3 * Server support for kernels:
4 2.6.18-238.19.1.el5 (RHEL5)
5 2.6.32-131.17.1.el6 (RHEL6)
6 * Client support for unpatched kernels:
7 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
8 2.6.18-238.19.1.el5 (RHEL5)
9 2.6.32-131.17.1.el6 (RHEL6)
10 2.6.29.4-167.fc11 (FC11)
11 2.6.16.60-0.69.1 (SLES10)
13 * Recommended e2fsprogs version: 1.41.90.wc3
14 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
15 Lustre filesystem with 4K stack may cause a stack overflow. For more
16 information, please refer to bugzilla 17630.
17 * NFSv4 reexport requires Lustre client on the re-exporting nodes to be
18 mounted with "32bitapi" mount option
20 Severity : enhancement
22 Description: Update RHEL6.1 kernel to 2.6.32-131.17.1.el6
24 --------------------------------------------------------------------------------
26 09-16-2011 Whamcloud, Inc.
28 * Server support for kernels:
29 2.6.18-238.19.1.el5 (RHEL5)
30 2.6.32-131.6.1.el6 (RHEL6)
31 * Client support for unpatched kernels:
32 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
33 2.6.18-238.19.1.el5 (RHEL5)
34 2.6.32-131.6.1.el6 (RHEL6)
35 2.6.29.4-167.fc11 (FC11)
36 2.6.16.60-0.69.1 (SLES10)
38 * Recommended e2fsprogs version: 1.41.90.wc3
39 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
40 Lustre filesystem with 4K stack may cause a stack overflow. For more
41 information, please refer to bugzilla 17630.
42 * NFSv4 reexport requires Lustre client on the re-exporting nodes to be
43 mounted with "32bitapi" mount option
45 Severity : enhancement
46 Description: Update RHEL6.1 kernel to 2.6.32-131.6.1.el6
48 Severity : enhancement
49 Description: Update RHEL5.6 kernel to 2.6.18-238.19.1.el5
51 Severity : enhancement
52 Description: Update RHEL5.6 kernel to 2.6.18-238.9.1.el5
55 Frequency : only with SLES10
57 Description: Use OFED "KMP" provided by Novell
58 Details : SLES10 SP3 ships with OFED in a separate "KMP" package.
59 Lustre is now built against this package. That means you need to
60 install the ofed-kmp package from Novell for the patchless client.
61 Note that the ofed-kmp that Novell ships may not exactly match the
62 kernel version but should still be compatible.
64 Severity : enhancement
66 Description: Update RHEL5.5 kernel to 2.6.18-194.17.1.el5
67 Update OEL5.5 kernel to 2.6.18-194.17.1.0.1.el5.
69 --------------------------------------------------------------------------------
71 2010-07-15 Oracle, Inc.
73 * Server support for kernels:
74 2.6.18-164.11.1.el5 (RHEL 5)
75 2.6.18-164.11.1.0.1.el5 (OEL 5)
76 * Client support for unpatched kernels:
77 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
78 2.6.18-164.11.1.el5 (RHEL 5),
79 2.6.18-164.11.1.0.1.el5 (OEL 5)
80 2.6.16.60-0.42.8 (SLES 10),
82 2.6.29.4-167.fc11 (FC11)
83 * Recommended e2fsprogs version: 1.41.10-sun2
85 Severity : enhancement
87 Description: Distribute lfs_migrate script with lustre
88 Details : The lfs_migrate mentioned in the manual is now included in the
89 lustre tree in lustre/scripts and will distributed in lustre RPMs.
91 Severity : enhancement
93 Description: Failover nid list added to import proc file.
95 Severity : enhancement
97 Description: "som_preview" mount option required to use SOM
98 Details : The clients will need to use som_preview mount option
102 Description: Lustre does not do 1MB IOs to HW RAID
103 Details : Bump MAX_PHYS/HW_SEGMENTS and SG_ALL to 256 in the RHEL5 kernel.
105 Severity : enhancement
107 Description: lfs quota output is not very convenient for awk/sed-parsing
108 Details : Some positions in lfs quota output table could be empty or
109 non-empty which made it hard to parse it with scripts, now
110 a dash is put instead of space where there is not supposed
115 Description: Update to OFED 1.5.1
116 Details : OFED 1.5.1 will be used in environments that don't use
117 vendor supplied OFED.
121 Description: Remove lnet. prefix from lctl params display
122 Details : Also renamed lnet.memused to lnet_memused to
127 Description: Use conf_param -d to remove permanant settings.
131 Description: Add list_param -R
132 Details : lctl list_param now supports recursive mode.
135 Bugzilla : 21829, 22312
136 Description: Fix llobdstat.
137 Details : 1. Now updating the write_bytes counter
138 2. Port script fixes from 1.8.
142 Description: decrease the usage of memory on clients.
143 Details : 1. On clients, recycle dentries and inodes unused.
144 2. Delete the code related to ll_deathrow(att 6215 in bug 1443). It
149 Description: llobdstat fix and enhancement
150 Details : add a counter to set a limit to how many samples will be returned
151 fix a wildcard in the path to limit to obdfilter stats only
155 Description: lfs check only allowed for root
156 Details : Code cleanups and added a regresion test of non-root use of
161 Description: MDS Oopses in ptlrpc_dispatch_difficult_reply()/__list_add()
162 Details : Fix hr->hr_index overload.
166 Description: Fix FIEMAP functionality.
170 Description: FC11 patchless client support added
174 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and
175 OEL5.4 kernel to 2.6.18-164.11.1.0.1.el5.
179 Description: New acc-sm script and various test-framework enhancments.
183 Description: Previously landed patch was reverted.
187 Description: Update SLES10 kernel to 2.6.16.60-0.42.8 (patchless)
191 Description: Add ldiskfs maxdirsize mount option.
195 Description: Fix mkfs.lustre support for 16TB LUNs.
199 Description: Print mmp_check_interval
200 Details : print mmp_check_interval and make it possible to abort
201 mount operation in case it takes too long.
205 Description: Add cascading_rw to lustre mpi tests.
209 Description: Store SOM data in EA.
210 Details : Also several SOM related cleanups.
214 Description: Add refcount to md_open_data
218 Description: Use refcount on lov_request_set to prevent memory corruption.
222 Description: Statahead: use parent dir's i_mutex for synchronization.
226 Description: loadgen improvements.
230 Description: Enable peer health detection in ksocklnd and ko2iblnd.
234 Description: Control DCACHE_LUSTRE_INVALID flag with MDS_INODE_LOOKCUP lock
238 Description: Fix race on obd_devcie:md_stats
242 Description: Fix SLES 11 client builds.
243 Details : Resolved uninitalized pointer gcc warnings.
247 Description: Allow non-root access for "lfs check".
248 Details : Added a check in obd_class_ioctl() for OBD_IOC_PING_TARGET.
252 Description: Clear imp_force_reconnect correctly in ptlrpc_connect_interpret()
254 Severity : enhancement
256 Description: Update kernel to RHEL5.4 2.6.18-164.6.1.el5 and
257 OEL5 2.6.18-164.6.1.0.1.el5(Both in-kernel OFED enabled).
261 Description: Sanity test_27p fails.
262 Details : OBD_FAIL_OST_ENOSPC has been moved to filter_statfs().
263 Sanity tests 27 were cleaned up and improved.
267 Description: Unify req format on client/servers
268 Details : Use new req_capsule API [almost] everywhere instead of old PTLRPC
269 buffers and swabbers approach..
274 Description: drop join file stuff.
276 Severity : enhancement
278 Description: Add 2.6.27 and SLES11 (patchless client) support
282 Description: Fix in ptlrpc_expire_one_request() to print the signed time
285 Severity : enhancement
287 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
290 Severity : enhancement
292 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
294 Severity : enhancement
296 Description: correctly handle big reply message.
297 Details : send LNet event if reply is bigger then buffer and adjust this buffer
302 Description: Drop unnecessary __GFP_NOMEMALLOC flag from filter_get_page()
304 Severity : enhancement
306 Description: Add support for OEL5.
310 Description: Conf-sanity.sh 50g test - deactivated OST should not cause a panic.
312 Severity : enhancement
314 Description: "weak-modules" support
315 Details : Implement "weak-modules" support which enables kernel modules
316 to be used with any kernel that implements the same kABI. In
317 order to achieve this modules are now installed in
318 /lib/modules/$(uname -r)/updates/kernel on all distributions.
322 Description: client could not reconnect to OST because of an active request.
323 Details : The client now retries to reconnect to the same server, if a
324 connect request failed with EBUSY or -EAGAIN. Test case 35b
325 conf-sanity.sh has been created.
329 Description: Don't allow make backward step on assignin osc next id.
330 Details : race between allocation next id and ll_sync thread can be cause
331 of set wrong osc next id and can be kill valid ost objects.
335 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
339 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
343 Description: Error handling in osc_statfs_interpret() has been improved.
344 Details : Check in osc_statfs_interpret() for EBADR.
348 Description: Do not send statfs() requests to OSTs disabled by administrator.
349 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
351 Severity : enhancement
353 Description: Added sanity-quota test to for lfs(1) to work with numeric uid/gid.
355 Severity : enhancement
357 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
359 Severity : enhancement
361 Description: Enhancement for lfs(1) command to use numeric uid/gid.
365 Description: Improved cleanup procedure for conf-sanity test_46a.
369 Description: Do not start bulk transfer, if client tried to reconnect.
370 Details : Do not start bulk transfer, if client tried to reconnect,
371 but failed due to active rpcs. Check for exp_abort_active_req
372 has been implemented.
376 Description: Prevent inconsistences between linux and lustre mount structures.
377 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
378 Make the sleep interruptible.
380 Severity : enhancement
382 Description: provide server to client comms path
383 Details : server to client communications path via new LDLM_SET_INFO rpc
387 Description: truncate starts GFP_FS allocation under transaction causing deadlock
388 Details : ldiskfs_truncate calls grab_cache_page which may start page
389 allocation under an open transaction. This may lead to
390 calling prune_icache with consequent lustre reentrance.
394 Description: send correctly lsm on open replay
395 Details : MDS is trust to LSM size on replay open, but client can set wrong size
400 Description: enable client interface failover
401 Details : When a child reconnects from another NID, properly update export
402 nid hash position and ldlm reverse import.
404 Severity : enhancement
406 Description: Communicate OST degraded/readonly state via statfs to MDS
407 Details : Flags in the statfs returned from OSTs indicate whether the
408 OST is in a degraded RAID state, or if the filesystem has
409 turned read-only after a filesystem error is detected.
414 Description: don't panic if EPROTO was hit when reading symlink
415 Details : correctly handling request reference in error cases.
417 Severity : enhancement
419 Description: Add LustreNetLink, a kernel-userspace communcation path.
421 Severity : enhancement
423 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
428 Description: access to llog context before init.
429 Details : move handling CATALOGS file at osc layer and forbid access to llog
433 Frequency : rare, only when client is writing to OST on local node
435 Description: Avoid deadlock for local client writes
436 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
437 memory freeing context. This allows OST threads to set the
438 PF_MEMALLOC flag on task structures in order to allocate memory
439 from reserved pools and complete IO.
440 Use GFP_HIGHUSER for OST allocations for non-local client writes,
441 so that the OST threads generate memory pressure and allow
442 inactive pages to be reclaimed.
444 Severity : enhancement
446 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
451 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
452 Details : move ldlm namespace creation in setup phase to avoid grab
453 _lprocfs_lock with cli_sem held.
457 Description: Temporarily disable grant shrink.
458 Details : Disable the feature for debugging.
461 Frequency : only when formatting test filesystems
463 Description: Unable to run several mkfs.lustre on loop devices at the same time
464 Details : mkfs.lustre returns error 256 on the concurrent loop devices
465 formatting. The solution is to proper handle the error.
467 Severity : enhancement
469 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
471 Severity : enhancement
473 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
478 Description: resolve race between obd_disconnect and class_disconnect_exports
479 Details : if obd_disconnect will be called to already disconnected export he
480 forget release one reference and osc module can't unloaded.
482 Severity : enhancement
484 Description: Allow tuning service thread via /proc
485 Details : For each service a new
486 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
487 created that can be used to set min/max thread counts, and get the
488 current number of running threads.
493 Descriptoin: don't return error if have partially created objects for file.
494 Details : lov_update_create_set uses set->set_success as index for created
495 objects, so if some requests will be failed, they will have hole
496 at end of array and qos_shrink_lsm for allocate correct lsm.
498 Severity : enhancement
500 Description: Update OFED support to 1.4.1
505 Description: Reduce small size read RPC
506 Details : Set read-ahead limit for every file and only do read-ahead when
507 available read-ahead pages are bigger than 1M to avoid small size
510 Severity : enhancement
512 Description: move AT tunable parameters for more consistent usage
513 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
515 Severity : enhancement
517 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
518 Details : allow skip disconnected ost for send statfs request and hide error
524 Description: fix racy locking of mballoc block bitmaps causing BUG
525 Details : The locking of the mballoc buddy bitmap and the in-memory
526 block bitmap was using two different spin locks in some
527 cases. This made it possible to incorrectly access the
528 mballoc bitmap while another process was modifying it,
529 causing a sanity assertion to fail. While no on-disk corruption
530 was reported, there was some risk of this happening.
533 Frequency : rare, on llog test 6
535 Descriptoin: don't allow connect to already connected import
536 Details : allowing connect to already connected import is hide connecting problem.
541 Descriptoin: allow kill process which wait statahead result
542 Details : for some reasons 'ls' can stick in waiting result from statahead,
543 in this case need way for kill this process.
545 Severity : enhancement
547 Description: Add state history info file, enhance import info file
548 Details : Track import connection state changes in a new osc/mdc proc file;
549 add overview-type data to the osc/mdc import proc file.
551 Severity : enhancement
553 Description: MDS create should not wait for statfs RPC while holding DLM lock.
555 Severity : enhancement
557 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
562 Descriptoin: OST grant too much space to client even there are not enough space.
563 Details : Client will shrink its grant cache to OST if there are no write
564 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
565 this grant cache if there are already not enough avaible space
566 (left_space < total_clients * 32M).
569 Frequency : start MDS on uncleanly shutdowned MDS device
571 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
572 Details : stay in waiting mds<>ost recovery finished produce random bugs
573 due race between two ll_sync thread for one lov target. send
574 ACTIVATE event only if connect realy finished and import have
578 Frequency : rare, connect and disconnect target at same time
580 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
581 Details : don't call obd_disconnect under lov_lock. this long time
582 operation and can block ptlrpcd which answer to connect request.
587 Descriptoin: don't lose wakeup for imp_recovery_waitq
588 Details : recover_import_no_retry or invalidate_import and import_close can
589 both sleep on imp_recovery_waitq, but we was send only one wakeup
593 Frequency : always with long access acl
595 Descriptoin: mds can't pack reply with long acl.
596 Details : mds don't control size of acl but they limited by reint/getattr
599 Severity : enhancement
601 Description: Update to SLES10 kernel-2.6.16.60-0.33.
603 Severity : enhancement
605 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
608 Frequency : start MDS on uncleanly shutdowned MDS device
610 Descriptoin: aborting recovery hang on MDS
611 Details : don't throttle destroy RPCs for the MDT.
614 Frequency : on remount
616 Description: external journal device not working after the remount
617 Details : clear dev_rdonly flag for external journal devices in
623 Description: shutdown vs evict race
624 Details : client_disconnect_export vs connect request race.
625 if client will evicted at this time - we start invalidate
626 thread without referece to import and import can be freed
630 Frequency : rare, need acl's on inode.
632 Description: client can't handle ost additional correctly
633 Details : if ost was added after client connected to mds client can have
634 hit lnet_try_match_md ... to big messages to wide striped files.
635 in this case need teach client to handle config events about add
636 lov target and update client max ea size at that event.
638 Severity : enhancement
640 Description: Changelogs
641 Details : Changelogs are a lightweight mechanism to track filesystem
642 metadata and namespace changes. The changelog is recorded
643 permanently on the MDTs, and is periodically "consumed" / purged
644 when records are no longer needed.
646 Severity : enhancement
648 Description: compact fld format with extents
649 Details : Store range of seq rather than every seq in FLD. Seq
650 controller update FLD rather than clients. In Case of CMD, mdt0
651 has FLD, all other metadata server act as non persistent proxy
652 for FLD queries and cache fld entries in fld cache.
657 Description: don't skip ost target if they assigned to file
658 Details : Drop slow OSCs if we can, but not for requested start idx.
659 This means "if OSC is slow and it is not the requested
660 start OST, then it can be skipped, otherwise skip it only
661 if it is inactive/recovering/out-of-space.
665 Description: more cleanup in mds_lov
666 Details : not send LOV EA under replay, we can't know about they size at this
667 time. Don't allow client connect to mds before any ost connected,
668 for avoid problems with LOV EA size and returning EIO to client.
670 Severity : enhancement
672 Description: Interoperability at server side (Disk interoperability)
674 Severity : enhancement
676 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
678 Severity : enhancement
680 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
682 Severity : enhancement
684 Description: New client IO stack (CLIO).
686 Severity : enhancement
688 Description: Commit on sharing. Eliminate inter-client dependencies between
689 uncommitted transactions by doing transaction commits.
690 Thereby clients may recovery independently.
693 Frequency : Create a symlink file with a very long name
695 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
696 Details : If there is no extra space in the request for early cancels,
697 ldlm_req_handles_avail() returns 0 instead of a negative value.
699 Severity : enhancement
701 Description: Add /proc entry for import status
702 Details : The mdc, osc, and mgc import directories now have
703 an import directory that contains useful import data for debugging
706 Severity : enhancement
708 Description: Re-disable certain /proc logging
709 Details : Enable and disable client's offset_stats, extents_stats and
710 extents_stats_per_process stats logging on the fly.
713 Frequency : Only on FC kernels 2.6.22+
715 Description: oops in statahead
716 Details : Do not drop reference count for the dentry from VFS when lookup,
717 VFS will do that by itself.
719 Severity : enhancement
721 Description: Generic /proc file permissions
722 Details : Set /Proc file permissions in a more generic way to enable non-
723 root users operate on some /proc files.
727 Description: Hitting mdc_commit_close() ASSERTION
728 Details : Properly handle request reference release in
729 ll_release_openhandle().
733 Description: quota recovery deadlock during mds failover
734 Details : This patch includes att18982, att18236, att18237 in bz14840.
736 1. fix osts hang when mds does failover with quotaon
737 2. prevent watchdog storm when osts threads wait for the
742 Frequency : only patchless client
743 Description: add workaround for race between add/remove dentry from hash
745 Severity : enhancement
747 Description: Allow OST glimpses to return PW locks
751 Description: LBUG when llog conf file is full
752 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
756 Description: Prevent import from entering FULL state when server in recovery
760 Description: service mount cannot take device name with ":"
761 Details : Only when device name contains ":/" will mount treat it as
767 Description: replace ptlrpcd with the statahead thread to interpret the async
768 statahead RPC callback
772 Frequency : on recovery
773 Description: I/O failures after umount during fail back
774 Details : if client reconnected to restarted server we need join to recovery
775 instead of find server handler is changed and process self
776 eviction with cancel all locks.
778 Severity : enhancement
780 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
782 Severity : enhancement
784 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
786 Severity : enhancement
788 Description: Upcall on Lustre log has been dumped
789 Details : Allow for a user mode script to be called once a Lustre log has
790 been dumped. It passes the filename of the dumped log to the
791 script, the location of the script can be specified via
792 /proc/sys/lnet/debug_log_upcall.
797 Description: avoid idr_remove called for id which is not allocated.
798 Details : Move assigment s_dev for clustered nfs to end of initialization,
799 to avoid problem with error handling.
804 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
805 Details : When connection is reused this not moved from CONN_UNUSED_HASH
806 into CONN_USED_HASH and this prodice warning when put connection
807 again in unused hash.
810 Severity : enhancement
812 Description: File striping can now be set to use an arbitrary pool of OSTs.
814 Severity : enhancement
816 Description: Export bytes_read/bytes_write count on OSC/OST.
820 Description: Early reply size mismatch, MGC loses connection
821 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
822 the connect flags are properly negotiated.
827 Description: quotas are not honored with O_DIRECT
828 Details : all writes with the flag O_DIRECT will use grants which leads to
829 this problem. Now using OBD_BRW_SYNC to guard this.
833 Description: add quota statistics
834 Details : 1. sort out quota proc entries and proc code.
835 2. add quota statistics
837 Severity : enhancement
839 Description: enable quota support for HEAD.
843 Description: Properly propagate oinfo flags from lov to osc for statfs
844 Details : restore missing copy oi_flags to lov requests.
846 Severity : enhancement
848 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
849 : l_getgroups(8), lst(8), routerstat(8)
850 Details : included man pages for llobdstat(8), llstat(8),
851 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
853 Severity : enhancement
855 Description: configure's --enable-quota should check the
856 : kernel .config for CONFIG_QUOTA
857 Details : configure is terminated if --enable-quota is passed but
858 : no quota support is in kernel
862 Description: Remove portals compatibility
863 Details : Remove portals compatibility, not interoperable with releases
868 Description: Resolve device initialization race
869 Details : Prevent proc handler from accessing devices added to the
870 obd_devs array but yet be intialized.
872 Severity : enhancement
874 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
876 Severity : enhancement
878 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
883 Description: Using wrong pointer in osc_brw_prep_request
884 Details : Access to array[-1] can produce panic if kernel compiled with
885 CONFIG_PAGE_ALLOC enabled
889 Description: Client runs out of low memory
890 Details : Consider only lowmem when counting initial number of llap pages
894 Description: *optional* service tags registration
895 Details : if the "service tags" package is installed on a Lustre node
896 When the filesystem is mounted, a local-node service tag will
897 be created. See http://inventory.sun.com/ for more information
898 about the Service Tags asset management system.
902 Description: Kernel BUG tries to release flock
903 Details : Lustre does not destroy flock lock before last reference goes
904 away. So always drop flock locks when client is evicted and
905 perform unlock regardless of successfulness of speaking to MDS.
909 Description: add refcount protection for osc callbacks, avoid panic on shutdown
913 Description: sanity test 65a fails if stripecount of -1 is set
914 Details : handle -1 striping on filesystem in ll_dirstripe_verify
919 Description: ASSERTION(CheckWriteback(page,cmd)) failed
920 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
923 Severity : enhancement
925 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
929 Description: do not process already freed flock
930 Details : flock can possibly be freed by another thread before it reaches
931 to ldlm_flock_completion_ast.
935 Description: LBUG during stress test
936 Details : Need properly lock accesses the flock deadlock detection list.
940 Description: oops in page fault handler
941 Details : kernel page fault handler can return two special 'pages' in error
942 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
946 Description: timeout with invalidate import.
947 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
948 should be handled by ptlrpcd. This produce long age waiting and
949 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
951 Severity : enhancement
953 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
957 Description: enable MGS and MDT services start separately
958 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
959 a co-located MGS without starting the MGS, which is a complement
960 to 'nosvc' mount option.
964 Frequency : after recovery
965 Description: precreate to many object's after del orphan.
966 Details : del orphan st in oscc last_id == next_id and this triger growing
967 count of precreated objects. Set flag LOW to skip increase count
968 of precreated objects.
972 Frequency : rare, on clear nid stats
973 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
974 Details : when clean nid stats sometimes try destroy live entry,
975 and this produce panic in free.
979 Description: Stack overflow during MDS log replay
980 ease stack pressure by using a thread dealing llog_process.
984 Description: wait until IO finished before start new when do lock cancel.
985 Details : VM protocol want old IO finished before start new, in this case
986 need wait until PG_writeback is cleared until check dirty flag and
987 call writepages in lock cancel callback.
989 Severity : enhancement
991 Description: using special macro for print time and cleanup in includes.
995 Description: mds_mfd_close() ASSERTION(rc == 0)
996 Details : In mds_mfd_close(), we need protect inode's writecount change
997 within its orphan write semaphore to prevent possible races.
