1 06-11-2011 Whamcloud, Inc.
3 * Server support for kernels:
4 2.6.18-308.4.1.el5 (RHEL5)
5 2.6.32-220.17.1.el6 (RHEL6)
6 * Client support for unpatched kernels:
7 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
8 2.6.18-308.4.1.el5 (RHEL5)
9 2.6.32-220.13.1.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: Kernel update [RHEL6.2 2.6.32-220.17.1]
24 Severity : enhancement
26 Description: Kernel update [RHEL5.8 2.6.18-308.4.1.el5]
28 Severity : enhancement
30 Description: Kernel update [RHEL6.2 2.6.32-220.13.1]
32 Severity : enhancement
34 Description: Kernel update [RHEL6.2 2.6.32-220.4.2]
36 Severity : enhancement
38 Description: Kernel update [RHEL5.7 2.6.18-274.18.1.el5]
40 Severity : enhancement
42 Description: Update RHEL6.2 kernel to 2.6.32-220.el6
44 Severity : enhancement
46 Description: Kernel update [RHEL5.7 2.6.18-274.12.1.el5]
48 Severity : enhancement
49 Description: Update RHEL6.1 kernel to 2.6.32-131.6.1.el6
51 Severity : enhancement
52 Description: Update RHEL5.6 kernel to 2.6.18-238.19.1.el5
54 Severity : enhancement
55 Description: Update RHEL5.6 kernel to 2.6.18-238.9.1.el5
58 Frequency : only with SLES10
60 Description: Use OFED "KMP" provided by Novell
61 Details : SLES10 SP3 ships with OFED in a separate "KMP" package.
62 Lustre is now built against this package. That means you need to
63 install the ofed-kmp package from Novell for the patchless client.
64 Note that the ofed-kmp that Novell ships may not exactly match the
65 kernel version but should still be compatible.
67 Severity : enhancement
69 Description: Update RHEL5.5 kernel to 2.6.18-194.17.1.el5
70 Update OEL5.5 kernel to 2.6.18-194.17.1.0.1.el5.
72 Severity : enhancement
74 Description: Update RHEL6.2 kernel to 2.6.32-220.7.1.el6
76 --------------------------------------------------------------------------------
78 2010-07-15 Oracle, Inc.
80 * Server support for kernels:
81 2.6.18-164.11.1.el5 (RHEL 5)
82 2.6.18-164.11.1.0.1.el5 (OEL 5)
83 * Client support for unpatched kernels:
84 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
85 2.6.18-164.11.1.el5 (RHEL 5),
86 2.6.18-164.11.1.0.1.el5 (OEL 5)
87 2.6.16.60-0.42.8 (SLES 10),
89 2.6.29.4-167.fc11 (FC11)
90 * Recommended e2fsprogs version: 1.41.10-sun2
92 Severity : enhancement
94 Description: Distribute lfs_migrate script with lustre
95 Details : The lfs_migrate mentioned in the manual is now included in the
96 lustre tree in lustre/scripts and will distributed in lustre RPMs.
98 Severity : enhancement
100 Description: Failover nid list added to import proc file.
102 Severity : enhancement
104 Description: "som_preview" mount option required to use SOM
105 Details : The clients will need to use som_preview mount option
109 Description: Lustre does not do 1MB IOs to HW RAID
110 Details : Bump MAX_PHYS/HW_SEGMENTS and SG_ALL to 256 in the RHEL5 kernel.
112 Severity : enhancement
114 Description: lfs quota output is not very convenient for awk/sed-parsing
115 Details : Some positions in lfs quota output table could be empty or
116 non-empty which made it hard to parse it with scripts, now
117 a dash is put instead of space where there is not supposed
122 Description: Update to OFED 1.5.1
123 Details : OFED 1.5.1 will be used in environments that don't use
124 vendor supplied OFED.
128 Description: Remove lnet. prefix from lctl params display
129 Details : Also renamed lnet.memused to lnet_memused to
134 Description: Use conf_param -d to remove permanant settings.
138 Description: Add list_param -R
139 Details : lctl list_param now supports recursive mode.
142 Bugzilla : 21829, 22312
143 Description: Fix llobdstat.
144 Details : 1. Now updating the write_bytes counter
145 2. Port script fixes from 1.8.
149 Description: decrease the usage of memory on clients.
150 Details : 1. On clients, recycle dentries and inodes unused.
151 2. Delete the code related to ll_deathrow(att 6215 in bug 1443). It
156 Description: llobdstat fix and enhancement
157 Details : add a counter to set a limit to how many samples will be returned
158 fix a wildcard in the path to limit to obdfilter stats only
162 Description: lfs check only allowed for root
163 Details : Code cleanups and added a regresion test of non-root use of
168 Description: MDS Oopses in ptlrpc_dispatch_difficult_reply()/__list_add()
169 Details : Fix hr->hr_index overload.
173 Description: Fix FIEMAP functionality.
177 Description: FC11 patchless client support added
181 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and
182 OEL5.4 kernel to 2.6.18-164.11.1.0.1.el5.
186 Description: New acc-sm script and various test-framework enhancments.
190 Description: Previously landed patch was reverted.
194 Description: Update SLES10 kernel to 2.6.16.60-0.42.8 (patchless)
198 Description: Add ldiskfs maxdirsize mount option.
202 Description: Fix mkfs.lustre support for 16TB LUNs.
206 Description: Print mmp_check_interval
207 Details : print mmp_check_interval and make it possible to abort
208 mount operation in case it takes too long.
212 Description: Add cascading_rw to lustre mpi tests.
216 Description: Store SOM data in EA.
217 Details : Also several SOM related cleanups.
221 Description: Add refcount to md_open_data
225 Description: Use refcount on lov_request_set to prevent memory corruption.
229 Description: Statahead: use parent dir's i_mutex for synchronization.
233 Description: loadgen improvements.
237 Description: Enable peer health detection in ksocklnd and ko2iblnd.
241 Description: Control DCACHE_LUSTRE_INVALID flag with MDS_INODE_LOOKCUP lock
245 Description: Fix race on obd_devcie:md_stats
249 Description: Fix SLES 11 client builds.
250 Details : Resolved uninitalized pointer gcc warnings.
254 Description: Allow non-root access for "lfs check".
255 Details : Added a check in obd_class_ioctl() for OBD_IOC_PING_TARGET.
259 Description: Clear imp_force_reconnect correctly in ptlrpc_connect_interpret()
261 Severity : enhancement
263 Description: Update kernel to RHEL5.4 2.6.18-164.6.1.el5 and
264 OEL5 2.6.18-164.6.1.0.1.el5(Both in-kernel OFED enabled).
268 Description: Sanity test_27p fails.
269 Details : OBD_FAIL_OST_ENOSPC has been moved to filter_statfs().
270 Sanity tests 27 were cleaned up and improved.
274 Description: Unify req format on client/servers
275 Details : Use new req_capsule API [almost] everywhere instead of old PTLRPC
276 buffers and swabbers approach..
281 Description: drop join file stuff.
283 Severity : enhancement
285 Description: Add 2.6.27 and SLES11 (patchless client) support
289 Description: Fix in ptlrpc_expire_one_request() to print the signed time
292 Severity : enhancement
294 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
297 Severity : enhancement
299 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
301 Severity : enhancement
303 Description: correctly handle big reply message.
304 Details : send LNet event if reply is bigger then buffer and adjust this buffer
309 Description: Drop unnecessary __GFP_NOMEMALLOC flag from filter_get_page()
311 Severity : enhancement
313 Description: Add support for OEL5.
317 Description: Conf-sanity.sh 50g test - deactivated OST should not cause a panic.
319 Severity : enhancement
321 Description: "weak-modules" support
322 Details : Implement "weak-modules" support which enables kernel modules
323 to be used with any kernel that implements the same kABI. In
324 order to achieve this modules are now installed in
325 /lib/modules/$(uname -r)/updates/kernel on all distributions.
329 Description: client could not reconnect to OST because of an active request.
330 Details : The client now retries to reconnect to the same server, if a
331 connect request failed with EBUSY or -EAGAIN. Test case 35b
332 conf-sanity.sh has been created.
336 Description: Don't allow make backward step on assignin osc next id.
337 Details : race between allocation next id and ll_sync thread can be cause
338 of set wrong osc next id and can be kill valid ost objects.
342 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
346 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
350 Description: Error handling in osc_statfs_interpret() has been improved.
351 Details : Check in osc_statfs_interpret() for EBADR.
355 Description: Do not send statfs() requests to OSTs disabled by administrator.
356 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
358 Severity : enhancement
360 Description: Added sanity-quota test to for lfs(1) to work with numeric uid/gid.
362 Severity : enhancement
364 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
366 Severity : enhancement
368 Description: Enhancement for lfs(1) command to use numeric uid/gid.
372 Description: Improved cleanup procedure for conf-sanity test_46a.
376 Description: Do not start bulk transfer, if client tried to reconnect.
377 Details : Do not start bulk transfer, if client tried to reconnect,
378 but failed due to active rpcs. Check for exp_abort_active_req
379 has been implemented.
383 Description: Prevent inconsistences between linux and lustre mount structures.
384 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
385 Make the sleep interruptible.
387 Severity : enhancement
389 Description: provide server to client comms path
390 Details : server to client communications path via new LDLM_SET_INFO rpc
394 Description: truncate starts GFP_FS allocation under transaction causing deadlock
395 Details : ldiskfs_truncate calls grab_cache_page which may start page
396 allocation under an open transaction. This may lead to
397 calling prune_icache with consequent lustre reentrance.
401 Description: send correctly lsm on open replay
402 Details : MDS is trust to LSM size on replay open, but client can set wrong size
407 Description: enable client interface failover
408 Details : When a child reconnects from another NID, properly update export
409 nid hash position and ldlm reverse import.
411 Severity : enhancement
413 Description: Communicate OST degraded/readonly state via statfs to MDS
414 Details : Flags in the statfs returned from OSTs indicate whether the
415 OST is in a degraded RAID state, or if the filesystem has
416 turned read-only after a filesystem error is detected.
421 Description: don't panic if EPROTO was hit when reading symlink
422 Details : correctly handling request reference in error cases.
424 Severity : enhancement
426 Description: Add LustreNetLink, a kernel-userspace communcation path.
428 Severity : enhancement
430 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
435 Description: access to llog context before init.
436 Details : move handling CATALOGS file at osc layer and forbid access to llog
440 Frequency : rare, only when client is writing to OST on local node
442 Description: Avoid deadlock for local client writes
443 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
444 memory freeing context. This allows OST threads to set the
445 PF_MEMALLOC flag on task structures in order to allocate memory
446 from reserved pools and complete IO.
447 Use GFP_HIGHUSER for OST allocations for non-local client writes,
448 so that the OST threads generate memory pressure and allow
449 inactive pages to be reclaimed.
451 Severity : enhancement
453 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
458 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
459 Details : move ldlm namespace creation in setup phase to avoid grab
460 _lprocfs_lock with cli_sem held.
464 Description: Temporarily disable grant shrink.
465 Details : Disable the feature for debugging.
468 Frequency : only when formatting test filesystems
470 Description: Unable to run several mkfs.lustre on loop devices at the same time
471 Details : mkfs.lustre returns error 256 on the concurrent loop devices
472 formatting. The solution is to proper handle the error.
474 Severity : enhancement
476 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
478 Severity : enhancement
480 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
485 Description: resolve race between obd_disconnect and class_disconnect_exports
486 Details : if obd_disconnect will be called to already disconnected export he
487 forget release one reference and osc module can't unloaded.
489 Severity : enhancement
491 Description: Allow tuning service thread via /proc
492 Details : For each service a new
493 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
494 created that can be used to set min/max thread counts, and get the
495 current number of running threads.
500 Descriptoin: don't return error if have partially created objects for file.
501 Details : lov_update_create_set uses set->set_success as index for created
502 objects, so if some requests will be failed, they will have hole
503 at end of array and qos_shrink_lsm for allocate correct lsm.
505 Severity : enhancement
507 Description: Update OFED support to 1.4.1
512 Description: Reduce small size read RPC
513 Details : Set read-ahead limit for every file and only do read-ahead when
514 available read-ahead pages are bigger than 1M to avoid small size
517 Severity : enhancement
519 Description: move AT tunable parameters for more consistent usage
520 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
522 Severity : enhancement
524 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
525 Details : allow skip disconnected ost for send statfs request and hide error
531 Description: fix racy locking of mballoc block bitmaps causing BUG
532 Details : The locking of the mballoc buddy bitmap and the in-memory
533 block bitmap was using two different spin locks in some
534 cases. This made it possible to incorrectly access the
535 mballoc bitmap while another process was modifying it,
536 causing a sanity assertion to fail. While no on-disk corruption
537 was reported, there was some risk of this happening.
540 Frequency : rare, on llog test 6
542 Descriptoin: don't allow connect to already connected import
543 Details : allowing connect to already connected import is hide connecting problem.
548 Descriptoin: allow kill process which wait statahead result
549 Details : for some reasons 'ls' can stick in waiting result from statahead,
550 in this case need way for kill this process.
552 Severity : enhancement
554 Description: Add state history info file, enhance import info file
555 Details : Track import connection state changes in a new osc/mdc proc file;
556 add overview-type data to the osc/mdc import proc file.
558 Severity : enhancement
560 Description: MDS create should not wait for statfs RPC while holding DLM lock.
562 Severity : enhancement
564 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
569 Descriptoin: OST grant too much space to client even there are not enough space.
570 Details : Client will shrink its grant cache to OST if there are no write
571 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
572 this grant cache if there are already not enough avaible space
573 (left_space < total_clients * 32M).
576 Frequency : start MDS on uncleanly shutdowned MDS device
578 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
579 Details : stay in waiting mds<>ost recovery finished produce random bugs
580 due race between two ll_sync thread for one lov target. send
581 ACTIVATE event only if connect realy finished and import have
585 Frequency : rare, connect and disconnect target at same time
587 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
588 Details : don't call obd_disconnect under lov_lock. this long time
589 operation and can block ptlrpcd which answer to connect request.
594 Descriptoin: don't lose wakeup for imp_recovery_waitq
595 Details : recover_import_no_retry or invalidate_import and import_close can
596 both sleep on imp_recovery_waitq, but we was send only one wakeup
600 Frequency : always with long access acl
602 Descriptoin: mds can't pack reply with long acl.
603 Details : mds don't control size of acl but they limited by reint/getattr
606 Severity : enhancement
608 Description: Update to SLES10 kernel-2.6.16.60-0.33.
610 Severity : enhancement
612 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
615 Frequency : start MDS on uncleanly shutdowned MDS device
617 Descriptoin: aborting recovery hang on MDS
618 Details : don't throttle destroy RPCs for the MDT.
621 Frequency : on remount
623 Description: external journal device not working after the remount
624 Details : clear dev_rdonly flag for external journal devices in
630 Description: shutdown vs evict race
631 Details : client_disconnect_export vs connect request race.
632 if client will evicted at this time - we start invalidate
633 thread without referece to import and import can be freed
637 Frequency : rare, need acl's on inode.
639 Description: client can't handle ost additional correctly
640 Details : if ost was added after client connected to mds client can have
641 hit lnet_try_match_md ... to big messages to wide striped files.
642 in this case need teach client to handle config events about add
643 lov target and update client max ea size at that event.
645 Severity : enhancement
647 Description: Changelogs
648 Details : Changelogs are a lightweight mechanism to track filesystem
649 metadata and namespace changes. The changelog is recorded
650 permanently on the MDTs, and is periodically "consumed" / purged
651 when records are no longer needed.
653 Severity : enhancement
655 Description: compact fld format with extents
656 Details : Store range of seq rather than every seq in FLD. Seq
657 controller update FLD rather than clients. In Case of CMD, mdt0
658 has FLD, all other metadata server act as non persistent proxy
659 for FLD queries and cache fld entries in fld cache.
