying to make a network
4280 connection to a disjoint part of the IP space. It was doing the
4281 math incorrectly for one set of cases.
4286 Description: Fix for potential infinite loop processing records in an llog.
4287 Details : If an llog record is corrupted/zeroed, it is possible to loop
4288 forever in llog_process(). Validate the llog record length
4289 and skip the remainder of the block on error.
4292 Frequency : occasional (liblustre only)
4294 Description: liblustre could not open files whose last component is a symlink
4295 Details : sysio_path_walk() would incorrectly pass the open intent to
4296 intermediate path components.
4299 Frequency : rare (liblustre only with non-standard tuning)
4300 Bugzilla : 7201 (7350)
4301 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4302 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4303 after it has been released, only do this for referenced locks
4305 Severity : enhancement
4307 Description: specifying an (invalid) directory default stripe_size of -1
4308 would reset the directory default striping
4309 Details : stripe_size -1 was used internally to signal directory stripe
4310 removal, now use "all default" to signal dir stripe removal
4311 as a directory striping of "all default" is not useful
4314 Frequency : common for large clusters running liblustre clients
4316 Description: doing an ls when liblustre clients are running is slow
4317 Details : sending a glimpse AST to a liblustre client waits for every AST
4318 to time out, as liblustre clients will not respond. Since they
4319 cannot cache data we refresh the OST lock LVB from disk instead.
4321 Severity : enhancement
4323 Description: doing an ls at the same time as file IO can be slow
4324 Details : enqueue and other "small" requests can be blocked behind many
4325 large IO requests. Create a new OST IO portal for non-IO
4326 requests so they can be processed faster.
4329 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4331 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4332 created file with mode 000
4333 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4334 allow an ftruncate() as a non-root user to a file with mode 000.
4335 Lustre now respects this flag to disable mode checks when
4336 truncating a file owned by the user
4339 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4342 Description: Revoking locks from clients that went dead or catatonic might take
4344 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4345 cancellation of such locks instant on servers without waiting for
4346 any reply from clients. Clients drops these locks when cancel
4347 notification from server is received without replying.
4350 Frequency : liblustre-only, when liblustre client dies or becomes busy
4352 Description: Doing ls on Linux clients can take a long time with active
4354 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4355 granting such locks to it in the first place if possible. Locks
4356 are taken by proxy on the OST during the read or write and
4357 dropped immediately afterward. Add connect flags handling, do
4358 not grant locks to liblustre clients for glimpse ASTs.
4360 Severity : enhancement
4362 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4363 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4364 sequential reads and behaves badly with applications doing random
4365 reads. Improve it by reading ahead at least read region, and
4366 avoiding excessive large RPC for small reads.
4368 Severity : enhancement
4370 Description: Creating more than 1000 files for a single job may cause a load
4371 imbalance on the OSTs if there are also a large number of OSTs.
4372 Details : qos_prep_create() uses an OST index reseed value that is an
4373 even multiple of the number of available OSTs so that if the
4374 reseed happens in the middle of the object allocation it will
4375 still utilize the OSTs as uniformly as possible.
4380 Description: OST or MDS may oops in ping_evictor_main()
4381 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
4382 but doesn't restart at beginning of obd_exports_timed list
4385 Severity : enhancement
4387 Description: improve by-nid export eviction on the MDS and OST
4388 Details : allow multiple exports with the same NID to be evicted at one
4389 time without re-searching the exports list.
4392 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
4394 Description: MDS may oops in groups_free()
4395 Details : in rare race conditions a newly allocated group_info struct is
4396 freed again, and this can be NULL. The 2.4 compatibility code
4397 for groups_free() checked for a NULL pointer, but 2.6 did not.
4400 Frequency : common for liblustre clients doing little filesystem IO
4401 Bugzilla : 9352, 7313
4402 Description: server may evict liblustre clients accessing contended locks
4403 Details : if a client is granted a lock or receives a completion AST
4404 with a blocking AST already set it would not reply to the AST
4405 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
4408 Frequency : lfs setstripe, only systems with more than 160 OSTs
4410 Description: unable to set striping with a starting offset beyond OST 160
4411 Details : llapi_create_file() incorrectly limited the starting stripe
4412 index to the maximum single-file stripe count.
4415 Frequency : LDAP users only
4417 Description: lconf did not handle in-kernel recovery with LDAP properly
4418 Details : lconf/LustreDB get_refs() is searching the wrong namespace
4420 Severity : enhancement
4422 Description: bind OST threads to NUMA nodes to improve performance
4423 Details : all OST threads are uniformly bound to CPUs on a single NUMA
4424 node and do their allocations there to localize memory access
4426 Severity : enhancement
4428 Description: llmount can determine client NID directly from Myrinet (GM)
4429 Details : the client NID code from gmnalnid was moved directly into
4430 llmount, removing the need to use this or specifying the
4431 client NID explicitly when mounting GM clients with zeroconf
4434 Frequency : if client is started with down MDS
4436 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
4437 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
4438 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
4439 ptlrpc_import_delayed_req() because we didn't check if the
4440 request was interrupted, and we also didn't break out of the
4441 event loop if there was no timeout
4446 Description: data loss during non-page-aligned writes to a single file from
4447 both multiple nodes and multiple threads on one node at same time
4448 Details : updates to KMS and lsm weren't protected by common lock. Resulting
4449 inconsistency led to false short-reads, that were cached and later
4450 used by ->prepare_write() to fill in partially written page,
4451 leading to data loss.
4454 Frequency : always, if lconf --abort_recovery used
4456 Description: lconf --abort_recovery fails with 'Operation not supported'
4457 Details : lconf was attempting to abort recovery on the MDT device and not
4460 Severity : enhancement
4462 Description: remove cleanup logs
4463 Details : replace lconf-generated cleanup logs with lustre internal
4464 cleanup routines. Eliminates the need for client-cleanup and
4467 Severity : enhancement
4469 Description: add support for EAs (user and system) on lustre filesystems
4470 Details : it is now possible to store extended attributes in the Lustre
4471 client filesystem, and with the user_xattr mount option it
4472 is possible to allow users to store EAs on their files also
4474 Severity : enhancement
4476 Description: Add possibility (config option) to show minimal available OST free
4478 Details : When compiled with --enable-mindf configure option, statfs(2)
4479 (and so, df) will return least minimal free space available from
4480 all OSTs as amount of free space on FS, instead of summary of
4481 free spaces of all OSTs.
4483 Severity : enhancement
4485 Description: do not expand extent locks acquired on OST-side
4486 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
4487 by server: they are not cached anyway.
4490 Frequency : when mmap is used/binaries executed from Lustre
4492 Description: Unmmap pages before throwing them away from read cache.
4493 Details : llap_shrink cache now attempts to unmap pages before discarding
4494 them (if unmapping failed - do not discard). SLES9 kernel has
4495 extra checks that trigger if this unmapping is not done first.
4500 Description: lconf didn't resolve symlinks before checking to see whether a
4501 given mountpoint was already in use
4504 Frequency : when migrating failover services
4505 Bugzilla : 6395, 9514
4506 Description: When migrating a subset of services from a node (e.g. failback
4507 from a failover service node) the remaining services would
4508 time out and evict clients.
4509 Details : lconf --force (implied by --failover) sets the global obd_timeout
4510 to 5 seconds in order to quickly disconnect, but this caused
4511 other RPCs to time out too quickly. Do not change the global
4512 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
4514 Severity : enhancement
4515 Frequency : if MDS is started with down OST
4516 Bugzilla : 9439,5706
4517 Description: Allow startup/shutdown of an MDS without depending on the
4518 availability of the OSTs.
4519 Details : Asynchronously call mds_lov_synchronize during MDS startup.
4520 Add appropriate locking and lov-osc refcounts for safe
4521 cleaning. Add osc abort_inflight calls in case the
4522 synchronize never started.
4525 Frequency : occasional (Cray XT3 only)
4527 Description: root not authorized to access files in CRAY_PORTALS environment
4528 Details : The client process capabilities were not honoured on the MDS in
4529 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
4530 been accessed by an authorized user then root was able to access
4531 the file on the local client also. The root user capabilities
4532 are now allowed on the MDS, as this environment has secure UID.
4535 Frequency : occasional
4537 Description: ldiskfs "too long searching" message happens too often
4538 Details : A debugging message (otherwise harmless) prints too often on
4539 the OST console. This has been reduced to only happen when
4540 there are fragmentation problems on the filesystem.
4545 Description: Division by zero in statfs when all OSCs are inactive
4546 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
4547 lov_statfs divides by value returned by lov_get_stripecnt().
4551 Bugzilla : 9489, 3273
4552 Description: First write from each client to each OST was only 4kB in size,
4553 to initialize client writeback cache, which caused sub-optimal
4554 RPCs and poor layout on disk for the first writen file.
4555 Details : Clients now request an initial cache grant at (re)connect time
4556 and so that they can start streaming writes to the cache right
4557 away and always do full-sized RPCs if there is enough data.
4558 If the OST is rebooted the client also re-establishes its grant
4559 so that client cached writes will be honoured under the grant.