1001 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
1005 Frequency : rare, on shutdown ost
1006 Description: don't hit live lock with umount ost.
1007 Details : shrink_dcache_parent can be in long loop with destroy dentries,
1008 use shrink_dcache_sb instead.
1012 Description: don't panic with use echo client
1013 Details : echo client pass NULL as client nid pointer and this produce null
1014 pointer dereference.
1018 Description: fix build on ppc32
1019 Details : compile code with -m64 flag produce wrong object file for ppc32.
1023 Description: add message levels for liblustreapi
1027 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
1028 Details : Move the 'good_osts' check before the 'total_bavail' check. This
1029 will result in an -EAGAIN and in the exit call path we call
1030 alloc_rr() which will with increasing aggressiveness attempt to
1031 aquire precreated objects on the minimum number of required OSCs.
1035 Description: Use old size assignment to avoid deadlock
1036 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
1037 the scheduling while holding a spinlock. We do not need locking
1038 for size in ll_update_inode() because size is only updated from
1039 the MDS for directories or files without objects, so there is no
1040 other place to do the update, and concurrent access to such inodes
1041 are protected by the inode lock.
1045 Description: resolve "_IOWR redefined" build error on SLES10
1049 Description: dump the memory debugging after all modules are unloaded to
1050 suppress false negative in conf_sanity test 39
1052 Severity : enhancement
1054 Description: build kernel-ib packages for OFED 1.3 in our release cycle
1059 Description: fix SLES kernel versioning
1060 Details : the kernel version for our SLES 10 kernel did not include a "-"
1061 before the "smp" at the end. while this was not a problem in
1062 general, it did mean that software trying to use the kernel
1063 version to try to detect a vendor specific kernel would fail.
1064 this was most evident by the OFED build scripts.
1068 Description: Don't update lov_desc members until making sure they are valid
1069 Details : When updating lov_desc members via proc fs, need fix their
1070 validities before doing the real update.
1074 Description: don't put request into delay list while invalidate in flight.
1075 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
1076 import in flight. this produce timeout for invalidate and sometimes
1077 can cause stale data.
1081 Frequency : on ppc only
1082 Description: not convert ost objects for directory because it's not exist.
1083 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
1087 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
1089 Severity : enhancement
1091 Description: Update to support 2.6.22.14 vanilla kernel.
1095 Frequency : rare, on recovery
1096 Description: read procfs can produce deadlock in some situation
1097 Details : Holding lprocfs lock which send rpc can produce block for destroy
1098 obd objects and this also block reconnect with -EALREADY. This
1099 isn't fix all lprocfs bugs - but make it rare.
1101 Severity : enhancement
1103 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
1106 Frequency : frequent on X2 node
1108 Description: mdc_set_open_replay_data LBUG
1109 Details : Set replay data for requests that are eligible for replay.
1113 Description: lustre_mgs: operation 101 on unconnected MGS
1114 Details : When MGC is disconnected from MGS long enough, MGS will evict the
1115 MGC, and late on MGC cannot successfully connect to MGS and a lot
1116 of the error messages complaining that MGS is not connected.
1120 Frequency : on network error
1121 Description: panic with double free request if network error
1122 Details : mdc_finish_enqueue is finish request if any network error ocuring,
1123 but it's true only for synchronus enqueue, for async enqueue
1124 (via ptlrpcd) this incorrect and ptlrpcd want finish request
1127 Severity : enhancement
1129 Description: client-side metadata stat-ahead during readdir(directory readahead)
1130 Details : perform client-side metadata stat-ahead when the client detects
1131 readdir and sequential stat of dir entries therein
1134 Frequency : on start mds
1136 Description: Implement get_info(last_id) in obdfilter.
1139 Frequency : occasional
1141 Description: Correctly check stale fid, not start epoch if ost not support SOM
1142 Details : open with flag O_CREATE need set old fid in op_fid3 because
1143 op_fid2 was overwritten with new generated fid, but mds can answer
1144 with one of these two fids and both is not stale. Setattr
1145 incorrectly started an epoch and assume will be called
1146 done_writing, but without SOM done_writing ever being called.
1149 Frequency : rare, depends on device drivers and load
1151 Description: MDS or OSS nodes crash due to stack overflow
1152 Details : Code changes in 1.8.0 increased the stack usage of some functions.
1153 In some cases, in conjunction with device drivers that use a lot
1154 of stack the MDS (or possibly OSS) service threads could overflow
1155 the stack. One change which was identified to consume additional
1156 stack has been reworked to avoid the extra stack usage.
1159 Frequency : occasional
1161 Description: Do not fail import if osc_interpret_create gets -EAGAIN
1162 Details : If osc_interpret_create got -EAGAIN it immediately exits and
1163 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
1164 oscc_has_objects and see OSC has no objests and call
1165 oscc_internal_create to resend create request.
1167 Severity : enhancement
1169 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
1171 Severity : enhancement
1173 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
1176 Frequency : very rare
1178 Description: Fix replay if there is an un-replied request and open
1179 Details : In some cases, older replay request will revert the
1180 mcd->mcd_last_xid on MDS which is used to record the client's
1181 latest sent request.
1183 Severity : enhancement
1185 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
1187 Severity : enhancement
1189 Description: Add rhel5 support to HEAD.
1191 Serverity : enhancement
1193 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
1198 Description: Don't allow skipping OSTs if index has been specified.
1199 Details : Don't allow skipping OSTs if index has been specified, make
1200 locking in internal create lots better.
1204 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
1205 Details : access to bitfield in structure is always rounded to long
1206 and this produce problem with not atomic change any bit.
1210 Description: Lustre make rpms failed.
1211 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
1212 builds Lustre rpms from tarball.
1215 Frequency : rare on shutdown ost
1217 Description: If llog cancel was not send before clean_exports phase, this can
1218 produce deadlock in llog code.
1219 Details : If llog thread has last reference to obd and call class_import_put
1220 this produce deadlock because llog_cleanup_commit_master wait when
1221 last llog_commit_thread exited, but this never success because was
1222 called from llog_commit_thread.
1226 Description: allow userland application know is lost one of stripes.
1227 Details : fill lvb_blocks with error code on ost and return it to
1228 application if error flag found.
1232 Description: NULL lov_tgts causing MDS oops
1233 Details : more safe checks for NULL lov_tgts for avoid oops.
1235 Severity : enhancement
1237 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
1241 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
1242 Details : If an OST is down the MDS will hang indefinitely in
1243 obd_statfs_rqset() waiting for the statfs data. While for
1244 MDS QOS usage of statfs, it should not stuck in waiting.
1246 Severity : enhancement
1248 Description: remote_acl support
1249 Details : Support ACL-based permission check for remote user.
1250 Support setfacl/getfacl for remote user with the utils
1251 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
1252 the system "{s,g}etfacl" utils.
1254 Severity : enhancement
1256 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
1258 Severity : enhancement
1260 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
1264 Description: make mgs_setparam() handle fsname containing dash
1265 Details : fsname containing a dash does not work with lctl conf_param
1267 Severity : enhancement
1269 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
1271 Severity : enhancement
1273 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
1277 Frequency : rare, at shutdown
1278 Description: access already free / zero obd_namespace.
1279 Details : if client_disconnect_export was called without force flag set,
1280 and exist connect request in flight, this can produce access to
1281 NULL pointer (or already free pointer) when connect_interpret
1282 store ocd flags in obd_namespace.
1286 Frequency : only at startup
1287 Description: not alloc memory with spinlock held.
1288 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
1289 if any spinlock held.
1291 Severity : enhancement
1293 Description: make lustre randomly fail allocating memory
1294 Details : Make lustre randomly failed allocating memory for testing purpose.
1296 Severity : enhancement
1298 Description: lost problems with lov objid file
1299 Details : Fixes some scability and access to not inited memory problems
1300 in work with lov objdid file.
1305 Description: lfs find does not continue on file error
1306 Details : Continue other files processing when a file/dir is absent.
1310 Description: Inconsistent usage of lustre_pack_reply()
1311 Details : Standardize the usage of lustre_pack_reply() such that it
1312 always generate a CERROR on failure.
1315 Frequency : occasional
1317 Description: MDS hang or stay in waiting lock
1318 Details : If client receive lock with CBPENDING flag ldlm need send lock
1319 cancel as separate rpc, to avoid situation when cancel request
1320 can't processed due all i/o threads stay in wait lock.
1324 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1328 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1332 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1336 Description: Update to RHEL4 latest kernel.
1338 Severity : enhancement
1340 Description: Build SLES10 patchless client fails
1341 Details : The configure was broken by run ./configure with
1342 --with-linux-obj=.... argument for patchless client. When the
1343 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1344 can't search header adequately. Use absolute path such as
1345 -I($LINUX)/include instead.
1349 Description: interrupt oig_wait produce painc on resend.
1350 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1351 private set, and this produce situation when rq_ptlrpcd_data not
1352 copyed to new allocated request and triggered LBUG on assert
1353 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1354 ptlrpcd set if request is changed and can be safety changed to use
1359 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1360 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1361 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1362 these macros and replace them with their difinitions instead.
1366 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1367 Details : Update which_patch & target file for SLES10 latest kernel.
1370 Description: Remove client patches from SLES 10 kernel.
1371 Details : This causes SLES 10 clients to behave as patchless clients
1372 even on a Lustre-patched (server) kernel.
1374 Severity : enhancement
1376 Description: self-adjustable client's lru lists
1377 Details : use adaptive algorithm for managing client cached locks lru
1378 lists according to current server load, other client's work
1379 pattern, memory activities, etc. Both, server and client
1380 side namespaces provide number of proc tunables for controlling
1383 Severity : enhancement
1385 Description: light-weight GSS support
1386 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
1387 and reduce performance overhead.
1389 Severity : enhancement
1391 Description: Linux keyring support
1392 Details : Support using service of Linux keyring for Lustre GSS internal
1393 context refresh/cache mechanism.
1397 Description: Fix errors in lfs documentation
1398 Details : Fixes man pages
1402 Description: don't use GFP_* in generic Lustre code.
1403 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
1407 Description: obdclass is limited by single OBD_ALLOC(idarray)
1408 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
1412 Description: warnings with build patchless client with vanila 2.6.19 and up
1413 Details : change old ctl_table style and replace ctl_table/ctl_table_header
1414 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
1418 Frequency : Only for SLES
1419 Description: sanity_quota fail test_1
1420 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
1421 file, which makes TSTID[2] inunique.
1425 Frequency : Only for SLES9
1426 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
1427 Details : Update target/ChangeLog/which_patch .
1431 Frequency : Only for test_10 in sanity-quota.sh
1432 Description: a bug in quota test code
1433 Details : checking a test flag in a wrong place
1437 Frequency : Only for test_13 in sanity-quota.sh
1438 Description: a bug in quota test script
1439 Details : shouldn't check overfull allocated quota
1443 Description: jbd statistics
1444 Details : Port older jbd statistics patch for sles10
1446 Severity : enhancement
1448 Description: Add printing inode info into message about error in writepage.
1453 Description: reply_lock_interpret crash due to race with it and lock cancel.
1454 Details : Do not replay locks that are being cancelled. Do not reference
1455 locks by their address during replay, just by their handle.
1459 Frequency : When flocks are used.
1460 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
1461 Details : Flock locks might destroy just granted lock if it could be merged
1462 with another existing flock, this is done in completion handler,
1463 so teach ldlm_cli_enquque_fini that this is a valid case for
1469 Description: Oops in read and write path when failing to allocate lock.
1470 Details : Check if lock allocation failed and return error back.
1474 Description: lstripe command fails for valid OST index
1475 Details : The stripe offset is compared to lov->desc.ld_tgt_count
1476 instead of lov->desc.ld_active_tgt_count.
1480 Description: sanity.sh failed test 103
1481 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
1486 Description: df doesn't work properly if diskfs blocksize != 4K
1487 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
1491 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
1492 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
1494 Severity : enhancement
1496 Description: Adaptive timeouts
1497 Details : RPC timeouts adapt to changing server load and network
1498 conditions to reduce resend attempts and improve recovery time.
1502 Description: llapi_file_create() does not allow some changes
1503 Details : add llapi_file_open() that allows specifying the mode and
1504 open flags, and also returns an open file handle.
1508 Description: merge and cleanup kernel patches.
1509 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
1513 Description: Add journal checksum support.(Kernel part)
1514 Details : The journal checksum feature adds two new flags i.e
1515 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
1516 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
1517 indicates that the commit block contains the checksum for
1518 the blocks described by the descriptor blocks. Now commit
1519 record can be sent to disk without waiting for descriptor
1520 blocks to be written to disk. This behavior is controlled
1521 using JBD2_FEATURE_ASYNC_COMMIT flag.
1525 Description: OSS needs mutliple precreate threads
1526 Details : Add ability to start more than one create thread per OSS.
1530 Description: Sanity.sh test_65k failed: setstripe should have succeeded
1531 Details : execute lfs setstripe on client
1535 Description: mds_obd_create error creating tmp object
1536 Details : When the user sets quota on root, llog will be affected and can't
1537 create files and write files.
1540 Frequency : Always on ia64 patchless client, and possibly others.
1542 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
1543 Details : This allows the patchless client to be loaded on architectures
1544 without this export.
1548 Description: RedHat Update kernel for RHEL5
1549 Details : Add the kernel config file for RHEL5.
1553 Description: RedHat Update kernel for RHEL5
1554 Details : Modify the kernel config file more closer RHEL5.
1558 Description: Build failure against Centos5 (RHEL5)
1559 Details : Define PAGE_SIZE when it isn't present.
1563 Description: test_7 on sanity-quota.sh can't be used on separate machines
1564 Details : add facet to handle it
1568 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
1569 Details : with NFS, the anon dentry's parent was set to itself in
1570 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
1571 corresponding dentry other than use rec->ur_name.
1573 Severity : enhancement
1575 Description: lfs setstripe enhancement
1576 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
1580 Description: enable data checksumming by default
1581 Details : enable checksum by default, allow --disable-checksum
1582 configure option and "-o nochecksum" mount option. Checksums
1583 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
1587 Description: lustre support for RHEL5
1588 Details : Add support for RHEL5.
1592 Description: Client eviction due to failover config
1593 Details : after a connection loss, the lustre client should attempt to
1594 reconnect to the last active server first before trying the
1595 other potential connections.
1599 Description: when mds and osts use different quota unit(32bit and 64bit),
1600 quota will be released repeatly.
1601 Details : void sending multiple quota reqs to mds, which will keep the
1602 status between the reqs.
1606 Description: osts not allocated evenly to files
1607 Details : change the condition to increase offset_idx
1610 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
1612 Description: Data corruption for OSTs that are formatted larger than 2TB
1614 Details : When generating the bio request for lustre file writes the
1615 sector number would overflow a temporary variable before being
1616 used for the IO. The data reads correctly from Lustre (which
1617 will overflow in a similar manner) but other file data or
1618 filesystem metadata may be corrupted in some cases.
1622 Description: Tune the kernel for good SCSI performance.
1623 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
1624 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
1629 Description: rewrite ext2-derived code in obdclass/uuid.c
1630 Details : rewrite inherited code (uuid parsing code from ext2 utils)
1631 from scratch preserving functionality.
1635 Description: "lfs find -obd UUID" prints directories
1636 Details : "lfs find -obd UUID" will return all directory names instead
1637 of just file names. It is incorrect because the directories
1638 do not reside on the OSTs.
1642 Description: use i_size_read and i_size_write in 2.6 port
1643 Details : replace inode->i_size access with i_size_read/write()
1646 Frequency : when removing large files
1648 Description: scheduling issue during removal of large Lustre files
1649 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
1650 It causes scheduling issues when removing large files (17TB in the
1654 Frequency : only with liblustre clients on XT3
1656 Description: evictions taking too long
1657 Details : allow llrd to evict clients directly on OSTs
1660 Frequency : only on ppc
1662 Description: /proc/fs/lustre/devices broken on ppc
1663 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
1664 We should make sure the type of 'index' is loff_t and then cast
1665 explicitly as needed below. Do not assign an explicitly cast
1669 Frequency : only for rhel5
1671 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
1672 Details : Modify the target file & which_kernel.
1674 Severity : enhancement
1676 Description: omit set fsid for export NFS
1677 Details : fix set/restore device id for avoid EMFILE error and mark lustre
1678 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
1682 Frequency : Always, for kernels after 2.6.16
1683 Description: Fix warning idr_remove called for id=.. which is not allocated.
1684 Details : Last kernels save old s_dev before kill super and not allow
1685 to restore from callback - restore it before call kill_anon_super.
1689 Description: buffer overruns could theoretically occur
1690 Details : llapi_semantic_traverse() modifies the "path" argument by
1691 appending values to the end of the origin string, and a
1692 overrun may occur. Adding buffer overrun check in liblustreapi.
1696 Description: Fix error on 'ls .' at the top of the Lustre mount.
1697 Details : Don't revalidate dentry if it is a root dentry.
1701 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
1702 Details : Modify vm-tunables-rhel4.patch.
1706 Description: Kernel config for 2.6.18-vanilla.
1707 Details : Modify targets/2.6-vanilla.target.in.
1708 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
1709 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
1710 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
1711 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
1715 Description: improve handling recoverable errors
1716 Details : if request processig with error which can be recoverable on server
1717 request should be resend, otherwise page released from cache and
1722 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
1723 Details : Modify target file & which_patch.
1724 A flaw was found in the IA32 system call emulation provided
1725 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
1726 value could be stored in the %RAX register, which could trigger an
1727 out-of-bounds system call table access. An untrusted local user
1728 could exploit this flaw to run code in the kernel
1729 (ie a root privilege escalation). (CVE-2007-4573).
1733 Description: O_DIRECT bypasses client statistics.
1734 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
1735 read_ahead_stats were updated. copied the stats section from
1736 osc_send_oap_rpc() into async_internal().
1740 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
1744 Description: change order of libsysio includes
1745 Details : '#include sysio.h' should always come before '#include xtio.h'
1749 Description: handle "serious error: objid * already exists" more gracefully
1750 Details : If LAST_ID value on disk is smaller than the objects existing in
1751 the O/0/d* directories, it indicates disk corruption and causes an
1752 LBUG(). If the object is 0-length, then we should use the existing
1753 object. This will help to avoid a full fsck in most cases.
1755 Severity : enhancement
1757 Description: adapt the lustre_config script to support the upgrade case
1758 Details : Add "-u" option for lustre_config script to support upgrading 1.4
1759 server targets to 1.6 in parallel.
1763 Description: To avoid grant space > avaible space when the disk is almost
1764 full. Without this patch you might see the error "grant XXXX >
1765 available" or some LBUG about grant, when the disk is almost
1767 Details : In filter_check_grant, for non_grant cache write, we should
1768 check the left space by if (*left > ungranted + bytes), instead
1769 of (*left > ungranted), because only we are sure the left space
1770 is enough for another "bytes", then the ungrant space should be
1771 increase. In client, we should update cl_avail_grant only there
1772 is OBD_MD_FLGRANT in the reply.
1776 Description: Update RHEL 4 kernel to fix local root privilege escalation.
1777 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
1778 described in CVE-2007-4573. This problem could allow untrusted
1779 local users to gain root access.
1782 Frequency : when using O_DIRECT and quotas
1784 Description: Incorrect file ownership on O_DIRECT output files
1785 Details : block usage reported by 'lfs quota' does not take into account
1786 files that have been written with O_DIRECT.
1791 Description: touch file failed when fs is not full
1792 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
1793 otherwise we can get ENOSP while fs is not full.
1797 Description: parallel lock callbacks
1798 Details : Instead of sending blocking and completion callbacks as separated
1799 requests, adding them to a set and sending in parallel.
1802 Frequency : only for Cray XT3
1803 Bugzilla : 12829/13455
1804 Description: Changing primary group doesn't change the group lustre assigns to
1806 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
1807 overridden with the primary group provided by the group upcall,
1808 whereas the supplied fsgid can be trusted if it is in the list of
1809 supplementary groups returned by the group upcall.
1811 Severity : enhancement
1813 Description: Allow masking D_WARNING, D_ERROR messages from console
1814 Details : Console messages can now be disabled via lnet.printk.
1818 Description: User code with malformed file open parameter crashes client node
1819 Details : Before packing join_file req, all the related reference should be
1820 checked carefully in case some malformed flags cause fake
1821 join_file req on client.
1825 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
1826 Details : ldlm_completion_ast() assumes that a lock is granted when the req
1827 mode is equal to the granted mode. However, it should also check
1828 that LDLM_FL_CP_REQD is not set.
1832 Description: Heavy nfs access might result in deadlocks
1833 Details : After ELC code landed, it is now improper to enqueue any mds
1834 locks under och_sem, because enqueue might want to decide to
1835 cancel open locks for same inode we are holding och_sem for.
1839 Description: Client eviction while running blogbench
1840 Details : A lot of unlink operations with concurrent I/O can lead to a
1841 deadlock causing evictions. To address the problem, the number of
1842 oustanding OST_DESTROY requests is now throttled to
1843 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
1844 callbacks are processed in priority.
1848 Description: enable ACLs on MDS by default
1849 Details : ACLs must be enabled on MDS by default.
1852 Frequency : PPC/PPC64 only
1854 Description: conflicts between asm-ppc64/types.h and lustre_types.h
1855 Details : fix duplicated definitions between asm-ppc64/types.h and
1856 lustre_types.h on PPC.
1859 Frequency : PPC/PPC64 only
1861 Description: asm-ppc/segment.h does not exist
1862 Details : fix compile issue on PPC.
1866 Description: better handle error messages in extents code
1869 Frequency : RHEL4 only
1871 Description: mkfs is very slow on IA64/RHEL4
1872 Details : A performance regression has been discovered in the MPT Fusion
1873 driver between versions 3.02.73rh and 3.02.99.00rh. As a
1874 consequence, we have downgraded the MPT Fusion driver in the RHEL4
1875 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
1877 Severity : enhancement
1879 Description: SNMP support enhancement
1880 Details : Adding total number of sampled request for an MDS node in snmp
1883 Severity : enhancement
1885 Description: Optimize ldlm waiting list processing for PR extent locks
1886 Details : When processing waiting list for read extent lock and meeting read
1887 lock that is same or wider to it that is not contended, skip
1888 processing rest of the list and immediatelly return current
1889 status of conflictness, since we are guaranteed there are no
1890 conflicting locks in the rest of the list.
1894 Description: Time out and refuse to reconnect
1895 Details : When the failover node is the primary node, it is possible
1896 to have two identical connections in imp_conn_list. We must
1897 compare not conn's pointers but NIDs, otherwise we can defeat
1898 connection throttling.
1902 Description: port llog fixes from b1_6 into HEAD
1903 Details : Port llog reference couting and some llog cleanups from b1_6
1904 (bug 10800) into HEAD, for protect from panic and access to already
1905 free llog structures.
1909 Description: Detect stride IO mode in read-ahead
1910 Details : When a client does stride read, read-ahead should detect that and
1911 read-ahead pages according to the detected stride pattern.
1915 Description: data checksumming impacts single node performance
1916 Details : add support for several checksum algorithm. Currently, only CRC32
1917 and Adler-32 are supported. The checksum type can be changed on
1918 the fly via /proc/fs/lustre/osc/*/checksum_type.
1922 Description: use adler32 for page checksums
1923 Details : when available, use the Adler-32 algorithm instead of CRC32 for
1928 Description: build for x2 fails
1929 Details : fix compile issue on Cray systems.
1933 Description: Properly match for duplicate locks
1934 Details : Due to different lock order from skiplists code, we need to
1935 traverse entire list for now
1938 Frequency : only on PPC/SLES10
1940 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
1941 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
1942 be defined. Add a hack in mkfs_lustre.c to work around this compile
1947 Description: LASSERT on MDS when client holding flock lock dies
1948 Details : ldlm pool logic depends on number of granted locks equal to
1949 number of released locks which is not true for flock locks, so
1950 just exclude such locks from consideration.