664 Description: don't skip ost target if they assigned to file
665 Details : Drop slow OSCs if we can, but not for requested start idx.
666 This means "if OSC is slow and it is not the requested
667 start OST, then it can be skipped, otherwise skip it only
668 if it is inactive/recovering/out-of-space.
672 Description: more cleanup in mds_lov
673 Details : not send LOV EA under replay, we can't know about they size at this
674 time. Don't allow client connect to mds before any ost connected,
675 for avoid problems with LOV EA size and returning EIO to client.
677 Severity : enhancement
679 Description: Interoperability at server side (Disk interoperability)
681 Severity : enhancement
683 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
685 Severity : enhancement
687 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
689 Severity : enhancement
691 Description: New client IO stack (CLIO).
693 Severity : enhancement
695 Description: Commit on sharing. Eliminate inter-client dependencies between
696 uncommitted transactions by doing transaction commits.
697 Thereby clients may recovery independently.
700 Frequency : Create a symlink file with a very long name
702 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
703 Details : If there is no extra space in the request for early cancels,
704 ldlm_req_handles_avail() returns 0 instead of a negative value.
706 Severity : enhancement
708 Description: Add /proc entry for import status
709 Details : The mdc, osc, and mgc import directories now have
710 an import directory that contains useful import data for debugging
713 Severity : enhancement
715 Description: Re-disable certain /proc logging
716 Details : Enable and disable client's offset_stats, extents_stats and
717 extents_stats_per_process stats logging on the fly.
720 Frequency : Only on FC kernels 2.6.22+
722 Description: oops in statahead
723 Details : Do not drop reference count for the dentry from VFS when lookup,
724 VFS will do that by itself.
726 Severity : enhancement
728 Description: Generic /proc file permissions
729 Details : Set /Proc file permissions in a more generic way to enable non-
730 root users operate on some /proc files.
734 Description: Hitting mdc_commit_close() ASSERTION
735 Details : Properly handle request reference release in
736 ll_release_openhandle().
740 Description: quota recovery deadlock during mds failover
741 Details : This patch includes att18982, att18236, att18237 in bz14840.
743 1. fix osts hang when mds does failover with quotaon
744 2. prevent watchdog storm when osts threads wait for the
749 Frequency : only patchless client
750 Description: add workaround for race between add/remove dentry from hash
752 Severity : enhancement
754 Description: Allow OST glimpses to return PW locks
758 Description: LBUG when llog conf file is full
759 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
763 Description: Prevent import from entering FULL state when server in recovery
767 Description: service mount cannot take device name with ":"
768 Details : Only when device name contains ":/" will mount treat it as
774 Description: replace ptlrpcd with the statahead thread to interpret the async
775 statahead RPC callback
779 Frequency : on recovery
780 Description: I/O failures after umount during fail back
781 Details : if client reconnected to restarted server we need join to recovery
782 instead of find server handler is changed and process self
783 eviction with cancel all locks.
785 Severity : enhancement
787 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
789 Severity : enhancement
791 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
793 Severity : enhancement
795 Description: Upcall on Lustre log has been dumped
796 Details : Allow for a user mode script to be called once a Lustre log has
797 been dumped. It passes the filename of the dumped log to the
798 script, the location of the script can be specified via
799 /proc/sys/lnet/debug_log_upcall.
804 Description: avoid idr_remove called for id which is not allocated.
805 Details : Move assigment s_dev for clustered nfs to end of initialization,
806 to avoid problem with error handling.
811 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
812 Details : When connection is reused this not moved from CONN_UNUSED_HASH
813 into CONN_USED_HASH and this prodice warning when put connection
814 again in unused hash.
817 Severity : enhancement
819 Description: File striping can now be set to use an arbitrary pool of OSTs.
821 Severity : enhancement
823 Description: Export bytes_read/bytes_write count on OSC/OST.
827 Description: Early reply size mismatch, MGC loses connection
828 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
829 the connect flags are properly negotiated.
834 Description: quotas are not honored with O_DIRECT
835 Details : all writes with the flag O_DIRECT will use grants which leads to
836 this problem. Now using OBD_BRW_SYNC to guard this.
840 Description: add quota statistics
841 Details : 1. sort out quota proc entries and proc code.
842 2. add quota statistics
844 Severity : enhancement
846 Description: enable quota support for HEAD.
850 Description: Properly propagate oinfo flags from lov to osc for statfs
851 Details : restore missing copy oi_flags to lov requests.
853 Severity : enhancement
855 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
856 : l_getgroups(8), lst(8), routerstat(8)
857 Details : included man pages for llobdstat(8), llstat(8),
858 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
860 Severity : enhancement
862 Description: configure's --enable-quota should check the
863 : kernel .config for CONFIG_QUOTA
864 Details : configure is terminated if --enable-quota is passed but
865 : no quota support is in kernel
869 Description: Remove portals compatibility
870 Details : Remove portals compatibility, not interoperable with releases
875 Description: Resolve device initialization race
876 Details : Prevent proc handler from accessing devices added to the
877 obd_devs array but yet be intialized.
879 Severity : enhancement
881 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
883 Severity : enhancement
885 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
890 Description: Using wrong pointer in osc_brw_prep_request
891 Details : Access to array[-1] can produce panic if kernel compiled with
892 CONFIG_PAGE_ALLOC enabled
896 Description: Client runs out of low memory
897 Details : Consider only lowmem when counting initial number of llap pages
901 Description: *optional* service tags registration
902 Details : if the "service tags" package is installed on a Lustre node
903 When the filesystem is mounted, a local-node service tag will
904 be created. See http://inventory.sun.com/ for more information
905 about the Service Tags asset management system.
909 Description: Kernel BUG tries to release flock
910 Details : Lustre does not destroy flock lock before last reference goes
911 away. So always drop flock locks when client is evicted and
912 perform unlock regardless of successfulness of speaking to MDS.
916 Description: add refcount protection for osc callbacks, avoid panic on shutdown
920 Description: sanity test 65a fails if stripecount of -1 is set
921 Details : handle -1 striping on filesystem in ll_dirstripe_verify
926 Description: ASSERTION(CheckWriteback(page,cmd)) failed
927 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
930 Severity : enhancement
932 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
936 Description: do not process already freed flock
937 Details : flock can possibly be freed by another thread before it reaches
938 to ldlm_flock_completion_ast.
942 Description: LBUG during stress test
943 Details : Need properly lock accesses the flock deadlock detection list.
947 Description: oops in page fault handler
948 Details : kernel page fault handler can return two special 'pages' in error
949 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
953 Description: timeout with invalidate import.
954 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
955 should be handled by ptlrpcd. This produce long age waiting and
956 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
958 Severity : enhancement
960 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
964 Description: enable MGS and MDT services start separately
965 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
966 a co-located MGS without starting the MGS, which is a complement
967 to 'nosvc' mount option.
971 Frequency : after recovery
972 Description: precreate to many object's after del orphan.
973 Details : del orphan st in oscc last_id == next_id and this triger growing
974 count of precreated objects. Set flag LOW to skip increase count
975 of precreated objects.
979 Frequency : rare, on clear nid stats
980 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
981 Details : when clean nid stats sometimes try destroy live entry,
982 and this produce panic in free.
986 Description: Stack overflow during MDS log replay
987 ease stack pressure by using a thread dealing llog_process.
991 Description: wait until IO finished before start new when do lock cancel.
992 Details : VM protocol want old IO finished before start new, in this case
993 need wait until PG_writeback is cleared until check dirty flag and
994 call writepages in lock cancel callback.
996 Severity : enhancement
998 Description: using special macro for print time and cleanup in includes.
1002 Description: mds_mfd_close() ASSERTION(rc == 0)
1003 Details : In mds_mfd_close(), we need protect inode's writecount change
1004 within its orphan write semaphore to prevent possible races.
1008 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
1012 Frequency : rare, on shutdown ost
1013 Description: don't hit live lock with umount ost.
1014 Details : shrink_dcache_parent can be in long loop with destroy dentries,
1015 use shrink_dcache_sb instead.
1019 Description: don't panic with use echo client
1020 Details : echo client pass NULL as client nid pointer and this produce null
1021 pointer dereference.
1025 Description: fix build on ppc32
1026 Details : compile code with -m64 flag produce wrong object file for ppc32.
1030 Description: add message levels for liblustreapi
1034 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
1035 Details : Move the 'good_osts' check before the 'total_bavail' check. This
1036 will result in an -EAGAIN and in the exit call path we call
1037 alloc_rr() which will with increasing aggressiveness attempt to
1038 aquire precreated objects on the minimum number of required OSCs.
1042 Description: Use old size assignment to avoid deadlock
1043 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
1044 the scheduling while holding a spinlock. We do not need locking
1045 for size in ll_update_inode() because size is only updated from
1046 the MDS for directories or files without objects, so there is no
1047 other place to do the update, and concurrent access to such inodes
1048 are protected by the inode lock.
1052 Description: resolve "_IOWR redefined" build error on SLES10
1056 Description: dump the memory debugging after all modules are unloaded to
1057 suppress false negative in conf_sanity test 39
1059 Severity : enhancement
1061 Description: build kernel-ib packages for OFED 1.3 in our release cycle
1066 Description: fix SLES kernel versioning
1067 Details : the kernel version for our SLES 10 kernel did not include a "-"
1068 before the "smp" at the end. while this was not a problem in
1069 general, it did mean that software trying to use the kernel
1070 version to try to detect a vendor specific kernel would fail.
1071 this was most evident by the OFED build scripts.
1075 Description: Don't update lov_desc members until making sure they are valid
1076 Details : When updating lov_desc members via proc fs, need fix their
1077 validities before doing the real update.
1081 Description: don't put request into delay list while invalidate in flight.
1082 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
1083 import in flight. this produce timeout for invalidate and sometimes
1084 can cause stale data.
1088 Frequency : on ppc only
1089 Description: not convert ost objects for directory because it's not exist.
1090 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
1094 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
1096 Severity : enhancement
1098 Description: Update to support 2.6.22.14 vanilla kernel.
1102 Frequency : rare, on recovery
1103 Description: read procfs can produce deadlock in some situation
1104 Details : Holding lprocfs lock which send rpc can produce block for destroy
1105 obd objects and this also block reconnect with -EALREADY. This
1106 isn't fix all lprocfs bugs - but make it rare.
1108 Severity : enhancement
1110 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
1113 Frequency : frequent on X2 node
1115 Description: mdc_set_open_replay_data LBUG
1116 Details : Set replay data for requests that are eligible for replay.
1120 Description: lustre_mgs: operation 101 on unconnected MGS
1121 Details : When MGC is disconnected from MGS long enough, MGS will evict the
1122 MGC, and late on MGC cannot successfully connect to MGS and a lot
1123 of the error messages complaining that MGS is not connected.
1127 Frequency : on network error
1128 Description: panic with double free request if network error
1129 Details : mdc_finish_enqueue is finish request if any network error ocuring,
1130 but it's true only for synchronus enqueue, for async enqueue
1131 (via ptlrpcd) this incorrect and ptlrpcd want finish request
1134 Severity : enhancement
1136 Description: client-side metadata stat-ahead during readdir(directory readahead)
1137 Details : perform client-side metadata stat-ahead when the client detects
1138 readdir and sequential stat of dir entries therein
1141 Frequency : on start mds
1143 Description: Implement get_info(last_id) in obdfilter.
1146 Frequency : occasional
1148 Description: Correctly check stale fid, not start epoch if ost not support SOM
1149 Details : open with flag O_CREATE need set old fid in op_fid3 because
1150 op_fid2 was overwritten with new generated fid, but mds can answer
1151 with one of these two fids and both is not stale. Setattr
1152 incorrectly started an epoch and assume will be called
1153 done_writing, but without SOM done_writing ever being called.
1156 Frequency : rare, depends on device drivers and load
1158 Description: MDS or OSS nodes crash due to stack overflow
1159 Details : Code changes in 1.8.0 increased the stack usage of some functions.
1160 In some cases, in conjunction with device drivers that use a lot
1161 of stack the MDS (or possibly OSS) service threads could overflow
1162 the stack. One change which was identified to consume additional
1163 stack has been reworked to avoid the extra stack usage.
1166 Frequency : occasional
1168 Description: Do not fail import if osc_interpret_create gets -EAGAIN
1169 Details : If osc_interpret_create got -EAGAIN it immediately exits and
1170 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
1171 oscc_has_objects and see OSC has no objests and call
1172 oscc_internal_create to resend create request.
1174 Severity : enhancement
1176 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
1178 Severity : enhancement
1180 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
1183 Frequency : very rare
1185 Description: Fix replay if there is an un-replied request and open
1186 Details : In some cases, older replay request will revert the
1187 mcd->mcd_last_xid on MDS which is used to record the client's
1188 latest sent request.
1190 Severity : enhancement
1192 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
1194 Severity : enhancement
1196 Description: Add rhel5 support to HEAD.
1198 Serverity : enhancement
1200 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
1205 Description: Don't allow skipping OSTs if index has been specified.
1206 Details : Don't allow skipping OSTs if index has been specified, make
1207 locking in internal create lots better.
1211 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
1212 Details : access to bitfield in structure is always rounded to long
1213 and this produce problem with not atomic change any bit.
1217 Description: Lustre make rpms failed.
1218 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
1219 builds Lustre rpms from tarball.
1222 Frequency : rare on shutdown ost
1224 Description: If llog cancel was not send before clean_exports phase, this can
1225 produce deadlock in llog code.
1226 Details : If llog thread has last reference to obd and call class_import_put
1227 this produce deadlock because llog_cleanup_commit_master wait when
1228 last llog_commit_thread exited, but this never success because was
1229 called from llog_commit_thread.
1233 Description: allow userland application know is lost one of stripes.
1234 Details : fill lvb_blocks with error code on ost and return it to
1235 application if error flag found.
1239 Description: NULL lov_tgts causing MDS oops
1240 Details : more safe checks for NULL lov_tgts for avoid oops.
1242 Severity : enhancement
1244 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
1248 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
1249 Details : If an OST is down the MDS will hang indefinitely in
1250 obd_statfs_rqset() waiting for the statfs data. While for
1251 MDS QOS usage of statfs, it should not stuck in waiting.
1253 Severity : enhancement
1255 Description: remote_acl support
1256 Details : Support ACL-based permission check for remote user.
1257 Support setfacl/getfacl for remote user with the utils
1258 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
1259 the system "{s,g}etfacl" utils.
1261 Severity : enhancement
1263 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
1265 Severity : enhancement
1267 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
1271 Description: make mgs_setparam() handle fsname containing dash
1272 Details : fsname containing a dash does not work with lctl conf_param
1274 Severity : enhancement
1276 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
1278 Severity : enhancement
1280 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
1284 Frequency : rare, at shutdown
1285 Description: access already free / zero obd_namespace.
1286 Details : if client_disconnect_export was called without force flag set,
1287 and exist connect request in flight, this can produce access to
1288 NULL pointer (or already free pointer) when connect_interpret
1289 store ocd flags in obd_namespace.
1293 Frequency : only at startup
1294 Description: not alloc memory with spinlock held.
1295 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
1296 if any spinlock held.
1298 Severity : enhancement
1300 Description: make lustre randomly fail allocating memory
1301 Details : Make lustre randomly failed allocating memory for testing purpose.
1303 Severity : enhancement
1305 Description: lost problems with lov objid file
1306 Details : Fixes some scability and access to not inited memory problems
1307 in work with lov objdid file.
1312 Description: lfs find does not continue on file error
1313 Details : Continue other files processing when a file/dir is absent.
1317 Description: Inconsistent usage of lustre_pack_reply()
1318 Details : Standardize the usage of lustre_pack_reply() such that it
1319 always generate a CERROR on failure.