4564 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
4565 Details : Now I/O RPCs go to different portal number and (presumably) fast
4566 lock requests (and glimses) and other RPCs get their own service
4567 threads pool that should be able to service those RPCs
4570 Severity : enhancement
4572 Description: Ability to exchange lustre version between client and servers and
4573 issue warnings at client side if client is too old. Also for
4574 liblustre clients there is ability to refuse connection of too old
4576 Details : New 'version' field is added to connect data structure that is
4577 filled with version info. That info is later checked by server and
4581 Frequency : rare, liblustre only.
4582 Bugzilla : 9296, 9581
4583 Description: Two simultaneous writes from liblustre at offset within same page
4584 might proceed at the same time overwriting eachother with stale
4586 Details : I/O lock withing llu_file_prwv was released too early, before data
4587 actually was hitting the wire. Extended lock-holding time until
4588 server acknowledges receiving data.
4591 Frequency : extremely rare. Never observed in practice.
4593 Description: avoid generating lustre_handle cookie of 0.
4594 Details : class_handle_hash() generates handle cookies by incrementing
4595 global counter, and can hit 0 occasionaly (this is unlikely, but
4596 not impossible, because initial value of cookie counter is
4597 selected randonly). Value of 0 is used as a sentinel meaning
4598 "unassigned handle" --- avoid it. Also coalesce two critical
4599 sections in this function into one.
4601 Severity : enhancement
4603 Description: allow liblustre clients to delegate truncate locking to OST
4604 Details : To avoid overhead of locking, liblustre client instructs OST to
4605 take extent lock in ost_punch() on client's behalf. New connection
4606 flag is added to handle backward compatibility.
4608 Severity : enhancement
4609 Bugzilla : 4928, 7341, 9758
4610 Description: allow number of OST service threads to be specified
4611 Details : a module parameter allows the number of OST service threads
4612 to be specified via "options ost ost_num_threads={N}" in the
4613 OSS's /etc/modules.conf or /etc/modprobe.conf.
4617 Bugzilla : 6146, 9635, 9895
4618 Description: servers crash with bad pointer in target_handle_connect()
4619 Details : In rare cases when a client is reconnecting it was possible that
4620 the connection request was the last reference for that export.
4621 We would temporarily drop the export reference and get a new
4622 one, but this may have been the last reference and the export
4623 was just destroyed. Get new reference before dropping old one.
4625 Severity : enhancement
4626 Frequency : if client is started with failover MDS
4628 Description: Allow multiple MDS hostnames in the mount command
4629 Details : Try to read the configuration from all specified MDS
4630 hostnames during a client mount in case the "primary"
4633 Severity : enhancement
4635 Description: Stop sending data to evicted clients as soon as possible.
4636 Details : Check if the client we are about to send or are sending data to
4637 was evicted already. (Check is done every second of waiting,
4638 for which l_wait_event interface was extended to allow checking
4639 of exit condition at specified intervals).
4642 Frequency : rare, normally only when NFS exporting is done from client
4644 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
4646 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
4649 Frequency : always, if srand() is called before liblustre initialization
4651 Description: Liblustre uses system PRNG disturbing its usage by user application
4652 Details : Introduce internal to lustre fast and high-quality PRNG for
4653 lustre usage and make liblustre and some other places in generic
4654 lustre code to use it.
4656 Severity : enhancement
4657 Bugzilla : 9477, 9557, 9870
4658 Description: Verify that the MDS configuration logs are updated when xml is
4659 Details : Check if the .xml configuration logs are newer than the config
4660 logs stored on the MDS and report an error if this is the case.
4661 Request --write-conf, or allow starting with --old_conf.
4663 Severity : enhancement
4665 Description: Handle symlinks in the path when checking if Lustre is mounted.
4666 Details : Resolve intermediate symlinks when checking if a client has
4667 mounted a filesystem to avoid duplicate client mounts.
4672 Description: lconf can hit an error exception but still return success.
4673 Details : The lconf command catches the Command error exception at the top
4674 level script context and will exit with the associated exit
4675 status, but doesn't ensure that this exit status is non-zero.
4680 Description: failure of ptlrpc thread startup can cause oops
4681 Details : Starting a ptlrpc service thread can fail if there are a large
4682 number of threads or the server memory is very fragmented.
4683 Handle this without oopsing.
4686 Frequency : always, only if liblustre and non-default acceptor port was used
4688 Description: liblustre cannot connect to servers with non-default acceptor port
4689 Details : tcpnal_set_default_params() was not called and was therefore
4690 ignoring the environment varaible TCPNAL_PORT, as well as other
4691 TCPNAL_ environment variables
4696 Description: two objects could be created on the same OST for a single file
4697 Details : If an OST is down, in some cases it was possible to create two
4698 objects on a single OST for a single file. No problems other
4699 than potential performance impact and spurious error messages.
4703 Bugzilla : 5681, 9562
4704 Description: Client may oops in ll_unhash_aliases
4705 Details : Client dcache may become inconsistent in race condition.
4706 In some cases "getcwd" can fail if the current directory is
4712 Description: Inode refcounting problems in NFS export code
4713 Details : link_raw functions used to call d_instantiate without obtaining
4714 extra inode reference first.
4718 Bugzilla : 9942, 9903
4719 Description: Referencing freed requests leading to crash, memleaks with NFS.
4720 Details : We used to require that call to ll_revalidate_it was always
4721 followed by ll_lookup_it. Also with revalidate_special() it is
4722 possible to call ll_revalidate_it() twice for the same dentry
4723 even if first occurence returned success. This fix changes semantic
4724 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
4725 reference on a request referred from the intent.
4726 ll_intent_release() then releases such a request.
4729 Frequency : rare, normally benchmark loads only
4731 Description: unlinked inodes were kept in memory on the client
4732 Details : If a client is repeatedly creating and unlinking files it
4733 can accumulate a lot of stale inodes in the inode slab cache.
4734 If there is no other client load running this can cause the
4735 client node to run out of memory. Instead flush old inodes
4736 from client cache that have the same inode number as a new inode.
4739 Frequency : SLES9 2.6.5 kernel and long filenames only
4740 Bugzilla : 9969, 10379
4741 Description: utime reports stale NFS file handle
4742 Details : SLES9 uses out-of-dentry names in some cases, which confused
4743 the lustre dentry revalidation. Change it to always use the
4747 Frequency : rare, unless heavy write-truncate concurrency is continuous
4748 Bugzilla : 4180, 6984, 7171, 9963, 9331
4749 Description: OST becomes very slow and/or deadlocked during object unlink
4750 Details : filter_destroy() was holding onto the parent directory lock
4751 while truncating+unlinking objects. For very large objects this
4752 may block other threads for a long time and slow overall OST
4753 responsiveness. It may also be possible to get a lock ordering
4754 deadlock in this case, or run out of journal credits because of
4755 the combined truncate+unlink. Solution is to do object truncate
4756 first in one transaction without parent lock, and then do the
4757 final unlink in a new transaction with the parent lock. This
4758 reduces the lock hold time dramatically.
4761 Frequency : rare, 2.4 kernels only
4763 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
4764 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
4765 before unmounting their filesystem in order to avoid deadlock.
4766 The kernel_locked() function in 2.4 kernels only checks whether
4767 the kernel lock is held, not whether it is this process that is
4768 holding it as 2.6 kernels do.
4773 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
4774 Details : The client ptlrpc code may be trying to reconnect to a down
4775 server before a previous connection attempt has timed out.
4776 Increase the reconnect interval to be longer than the connection
4777 timeout interval to avoid sending duplicate connections to
4781 Frequency : echo_client brw_test command
4783 Description: fix echo_client to work with OST preallocated code
4784 Details : OST preallocation code (5137) didn't take echo_client IO path
4785 into account: echo_client calls filter methods outside of any
4786 OST thread and, hence, there is no per-thread preallocated
4787 pages and buffers to use. Solution: hijack pga pages for IO. As
4788 a byproduct, this avoids unnecessary data copying.
4792 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
4793 Description: Client can oops in mdc_commit_close() after open replay
4794 Details : It was possible for the MDS to return an open request with no
4795 transaction number in mds_finish_transno() if the client was
4796 evicted, but without actually returning an error. Clients
4797 would later try to replay that open and may trip an assertion
4798 Simplify the client close codepath, and always return an error
4799 from the MDS in case the open is not successful.
4802 Frequency : rare, 2.6 OSTs only
4804 Description: OST may deadlock under high load on fragmented files
4805 Details : If there was a heavy load and highly-fragmented OST filesystems
4806 it was possible to have all the OST threads deadlock waiting on
4807 allocation of biovecs, because the biovecs were not released
4808 until the entire RPC IO was completed. Instead, release biovecs
4809 as soon as they are complete to ensure forward IO progress.
4811 Severity : enhancement
4813 Description: Support for specifying external journal device at mount
4814 Details : If an OST or MDS device is formatted with an external journal
4815 device, this device major/minor is stored in the ext3 superblock
4816 and may not be valid for failover. Allow detecting and
4817 specifying the external journal at mount time.
4822 Description: Mounting an MDS with pending unlinked files may cause oops
4823 Details : target_finish_recovery() calls mds_postrecov() which returned
4824 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
4825 considers this an error and immediately begins cleaning up the
4826 lov, just after starting the mds_lov process
4828 Severity : enhancement
4830 Description: Implement 'lfs df' to report actual free space on per-OST basis
4831 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
4832 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
4833 usage of objects; '-h' to report in human readable format.
4835 ------------------------------------------------------------------------------
4837 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
4844 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
4845 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
4846 at startup during orphan unlinking. Instead, skip buddy update
4847 before setup, as it will be regenerated after recovery is complete.
4852 Description: If an OST is inactive, its locks might reference stale inodes.
4853 Details : lov_change_cbdata() must iterate over all namespaces, even if
4854 they are inactive to clear inode references from the lock.