1954 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
1955 Details : Use fsfilt_sync() for both the whole filesystem sync and
1956 individual file sync to eliminate dangerous inode locking
1957 with I_LOCK that can lead to a deadlock.
1961 Description: performance in 1.6.3
1962 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
1963 sure that 1MB requests are not fragmented by the block layer.
1965 Severity : enhancement
1967 Description: organize the server-side client stats on per-nid basis
1968 Details : Change the structure of stats under obdfilter and mds to
1976 The "uuid"s file would list the uuids of _active_ exports.
1977 And the clear entry is to clear all stats and stale nids.
1979 Severity : enhancement
1981 Description: eliminate client locks in face of contention
1982 Details : file contention detection and lockless i/o implementation
1983 for contended files.
1987 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
1990 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
1992 Description: MDS may lose file striping (and hence file data) in some cases
1993 Details : If there are additional extended attributes stored on the MDS,
1994 in particular ACLs, SELinux, or user attributes (if user_xattr
1995 is specified for the client mount options) then there is a risk
1996 of attribute loss. Additionally, the Lustre file striping
1997 needs to be larger than default (e.g. striped over all OSTs),
1998 and an additional attribute must be stored initially in the
1999 inode and then increase in size enough to be moved to the
2000 external attribute block (e.g. ACL growing in size) for file
2005 Description: skiplist implementation simplification
2006 Details : skiplists are used to group compatible locks on granted list
2007 that was implemented as tracking first and last lock of each
2008 lock group the patch changes that to using doubly linked lists
2012 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
2013 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
2014 because LASSERT on both the data supplied by a client, and the
2015 data on disk is dangerous and incorrect.
2017 Severity : enhancement
2019 Description: Slow truncate/writes to huge files at high offsets.
2020 Details : Directly associate cached pages to lock that protect those pages,
2021 this allows us to quickly find what pages to write and remove
2022 once lock callback is received.
2026 Description: more ldlm soft lockups
2027 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
2028 starve other threads from the resource lock for a long time in
2029 case of long waiting queue, so change the debug level from
2030 D_OTHER to the less frequently used D_INFO.
2032 Severity : enhancement
2034 Description: add -gid, -group, -uid, -user options to lfs find
2038 Description: Hung threads in invalidate_inode_pages2_range
2039 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
2040 once one is completed. As a result, some RPCs are stuck in the
2041 queue and are never sent.
2045 Description: filter threads hungs on waiting journal commit
2046 Details : Cleanup filter group llog code, then only filter group llog will
2047 be only created in the MDS/OST syncing process.
2051 Description: Procfs and llog threads access destoryed import sometimes.
2052 Details : Sync the import destoryed process with procfs and llog threads by
2053 the import refcount and semaphore.
2055 Severity : enhancement
2057 Description: openlock cache of b1_6 port to HEAD
2062 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
2063 Details : If insertion of an extent fails, then discard the inode
2064 preallocation and free data blocks else it can lead to duplicate
2069 Description: don't always update ctime in ext3_xattr_set_handle()
2070 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
2071 In some cases the ctime should not be updated, for example for
2072 2.0->1.8 compatibility it is necessary to delete an xattr and it
2073 should not update the ctime.
2077 Bugzilla : 15713/16362
2078 Description: Assertion in iopen_connect_dentry in 1.6.3
2079 Details : looking up an inode via iopen with the wrong generation number can
2080 populate the dcache with a disconneced dentry while the inode
2081 number is in the process of being reallocated. This causes an
2082 assertion failure in iopen since the inode's dentry list contains
2083 both a connected and disconnected dentry.
2087 Description: assertion failure in ldlm_handle2lock()
2088 Details : fix a race between class_handle_unhash() and class_handle2object()
2089 introduced in lustre 1.6.5 by bug 13622.
2094 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
2095 Details : remove the limit of 256 scsi disks in the sd_iostat patch
2100 Description: oops in sd_iostats_seq_show()
2101 Details : unloading/reloading the scsi low level driver triggers a kernel
2102 bug when trying to access the sd iostat file.
2107 Description: Kernel panics during QLogic driver reload
2108 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
2109 patch, causing memory corruption.
2114 Description: journal_dev option does not work in b1_6
2115 Details : pass mount option during pre-mount.
2117 Severity : enhancement
2119 Description: Add a FIEMAP(FIle Extent MAP) ioctl
2120 Details : FIEMAP ioctl will allow an application to efficiently fetch the
2121 extent information of a file. It can be used to map logical blocks
2122 in a file to physical blocks in the block device.
2126 Description: LDLM soft lockups - improvement
2127 Details : It is be possible to send the lock handle along with each read
2128 or write request because the client is already doing a lock match
2129 itself so there isn't any reason the OST should have to re-do that
2133 Frequency : only on Cray X2
2135 Description: X2 build failures
2136 Details : fix build failures on Cray X2.
2140 Description: xid & resent requests
2141 Details : Initialize RPC XID from clock at startup (randomly if clock is
2144 Severity : enhancement
2146 Description: Add lustre_start utility to start or stop multiple Lustre servers
2151 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
2152 Details : In case of memory pressure, list_del() can be called twice on
2153 req->rq_history_list, causing a kernel oops.
2157 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
2158 Details : fix stack overflow in the distributed lock manager by defering
2159 export eviction after a failed AST to the elt thread instead of
2160 handling it in the dlm interpret routine.
2164 Description: Convert some comments to new format.
2165 Details : Update documenting comments to match doxygen conventions.
2169 Description: Grammar fixes.
2170 Details : A couple of trivial spelling fixes.
2174 Description: OSD_COUNTERS-mandatory
2175 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
2176 to simplify the code.
2180 Description: simplify cmm_device freeing logic.
2181 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
2185 Description: Add lockdep support to dt_object_operations locking interface.
2186 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
2187 parameter indicating lock ordering. Update mdd code to use new
2192 Description: Introduce failloc constants for lockless IO tests.
2193 Details : Add two new failloc constants to test lockless IO. Only one of
2194 them in implemented---another is checked in yet to be landed
2199 Description: Add lockdep support for inode mutex.
2200 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
2201 in the situations where Lustre has to lock more than one inode
2206 Description: Add optional invariants checking support.
2207 Details : Add new LINVRNT() macro, optional on new --enable-invariants
2208 configure switch. This macro is to be used for consistency and
2209 sanity checks that are too expensive to be left in `production'
2214 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
2215 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
2220 Description: Add lockdep support for ldlm_lock and ldlm_resource.
2221 Details : Use spin_lock_nested() in (the only) situation where more than
2222 one ldlm_lock is locked simultaneously. Also, fix possible
2223 dead-lock in ldlm_lock_change_resource() by enforcing particular
2228 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
2229 Details : Instead of specifying each ldlm_lock call-back through separate
2230 parameter, wrap them into struct ldlm_callback_suite.
2234 Description: Kill join_lru obd method and its callers.
2235 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
2236 in memory---a requirement implemented through ->o_join_lru() obd
2237 method in HEAD. Remove this method and its users.
2241 Description: Add asynchronous ldlm ENQUEUE completion handler.
2242 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
2243 case that stock ldlm_completion_ast() cannot handle as it waits
2244 until lock is granted. Introduce new ldlm_completion_ast_async()
2245 for this. Also comment ldlm_completion_ast().
2249 Description: ldlm_error <-> errno conversion.
2250 Details : Add functions to map (rather arbitrary) between LDLM error codes
2251 and standard errno values. CLIO needs this to prevent LDLM specific
2252 constants from escaping ldlm and osc.
2256 Description: Kill unused ldlm_handle2lock_ns() function.
2257 Details : Kill unused ldlm_handle2lock_ns() function.
2261 Description: Add lu_ref support to ldlm_lock
2262 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
2263 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
2264 and ldlm_resource. LDLM interface has to be changed, because code that
2265 releases a reference on a lock, has to "know" what reference this is.
2266 In the most frequent case
2268 lock = ldlm_handle2lock(handle);
2270 LDLM_LOCK_PUT(lock);
2272 no changes are required. When any other reference (received _not_ from
2273 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
2274 instead of LDLM_LOCK_PUT().
2276 Arguably, changes are pervasive, and interface requires some discipline
2277 for proper use. On the other hand, it was very instrumental in finding
2278 a few leaked lock references.
2282 Description: Add ldlm_lock_addref_try().
2283 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
2284 attempts to addref a lock that might be being canceled
2289 Description: Add ldlm_weigh_callback().
2290 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
2291 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
2292 hard-coded `number of pages' logic.
2296 Description: Add lockdep annotations to llog code.
2297 Details : Use appropriately tagged _nested() locking calls in the places
2298 where llog takes more than one ->lgh_lock lock.
2302 Description: Add loi_kms_set().
2303 Details : Wrap kms updates into a helper function.
2307 Description: Constify instances of struct lsm_operations.
2308 Details : Constify instances of struct lsm_operations.
2312 Description: lu_conf support.
2313 Details : On a server, a file system object is uniquely identified
2314 by a fid, which is sufficient to locate and load all object
2315 state (inode). On a client, on the other hand, more data are
2316 necessary instantiate an object. Change lu_object_find() and
2317 friends to take additional `lu_conf' argument describing object.
2318 Typically this includes layout information.
2322 Description: lu_context fixes.
2323 Details : Introduce new lu_context functions that are needed on the client
2324 side, where some system threads (ptlrpcd) are shared by multiple
2325 modules, and so cannot be stopped during module shutdown.
2329 Description: Add start and stop methods to lu_device_type_operations.
2330 Details : Introduce two new methods in lu_device_type_operations, that are
2331 invoked when first instance of a given type is created and last one
2332 is destroyed respectively. This is need by CLIO.
2336 Description: Add lu_ref support to struct lu_device.
2337 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
2338 track leaked references.
2342 Description: Introduce lu_kmem_descr.
2343 Details : lu_kmem_descr and its companion interface allow to create
2344 and destroy a number of kmem caches at once.
2348 Description: Fix lu_object finalization race.
2349 Details : Fix a race between lu_object_find() finding an object and its
2350 concurrent finalization. This race is (most likely) not possible
2351 on the server, but might happen on the client.
2355 Description: Introduce lu_ref interface.
2356 Details : lu_ref is a debugging module allowing to track references to
2357 a given object. It is quite cpu expensive, and has to be
2358 explicitly enabled with --enable-lu_ref. See usage description
2363 Description: Factor lu_site procfs stats into a separate function.
2364 Details : Separate lu_site stats printing code into a separate function
2365 that can be reused on a client.
2369 Description: Constify instances of struct {lu,dt,md}_device_operations.
2370 Details : Constify instances of struct {lu,dt,md}_device_operations.
2374 Description: Introduce struct md_site and move meta-data specific parts of
2375 struct lu_site here.
2376 Details : Move md-specific fields out of struct lu_site into special struct
2377 md_site, so that lu_site can be used on a client.
2381 Description: Kill mdd_lov_destroy().
2382 Details : Remove unused mdd code.
2386 Description: Add st_block checking to multistat.c.
2387 Details : Add st_block checking to multistat.c.
2391 Description: Add lu_ref support to struct obd_device.
2392 Details : Add lu_ref tracking to obd_device.
2396 Description: Kill obd_set_fail_loc().
2397 Details : Remove unused code.
2401 Description: Add special type for ptlrpc_request interpret functions.
2402 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
2403 there. Actual usage will be in CLIO.
2407 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
2408 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
2409 doesn't work with lockdep.
2413 Description: Add rwv.c test program.
2414 Details : New testing program exercising readv(2) and writev(2) (Qian).
2418 Description: Add sendfile.c test program.
2419 Details : New testing program exercising sendfile(2) (Jay).
2423 Description: Ratelimit a message that can be very frequent.
2424 Details : Ratelimit a memory allocation failure message that can
2429 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2430 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2434 Description: Liblustre build fixes.
2435 Details : Liblustre build fixes.
2439 Description: libcfs: add cfs_{need,cond}_resched() interface.
2440 Details : libcfs: add cfs_{need,cond}_resched() definition and
2441 implementations for Linux, NT, and liblustre.
2443 Severity : enhancement
2445 Description: More exported tunables for mballoc
2446 Details : Add support for tunable preallocation window and new tunables for
2447 large/small requests
2451 Description: Detect corruption of block bitmap and checking for preallocations
2452 Details : Checks validity of on-disk block bitmap. Also it does better
2453 checking of number of applied preallocations. When corruption is
2454 found, it turns filesystem readonly to prevent further corruptions.
2458 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
2459 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
2460 window to 0. If it does hit cache miss, and read-pattern is still
2461 stride-io mode, does not reset the stride window, but also does not
2462 increase the stride window length in this case.
2466 Frequency : only for big-endian servers
2467 Description: Check if system is big-endian while mounting fs with extents feature
2468 Details : Mounting a filesystem with extents feature will fail on big-endian
2469 systems since ext3-based ldiskfs is not supported on big-endian
2470 systems. This can be over-riden with "bigendian_extents" mount option.
2472 Severity : enhancement
2474 Description: The root squash functionality
2475 Details : A security feature, which is to prevent users from being able
2476 to mount lustre on their desktop, run as root, and delete
2477 all of the files in the filesystem. The goal is accomplished by
2478 remapping user id (UID) and group id (GID) of the root user to
2479 a UID and GID specified by the system administartor via Lustre
2480 configuration management server (MGS). The functionality also
2481 allows to specify sets of clients for which the remapping does
2486 Description: Excessive recovery window
2487 Details : With AT enabled, the recovery window can be excessively long (6000+
2488 seconds). To address this problem, we no longer use
2489 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
2490 timeout no longer depends on the service time, it is set to
2491 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
2492 time via pb_service_time.
2496 Description: Watchdog triggered on MDS failover
2497 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
2498 the OSTs know that the MDS "UUID" can be reused for the same export
2499 from a different NID, so we do not need to wait for the export to be
2503 Frequency : rare, only if using MMP with Linux RAID
2505 Description: MMP doesn't work with Linux RAID
2506 Details : While using HA for Lustre servers with Linux RAID, it is possible
2507 that MMP will not detect multiple mounts. To make this work we
2508 need to unplug the device queue in RAID when the MMP block is being
2509 written. Also while reading the MMP block, we should read it from
2510 disk and not the cached one.
2512 Severity : enhancement
2514 Description: open file using fid
2515 Details : A file can be opened using just its fid, like
2516 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
2519 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
2522 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
2523 Details : add proper locking for creation of jbd_4k slab cache
2527 Description: MMP check in ext3_remount() fails without displaying any error
2528 Details : When multiple mount protection fails during remount, proper error
2531 Severity : enhancement
2533 Description: Allow stripe size to be up to 4G-64k
2534 Details : Fix math logic to allow large stripe sizes.
2538 Description: add check for >8TB ldiskfs filesystems
2539 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
2540 Don't allow >8TB ldiskfs filesystems to be mounted without
2541 force_over_8tb mount option
2546 Descriptoin: OSS complains "too many blocks" and crashes
2547 Details : In filter_commitrw_write():
2548 1. lquota_chkquota();
2549 2. obdo_from_inode();
2550 3. lquota_pending_commit();
2551 "2." has changed uid and gid in this case, then caused
2552 lquota_chkquota() and lquota_pending_commit() to use
2553 different uid and gid. That is the root of the bug.
2557 Descriptoin: getstripe doesn't report stripe count = -1 sanely
2558 Details : Detect and print out the 'stripe count = -1' case properly.
2562 Description: Out or order replies might be lost on replay
2563 Details : In ptlrpc_retain_replayable_request if we cannot find retained
2564 request with tid smaller then one currently being added, add it
2565 to the start, not end of the list.
2567 --------------------------------------------------------------------------------
2569 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
2571 * Support for kernels:
2572 2.6.5-7.283 (SLES 9),
2573 2.6.9-55.EL (RHEL 4),
2574 2.6.16.46-0.14 (SLES 10),
2575 2.6.18.8 vanilla (kernel.org)
2576 * Client support for unpatched kernels:
2577 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2578 2.6.16 - 2.6.22 vanilla (kernel.org)
2579 * Due to recently discovered recovery problems, we do not recommend
2580 using patchless RHEL 4 clients with this or any earlier release.
2581 * Recommended e2fsprogs version: 1.39.cfs8
2582 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2583 * Starting with this release, the ldiskfs backing filesystem required
2584 by Lustre is now in its own package, lustre-ldiskfs. This package
2585 should be installed. It is versioned separately from Lustre and
2586 may be released separately in future.
2590 Description: block reactivating mgc import until all deactivates complete
2591 Details : Fix race when failing back MDT/MGS to itself (testing)
2594 Frequency : at statup only
2596 Description: mds_lov_synchronize race leads to various problems
2597 Details : simultaneous MDT->OST connections at startup can cause the
2598 sync to abort, leaving the OSC in a bad state.
2600 Severity : enhancement
2602 Description: add optional extra BUILD_VERSION info
2603 Details : add a new environment variable (namely LUSTRE_VERS) which allows
2604 to override the lustre version.
2607 Frequency : 2.6.18 servers only
2609 Description: ll_kern_mount() doesn't release the module reference
2610 Details : The ldiskfs module reference count never drops down to 0
2611 because ll_kern_mount() doesn't release the module reference.
2616 Description: server LBUG when using old ost_num_threads parameter
2617 Details : Accept the old ost_num_threads parameter but warn that it
2618 is deprecated, and fix an off-by-one error that caused an LBUG.
2623 Description: Transient SCSI error results in persistent IO issue
2624 Details : iobuf->dr_error is not reinitialized to 0 between two
2628 Frequency : sometimes when underlying device returns I/O errors
2630 Description: OSTs not going read-only during write failures
2631 Details : OSTs are not remounted read-only when the journal commit threads
2632 get I/O errors because fsfilt_ext3 calls journal_start/stop()
2633 instead of the ext3 wrappers.
2637 Description: poor connect scaling with increasing client count
2638 Details : Don't run filter_grant_sanity_check for more than 100 exports
2639 to improve scaling for large numbers of clients.
2642 Frequency : SLES10 only
2644 Description: sanity-quota.sh quotacheck failed: rc = -22
2645 Details : Quotas cannot be enabled on SLES10.
2648 Frequency : liblustre clients only
2650 Description: getdirentries does not give error when run on compute nodes
2651 Details : getdirentries does not fail when the size specified as an argument
2652 is too small to contain at least one entry
2654 Severity : enhancement
2656 Description: Add LNET router traceability for debug purposes
2657 Details : If a checksum failure occurs with a router as part of the
2658 IO path, the NID of the last router that forwarded the bulk data
2659 is printed so it can be identified.
2664 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
2665 Details : Due to a race condition, liblustre clients were occasionally
2666 evicted incorrectly.
2668 Severity : enhancement
2670 Description: lfs setstripe use optional parameters instead of postional
2673 Severity : enhancement
2675 Description: Nanosecond timestamp support for ldiskfs
2676 Details : The on-disk ldiskfs filesystem has added support for nanosecond
2677 resolution timestamps. There is not yet support for this at
2678 the Lustre filesystem level.
2681 Frequency : during server recovery
2683 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
2684 Details : request with rq_no_resend flag not awake l_wait_event if they get
2688 Frequency : nfs export on patchless client
2690 Description: connectathon hang when test nfs export over patchless client
2691 Details : Disconnected dentry cannot be found with lookup, so we do not need
2692 to unhash it or make it invalid
2695 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
2696 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
2697 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
2702 Description: open req refcounting wrong on reconnect
2703 Details : If reconnect happened between getting open reply from server and
2704 call to mdc_set_replay_data in ll_file_open, we will schedule
2705 replay for unreferenced request that we are about to free.
2706 Subsequent close will crash in variety of ways.
2707 Check that request is still eligible for replay in
2708 mdc_set_replay_data().
2713 Description: disable writes to filesystem when reading health_check file
2714 Details : the default for reading the health_check proc file has changed
2715 to NOT do a journal transaction and write to disk, because this
2716 can cause reads of the /proc file to hang and block HA state
2717 checking on a healthy but otherwise heavily loaded system. It
2718 is possible to return to the previous behaviour during configure
2719 with --enable-health-write.
2721 Severity : enhancement
2723 Description: 64-bit inode version
2724 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
2725 to the inode. This will be required for version-based recovery.
2730 Description: MDS fails to start if a duplicate client export is detected
2731 Details : in some rare cases it was possible for a client to connect to
2732 an MDS multiple times. Upon recovery the MDS would detect this
2733 and fail during startup. Handle this more gracefully.
2735 Severity : enhancement
2737 Description: Add -o localflock option to simulate old noflock
2739 Details : This will achieve local-only flock/fcntl locks
2745 Description: log_commit_thread vs filter_destroy race leads to crash
2746 Details : Take import reference before releasing llog record semaphore
2751 Description: Wrong request locking in request set processing
2752 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
2753 imp_delayed_list, in this place should be used imp_lock.
2756 Frequency : when reconnection
2758 Description: Grant Leak when osc reconnect to OST
2759 Details : When osc reconnect ost, OST(filter) should check whether it
2760 should grant more space to client by comparing fed_grant and
2761 cl_avail_grant, and return the granted space to client instead
2762 of "new granted" space, because client will call osc_init_grant
2763 to update the client grant space info.
2766 Frequency : when client reconnect to OST
2768 Description: Grant Leak when osc do resend and replay bulk write
2769 Details : When osc reconnect to OST, OST(filter)should clear grant info of
2770 bulk write request, because the grant info while be sync between
2771 OSC and OST when reconnect, and we should ignore the grant info
2772 these of resend/replay write req.
2777 Description: Grant space more than avaiable left space sometimes.
2778 Details : When then OST is about to be full, if two bulk writing from
2779 different clients came to OST. Accord the avaliable space of the
2780 OST, the first req should be permitted, and the second one
2781 should be denied by ENOSPC. But if the seconde arrived before
2782 the first one is commited. The OST might wrongly permit second
2783 writing, which will cause grant space > avaiable space.
2786 Frequency : when client is evicted
2788 Description: Grant might be wrongly erased when osc is evicted by OST
2789 Details : when the import is evicted by server, it will fork another
2790 thread ptlrpc_invalidate_import_thread to invalidate the
2791 import, where the grant will be set to 0. While the original
2792 thread will update the grant it got when connecting. So if
2793 the former happened latter, the grant will be wrongly errased
2794 because of this race.
2799 Description: Checking Stale with correct fid
2800 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
2801 to check whether it is stale, because sometimes, we want the
2802 enqueue happened anyway, and op_data.fid2 will not be initialized.
2804 Severity : enhancement
2806 Description: update patchless client
2807 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
2810 Frequency : only with 2.4 kernel
2812 Description: random memory corruption
2813 Details : size of struct ll_inode_info is to big for union inode.u and this
2814 can be cause of random memory corruption.
2819 Description: Memory leak in recovery
2820 Details : Lov_mds_md was not free in an error handler in mds_create_object.
2821 It should also check obd_fail before fsfilt_start, otherwise if
2822 fsfilt_start return -EROFS,(failover mds during mds recovery).
2823 then the req will return with repmsg->transno = 0 and rc = EROFS.
2824 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
2825 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
2826 be freed no matter whether fsfilt_commit success or not.
2829 Frequency : only with huge count clients
2831 Description: Prevents from taking the superblock lock in llap_from_page for
2833 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
2834 for llap_from_page call in ll_removepage prevents from taking the
2835 superblock lock for a soon died page.
2840 Description: Not check open intent error before release open handle
2841 Details : in some rare cases, the open intent error is not checked before
2842 release open handle, which may cause
2843 ASSERTION(open_req->rq_transno != 0), because it tries to release
2844 the failed open handle.
2849 Description: Set cat log bitmap only after create log success.
2850 Details : in some rare cases, the cat log bitmap is set too early. and it
2851 should be set only after create log success.