1322 Frequency : occasional
1324 Description: MDS hang or stay in waiting lock
1325 Details : If client receive lock with CBPENDING flag ldlm need send lock
1326 cancel as separate rpc, to avoid situation when cancel request
1327 can't processed due all i/o threads stay in wait lock.
1331 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1335 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1339 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1343 Description: Update to RHEL4 latest kernel.
1345 Severity : enhancement
1347 Description: Build SLES10 patchless client fails
1348 Details : The configure was broken by run ./configure with
1349 --with-linux-obj=.... argument for patchless client. When the
1350 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1351 can't search header adequately. Use absolute path such as
1352 -I($LINUX)/include instead.
1356 Description: interrupt oig_wait produce painc on resend.
1357 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1358 private set, and this produce situation when rq_ptlrpcd_data not
1359 copyed to new allocated request and triggered LBUG on assert
1360 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1361 ptlrpcd set if request is changed and can be safety changed to use
1366 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1367 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1368 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1369 these macros and replace them with their difinitions instead.
1373 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1374 Details : Update which_patch & target file for SLES10 latest kernel.
1377 Description: Remove client patches from SLES 10 kernel.
1378 Details : This causes SLES 10 clients to behave as patchless clients
1379 even on a Lustre-patched (server) kernel.
1381 Severity : enhancement
1383 Description: self-adjustable client's lru lists
1384 Details : use adaptive algorithm for managing client cached locks lru
1385 lists according to current server load, other client's work
1386 pattern, memory activities, etc. Both, server and client
1387 side namespaces provide number of proc tunables for controlling
1390 Severity : enhancement
1392 Description: light-weight GSS support
1393 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
1394 and reduce performance overhead.
1396 Severity : enhancement
1398 Description: Linux keyring support
1399 Details : Support using service of Linux keyring for Lustre GSS internal
1400 context refresh/cache mechanism.
1404 Description: Fix errors in lfs documentation
1405 Details : Fixes man pages
1409 Description: don't use GFP_* in generic Lustre code.
1410 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
1414 Description: obdclass is limited by single OBD_ALLOC(idarray)
1415 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
1419 Description: warnings with build patchless client with vanila 2.6.19 and up
1420 Details : change old ctl_table style and replace ctl_table/ctl_table_header
1421 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
1425 Frequency : Only for SLES
1426 Description: sanity_quota fail test_1
1427 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
1428 file, which makes TSTID[2] inunique.
1432 Frequency : Only for SLES9
1433 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
1434 Details : Update target/ChangeLog/which_patch .
1438 Frequency : Only for test_10 in sanity-quota.sh
1439 Description: a bug in quota test code
1440 Details : checking a test flag in a wrong place
1444 Frequency : Only for test_13 in sanity-quota.sh
1445 Description: a bug in quota test script
1446 Details : shouldn't check overfull allocated quota
1450 Description: jbd statistics
1451 Details : Port older jbd statistics patch for sles10
1453 Severity : enhancement
1455 Description: Add printing inode info into message about error in writepage.
1460 Description: reply_lock_interpret crash due to race with it and lock cancel.
1461 Details : Do not replay locks that are being cancelled. Do not reference
1462 locks by their address during replay, just by their handle.
1466 Frequency : When flocks are used.
1467 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
1468 Details : Flock locks might destroy just granted lock if it could be merged
1469 with another existing flock, this is done in completion handler,
1470 so teach ldlm_cli_enquque_fini that this is a valid case for
1476 Description: Oops in read and write path when failing to allocate lock.
1477 Details : Check if lock allocation failed and return error back.
1481 Description: lstripe command fails for valid OST index
1482 Details : The stripe offset is compared to lov->desc.ld_tgt_count
1483 instead of lov->desc.ld_active_tgt_count.
1487 Description: sanity.sh failed test 103
1488 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
1493 Description: df doesn't work properly if diskfs blocksize != 4K
1494 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
1498 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
1499 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
1501 Severity : enhancement
1503 Description: Adaptive timeouts
1504 Details : RPC timeouts adapt to changing server load and network
1505 conditions to reduce resend attempts and improve recovery time.
1509 Description: llapi_file_create() does not allow some changes
1510 Details : add llapi_file_open() that allows specifying the mode and
1511 open flags, and also returns an open file handle.
1515 Description: merge and cleanup kernel patches.
1516 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
1520 Description: Add journal checksum support.(Kernel part)
1521 Details : The journal checksum feature adds two new flags i.e
1522 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
1523 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
1524 indicates that the commit block contains the checksum for
1525 the blocks described by the descriptor blocks. Now commit
1526 record can be sent to disk without waiting for descriptor
1527 blocks to be written to disk. This behavior is controlled
1528 using JBD2_FEATURE_ASYNC_COMMIT flag.
1532 Description: OSS needs mutliple precreate threads
1533 Details : Add ability to start more than one create thread per OSS.
1537 Description: Sanity.sh test_65k failed: setstripe should have succeeded
1538 Details : execute lfs setstripe on client
1542 Description: mds_obd_create error creating tmp object
1543 Details : When the user sets quota on root, llog will be affected and can't
1544 create files and write files.
1547 Frequency : Always on ia64 patchless client, and possibly others.
1549 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
1550 Details : This allows the patchless client to be loaded on architectures
1551 without this export.
1555 Description: RedHat Update kernel for RHEL5
1556 Details : Add the kernel config file for RHEL5.
1560 Description: RedHat Update kernel for RHEL5
1561 Details : Modify the kernel config file more closer RHEL5.
1565 Description: Build failure against Centos5 (RHEL5)
1566 Details : Define PAGE_SIZE when it isn't present.
1570 Description: test_7 on sanity-quota.sh can't be used on separate machines
1571 Details : add facet to handle it
1575 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
1576 Details : with NFS, the anon dentry's parent was set to itself in
1577 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
1578 corresponding dentry other than use rec->ur_name.
1580 Severity : enhancement
1582 Description: lfs setstripe enhancement
1583 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
1587 Description: enable data checksumming by default
1588 Details : enable checksum by default, allow --disable-checksum
1589 configure option and "-o nochecksum" mount option. Checksums
1590 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
1594 Description: lustre support for RHEL5
1595 Details : Add support for RHEL5.
1599 Description: Client eviction due to failover config
1600 Details : after a connection loss, the lustre client should attempt to
1601 reconnect to the last active server first before trying the
1602 other potential connections.
1606 Description: when mds and osts use different quota unit(32bit and 64bit),
1607 quota will be released repeatly.
1608 Details : void sending multiple quota reqs to mds, which will keep the
1609 status between the reqs.
1613 Description: osts not allocated evenly to files
1614 Details : change the condition to increase offset_idx
1617 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
1619 Description: Data corruption for OSTs that are formatted larger than 2TB
1621 Details : When generating the bio request for lustre file writes the
1622 sector number would overflow a temporary variable before being
1623 used for the IO. The data reads correctly from Lustre (which
1624 will overflow in a similar manner) but other file data or
1625 filesystem metadata may be corrupted in some cases.
1629 Description: Tune the kernel for good SCSI performance.
1630 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
1631 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
1636 Description: rewrite ext2-derived code in obdclass/uuid.c
1637 Details : rewrite inherited code (uuid parsing code from ext2 utils)
1638 from scratch preserving functionality.
1642 Description: "lfs find -obd UUID" prints directories
1643 Details : "lfs find -obd UUID" will return all directory names instead
1644 of just file names. It is incorrect because the directories
1645 do not reside on the OSTs.
1649 Description: use i_size_read and i_size_write in 2.6 port
1650 Details : replace inode->i_size access with i_size_read/write()
1653 Frequency : when removing large files
1655 Description: scheduling issue during removal of large Lustre files
1656 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
1657 It causes scheduling issues when removing large files (17TB in the
1661 Frequency : only with liblustre clients on XT3
1663 Description: evictions taking too long
1664 Details : allow llrd to evict clients directly on OSTs
1667 Frequency : only on ppc
1669 Description: /proc/fs/lustre/devices broken on ppc
1670 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
1671 We should make sure the type of 'index' is loff_t and then cast
1672 explicitly as needed below. Do not assign an explicitly cast
1676 Frequency : only for rhel5
1678 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
1679 Details : Modify the target file & which_kernel.
1681 Severity : enhancement
1683 Description: omit set fsid for export NFS
1684 Details : fix set/restore device id for avoid EMFILE error and mark lustre
1685 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
1689 Frequency : Always, for kernels after 2.6.16
1690 Description: Fix warning idr_remove called for id=.. which is not allocated.
1691 Details : Last kernels save old s_dev before kill super and not allow
1692 to restore from callback - restore it before call kill_anon_super.
1696 Description: buffer overruns could theoretically occur
1697 Details : llapi_semantic_traverse() modifies the "path" argument by
1698 appending values to the end of the origin string, and a
1699 overrun may occur. Adding buffer overrun check in liblustreapi.
1703 Description: Fix error on 'ls .' at the top of the Lustre mount.
1704 Details : Don't revalidate dentry if it is a root dentry.
1708 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
1709 Details : Modify vm-tunables-rhel4.patch.
1713 Description: Kernel config for 2.6.18-vanilla.
1714 Details : Modify targets/2.6-vanilla.target.in.
1715 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
1716 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
1717 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
1718 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
1722 Description: improve handling recoverable errors
1723 Details : if request processig with error which can be recoverable on server
1724 request should be resend, otherwise page released from cache and
1729 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
1730 Details : Modify target file & which_patch.
1731 A flaw was found in the IA32 system call emulation provided
1732 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
1733 value could be stored in the %RAX register, which could trigger an
1734 out-of-bounds system call table access. An untrusted local user
1735 could exploit this flaw to run code in the kernel
1736 (ie a root privilege escalation). (CVE-2007-4573).
1740 Description: O_DIRECT bypasses client statistics.
1741 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
1742 read_ahead_stats were updated. copied the stats section from
1743 osc_send_oap_rpc() into async_internal().
1747 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
1751 Description: change order of libsysio includes
1752 Details : '#include sysio.h' should always come before '#include xtio.h'
1756 Description: handle "serious error: objid * already exists" more gracefully
1757 Details : If LAST_ID value on disk is smaller than the objects existing in
1758 the O/0/d* directories, it indicates disk corruption and causes an
1759 LBUG(). If the object is 0-length, then we should use the existing
1760 object. This will help to avoid a full fsck in most cases.
1762 Severity : enhancement
1764 Description: adapt the lustre_config script to support the upgrade case
1765 Details : Add "-u" option for lustre_config script to support upgrading 1.4
1766 server targets to 1.6 in parallel.
1770 Description: To avoid grant space > avaible space when the disk is almost
1771 full. Without this patch you might see the error "grant XXXX >
1772 available" or some LBUG about grant, when the disk is almost
1774 Details : In filter_check_grant, for non_grant cache write, we should
1775 check the left space by if (*left > ungranted + bytes), instead
1776 of (*left > ungranted), because only we are sure the left space
1777 is enough for another "bytes", then the ungrant space should be
1778 increase. In client, we should update cl_avail_grant only there
1779 is OBD_MD_FLGRANT in the reply.
1783 Description: Update RHEL 4 kernel to fix local root privilege escalation.
1784 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
1785 described in CVE-2007-4573. This problem could allow untrusted
1786 local users to gain root access.
1789 Frequency : when using O_DIRECT and quotas
1791 Description: Incorrect file ownership on O_DIRECT output files
1792 Details : block usage reported by 'lfs quota' does not take into account
1793 files that have been written with O_DIRECT.
1798 Description: touch file failed when fs is not full
1799 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
1800 otherwise we can get ENOSP while fs is not full.
1804 Description: parallel lock callbacks
1805 Details : Instead of sending blocking and completion callbacks as separated
1806 requests, adding them to a set and sending in parallel.
1809 Frequency : only for Cray XT3
1810 Bugzilla : 12829/13455
1811 Description: Changing primary group doesn't change the group lustre assigns to
1813 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
1814 overridden with the primary group provided by the group upcall,
1815 whereas the supplied fsgid can be trusted if it is in the list of
1816 supplementary groups returned by the group upcall.
1818 Severity : enhancement
1820 Description: Allow masking D_WARNING, D_ERROR messages from console
1821 Details : Console messages can now be disabled via lnet.printk.
1825 Description: User code with malformed file open parameter crashes client node
1826 Details : Before packing join_file req, all the related reference should be
1827 checked carefully in case some malformed flags cause fake
1828 join_file req on client.
1832 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
1833 Details : ldlm_completion_ast() assumes that a lock is granted when the req
1834 mode is equal to the granted mode. However, it should also check
1835 that LDLM_FL_CP_REQD is not set.
1839 Description: Heavy nfs access might result in deadlocks
1840 Details : After ELC code landed, it is now improper to enqueue any mds
1841 locks under och_sem, because enqueue might want to decide to
1842 cancel open locks for same inode we are holding och_sem for.
1846 Description: Client eviction while running blogbench
1847 Details : A lot of unlink operations with concurrent I/O can lead to a
1848 deadlock causing evictions. To address the problem, the number of
1849 oustanding OST_DESTROY requests is now throttled to
1850 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
1851 callbacks are processed in priority.
1855 Description: enable ACLs on MDS by default
1856 Details : ACLs must be enabled on MDS by default.
1859 Frequency : PPC/PPC64 only
1861 Description: conflicts between asm-ppc64/types.h and lustre_types.h
1862 Details : fix duplicated definitions between asm-ppc64/types.h and
1863 lustre_types.h on PPC.
1866 Frequency : PPC/PPC64 only
1868 Description: asm-ppc/segment.h does not exist
1869 Details : fix compile issue on PPC.
1873 Description: better handle error messages in extents code
1876 Frequency : RHEL4 only
1878 Description: mkfs is very slow on IA64/RHEL4
1879 Details : A performance regression has been discovered in the MPT Fusion
1880 driver between versions 3.02.73rh and 3.02.99.00rh. As a
1881 consequence, we have downgraded the MPT Fusion driver in the RHEL4
1882 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
1884 Severity : enhancement
1886 Description: SNMP support enhancement
1887 Details : Adding total number of sampled request for an MDS node in snmp
1890 Severity : enhancement
1892 Description: Optimize ldlm waiting list processing for PR extent locks
1893 Details : When processing waiting list for read extent lock and meeting read
1894 lock that is same or wider to it that is not contended, skip
1895 processing rest of the list and immediatelly return current
1896 status of conflictness, since we are guaranteed there are no
1897 conflicting locks in the rest of the list.
1901 Description: Time out and refuse to reconnect
1902 Details : When the failover node is the primary node, it is possible
1903 to have two identical connections in imp_conn_list. We must
1904 compare not conn's pointers but NIDs, otherwise we can defeat
1905 connection throttling.
1909 Description: port llog fixes from b1_6 into HEAD
1910 Details : Port llog reference couting and some llog cleanups from b1_6
1911 (bug 10800) into HEAD, for protect from panic and access to already
1912 free llog structures.
1916 Description: Detect stride IO mode in read-ahead
1917 Details : When a client does stride read, read-ahead should detect that and
1918 read-ahead pages according to the detected stride pattern.
1922 Description: data checksumming impacts single node performance
1923 Details : add support for several checksum algorithm. Currently, only CRC32
1924 and Adler-32 are supported. The checksum type can be changed on
1925 the fly via /proc/fs/lustre/osc/*/checksum_type.
1929 Description: use adler32 for page checksums
1930 Details : when available, use the Adler-32 algorithm instead of CRC32 for
1935 Description: build for x2 fails
1936 Details : fix compile issue on Cray systems.
1940 Description: Properly match for duplicate locks
1941 Details : Due to different lock order from skiplists code, we need to
1942 traverse entire list for now
1945 Frequency : only on PPC/SLES10
1947 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
1948 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
1949 be defined. Add a hack in mkfs_lustre.c to work around this compile
1954 Description: LASSERT on MDS when client holding flock lock dies
1955 Details : ldlm pool logic depends on number of granted locks equal to
1956 number of released locks which is not true for flock locks, so
1957 just exclude such locks from consideration.