4856 Severity : enhancement
4857 Frequency : occasional, if non-standard max_dirty_mb used
4859 Description: Client will block write RPCs if not enough grant
4860 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
4861 then the client will block writes while waiting for another RPC
4862 to complete instead of consuming its dirty limit. With change
4863 we get improved performance when max_dirty_mb is small.
4865 Severity : enhancement
4866 Bugzilla : 3389, 6253
4867 Description: Add support for supplementary groups on the MDS.
4868 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
4869 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
4870 lookups for user supplementary groups into a cache.
4874 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
4875 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
4876 for liblustre O_EXCL is enforced by MDS.
4880 Description: readdir never returns NULL in liblustre.
4881 Details : Corrected llu_iop_getdirentries logic, to return offset of next
4882 dentry in struct dirent.
4886 Frequency : liblustre only, depends on application IO pattern
4887 Description: liblustre clients evicted if not contacting servers
4888 Details : Don't put liblustre clients into the ping_evictor list, so
4889 they will not be evicted by the pinger ever.
4891 Severity : enhancement
4893 Description: Add ability to evict clients by NID from MDS.
4894 Details : By echoing "nid:$NID" string into
4895 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
4896 $NID would be instantly evicted from this MDS and from all active
4897 OSTs connected to it.
4901 Description: Do not query file size twice, somewhat slowing stat(2) calls.
4902 Details : lookup_it_finish() used to query file size from OSTs that was not
4907 Description: service threads change working directory to that of init
4908 Details : Starting lustre service threads may pin the working directory
4909 of the parent thread, making that filesystem busy. Threads
4910 now change to the working directory of init to avoid this.
4914 Frequency : during shutdown only
4915 Description: shutdown with a failed MDS or OST can cause unmount to hang
4916 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
4921 Frequency : 2.6 only
4922 Description: chmod/chown may include an extra supplementary group
4923 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
4924 supplementary group and if none is specified this is used.
4927 Bugzilla : 5479 (6816)
4929 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
4930 Details : If lookup is in progress on a file that is unlinked we might try
4931 to revalidate the inode and fail in revalidate after lookup is
4932 complete and ll_file_open() enqueues the open again but
4933 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
4936 Frequency : always, if lconf --abort_recovery used
4938 Description: lconf --abort_recovery fails with 'Operation not supported'
4939 Details : lconf was attempting to abort recovery on the MDT device and not
4942 ------------------------------------------------------------------------------
4944 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
4949 Frequency : rare (only unsupported configurations with a node running as an
4951 Bugzilla : 6514, 5137
4952 Description: Mounting a Lustre file system on a node running as an OST could
4954 Details : OSTs now preallocates memory needed to write out data at
4955 startup, instead of when needed, to avoid having to
4956 allocate memory in possibly low memory situations.
4957 Specifically, if the file system is mounted on on OST,
4958 memory pressure could force it to try to write out data,
4959 which it needed to allocate memory to do. Due to the low
4960 memory, it would be unable to do so and the node would
4961 become unresponsive.
4963 Severity : enhancement
4965 Description: Addition of lconf --service command line option
4966 Details : lconf now accepts a '--service <arg>' option, which is
4967 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
4969 Severity : enhancement
4971 Description: Failover mode is now the default for OSTs.
4972 Details : By default, OSTs will now run in failover mode. To return to
4973 the old behaviour, add '--failout' to the lmc line for OSTs.
4975 Severity : enhancement
4977 Description: Health checks are now provided for MDS and OSTs
4978 Details : Additional detailed health check information on MSD and OSTs
4979 is now provided through the procfs health_check value.
4982 Frequency : occasional, depends on IO load
4984 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
4985 after numerous create/delete cycles
4986 Details : The ext3 inode allocation policy would not allocate new inodes
4987 very well on the OSTs because there are no new directories
4988 being created. Instead we look for groups with free space if
4989 the parent directories are nearly full.
4994 Description: Network or server problems during mount may cause partially
4995 mounted clients instead of returning an error.
4996 Details : The config llog parsing code may overwrite the error return
4997 code during mount error handling, returning success instead
5003 Description: MDS can fail to allocate large reply buffers
5004 Details : After long uptimes the MDS can fail to allocate large reply
5005 buffers (e.g. zconf client mount config records) due to memory
5006 fragmentation or consumption by the buffer cache. Preallocate
5007 some large reply buffers so that these replies can be sent even
5008 under memory pressure.
5012 Frequency : rare (liblustre)
5013 Description: fsx running with liblustre complained that using truncate() to
5014 extend the file doesn't work. This patch corrects that issue.
5015 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
5016 ATTR_SIZE and lsm use in llu_setattr_raw.
5020 Frequency : rare, only 2.6 kernels
5021 Description: Unusual file access patterns on the MDS may result in inode
5022 data being lost in very rare circumstances.
5023 Details : Bad interaction between the ea-in-inode patch and the "no-read"
5024 code in the 2.6 kernel caused the inode and/or EA data not to
5025 be read from disk, causing single-file corruption.
5029 Frequency : rare, only 2.6 filesystems using extents
5030 Description: Heavy concurrent write and delete load may cause data corruption.
5031 Details : It was possible under high-load situations to have an extent
5032 metadata block in the block device cache from a just-unlinked
5033 file overwrite a newly-allocated data block. We now unmap any
5034 metadata buffers that alias just-allocated data blocks.
5038 Frequency : filesystems with default stripe_count larger than 77
5039 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
5040 Details : lconf specifies an inode size of 4096 bytes when the default
5041 stripe_count is larger than 77. This conflicts with the default
5042 inode density of 1 per 4096 bytes. Allocate smaller inodes in
5043 this case to avoid pinning too much memory for large EAs.
5045 ------------------------------------------------------------------------------
5047 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
5052 Frequency : rare (extremely heavy IO load with hundreds of clients)
5054 Description: Client is evicted, gets IO error writing to file
5055 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
5056 caused clients to be evicted for AST timeouts. The fixes in
5057 bug 5192 mean we no longer need to have such short AST timeouts
5058 so ldlm_timeout has been increased.
5061 Frequency : occasional during --force or --failover shutdown under load
5062 Bugzilla : 5949, 4834
5063 Description: Server oops/LBUG if stopped with --force or --failover under load
5064 Details : a collection of import/export refcount and cleanup ordering
5065 issues fixed for safer force cleanup
5068 Frequency : only filesystems larger than 120 OSTs
5069 Bugzilla : 5990, 6223
5070 Description: lfs getstripe would oops on a very large filesystem
5071 Details : lov_getconfig used kfree on vmalloc'd memory
5074 Frequency : only filesystems exporting via NFS to Solaris 10 clients
5075 Bugzilla : 6242, 6243
5076 Description: reading from files that had been truncated to a non-zero size
5077 but never opened returned no data
5078 Details : ll_file_read() reads zeros from no-object files to EOF
5083 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
5084 Details : There's little harm in aborting MDS/OSS recovery and letting it
5085 try again, so I removed the LASSERT and return an error instead.
5087 Severity : enhancement
5089 Description: New debugging infrastructure for tracking down data corruption
5090 Details : The I/O checksum code was replaced to: (a) control it at runtime,
5091 (b) cover more of the client-side code path, and (c) try to narrow
5092 down where problems occurred
5096 Bugzilla : 3819, 4364, 4397, 6313
5097 Description: Racing close and eviction MDS could cause assertion in mds_close
5098 Details : It was possible to get multiple mfd references during close and
5099 client eviction, leading to one thread referencing a freed mfd.
5101 Severity: : enhancement
5102 Bugzilla : 3262, 6359
5103 Description: Attempts to reconnect to servers are now more aggressive.
5104 Details : This builds on the enhanced upcall-less recovery that was added
5105 in 1.4.2. When trying to reconnect to servers, clients will
5106 now try each server in the failover group every 10 seconds. By
5107 default, clients would previously try one server every 25 seconds.
5112 Description: After recovery, certain operations trigger a failed
5113 assertion on a client.
5114 Details : Failing over an mds, using lconf -d --failover, while a
5115 client was doing a readdir() call would cause the client to
5116 LBUG after recovery completed and the readdir() was resent.
5118 Severity : enhancement
5120 Description: Default groups are now added by lconf
5121 Details : You can now run lconf --group <servicename> without having to
5122 manually add groups with lmc.
5125 Frequency : occasional
5127 Description: Nodes with an elan id of 0 trigger a failed assertion
5130 Frequency : always when accessing e.g. tty/console device nodes
5132 Description: tty and some other devices nodes cannot be used on lustre
5133 Details : file's private_data field is used by device data and lustre
5134 values in there got lost. New field was added to struct file to
5135 store fs-specific private data.
5138 Frequency : when exporting Lustre via NFS
5140 Description: NFSD failed occasionally when looking up a path component
5141 Details : NFSD is looking up ".." which was broken in ext3 directories
5142 that had grown large enough to become hashed.
5145 Frequency : Clusters with multiple interfaces not on the same subnet
5147 Description: Nodes will repeatedly try to reconnect to an interface which it
5148 cannot reach and report an error to the log.
5149 Details : Extra peer list entries will be created by lconf with some peers
5150 unreachable. lconf now validates the peer before adding it.
5153 Frequency : Only if a default stripe is set on the filesystem root.
5155 Description: Setting a default stripe on the filesystem root prevented the
5156 filesystem from being remounted.
5157 Details : The client was sending extra request flags in the root getattr
5158 request and did not allocate a reply buffer for the dir EA.