2855 Description: Accessing a block bevice can re-enable I/O when Lustre is
2856 tearing down a device.
2857 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
2862 Description: service threads may hog cpus when there are a lot of requests
2863 Details : Insert cond_resched to give other threads a chance to use some CPU
2868 Description: the cat log was not initialized in recovery
2869 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
2870 unlink log on mds will not be initialized until mds post
2871 recovery. And also in mds post recovery, the unlink log will
2872 initialization will be done asynchronausly, so there will be race
2873 between add unlink log and unlink log initialization.
2877 Description: brw_stats were being printed incorrectly
2878 Details : brw_stats were being printed as log2 but all of them were not
2879 recorded as log2. Also remove some code duplication arising from
2880 filter_tally_{read,write}.
2884 Frequency : rare, only in recovery.
2885 Description: ASSERTION(req->rq_type != LI_POISON) failed
2886 Details : imp_lock should be held while iterating over imp_sending_list for
2887 prevent destroy request after get timeout in ptlrpc_queue_wait.
2891 Description: replay-single.sh test 52 fails
2892 Details : A lock's skiplist need to be cleanup when it being unlinked
2893 from its resource list.
2897 Description: Short directio read returns full requested size rather than
2899 Details : Direct I/O operations should return actual amount of bytes
2900 transferred rather than requested size.
2902 Severity : enhancement
2904 Description: metadata RPC reduction (e.g. for rm performance)
2905 Details : decrease the amount of synchronous RPC between clients and servers
2906 by canceling conflicing lock before the operation on the client
2907 and packing thier handles into the main operation RPC to server.
2909 Severity : enhancement
2911 Description: Async OSC create to avoid the blocking unnecessarily.
2912 Details : If a OST has no remain object, system will block on the creating
2913 when need to create a new object on this OST. Now, ways use
2914 pre-created objects when available, instead of blocking on an
2915 empty osc while others are not empty. If we must block, we block
2916 for the shortest possible period of time.
2920 Description: improve handling recoverable errors
2921 Details : if request processig with error which can be recoverable on server
2922 request should be resend, otherwise page released from cache and
2925 Severity : enhancement
2927 Description: refine locking for avoid write wrong info into lov_objid file
2928 Details : fix possible races with add new target and write/update data in
2931 --------------------------------------------------------------------------------
2933 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
2938 Frequency : on some architectures
2940 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
2941 Details : Uninitialized flags sometimes cause configuration commands to
2945 Frequency : patchless clients only
2947 Description: missing __iget() symbol export
2948 Details : The __iget() symbol export is missing. To avoid the need for
2949 this on patchless clients the deathrow inode reaper is turned
2950 off, and we depend on the VM to clean up old inodes. This
2951 dependency was during via the fix for bug 12181.
2953 --------------------------------------------------------------------------------
2955 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
2957 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
2958 INTEROPERATE with older versions automatically. In many cases a
2959 special upgrade step is needed. Please read the
2960 user documentation before upgrading any part of a live system.
2961 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
2962 version will not interoperate with 1.6 betas before beta5 (1.5.95).
2963 * WARNING: Lustre configuration and startup changes are required with
2964 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
2968 Severity : enhancement
2970 Description: MountConf
2971 Details : Lustre configuration is now managed via mkfs and mount
2972 commands instead of lmc and lconf. New obd types (MGS, MGC)
2973 are added for dynamic configuration management. See
2974 https://mail.clusterfs.com/wikis/lustre/MountConf for
2977 Severity : enhancement
2979 Description: dynamic OST addition
2980 Details : OSTs can now be added to a live filesystem
2982 Severity : enhancement
2984 Description: startup order invariance
2985 Details : MDTs and OSTs can be started in any order. Clients only
2986 require the MDT to complete startup.
2988 Severity : enhancement
2990 Description: parallel, asynchronous orphan cleanup
2991 Details : orphan cleanup is now performed in separate threads for each
2992 OST, allowing parallel non-blocking operation.
2994 Severity : enhancement
2996 Description: optimized stripe assignment
2997 Details : stripe assignments are now made based on ost space available,
2998 ost previous usage, and OSS previous usage, in order to try
2999 to optimize storage space and networking resources.
3001 Severity : enhancement
3003 Description: Permanently set tunables
3004 Details : All writable /proc/fs/lustre tunables can now be permanently
3005 set on a per-server basis, at mkfs time or on a live system.
3007 Severity : enhancement
3009 Description: Lustre message v2
3010 Details : Add lustre message format v2.
3012 Severity : enhancement
3014 Description: client OST exclusion list
3015 Details : Clients can be started with a list of OSTs that should be
3016 declared "inactive" for known non-responsive OSTs.
3020 Description: SPEC SFS validation failure on NFS v2 over lustre.
3021 Details : Changes the blocksize for regular files to be 2x RPC size,
3022 and not depend on stripe size.
3024 Severity : enhancement
3026 Description: Multiple MD RPCs in flight.
3027 Details : Further unserialise some read-only MDS RPCs - learn about intents.
3028 To avoid overly-overloading MDS, introduce a limit on number of
3029 MDS RPCs in flight for a single client and add /proc controls
3030 to adjust this limit.
3032 Severity : enhancement
3034 Description: client read/write statistics
3035 Details : Add client read/write call usage stats for performance
3036 analysis of user processes.
3037 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
3038 file access. extents_stats shows chunk size distribution.
3039 extents_stats_per_process show chunk size distribution per
3042 Severity : enhancement
3044 Description: mds statistics
3045 Details : Add detailed mds operations statistics in
3046 /proc/fs/lustre/mds/*/stats.
3050 Description: Failure of copying files with lustre special EAs.
3051 Details : Client side always return success for setxattr call for lustre
3052 special xattr (currently only "trusted.lov").
3057 Description: Refcount LNET uuids
3058 Details : The global LNET uuid list grew linearly with every startup;
3059 refcount repeated list entries instead of always adding to
3062 Severity : enhancement
3064 Description: Dynamic service threads
3065 Details : Within a small range, start extra service threads
3066 automatically when the request queue builds up.
3069 Frequency : mixed-endian client/server environments
3071 Description: mixed-endian crashes
3072 Details : The new msg_v2 system had some failures in mixed-endian
3075 Severity : enhancement
3077 Description: Easy OST removal
3078 Details : OSTs can be permanently deactivated with e.g. 'lctl
3079 conf_param lustre-OST0001.osc.active=0'
3081 Severity : enhancement
3083 Description: MGS proc entries
3084 Details : Added basic proc entries for the MGS showing what filesystems
3087 Severity : enhancement
3089 Description: provide MGS failover
3090 Details : Added config lock reacquisition after MGS server failover.
3092 Severity : enhancement
3094 Description: add Linux 2.4 support
3095 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
3099 Description: a large application tries to do I/O to the same resource and dies
3100 in the middle of it.
3101 Details : Check the req->rq_arrival time after the call to
3102 ost_brw_lock_get(), but before we do anything about
3103 processing it & sending the BULK transfer request. This
3104 should help move old stale pending locks off the queue as
3105 quickly as obd_timeout.
3108 Frequency : when an incorrect nid is specified during startup
3110 Description: ptlrpc connect to non-existant node causes kernel crash
3111 Details : LNET can't be re-entered from an event callback, which
3112 happened when we expire a message after the export has been
3113 cleaned up. Instead, hand the zombie cleanup off to another
3116 Severity : enhancement
3118 Description: plain/inodebits lock performance improvement
3119 Details : Grouping plain/inodebits in granted list by their request modes
3120 and bits policy, thus improving the performance of search through
3124 Frequency : only if OST filesystem is corrupted
3126 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3127 Details : for a short time RPCs with bulk IO are in the replay list,
3128 but replay of bulk IOs is unimplemented. If the OST filesystem
3129 is corrupted due to disk cache incoherency and then replay is
3130 started it is possible to trip an assertion. Avoid putting
3131 committed RPCs into the replay list at all to avoid this issue.
3134 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3136 Description: System hang on startup
3137 Details : This bug allowed the liblustre (e.g. catamount) client to
3138 return to the app before handling all startup RPCs. This
3139 could leave the node unresponsive to lustre network traffic
3140 and manifested as a server ptllnd timeout.
3142 Severity : enhancement
3144 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3145 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
3146 boolean to control whether to print peer debug info when a
3147 client's RPC times out.
3150 Frequency : only for kernels with patches from Lustre below 1.4.3
3152 Description: Remove old rdonly API
3153 Details : Remove old rdonly API which unused from at least lustre 1.4.3
3156 Frequency : only for devices with external journals
3158 Description: Set external device read-only also
3159 Details : During a commanded failover stop, we set the disk device
3160 read-only while the server shuts down. We now also set any
3161 external journal device read-only at the same time.
3164 Frequency : when upgrading from 1.4 while trying to change parameters
3166 Description: The wrong (new) MDC name was used when setting parameters for
3167 upgraded MDT's. Also allows changing of OSC (and MDC)
3168 parameters if --writeconf is specified at tunefs upgrade time.
3171 Frequency : when setting specific ost indicies
3173 Description: QOS code breaks on skipped indicies
3174 Details : Add checks for missing OST indicies in the QOS code, so OSTs
3175 created with --index need not be sequential.
3177 Severity : enhancement
3179 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3180 Details : The uninit_groups feature works in conjunction with the kernel
3181 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3182 up the pass1 processing of e2fsck. This is a read-only feature
3183 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3184 mount filesystems that have had this feature enabled.
3186 Severity : enhancement
3188 Description: Improve multi-block allocation algorithm to avoid fragmentation
3189 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3190 allocation locality and avoid filesystem fragmentation.
3192 ------------------------------------------------------------------------------
3194 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
3196 * Support for kernels:
3197 2.4.21-47.0.1.EL (RHEL 3)
3198 2.6.5-7.283 (SLES 9)
3199 2.6.9-42.0.10.EL (RHEL 4)
3200 2.6.12.6 vanilla (kernel.org)
3201 2.6.16.27-0.9 (SLES 10)
3202 * Recommended e2fsprogs version: 1.39.cfs5
3204 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3208 Frequency : occasional, depends on client load and configuration
3209 Bugzilla : 12181, 12203
3210 Description: data loss for recently-modified files
3212 Details : In some cases it is possible that recently written or created
3213 files may not be written to disk in a timely manner (this should
3214 normally be within 30s unless client IO load is very high).
3215 The problem appears as zero-length files or files that are a
3216 multiple of 1MB in size after a client crash or client eviction
3217 that are missing data at the end of the file.
3219 This problem is more likely to be hit on clients where files are
3220 repeatedly created and unlinked in the same directory, clients
3221 have a large amount of RAM, have many CPUs, the filesystem has
3222 many OSTs, the clients are rebooted frequently, and/or the files
3223 are not accessed by other nodes after being written.
3225 The presence of the problem can be detected by looking at
3226 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
3227 smaller than the second (nr_unused) then dirty files will not
3228 be flushed automatically to disk. "sync; sleep 10" should be
3229 run several times on the node before unmounting it to update
3230 Lustre (this is also safe to run on nodes without this problem).
3232 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
3233 that can cause this same problem, so customers using that kernel
3234 also need to update the kernel in addition to Lustre. In order
3235 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
3237 It is normal that files written just before a client crash (less
3238 than 30s) may not yet have been flushed to disk, even for local
3242 Frequency : frequent on thin XT3 nodes
3244 Description: UUID collision on thin XT3 Linux nodes
3245 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
3246 randomly, since we relied on an insufficiently-seeded PRNG.
3251 Description: OSS hangs after "All ost request buffers busy"
3252 Details : A deadlock between quota and journal operations caused OSS
3253 hangs after printing "All ost request buffers busy."
3256 Frequency : always on liblustre builds
3258 Description: Cleanup compiler warnings on liblustre
3261 Frequency : always on liblustre builds on XT3
3263 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
3267 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
3268 > 32000 subdirectories
3269 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
3270 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
3271 subdirectory count crosses 32000. This will aid e2fsck to
3272 correctly handle more than 32000 subdirectories.
3275 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3277 Description: System hang on startup
3278 Details : This bug allowed the liblustre (e.g. catamount) client to
3279 return to the app before handling all startup RPCs. This
3280 could leave the node unresponsive to lustre network traffic
3281 and manifested as a server ptllnd timeout.
3283 Severity : enhancement
3285 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3286 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
3287 boolean to control whether to print peer debug info when a
3288 client's RPC times out.
3293 Description: make O_SYNC working on 2.6 kernels
3294 Details : 2.6 kernels use different method for mark pages for write,
3295 so need add a code to lustre for O_SYNC work.
3300 Description: Failure to close file and release space on NFS
3301 Details : Put inode details into lock acquired in ll_intent_file_open.
3302 Use mdc_intent_lock in ll_intent_open to properly
3303 detect all kind of errors unhandled by mdc_enqueue.
3308 Description: proc file read during shutdown sometimes raced obd removal,
3310 Details : Add lock to prevent obd access after proc file removal.
3313 Frequency : Only for files larger than 4GB on 32-bit clients.
3315 Description: improperly doing page alignment of locks
3316 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
3317 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3322 Description: RPCs being resent when they shouldn't be
3323 Details : Some RPCs that should not be resent are being resent. This
3324 can cause inconsistencies in the RPC state machine. Do not
3325 resend such requests.
3328 Frequency : rare, only with NFS export
3330 Description: Crash on NFS re-export node
3331 Details : under very unusual load conditions an assertion is hit in
3332 ll_intent_file_open()
3335 Frequency : only if OST filesystem is corrupted
3337 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3338 Details : for a short time RPCs with bulk IO are in the replay list,
3339 but replay of bulk IOs is unimplemented. If the OST filesystem
3340 is corrupted due to disk cache incoherency and then replay is
3341 started it is possible to trip an assertion. Avoid putting
3342 committed RPCs into the replay list at all to avoid this issue.
3347 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3348 Details : Large single O_DIRECT read and write calls can fail to allocate
3349 a sufficiently large buffer to process the request. In case of
3350 allocation failure the allocation is retried with a smaller
3351 buffer and broken into smaller requests.
3353 Severity : enhancement
3355 Description: Add -o localflock option to simulate old noflock behaviour.
3356 Details : This will achieve local-only flock/fcntl locks coherentness.
3361 Description: versioning check is incomplete
3362 Details : Checking the version difference of client vs. server, report
3363 error if the gap is too big.
3368 Description: add support PG_writeback bit
3369 Details : add support for PG_writeback bit for Lustre, for more carefull
3370 work with page cache in 2.6 kernel. This also fix some deadlocks
3371 and remove hack for work O_SYNC with 2.6 kernel.
3373 Severity : enhancement
3375 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3376 Details : The uninit_groups feature works in conjunction with the kernel
3377 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3378 up the pass1 processing of e2fsck. This is a read-only feature
3379 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3380 mount filesystems that have had this feature enabled.
3382 Severity : enhancement
3384 Description: Improve multi-block allocation algorithm to avoid fragmentation
3385 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3386 allocation locality and avoid filesystem fragmentation.
3388 ------------------------------------------------------------------------------
3390 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3392 * Support for kernels:
3393 2.6.9-42.0.3.EL (RHEL 4)
3394 2.6.5-7.276 (SLES 9)
3395 2.4.21-47.0.1.EL (RHEL 3)
3396 2.6.12.6 vanilla (kernel.org)
3397 2.6.16.21-0.8 (SLES10)
3398 * Recommended e2fsprogs version: 1.39.cfs2-0
3400 * The backwards-compatible /proc/sys/portals symlink has been removed
3401 in this release. Before upgrading, please ensure that you change
3402 any configuration scripts or /etc/sysctl.conf files that access
3403 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3404 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3405 in advance of the upgrade on any system running Lustre 1.4.6 or
3406 newer, since /proc/sys/lnet was added in that version.
3407 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3411 Frequency : only when quota is used
3413 Description: avoid scanning export list for quota master
3414 Details : Change the algorithms to avoid scanning export list in order
3415 to improve the efficiency.
3418 Frequency : MDS failover only, very rarely
3420 Description: "went back in time" messages on mds failover
3421 Details : The greatest transno may be lost when the current operation
3422 finishes with an error (transno==0) and the client's last_rcvd
3423 record is over-written. Save the greatest transno in the
3424 mds_last_transno for this case.
3427 Frequency : always for specific kernels and striping counts
3429 Description: client may get "Matching packet too big" without ACL support
3430 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
3431 when trying to access files in certain configurations. The
3432 clients should in fact be denied when mounting because they do
3433 not understand ACLs.
3436 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
3438 Description: many clients connecting with IO in progress causes connect timeouts
3439 Details : Avoid synchronous journal commits to avoid delays caused by many
3440 clients connecting/disconnecting when bulk IO is in progress.
3441 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
3442 OST_IO_PORTAL to avoid delays behind potentially many pending
3446 Frequency : occasionally with multiple writers to a single file
3448 Description: shared writes to file may result in wrong size reported by stat()
3449 Details : Allow growing of kms when extent lock is cancelled
3452 Frequency : always with random mmap IO to multi-striped file
3454 Description: mmap write might be lost if we are writing to a 'hole' in stripe
3455 Details : Only if the hole is at the end of OST object so that kms is too
3456 small. Fix is to increase kms accordingly in ll_nopage.
3459 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
3461 Description: writes to a missing object would leak memory on the OST
3462 Details : If there is an inconsistency between the MDS and OST filesystems,
3463 such that the MDS references an object that doesn't exist, writes
3464 to that object will leak memory due to incorrect cleanup in the
3465 error handling path, eventually running out of memory on the OST.
3470 Description: Creating too long symlink causes lustre errors
3471 Details : Check symlink and name lengths before sending requests to MDS.
3474 Frequency : only if flock is enabled (not on by default)
3476 Description: posix locks not released on fd closure on 2.6.9+
3477 Details : We failed to add posix locks to list of inode locks on 2.6.9+
3478 kernels, this caused such locks not to be released on fd close and
3479 then assertions on fs unmount about still used locks.
3482 Frequency : MDS failover only, very rarely
3484 Description: clients may get ASSERTION(granted_lock != NULL)
3485 Details : When request was taking a long time, and a client was resending
3486 a getattr by name lock request. The were multiple lock requests
3487 with the same client lock handle and
3488 mds_getattr_name->fixup_handle_for_resent_request found one of the
3489 lock handles but later failed with ASSERTION(granted_lock != NULL).
3494 Description: handle->h_buffer_credits > 0, assertion failure
3495 Details : h_buffer_credits is zero after truncate, causing assertion
3496 failure. This patch extends the transaction or creates a new
3500 Frequency : NFS re-export or patchless client
3501 Bugzilla : 11179, 10796
3502 Description: Crash on NFS re-export node (__d_move)
3503 Details : We do not want to hash the dentry if we don't have a lock.
3504 But if this dentry is later used in d_move, we'd hit uninitialised
3505 list head d_hash, so we just do this to init d_hash field but
3506 leave dentry unhashed.
3509 Frequency : NFS re-export or patchless client
3511 Description: NFS exports has problem with symbolic link
3512 Details : lustre client didn't properly install dentry when re-exported
3513 to NFS or running patchless client.
3516 Frequency : NFS re-export or patchless client
3518 Description: Various nfs/patchless fixes.
3519 Details : fixes reuse disconected alias for lookup process - this fixes
3520 warning "find_exported_dentry: npd != pd",
3521 fix permission error with open files at nfs.
3522 fix apply umask when do revalidate.
3525 Frequency : occasional
3527 Description: Crash on NFS re-export node
3528 Details : calling clear_page() on the wrong pointer triggered oops in
3529 generic_mapping_read().
3532 Frequency : rarely, using O_DIRECT IO
3534 Description: unaligned directio crashes client with LASSERT
3535 Details : check for unaligned buffers before trying any requests.
3538 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
3540 Description: stale data returned from RAID cache
3541 Details : If only a small amount of IO is done to the RAID device before
3542 reading it again it is possible to get stale data from the RAID
3543 cache instead of reading it from disk.
3546 Frequency : always for sles10 kernel
3548 Description: sles10 support
3549 Details : ll_follow_link: compile fixes and using of nd_set_link
3550 under newer kernels.
3553 Frequency : depends on arch, kernel and compiler version, always on sles10
3556 Description: recursive or deep enough symlinks cause stack overflow
3557 Details : getting rid of large stack-allocated variable in
3561 Frequency : depends on hardware
3563 Description: lustre write performance loss in the SLES10 kernel
3564 Details : the performance loss is caused by using of write barriers in the
3565 ext3 code. The SLES10 kernel turns barrier support on by
3566 default. The fix is to undo that change for ldiskfs.
3568 ------------------------------------------------------------------------------
3570 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
3572 * Support for kernels:
3573 2.6.9-42.0.3EL (RHEL 4)
3574 2.6.5-7.276 (SLES 9)
3575 2.4.21-47.0.1.EL (RHEL 3)
3576 2.6.12.6 vanilla (kernel.org)
3580 Frequency : quota enabled and large files being deleted
3582 Description: releasing more than 4GB of quota at once hangs OST
3583 Details : If a user deletes more than 4GB of files on a single OST it
3584 will cause the OST to spin in an infinite loop. Release
3585 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3590 Description: statfs data retrieved from /proc may be stale or zero
3591 Details : When reading per-device statfs data from /proc, in the
3592 {kbytes,files}_{total,free,avail} files, it may appear
3593 as zero or be out of date.
3596 Frequency : systems with MD RAID1 external journal devices
3598 Description: lconf's call to blkid is confused by RAID1 journal devices
3599 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3600 of returning all block devices that match the journal UUID.
3603 Frequency : always, for aggregate stripe size over 4GB
3605 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
3606 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
3607 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
3610 Frequency : always if "lfs find" used on a local file/directory
3612 Description: "lfs find" segfaults if used on a local file/directory
3613 Details : The case where a directory component was not specified wasn't
3614 handled correctly. Handle this properly.
3617 Frequency : always on ppc64
3619 Description: the write to an ext3 filesystem mounted with mballoc got stuck
3620 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3621 perform endianness conversion.
3624 Frequency : rarely (truncate to non-zero file size after write under load)
3625 Bugzilla : 10730, 10687
3626 Description: Files padded with zeros to next 4K multiple
3627 Details : With filesystems mounted using the "extents" option (2.6 kernels)
3628 it is possible that files that are truncated to a non-zero size
3629 immediately after being written are filled with zero bytes beyond
3630 the truncated size. No file data is lost.
3632 Severity : enhancement
3634 Description: Allow recovery/failover for liblustre clients.
3635 Details : liblustre clients were unaware of failover configurations until
3638 Severity : enhancement
3640 Description: user file locks should fail when not mounting with flock option
3641 Details : Set up an error-returning stub in ll_file_operations.lock field
3642 to prevent incorrect behaviour when client is mounted without
3643 flock option. Also, set up properly f_op->flock field for
3647 Frequency : always on ia64
3649 Description: "lfs df" loops on printing out MDS statfs information
3650 Details : The obd_ioctl_data was not initialized and in some systems
3651 this caused a failure during the ioctl that did not return
3652 an error. Initialize the struct and return an error on failure.
3655 Frequency : SLES 9 only
3657 Description: Error of copying files with lustre special EAs as root
3658 Details : Client side always return success for setxattr call for lustre
3659 special xattr (currently only "trusted.lov").
3662 Frequency : rarely on clusters with both ia64+i386 clients
3664 Description: ia64+i686 clients doing shared IO on the same file may LBUG
3665 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
3666 clients are doing concurrent writes to the same file it is
3667 possible that the ia64 clients may LASSERT because the OST
3668 extent locks are not PAGE_SIZE aligned. Ensure that grown
3669 locks are always aligned on the request boundary.
3672 Frequency : specific use, occasional
3674 Description: Overwriting in use executable truncates on-disk binary image
3675 Details : If one node attempts to overwrite an executable in use by
3676 another node, we now correctly return ETXTBSY instead of
3677 truncating the file.