1961 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
1962 Details : Use fsfilt_sync() for both the whole filesystem sync and
1963 individual file sync to eliminate dangerous inode locking
1964 with I_LOCK that can lead to a deadlock.
1968 Description: performance in 1.6.3
1969 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
1970 sure that 1MB requests are not fragmented by the block layer.
1972 Severity : enhancement
1974 Description: organize the server-side client stats on per-nid basis
1975 Details : Change the structure of stats under obdfilter and mds to
1983 The "uuid"s file would list the uuids of _active_ exports.
1984 And the clear entry is to clear all stats and stale nids.
1986 Severity : enhancement
1988 Description: eliminate client locks in face of contention
1989 Details : file contention detection and lockless i/o implementation
1990 for contended files.
1994 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
1997 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
1999 Description: MDS may lose file striping (and hence file data) in some cases
2000 Details : If there are additional extended attributes stored on the MDS,
2001 in particular ACLs, SELinux, or user attributes (if user_xattr
2002 is specified for the client mount options) then there is a risk
2003 of attribute loss. Additionally, the Lustre file striping
2004 needs to be larger than default (e.g. striped over all OSTs),
2005 and an additional attribute must be stored initially in the
2006 inode and then increase in size enough to be moved to the
2007 external attribute block (e.g. ACL growing in size) for file
2012 Description: skiplist implementation simplification
2013 Details : skiplists are used to group compatible locks on granted list
2014 that was implemented as tracking first and last lock of each
2015 lock group the patch changes that to using doubly linked lists
2019 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
2020 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
2021 because LASSERT on both the data supplied by a client, and the
2022 data on disk is dangerous and incorrect.
2024 Severity : enhancement
2026 Description: Slow truncate/writes to huge files at high offsets.
2027 Details : Directly associate cached pages to lock that protect those pages,
2028 this allows us to quickly find what pages to write and remove
2029 once lock callback is received.
2033 Description: more ldlm soft lockups
2034 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
2035 starve other threads from the resource lock for a long time in
2036 case of long waiting queue, so change the debug level from
2037 D_OTHER to the less frequently used D_INFO.
2039 Severity : enhancement
2041 Description: add -gid, -group, -uid, -user options to lfs find
2045 Description: Hung threads in invalidate_inode_pages2_range
2046 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
2047 once one is completed. As a result, some RPCs are stuck in the
2048 queue and are never sent.
2052 Description: filter threads hungs on waiting journal commit
2053 Details : Cleanup filter group llog code, then only filter group llog will
2054 be only created in the MDS/OST syncing process.
2058 Description: Procfs and llog threads access destoryed import sometimes.
2059 Details : Sync the import destoryed process with procfs and llog threads by
2060 the import refcount and semaphore.
2062 Severity : enhancement
2064 Description: openlock cache of b1_6 port to HEAD
2069 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
2070 Details : If insertion of an extent fails, then discard the inode
2071 preallocation and free data blocks else it can lead to duplicate
2076 Description: don't always update ctime in ext3_xattr_set_handle()
2077 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
2078 In some cases the ctime should not be updated, for example for
2079 2.0->1.8 compatibility it is necessary to delete an xattr and it
2080 should not update the ctime.
2084 Bugzilla : 15713/16362
2085 Description: Assertion in iopen_connect_dentry in 1.6.3
2086 Details : looking up an inode via iopen with the wrong generation number can
2087 populate the dcache with a disconneced dentry while the inode
2088 number is in the process of being reallocated. This causes an
2089 assertion failure in iopen since the inode's dentry list contains
2090 both a connected and disconnected dentry.
2094 Description: assertion failure in ldlm_handle2lock()
2095 Details : fix a race between class_handle_unhash() and class_handle2object()
2096 introduced in lustre 1.6.5 by bug 13622.
2101 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
2102 Details : remove the limit of 256 scsi disks in the sd_iostat patch
2107 Description: oops in sd_iostats_seq_show()
2108 Details : unloading/reloading the scsi low level driver triggers a kernel
2109 bug when trying to access the sd iostat file.
2114 Description: Kernel panics during QLogic driver reload
2115 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
2116 patch, causing memory corruption.
2121 Description: journal_dev option does not work in b1_6
2122 Details : pass mount option during pre-mount.
2124 Severity : enhancement
2126 Description: Add a FIEMAP(FIle Extent MAP) ioctl
2127 Details : FIEMAP ioctl will allow an application to efficiently fetch the
2128 extent information of a file. It can be used to map logical blocks
2129 in a file to physical blocks in the block device.
2133 Description: LDLM soft lockups - improvement
2134 Details : It is be possible to send the lock handle along with each read
2135 or write request because the client is already doing a lock match
2136 itself so there isn't any reason the OST should have to re-do that
2140 Frequency : only on Cray X2
2142 Description: X2 build failures
2143 Details : fix build failures on Cray X2.
2147 Description: xid & resent requests
2148 Details : Initialize RPC XID from clock at startup (randomly if clock is
2151 Severity : enhancement
2153 Description: Add lustre_start utility to start or stop multiple Lustre servers
2158 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
2159 Details : In case of memory pressure, list_del() can be called twice on
2160 req->rq_history_list, causing a kernel oops.
2164 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
2165 Details : fix stack overflow in the distributed lock manager by defering
2166 export eviction after a failed AST to the elt thread instead of
2167 handling it in the dlm interpret routine.
2171 Description: Convert some comments to new format.
2172 Details : Update documenting comments to match doxygen conventions.
2176 Description: Grammar fixes.
2177 Details : A couple of trivial spelling fixes.
2181 Description: OSD_COUNTERS-mandatory
2182 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
2183 to simplify the code.
2187 Description: simplify cmm_device freeing logic.
2188 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
2192 Description: Add lockdep support to dt_object_operations locking interface.
2193 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
2194 parameter indicating lock ordering. Update mdd code to use new
2199 Description: Introduce failloc constants for lockless IO tests.
2200 Details : Add two new failloc constants to test lockless IO. Only one of
2201 them in implemented---another is checked in yet to be landed
2206 Description: Add lockdep support for inode mutex.
2207 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
2208 in the situations where Lustre has to lock more than one inode
2213 Description: Add optional invariants checking support.
2214 Details : Add new LINVRNT() macro, optional on new --enable-invariants
2215 configure switch. This macro is to be used for consistency and
2216 sanity checks that are too expensive to be left in `production'
2221 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
2222 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
2227 Description: Add lockdep support for ldlm_lock and ldlm_resource.
2228 Details : Use spin_lock_nested() in (the only) situation where more than
2229 one ldlm_lock is locked simultaneously. Also, fix possible
2230 dead-lock in ldlm_lock_change_resource() by enforcing particular
2235 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
2236 Details : Instead of specifying each ldlm_lock call-back through separate
2237 parameter, wrap them into struct ldlm_callback_suite.
2241 Description: Kill join_lru obd method and its callers.
2242 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
2243 in memory---a requirement implemented through ->o_join_lru() obd
2244 method in HEAD. Remove this method and its users.
2248 Description: Add asynchronous ldlm ENQUEUE completion handler.
2249 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
2250 case that stock ldlm_completion_ast() cannot handle as it waits
2251 until lock is granted. Introduce new ldlm_completion_ast_async()
2252 for this. Also comment ldlm_completion_ast().
2256 Description: ldlm_error <-> errno conversion.
2257 Details : Add functions to map (rather arbitrary) between LDLM error codes
2258 and standard errno values. CLIO needs this to prevent LDLM specific
2259 constants from escaping ldlm and osc.
2263 Description: Kill unused ldlm_handle2lock_ns() function.
2264 Details : Kill unused ldlm_handle2lock_ns() function.
2268 Description: Add lu_ref support to ldlm_lock
2269 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
2270 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
2271 and ldlm_resource. LDLM interface has to be changed, because code that
2272 releases a reference on a lock, has to "know" what reference this is.
2273 In the most frequent case
2275 lock = ldlm_handle2lock(handle);
2277 LDLM_LOCK_PUT(lock);
2279 no changes are required. When any other reference (received _not_ from
2280 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
2281 instead of LDLM_LOCK_PUT().
2283 Arguably, changes are pervasive, and interface requires some discipline
2284 for proper use. On the other hand, it was very instrumental in finding
2285 a few leaked lock references.
2289 Description: Add ldlm_lock_addref_try().
2290 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
2291 attempts to addref a lock that might be being canceled
2296 Description: Add ldlm_weigh_callback().
2297 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
2298 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
2299 hard-coded `number of pages' logic.
2303 Description: Add lockdep annotations to llog code.
2304 Details : Use appropriately tagged _nested() locking calls in the places
2305 where llog takes more than one ->lgh_lock lock.
2309 Description: Add loi_kms_set().
2310 Details : Wrap kms updates into a helper function.
2314 Description: Constify instances of struct lsm_operations.
2315 Details : Constify instances of struct lsm_operations.
2319 Description: lu_conf support.
2320 Details : On a server, a file system object is uniquely identified
2321 by a fid, which is sufficient to locate and load all object
2322 state (inode). On a client, on the other hand, more data are
2323 necessary instantiate an object. Change lu_object_find() and
2324 friends to take additional `lu_conf' argument describing object.
2325 Typically this includes layout information.
2329 Description: lu_context fixes.
2330 Details : Introduce new lu_context functions that are needed on the client
2331 side, where some system threads (ptlrpcd) are shared by multiple
2332 modules, and so cannot be stopped during module shutdown.
2336 Description: Add start and stop methods to lu_device_type_operations.
2337 Details : Introduce two new methods in lu_device_type_operations, that are
2338 invoked when first instance of a given type is created and last one
2339 is destroyed respectively. This is need by CLIO.
2343 Description: Add lu_ref support to struct lu_device.
2344 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
2345 track leaked references.
2349 Description: Introduce lu_kmem_descr.
2350 Details : lu_kmem_descr and its companion interface allow to create
2351 and destroy a number of kmem caches at once.
2355 Description: Fix lu_object finalization race.
2356 Details : Fix a race between lu_object_find() finding an object and its
2357 concurrent finalization. This race is (most likely) not possible
2358 on the server, but might happen on the client.
2362 Description: Introduce lu_ref interface.
2363 Details : lu_ref is a debugging module allowing to track references to
2364 a given object. It is quite cpu expensive, and has to be
2365 explicitly enabled with --enable-lu_ref. See usage description
2370 Description: Factor lu_site procfs stats into a separate function.
2371 Details : Separate lu_site stats printing code into a separate function
2372 that can be reused on a client.
2376 Description: Constify instances of struct {lu,dt,md}_device_operations.
2377 Details : Constify instances of struct {lu,dt,md}_device_operations.
2381 Description: Introduce struct md_site and move meta-data specific parts of
2382 struct lu_site here.
2383 Details : Move md-specific fields out of struct lu_site into special struct
2384 md_site, so that lu_site can be used on a client.
2388 Description: Kill mdd_lov_destroy().
2389 Details : Remove unused mdd code.
2393 Description: Add st_block checking to multistat.c.
2394 Details : Add st_block checking to multistat.c.
2398 Description: Add lu_ref support to struct obd_device.
2399 Details : Add lu_ref tracking to obd_device.
2403 Description: Kill obd_set_fail_loc().
2404 Details : Remove unused code.
2408 Description: Add special type for ptlrpc_request interpret functions.
2409 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
2410 there. Actual usage will be in CLIO.
2414 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
2415 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
2416 doesn't work with lockdep.
2420 Description: Add rwv.c test program.
2421 Details : New testing program exercising readv(2) and writev(2) (Qian).
2425 Description: Add sendfile.c test program.
2426 Details : New testing program exercising sendfile(2) (Jay).
2430 Description: Ratelimit a message that can be very frequent.
2431 Details : Ratelimit a memory allocation failure message that can
2436 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2437 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2441 Description: Liblustre build fixes.
2442 Details : Liblustre build fixes.
2446 Description: libcfs: add cfs_{need,cond}_resched() interface.
2447 Details : libcfs: add cfs_{need,cond}_resched() definition and
2448 implementations for Linux, NT, and liblustre.
2450 Severity : enhancement
2452 Description: More exported tunables for mballoc
2453 Details : Add support for tunable preallocation window and new tunables for
2454 large/small requests
2458 Description: Detect corruption of block bitmap and checking for preallocations
2459 Details : Checks validity of on-disk block bitmap. Also it does better
2460 checking of number of applied preallocations. When corruption is
2461 found, it turns filesystem readonly to prevent further corruptions.
2465 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
2466 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
2467 window to 0. If it does hit cache miss, and read-pattern is still
2468 stride-io mode, does not reset the stride window, but also does not
2469 increase the stride window length in this case.
2473 Frequency : only for big-endian servers
2474 Description: Check if system is big-endian while mounting fs with extents feature
2475 Details : Mounting a filesystem with extents feature will fail on big-endian
2476 systems since ext3-based ldiskfs is not supported on big-endian
2477 systems. This can be over-riden with "bigendian_extents" mount option.
2479 Severity : enhancement
2481 Description: The root squash functionality
2482 Details : A security feature, which is to prevent users from being able
2483 to mount lustre on their desktop, run as root, and delete
2484 all of the files in the filesystem. The goal is accomplished by
2485 remapping user id (UID) and group id (GID) of the root user to
2486 a UID and GID specified by the system administartor via Lustre
2487 configuration management server (MGS). The functionality also
2488 allows to specify sets of clients for which the remapping does
2493 Description: Excessive recovery window
2494 Details : With AT enabled, the recovery window can be excessively long (6000+
2495 seconds). To address this problem, we no longer use
2496 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
2497 timeout no longer depends on the service time, it is set to
2498 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
2499 time via pb_service_time.
2503 Description: Watchdog triggered on MDS failover
2504 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
2505 the OSTs know that the MDS "UUID" can be reused for the same export
2506 from a different NID, so we do not need to wait for the export to be
2510 Frequency : rare, only if using MMP with Linux RAID
2512 Description: MMP doesn't work with Linux RAID
2513 Details : While using HA for Lustre servers with Linux RAID, it is possible
2514 that MMP will not detect multiple mounts. To make this work we
2515 need to unplug the device queue in RAID when the MMP block is being
2516 written. Also while reading the MMP block, we should read it from
2517 disk and not the cached one.
2519 Severity : enhancement
2521 Description: open file using fid
2522 Details : A file can be opened using just its fid, like
2523 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
2526 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
2529 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
2530 Details : add proper locking for creation of jbd_4k slab cache
2534 Description: MMP check in ext3_remount() fails without displaying any error
2535 Details : When multiple mount protection fails during remount, proper error
2538 Severity : enhancement
2540 Description: Allow stripe size to be up to 4G-64k
2541 Details : Fix math logic to allow large stripe sizes.
2545 Description: add check for >8TB ldiskfs filesystems
2546 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
2547 Don't allow >8TB ldiskfs filesystems to be mounted without
2548 force_over_8tb mount option
2553 Descriptoin: OSS complains "too many blocks" and crashes
2554 Details : In filter_commitrw_write():
2555 1. lquota_chkquota();
2556 2. obdo_from_inode();
2557 3. lquota_pending_commit();
2558 "2." has changed uid and gid in this case, then caused
2559 lquota_chkquota() and lquota_pending_commit() to use
2560 different uid and gid. That is the root of the bug.
2564 Descriptoin: getstripe doesn't report stripe count = -1 sanely
2565 Details : Detect and print out the 'stripe count = -1' case properly.
2569 Description: Out or order replies might be lost on replay
2570 Details : In ptlrpc_retain_replayable_request if we cannot find retained
2571 request with tid smaller then one currently being added, add it
2572 to the start, not end of the list.