5161 Frequency : occasional, higher if lots of files are accessed by one client
5162 Bugzilla : 6159, 6097
5163 Description: Client trips assertion regarding lsm mismatch/magic
5164 Details : While revalidating inodes the VFS looks up inodes with ifind()
5165 and in rare cases can find an inode that is being freed.
5166 The ll_test_inode() code will free the lsm during ifind()
5167 when it finds an existing inode and then the VFS later attaches
5168 this free lsm to a new inode.
5172 Bugzilla : 6422, 7030
5173 Description: MDS deadlock between mkdir and client eviction
5174 Details : Creating a new file via mkdir or mknod (starting a transaction
5175 and getting the ns lock) can deadlock with client eviction
5176 (gets ns lock and trying to finish a synchronous transaction).
5179 Frequency : occasional
5180 Description: While starting a server, the fsfilt_ext3 module could not be
5182 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
5183 kernels. Previously, lconf would still use the ext3 name
5184 when trying to load modules. Now, it will correctly use
5185 ext3 on 2.4 and ldiskfs on 2.6.
5187 Severity : enhancement
5188 Description: The default stripe count has been changed to 1
5189 Details : The interpretation of the default stripe count (0, to lfs
5190 or lmc) has been changed to mean striping across a single
5191 OST, rather than all available. For general usage we have
5192 found a stripe count of 1 or 2 works best.
5194 Severity : enhancement
5195 Description: Add support for compiling against Cray portals.
5196 Details : Conditional compiling for some areas that are different
5200 Frequency : occasional
5201 Bugzilla : 6409, 6834
5202 Description: Creating files with an explicit stripe count may lead to
5203 a failed assertion on the MDS
5204 Details : If some OSTs are full or unavailable, creating files may
5205 trigger a failed assertion on the MDS. Now, Lustre will
5206 try to use other servers or return an error to the
5210 Frequency : occasional
5212 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
5213 caused lock timeout during readahead (since 1.4.2).
5214 Details : Processes doing readahead might match a lock that hasn't been
5215 granted yet if there are overlapping and conflicting lock
5216 requests. The readahead process waits on ungranted lock
5217 (original lock is CBPENDING), while OST waits for that process
5218 to cancel CBPENDING read lock and eventually evicts client.
5220 Severity : enhancement
5222 Description: Initial enabling of flock support for clients
5223 Details : Implements fcntl advisory locking and file status functions.
5224 This feature is provided as an optional mount flag (default
5225 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
5226 locking are implemented yet, and those that are are not guaranteed
5227 to be completely correct in production environments.
5228 mount -t lustre -o [flock|noflock] ...
5231 Frequency : occasional
5233 Description: OSTs running 2.4 kernels but with extents enabled might trip an
5234 assertion in the ext3 JBD (journaling) layer.
5235 Details : The b_committed_data struct is protected by the big kernel lock
5236 in 2.4 kernels, serializing journal_commit_transaction() and
5237 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
5238 access to this struct. In 2.6 kernels there is finer grained
5239 locking to improve SMP performance of the JBD layer.
5243 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
5248 Description: merge and cleanup kernel patches.
5251 -----------------------------------------------------------------------------
5253 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
5255 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
5256 versions of Lustre. Please update all servers and clients to
5257 version 1.4.2 or later at the same time. You must also run
5258 "lconf --write-conf {config}.xml" on the MDS while it is stopped
5259 to update the configuration logs.
5261 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
5262 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
5263 - swab llog records of type '0' so we get proper header size/idx (5861)
5264 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
5265 - fix rename of one directory over another leaking an inode (5953)
5266 - avoid SetPageDirty on 2.6 (5981)
5267 - don't re-add just-being-destroyed locks to the waiting list (5653)
5268 - when creating new directories, inherit the parent's custom
5269 striping settings if present parent (3048)
5270 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
5271 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
5272 - don't hold client locks on temporary worklist from l_lru (5666)
5273 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
5274 - automatically evict dead clients (5921)
5275 - Update file size properly in create+truncate+fstat case (6196)
5276 - Do not unhash mountpoint dentries, do not allow removal of
5278 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
5279 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
5280 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
5281 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
5282 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
5283 - Correctly handle reads of files with no objects (6243)
5284 - lctl recover will also mark a device active if deactivate used (5933)
5286 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
5287 - allow --write-conf on an MDS with different nettype than client (5619)
5288 - don't write config llogs to MDS for mounts not from that MDS (5617)
5289 - lconf should create multiple TCP connections from a client (5201)
5290 - init scripts are now turned off by default; run chkconfig --on
5291 lustre and chkconfig --on lustrefs to use them
5292 - upcalls are no longer needed for clients to recover to failover
5294 - add --abort-recovery option to lconf to abort recovery on device
5296 - add support for an arbitrary number of OSTs (3026)
5297 - Quota support protocol changes.
5298 - forward compatibility changes to wire structs (6007)
5299 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5300 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5301 - improved "lustre status" script
5302 - initialize blocksize for non-regular files (6062)
5303 - added --disable-server and --disable-client configure options (5782)
5304 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5305 - Vanilla 2.4.29 support
5306 - increase maximum number of obd devices to 520 (6242)
5307 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5308 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5309 SLES 9 2.6.5 kernels
5310 - limit stripes per file to 160 (the maximum EA size) (6093)
5312 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5315 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5316 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5317 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5318 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5319 - small ext3 extents cleanups and fixes (5733)
5320 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5321 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5322 - avoid needless client->OST connect, fix handle mismatch (5317)
5323 - fix DLM error path that led to out-of-sync client, long delays (5779)
5324 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5325 - fix several locking issues related to i_size (5492,5624,5654,5672)
5326 - don't move pending lock onto export if it is already evicted (5683)
5327 - fix kernel oops when creating .foo in unlinked directory (5548)
5328 - fix deadlock in obdfilter statistics vs. object create (5811)
5329 - use time_{before,after} to avoid timer jiffies wrap (5882)
5330 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5331 - Do not leak request if server was not able to process it (5154)
5332 - If mds_open unable to find parent dir, make that negative lookup(5154)
5333 - don't create new directories with extent-mapping (5909, 5936)
5335 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5336 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5337 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5338 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5339 - support for RHEL4 kernel on i686 (5773)
5340 - provide error messages when incompatible logs are encountered (5898)
5342 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5343 * version 1.4.0.10 (1.4.1 release candidate 1)
5345 - don't keep a lock reference when lock is not granted (4238)
5346 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5347 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5348 - drop import inflight refcount on signal_completed_replay error (5255)
5349 - unlock page after async write error during send (3677)
5350 - handle missing objects in filter_preprw_read properly (5265)
5351 - no transno return for symlink open, don't save no-trasno open (3440)
5352 - don't try to complete elan receive that already failed (4012)
5353 - free RPC server reply state on error (5406)
5354 - clean up thread from ptlrpc_start_thread() on error (5160)
5355 - readahead could read extra page into cache that wasn't ejected (5388)
5356 - prevent races in class_attach/setup/cleanup/detach (5260)
5357 - don't dereference de->d_inode after l_dput of de (5458)
5358 - use "int" for stripe value returned from lock_to_stripe (5544)
5359 - mballoc allocation and error-checking fixes in 2.6 (5504)
5360 - block device patches to fix I/O request sizes in 2.6 (5482)
5361 - look up hostnames for IB nals (5602)
5362 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5363 - don't start multiple acceptors for the same port (5277)
5364 - fix incorrect LASSERT in mds_getattr_name (5635)
5365 - export a proc file for general "ping" checking (5628)
5366 - fix "lfs check" to not block when the MDS is down (5628)
5368 - service request history (4965)
5369 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5370 - create an "evict_client" /proc entry on OSTs, like the MDS has
5371 - fix mount usage message, return errors per mount(8) (5168)
5372 - change grep [] to grep "[]" in tests so they work in more UMLs
5373 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5374 - remove ancient LOV_MAGIC_V0 EA support (5047)
5375 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5376 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5377 - allow readahead allocations to fail when low on memory (5383)
5378 - mmap locking landed again, after considerable improvement (2828)
5379 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
5381 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
5384 - send OST transaction number in read/write reply to free req (4966)
5385 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5386 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5387 - the watchdog thread now runs as interruptible (5246)
5388 - flock/lockf fixes (but it's still disabled, pending 5135)
5389 - don't use EXT3 constants in llite code (5094)
5390 - memory shortage at startup could cause assertion (5176)
5392 - reorganization of lov code