3679 Severity : enhancement
3681 Description: Async OSC create to avoid the blocking unnecessarily.
3682 Details : If a OST has no remain object, system will block on the creating
3683 when need to create a new object on this OST. Now, ways use
3684 pre-created objects when available, instead of blocking on an
3685 empty osc while others are not empty. If we must block, we block
3686 for the shortest possible period of time.
3691 Description: chmod on Lustre root is propagated to other clients
3692 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
3693 invalid by the follow_mount time.
3696 Frequency : liblustre clients only
3698 Description: Race in 'instant cancel' lock handling could lead to such locks
3699 never to be granted in case of SMP MDS
3700 Details : Do not destroy not yet granted but cbpending locks in
3704 Frequency : replay/resend of open
3706 Description: non null lock assetion failure in mds_intent_policy
3707 Details : Trying to replay/resend lockless open requests resulted in
3708 mds_open() returning 0 with no lock. Now it sets a flag if
3709 a lock is going to be returned.
3711 Severity : enhancement
3713 Description: Checksum enhancements
3714 Details : New checksum enhancements allow for resending RPCs that failed
3717 Severity : enhancement
3719 Description: Tunables on number of dirty pages in cacche
3720 Details : Allow to set limit on number of dirty pages cached.
3725 Description: client crash on unmount - lock still has references
3726 Details : In some error handling cases it was possible to leak a lock
3727 reference on a client while accessing a file. This was not
3728 harmful to the client during operation, but would cause the
3729 client to crash when the filesystem is unmounted.
3732 Frequency : specific case, rare
3734 Description: ETXTBSY on mds though file not in use by client
3735 Details : ETXTBSY is no longer incorrectly returned when attempting to
3736 chmod or chown a directory that the user previously tried to
3737 execute or a currently-executing binary.
3740 Frequency : extremely rare except on liblustre-based clients
3742 Description: Lustre space not freed when files are deleted
3743 Details : Clean up open-unlinked files after client eviction. Previously
3744 the unlink was skipped and the files remained as orphans.
3749 Description: OST failure "would be an LBUG" in waiting_locks_callback()
3750 Details : In some cases it was possible to send a blocking callback to
3751 a client doing a glimpse, even though that client didn't get
3752 a lock granted. When the glimpse lock is cancelled on the OST
3753 the freed lock is left on the waiting list and corrupted the list.
3756 Frequency : all core dumps
3758 Description: Broke core dumps to lustre
3759 Details : Negative dentry may be unhashed if parent does not have UPDATE
3760 lock, but some callers, e.g. do_coredump, expect dentry to be
3761 hashed after successful create, hash it in ll_create_it.
3763 ------------------------------------------------------------------------------
3765 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
3767 * Support for kernels:
3768 2.6.9-42.0.2.EL (RHEL 4)
3769 2.6.5-7.276 (SLES 9)
3770 2.4.21-40.EL (RHEL 3)
3771 2.6.12.6 vanilla (kernel.org)
3775 Frequency : always on RHEL 3
3777 Description: Number of open files grows over time
3778 Details : The number of open files grows over time, whether or not
3779 Lustre is started. This was due to a filp leak introduced
3780 by one of our kernel patches.
3782 ------------------------------------------------------------------------------
3784 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
3786 * Support for kernels:
3787 2.6.9-42.EL (RHEL 4)
3788 2.6.5-7.276 (SLES 9)
3789 2.4.21-40.EL (RHEL 3)
3790 2.6.12.6 vanilla (kernel.org)
3795 Bugzilla : 5719, 9635, 9792, 9684
3796 Description: OST (or MDS) trips assertions in (re)connection under heavy load
3797 Details : If a server is under heavy load and cannot reply to new
3798 connection requests before the client resends the (re)connect,
3799 the connection handling code can behave badly if two service
3800 threads are concurrently handing separate (re)connections from
3801 the same client. Add better locking to the connection handling
3802 code, and ensure that only a single connection will be processed
3803 for a given client UUID, even if the lock is dropped.
3805 Severity : enhancement
3807 Description: add TCP zero-copy support to kernel
3808 Details : Add support to the kernel TCP stack to allow zero-copy bulk
3809 sends if the hardware supports scatter-gather and checksumming.
3810 This allows socklnd to do client-write and server-read more
3811 efficiently and reduce CPU utilization from skbuf copying.
3814 Frequency : only if NFS exporting from client
3816 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
3817 Details : If setting attributes on a file created under NFS that had
3818 never been opened it would be possible to oops the client
3819 if the file had no objects.
3822 Frequency : always for liblustre
3824 Description: liblustre client does MDS+OSTs setattr RPC for each write
3825 Details : When doing a write from a liblustre client, the client
3826 incorrectly issued an RPC to the MDS and each OST the file was
3827 striped over in order to update the timestamps. When writing
3828 with small chunks and many clients this could overwhelm the MDS
3829 with RPCs. In all cases it would slow down the write because
3830 these RPCs are unnecessary.
3832 Severity : enhancement
3834 Description: allow number of MDS service threads to be changed at module load
3835 Details : It is now possible to change the number of MDS service threads
3836 running. Adding "options mds mds_num_threads={N}" to the MDS's
3837 /etc/modprobe.conf will set the number of threads for the next
3838 time Lustre is restarted (assuming the "mds" module is also
3839 reloaded at that time). The default number of threads will
3840 stay the same, 32 for most systems.
3845 Description: OST crash if filesystem is unformatted or corrupt
3846 Details : If an OST is started on a device that has never been formatted
3847 or if the filesystem is corrupt and cannot even mount then the
3848 error handling cleanup routines would dereference a NULL pointer.
3853 Description: NULL pointer deref in llap_from_page.
3854 Details : get_cache_page_nowait can return a page with NULL (or otherwise
3855 incorrect) mapping if the page was truncated/reclaimed while it was
3856 searched for. Check for this condition and skip such pages when
3857 doing readahead. Introduce extra check to llap_from_page() to
3858 verify page->mapping->host is non-NULL (so page is not anonymous).
3861 Frequency : Sometimes when using sys_sendfile
3863 Description: "page not covered by a lock" warnings from ll_readpage
3864 Details : sendfile called ll_readpage without right page locks present.
3865 Now we introduced ll_file_sendfile that does necessary locking
3866 around call to generic_file_sendfile() much like we do in
3870 Frequency : with certain MDS communication failures at client mount time
3872 Description: NULL pointer deref after failed client mount
3873 Details : a client connection request may delayed by the network layer
3874 and not be sent until after the PTLRPC layer has timed out the
3875 request. If the client fails the mount immediately it will try
3876 to clean up before the network times out the request. Add a
3877 reference from the request import to the obd device and delay
3878 the cleanup until the network drops the request.
3881 Frequency : occasionally during client (re)connect
3883 Description: assertion failure during client (re)connect
3884 Details : processing a client connection request may be delayed by the
3885 client or server longer than the client connect timeout. This
3886 causes the client to resend the connection request. If the
3887 original connection request is replied in this interval, the
3888 client may trip an assertion failure in ptlrpc_connect_interpret()
3889 which thought it would be the only running connect process.
3892 Frequency : only with obd_echo servers and clients that are rebooted
3894 Description: kernel BUG accessing uninitialized data structure
3895 Details : When running an obd_echo server it did not start the ping_evictor
3896 thread, and when a client was evicted an uninitialized data
3897 structure was accessed. Start the ping_evictor in the RPC
3898 service startup instead of the OBD startup.
3900 Severity : enhancement
3901 Bugzilla : 10193 (patchless)
3902 Description: Remove dependency on various unexported kernel interfaces.
3903 Details : No longer need reparent_to_init, exit_mm, exit_files,
3904 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
3907 Frequency : rare (only users of deprecated and unsupported LDAP config)
3909 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
3910 Details : LDAP apparently contains 'lustreName' attributes instead of
3911 'name'. A simple remapping of the name is sufficient.
3914 Frequency : rare (only with non-default dump_on_timeout debug enabled)
3916 Description: waiting_locks_callback trips kernel BUG if client is evicted
3917 Details : Running with the dump_on_timeout debug flag turned on makes
3918 it possible that the waiting_locks_callback() can try to dump
3919 the Lustre kernel debug logs from an interrupt handler. Defer
3920 this log dumping to the expired_lock_main() thread.
3922 Severity : enhancement
3924 Description: Support NFS exporting on 2.6 kernels.
3925 Details : Implement non-rawops metadata methods for NFS server to use without
3926 changing NFS server code.
3929 Frequency : very rare (synthetic metadata workload only)
3931 Description: two racing renames might cause an MDS thread to deadlock
3932 Details : Running the "racer" program may cause one MDS thread to rename
3933 a file from being the source of a rename to being the target of
3934 a rename at exactly the same time that another thread is doing
3935 so, and the second thread has already enqueued these locks after
3936 doing a lookup of the target and is trying to relock them in
3937 order. Ensure that we don't try to re-lock the same resource.
3940 Frequency : only very large systems with liblustre clients
3942 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
3943 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
3944 all OSTs in parallel. This allows the eviction of stale liblustre
3945 clients to proceed much faster than if they were done in series,
3946 and also offers similar improvements for other set_info RPCs.
3951 Description: excessive CPU usage during initial read phase on client
3952 Details : During the initial read phase on a client, it would agressively
3953 retry readahead on the file, consuming too much CPU and impacting
3954 performance (since 1.4.5.8). Improve the readahead algorithm
3955 to avoid this, and also improve some other common cases (read
3956 of small files in particular, where "small" is files smaller than
3957 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
3962 Description: MDS crash when receiving packet with unknown intent.
3963 Details : Do not LBUG in unknown intent case, just return -EFAULT
3965 Severity : enhancement
3966 Bugzilla : 9293, 9385
3967 Description: MDS RPCs are serialised on client. This is unnecessary for some.
3968 Details : Do not serialize getattr (non-intent version) and statfs.
3971 Frequency : occasional, when OST network is overloaded/intermittent
3973 Description: client evicted by OST after bulk IO timeout
3974 Details : If a client sends a bulk IO request (read or write) the OST
3975 may evict the client if it is unresposive to its data GET/PUT
3976 request. This is incorrect if the network is overloaded (takes
3977 too long to transfer the RPC data) or dropped the OST GET/PUT
3978 request. There is no need to evict the client at all, since
3979 the pinger and/or lock callbacks will handle this, and the
3980 client can restart the bulk request.
3983 Frequency : Always when mmapping file with no objects
3985 Description: client crashes when mmapping file with no objects
3986 Details : Check that we actually have objects in a file before doing any
3987 operations on objects in ll_vm_open, ll_vm_close and
3993 Description: Request leak when working with deleted CWD
3994 Details : Introduce advanced request refcount tracking for requests
3995 referenced from lustre intent.
3997 Severity : Enhancement
3999 Description: Cache open file handles on client.
4000 Details : MDS now will return special lock along with openhandle, if
4001 requested and client is allowed to hold openhandle, even if unused,
4002 until such a lock is revoked. Helps NFS a lot, since NFS is opening
4003 closing files for every read/write openration.
4005 Severity : Enhancement
4007 Description: Cache open negative dentries on client when possible.
4008 Details : Guard negative dentries with UPDATE lock on parent dir, drop
4009 negative dentries on lock revocation.
4014 Description: Remounting a client read-only wasn't possible with a zconf mount
4015 Details : It wasn't possible to remount a client read-only with llmount.
4017 Severity : enhancement
4018 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
4019 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
4020 MPICH is included to add Lustre-specific ADIO interfaces.
4021 This is based closely on the UFS ADIO layer and only differs
4022 in file creation, in order to allow the OST striping to be set.
4023 This is user-contributed code and not supported by CFS.
4028 Description: extended inode attributes (immutable, append-only) work improperly
4029 when 2.4 and 2.6 kernels are used on client/server or vice versa
4030 Details : Introduce kernel-independent values for these flags.
4032 Severity : enhancement
4035 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
4036 Details : Many of the /proc/ tunables can only be tuned at a megabyte
4037 granularity. Now, Fractional MB granularity is be supported,
4038 this is very useful for low memory system.
4040 Severity : enhancement
4042 Description: Getattr by fid
4043 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
4044 locks, avoids extra getattr rpc requests to MDS, allows '/' to
4045 have locks and avoids getattr rpc requests for it on every stat.
4048 Frequency : Always, for filesystems larger than 2TB
4050 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4051 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
4052 filesystems larger than 2TB to be mounted with mballoc enabled.
4055 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
4057 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4058 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
4059 filesystems larger than 2TB is mounted then the kernel will
4060 silently corrupt the start of the filesystem. CONFIG_LBD is
4061 enabled for all CFS-supported kernels, but the possibility of
4062 this happening with a modified kernel config exists.
4064 Severity : enhancement
4066 Description: add client O_DIRECT support for 2.6 kernels
4067 Details : It is now possible to do O_DIRECT reads and writes to files
4068 in the Lustre client mountpoint on 2.6 kernel clients.
4070 Severity : enhancement
4072 Description: parallel glimpse, setattr, statfs, punch, destroy requests
4073 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
4074 parallel, not waiting for response from every OST before sending
4075 a rpc to the next OST.
4080 Description: setattr vs write race when updating file timestamps
4081 Details : Client processes that update a file timestamp into the past
4082 right after writing to the file (e.g. tar) it is possible that
4083 the updated file modification time can be reset to the current
4084 time due to a race between processing the setattr and write RPC.
4086 Severity : enhancement
4088 Description: Bring 'lfs find' closer in line with regular Linux find.
4089 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
4090 -print0 options and obtains all the needed info through the lustre
4093 Severity : enhancement
4095 Description: support up to 1024 configured devices on one node
4096 Details : change obd_dev array from statically allocated to dynamically
4097 allocated structs as they are first used to reduce memory usage
4102 Description: Flush dirty partially truncated pages during truncate
4103 Details : Immediatelly flush partially truncated pages in filter_setattr,
4104 this way we completely avoid having any pages in page cache on OST
4105 and can retire ugly workarounds during writes to flush such pages.
4110 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
4111 Details : protect inode from truncation within vfs_unlink() context
4112 just take a reference before calling vfs_unlink() and release it
4113 when parent's i_sem is free.
4118 Description: last_id value checked outside lock on OST caused LASSERT failure
4119 Details : If there were multiple MDS->OST object precreate requests in
4120 flight, it was possible that the OST's last object id was checked
4121 outside a lock and incorrectly tripped an assertion. Move checks
4122 inside locks, and discard old precreate requests.
4125 Frequency : always, if extents are used on OSTs
4127 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
4128 Details : OSTs using the extents format would not zero the high 16 bits of
4129 the index physical block number. This is not a problem for any
4130 OST filesystems smaller than 16TB, and no kernels support ext3
4131 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
4132 new/modified files) and can be fixed for existing filesystems
4133 with e2fsprogs-1.39-cfs1.
4138 Description: import connection selection may be incorrect if timer wraps
4139 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
4140 import connections to be ignored if the 32-bit jiffies counter
4141 wraps. Use a 64-bit jiffies counter.
4144 Frequency : very large clusters immediately after boot
4146 Description: LNET request buffers exhausted under heavy short-term load
4147 Details : If a large number of client requests are generated on a service
4148 that has previously never seen so many requests it is possible
4149 that the request buffer growth cannot keep up with the spike in
4150 demand. Instead of dropping incoming requests, they are held in
4151 the LND until the RPC service can accept more requests.
4154 Frequency : Sometimes during replay
4156 Description: Assertion failure in ll_local_open after replay.
4157 Details : If replay happened on an open request reply before we were able
4158 to set replay handler, reply will become not swabbed tripping the
4159 assertion in ll_local_open. Now we set the handler right after
4160 recognising of open request
4163 Frequency : very rare
4165 Description: kernel reports "badness in vsnprintf"
4166 Details : Reading from the "recovery_status" /proc file in small chunks
4167 may cause a negative length in lprocfs_obd_rd_recovery_status()
4168 call to vsnprintf() (which is otherwise harmless). Exit early
4169 if there is no more space in the output buffer.
4171 Severity : enhancement
4173 Description: clear OBD RPC statistics by writing to them
4174 Details : It is now possible to clear the OBD RPC statistics by writing
4175 to the "stats" file.
4180 Description: Client mtime is not the same on different clients after utimes
4181 Details : In some cases, the client was using the utimes() syscall on
4182 a file cached on another node. The clients now validate the
4183 ctime from the MDS + OSTs to determine which one is right.
4188 Description: Inability to activate failout mode
4189 Details : lconf script incorrectly assumed that in pythong string's numeric
4190 value is used in comparisons.
4193 Frequency : always with multiple stripes per file
4195 Description: Inefficient object allocation for mutli-stripe files
4196 Details : When selecting which OSTs to stripe files over, for files with
4197 a stripe count that divides evenly into the number of OSTs,
4198 the MDS is always picking the same starting OST for each file.
4199 Return the OST selection heuristic to the original design.
4204 Description: mount failures may take full timeout to return an error
4205 Details : Under some heavy load conditions it is possible that a
4206 failed mount can wait for the full obd_timeout interval,
4207 possibly several minutes, before reporting an error.
4208 Instead return an error as soon as the status is known.
4210 Frequency : quota enabled and large files being deleted
4212 Description: releasing more than 4GB of quota at once hangs OST
4213 Details : If a user deletes more than 4GB of files on a single OST it
4214 will cause the OST to spin in an infinite loop. Release
4215 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4220 Description: statfs data retrieved from /proc may be stale or zero
4221 Details : When reading per-device statfs data from /proc, in the
4222 {kbytes,files}_{total,free,avail} files, it may appear
4223 as zero or be out of date.
4226 Frequency : systems with MD RAID1 external journal devices
4228 Description: lconf's call to blkid is confused by RAID1 journal devices
4229 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4230 of returning all block devices that match the journal UUID.
4233 Frequency : always, for aggregate stripe size over 4GB
4235 Description: assertion fails when trying to use 4GB stripe size
4236 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
4237 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
4240 Frequency : always on ppc64
4242 Description: the first write on an ext3 filesystem with mballoc got stuck
4243 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4244 perform endianness conversion.
4246 ------------------------------------------------------------------------------
4248 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4250 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
4251 INTEROPERATE with older versions automatically. Please read the
4252 user documentation before upgrading any part of a live system.
4253 * WARNING: Lustre networking configuration changes are required with
4254 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
4257 * Support for kernels:
4258 2.6.9-22.0.2.EL (RHEL 4)
4259 2.6.5-7.244 (SLES 9)
4260 2.6.12.6 vanilla (kernel.org)
4263 Severity : enhancement
4264 Bugzilla : 7981/8208
4265 Description: Introduced Lustre Networking (LNET)
4266 Details : LNET is new networking infrastructure for Lustre, it includes
4267 a reorganized network configuration mode (see the user
4268 documentation for full details) as well as support for routing
4269 between different network fabrics. Lustre Networking Devices
4270 (LNDs) for the supported network fabrics have also been
4271 created for this new infrastructure.
4273 Severity : enhancement
4274 Description: Introduced Access control lists
4275 Details : clients can set ACLs on files and directories in order to have
4276 more fine-grained permissions than the standard Unix UGO+RWX.
4277 The MDS must be started with the "-o acl" mount option.
4279 Severity : enhancement
4280 Description: Introduced filesystem quotas
4281 Details : Administrators may now establish per-user quotas on the
4284 Severity : enhancement
4286 Description: Configuration change for the XT3
4287 The PTLLND is now used to run Lustre over Portals on the XT3
4288 The configure option(s) --with-cray-portals are no longer used.
4289 Rather --with-portals=<path-to-portals-includes> is used to
4290 enable building on the XT3. In addition to enable XT3 specific
4291 features the option --enable-cray-xt3 must be used.
4296 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
4297 Details : A race between allocating a new llcd and re-getting the llcd_lock
4298 allowed another thread to grab newly-allocated llcd.
4300 Severity : enhancement
4302 Description: 2.6 OST async journal commit and locking fix to improve performance
4303 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
4304 2.6 kernels are now async as they already were in 2.4 kernels so
4305 that they can commit concurrently with the network bulk transfer.
4306 For block-allocated files the filter allocation semaphore is held
4307 to avoid filesystem fragmentation during allocation. BKL lock
4308 removed for 2.6 xattr operations where it is no longer needed.
4313 Description: lconf incorrectly determined whether two IP networks could talk
4314 Details : In some more complicated routing and multiple-network
4315 configurations, lconf will avoid trying to make a network
4316 connection to a disjoint part of the IP space. It was doing the
4317 math incorrectly for one set of cases.
4322 Description: Fix for potential infinite loop processing records in an llog.
4323 Details : If an llog record is corrupted/zeroed, it is possible to loop
4324 forever in llog_process(). Validate the llog record length
4325 and skip the remainder of the block on error.
4328 Frequency : occasional (liblustre only)
4330 Description: liblustre could not open files whose last component is a symlink
4331 Details : sysio_path_walk() would incorrectly pass the open intent to
4332 intermediate path components.
4335 Frequency : rare (liblustre only with non-standard tuning)
4336 Bugzilla : 7201 (7350)
4337 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4338 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4339 after it has been released, only do this for referenced locks
4341 Severity : enhancement
4343 Description: specifying an (invalid) directory default stripe_size of -1
4344 would reset the directory default striping
4345 Details : stripe_size -1 was used internally to signal directory stripe
4346 removal, now use "all default" to signal dir stripe removal
4347 as a directory striping of "all default" is not useful
4350 Frequency : common for large clusters running liblustre clients
4352 Description: doing an ls when liblustre clients are running is slow
4353 Details : sending a glimpse AST to a liblustre client waits for every AST
4354 to time out, as liblustre clients will not respond. Since they
4355 cannot cache data we refresh the OST lock LVB from disk instead.
4357 Severity : enhancement
4359 Description: doing an ls at the same time as file IO can be slow
4360 Details : enqueue and other "small" requests can be blocked behind many
4361 large IO requests. Create a new OST IO portal for non-IO
4362 requests so they can be processed faster.
4365 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4367 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4368 created file with mode 000
4369 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4370 allow an ftruncate() as a non-root user to a file with mode 000.
4371 Lustre now respects this flag to disable mode checks when
4372 truncating a file owned by the user
4375 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4378 Description: Revoking locks from clients that went dead or catatonic might take
4380 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4381 cancellation of such locks instant on servers without waiting for
4382 any reply from clients. Clients drops these locks when cancel
4383 notification from server is received without replying.
4386 Frequency : liblustre-only, when liblustre client dies or becomes busy
4388 Description: Doing ls on Linux clients can take a long time with active
4390 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4391 granting such locks to it in the first place if possible. Locks
4392 are taken by proxy on the OST during the read or write and
4393 dropped immediately afterward. Add connect flags handling, do
4394 not grant locks to liblustre clients for glimpse ASTs.
4396 Severity : enhancement
4398 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4399 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4400 sequential reads and behaves badly with applications doing random
4401 reads. Improve it by reading ahead at least read region, and
4402 avoiding excessive large RPC for small reads.
4404 Severity : enhancement
4406 Description: Creating more than 1000 files for a single job may cause a load
4407 imbalance on the OSTs if there are also a large number of OSTs.
4408 Details : qos_prep_create() uses an OST index reseed value that is an
4409 even multiple of the number of available OSTs so that if the
4410 reseed happens in the middle of the object allocation it will
4411 still utilize the OSTs as uniformly as possible.
4416 Description: OST or MDS may oops in ping_evictor_main()
4417 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
4418 but doesn't restart at beginning of obd_exports_timed list
4421 Severity : enhancement
4423 Description: improve by-nid export eviction on the MDS and OST
4424 Details : allow multiple exports with the same NID to be evicted at one
4425 time without re-searching the exports list.