2574 --------------------------------------------------------------------------------
2576 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
2578 * Support for kernels:
2579 2.6.5-7.283 (SLES 9),
2580 2.6.9-55.EL (RHEL 4),
2581 2.6.16.46-0.14 (SLES 10),
2582 2.6.18.8 vanilla (kernel.org)
2583 * Client support for unpatched kernels:
2584 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2585 2.6.16 - 2.6.22 vanilla (kernel.org)
2586 * Due to recently discovered recovery problems, we do not recommend
2587 using patchless RHEL 4 clients with this or any earlier release.
2588 * Recommended e2fsprogs version: 1.39.cfs8
2589 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2590 * Starting with this release, the ldiskfs backing filesystem required
2591 by Lustre is now in its own package, lustre-ldiskfs. This package
2592 should be installed. It is versioned separately from Lustre and
2593 may be released separately in future.
2597 Description: block reactivating mgc import until all deactivates complete
2598 Details : Fix race when failing back MDT/MGS to itself (testing)
2601 Frequency : at statup only
2603 Description: mds_lov_synchronize race leads to various problems
2604 Details : simultaneous MDT->OST connections at startup can cause the
2605 sync to abort, leaving the OSC in a bad state.
2607 Severity : enhancement
2609 Description: add optional extra BUILD_VERSION info
2610 Details : add a new environment variable (namely LUSTRE_VERS) which allows
2611 to override the lustre version.
2614 Frequency : 2.6.18 servers only
2616 Description: ll_kern_mount() doesn't release the module reference
2617 Details : The ldiskfs module reference count never drops down to 0
2618 because ll_kern_mount() doesn't release the module reference.
2623 Description: server LBUG when using old ost_num_threads parameter
2624 Details : Accept the old ost_num_threads parameter but warn that it
2625 is deprecated, and fix an off-by-one error that caused an LBUG.
2630 Description: Transient SCSI error results in persistent IO issue
2631 Details : iobuf->dr_error is not reinitialized to 0 between two
2635 Frequency : sometimes when underlying device returns I/O errors
2637 Description: OSTs not going read-only during write failures
2638 Details : OSTs are not remounted read-only when the journal commit threads
2639 get I/O errors because fsfilt_ext3 calls journal_start/stop()
2640 instead of the ext3 wrappers.
2644 Description: poor connect scaling with increasing client count
2645 Details : Don't run filter_grant_sanity_check for more than 100 exports
2646 to improve scaling for large numbers of clients.
2649 Frequency : SLES10 only
2651 Description: sanity-quota.sh quotacheck failed: rc = -22
2652 Details : Quotas cannot be enabled on SLES10.
2655 Frequency : liblustre clients only
2657 Description: getdirentries does not give error when run on compute nodes
2658 Details : getdirentries does not fail when the size specified as an argument
2659 is too small to contain at least one entry
2661 Severity : enhancement
2663 Description: Add LNET router traceability for debug purposes
2664 Details : If a checksum failure occurs with a router as part of the
2665 IO path, the NID of the last router that forwarded the bulk data
2666 is printed so it can be identified.
2671 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
2672 Details : Due to a race condition, liblustre clients were occasionally
2673 evicted incorrectly.
2675 Severity : enhancement
2677 Description: lfs setstripe use optional parameters instead of postional
2680 Severity : enhancement
2682 Description: Nanosecond timestamp support for ldiskfs
2683 Details : The on-disk ldiskfs filesystem has added support for nanosecond
2684 resolution timestamps. There is not yet support for this at
2685 the Lustre filesystem level.
2688 Frequency : during server recovery
2690 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
2691 Details : request with rq_no_resend flag not awake l_wait_event if they get
2695 Frequency : nfs export on patchless client
2697 Description: connectathon hang when test nfs export over patchless client
2698 Details : Disconnected dentry cannot be found with lookup, so we do not need
2699 to unhash it or make it invalid
2702 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
2703 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
2704 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
2709 Description: open req refcounting wrong on reconnect
2710 Details : If reconnect happened between getting open reply from server and
2711 call to mdc_set_replay_data in ll_file_open, we will schedule
2712 replay for unreferenced request that we are about to free.
2713 Subsequent close will crash in variety of ways.
2714 Check that request is still eligible for replay in
2715 mdc_set_replay_data().
2720 Description: disable writes to filesystem when reading health_check file
2721 Details : the default for reading the health_check proc file has changed
2722 to NOT do a journal transaction and write to disk, because this
2723 can cause reads of the /proc file to hang and block HA state
2724 checking on a healthy but otherwise heavily loaded system. It
2725 is possible to return to the previous behaviour during configure
2726 with --enable-health-write.
2728 Severity : enhancement
2730 Description: 64-bit inode version
2731 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
2732 to the inode. This will be required for version-based recovery.
2737 Description: MDS fails to start if a duplicate client export is detected
2738 Details : in some rare cases it was possible for a client to connect to
2739 an MDS multiple times. Upon recovery the MDS would detect this
2740 and fail during startup. Handle this more gracefully.
2742 Severity : enhancement
2744 Description: Add -o localflock option to simulate old noflock
2746 Details : This will achieve local-only flock/fcntl locks
2752 Description: log_commit_thread vs filter_destroy race leads to crash
2753 Details : Take import reference before releasing llog record semaphore
2758 Description: Wrong request locking in request set processing
2759 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
2760 imp_delayed_list, in this place should be used imp_lock.
2763 Frequency : when reconnection
2765 Description: Grant Leak when osc reconnect to OST
2766 Details : When osc reconnect ost, OST(filter) should check whether it
2767 should grant more space to client by comparing fed_grant and
2768 cl_avail_grant, and return the granted space to client instead
2769 of "new granted" space, because client will call osc_init_grant
2770 to update the client grant space info.
2773 Frequency : when client reconnect to OST
2775 Description: Grant Leak when osc do resend and replay bulk write
2776 Details : When osc reconnect to OST, OST(filter)should clear grant info of
2777 bulk write request, because the grant info while be sync between
2778 OSC and OST when reconnect, and we should ignore the grant info
2779 these of resend/replay write req.
2784 Description: Grant space more than avaiable left space sometimes.
2785 Details : When then OST is about to be full, if two bulk writing from
2786 different clients came to OST. Accord the avaliable space of the
2787 OST, the first req should be permitted, and the second one
2788 should be denied by ENOSPC. But if the seconde arrived before
2789 the first one is commited. The OST might wrongly permit second
2790 writing, which will cause grant space > avaiable space.
2793 Frequency : when client is evicted
2795 Description: Grant might be wrongly erased when osc is evicted by OST
2796 Details : when the import is evicted by server, it will fork another
2797 thread ptlrpc_invalidate_import_thread to invalidate the
2798 import, where the grant will be set to 0. While the original
2799 thread will update the grant it got when connecting. So if
2800 the former happened latter, the grant will be wrongly errased
2801 because of this race.
2806 Description: Checking Stale with correct fid
2807 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
2808 to check whether it is stale, because sometimes, we want the
2809 enqueue happened anyway, and op_data.fid2 will not be initialized.
2811 Severity : enhancement
2813 Description: update patchless client
2814 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
2817 Frequency : only with 2.4 kernel
2819 Description: random memory corruption
2820 Details : size of struct ll_inode_info is to big for union inode.u and this
2821 can be cause of random memory corruption.
2826 Description: Memory leak in recovery
2827 Details : Lov_mds_md was not free in an error handler in mds_create_object.
2828 It should also check obd_fail before fsfilt_start, otherwise if
2829 fsfilt_start return -EROFS,(failover mds during mds recovery).
2830 then the req will return with repmsg->transno = 0 and rc = EROFS.
2831 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
2832 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
2833 be freed no matter whether fsfilt_commit success or not.
2836 Frequency : only with huge count clients
2838 Description: Prevents from taking the superblock lock in llap_from_page for
2840 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
2841 for llap_from_page call in ll_removepage prevents from taking the
2842 superblock lock for a soon died page.
2847 Description: Not check open intent error before release open handle
2848 Details : in some rare cases, the open intent error is not checked before
2849 release open handle, which may cause
2850 ASSERTION(open_req->rq_transno != 0), because it tries to release
2851 the failed open handle.
2856 Description: Set cat log bitmap only after create log success.
2857 Details : in some rare cases, the cat log bitmap is set too early. and it
2858 should be set only after create log success.
2862 Description: Accessing a block bevice can re-enable I/O when Lustre is
2863 tearing down a device.
2864 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
2869 Description: service threads may hog cpus when there are a lot of requests
2870 Details : Insert cond_resched to give other threads a chance to use some CPU
2875 Description: the cat log was not initialized in recovery
2876 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
2877 unlink log on mds will not be initialized until mds post
2878 recovery. And also in mds post recovery, the unlink log will
2879 initialization will be done asynchronausly, so there will be race
2880 between add unlink log and unlink log initialization.
2884 Description: brw_stats were being printed incorrectly
2885 Details : brw_stats were being printed as log2 but all of them were not
2886 recorded as log2. Also remove some code duplication arising from
2887 filter_tally_{read,write}.
2891 Frequency : rare, only in recovery.
2892 Description: ASSERTION(req->rq_type != LI_POISON) failed
2893 Details : imp_lock should be held while iterating over imp_sending_list for
2894 prevent destroy request after get timeout in ptlrpc_queue_wait.
2898 Description: replay-single.sh test 52 fails
2899 Details : A lock's skiplist need to be cleanup when it being unlinked
2900 from its resource list.
2904 Description: Short directio read returns full requested size rather than
2906 Details : Direct I/O operations should return actual amount of bytes
2907 transferred rather than requested size.
2909 Severity : enhancement
2911 Description: metadata RPC reduction (e.g. for rm performance)
2912 Details : decrease the amount of synchronous RPC between clients and servers
2913 by canceling conflicing lock before the operation on the client
2914 and packing thier handles into the main operation RPC to server.
2916 Severity : enhancement
2918 Description: Async OSC create to avoid the blocking unnecessarily.
2919 Details : If a OST has no remain object, system will block on the creating
2920 when need to create a new object on this OST. Now, ways use
2921 pre-created objects when available, instead of blocking on an
2922 empty osc while others are not empty. If we must block, we block
2923 for the shortest possible period of time.
2927 Description: improve handling recoverable errors
2928 Details : if request processig with error which can be recoverable on server
2929 request should be resend, otherwise page released from cache and
2932 Severity : enhancement
2934 Description: refine locking for avoid write wrong info into lov_objid file
2935 Details : fix possible races with add new target and write/update data in
2938 --------------------------------------------------------------------------------
2940 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
2945 Frequency : on some architectures
2947 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
2948 Details : Uninitialized flags sometimes cause configuration commands to
2952 Frequency : patchless clients only
2954 Description: missing __iget() symbol export
2955 Details : The __iget() symbol export is missing. To avoid the need for
2956 this on patchless clients the deathrow inode reaper is turned
2957 off, and we depend on the VM to clean up old inodes. This
2958 dependency was during via the fix for bug 12181.
2960 --------------------------------------------------------------------------------
2962 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
2964 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
2965 INTEROPERATE with older versions automatically. In many cases a
2966 special upgrade step is needed. Please read the
2967 user documentation before upgrading any part of a live system.
2968 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
2969 version will not interoperate with 1.6 betas before beta5 (1.5.95).
2970 * WARNING: Lustre configuration and startup changes are required with
2971 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
2975 Severity : enhancement
2977 Description: MountConf
2978 Details : Lustre configuration is now managed via mkfs and mount
2979 commands instead of lmc and lconf. New obd types (MGS, MGC)
2980 are added for dynamic configuration management. See
2981 https://mail.clusterfs.com/wikis/lustre/MountConf for
2984 Severity : enhancement
2986 Description: dynamic OST addition
2987 Details : OSTs can now be added to a live filesystem
2989 Severity : enhancement
2991 Description: startup order invariance
2992 Details : MDTs and OSTs can be started in any order. Clients only
2993 require the MDT to complete startup.
2995 Severity : enhancement
2997 Description: parallel, asynchronous orphan cleanup
2998 Details : orphan cleanup is now performed in separate threads for each
2999 OST, allowing parallel non-blocking operation.
3001 Severity : enhancement
3003 Description: optimized stripe assignment
3004 Details : stripe assignments are now made based on ost space available,
3005 ost previous usage, and OSS previous usage, in order to try
3006 to optimize storage space and networking resources.
3008 Severity : enhancement
3010 Description: Permanently set tunables
3011 Details : All writable /proc/fs/lustre tunables can now be permanently
3012 set on a per-server basis, at mkfs time or on a live system.
3014 Severity : enhancement
3016 Description: Lustre message v2
3017 Details : Add lustre message format v2.
3019 Severity : enhancement
3021 Description: client OST exclusion list
3022 Details : Clients can be started with a list of OSTs that should be
3023 declared "inactive" for known non-responsive OSTs.
3027 Description: SPEC SFS validation failure on NFS v2 over lustre.
3028 Details : Changes the blocksize for regular files to be 2x RPC size,
3029 and not depend on stripe size.
3031 Severity : enhancement
3033 Description: Multiple MD RPCs in flight.
3034 Details : Further unserialise some read-only MDS RPCs - learn about intents.
3035 To avoid overly-overloading MDS, introduce a limit on number of
3036 MDS RPCs in flight for a single client and add /proc controls
3037 to adjust this limit.
3039 Severity : enhancement
3041 Description: client read/write statistics
3042 Details : Add client read/write call usage stats for performance
3043 analysis of user processes.
3044 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
3045 file access. extents_stats shows chunk size distribution.
3046 extents_stats_per_process show chunk size distribution per
3049 Severity : enhancement
3051 Description: mds statistics
3052 Details : Add detailed mds operations statistics in
3053 /proc/fs/lustre/mds/*/stats.
3057 Description: Failure of copying files with lustre special EAs.
3058 Details : Client side always return success for setxattr call for lustre
3059 special xattr (currently only "trusted.lov").
3064 Description: Refcount LNET uuids
3065 Details : The global LNET uuid list grew linearly with every startup;
3066 refcount repeated list entries instead of always adding to
3069 Severity : enhancement
3071 Description: Dynamic service threads
3072 Details : Within a small range, start extra service threads
3073 automatically when the request queue builds up.
3076 Frequency : mixed-endian client/server environments
3078 Description: mixed-endian crashes
3079 Details : The new msg_v2 system had some failures in mixed-endian
3082 Severity : enhancement
3084 Description: Easy OST removal
3085 Details : OSTs can be permanently deactivated with e.g. 'lctl
3086 conf_param lustre-OST0001.osc.active=0'
3088 Severity : enhancement
3090 Description: MGS proc entries
3091 Details : Added basic proc entries for the MGS showing what filesystems
3094 Severity : enhancement
3096 Description: provide MGS failover
3097 Details : Added config lock reacquisition after MGS server failover.
3099 Severity : enhancement
3101 Description: add Linux 2.4 support
3102 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
3106 Description: a large application tries to do I/O to the same resource and dies
3107 in the middle of it.
3108 Details : Check the req->rq_arrival time after the call to
3109 ost_brw_lock_get(), but before we do anything about
3110 processing it & sending the BULK transfer request. This
3111 should help move old stale pending locks off the queue as
3112 quickly as obd_timeout.