5393 - single portals codebase
5395 - add extents/mballoc support (5025)
5396 - direct I/O reads in the obdfilter (4048)
5397 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
5399 tbd Cluster File Systems, Inc. <info@clusterfs.com>
5402 - send OST transaction number in read/write reply to free req (4966)
5403 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5404 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5405 - the watchdog thread now runs as interruptible (5246)
5406 - handle missing objects in filter_preprw_read properly (5265)
5407 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5408 - drop import inflight refcount on signal_completed_replay error (5255)
5409 - unlock page after async write error during send (3677)
5410 - return original error code on reconstructed replies (3761)
5411 - no transno return for symlink open, don't save no-trasno open (3440)
5413 - add pid to ldlm debugging output (4922)
5414 - bump the watchdog timeouts -- we can't handle 30sec yet
5415 - extra debugging for orphan dentry/inode bug (5259)
5417 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
5420 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
5421 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
5422 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
5423 - fix NULL dereference in /proc/sys/portals/routes (4827)
5424 - allow failed mdc_close() operations to be interrupted (4561)
5425 - stop precreate on OST before MDS would time out on it (4778)
5426 - don't send partial-page writes before EOF from client (4410)
5427 - discard client grant for sub-page writes on large-page clients (4520)
5428 - don't free dentries not owned by NFS code, check generation (4806)
5429 - fix lsm leak if mds_create_objects() fails (4801)
5430 - limit debug_daemon file size, always print CERROR messages (4789)
5431 - use transno after validating reply (3892)
5432 - process timed out requests if import state changes (3754)
5433 - update mtime on OST during writes, return in glimpse (4829)
5434 - add mkfsoptions to LDAP (4679)
5435 - use ->max_readahead method instead of zapping global ra (5039)
5436 - don't interrupt __l_wait_event() during strace
5438 - add software watchdogs to catch hung threads quickly (4941)
5439 - make lustrefs init script start after nfs is mounted
5440 - fix CWARN/ERROR duplication (4930)
5441 - return async write errors to application if possible (2248)
5442 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
5443 - print NAL number in %x format (4645)
5444 - update barely-supported suse-2.4.21-171 series (4842)
5445 - support for sles 9 %post scripts
5446 - support for building 2.6 kernel-source packages
5447 - support for sles km_* packages
5449 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
5452 - ignore -ENOENT errors in osc_destroy (3639)
5453 - notify osc create thread that OSC is being cleaned up (4600)
5454 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
5455 - reconstruct ost_handle() like mds_handle() (4657)
5456 - create a new thread to do import eviction to avoid deadlock (3969)
5457 - let lconf resolve symlinked-to devices (4629)
5458 - don't unlink "objects" from directory with default EA (4554)
5459 - hold socknal file ref over connect in case target is down (4394)
5460 - allow more than 32000 subdirectories in a single directory (3244)
5461 - fix blocks count for O_DIRECT writes (3751)
5462 - OST returns ENOSPC from object create when no space left (4539)
5463 - don't send truncate RPC if file size isn't changing (4410)
5464 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
5465 - bind to privileged port in socknal and tcpnal (3689)
5467 - rate limit CERROR/CWARN console message to avoid overload (4519)
5468 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
5469 - basic mmap support (3918)
5470 - kernel patch series update from b1_4 (4711)
5472 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
5475 - avoid crash during MDS cleanup with OST shut down (2775)
5476 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
5477 - don't use bad inodes on the MDS (3744)
5478 - dynamic object preallocation to improve recovery speed (4236)
5479 - don't hold spinlock over lock dumping or change debug flags (4401)
5480 - don't zero obd_dev when it is force cleaned (3651)
5481 - print grants to console if they go negative (4431)
5482 - "lctl deactivate" will stop automatic recovery attempts (3406)
5483 - look for existing locks in ldlm_handle_enqueue() (3764)
5484 - don't resolve lock handle twice in recovery avoiding race (4401)
5485 - revalidate should check working dir is a directory (4134)
5487 - don't always mark "slow" obdfilter messages as errors (4418)
5489 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
5492 - don't close LustreDB during write_conf until it is done (3860)
5493 - fix typo in lconf for_each_profile (3821)
5494 - allow dumping logs from multiple threads at one time (3820)
5495 - don't allow multiple threads in OSC recovery (3812)
5496 - fix debug_size parameters (3864)
5497 - fix mds_postrecov to initialize import for llog ctxt (3121)
5498 - replace config semaphore with spinlock (3306)
5499 - be sure to send a reply for a CANCEL rpc with bad export (3863)
5500 - don't allow enqueue to complete on a destroyed export (3822)
5501 - down write_lock before checking llog header bitmap (3825)
5502 - recover from lock replay timeout (3764)
5503 - up llog sem before sending rpc (3652)
5504 - reduce ns lock hold times when setting kms (3267)
5505 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
5506 - fix NULL deref and obd_dev leak on setup error (3312)
5507 - replace some LBUG about llog ops with error handling (3841)
5508 - don't match INVALID dentries from d_lookup and spin (3784)
5509 - hold dcache_lock while marking dentries INVALID and hashing (4255)
5510 - fix invalid assertion in ptlrpc_set_wait (3880)
5512 - add libwrap support for the TCP acceptor (3996)
5513 - add /proc/sys/portals/routes for non-root route listing (3994)
5514 - allow setting MDS UUID in .xml (2580)
5515 - print the stack of a process that LBUGs (4228)
5517 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
5520 - don't cleanup request in ll_file_open() on failed MDS open (3430)
5521 - make sure to unset replay flag from failed open requests (3440)
5522 - if default stripe count is 0, use OST count for inode size (3636)
5523 - update parent mtime/ctime on client for create/unlink (2611)
5524 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
5525 - free recovery state on server during a forced cleanup (3571)
5526 - unregister_reply for resent reqs (3063)
5527 - loop back devices mounting and status check on 2.6 (3563)
5528 - fix resource-creation race that can provoke i_size == 0 (3513)
5529 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
5530 - more debugging for page-accounting assertion (3746)
5531 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
5532 - avoid deadlock after precreation failure (3758)
5533 - fix race and lock order deadlock in orphan handling (3450, 3750)
5534 - add validity checks when grabbing inodes from l_ast_data (3599)
5536 - add /proc/.../recovery_status to obdfilter (3428)
5537 - lightweight CDEBUG infrastructure, debug daemon (3668)
5538 - change default OSC RPC parameters to be better on small clusters
5539 - turn off OST read cache for files smaller than 32MB
5540 - install man pages and include them in rpms (3100)
5541 - add new init script for (un)mounting lustre filesystems (2593)
5542 - run chkconfig in %post for init scripts (3701)
5543 - drop scimac NAL (unmaintained)
5545 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
5548 - clean kiobufs before and after use (3485)
5549 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
5550 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
5551 - change init script to fail more clearly if not run as root (1528)
5552 - allow clients to reconnect during replay (1742)
5553 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
5554 - don't do DNS lookups on NIDs too small for IP addresses (3442)
5555 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
5556 - fix cond_resched (3554)
5557 - only evict unfinished clients after recovery (3515)
5558 - allow bulk resend, prevent data loss (3570)
5559 - dynamic ptlrpc request buffer allocation (2102)
5560 - don't allow unlinking open directory if it isn't empty (2904)
5561 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
5562 - remove extraneous obd dereference causing LASSERT failure (3334)
5563 - don't use get_cycles() when creating temp. files on the mds (3156)
5564 - hold i_sem when setting i_size in ll_extent_lock() (3564)
5565 - handle EEXIST for set-stripe, set proper directory name (3336)
5567 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
5568 - fix ksocknal_fmb_callback() error messages (2918)
5570 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
5573 - don't copy lvb into (possibly NULL) reply on error (2983)
5574 - don't deref dentry after dput, don't free lvb on error (2922)
5575 - use the kms to determine writeback rpc length (2947)
5576 - increment oti_logcookies when osc is inactive (2948)
5577 - update client's i_blocks count via lvb messages (2543)
5578 - handle intent open/close of special files properly (1557)
5579 - mount MDS with errors=remount-ro, like obdfilter (2009)
5580 - initialize lock handle to avoid ASSERT on error cleanup (3057)
5581 - don't use cancelling-locks' kms values (2947)
5582 - use highest lock extent for kms, not last one (2925)
5583 - don't dereference ERR_PTR() dentry in error handling path (3107)
5584 - fix thread race in portals_debug_dumplog() (3122)
5585 - create lprocfs device entries at setup instead of at attach (1519)
5586 - common AST error handler, don't evict client on completion race (3145)
5587 - zero nameidata in detach_mnt in 2.6 (3118)
5588 - verify d_inode after revalidate_special is valid in 2.6 (3116)
5589 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
5590 - initialize RPC timeout timer earlier for 2.6 (3219)
5591 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
5592 - print nal/nid for unknown nid (3258)
5593 - additional checks for oscc recovery before doing precreate (3284)
5594 - fix ll_extent_lock() error return code for 64-bit systems (3043)
5595 - don't crash in mdc_close for bad permissions on open (3285)
5596 - zero i_rdev for non-device files (3147)
5597 - clear page->private before handing to FS, better assertion (3119)
5598 - tune the read pipeline (3236)
5599 - fix incorrect decref of invalidated dentry (2350)
5600 - provide read-ahead stats and refine rpc in flight stats (3328)
5601 - don't hold journal transaction open across create RPC (3313)
5602 - update atime on MDS at close time (3265)
5603 - close LDAP connection when recovering to avoid server load (3315)
5604 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
5605 - don't leak open file on MDS after open resend (3325)
5606 - serialize filter_precreate and filter_destroy_precreated (3329)
5607 - loop device shouldn't call sync_dev() for nul device (3092)