4428 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
4430 Description: MDS may oops in groups_free()
4431 Details : in rare race conditions a newly allocated group_info struct is
4432 freed again, and this can be NULL. The 2.4 compatibility code
4433 for groups_free() checked for a NULL pointer, but 2.6 did not.
4436 Frequency : common for liblustre clients doing little filesystem IO
4437 Bugzilla : 9352, 7313
4438 Description: server may evict liblustre clients accessing contended locks
4439 Details : if a client is granted a lock or receives a completion AST
4440 with a blocking AST already set it would not reply to the AST
4441 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
4444 Frequency : lfs setstripe, only systems with more than 160 OSTs
4446 Description: unable to set striping with a starting offset beyond OST 160
4447 Details : llapi_create_file() incorrectly limited the starting stripe
4448 index to the maximum single-file stripe count.
4451 Frequency : LDAP users only
4453 Description: lconf did not handle in-kernel recovery with LDAP properly
4454 Details : lconf/LustreDB get_refs() is searching the wrong namespace
4456 Severity : enhancement
4458 Description: bind OST threads to NUMA nodes to improve performance
4459 Details : all OST threads are uniformly bound to CPUs on a single NUMA
4460 node and do their allocations there to localize memory access
4462 Severity : enhancement
4464 Description: llmount can determine client NID directly from Myrinet (GM)
4465 Details : the client NID code from gmnalnid was moved directly into
4466 llmount, removing the need to use this or specifying the
4467 client NID explicitly when mounting GM clients with zeroconf
4470 Frequency : if client is started with down MDS
4472 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
4473 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
4474 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
4475 ptlrpc_import_delayed_req() because we didn't check if the
4476 request was interrupted, and we also didn't break out of the
4477 event loop if there was no timeout
4482 Description: data loss during non-page-aligned writes to a single file from
4483 both multiple nodes and multiple threads on one node at same time
4484 Details : updates to KMS and lsm weren't protected by common lock. Resulting
4485 inconsistency led to false short-reads, that were cached and later
4486 used by ->prepare_write() to fill in partially written page,
4487 leading to data loss.
4490 Frequency : always, if lconf --abort_recovery used
4492 Description: lconf --abort_recovery fails with 'Operation not supported'
4493 Details : lconf was attempting to abort recovery on the MDT device and not
4496 Severity : enhancement
4498 Description: remove cleanup logs
4499 Details : replace lconf-generated cleanup logs with lustre internal
4500 cleanup routines. Eliminates the need for client-cleanup and
4503 Severity : enhancement
4505 Description: add support for EAs (user and system) on lustre filesystems
4506 Details : it is now possible to store extended attributes in the Lustre
4507 client filesystem, and with the user_xattr mount option it
4508 is possible to allow users to store EAs on their files also
4510 Severity : enhancement
4512 Description: Add possibility (config option) to show minimal available OST free
4514 Details : When compiled with --enable-mindf configure option, statfs(2)
4515 (and so, df) will return least minimal free space available from
4516 all OSTs as amount of free space on FS, instead of summary of
4517 free spaces of all OSTs.
4519 Severity : enhancement
4521 Description: do not expand extent locks acquired on OST-side
4522 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
4523 by server: they are not cached anyway.
4526 Frequency : when mmap is used/binaries executed from Lustre
4528 Description: Unmmap pages before throwing them away from read cache.
4529 Details : llap_shrink cache now attempts to unmap pages before discarding
4530 them (if unmapping failed - do not discard). SLES9 kernel has
4531 extra checks that trigger if this unmapping is not done first.
4536 Description: lconf didn't resolve symlinks before checking to see whether a
4537 given mountpoint was already in use
4540 Frequency : when migrating failover services
4541 Bugzilla : 6395, 9514
4542 Description: When migrating a subset of services from a node (e.g. failback
4543 from a failover service node) the remaining services would
4544 time out and evict clients.
4545 Details : lconf --force (implied by --failover) sets the global obd_timeout
4546 to 5 seconds in order to quickly disconnect, but this caused
4547 other RPCs to time out too quickly. Do not change the global
4548 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
4550 Severity : enhancement
4551 Frequency : if MDS is started with down OST
4552 Bugzilla : 9439,5706
4553 Description: Allow startup/shutdown of an MDS without depending on the
4554 availability of the OSTs.
4555 Details : Asynchronously call mds_lov_synchronize during MDS startup.
4556 Add appropriate locking and lov-osc refcounts for safe
4557 cleaning. Add osc abort_inflight calls in case the
4558 synchronize never started.
4561 Frequency : occasional (Cray XT3 only)
4563 Description: root not authorized to access files in CRAY_PORTALS environment
4564 Details : The client process capabilities were not honoured on the MDS in
4565 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
4566 been accessed by an authorized user then root was able to access
4567 the file on the local client also. The root user capabilities
4568 are now allowed on the MDS, as this environment has secure UID.
4571 Frequency : occasional
4573 Description: ldiskfs "too long searching" message happens too often
4574 Details : A debugging message (otherwise harmless) prints too often on
4575 the OST console. This has been reduced to only happen when
4576 there are fragmentation problems on the filesystem.
4581 Description: Division by zero in statfs when all OSCs are inactive
4582 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
4583 lov_statfs divides by value returned by lov_get_stripecnt().
4587 Bugzilla : 9489, 3273
4588 Description: First write from each client to each OST was only 4kB in size,
4589 to initialize client writeback cache, which caused sub-optimal
4590 RPCs and poor layout on disk for the first writen file.
4591 Details : Clients now request an initial cache grant at (re)connect time
4592 and so that they can start streaming writes to the cache right
4593 away and always do full-sized RPCs if there is enough data.
4594 If the OST is rebooted the client also re-establishes its grant
4595 so that client cached writes will be honoured under the grant.
4600 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
4601 Details : Now I/O RPCs go to different portal number and (presumably) fast
4602 lock requests (and glimses) and other RPCs get their own service
4603 threads pool that should be able to service those RPCs
4606 Severity : enhancement
4608 Description: Ability to exchange lustre version between client and servers and
4609 issue warnings at client side if client is too old. Also for
4610 liblustre clients there is ability to refuse connection of too old
4612 Details : New 'version' field is added to connect data structure that is
4613 filled with version info. That info is later checked by server and
4617 Frequency : rare, liblustre only.
4618 Bugzilla : 9296, 9581
4619 Description: Two simultaneous writes from liblustre at offset within same page
4620 might proceed at the same time overwriting eachother with stale
4622 Details : I/O lock withing llu_file_prwv was released too early, before data
4623 actually was hitting the wire. Extended lock-holding time until
4624 server acknowledges receiving data.
4627 Frequency : extremely rare. Never observed in practice.
4629 Description: avoid generating lustre_handle cookie of 0.
4630 Details : class_handle_hash() generates handle cookies by incrementing
4631 global counter, and can hit 0 occasionaly (this is unlikely, but
4632 not impossible, because initial value of cookie counter is
4633 selected randonly). Value of 0 is used as a sentinel meaning
4634 "unassigned handle" --- avoid it. Also coalesce two critical
4635 sections in this function into one.
4637 Severity : enhancement
4639 Description: allow liblustre clients to delegate truncate locking to OST
4640 Details : To avoid overhead of locking, liblustre client instructs OST to
4641 take extent lock in ost_punch() on client's behalf. New connection
4642 flag is added to handle backward compatibility.
4644 Severity : enhancement
4645 Bugzilla : 4928, 7341, 9758
4646 Description: allow number of OST service threads to be specified
4647 Details : a module parameter allows the number of OST service threads
4648 to be specified via "options ost ost_num_threads={N}" in the
4649 OSS's /etc/modules.conf or /etc/modprobe.conf.
4653 Bugzilla : 6146, 9635, 9895
4654 Description: servers crash with bad pointer in target_handle_connect()
4655 Details : In rare cases when a client is reconnecting it was possible that
4656 the connection request was the last reference for that export.
4657 We would temporarily drop the export reference and get a new
4658 one, but this may have been the last reference and the export
4659 was just destroyed. Get new reference before dropping old one.
4661 Severity : enhancement
4662 Frequency : if client is started with failover MDS
4664 Description: Allow multiple MDS hostnames in the mount command
4665 Details : Try to read the configuration from all specified MDS
4666 hostnames during a client mount in case the "primary"
4669 Severity : enhancement
4671 Description: Stop sending data to evicted clients as soon as possible.
4672 Details : Check if the client we are about to send or are sending data to
4673 was evicted already. (Check is done every second of waiting,
4674 for which l_wait_event interface was extended to allow checking
4675 of exit condition at specified intervals).
4678 Frequency : rare, normally only when NFS exporting is done from client
4680 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
4682 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
4685 Frequency : always, if srand() is called before liblustre initialization
4687 Description: Liblustre uses system PRNG disturbing its usage by user application
4688 Details : Introduce internal to lustre fast and high-quality PRNG for
4689 lustre usage and make liblustre and some other places in generic
4690 lustre code to use it.
4692 Severity : enhancement
4693 Bugzilla : 9477, 9557, 9870
4694 Description: Verify that the MDS configuration logs are updated when xml is
4695 Details : Check if the .xml configuration logs are newer than the config
4696 logs stored on the MDS and report an error if this is the case.
4697 Request --write-conf, or allow starting with --old_conf.
4699 Severity : enhancement
4701 Description: Handle symlinks in the path when checking if Lustre is mounted.
4702 Details : Resolve intermediate symlinks when checking if a client has
4703 mounted a filesystem to avoid duplicate client mounts.
4708 Description: lconf can hit an error exception but still return success.
4709 Details : The lconf command catches the Command error exception at the top
4710 level script context and will exit with the associated exit
4711 status, but doesn't ensure that this exit status is non-zero.
4716 Description: failure of ptlrpc thread startup can cause oops
4717 Details : Starting a ptlrpc service thread can fail if there are a large
4718 number of threads or the server memory is very fragmented.
4719 Handle this without oopsing.
4722 Frequency : always, only if liblustre and non-default acceptor port was used
4724 Description: liblustre cannot connect to servers with non-default acceptor port
4725 Details : tcpnal_set_default_params() was not called and was therefore
4726 ignoring the environment varaible TCPNAL_PORT, as well as other
4727 TCPNAL_ environment variables
4732 Description: two objects could be created on the same OST for a single file
4733 Details : If an OST is down, in some cases it was possible to create two
4734 objects on a single OST for a single file. No problems other
4735 than potential performance impact and spurious error messages.
4739 Bugzilla : 5681, 9562
4740 Description: Client may oops in ll_unhash_aliases
4741 Details : Client dcache may become inconsistent in race condition.
4742 In some cases "getcwd" can fail if the current directory is
4748 Description: Inode refcounting problems in NFS export code
4749 Details : link_raw functions used to call d_instantiate without obtaining
4750 extra inode reference first.
4754 Bugzilla : 9942, 9903
4755 Description: Referencing freed requests leading to crash, memleaks with NFS.
4756 Details : We used to require that call to ll_revalidate_it was always
4757 followed by ll_lookup_it. Also with revalidate_special() it is
4758 possible to call ll_revalidate_it() twice for the same dentry
4759 even if first occurence returned success. This fix changes semantic
4760 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
4761 reference on a request referred from the intent.
4762 ll_intent_release() then releases such a request.
4765 Frequency : rare, normally benchmark loads only
4767 Description: unlinked inodes were kept in memory on the client
4768 Details : If a client is repeatedly creating and unlinking files it
4769 can accumulate a lot of stale inodes in the inode slab cache.
4770 If there is no other client load running this can cause the
4771 client node to run out of memory. Instead flush old inodes
4772 from client cache that have the same inode number as a new inode.
4775 Frequency : SLES9 2.6.5 kernel and long filenames only
4776 Bugzilla : 9969, 10379
4777 Description: utime reports stale NFS file handle
4778 Details : SLES9 uses out-of-dentry names in some cases, which confused
4779 the lustre dentry revalidation. Change it to always use the
4783 Frequency : rare, unless heavy write-truncate concurrency is continuous
4784 Bugzilla : 4180, 6984, 7171, 9963, 9331
4785 Description: OST becomes very slow and/or deadlocked during object unlink
4786 Details : filter_destroy() was holding onto the parent directory lock
4787 while truncating+unlinking objects. For very large objects this
4788 may block other threads for a long time and slow overall OST
4789 responsiveness. It may also be possible to get a lock ordering
4790 deadlock in this case, or run out of journal credits because of
4791 the combined truncate+unlink. Solution is to do object truncate
4792 first in one transaction without parent lock, and then do the
4793 final unlink in a new transaction with the parent lock. This
4794 reduces the lock hold time dramatically.
4797 Frequency : rare, 2.4 kernels only
4799 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
4800 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
4801 before unmounting their filesystem in order to avoid deadlock.
4802 The kernel_locked() function in 2.4 kernels only checks whether
4803 the kernel lock is held, not whether it is this process that is
4804 holding it as 2.6 kernels do.
4809 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
4810 Details : The client ptlrpc code may be trying to reconnect to a down
4811 server before a previous connection attempt has timed out.
4812 Increase the reconnect interval to be longer than the connection
4813 timeout interval to avoid sending duplicate connections to
4817 Frequency : echo_client brw_test command
4819 Description: fix echo_client to work with OST preallocated code
4820 Details : OST preallocation code (5137) didn't take echo_client IO path
4821 into account: echo_client calls filter methods outside of any
4822 OST thread and, hence, there is no per-thread preallocated
4823 pages and buffers to use. Solution: hijack pga pages for IO. As
4824 a byproduct, this avoids unnecessary data copying.
4828 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
4829 Description: Client can oops in mdc_commit_close() after open replay
4830 Details : It was possible for the MDS to return an open request with no
4831 transaction number in mds_finish_transno() if the client was
4832 evicted, but without actually returning an error. Clients
4833 would later try to replay that open and may trip an assertion
4834 Simplify the client close codepath, and always return an error
4835 from the MDS in case the open is not successful.
4838 Frequency : rare, 2.6 OSTs only
4840 Description: OST may deadlock under high load on fragmented files
4841 Details : If there was a heavy load and highly-fragmented OST filesystems
4842 it was possible to have all the OST threads deadlock waiting on
4843 allocation of biovecs, because the biovecs were not released
4844 until the entire RPC IO was completed. Instead, release biovecs
4845 as soon as they are complete to ensure forward IO progress.
4847 Severity : enhancement
4849 Description: Support for specifying external journal device at mount
4850 Details : If an OST or MDS device is formatted with an external journal
4851 device, this device major/minor is stored in the ext3 superblock
4852 and may not be valid for failover. Allow detecting and
4853 specifying the external journal at mount time.
4858 Description: Mounting an MDS with pending unlinked files may cause oops
4859 Details : target_finish_recovery() calls mds_postrecov() which returned
4860 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
4861 considers this an error and immediately begins cleaning up the
4862 lov, just after starting the mds_lov process
4864 Severity : enhancement
4866 Description: Implement 'lfs df' to report actual free space on per-OST basis
4867 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
4868 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
4869 usage of objects; '-h' to report in human readable format.
4871 ------------------------------------------------------------------------------
4873 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
4880 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
4881 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
4882 at startup during orphan unlinking. Instead, skip buddy update
4883 before setup, as it will be regenerated after recovery is complete.
4888 Description: If an OST is inactive, its locks might reference stale inodes.
4889 Details : lov_change_cbdata() must iterate over all namespaces, even if
4890 they are inactive to clear inode references from the lock.
4892 Severity : enhancement
4893 Frequency : occasional, if non-standard max_dirty_mb used
4895 Description: Client will block write RPCs if not enough grant
4896 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
4897 then the client will block writes while waiting for another RPC
4898 to complete instead of consuming its dirty limit. With change
4899 we get improved performance when max_dirty_mb is small.
4901 Severity : enhancement
4902 Bugzilla : 3389, 6253
4903 Description: Add support for supplementary groups on the MDS.
4904 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
4905 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
4906 lookups for user supplementary groups into a cache.
4910 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
4911 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
4912 for liblustre O_EXCL is enforced by MDS.
4916 Description: readdir never returns NULL in liblustre.
4917 Details : Corrected llu_iop_getdirentries logic, to return offset of next
4918 dentry in struct dirent.
4922 Frequency : liblustre only, depends on application IO pattern
4923 Description: liblustre clients evicted if not contacting servers
4924 Details : Don't put liblustre clients into the ping_evictor list, so
4925 they will not be evicted by the pinger ever.
4927 Severity : enhancement
4929 Description: Add ability to evict clients by NID from MDS.
4930 Details : By echoing "nid:$NID" string into
4931 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
4932 $NID would be instantly evicted from this MDS and from all active
4933 OSTs connected to it.
4937 Description: Do not query file size twice, somewhat slowing stat(2) calls.
4938 Details : lookup_it_finish() used to query file size from OSTs that was not
4943 Description: service threads change working directory to that of init
4944 Details : Starting lustre service threads may pin the working directory
4945 of the parent thread, making that filesystem busy. Threads
4946 now change to the working directory of init to avoid this.
4950 Frequency : during shutdown only
4951 Description: shutdown with a failed MDS or OST can cause unmount to hang
4952 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
4957 Frequency : 2.6 only
4958 Description: chmod/chown may include an extra supplementary group
4959 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
4960 supplementary group and if none is specified this is used.
4963 Bugzilla : 5479 (6816)
4965 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
4966 Details : If lookup is in progress on a file that is unlinked we might try
4967 to revalidate the inode and fail in revalidate after lookup is
4968 complete and ll_file_open() enqueues the open again but
4969 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
4972 Frequency : always, if lconf --abort_recovery used
4974 Description: lconf --abort_recovery fails with 'Operation not supported'
4975 Details : lconf was attempting to abort recovery on the MDT device and not
4978 ------------------------------------------------------------------------------
4980 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
4985 Frequency : rare (only unsupported configurations with a node running as an
4987 Bugzilla : 6514, 5137
4988 Description: Mounting a Lustre file system on a node running as an OST could
4990 Details : OSTs now preallocates memory needed to write out data at
4991 startup, instead of when needed, to avoid having to
4992 allocate memory in possibly low memory situations.
4993 Specifically, if the file system is mounted on on OST,
4994 memory pressure could force it to try to write out data,
4995 which it needed to allocate memory to do. Due to the low
4996 memory, it would be unable to do so and the node would
4997 become unresponsive.
4999 Severity : enhancement
5001 Description: Addition of lconf --service command line option
5002 Details : lconf now accepts a '--service <arg>' option, which is
5003 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
5005 Severity : enhancement
5007 Description: Failover mode is now the default for OSTs.
5008 Details : By default, OSTs will now run in failover mode. To return to
5009 the old behaviour, add '--failout' to the lmc line for OSTs.
5011 Severity : enhancement
5013 Description: Health checks are now provided for MDS and OSTs
5014 Details : Additional detailed health check information on MSD and OSTs
5015 is now provided through the procfs health_check value.
5018 Frequency : occasional, depends on IO load
5020 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
5021 after numerous create/delete cycles
5022 Details : The ext3 inode allocation policy would not allocate new inodes
5023 very well on the OSTs because there are no new directories
5024 being created. Instead we look for groups with free space if
5025 the parent directories are nearly full.
5030 Description: Network or server problems during mount may cause partially
5031 mounted clients instead of returning an error.
5032 Details : The config llog parsing code may overwrite the error return
5033 code during mount error handling, returning success instead
5039 Description: MDS can fail to allocate large reply buffers
5040 Details : After long uptimes the MDS can fail to allocate large reply
5041 buffers (e.g. zconf client mount config records) due to memory
5042 fragmentation or consumption by the buffer cache. Preallocate
5043 some large reply buffers so that these replies can be sent even
5044 under memory pressure.
5048 Frequency : rare (liblustre)
5049 Description: fsx running with liblustre complained that using truncate() to
5050 extend the file doesn't work. This patch corrects that issue.
5051 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
5052 ATTR_SIZE and lsm use in llu_setattr_raw.
5056 Frequency : rare, only 2.6 kernels
5057 Description: Unusual file access patterns on the MDS may result in inode
5058 data being lost in very rare circumstances.
5059 Details : Bad interaction between the ea-in-inode patch and the "no-read"
5060 code in the 2.6 kernel caused the inode and/or EA data not to
5061 be read from disk, causing single-file corruption.
5065 Frequency : rare, only 2.6 filesystems using extents
5066 Description: Heavy concurrent write and delete load may cause data corruption.
5067 Details : It was possible under high-load situations to have an extent
5068 metadata block in the block device cache from a just-unlinked
5069 file overwrite a newly-allocated data block. We now unmap any
5070 metadata buffers that alias just-allocated data blocks.
5074 Frequency : filesystems with default stripe_count larger than 77
5075 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
5076 Details : lconf specifies an inode size of 4096 bytes when the default
5077 stripe_count is larger than 77. This conflicts with the default
5078 inode density of 1 per 4096 bytes. Allocate smaller inodes in
5079 this case to avoid pinning too much memory for large EAs.
5081 ------------------------------------------------------------------------------
5083 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
5088 Frequency : rare (extremely heavy IO load with hundreds of clients)
5090 Description: Client is evicted, gets IO error writing to file
5091 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
5092 caused clients to be evicted for AST timeouts. The fixes in
5093 bug 5192 mean we no longer need to have such short AST timeouts
5094 so ldlm_timeout has been increased.
5097 Frequency : occasional during --force or --failover shutdown under load
5098 Bugzilla : 5949, 4834
5099 Description: Server oops/LBUG if stopped with --force or --failover under load
5100 Details : a collection of import/export refcount and cleanup ordering
5101 issues fixed for safer force cleanup
5104 Frequency : only filesystems larger than 120 OSTs
5105 Bugzilla : 5990, 6223
5106 Description: lfs getstripe would oops on a very large filesystem
5107 Details : lov_getconfig used kfree on vmalloc'd memory
5110 Frequency : only filesystems exporting via NFS to Solaris 10 clients
5111 Bugzilla : 6242, 6243
5112 Description: reading from files that had been truncated to a non-zero size
5113 but never opened returned no data
5114 Details : ll_file_read() reads zeros from no-object files to EOF
5119 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
5120 Details : There's little harm in aborting MDS/OSS recovery and letting it
5121 try again, so I removed the LASSERT and return an error instead.
5123 Severity : enhancement
5125 Description: New debugging infrastructure for tracking down data corruption
5126 Details : The I/O checksum code was replaced to: (a) control it at runtime,
5127 (b) cover more of the client-side code path, and (c) try to narrow
5128 down where problems occurred
5132 Bugzilla : 3819, 4364, 4397, 6313
5133 Description: Racing close and eviction MDS could cause assertion in mds_close
5134 Details : It was possible to get multiple mfd references during close and
5135 client eviction, leading to one thread referencing a freed mfd.
5137 Severity: : enhancement
5138 Bugzilla : 3262, 6359
5139 Description: Attempts to reconnect to servers are now more aggressive.
5140 Details : This builds on the enhanced upcall-less recovery that was added
5141 in 1.4.2. When trying to reconnect to servers, clients will
5142 now try each server in the failover group every 10 seconds. By
5143 default, clients would previously try one server every 25 seconds.
5148 Description: After recovery, certain operations trigger a failed
5149 assertion on a client.
5150 Details : Failing over an mds, using lconf -d --failover, while a
5151 client was doing a readdir() call would cause the client to
5152 LBUG after recovery completed and the readdir() was resent.
5154 Severity : enhancement
5156 Description: Default groups are now added by lconf
5157 Details : You can now run lconf --group <servicename> without having to
5158 manually add groups with lmc.
5161 Frequency : occasional
5163 Description: Nodes with an elan id of 0 trigger a failed assertion
5166 Frequency : always when accessing e.g. tty/console device nodes
5168 Description: tty and some other devices nodes cannot be used on lustre
5169 Details : file's private_data field is used by device data and lustre
5170 values in there got lost. New field was added to struct file to
5171 store fs-specific private data.