3115 Frequency : when an incorrect nid is specified during startup
3117 Description: ptlrpc connect to non-existant node causes kernel crash
3118 Details : LNET can't be re-entered from an event callback, which
3119 happened when we expire a message after the export has been
3120 cleaned up. Instead, hand the zombie cleanup off to another
3123 Severity : enhancement
3125 Description: plain/inodebits lock performance improvement
3126 Details : Grouping plain/inodebits in granted list by their request modes
3127 and bits policy, thus improving the performance of search through
3131 Frequency : only if OST filesystem is corrupted
3133 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3134 Details : for a short time RPCs with bulk IO are in the replay list,
3135 but replay of bulk IOs is unimplemented. If the OST filesystem
3136 is corrupted due to disk cache incoherency and then replay is
3137 started it is possible to trip an assertion. Avoid putting
3138 committed RPCs into the replay list at all to avoid this issue.
3141 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3143 Description: System hang on startup
3144 Details : This bug allowed the liblustre (e.g. catamount) client to
3145 return to the app before handling all startup RPCs. This
3146 could leave the node unresponsive to lustre network traffic
3147 and manifested as a server ptllnd timeout.
3149 Severity : enhancement
3151 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3152 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
3153 boolean to control whether to print peer debug info when a
3154 client's RPC times out.
3157 Frequency : only for kernels with patches from Lustre below 1.4.3
3159 Description: Remove old rdonly API
3160 Details : Remove old rdonly API which unused from at least lustre 1.4.3
3163 Frequency : only for devices with external journals
3165 Description: Set external device read-only also
3166 Details : During a commanded failover stop, we set the disk device
3167 read-only while the server shuts down. We now also set any
3168 external journal device read-only at the same time.
3171 Frequency : when upgrading from 1.4 while trying to change parameters
3173 Description: The wrong (new) MDC name was used when setting parameters for
3174 upgraded MDT's. Also allows changing of OSC (and MDC)
3175 parameters if --writeconf is specified at tunefs upgrade time.
3178 Frequency : when setting specific ost indicies
3180 Description: QOS code breaks on skipped indicies
3181 Details : Add checks for missing OST indicies in the QOS code, so OSTs
3182 created with --index need not be sequential.
3184 Severity : enhancement
3186 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3187 Details : The uninit_groups feature works in conjunction with the kernel
3188 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3189 up the pass1 processing of e2fsck. This is a read-only feature
3190 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3191 mount filesystems that have had this feature enabled.
3193 Severity : enhancement
3195 Description: Improve multi-block allocation algorithm to avoid fragmentation
3196 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3197 allocation locality and avoid filesystem fragmentation.
3199 ------------------------------------------------------------------------------
3201 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
3203 * Support for kernels:
3204 2.4.21-47.0.1.EL (RHEL 3)
3205 2.6.5-7.283 (SLES 9)
3206 2.6.9-42.0.10.EL (RHEL 4)
3207 2.6.12.6 vanilla (kernel.org)
3208 2.6.16.27-0.9 (SLES 10)
3209 * Recommended e2fsprogs version: 1.39.cfs5
3211 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3215 Frequency : occasional, depends on client load and configuration
3216 Bugzilla : 12181, 12203
3217 Description: data loss for recently-modified files
3219 Details : In some cases it is possible that recently written or created
3220 files may not be written to disk in a timely manner (this should
3221 normally be within 30s unless client IO load is very high).
3222 The problem appears as zero-length files or files that are a
3223 multiple of 1MB in size after a client crash or client eviction
3224 that are missing data at the end of the file.
3226 This problem is more likely to be hit on clients where files are
3227 repeatedly created and unlinked in the same directory, clients
3228 have a large amount of RAM, have many CPUs, the filesystem has
3229 many OSTs, the clients are rebooted frequently, and/or the files
3230 are not accessed by other nodes after being written.
3232 The presence of the problem can be detected by looking at
3233 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
3234 smaller than the second (nr_unused) then dirty files will not
3235 be flushed automatically to disk. "sync; sleep 10" should be
3236 run several times on the node before unmounting it to update
3237 Lustre (this is also safe to run on nodes without this problem).
3239 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
3240 that can cause this same problem, so customers using that kernel
3241 also need to update the kernel in addition to Lustre. In order
3242 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
3244 It is normal that files written just before a client crash (less
3245 than 30s) may not yet have been flushed to disk, even for local
3249 Frequency : frequent on thin XT3 nodes
3251 Description: UUID collision on thin XT3 Linux nodes
3252 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
3253 randomly, since we relied on an insufficiently-seeded PRNG.
3258 Description: OSS hangs after "All ost request buffers busy"
3259 Details : A deadlock between quota and journal operations caused OSS
3260 hangs after printing "All ost request buffers busy."
3263 Frequency : always on liblustre builds
3265 Description: Cleanup compiler warnings on liblustre
3268 Frequency : always on liblustre builds on XT3
3270 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
3274 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
3275 > 32000 subdirectories
3276 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
3277 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
3278 subdirectory count crosses 32000. This will aid e2fsck to
3279 correctly handle more than 32000 subdirectories.
3282 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3284 Description: System hang on startup
3285 Details : This bug allowed the liblustre (e.g. catamount) client to
3286 return to the app before handling all startup RPCs. This
3287 could leave the node unresponsive to lustre network traffic
3288 and manifested as a server ptllnd timeout.
3290 Severity : enhancement
3292 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3293 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
3294 boolean to control whether to print peer debug info when a
3295 client's RPC times out.
3300 Description: make O_SYNC working on 2.6 kernels
3301 Details : 2.6 kernels use different method for mark pages for write,
3302 so need add a code to lustre for O_SYNC work.
3307 Description: Failure to close file and release space on NFS
3308 Details : Put inode details into lock acquired in ll_intent_file_open.
3309 Use mdc_intent_lock in ll_intent_open to properly
3310 detect all kind of errors unhandled by mdc_enqueue.
3315 Description: proc file read during shutdown sometimes raced obd removal,
3317 Details : Add lock to prevent obd access after proc file removal.
3320 Frequency : Only for files larger than 4GB on 32-bit clients.
3322 Description: improperly doing page alignment of locks
3323 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
3324 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3329 Description: RPCs being resent when they shouldn't be
3330 Details : Some RPCs that should not be resent are being resent. This
3331 can cause inconsistencies in the RPC state machine. Do not
3332 resend such requests.
3335 Frequency : rare, only with NFS export
3337 Description: Crash on NFS re-export node
3338 Details : under very unusual load conditions an assertion is hit in
3339 ll_intent_file_open()
3342 Frequency : only if OST filesystem is corrupted
3344 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3345 Details : for a short time RPCs with bulk IO are in the replay list,
3346 but replay of bulk IOs is unimplemented. If the OST filesystem
3347 is corrupted due to disk cache incoherency and then replay is
3348 started it is possible to trip an assertion. Avoid putting
3349 committed RPCs into the replay list at all to avoid this issue.
3354 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3355 Details : Large single O_DIRECT read and write calls can fail to allocate
3356 a sufficiently large buffer to process the request. In case of
3357 allocation failure the allocation is retried with a smaller
3358 buffer and broken into smaller requests.
3360 Severity : enhancement
3362 Description: Add -o localflock option to simulate old noflock behaviour.
3363 Details : This will achieve local-only flock/fcntl locks coherentness.
3368 Description: versioning check is incomplete
3369 Details : Checking the version difference of client vs. server, report
3370 error if the gap is too big.
3375 Description: add support PG_writeback bit
3376 Details : add support for PG_writeback bit for Lustre, for more carefull
3377 work with page cache in 2.6 kernel. This also fix some deadlocks
3378 and remove hack for work O_SYNC with 2.6 kernel.
3380 Severity : enhancement
3382 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3383 Details : The uninit_groups feature works in conjunction with the kernel
3384 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3385 up the pass1 processing of e2fsck. This is a read-only feature
3386 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3387 mount filesystems that have had this feature enabled.
3389 Severity : enhancement
3391 Description: Improve multi-block allocation algorithm to avoid fragmentation
3392 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3393 allocation locality and avoid filesystem fragmentation.
3395 ------------------------------------------------------------------------------
3397 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3399 * Support for kernels:
3400 2.6.9-42.0.3.EL (RHEL 4)
3401 2.6.5-7.276 (SLES 9)
3402 2.4.21-47.0.1.EL (RHEL 3)
3403 2.6.12.6 vanilla (kernel.org)
3404 2.6.16.21-0.8 (SLES10)
3405 * Recommended e2fsprogs version: 1.39.cfs2-0
3407 * The backwards-compatible /proc/sys/portals symlink has been removed
3408 in this release. Before upgrading, please ensure that you change
3409 any configuration scripts or /etc/sysctl.conf files that access
3410 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3411 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3412 in advance of the upgrade on any system running Lustre 1.4.6 or
3413 newer, since /proc/sys/lnet was added in that version.
3414 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3418 Frequency : only when quota is used
3420 Description: avoid scanning export list for quota master
3421 Details : Change the algorithms to avoid scanning export list in order
3422 to improve the efficiency.
3425 Frequency : MDS failover only, very rarely
3427 Description: "went back in time" messages on mds failover
3428 Details : The greatest transno may be lost when the current operation
3429 finishes with an error (transno==0) and the client's last_rcvd
3430 record is over-written. Save the greatest transno in the
3431 mds_last_transno for this case.
3434 Frequency : always for specific kernels and striping counts
3436 Description: client may get "Matching packet too big" without ACL support
3437 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
3438 when trying to access files in certain configurations. The
3439 clients should in fact be denied when mounting because they do
3440 not understand ACLs.
3443 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
3445 Description: many clients connecting with IO in progress causes connect timeouts
3446 Details : Avoid synchronous journal commits to avoid delays caused by many
3447 clients connecting/disconnecting when bulk IO is in progress.
3448 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
3449 OST_IO_PORTAL to avoid delays behind potentially many pending
3453 Frequency : occasionally with multiple writers to a single file
3455 Description: shared writes to file may result in wrong size reported by stat()
3456 Details : Allow growing of kms when extent lock is cancelled
3459 Frequency : always with random mmap IO to multi-striped file
3461 Description: mmap write might be lost if we are writing to a 'hole' in stripe
3462 Details : Only if the hole is at the end of OST object so that kms is too
3463 small. Fix is to increase kms accordingly in ll_nopage.
3466 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
3468 Description: writes to a missing object would leak memory on the OST
3469 Details : If there is an inconsistency between the MDS and OST filesystems,
3470 such that the MDS references an object that doesn't exist, writes
3471 to that object will leak memory due to incorrect cleanup in the
3472 error handling path, eventually running out of memory on the OST.
3477 Description: Creating too long symlink causes lustre errors
3478 Details : Check symlink and name lengths before sending requests to MDS.
3481 Frequency : only if flock is enabled (not on by default)
3483 Description: posix locks not released on fd closure on 2.6.9+
3484 Details : We failed to add posix locks to list of inode locks on 2.6.9+
3485 kernels, this caused such locks not to be released on fd close and
3486 then assertions on fs unmount about still used locks.
3489 Frequency : MDS failover only, very rarely
3491 Description: clients may get ASSERTION(granted_lock != NULL)
3492 Details : When request was taking a long time, and a client was resending
3493 a getattr by name lock request. The were multiple lock requests
3494 with the same client lock handle and
3495 mds_getattr_name->fixup_handle_for_resent_request found one of the
3496 lock handles but later failed with ASSERTION(granted_lock != NULL).
3501 Description: handle->h_buffer_credits > 0, assertion failure
3502 Details : h_buffer_credits is zero after truncate, causing assertion
3503 failure. This patch extends the transaction or creates a new
3507 Frequency : NFS re-export or patchless client
3508 Bugzilla : 11179, 10796
3509 Description: Crash on NFS re-export node (__d_move)
3510 Details : We do not want to hash the dentry if we don't have a lock.
3511 But if this dentry is later used in d_move, we'd hit uninitialised
3512 list head d_hash, so we just do this to init d_hash field but
3513 leave dentry unhashed.
3516 Frequency : NFS re-export or patchless client
3518 Description: NFS exports has problem with symbolic link
3519 Details : lustre client didn't properly install dentry when re-exported
3520 to NFS or running patchless client.
3523 Frequency : NFS re-export or patchless client
3525 Description: Various nfs/patchless fixes.
3526 Details : fixes reuse disconected alias for lookup process - this fixes
3527 warning "find_exported_dentry: npd != pd",
3528 fix permission error with open files at nfs.
3529 fix apply umask when do revalidate.
3532 Frequency : occasional
3534 Description: Crash on NFS re-export node
3535 Details : calling clear_page() on the wrong pointer triggered oops in
3536 generic_mapping_read().
3539 Frequency : rarely, using O_DIRECT IO
3541 Description: unaligned directio crashes client with LASSERT
3542 Details : check for unaligned buffers before trying any requests.
3545 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
3547 Description: stale data returned from RAID cache
3548 Details : If only a small amount of IO is done to the RAID device before
3549 reading it again it is possible to get stale data from the RAID
3550 cache instead of reading it from disk.
3553 Frequency : always for sles10 kernel
3555 Description: sles10 support
3556 Details : ll_follow_link: compile fixes and using of nd_set_link
3557 under newer kernels.
3560 Frequency : depends on arch, kernel and compiler version, always on sles10
3563 Description: recursive or deep enough symlinks cause stack overflow
3564 Details : getting rid of large stack-allocated variable in
3568 Frequency : depends on hardware
3570 Description: lustre write performance loss in the SLES10 kernel
3571 Details : the performance loss is caused by using of write barriers in the
3572 ext3 code. The SLES10 kernel turns barrier support on by
3573 default. The fix is to undo that change for ldiskfs.
3575 ------------------------------------------------------------------------------
3577 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
3579 * Support for kernels:
3580 2.6.9-42.0.3EL (RHEL 4)
3581 2.6.5-7.276 (SLES 9)
3582 2.4.21-47.0.1.EL (RHEL 3)
3583 2.6.12.6 vanilla (kernel.org)
3587 Frequency : quota enabled and large files being deleted
3589 Description: releasing more than 4GB of quota at once hangs OST
3590 Details : If a user deletes more than 4GB of files on a single OST it
3591 will cause the OST to spin in an infinite loop. Release
3592 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3597 Description: statfs data retrieved from /proc may be stale or zero
3598 Details : When reading per-device statfs data from /proc, in the
3599 {kbytes,files}_{total,free,avail} files, it may appear
3600 as zero or be out of date.
3603 Frequency : systems with MD RAID1 external journal devices
3605 Description: lconf's call to blkid is confused by RAID1 journal devices
3606 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3607 of returning all block devices that match the journal UUID.
3610 Frequency : always, for aggregate stripe size over 4GB
3612 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
3613 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
3614 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
3617 Frequency : always if "lfs find" used on a local file/directory
3619 Description: "lfs find" segfaults if used on a local file/directory
3620 Details : The case where a directory component was not specified wasn't
3621 handled correctly. Handle this properly.
3624 Frequency : always on ppc64
3626 Description: the write to an ext3 filesystem mounted with mballoc got stuck
3627 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3628 perform endianness conversion.
3631 Frequency : rarely (truncate to non-zero file size after write under load)
3632 Bugzilla : 10730, 10687
3633 Description: Files padded with zeros to next 4K multiple
3634 Details : With filesystems mounted using the "extents" option (2.6 kernels)
3635 it is possible that files that are truncated to a non-zero size
3636 immediately after being written are filled with zero bytes beyond
3637 the truncated size. No file data is lost.
3639 Severity : enhancement
3641 Description: Allow recovery/failover for liblustre clients.
3642 Details : liblustre clients were unaware of failover configurations until
3645 Severity : enhancement
3647 Description: user file locks should fail when not mounting with flock option
3648 Details : Set up an error-returning stub in ll_file_operations.lock field
3649 to prevent incorrect behaviour when client is mounted without
3650 flock option. Also, set up properly f_op->flock field for
3654 Frequency : always on ia64
3656 Description: "lfs df" loops on printing out MDS statfs information
3657 Details : The obd_ioctl_data was not initialized and in some systems
3658 this caused a failure during the ioctl that did not return
3659 an error. Initialize the struct and return an error on failure.