5608 - clear page cache after eviction (2766)
5609 - resynchronize MDS->OST in background (2824)
5610 - refuse to mount the same filesystem twice on same mountpoint (3394)
5611 - allow llmount to create routes for mounting behind routers (3320)
5612 - push lock cancellation to blocking thread for glimpse ASTs (3409)
5613 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
5614 - fix rare problem with rename on htree directories (3417)
5616 - allow default OST striping configuration per directory (1414)
5617 - fix compilation for qswnal for 2.6 kernels (3125)
5618 - increase maximum number of MDS request buffers for large systems
5619 - change liblustreapi to be useful for external progs like lfsck (3098)
5620 - increase local configuration timeout for slow disks (3353)
5621 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
5623 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5626 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
5627 - don't overwrite extent policy data in reply if lock was blocked (2901)
5628 - drop filter export grants atomically with removal from device (2663)
5629 - del obd_self_export from work_list in class_disconnect_exports (2908)
5630 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
5631 - swab reply message in mdc_close, other PPC fixes (2464)
5632 - fix destroying of named logs (2325)
5633 - overwrite old logs when running lconf --write_conf (2264)
5634 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
5635 - fix race in target_handle_connect (2898)
5636 - mds_reint_create() should take same inode create lock (2926)
5637 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
5638 - don't close files for self_export to avoid uninitialized obd (2936)
5639 - allow MDS with the same name as client node (2939)
5640 - hold dentry reference for closed log files for unlink (2325)
5641 - reserve space for all logs during transactions (2059)
5642 - don't evict page beyond end of stripe extent (2925)
5643 - don't oops on a deleted current working directory (2399)
5644 - handle hard links to targets without a parent properly (2517)
5645 - don't dereference NULL lock when racing during eviction (2867)
5646 - don't grow lock extents when lots of conflicting locks (2919)
5648 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
5651 - account for cache space usage on clients to avoid data loss (974)
5652 - lfsck support in lustre kernel code (2349)
5653 - reduce journal credits needed for BRW writes (2370)
5654 - orphan handling to avoid losing space on client/server crashes
5655 - ptlrpcd can be blocked, stopping ALL progress (2477)
5656 - use lock value blocks to assist in proper KMS, faster stat (1021)
5657 - takes i_sem instead of DLM locks internally on obdfilter (2720)
5658 - recovery for initial connections (2355)
5659 - fixes for mds_cleanup_orphans (1934)
5660 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
5661 - block all file creations until orphan recovery completes (1901)
5662 - client remove rq_connection from request struct (2423)
5663 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
5664 - recovery timer race (2670)
5665 - mdc_close recovey bug (2532)
5666 - ptlrpc cleanup bug (2710)
5667 - mds timeout on local locks (2588)
5668 - namespace lock held during RPCs (2431)
5669 - handle interrupted sync write properly (2503)
5670 - don't try to handle a message that hasn't been replied to (2699)
5671 - client assert failure during cleanup after abort recovery (2701)
5672 - leak mdc device after failed mount (2712)
5673 - ptlrpc_check_set allows timedout requests to complete (2714)
5674 - wait for inflight reqs when ptlrpcd finishes (2710)
5675 - make sure unregistered services are removed from the srv_list
5676 - reset bulk XID's when resending them (caught by 1138 test)
5677 - unregister_bulk after timeout
5678 - fix lconf error (2694)
5679 - handle write after unfinished setstripe, stripe-only getstripe (2388)
5680 - readahead locks pages, leaves pending causing memory pressure (2673)
5681 - increase OST request buffers to 4096 on large machines (2729)
5682 - fix up permission of existing directories in simple_mkdir (2661)
5683 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
5684 - don't assign transno to errored transactions (2742)
5685 - don't delete objects on OST if given a bogus objid from MDS (2751)
5686 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
5687 - if rq_no_resend, then timeout request after recovery (2432)
5688 - fix MDS llog_logid record size, 64-bit array alignment (2733)
5689 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
5690 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
5691 - MDS recovery shouldn't delete objects that it has given out (2730)
5692 - if enqueue arrives after completion, don't clobber LVB (2819)
5693 - don't unlock pages twice when trigger_group_io returns error (2814)
5694 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
5695 - don't write pages to disk if there was an error (1450)
5696 - don't ping imports that have recovery disabled (2676)
5697 - take buffered bytes into account when balancing socknal conn (2817)
5698 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
5699 - reconnect unlink llog connection after MDS reconnects to OST (2816)
5700 - remove little-endian swabbing of llog records (1987)
5701 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
5702 - retry reposting request buffers if they fail (1191)
5703 - grow extent at grant time to avoid granting a revoked lock (2809)
5704 - lock revoke doesn't evict page if covered by a second lock (2765)
5705 - disable VM readahead to avoid reading outside lock extents (2805)
5707 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
5708 - updated kernel patches for hp-2.4.20 kernel (2681)
5710 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
5713 - fix truncated write corruption (2366)
5714 - fix for failed assertion in iopen_connect_dentry (1792,2517)
5716 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
5717 - ensure len(uuid) < 37 in lmc (1171)
5718 - fix ia64 OOPS in llog_test (2255)
5719 - zero end of page at obdfilter for partial page writes (2648)
5720 - don't leave stale dentries around after renames (bug 2428)
5721 - fix timeouts when evicting a client with a single lock held (2642)
5722 - set deadline for the initial HELLO message to drain (2634)
5723 - print out dotted-quad IP addresses in the socknal (2302)
5725 - additional debugging for MDS client eviction problem (2443)
5726 - fix mkfsoptions support for osts (2603, 2604)
5728 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
5731 - add series for the vanilla 2.6.0 kernel
5732 - add series for the vanilla 2.4.24 kernel
5733 - add series for a cray x86/64 UL kernel drop
5734 - fix xattr patches for the vanilla 2.4.19 series
5736 - generate true UUIDs in lmc (1171)
5737 - have portals stack dumping break in UML (2466)
5738 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
5739 - allocate with _NOFS in ldlm to avoid deadlock (1933)
5740 - wake callback waiting threads on client eviction (2460)
5741 - Add --ptldebug and --subsystem to lmc (1719)
5742 - update assertion to allow safe interrupt allocation
5743 - set rq_no_resend for cancel requests (2432)
5744 - recalculate ptlrpcd timeout after resend (2494)
5745 - call vfs_rmdir when removing pending directories (2368)
5746 - fix renaming a file to itself (2429)
5747 - lmc creates a default one-stripe lov (2454)
5748 - expand procfs space to handle large clusters (2326)
5749 - increase UML stack to avoid overflow
5750 - update lconf's list of debug and subsystem masks
5751 - fix lfs find --obd (2510)
5752 - /proc tunable for disabling filter read caching (2591)
5753 - stop rpm packages from altering slapd.conf (2301)
5754 - disable nagle in the socknal under 0conf (2578)
5755 - choose mds inode size based on stripe count (2572)
5756 - fix kernel-source rpm problems (2516)
5758 - add --disable-doc to avoid pdf generation (2421)
5759 - update documentation, tests, type-os, comments
5760 - avoid format warnings on ia64
5761 - remove the TOE NAL
5762 - tiny code cleanups by removing unused fields
5764 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
5767 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
5768 - remove the most blatant lies from BUILDING file (2371)
5769 - change default debug level to reasonable production setting
5770 - reduce client side cache size to reduce cache flush time
5771 - reduce max RPCs in flight to avoid unnecessary file fragmentation
5772 - make TCP zerocopy and pinger support enabled by default (2476)
5773 - sync writes completed after process exits caused crashes (2319)
5774 - maintain correct mount count on the MDS (2356)
5775 - backout 1557, because 2316 wasn't really fixed
5776 - better file I/O statistics gathering in /proc
5777 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
5778 - another kernel patch to fix zero-copy TCP function export
5779 - don't take duplicate lock when processing re-sent getattr (2420)
5780 - lctl uses obd_self_export instead of creating new conn (2353)
5781 - MDS/OST recovery case which requires object creation asserted (2425)
5782 - move lfs from /usr/sbin to /usr/bin in packages
5783 - fix race between mds_client_add and mds_client_free (2417)
5784 - use kmalloc instead of slabs in portals (2430)
5785 - don't create duplicate records when a failover MDS is present (2442)
5786 - remove unnecessary mount age check (2332)
5787 - don't remove directory inodes from locks prematurely (2451)
5788 - don't break if MDS service name is the same as hostname (2103)
5789 - fix races in client write RPC generation when cache full (2482)
5791 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
5794 - remove now-unused request->rq_obd (278)
5795 - if an allocation fails, print out how much memory we've used (1933)
5796 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
5797 - add the "configurable stack size" patch to most series files (1256)
5798 - ability to write large log records, for 100+ OST configs (2306)
5799 - fix NULL deref when filter_prep fails (2314)
5800 - fix operator precedence error in filter_sync
5801 - dynamic allocation of socknal TX descriptors (2315)
5802 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
5803 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
5804 - ensure that CWARN messages in Portals always get to the syslog
5805 - __init/__exit are not for prototype decls (ldlm_init/exit)
5806 - x86-64 compile warning fixes
5807 - fix gateway LMC keyword conflict (2318)
5808 - fix MDS lock inversions in getattr/reint paths (1844)
5809 - fix a rare lock re-ordering bug, which caused deadlock (2322)
5810 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
5811 - DLM race condition prevented some lock evictions (2328)
5812 - ENOMEM detection and retry on socknal sends (2230)
5813 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
5814 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
5815 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
5816 - better error messages when a client is rejected during recovery (1505)