5174 Frequency : when exporting Lustre via NFS
5176 Description: NFSD failed occasionally when looking up a path component
5177 Details : NFSD is looking up ".." which was broken in ext3 directories
5178 that had grown large enough to become hashed.
5181 Frequency : Clusters with multiple interfaces not on the same subnet
5183 Description: Nodes will repeatedly try to reconnect to an interface which it
5184 cannot reach and report an error to the log.
5185 Details : Extra peer list entries will be created by lconf with some peers
5186 unreachable. lconf now validates the peer before adding it.
5189 Frequency : Only if a default stripe is set on the filesystem root.
5191 Description: Setting a default stripe on the filesystem root prevented the
5192 filesystem from being remounted.
5193 Details : The client was sending extra request flags in the root getattr
5194 request and did not allocate a reply buffer for the dir EA.
5197 Frequency : occasional, higher if lots of files are accessed by one client
5198 Bugzilla : 6159, 6097
5199 Description: Client trips assertion regarding lsm mismatch/magic
5200 Details : While revalidating inodes the VFS looks up inodes with ifind()
5201 and in rare cases can find an inode that is being freed.
5202 The ll_test_inode() code will free the lsm during ifind()
5203 when it finds an existing inode and then the VFS later attaches
5204 this free lsm to a new inode.
5208 Bugzilla : 6422, 7030
5209 Description: MDS deadlock between mkdir and client eviction
5210 Details : Creating a new file via mkdir or mknod (starting a transaction
5211 and getting the ns lock) can deadlock with client eviction
5212 (gets ns lock and trying to finish a synchronous transaction).
5215 Frequency : occasional
5216 Description: While starting a server, the fsfilt_ext3 module could not be
5218 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
5219 kernels. Previously, lconf would still use the ext3 name
5220 when trying to load modules. Now, it will correctly use
5221 ext3 on 2.4 and ldiskfs on 2.6.
5223 Severity : enhancement
5224 Description: The default stripe count has been changed to 1
5225 Details : The interpretation of the default stripe count (0, to lfs
5226 or lmc) has been changed to mean striping across a single
5227 OST, rather than all available. For general usage we have
5228 found a stripe count of 1 or 2 works best.
5230 Severity : enhancement
5231 Description: Add support for compiling against Cray portals.
5232 Details : Conditional compiling for some areas that are different
5236 Frequency : occasional
5237 Bugzilla : 6409, 6834
5238 Description: Creating files with an explicit stripe count may lead to
5239 a failed assertion on the MDS
5240 Details : If some OSTs are full or unavailable, creating files may
5241 trigger a failed assertion on the MDS. Now, Lustre will
5242 try to use other servers or return an error to the
5246 Frequency : occasional
5248 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
5249 caused lock timeout during readahead (since 1.4.2).
5250 Details : Processes doing readahead might match a lock that hasn't been
5251 granted yet if there are overlapping and conflicting lock
5252 requests. The readahead process waits on ungranted lock
5253 (original lock is CBPENDING), while OST waits for that process
5254 to cancel CBPENDING read lock and eventually evicts client.
5256 Severity : enhancement
5258 Description: Initial enabling of flock support for clients
5259 Details : Implements fcntl advisory locking and file status functions.
5260 This feature is provided as an optional mount flag (default
5261 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
5262 locking are implemented yet, and those that are are not guaranteed
5263 to be completely correct in production environments.
5264 mount -t lustre -o [flock|noflock] ...
5267 Frequency : occasional
5269 Description: OSTs running 2.4 kernels but with extents enabled might trip an
5270 assertion in the ext3 JBD (journaling) layer.
5271 Details : The b_committed_data struct is protected by the big kernel lock
5272 in 2.4 kernels, serializing journal_commit_transaction() and
5273 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
5274 access to this struct. In 2.6 kernels there is finer grained
5275 locking to improve SMP performance of the JBD layer.
5279 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
5284 Description: merge and cleanup kernel patches.
5287 -----------------------------------------------------------------------------
5289 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
5291 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
5292 versions of Lustre. Please update all servers and clients to
5293 version 1.4.2 or later at the same time. You must also run
5294 "lconf --write-conf {config}.xml" on the MDS while it is stopped
5295 to update the configuration logs.
5297 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
5298 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
5299 - swab llog records of type '0' so we get proper header size/idx (5861)
5300 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
5301 - fix rename of one directory over another leaking an inode (5953)
5302 - avoid SetPageDirty on 2.6 (5981)
5303 - don't re-add just-being-destroyed locks to the waiting list (5653)
5304 - when creating new directories, inherit the parent's custom
5305 striping settings if present parent (3048)
5306 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
5307 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
5308 - don't hold client locks on temporary worklist from l_lru (5666)
5309 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
5310 - automatically evict dead clients (5921)
5311 - Update file size properly in create+truncate+fstat case (6196)
5312 - Do not unhash mountpoint dentries, do not allow removal of
5314 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
5315 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
5316 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
5317 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
5318 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
5319 - Correctly handle reads of files with no objects (6243)
5320 - lctl recover will also mark a device active if deactivate used (5933)
5322 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
5323 - allow --write-conf on an MDS with different nettype than client (5619)
5324 - don't write config llogs to MDS for mounts not from that MDS (5617)
5325 - lconf should create multiple TCP connections from a client (5201)
5326 - init scripts are now turned off by default; run chkconfig --on
5327 lustre and chkconfig --on lustrefs to use them
5328 - upcalls are no longer needed for clients to recover to failover
5330 - add --abort-recovery option to lconf to abort recovery on device
5332 - add support for an arbitrary number of OSTs (3026)
5333 - Quota support protocol changes.
5334 - forward compatibility changes to wire structs (6007)
5335 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5336 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5337 - improved "lustre status" script
5338 - initialize blocksize for non-regular files (6062)
5339 - added --disable-server and --disable-client configure options (5782)
5340 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5341 - Vanilla 2.4.29 support
5342 - increase maximum number of obd devices to 520 (6242)
5343 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5344 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5345 SLES 9 2.6.5 kernels
5346 - limit stripes per file to 160 (the maximum EA size) (6093)
5348 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5351 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5352 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5353 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5354 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5355 - small ext3 extents cleanups and fixes (5733)
5356 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5357 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5358 - avoid needless client->OST connect, fix handle mismatch (5317)
5359 - fix DLM error path that led to out-of-sync client, long delays (5779)
5360 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5361 - fix several locking issues related to i_size (5492,5624,5654,5672)
5362 - don't move pending lock onto export if it is already evicted (5683)
5363 - fix kernel oops when creating .foo in unlinked directory (5548)
5364 - fix deadlock in obdfilter statistics vs. object create (5811)
5365 - use time_{before,after} to avoid timer jiffies wrap (5882)
5366 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5367 - Do not leak request if server was not able to process it (5154)
5368 - If mds_open unable to find parent dir, make that negative lookup(5154)
5369 - don't create new directories with extent-mapping (5909, 5936)
5371 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5372 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5373 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5374 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5375 - support for RHEL4 kernel on i686 (5773)
5376 - provide error messages when incompatible logs are encountered (5898)
5378 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5379 * version 1.4.0.10 (1.4.1 release candidate 1)
5381 - don't keep a lock reference when lock is not granted (4238)
5382 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5383 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5384 - drop import inflight refcount on signal_completed_replay error (5255)
5385 - unlock page after async write error during send (3677)
5386 - handle missing objects in filter_preprw_read properly (5265)
5387 - no transno return for symlink open, don't save no-trasno open (3440)
5388 - don't try to complete elan receive that already failed (4012)
5389 - free RPC server reply state on error (5406)
5390 - clean up thread from ptlrpc_start_thread() on error (5160)
5391 - readahead could read extra page into cache that wasn't ejected (5388)
5392 - prevent races in class_attach/setup/cleanup/detach (5260)
5393 - don't dereference de->d_inode after l_dput of de (5458)
5394 - use "int" for stripe value returned from lock_to_stripe (5544)
5395 - mballoc allocation and error-checking fixes in 2.6 (5504)
5396 - block device patches to fix I/O request sizes in 2.6 (5482)
5397 - look up hostnames for IB nals (5602)
5398 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5399 - don't start multiple acceptors for the same port (5277)
5400 - fix incorrect LASSERT in mds_getattr_name (5635)
5401 - export a proc file for general "ping" checking (5628)
5402 - fix "lfs check" to not block when the MDS is down (5628)
5404 - service request history (4965)
5405 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5406 - create an "evict_client" /proc entry on OSTs, like the MDS has
5407 - fix mount usage message, return errors per mount(8) (5168)
5408 - change grep [] to grep "[]" in tests so they work in more UMLs
5409 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5410 - remove ancient LOV_MAGIC_V0 EA support (5047)
5411 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5412 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5413 - allow readahead allocations to fail when low on memory (5383)
5414 - mmap locking landed again, after considerable improvement (2828)
5415 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
5417 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
5420 - send OST transaction number in read/write reply to free req (4966)
5421 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5422 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5423 - the watchdog thread now runs as interruptible (5246)
5424 - flock/lockf fixes (but it's still disabled, pending 5135)
5425 - don't use EXT3 constants in llite code (5094)
5426 - memory shortage at startup could cause assertion (5176)
5428 - reorganization of lov code
5429 - single portals codebase
5431 - add extents/mballoc support (5025)
5432 - direct I/O reads in the obdfilter (4048)
5433 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
5435 tbd Cluster File Systems, Inc. <info@clusterfs.com>
5438 - send OST transaction number in read/write reply to free req (4966)
5439 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5440 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5441 - the watchdog thread now runs as interruptible (5246)
5442 - handle missing objects in filter_preprw_read properly (5265)
5443 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5444 - drop import inflight refcount on signal_completed_replay error (5255)
5445 - unlock page after async write error during send (3677)
5446 - return original error code on reconstructed replies (3761)
5447 - no transno return for symlink open, don't save no-trasno open (3440)
5449 - add pid to ldlm debugging output (4922)
5450 - bump the watchdog timeouts -- we can't handle 30sec yet
5451 - extra debugging for orphan dentry/inode bug (5259)
5453 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
5456 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
5457 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
5458 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
5459 - fix NULL dereference in /proc/sys/portals/routes (4827)
5460 - allow failed mdc_close() operations to be interrupted (4561)
5461 - stop precreate on OST before MDS would time out on it (4778)
5462 - don't send partial-page writes before EOF from client (4410)
5463 - discard client grant for sub-page writes on large-page clients (4520)
5464 - don't free dentries not owned by NFS code, check generation (4806)
5465 - fix lsm leak if mds_create_objects() fails (4801)
5466 - limit debug_daemon file size, always print CERROR messages (4789)
5467 - use transno after validating reply (3892)
5468 - process timed out requests if import state changes (3754)
5469 - update mtime on OST during writes, return in glimpse (4829)
5470 - add mkfsoptions to LDAP (4679)
5471 - use ->max_readahead method instead of zapping global ra (5039)
5472 - don't interrupt __l_wait_event() during strace
5474 - add software watchdogs to catch hung threads quickly (4941)
5475 - make lustrefs init script start after nfs is mounted
5476 - fix CWARN/ERROR duplication (4930)
5477 - return async write errors to application if possible (2248)
5478 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
5479 - print NAL number in %x format (4645)
5480 - update barely-supported suse-2.4.21-171 series (4842)
5481 - support for sles 9 %post scripts
5482 - support for building 2.6 kernel-source packages
5483 - support for sles km_* packages
5485 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
5488 - ignore -ENOENT errors in osc_destroy (3639)
5489 - notify osc create thread that OSC is being cleaned up (4600)
5490 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
5491 - reconstruct ost_handle() like mds_handle() (4657)
5492 - create a new thread to do import eviction to avoid deadlock (3969)
5493 - let lconf resolve symlinked-to devices (4629)
5494 - don't unlink "objects" from directory with default EA (4554)
5495 - hold socknal file ref over connect in case target is down (4394)
5496 - allow more than 32000 subdirectories in a single directory (3244)
5497 - fix blocks count for O_DIRECT writes (3751)
5498 - OST returns ENOSPC from object create when no space left (4539)
5499 - don't send truncate RPC if file size isn't changing (4410)
5500 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
5501 - bind to privileged port in socknal and tcpnal (3689)
5503 - rate limit CERROR/CWARN console message to avoid overload (4519)
5504 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
5505 - basic mmap support (3918)
5506 - kernel patch series update from b1_4 (4711)
5508 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
5511 - avoid crash during MDS cleanup with OST shut down (2775)
5512 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
5513 - don't use bad inodes on the MDS (3744)
5514 - dynamic object preallocation to improve recovery speed (4236)
5515 - don't hold spinlock over lock dumping or change debug flags (4401)
5516 - don't zero obd_dev when it is force cleaned (3651)
5517 - print grants to console if they go negative (4431)
5518 - "lctl deactivate" will stop automatic recovery attempts (3406)
5519 - look for existing locks in ldlm_handle_enqueue() (3764)
5520 - don't resolve lock handle twice in recovery avoiding race (4401)
5521 - revalidate should check working dir is a directory (4134)
5523 - don't always mark "slow" obdfilter messages as errors (4418)
5525 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
5528 - don't close LustreDB during write_conf until it is done (3860)
5529 - fix typo in lconf for_each_profile (3821)
5530 - allow dumping logs from multiple threads at one time (3820)
5531 - don't allow multiple threads in OSC recovery (3812)
5532 - fix debug_size parameters (3864)
5533 - fix mds_postrecov to initialize import for llog ctxt (3121)
5534 - replace config semaphore with spinlock (3306)
5535 - be sure to send a reply for a CANCEL rpc with bad export (3863)
5536 - don't allow enqueue to complete on a destroyed export (3822)
5537 - down write_lock before checking llog header bitmap (3825)
5538 - recover from lock replay timeout (3764)
5539 - up llog sem before sending rpc (3652)
5540 - reduce ns lock hold times when setting kms (3267)
5541 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
5542 - fix NULL deref and obd_dev leak on setup error (3312)
5543 - replace some LBUG about llog ops with error handling (3841)
5544 - don't match INVALID dentries from d_lookup and spin (3784)
5545 - hold dcache_lock while marking dentries INVALID and hashing (4255)
5546 - fix invalid assertion in ptlrpc_set_wait (3880)
5548 - add libwrap support for the TCP acceptor (3996)
5549 - add /proc/sys/portals/routes for non-root route listing (3994)
5550 - allow setting MDS UUID in .xml (2580)
5551 - print the stack of a process that LBUGs (4228)
5553 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
5556 - don't cleanup request in ll_file_open() on failed MDS open (3430)
5557 - make sure to unset replay flag from failed open requests (3440)
5558 - if default stripe count is 0, use OST count for inode size (3636)
5559 - update parent mtime/ctime on client for create/unlink (2611)
5560 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
5561 - free recovery state on server during a forced cleanup (3571)
5562 - unregister_reply for resent reqs (3063)
5563 - loop back devices mounting and status check on 2.6 (3563)
5564 - fix resource-creation race that can provoke i_size == 0 (3513)
5565 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
5566 - more debugging for page-accounting assertion (3746)
5567 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
5568 - avoid deadlock after precreation failure (3758)
5569 - fix race and lock order deadlock in orphan handling (3450, 3750)
5570 - add validity checks when grabbing inodes from l_ast_data (3599)
5572 - add /proc/.../recovery_status to obdfilter (3428)
5573 - lightweight CDEBUG infrastructure, debug daemon (3668)
5574 - change default OSC RPC parameters to be better on small clusters
5575 - turn off OST read cache for files smaller than 32MB
5576 - install man pages and include them in rpms (3100)
5577 - add new init script for (un)mounting lustre filesystems (2593)
5578 - run chkconfig in %post for init scripts (3701)
5579 - drop scimac NAL (unmaintained)
5581 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
5584 - clean kiobufs before and after use (3485)
5585 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
5586 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
5587 - change init script to fail more clearly if not run as root (1528)
5588 - allow clients to reconnect during replay (1742)
5589 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
5590 - don't do DNS lookups on NIDs too small for IP addresses (3442)
5591 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
5592 - fix cond_resched (3554)
5593 - only evict unfinished clients after recovery (3515)
5594 - allow bulk resend, prevent data loss (3570)
5595 - dynamic ptlrpc request buffer allocation (2102)
5596 - don't allow unlinking open directory if it isn't empty (2904)
5597 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
5598 - remove extraneous obd dereference causing LASSERT failure (3334)
5599 - don't use get_cycles() when creating temp. files on the mds (3156)
5600 - hold i_sem when setting i_size in ll_extent_lock() (3564)
5601 - handle EEXIST for set-stripe, set proper directory name (3336)
5603 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
5604 - fix ksocknal_fmb_callback() error messages (2918)
5606 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
5609 - don't copy lvb into (possibly NULL) reply on error (2983)
5610 - don't deref dentry after dput, don't free lvb on error (2922)
5611 - use the kms to determine writeback rpc length (2947)
5612 - increment oti_logcookies when osc is inactive (2948)
5613 - update client's i_blocks count via lvb messages (2543)
5614 - handle intent open/close of special files properly (1557)
5615 - mount MDS with errors=remount-ro, like obdfilter (2009)
5616 - initialize lock handle to avoid ASSERT on error cleanup (3057)
5617 - don't use cancelling-locks' kms values (2947)
5618 - use highest lock extent for kms, not last one (2925)
5619 - don't dereference ERR_PTR() dentry in error handling path (3107)
5620 - fix thread race in portals_debug_dumplog() (3122)
5621 - create lprocfs device entries at setup instead of at attach (1519)
5622 - common AST error handler, don't evict client on completion race (3145)
5623 - zero nameidata in detach_mnt in 2.6 (3118)
5624 - verify d_inode after revalidate_special is valid in 2.6 (3116)
5625 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
5626 - initialize RPC timeout timer earlier for 2.6 (3219)
5627 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
5628 - print nal/nid for unknown nid (3258)
5629 - additional checks for oscc recovery before doing precreate (3284)
5630 - fix ll_extent_lock() error return code for 64-bit systems (3043)
5631 - don't crash in mdc_close for bad permissions on open (3285)
5632 - zero i_rdev for non-device files (3147)
5633 - clear page->private before handing to FS, better assertion (3119)
5634 - tune the read pipeline (3236)
5635 - fix incorrect decref of invalidated dentry (2350)
5636 - provide read-ahead stats and refine rpc in flight stats (3328)
5637 - don't hold journal transaction open across create RPC (3313)
5638 - update atime on MDS at close time (3265)
5639 - close LDAP connection when recovering to avoid server load (3315)
5640 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
5641 - don't leak open file on MDS after open resend (3325)
5642 - serialize filter_precreate and filter_destroy_precreated (3329)
5643 - loop device shouldn't call sync_dev() for nul device (3092)
5644 - clear page cache after eviction (2766)
5645 - resynchronize MDS->OST in background (2824)
5646 - refuse to mount the same filesystem twice on same mountpoint (3394)
5647 - allow llmount to create routes for mounting behind routers (3320)
5648 - push lock cancellation to blocking thread for glimpse ASTs (3409)
5649 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
5650 - fix rare problem with rename on htree directories (3417)
5652 - allow default OST striping configuration per directory (1414)
5653 - fix compilation for qswnal for 2.6 kernels (3125)
5654 - increase maximum number of MDS request buffers for large systems
5655 - change liblustreapi to be useful for external progs like lfsck (3098)
5656 - increase local configuration timeout for slow disks (3353)
5657 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
5659 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5662 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
5663 - don't overwrite extent policy data in reply if lock was blocked (2901)
5664 - drop filter export grants atomically with removal from device (2663)
5665 - del obd_self_export from work_list in class_disconnect_exports (2908)
5666 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
5667 - swab reply message in mdc_close, other PPC fixes (2464)
5668 - fix destroying of named logs (2325)
5669 - overwrite old logs when running lconf --write_conf (2264)
5670 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
5671 - fix race in target_handle_connect (2898)
5672 - mds_reint_create() should take same inode create lock (2926)
5673 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
5674 - don't close files for self_export to avoid uninitialized obd (2936)
5675 - allow MDS with the same name as client node (2939)
5676 - hold dentry reference for closed log files for unlink (2325)
5677 - reserve space for all logs during transactions (2059)
5678 - don't evict page beyond end of stripe extent (2925)
5679 - don't oops on a deleted current working directory (2399)
5680 - handle hard links to targets without a parent properly (2517)
5681 - don't dereference NULL lock when racing during eviction (2867)
5682 - don't grow lock extents when lots of conflicting locks (2919)
5684 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
5687 - account for cache space usage on clients to avoid data loss (974)
5688 - lfsck support in lustre kernel code (2349)
5689 - reduce journal credits needed for BRW writes (2370)
5690 - orphan handling to avoid losing space on client/server crashes
5691 - ptlrpcd can be blocked, stopping ALL progress (2477)
5692 - use lock value blocks to assist in proper KMS, faster stat (1021)
5693 - takes i_sem instead of DLM locks internally on obdfilter (2720)
5694 - recovery for initial connections (2355)
5695 - fixes for mds_cleanup_orphans (1934)
5696 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
5697 - block all file creations until orphan recovery completes (1901)
5698 - client remove rq_connection from request struct (2423)
5699 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
5700 - recovery timer race (2670)
5701 - mdc_close recovey bug (2532)
5702 - ptlrpc cleanup bug (2710)
5703 - mds timeout on local locks (2588)
5704 - namespace lock held during RPCs (2431)
5705 - handle interrupted sync write properly (2503)
5706 - don't try to handle a message that hasn't been replied to (2699)
5707 - client assert failure during cleanup after abort recovery (2701)
5708 - leak mdc device after failed mount (2712)
5709 - ptlrpc_check_set allows timedout requests to complete (2714)
5710 - wait for inflight reqs when ptlrpcd finishes (2710)
5711 - make sure unregistered services are removed from the srv_list
5712 - reset bulk XID's when resending them (caught by 1138 test)
5713 - unregister_bulk after timeout
5714 - fix lconf error (2694)
5715 - handle write after unfinished setstripe, stripe-only getstripe (2388)
5716 - readahead locks pages, leaves pending causing memory pressure (2673)
5717 - increase OST request buffers to 4096 on large machines (2729)
5718 - fix up permission of existing directories in simple_mkdir (2661)
5719 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
5720 - don't assign transno to errored transactions (2742)
5721 - don't delete objects on OST if given a bogus objid from MDS (2751)
5722 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
5723 - if rq_no_resend, then timeout request after recovery (2432)
5724 - fix MDS llog_logid record size, 64-bit array alignment (2733)
5725 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
5726 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
5727 - MDS recovery shouldn't delete objects that it has given out (2730)
5728 - if enqueue arrives after completion, don't clobber LVB (2819)
5729 - don't unlock pages twice when trigger_group_io returns error (2814)
5730 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
5731 - don't write pages to disk if there was an error (1450)
5732 - don't ping imports that have recovery disabled (2676)