3662 Frequency : SLES 9 only
3664 Description: Error of copying files with lustre special EAs as root
3665 Details : Client side always return success for setxattr call for lustre
3666 special xattr (currently only "trusted.lov").
3669 Frequency : rarely on clusters with both ia64+i386 clients
3671 Description: ia64+i686 clients doing shared IO on the same file may LBUG
3672 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
3673 clients are doing concurrent writes to the same file it is
3674 possible that the ia64 clients may LASSERT because the OST
3675 extent locks are not PAGE_SIZE aligned. Ensure that grown
3676 locks are always aligned on the request boundary.
3679 Frequency : specific use, occasional
3681 Description: Overwriting in use executable truncates on-disk binary image
3682 Details : If one node attempts to overwrite an executable in use by
3683 another node, we now correctly return ETXTBSY instead of
3684 truncating the file.
3686 Severity : enhancement
3688 Description: Async OSC create to avoid the blocking unnecessarily.
3689 Details : If a OST has no remain object, system will block on the creating
3690 when need to create a new object on this OST. Now, ways use
3691 pre-created objects when available, instead of blocking on an
3692 empty osc while others are not empty. If we must block, we block
3693 for the shortest possible period of time.
3698 Description: chmod on Lustre root is propagated to other clients
3699 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
3700 invalid by the follow_mount time.
3703 Frequency : liblustre clients only
3705 Description: Race in 'instant cancel' lock handling could lead to such locks
3706 never to be granted in case of SMP MDS
3707 Details : Do not destroy not yet granted but cbpending locks in
3711 Frequency : replay/resend of open
3713 Description: non null lock assetion failure in mds_intent_policy
3714 Details : Trying to replay/resend lockless open requests resulted in
3715 mds_open() returning 0 with no lock. Now it sets a flag if
3716 a lock is going to be returned.
3718 Severity : enhancement
3720 Description: Checksum enhancements
3721 Details : New checksum enhancements allow for resending RPCs that failed
3724 Severity : enhancement
3726 Description: Tunables on number of dirty pages in cacche
3727 Details : Allow to set limit on number of dirty pages cached.
3732 Description: client crash on unmount - lock still has references
3733 Details : In some error handling cases it was possible to leak a lock
3734 reference on a client while accessing a file. This was not
3735 harmful to the client during operation, but would cause the
3736 client to crash when the filesystem is unmounted.
3739 Frequency : specific case, rare
3741 Description: ETXTBSY on mds though file not in use by client
3742 Details : ETXTBSY is no longer incorrectly returned when attempting to
3743 chmod or chown a directory that the user previously tried to
3744 execute or a currently-executing binary.
3747 Frequency : extremely rare except on liblustre-based clients
3749 Description: Lustre space not freed when files are deleted
3750 Details : Clean up open-unlinked files after client eviction. Previously
3751 the unlink was skipped and the files remained as orphans.
3756 Description: OST failure "would be an LBUG" in waiting_locks_callback()
3757 Details : In some cases it was possible to send a blocking callback to
3758 a client doing a glimpse, even though that client didn't get
3759 a lock granted. When the glimpse lock is cancelled on the OST
3760 the freed lock is left on the waiting list and corrupted the list.
3763 Frequency : all core dumps
3765 Description: Broke core dumps to lustre
3766 Details : Negative dentry may be unhashed if parent does not have UPDATE
3767 lock, but some callers, e.g. do_coredump, expect dentry to be
3768 hashed after successful create, hash it in ll_create_it.
3770 ------------------------------------------------------------------------------
3772 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
3774 * Support for kernels:
3775 2.6.9-42.0.2.EL (RHEL 4)
3776 2.6.5-7.276 (SLES 9)
3777 2.4.21-40.EL (RHEL 3)
3778 2.6.12.6 vanilla (kernel.org)
3782 Frequency : always on RHEL 3
3784 Description: Number of open files grows over time
3785 Details : The number of open files grows over time, whether or not
3786 Lustre is started. This was due to a filp leak introduced
3787 by one of our kernel patches.
3789 ------------------------------------------------------------------------------
3791 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
3793 * Support for kernels:
3794 2.6.9-42.EL (RHEL 4)
3795 2.6.5-7.276 (SLES 9)
3796 2.4.21-40.EL (RHEL 3)
3797 2.6.12.6 vanilla (kernel.org)
3802 Bugzilla : 5719, 9635, 9792, 9684
3803 Description: OST (or MDS) trips assertions in (re)connection under heavy load
3804 Details : If a server is under heavy load and cannot reply to new
3805 connection requests before the client resends the (re)connect,
3806 the connection handling code can behave badly if two service
3807 threads are concurrently handing separate (re)connections from
3808 the same client. Add better locking to the connection handling
3809 code, and ensure that only a single connection will be processed
3810 for a given client UUID, even if the lock is dropped.
3812 Severity : enhancement
3814 Description: add TCP zero-copy support to kernel
3815 Details : Add support to the kernel TCP stack to allow zero-copy bulk
3816 sends if the hardware supports scatter-gather and checksumming.
3817 This allows socklnd to do client-write and server-read more
3818 efficiently and reduce CPU utilization from skbuf copying.
3821 Frequency : only if NFS exporting from client
3823 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
3824 Details : If setting attributes on a file created under NFS that had
3825 never been opened it would be possible to oops the client
3826 if the file had no objects.
3829 Frequency : always for liblustre
3831 Description: liblustre client does MDS+OSTs setattr RPC for each write
3832 Details : When doing a write from a liblustre client, the client
3833 incorrectly issued an RPC to the MDS and each OST the file was
3834 striped over in order to update the timestamps. When writing
3835 with small chunks and many clients this could overwhelm the MDS
3836 with RPCs. In all cases it would slow down the write because
3837 these RPCs are unnecessary.
3839 Severity : enhancement
3841 Description: allow number of MDS service threads to be changed at module load
3842 Details : It is now possible to change the number of MDS service threads
3843 running. Adding "options mds mds_num_threads={N}" to the MDS's
3844 /etc/modprobe.conf will set the number of threads for the next
3845 time Lustre is restarted (assuming the "mds" module is also
3846 reloaded at that time). The default number of threads will
3847 stay the same, 32 for most systems.
3852 Description: OST crash if filesystem is unformatted or corrupt
3853 Details : If an OST is started on a device that has never been formatted
3854 or if the filesystem is corrupt and cannot even mount then the
3855 error handling cleanup routines would dereference a NULL pointer.
3860 Description: NULL pointer deref in llap_from_page.
3861 Details : get_cache_page_nowait can return a page with NULL (or otherwise
3862 incorrect) mapping if the page was truncated/reclaimed while it was
3863 searched for. Check for this condition and skip such pages when
3864 doing readahead. Introduce extra check to llap_from_page() to
3865 verify page->mapping->host is non-NULL (so page is not anonymous).
3868 Frequency : Sometimes when using sys_sendfile
3870 Description: "page not covered by a lock" warnings from ll_readpage
3871 Details : sendfile called ll_readpage without right page locks present.
3872 Now we introduced ll_file_sendfile that does necessary locking
3873 around call to generic_file_sendfile() much like we do in
3877 Frequency : with certain MDS communication failures at client mount time
3879 Description: NULL pointer deref after failed client mount
3880 Details : a client connection request may delayed by the network layer
3881 and not be sent until after the PTLRPC layer has timed out the
3882 request. If the client fails the mount immediately it will try
3883 to clean up before the network times out the request. Add a
3884 reference from the request import to the obd device and delay
3885 the cleanup until the network drops the request.
3888 Frequency : occasionally during client (re)connect
3890 Description: assertion failure during client (re)connect
3891 Details : processing a client connection request may be delayed by the
3892 client or server longer than the client connect timeout. This
3893 causes the client to resend the connection request. If the
3894 original connection request is replied in this interval, the
3895 client may trip an assertion failure in ptlrpc_connect_interpret()
3896 which thought it would be the only running connect process.
3899 Frequency : only with obd_echo servers and clients that are rebooted
3901 Description: kernel BUG accessing uninitialized data structure
3902 Details : When running an obd_echo server it did not start the ping_evictor
3903 thread, and when a client was evicted an uninitialized data
3904 structure was accessed. Start the ping_evictor in the RPC
3905 service startup instead of the OBD startup.
3907 Severity : enhancement
3908 Bugzilla : 10193 (patchless)
3909 Description: Remove dependency on various unexported kernel interfaces.
3910 Details : No longer need reparent_to_init, exit_mm, exit_files,
3911 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
3914 Frequency : rare (only users of deprecated and unsupported LDAP config)
3916 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
3917 Details : LDAP apparently contains 'lustreName' attributes instead of
3918 'name'. A simple remapping of the name is sufficient.
3921 Frequency : rare (only with non-default dump_on_timeout debug enabled)
3923 Description: waiting_locks_callback trips kernel BUG if client is evicted
3924 Details : Running with the dump_on_timeout debug flag turned on makes
3925 it possible that the waiting_locks_callback() can try to dump
3926 the Lustre kernel debug logs from an interrupt handler. Defer
3927 this log dumping to the expired_lock_main() thread.
3929 Severity : enhancement
3931 Description: Support NFS exporting on 2.6 kernels.
3932 Details : Implement non-rawops metadata methods for NFS server to use without
3933 changing NFS server code.
3936 Frequency : very rare (synthetic metadata workload only)
3938 Description: two racing renames might cause an MDS thread to deadlock
3939 Details : Running the "racer" program may cause one MDS thread to rename
3940 a file from being the source of a rename to being the target of
3941 a rename at exactly the same time that another thread is doing
3942 so, and the second thread has already enqueued these locks after
3943 doing a lookup of the target and is trying to relock them in
3944 order. Ensure that we don't try to re-lock the same resource.
3947 Frequency : only very large systems with liblustre clients
3949 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
3950 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
3951 all OSTs in parallel. This allows the eviction of stale liblustre
3952 clients to proceed much faster than if they were done in series,
3953 and also offers similar improvements for other set_info RPCs.
3958 Description: excessive CPU usage during initial read phase on client
3959 Details : During the initial read phase on a client, it would agressively
3960 retry readahead on the file, consuming too much CPU and impacting
3961 performance (since 1.4.5.8). Improve the readahead algorithm
3962 to avoid this, and also improve some other common cases (read
3963 of small files in particular, where "small" is files smaller than
3964 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
3969 Description: MDS crash when receiving packet with unknown intent.
3970 Details : Do not LBUG in unknown intent case, just return -EFAULT
3972 Severity : enhancement
3973 Bugzilla : 9293, 9385
3974 Description: MDS RPCs are serialised on client. This is unnecessary for some.
3975 Details : Do not serialize getattr (non-intent version) and statfs.
3978 Frequency : occasional, when OST network is overloaded/intermittent
3980 Description: client evicted by OST after bulk IO timeout
3981 Details : If a client sends a bulk IO request (read or write) the OST
3982 may evict the client if it is unresposive to its data GET/PUT
3983 request. This is incorrect if the network is overloaded (takes
3984 too long to transfer the RPC data) or dropped the OST GET/PUT
3985 request. There is no need to evict the client at all, since
3986 the pinger and/or lock callbacks will handle this, and the
3987 client can restart the bulk request.
3990 Frequency : Always when mmapping file with no objects
3992 Description: client crashes when mmapping file with no objects
3993 Details : Check that we actually have objects in a file before doing any
3994 operations on objects in ll_vm_open, ll_vm_close and
4000 Description: Request leak when working with deleted CWD
4001 Details : Introduce advanced request refcount tracking for requests
4002 referenced from lustre intent.
4004 Severity : Enhancement
4006 Description: Cache open file handles on client.
4007 Details : MDS now will return special lock along with openhandle, if
4008 requested and client is allowed to hold openhandle, even if unused,
4009 until such a lock is revoked. Helps NFS a lot, since NFS is opening
4010 closing files for every read/write openration.
4012 Severity : Enhancement
4014 Description: Cache open negative dentries on client when possible.
4015 Details : Guard negative dentries with UPDATE lock on parent dir, drop
4016 negative dentries on lock revocation.
4021 Description: Remounting a client read-only wasn't possible with a zconf mount
4022 Details : It wasn't possible to remount a client read-only with llmount.
4024 Severity : enhancement
4025 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
4026 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
4027 MPICH is included to add Lustre-specific ADIO interfaces.
4028 This is based closely on the UFS ADIO layer and only differs
4029 in file creation, in order to allow the OST striping to be set.
4030 This is user-contributed code and not supported by CFS.
4035 Description: extended inode attributes (immutable, append-only) work improperly
4036 when 2.4 and 2.6 kernels are used on client/server or vice versa
4037 Details : Introduce kernel-independent values for these flags.
4039 Severity : enhancement
4042 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
4043 Details : Many of the /proc/ tunables can only be tuned at a megabyte
4044 granularity. Now, Fractional MB granularity is be supported,
4045 this is very useful for low memory system.
4047 Severity : enhancement
4049 Description: Getattr by fid
4050 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
4051 locks, avoids extra getattr rpc requests to MDS, allows '/' to
4052 have locks and avoids getattr rpc requests for it on every stat.
4055 Frequency : Always, for filesystems larger than 2TB
4057 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4058 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
4059 filesystems larger than 2TB to be mounted with mballoc enabled.
4062 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
4064 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4065 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
4066 filesystems larger than 2TB is mounted then the kernel will
4067 silently corrupt the start of the filesystem. CONFIG_LBD is
4068 enabled for all CFS-supported kernels, but the possibility of
4069 this happening with a modified kernel config exists.
4071 Severity : enhancement
4073 Description: add client O_DIRECT support for 2.6 kernels
4074 Details : It is now possible to do O_DIRECT reads and writes to files
4075 in the Lustre client mountpoint on 2.6 kernel clients.
4077 Severity : enhancement
4079 Description: parallel glimpse, setattr, statfs, punch, destroy requests
4080 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
4081 parallel, not waiting for response from every OST before sending
4082 a rpc to the next OST.
4087 Description: setattr vs write race when updating file timestamps
4088 Details : Client processes that update a file timestamp into the past
4089 right after writing to the file (e.g. tar) it is possible that
4090 the updated file modification time can be reset to the current
4091 time due to a race between processing the setattr and write RPC.
4093 Severity : enhancement
4095 Description: Bring 'lfs find' closer in line with regular Linux find.
4096 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
4097 -print0 options and obtains all the needed info through the lustre
4100 Severity : enhancement
4102 Description: support up to 1024 configured devices on one node
4103 Details : change obd_dev array from statically allocated to dynamically
4104 allocated structs as they are first used to reduce memory usage
4109 Description: Flush dirty partially truncated pages during truncate
4110 Details : Immediatelly flush partially truncated pages in filter_setattr,
4111 this way we completely avoid having any pages in page cache on OST
4112 and can retire ugly workarounds during writes to flush such pages.
4117 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
4118 Details : protect inode from truncation within vfs_unlink() context
4119 just take a reference before calling vfs_unlink() and release it
4120 when parent's i_sem is free.
4125 Description: last_id value checked outside lock on OST caused LASSERT failure
4126 Details : If there were multiple MDS->OST object precreate requests in
4127 flight, it was possible that the OST's last object id was checked
4128 outside a lock and incorrectly tripped an assertion. Move checks
4129 inside locks, and discard old precreate requests.
4132 Frequency : always, if extents are used on OSTs
4134 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
4135 Details : OSTs using the extents format would not zero the high 16 bits of
4136 the index physical block number. This is not a problem for any
4137 OST filesystems smaller than 16TB, and no kernels support ext3
4138 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
4139 new/modified files) and can be fixed for existing filesystems
4140 with e2fsprogs-1.39-cfs1.