5817 - avoid cancelling locks which were never granted, after failure (2330)
5818 - fix i_sem/journal inversion in mds_client_add (2333)
5819 - fix truncate/getattr lock cycle deadlock (2334)
5820 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
5821 - fix two rare exit paths which could leak an l_lock() ref (2321)
5822 - fencepost error in MDS/OST orphan recovery (2226)
5823 - make log record alignment 8 bytes (1988)
5824 - lstripe now fails when requested offset > ost_count (2237)
5825 - ensure that all kernel series have a complete list.h (1607)
5826 - fix crashes in special-file operations (2316)
5827 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
5829 - allow configurable automake binary, for testing new versions
5830 - small update to the lfs documentation
5832 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
5834 * fix negative export reference count in fsfilt_sync (2312)
5836 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
5837 * release candidate 0.9.1
5839 - orphans are moved into the PENDING directory for possible recovery
5840 - replayed opens now open by fid for orphan/rename safety (1042)
5841 - last close of an orphan inode generates a transno (683)
5842 - chdir() and mount() now pin the directory entry (1020)
5843 - avoid CERROR in normal ll_setattr_raw() error case (1500)
5844 - discard very old requests without processing them (1502)
5845 - remove some common, well-understood CERRORs (1505)
5846 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
5847 - clear "grant" flags in OST replies until OST grant code lands (1644)
5848 - fix read performance by not clobbering i_blksize on client (1598)
5849 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
5850 - make LRU size a /proc tunable, clears locks when reduced (707)
5851 - fix some lprocfs rot that prevented ptlbd from loading (1732)
5852 - server locks take references on exports now (1558)
5853 - build fixes for 2.4.20-rh trees (1663)
5854 - return an error from lov_create if all OSCs are inactive (1751)
5855 - fix import levels when a reconnect happens without a timeout (1597)
5856 - exit early from mds_open if we get a lookup error (1749)
5857 - partial page read at EOF wouldn't wait for disk before sending (1642)
5858 - avoid NULL deref in obdfilter when reading page past EOF (1592)
5859 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
5860 - fix LBUG in ll_it_open_error with rc = -2 (1861)
5861 - write/truncate lock inversion (1639)
5862 - Don't auto-load obdclass, portals modules during cleanup (1495)
5863 - fix timestamps from jumping to "now" (1763)
5864 - extra journal assertions (1648)
5865 - add an extra multiunlink test (1771)
5866 - fix read_record/write_record API (1776)
5867 - fix leak of offset_extent, possible incorrect i_size later (1772)
5868 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
5869 - Debugging for the kqswnal_get_idle_tx problems (1820)
5870 - Allow recovery to be attempted multiple times (1536)
5871 - Write out MDS last_rcvd file after it is first created (1600)
5872 - Fix tx_descriptor leak in failed transmit situations (1827)
5873 - ext3 journaling fixes for assertion failure after IO error (1871)
5874 - class_export_put() on freed export after completion AST error (1896)
5875 - Fix revalidate looping in VFS (1322)
5876 - Don't access a freed export during MDS_REINT timeout (1521)
5877 - Add open-unlink recovery support on the MDS (1673,1764)
5878 - Return an error if no MDS data was read from last_rcvd (1946)
5879 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
5880 - Don't setup a disk device that doesn't match exported UUID (317)
5881 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
5882 - avoid committing NULL handle in force close
5883 - local.sh is now a one-stripe LOV configuration
5884 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
5885 - fix ext3 htree duplicate directory entry corruption (1516)
5886 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
5887 - update to new LOV EA format (2097)
5888 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
5890 2003-06-15 Phil Schwan <phil@clusterfs.com>
5893 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
5894 - per-import/export recovery handling (958, 931, 959)
5895 - multiple last-rcvd slots, for serving multiple FSes (949)
5896 - connections are again shared between multiple imp/exports (963, 964)
5897 - "umount -f" would hang if any requests needed to be sent (393, 978)
5898 - avoid pinning large req buffer by copying for queued messages (989)
5899 - add "uuid" to "lctl device" command to help upcalls (991)
5900 - "open" RPCs with transnos would confuse recovery counters (1037)
5901 - do proper endian conversion of all wire messages (288, 340, 891)
5902 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
5903 - call waiting locks callback from LDLM recovery thread (1127, 1151)
5904 - fix ptlrpc_connection leak in target_handle_connect (1174)
5905 - fix import refcounting bug in OST and MDS cleanup (1134)
5906 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
5907 - fix very unlikely obd_types race condition (501)
5908 - remove osc_open hack for echo_client (1187)
5909 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
5910 - a failure in read_inode2 leads to deadlock (1139)
5911 - cancel ack-locks as soon as transaction is committed (1072)
5912 - fix major leaks and crashes in the bulk I/O path (937, 1057)
5913 - make sure to commitrw after any preprw to avoid deadlock (1162)
5914 - failing to execute a file in a lustre FS would lock inode (1203)
5915 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
5916 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
5917 - obd_uuid2tgt would walk off the end of the list (1255)
5918 - on IA64 the osc would give portals incorrect bulk size (1258)
5919 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
5920 - fix lock inversion caused by new llite matching code (1282)
5921 - limit the number of dirty pages on a client to 10MB (1286)
5922 - timed out locks were not being corrected cancelled (1289)
5923 - fix O_DIRECT above 4GB on IA-32 (1292)
5924 * major user-visible changes
5925 - fail out/fail over policy now controlled by the upcall (993)
5927 - add OBD_PING to check server availability and failure (954)
5928 - lustre messages are now sent in sending host order (288, 340, 891)
5929 - add eadatalen to MDS getattr reply (340)
5930 - OST read replies may contain second buffer, with per-page status (593)
5932 2003-03-11 Phil Schwan <phil@clusterfs.com>
5935 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
5936 - failed open()s could cause deadlock; fixed (867, 869)
5937 - stop cancelling OST locks when files are closed (481)
5938 - overlapping XID spaces caused network corruption (851, 853)
5939 - fix unsafe fsfilt counter arithmetic; change to atomic_t
5940 - setattr_raw added, to do single-RPC, server-side setattrs
5941 - lmc/lconf syntax change for OST UUIDs
5942 - fix crashy race condition between ptlrpc_free_req and osc_close
5943 - don't use request in mdc_enqueue if we hit a timeout (889)
5944 - don't set the inode i_size for regular files from the MDS (896)
5945 - handle out of order completion AST (842)
5946 - don't LBUG if a lock request times out after receiving AST (913)
5947 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
5948 - if a bad lock AST arrives, send an error instead of dropping entirely
5949 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
5950 - fix leak in bulk IO when only partially completed (899, 900, 926)
5951 - fix O_DIRECT for ia64 (55)
5952 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
5953 - C-z after timeout could hang a process forever; fixed (977)
5955 - client-side I/O cache (678, 924, 929, 941, 970)
5957 - READPAGE and SETATTRs which don't take server-side locks get
5960 2003-02-11 Phil Schwan <phil@clusterfs.com>
5963 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
5964 - Fix fsfilt_extN_readpage() to read a full page of directory
5965 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
5966 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
5967 - Fix imbalanced LOV object allocation and out-of-bound access (469)
5968 - Most intent operations were removed, in favour of a new RPC mode
5969 that does a single RPC to the server and bypasses most of the VFS
5970 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
5971 - Aggressively cancel local locks on DLM servers
5972 - mds_reint_unlink sends EA to the client if it's the last nlink.
5973 client uses that EA to unlink OST objects.
5974 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
5975 - recursive symlinks were fixed (439)
5976 - fixed NULL deref in DEBUG_REQ
5977 - filter_update_lastobjid no longer calls sync, which annoyed extN
5978 - fixed multi-client small-writes to a single file problem (445)
5979 - fixed mtime updates during file writes (607)
5980 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
5981 - fixed bug in obd_brw_read/write() (under guise of testing 367)
5982 - fixed Linux OST size reporting problem (444, 656)
5983 - OST now updates object mtime with writes or setattr (607, 619)
5984 - client verifies file size before zeroing page past EOF (445)
5985 - OST now writes last allocated objid to disk with allocation (108)
5986 - LOV on echo now works (409)
5988 - mds_reint_unlink sends a new buffer, with the EA included. this
5989 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
5990 set if a regular file was being unlinked, and it was the last link
5991 - use PtlGet from the target for bulk writes (315)
5992 - OST now updates object mtime with writes or setattr (607, 619)
5993 - LDLM now has a grant-time callback to revalidate locked items, if
5995 - Many MDS operations were reorganized to combat race conditions
5997 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
5998 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
6000 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
6003 - Fully reactivate OST imports after reconnection (512, others)
6004 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
6005 - More graceful error handling for truncating on dead OST (515)
6006 - Don't error out unless we're actually accessing dead stripes (474)
6007 - Fix garbage sizes when stripes are missing (410)
6008 - LRU counters were broken, causing constant lock purge (433, 432)
6009 - garbage on read from stripes with failed OSTs (441)
6010 - mark OSCs as active before reconnecting during recovery (438)
6011 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
6012 - lfind did not preserve OST order in output (443)
6013 - symlinks cause hung clients, incorrect data (439)
6014 - stop dereferencing request after dropping refcount (457)
6015 - don't LASSERT(spin_is_locked) on non-SMP (455)
6016 - fixes for many rename() bugs
6017 - fstat didn't correctly synchronize attributes (399)
6018 - server must handle lock cancellation during blocking AST prep (487)
6019 - bulk descriptors were free()d too soon (511)
6020 - fix paths in lconf, which would load incorrect modules (451, 507)