5733 - take buffered bytes into account when balancing socknal conn (2817)
5734 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
5735 - reconnect unlink llog connection after MDS reconnects to OST (2816)
5736 - remove little-endian swabbing of llog records (1987)
5737 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
5738 - retry reposting request buffers if they fail (1191)
5739 - grow extent at grant time to avoid granting a revoked lock (2809)
5740 - lock revoke doesn't evict page if covered by a second lock (2765)
5741 - disable VM readahead to avoid reading outside lock extents (2805)
5743 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
5744 - updated kernel patches for hp-2.4.20 kernel (2681)
5746 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
5749 - fix truncated write corruption (2366)
5750 - fix for failed assertion in iopen_connect_dentry (1792,2517)
5752 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
5753 - ensure len(uuid) < 37 in lmc (1171)
5754 - fix ia64 OOPS in llog_test (2255)
5755 - zero end of page at obdfilter for partial page writes (2648)
5756 - don't leave stale dentries around after renames (bug 2428)
5757 - fix timeouts when evicting a client with a single lock held (2642)
5758 - set deadline for the initial HELLO message to drain (2634)
5759 - print out dotted-quad IP addresses in the socknal (2302)
5761 - additional debugging for MDS client eviction problem (2443)
5762 - fix mkfsoptions support for osts (2603, 2604)
5764 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
5767 - add series for the vanilla 2.6.0 kernel
5768 - add series for the vanilla 2.4.24 kernel
5769 - add series for a cray x86/64 UL kernel drop
5770 - fix xattr patches for the vanilla 2.4.19 series
5772 - generate true UUIDs in lmc (1171)
5773 - have portals stack dumping break in UML (2466)
5774 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
5775 - allocate with _NOFS in ldlm to avoid deadlock (1933)
5776 - wake callback waiting threads on client eviction (2460)
5777 - Add --ptldebug and --subsystem to lmc (1719)
5778 - update assertion to allow safe interrupt allocation
5779 - set rq_no_resend for cancel requests (2432)
5780 - recalculate ptlrpcd timeout after resend (2494)
5781 - call vfs_rmdir when removing pending directories (2368)
5782 - fix renaming a file to itself (2429)
5783 - lmc creates a default one-stripe lov (2454)
5784 - expand procfs space to handle large clusters (2326)
5785 - increase UML stack to avoid overflow
5786 - update lconf's list of debug and subsystem masks
5787 - fix lfs find --obd (2510)
5788 - /proc tunable for disabling filter read caching (2591)
5789 - stop rpm packages from altering slapd.conf (2301)
5790 - disable nagle in the socknal under 0conf (2578)
5791 - choose mds inode size based on stripe count (2572)
5792 - fix kernel-source rpm problems (2516)
5794 - add --disable-doc to avoid pdf generation (2421)
5795 - update documentation, tests, type-os, comments
5796 - avoid format warnings on ia64
5797 - remove the TOE NAL
5798 - tiny code cleanups by removing unused fields
5800 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
5803 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
5804 - remove the most blatant lies from BUILDING file (2371)
5805 - change default debug level to reasonable production setting
5806 - reduce client side cache size to reduce cache flush time
5807 - reduce max RPCs in flight to avoid unnecessary file fragmentation
5808 - make TCP zerocopy and pinger support enabled by default (2476)
5809 - sync writes completed after process exits caused crashes (2319)
5810 - maintain correct mount count on the MDS (2356)
5811 - backout 1557, because 2316 wasn't really fixed
5812 - better file I/O statistics gathering in /proc
5813 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
5814 - another kernel patch to fix zero-copy TCP function export
5815 - don't take duplicate lock when processing re-sent getattr (2420)
5816 - lctl uses obd_self_export instead of creating new conn (2353)
5817 - MDS/OST recovery case which requires object creation asserted (2425)
5818 - move lfs from /usr/sbin to /usr/bin in packages
5819 - fix race between mds_client_add and mds_client_free (2417)
5820 - use kmalloc instead of slabs in portals (2430)
5821 - don't create duplicate records when a failover MDS is present (2442)
5822 - remove unnecessary mount age check (2332)
5823 - don't remove directory inodes from locks prematurely (2451)
5824 - don't break if MDS service name is the same as hostname (2103)
5825 - fix races in client write RPC generation when cache full (2482)
5827 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
5830 - remove now-unused request->rq_obd (278)
5831 - if an allocation fails, print out how much memory we've used (1933)
5832 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
5833 - add the "configurable stack size" patch to most series files (1256)
5834 - ability to write large log records, for 100+ OST configs (2306)
5835 - fix NULL deref when filter_prep fails (2314)
5836 - fix operator precedence error in filter_sync
5837 - dynamic allocation of socknal TX descriptors (2315)
5838 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
5839 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
5840 - ensure that CWARN messages in Portals always get to the syslog
5841 - __init/__exit are not for prototype decls (ldlm_init/exit)
5842 - x86-64 compile warning fixes
5843 - fix gateway LMC keyword conflict (2318)
5844 - fix MDS lock inversions in getattr/reint paths (1844)
5845 - fix a rare lock re-ordering bug, which caused deadlock (2322)
5846 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
5847 - DLM race condition prevented some lock evictions (2328)
5848 - ENOMEM detection and retry on socknal sends (2230)
5849 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
5850 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
5851 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
5852 - better error messages when a client is rejected during recovery (1505)
5853 - avoid cancelling locks which were never granted, after failure (2330)
5854 - fix i_sem/journal inversion in mds_client_add (2333)
5855 - fix truncate/getattr lock cycle deadlock (2334)
5856 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
5857 - fix two rare exit paths which could leak an l_lock() ref (2321)
5858 - fencepost error in MDS/OST orphan recovery (2226)
5859 - make log record alignment 8 bytes (1988)
5860 - lstripe now fails when requested offset > ost_count (2237)
5861 - ensure that all kernel series have a complete list.h (1607)
5862 - fix crashes in special-file operations (2316)
5863 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
5865 - allow configurable automake binary, for testing new versions
5866 - small update to the lfs documentation
5868 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
5870 * fix negative export reference count in fsfilt_sync (2312)
5872 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
5873 * release candidate 0.9.1
5875 - orphans are moved into the PENDING directory for possible recovery
5876 - replayed opens now open by fid for orphan/rename safety (1042)
5877 - last close of an orphan inode generates a transno (683)
5878 - chdir() and mount() now pin the directory entry (1020)
5879 - avoid CERROR in normal ll_setattr_raw() error case (1500)
5880 - discard very old requests without processing them (1502)
5881 - remove some common, well-understood CERRORs (1505)
5882 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
5883 - clear "grant" flags in OST replies until OST grant code lands (1644)
5884 - fix read performance by not clobbering i_blksize on client (1598)
5885 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
5886 - make LRU size a /proc tunable, clears locks when reduced (707)
5887 - fix some lprocfs rot that prevented ptlbd from loading (1732)
5888 - server locks take references on exports now (1558)
5889 - build fixes for 2.4.20-rh trees (1663)
5890 - return an error from lov_create if all OSCs are inactive (1751)
5891 - fix import levels when a reconnect happens without a timeout (1597)
5892 - exit early from mds_open if we get a lookup error (1749)
5893 - partial page read at EOF wouldn't wait for disk before sending (1642)
5894 - avoid NULL deref in obdfilter when reading page past EOF (1592)
5895 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
5896 - fix LBUG in ll_it_open_error with rc = -2 (1861)
5897 - write/truncate lock inversion (1639)
5898 - Don't auto-load obdclass, portals modules during cleanup (1495)
5899 - fix timestamps from jumping to "now" (1763)
5900 - extra journal assertions (1648)
5901 - add an extra multiunlink test (1771)
5902 - fix read_record/write_record API (1776)
5903 - fix leak of offset_extent, possible incorrect i_size later (1772)
5904 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
5905 - Debugging for the kqswnal_get_idle_tx problems (1820)
5906 - Allow recovery to be attempted multiple times (1536)
5907 - Write out MDS last_rcvd file after it is first created (1600)
5908 - Fix tx_descriptor leak in failed transmit situations (1827)
5909 - ext3 journaling fixes for assertion failure after IO error (1871)
5910 - class_export_put() on freed export after completion AST error (1896)
5911 - Fix revalidate looping in VFS (1322)
5912 - Don't access a freed export during MDS_REINT timeout (1521)
5913 - Add open-unlink recovery support on the MDS (1673,1764)
5914 - Return an error if no MDS data was read from last_rcvd (1946)
5915 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
5916 - Don't setup a disk device that doesn't match exported UUID (317)
5917 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
5918 - avoid committing NULL handle in force close
5919 - local.sh is now a one-stripe LOV configuration
5920 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
5921 - fix ext3 htree duplicate directory entry corruption (1516)
5922 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
5923 - update to new LOV EA format (2097)
5924 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
5926 2003-06-15 Phil Schwan <phil@clusterfs.com>
5929 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
5930 - per-import/export recovery handling (958, 931, 959)
5931 - multiple last-rcvd slots, for serving multiple FSes (949)
5932 - connections are again shared between multiple imp/exports (963, 964)
5933 - "umount -f" would hang if any requests needed to be sent (393, 978)
5934 - avoid pinning large req buffer by copying for queued messages (989)
5935 - add "uuid" to "lctl device" command to help upcalls (991)
5936 - "open" RPCs with transnos would confuse recovery counters (1037)
5937 - do proper endian conversion of all wire messages (288, 340, 891)
5938 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
5939 - call waiting locks callback from LDLM recovery thread (1127, 1151)
5940 - fix ptlrpc_connection leak in target_handle_connect (1174)
5941 - fix import refcounting bug in OST and MDS cleanup (1134)
5942 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
5943 - fix very unlikely obd_types race condition (501)
5944 - remove osc_open hack for echo_client (1187)
5945 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
5946 - a failure in read_inode2 leads to deadlock (1139)
5947 - cancel ack-locks as soon as transaction is committed (1072)
5948 - fix major leaks and crashes in the bulk I/O path (937, 1057)
5949 - make sure to commitrw after any preprw to avoid deadlock (1162)
5950 - failing to execute a file in a lustre FS would lock inode (1203)
5951 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
5952 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
5953 - obd_uuid2tgt would walk off the end of the list (1255)
5954 - on IA64 the osc would give portals incorrect bulk size (1258)
5955 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
5956 - fix lock inversion caused by new llite matching code (1282)
5957 - limit the number of dirty pages on a client to 10MB (1286)
5958 - timed out locks were not being corrected cancelled (1289)
5959 - fix O_DIRECT above 4GB on IA-32 (1292)
5960 * major user-visible changes
5961 - fail out/fail over policy now controlled by the upcall (993)
5963 - add OBD_PING to check server availability and failure (954)
5964 - lustre messages are now sent in sending host order (288, 340, 891)
5965 - add eadatalen to MDS getattr reply (340)
5966 - OST read replies may contain second buffer, with per-page status (593)
5968 2003-03-11 Phil Schwan <phil@clusterfs.com>
5971 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
5972 - failed open()s could cause deadlock; fixed (867, 869)
5973 - stop cancelling OST locks when files are closed (481)
5974 - overlapping XID spaces caused network corruption (851, 853)
5975 - fix unsafe fsfilt counter arithmetic; change to atomic_t
5976 - setattr_raw added, to do single-RPC, server-side setattrs
5977 - lmc/lconf syntax change for OST UUIDs
5978 - fix crashy race condition between ptlrpc_free_req and osc_close
5979 - don't use request in mdc_enqueue if we hit a timeout (889)
5980 - don't set the inode i_size for regular files from the MDS (896)
5981 - handle out of order completion AST (842)
5982 - don't LBUG if a lock request times out after receiving AST (913)
5983 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
5984 - if a bad lock AST arrives, send an error instead of dropping entirely
5985 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
5986 - fix leak in bulk IO when only partially completed (899, 900, 926)
5987 - fix O_DIRECT for ia64 (55)
5988 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
5989 - C-z after timeout could hang a process forever; fixed (977)
5991 - client-side I/O cache (678, 924, 929, 941, 970)
5993 - READPAGE and SETATTRs which don't take server-side locks get
5996 2003-02-11 Phil Schwan <phil@clusterfs.com>
5999 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
6000 - Fix fsfilt_extN_readpage() to read a full page of directory
6001 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
6002 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
6003 - Fix imbalanced LOV object allocation and out-of-bound access (469)
6004 - Most intent operations were removed, in favour of a new RPC mode
6005 that does a single RPC to the server and bypasses most of the VFS
6006 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
6007 - Aggressively cancel local locks on DLM servers
6008 - mds_reint_unlink sends EA to the client if it's the last nlink.
6009 client uses that EA to unlink OST objects.
6010 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
6011 - recursive symlinks were fixed (439)
6012 - fixed NULL deref in DEBUG_REQ
6013 - filter_update_lastobjid no longer calls sync, which annoyed extN
6014 - fixed multi-client small-writes to a single file problem (445)
6015 - fixed mtime updates during file writes (607)
6016 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
6017 - fixed bug in obd_brw_read/write() (under guise of testing 367)
6018 - fixed Linux OST size reporting problem (444, 656)
6019 - OST now updates object mtime with writes or setattr (607, 619)
6020 - client verifies file size before zeroing page past EOF (445)
6021 - OST now writes last allocated objid to disk with allocation (108)
6022 - LOV on echo now works (409)
6024 - mds_reint_unlink sends a new buffer, with the EA included. this
6025 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
6026 set if a regular file was being unlinked, and it was the last link
6027 - use PtlGet from the target for bulk writes (315)
6028 - OST now updates object mtime with writes or setattr (607, 619)
6029 - LDLM now has a grant-time callback to revalidate locked items, if
6031 - Many MDS operations were reorganized to combat race conditions
6033 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
6034 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
6036 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
6039 - Fully reactivate OST imports after reconnection (512, others)
6040 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
6041 - More graceful error handling for truncating on dead OST (515)
6042 - Don't error out unless we're actually accessing dead stripes (474)
6043 - Fix garbage sizes when stripes are missing (410)
6044 - LRU counters were broken, causing constant lock purge (433, 432)
6045 - garbage on read from stripes with failed OSTs (441)
6046 - mark OSCs as active before reconnecting during recovery (438)
6047 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
6048 - lfind did not preserve OST order in output (443)
6049 - symlinks cause hung clients, incorrect data (439)
6050 - stop dereferencing request after dropping refcount (457)
6051 - don't LASSERT(spin_is_locked) on non-SMP (455)
6052 - fixes for many rename() bugs
6053 - fstat didn't correctly synchronize attributes (399)
6054 - server must handle lock cancellation during blocking AST prep (487)
6055 - bulk descriptors were free()d too soon (511)
6056 - fix paths in lconf, which would load incorrect modules (451, 507)
6057 - fix confusing lconf 'host not found' error message (386)
6058 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
6059 - fix race condition in mdc_blocking_ast() for inode access (526)
6060 - fix lov_unpackmd() unpacking wrong number of stripes (537)
6061 - fix lov_set_osc_active() marking wrong OSC inactive (440)
6062 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
6063 - fix multiple writes of stripe MD to MDS (358, maybe 519)
6064 - fix lstripe in several ways (kernel side) (527)
6065 - fix request leak in ldlm_cli_enqueue (262)
6066 - incorrect OSC was marked inactive after OST failure
6067 - call mds_fs_cleanup before unmounting filesystem (524)
6068 - fix races between taking ns_lock and ldlm_lock_change_resource
6069 - fix races updating LOV export open file list
6070 - fix lov_enqueue error path, avoid decref-ing bad lock handle
6071 - fix recovery NULL deref in ldlm_cli_cancel_unused
6072 - fix some DLM races by using new hash table for lock handles (419)
6073 - permit the client to specify desired inodes, at replay
6074 - duplicate requests when we queue them for replay reintegration
6075 - fix last_rcvd offset calculation
6076 - sync after each recovered transaction, so we always make progress
6077 - never, not always, ERESTART requests without transnos
6078 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
6080 - skip replay if the MDS says that the client is already connected
6081 - don't check for a recovery-enabled export to match lctl's UUID
6082 - don't INC_USE_COUNT for phantom exports
6083 - don't crash when cleaning up phantom exports (567)
6084 - don't double-finish or set replay data for errored mdc_open requests
6085 - abort requests when they time out, so we don't get old replies
6086 - send/receive replies for AST messages again
6087 - if the client says that it doesn't have the lock, cancel it on the
6089 - if we timeout during I/O, don't try to cancel an in-use lock; instead
6090 mark it as destroyed, it will all work out when decref is called
6091 - fix module use counts (22, 581)
6093 - ASTs now expect a reply (server cancels lock on error reply)
6095 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
6098 - fix many simultaneous client startup (392)
6099 - fix dentry->d_it clobbering
6100 - credentials weren't being shipped for readdir/getattr operations
6101 - remove invalid assertions triggered during some concurrent MD
6103 - proper Lustre versions added (336, 389)
6104 - fix memory leak for create error case (398)
6105 - fix LOV locking bug that would get cli/srv out of sync
6106 - fix echo client over LOV (409)
6107 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
6108 - fix double-O_EXCL intent crash (424)
6109 - avoid sending multiple lock CANCELs (352)
6111 - MDS can do multi-client recovery (modulo bugs in new code)
6113 - many updates, edits, cleanups
6115 2002-11-18 Phil Schwan <phil@clusterfs.com>
6118 - fix null d_it dereference (346)
6119 - fix full OST/dbench hang (333)
6120 - fix permission problem with file removal (286)
6121 - fix removal of OSCs from LOV when they fail
6122 - fix NULL deref during bulk timeout (214)
6123 - fix problems related to multiple filesystems on one MDS (241)
6124 - fixed serious subtle metadata locking bugs
6125 - free locks on clients when inodes are removed due to memory
6127 - fix inode pointer in lock data (285)
6128 - partial support for multiple MDS on a single host (241)
6129 - data locks weren't cancelled at clear_inode time (290, 311)
6130 - intent locks could lead to unbounded lock growth (205)
6131 - added a maximum lock count, an LRU list, and a flusher
6132 - fix multiple rename (365)
6133 - properly abstracted the echo client
6134 - OSC locked 1 byte too many; fixed
6135 - rewrote brw callback code:
6136 - fixed recovery bugs related to LOVs (306)
6137 - fixed too-many-pages-in-one-write crash (191)
6138 - fixed (again) crash in sync_io_timeout (214)
6139 - probably fixed callback-related race (385)
6141 - Add capability to MDS protocol
6142 - LDLM cancellations and callbacks on different portals
6144 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
6147 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
6148 - defer open object destruction to close time (601981, 138)
6149 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
6150 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
6151 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
6152 - don't call obd_setattr in ll_file_release for destroyed objects
6153 * protocol change to lustre_msg: move |version| and add |flags|
6154 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
6155 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
6156 * added replay of create, unlink, link and rename operations during
6157 MDS failover; recovery should be much more robust now
6158 * remove failed OSCs from LOVs (only lov_create uses this so far)
6159 * the lustre-HOWTO was brought (more) up to date (582544)
6161 2002-10-23 Phil Schwan <phil@clusterfs.com>
6164 - in-use dentries weren't being reused properly (617851)
6165 - prevent multiple LDLM setup (599178)
6166 - fix LOV size calculations for truncate (617853)
6167 - fix client handling of MDS intent errors (POSIX)
6168 - fix permission bug in lovstripe.c test (624321)
6169 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
6170 - truncate past end of file could corrupt data
6171 - proper cleanup after timeouts, crashes, etc (592524, 550815)
6172 - a race in recovery could return ETIMEDOUT to apps (623947)
6173 - building outside the source directory was fixed
6174 * the lustre-HOWTO was brought (more) up to date (582544)
6175 * major progress was made on recovery functionality
6177 2002-10-10 Phil Schwan <phil@clusterfs.com>
6180 - recovery deadlock fix
6181 - rm -rf causes LBUG fix (617817)
6182 - file open by multiple tasks fix (618962)
6183 - directory permissions bugs (602707 and 620007)
6184 - journal_stop fixed with locking (611313)
6185 - O_APPEND failures resolved (618273, perhaps 614459)
6186 - lconf PATH fix (619770)
6187 - IA64 build fix (621450)
6188 - RPC buffer sizes scale with amount of memory
6190 2002-10-01 Phil Schwan <phil@clusterfs.com>
6193 - locks would be cancelled without throwing away data pages,
6194 resulting in inconsistent data (605627)
6195 - inode attributes were not always being refreshed (605627, 612449)
6196 - lconf now continues to cleanup after lctl reports an error
6197 - MDS now enforces user permissions (602707)
6198 - lprocfs cleanup fixed, but not yet enabled (614157)
6199 - fixed infinite server hang, should a client not respond to an AST
6200 - avoid going into recovery if user calls readlink() with a buffer
6201 that's too small (613941)
6202 - AST RPCs no longer require replies (614867) -- this may be changed
6203 - don't crash server if client sends an IOV that's too big (611336)
6204 - fixed lock conversion deadlock (611892)
6205 - fixed the following of symlinks (614622)
6206 * recovery: the server can remove locks from a client that dies, other
6207 clients can make progress
6208 * more extN patch fixes
6209 * compile-time configurable ptlrpc buffer allocations
6211 - collaborative read cache document
6212 - Lustre Lite Performance CDR document-in-progress
6214 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6217 - fix typo in patch-2.4.18
6219 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6222 - clear ptlrpc request each time in handle_incoming_request()
6223 - unlink of files now destroys the object on the OST
6225 2002-09-19 Peter Braam <braam@clusterfs.com>
6227 * add hard link support
6228 * change obdfile creation method
6229 * kernel patch changed
6231 2002-09-19 Peter Braam <braam@clusterfs.com>
6234 - stack overflow bug in extN fixed
6236 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
6238 * documentation updates
6239 - add man pages for config tools
6240 - update tests/README to describe testing with new config tools
6241 - finish metadata API descriptions
6242 * bug fixes and cleanups
6243 - statfs workaround for 16TB limit
6244 - LOV stripe allocation improved, can stripe on subset of OSTs
6245 - LOV file size/IO offset was wrong for files > 4GB in size
6246 - object EA data was being dropped, caused files to be unreadable
6247 - memory overflow with non-LOV OST caused memory corruption
6248 - fixed regression tests to work with new config tools, obdfilter
6249 - fixed bug when directory size became larger than 1 block
6250 - fixed bug (for single client case) when PWD was deleted
6251 - invalidate local directory pages when doing intent-based ops
6252 - avoid LDLM oops when lock callback contained bad data
6254 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
6256 * documentation updates
6257 * bug fixes and cleanups
6258 - configuration tools
6261 - 64-bit compile warnings
6262 - 64-bit internal statfs data
6264 * test_brw on persistent OST devices
6266 * lprocfs (disabled)
6268 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
6270 * documentation updates
6271 * bug fixes and cleanups
6272 * configuration tools
6274 2002-08-30 Peter J. Braam <braam@clusterfs.com>
6277 * many small fixes to 0_5_4
6278 * io/network handling
6279 * thinkos in MDS operations
6281 2002-08-24 Peter J. Braam <braam@clusterfs.com>
6284 * crucial basic fixes to 0.5.3
6285 * IOR, Iozone work over Elan
6288 2002-08-07 Phil Schwan <phil@clusterfs.com>
6289 * version 0_5_3, our first alpha
6290 * we use the new Portals iovs
6291 * documentation updates
6292 * bug fixes and cleanups
6293 * small changes in the DLM wire protocol
6295 2002-07-25 Peter J. Braam <braam@clusterfs.com>
6296 * version 0_5_1 with some initial stability,
6297 * locking on MD and file I/O.
6298 * documentation updates
6299 * several bug fixes since 0.5.0
6300 * small changes in wire protocol
6302 2002-07-18 Phil Schwan <phil@clusterfs.com>
6304 * delivered as Lustre Light Alpha
6305 * fixed a crash after handling invalid MDS requests
6306 * fixed directory pages for architectures with non-4k pages sizes
6308 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
6309 * release version v0_4_4
6310 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
6312 * Fixes a number of interruption problems with OST operations.
6313 * Update documentation for portals header changes
6314 * Move all wire protocol structs/defines to lustre_idl.h
6315 * Fixes symlink length bug.
6316 * Add tcpdump to repository.
6318 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
6319 * release version v0_4_3
6320 * Fixes statfs for inodes on extN.
6321 * Fixes bug in runtests which would delete /etc/hosts.
6322 * Use 64-bit object IDs wherever possible (not into VFS though)
6323 Remove ost_get_info, which is unused by lustre, and out of date.
6325 2002-07-03 Peter Braam <braam@clusterfs.com>
6326 * release version v0_4_2 Fixes a lookup error (type not passed)
6327 * move forward to head of Portals
6328 * move forward to latest Lustre kernel
6330 2002-06-25 Peter Braam <braam@clusterfs.com>
6331 * release version v0_4_1. Hopefully stable on single node use.