6021 - fix confusing lconf 'host not found' error message (386)
6022 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
6023 - fix race condition in mdc_blocking_ast() for inode access (526)
6024 - fix lov_unpackmd() unpacking wrong number of stripes (537)
6025 - fix lov_set_osc_active() marking wrong OSC inactive (440)
6026 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
6027 - fix multiple writes of stripe MD to MDS (358, maybe 519)
6028 - fix lstripe in several ways (kernel side) (527)
6029 - fix request leak in ldlm_cli_enqueue (262)
6030 - incorrect OSC was marked inactive after OST failure
6031 - call mds_fs_cleanup before unmounting filesystem (524)
6032 - fix races between taking ns_lock and ldlm_lock_change_resource
6033 - fix races updating LOV export open file list
6034 - fix lov_enqueue error path, avoid decref-ing bad lock handle
6035 - fix recovery NULL deref in ldlm_cli_cancel_unused
6036 - fix some DLM races by using new hash table for lock handles (419)
6037 - permit the client to specify desired inodes, at replay
6038 - duplicate requests when we queue them for replay reintegration
6039 - fix last_rcvd offset calculation
6040 - sync after each recovered transaction, so we always make progress
6041 - never, not always, ERESTART requests without transnos
6042 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
6044 - skip replay if the MDS says that the client is already connected
6045 - don't check for a recovery-enabled export to match lctl's UUID
6046 - don't INC_USE_COUNT for phantom exports
6047 - don't crash when cleaning up phantom exports (567)
6048 - don't double-finish or set replay data for errored mdc_open requests
6049 - abort requests when they time out, so we don't get old replies
6050 - send/receive replies for AST messages again
6051 - if the client says that it doesn't have the lock, cancel it on the
6053 - if we timeout during I/O, don't try to cancel an in-use lock; instead
6054 mark it as destroyed, it will all work out when decref is called
6055 - fix module use counts (22, 581)
6057 - ASTs now expect a reply (server cancels lock on error reply)
6059 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
6062 - fix many simultaneous client startup (392)
6063 - fix dentry->d_it clobbering
6064 - credentials weren't being shipped for readdir/getattr operations
6065 - remove invalid assertions triggered during some concurrent MD
6067 - proper Lustre versions added (336, 389)
6068 - fix memory leak for create error case (398)
6069 - fix LOV locking bug that would get cli/srv out of sync
6070 - fix echo client over LOV (409)
6071 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
6072 - fix double-O_EXCL intent crash (424)
6073 - avoid sending multiple lock CANCELs (352)
6075 - MDS can do multi-client recovery (modulo bugs in new code)
6077 - many updates, edits, cleanups
6079 2002-11-18 Phil Schwan <phil@clusterfs.com>
6082 - fix null d_it dereference (346)
6083 - fix full OST/dbench hang (333)
6084 - fix permission problem with file removal (286)
6085 - fix removal of OSCs from LOV when they fail
6086 - fix NULL deref during bulk timeout (214)
6087 - fix problems related to multiple filesystems on one MDS (241)
6088 - fixed serious subtle metadata locking bugs
6089 - free locks on clients when inodes are removed due to memory
6091 - fix inode pointer in lock data (285)
6092 - partial support for multiple MDS on a single host (241)
6093 - data locks weren't cancelled at clear_inode time (290, 311)
6094 - intent locks could lead to unbounded lock growth (205)
6095 - added a maximum lock count, an LRU list, and a flusher
6096 - fix multiple rename (365)
6097 - properly abstracted the echo client
6098 - OSC locked 1 byte too many; fixed
6099 - rewrote brw callback code:
6100 - fixed recovery bugs related to LOVs (306)
6101 - fixed too-many-pages-in-one-write crash (191)
6102 - fixed (again) crash in sync_io_timeout (214)
6103 - probably fixed callback-related race (385)
6105 - Add capability to MDS protocol
6106 - LDLM cancellations and callbacks on different portals
6108 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
6111 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
6112 - defer open object destruction to close time (601981, 138)
6113 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
6114 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
6115 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
6116 - don't call obd_setattr in ll_file_release for destroyed objects
6117 * protocol change to lustre_msg: move |version| and add |flags|
6118 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
6119 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
6120 * added replay of create, unlink, link and rename operations during
6121 MDS failover; recovery should be much more robust now
6122 * remove failed OSCs from LOVs (only lov_create uses this so far)
6123 * the lustre-HOWTO was brought (more) up to date (582544)
6125 2002-10-23 Phil Schwan <phil@clusterfs.com>
6128 - in-use dentries weren't being reused properly (617851)
6129 - prevent multiple LDLM setup (599178)
6130 - fix LOV size calculations for truncate (617853)
6131 - fix client handling of MDS intent errors (POSIX)
6132 - fix permission bug in lovstripe.c test (624321)
6133 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
6134 - truncate past end of file could corrupt data
6135 - proper cleanup after timeouts, crashes, etc (592524, 550815)
6136 - a race in recovery could return ETIMEDOUT to apps (623947)
6137 - building outside the source directory was fixed
6138 * the lustre-HOWTO was brought (more) up to date (582544)
6139 * major progress was made on recovery functionality
6141 2002-10-10 Phil Schwan <phil@clusterfs.com>
6144 - recovery deadlock fix
6145 - rm -rf causes LBUG fix (617817)
6146 - file open by multiple tasks fix (618962)
6147 - directory permissions bugs (602707 and 620007)
6148 - journal_stop fixed with locking (611313)
6149 - O_APPEND failures resolved (618273, perhaps 614459)
6150 - lconf PATH fix (619770)
6151 - IA64 build fix (621450)
6152 - RPC buffer sizes scale with amount of memory
6154 2002-10-01 Phil Schwan <phil@clusterfs.com>
6157 - locks would be cancelled without throwing away data pages,
6158 resulting in inconsistent data (605627)
6159 - inode attributes were not always being refreshed (605627, 612449)
6160 - lconf now continues to cleanup after lctl reports an error
6161 - MDS now enforces user permissions (602707)
6162 - lprocfs cleanup fixed, but not yet enabled (614157)
6163 - fixed infinite server hang, should a client not respond to an AST
6164 - avoid going into recovery if user calls readlink() with a buffer
6165 that's too small (613941)
6166 - AST RPCs no longer require replies (614867) -- this may be changed
6167 - don't crash server if client sends an IOV that's too big (611336)
6168 - fixed lock conversion deadlock (611892)
6169 - fixed the following of symlinks (614622)
6170 * recovery: the server can remove locks from a client that dies, other
6171 clients can make progress
6172 * more extN patch fixes
6173 * compile-time configurable ptlrpc buffer allocations
6175 - collaborative read cache document
6176 - Lustre Lite Performance CDR document-in-progress
6178 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6181 - fix typo in patch-2.4.18
6183 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6186 - clear ptlrpc request each time in handle_incoming_request()
6187 - unlink of files now destroys the object on the OST
6189 2002-09-19 Peter Braam <braam@clusterfs.com>
6191 * add hard link support
6192 * change obdfile creation method
6193 * kernel patch changed
6195 2002-09-19 Peter Braam <braam@clusterfs.com>
6198 - stack overflow bug in extN fixed
6200 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
6202 * documentation updates
6203 - add man pages for config tools
6204 - update tests/README to describe testing with new config tools
6205 - finish metadata API descriptions
6206 * bug fixes and cleanups
6207 - statfs workaround for 16TB limit
6208 - LOV stripe allocation improved, can stripe on subset of OSTs
6209 - LOV file size/IO offset was wrong for files > 4GB in size
6210 - object EA data was being dropped, caused files to be unreadable
6211 - memory overflow with non-LOV OST caused memory corruption
6212 - fixed regression tests to work with new config tools, obdfilter
6213 - fixed bug when directory size became larger than 1 block
6214 - fixed bug (for single client case) when PWD was deleted
6215 - invalidate local directory pages when doing intent-based ops
6216 - avoid LDLM oops when lock callback contained bad data
6218 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
6220 * documentation updates
6221 * bug fixes and cleanups
6222 - configuration tools
6225 - 64-bit compile warnings
6226 - 64-bit internal statfs data
6228 * test_brw on persistent OST devices
6230 * lprocfs (disabled)
6232 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
6234 * documentation updates
6235 * bug fixes and cleanups
6236 * configuration tools
6238 2002-08-30 Peter J. Braam <braam@clusterfs.com>
6241 * many small fixes to 0_5_4
6242 * io/network handling
6243 * thinkos in MDS operations
6245 2002-08-24 Peter J. Braam <braam@clusterfs.com>
6248 * crucial basic fixes to 0.5.3
6249 * IOR, Iozone work over Elan
6252 2002-08-07 Phil Schwan <phil@clusterfs.com>
6253 * version 0_5_3, our first alpha
6254 * we use the new Portals iovs
6255 * documentation updates
6256 * bug fixes and cleanups
6257 * small changes in the DLM wire protocol
6259 2002-07-25 Peter J. Braam <braam@clusterfs.com>
6260 * version 0_5_1 with some initial stability,
6261 * locking on MD and file I/O.
6262 * documentation updates
6263 * several bug fixes since 0.5.0
6264 * small changes in wire protocol
6266 2002-07-18 Phil Schwan <phil@clusterfs.com>
6268 * delivered as Lustre Light Alpha
6269 * fixed a crash after handling invalid MDS requests
6270 * fixed directory pages for architectures with non-4k pages sizes
6272 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
6273 * release version v0_4_4
6274 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
6276 * Fixes a number of interruption problems with OST operations.
6277 * Update documentation for portals header changes
6278 * Move all wire protocol structs/defines to lustre_idl.h
6279 * Fixes symlink length bug.
6280 * Add tcpdump to repository.
6282 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
6283 * release version v0_4_3
6284 * Fixes statfs for inodes on extN.
6285 * Fixes bug in runtests which would delete /etc/hosts.
6286 * Use 64-bit object IDs wherever possible (not into VFS though)
6287 Remove ost_get_info, which is unused by lustre, and out of date.
6289 2002-07-03 Peter Braam <braam@clusterfs.com>
6290 * release version v0_4_2 Fixes a lookup error (type not passed)
6291 * move forward to head of Portals
6292 * move forward to latest Lustre kernel
6294 2002-06-25 Peter Braam <braam@clusterfs.com>
6295 * release version v0_4_1. Hopefully stable on single node use.