4145 Description: import connection selection may be incorrect if timer wraps
4146 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
4147 import connections to be ignored if the 32-bit jiffies counter
4148 wraps. Use a 64-bit jiffies counter.
4151 Frequency : very large clusters immediately after boot
4153 Description: LNET request buffers exhausted under heavy short-term load
4154 Details : If a large number of client requests are generated on a service
4155 that has previously never seen so many requests it is possible
4156 that the request buffer growth cannot keep up with the spike in
4157 demand. Instead of dropping incoming requests, they are held in
4158 the LND until the RPC service can accept more requests.
4161 Frequency : Sometimes during replay
4163 Description: Assertion failure in ll_local_open after replay.
4164 Details : If replay happened on an open request reply before we were able
4165 to set replay handler, reply will become not swabbed tripping the
4166 assertion in ll_local_open. Now we set the handler right after
4167 recognising of open request
4170 Frequency : very rare
4172 Description: kernel reports "badness in vsnprintf"
4173 Details : Reading from the "recovery_status" /proc file in small chunks
4174 may cause a negative length in lprocfs_obd_rd_recovery_status()
4175 call to vsnprintf() (which is otherwise harmless). Exit early
4176 if there is no more space in the output buffer.
4178 Severity : enhancement
4180 Description: clear OBD RPC statistics by writing to them
4181 Details : It is now possible to clear the OBD RPC statistics by writing
4182 to the "stats" file.
4187 Description: Client mtime is not the same on different clients after utimes
4188 Details : In some cases, the client was using the utimes() syscall on
4189 a file cached on another node. The clients now validate the
4190 ctime from the MDS + OSTs to determine which one is right.
4195 Description: Inability to activate failout mode
4196 Details : lconf script incorrectly assumed that in pythong string's numeric
4197 value is used in comparisons.
4200 Frequency : always with multiple stripes per file
4202 Description: Inefficient object allocation for mutli-stripe files
4203 Details : When selecting which OSTs to stripe files over, for files with
4204 a stripe count that divides evenly into the number of OSTs,
4205 the MDS is always picking the same starting OST for each file.
4206 Return the OST selection heuristic to the original design.
4211 Description: mount failures may take full timeout to return an error
4212 Details : Under some heavy load conditions it is possible that a
4213 failed mount can wait for the full obd_timeout interval,
4214 possibly several minutes, before reporting an error.
4215 Instead return an error as soon as the status is known.
4217 Frequency : quota enabled and large files being deleted
4219 Description: releasing more than 4GB of quota at once hangs OST
4220 Details : If a user deletes more than 4GB of files on a single OST it
4221 will cause the OST to spin in an infinite loop. Release
4222 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4227 Description: statfs data retrieved from /proc may be stale or zero
4228 Details : When reading per-device statfs data from /proc, in the
4229 {kbytes,files}_{total,free,avail} files, it may appear
4230 as zero or be out of date.
4233 Frequency : systems with MD RAID1 external journal devices
4235 Description: lconf's call to blkid is confused by RAID1 journal devices
4236 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4237 of returning all block devices that match the journal UUID.
4240 Frequency : always, for aggregate stripe size over 4GB
4242 Description: assertion fails when trying to use 4GB stripe size
4243 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
4244 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
4247 Frequency : always on ppc64
4249 Description: the first write on an ext3 filesystem with mballoc got stuck
4250 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4251 perform endianness conversion.
4253 ------------------------------------------------------------------------------
4255 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4257 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
4258 INTEROPERATE with older versions automatically. Please read the
4259 user documentation before upgrading any part of a live system.
4260 * WARNING: Lustre networking configuration changes are required with
4261 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
4264 * Support for kernels:
4265 2.6.9-22.0.2.EL (RHEL 4)
4266 2.6.5-7.244 (SLES 9)
4267 2.6.12.6 vanilla (kernel.org)
4270 Severity : enhancement
4271 Bugzilla : 7981/8208
4272 Description: Introduced Lustre Networking (LNET)
4273 Details : LNET is new networking infrastructure for Lustre, it includes
4274 a reorganized network configuration mode (see the user
4275 documentation for full details) as well as support for routing
4276 between different network fabrics. Lustre Networking Devices
4277 (LNDs) for the supported network fabrics have also been
4278 created for this new infrastructure.
4280 Severity : enhancement
4281 Description: Introduced Access control lists
4282 Details : clients can set ACLs on files and directories in order to have
4283 more fine-grained permissions than the standard Unix UGO+RWX.
4284 The MDS must be started with the "-o acl" mount option.
4286 Severity : enhancement
4287 Description: Introduced filesystem quotas
4288 Details : Administrators may now establish per-user quotas on the
4291 Severity : enhancement
4293 Description: Configuration change for the XT3
4294 The PTLLND is now used to run Lustre over Portals on the XT3
4295 The configure option(s) --with-cray-portals are no longer used.
4296 Rather --with-portals=<path-to-portals-includes> is used to
4297 enable building on the XT3. In addition to enable XT3 specific
4298 features the option --enable-cray-xt3 must be used.
4303 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
4304 Details : A race between allocating a new llcd and re-getting the llcd_lock
4305 allowed another thread to grab newly-allocated llcd.
4307 Severity : enhancement
4309 Description: 2.6 OST async journal commit and locking fix to improve performance
4310 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
4311 2.6 kernels are now async as they already were in 2.4 kernels so
4312 that they can commit concurrently with the network bulk transfer.
4313 For block-allocated files the filter allocation semaphore is held
4314 to avoid filesystem fragmentation during allocation. BKL lock
4315 removed for 2.6 xattr operations where it is no longer needed.
4320 Description: lconf incorrectly determined whether two IP networks could talk
4321 Details : In some more complicated routing and multiple-network
4322 configurations, lconf will avoid trying to make a network
4323 connection to a disjoint part of the IP space. It was doing the
4324 math incorrectly for one set of cases.
4329 Description: Fix for potential infinite loop processing records in an llog.
4330 Details : If an llog record is corrupted/zeroed, it is possible to loop
4331 forever in llog_process(). Validate the llog record length
4332 and skip the remainder of the block on error.
4335 Frequency : occasional (liblustre only)
4337 Description: liblustre could not open files whose last component is a symlink
4338 Details : sysio_path_walk() would incorrectly pass the open intent to
4339 intermediate path components.
4342 Frequency : rare (liblustre only with non-standard tuning)
4343 Bugzilla : 7201 (7350)
4344 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4345 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4346 after it has been released, only do this for referenced locks
4348 Severity : enhancement
4350 Description: specifying an (invalid) directory default stripe_size of -1
4351 would reset the directory default striping
4352 Details : stripe_size -1 was used internally to signal directory stripe
4353 removal, now use "all default" to signal dir stripe removal
4354 as a directory striping of "all default" is not useful
4357 Frequency : common for large clusters running liblustre clients
4359 Description: doing an ls when liblustre clients are running is slow
4360 Details : sending a glimpse AST to a liblustre client waits for every AST
4361 to time out, as liblustre clients will not respond. Since they
4362 cannot cache data we refresh the OST lock LVB from disk instead.
4364 Severity : enhancement
4366 Description: doing an ls at the same time as file IO can be slow
4367 Details : enqueue and other "small" requests can be blocked behind many
4368 large IO requests. Create a new OST IO portal for non-IO
4369 requests so they can be processed faster.
4372 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4374 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4375 created file with mode 000
4376 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4377 allow an ftruncate() as a non-root user to a file with mode 000.
4378 Lustre now respects this flag to disable mode checks when
4379 truncating a file owned by the user
4382 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4385 Description: Revoking locks from clients that went dead or catatonic might take
4387 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4388 cancellation of such locks instant on servers without waiting for
4389 any reply from clients. Clients drops these locks when cancel
4390 notification from server is received without replying.
4393 Frequency : liblustre-only, when liblustre client dies or becomes busy
4395 Description: Doing ls on Linux clients can take a long time with active
4397 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4398 granting such locks to it in the first place if possible. Locks
4399 are taken by proxy on the OST during the read or write and
4400 dropped immediately afterward. Add connect flags handling, do
4401 not grant locks to liblustre clients for glimpse ASTs.
4403 Severity : enhancement
4405 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4406 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4407 sequential reads and behaves badly with applications doing random
4408 reads. Improve it by reading ahead at least read region, and
4409 avoiding excessive large RPC for small reads.
4411 Severity : enhancement
4413 Description: Creating more than 1000 files for a single job may cause a load
4414 imbalance on the OSTs if there are also a large number of OSTs.
4415 Details : qos_prep_create() uses an OST index reseed value that is an
4416 even multiple of the number of available OSTs so that if the
4417 reseed happens in the middle of the object allocation it will
4418 still utilize the OSTs as uniformly as possible.
4423 Description: OST or MDS may oops in ping_evictor_main()
4424 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
4425 but doesn't restart at beginning of obd_exports_timed list
4428 Severity : enhancement
4430 Description: improve by-nid export eviction on the MDS and OST
4431 Details : allow multiple exports with the same NID to be evicted at one
4432 time without re-searching the exports list.
4435 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
4437 Description: MDS may oops in groups_free()
4438 Details : in rare race conditions a newly allocated group_info struct is
4439 freed again, and this can be NULL. The 2.4 compatibility code
4440 for groups_free() checked for a NULL pointer, but 2.6 did not.
4443 Frequency : common for liblustre clients doing little filesystem IO
4444 Bugzilla : 9352, 7313
4445 Description: server may evict liblustre clients accessing contended locks
4446 Details : if a client is granted a lock or receives a completion AST
4447 with a blocking AST already set it would not reply to the AST
4448 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
4451 Frequency : lfs setstripe, only systems with more than 160 OSTs
4453 Description: unable to set striping with a starting offset beyond OST 160
4454 Details : llapi_create_file() incorrectly limited the starting stripe
4455 index to the maximum single-file stripe count.
4458 Frequency : LDAP users only
4460 Description: lconf did not handle in-kernel recovery with LDAP properly
4461 Details : lconf/LustreDB get_refs() is searching the wrong namespace
4463 Severity : enhancement
4465 Description: bind OST threads to NUMA nodes to improve performance
4466 Details : all OST threads are uniformly bound to CPUs on a single NUMA
4467 node and do their allocations there to localize memory access
4469 Severity : enhancement
4471 Description: llmount can determine client NID directly from Myrinet (GM)
4472 Details : the client NID code from gmnalnid was moved directly into
4473 llmount, removing the need to use this or specifying the
4474 client NID explicitly when mounting GM clients with zeroconf
4477 Frequency : if client is started with down MDS
4479 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
4480 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
4481 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
4482 ptlrpc_import_delayed_req() because we didn't check if the
4483 request was interrupted, and we also didn't break out of the
4484 event loop if there was no timeout
4489 Description: data loss during non-page-aligned writes to a single file from
4490 both multiple nodes and multiple threads on one node at same time
4491 Details : updates to KMS and lsm weren't protected by common lock. Resulting
4492 inconsistency led to false short-reads, that were cached and later
4493 used by ->prepare_write() to fill in partially written page,
4494 leading to data loss.
4497 Frequency : always, if lconf --abort_recovery used
4499 Description: lconf --abort_recovery fails with 'Operation not supported'
4500 Details : lconf was attempting to abort recovery on the MDT device and not
4503 Severity : enhancement
4505 Description: remove cleanup logs
4506 Details : replace lconf-generated cleanup logs with lustre internal
4507 cleanup routines. Eliminates the need for client-cleanup and
4510 Severity : enhancement
4512 Description: add support for EAs (user and system) on lustre filesystems
4513 Details : it is now possible to store extended attributes in the Lustre
4514 client filesystem, and with the user_xattr mount option it
4515 is possible to allow users to store EAs on their files also
4517 Severity : enhancement
4519 Description: Add possibility (config option) to show minimal available OST free
4521 Details : When compiled with --enable-mindf configure option, statfs(2)
4522 (and so, df) will return least minimal free space available from
4523 all OSTs as amount of free space on FS, instead of summary of
4524 free spaces of all OSTs.
4526 Severity : enhancement
4528 Description: do not expand extent locks acquired on OST-side
4529 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
4530 by server: they are not cached anyway.
4533 Frequency : when mmap is used/binaries executed from Lustre
4535 Description: Unmmap pages before throwing them away from read cache.
4536 Details : llap_shrink cache now attempts to unmap pages before discarding
4537 them (if unmapping failed - do not discard). SLES9 kernel has
4538 extra checks that trigger if this unmapping is not done first.
4543 Description: lconf didn't resolve symlinks before checking to see whether a
4544 given mountpoint was already in use
4547 Frequency : when migrating failover services
4548 Bugzilla : 6395, 9514
4549 Description: When migrating a subset of services from a node (e.g. failback
4550 from a failover service node) the remaining services would
4551 time out and evict clients.
4552 Details : lconf --force (implied by --failover) sets the global obd_timeout
4553 to 5 seconds in order to quickly disconnect, but this caused
4554 other RPCs to time out too quickly. Do not change the global
4555 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
4557 Severity : enhancement
4558 Frequency : if MDS is started with down OST
4559 Bugzilla : 9439,5706
4560 Description: Allow startup/shutdown of an MDS without depending on the
4561 availability of the OSTs.
4562 Details : Asynchronously call mds_lov_synchronize during MDS startup.
4563 Add appropriate locking and lov-osc refcounts for safe
4564 cleaning. Add osc abort_inflight calls in case the
4565 synchronize never started.
4568 Frequency : occasional (Cray XT3 only)
4570 Description: root not authorized to access files in CRAY_PORTALS environment
4571 Details : The client process capabilities were not honoured on the MDS in
4572 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
4573 been accessed by an authorized user then root was able to access
4574 the file on the local client also. The root user capabilities
4575 are now allowed on the MDS, as this environment has secure UID.
4578 Frequency : occasional
4580 Description: ldiskfs "too long searching" message happens too often
4581 Details : A debugging message (otherwise harmless) prints too often on
4582 the OST console. This has been reduced to only happen when
4583 there are fragmentation problems on the filesystem.
4588 Description: Division by zero in statfs when all OSCs are inactive
4589 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
4590 lov_statfs divides by value returned by lov_get_stripecnt().
4594 Bugzilla : 9489, 3273
4595 Description: First write from each client to each OST was only 4kB in size,
4596 to initialize client writeback cache, which caused sub-optimal
4597 RPCs and poor layout on disk for the first writen file.
4598 Details : Clients now request an initial cache grant at (re)connect time
4599 and so that they can start streaming writes to the cache right
4600 away and always do full-sized RPCs if there is enough data.
4601 If the OST is rebooted the client also re-establishes its grant
4602 so that client cached writes will be honoured under the grant.
4607 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
4608 Details : Now I/O RPCs go to different portal number and (presumably) fast
4609 lock requests (and glimses) and other RPCs get their own service
4610 threads pool that should be able to service those RPCs
4613 Severity : enhancement
4615 Description: Ability to exchange lustre version between client and servers and
4616 issue warnings at client side if client is too old. Also for
4617 liblustre clients there is ability to refuse connection of too old
4619 Details : New 'version' field is added to connect data structure that is
4620 filled with version info. That info is later checked by server and
4624 Frequency : rare, liblustre only.
4625 Bugzilla : 9296, 9581
4626 Description: Two simultaneous writes from liblustre at offset within same page
4627 might proceed at the same time overwriting eachother with stale
4629 Details : I/O lock withing llu_file_prwv was released too early, before data
4630 actually was hitting the wire. Extended lock-holding time until
4631 server acknowledges receiving data.
4634 Frequency : extremely rare. Never observed in practice.
4636 Description: avoid generating lustre_handle cookie of 0.
4637 Details : class_handle_hash() generates handle cookies by incrementing
4638 global counter, and can hit 0 occasionaly (this is unlikely, but
4639 not impossible, because initial value of cookie counter is
4640 selected randonly). Value of 0 is used as a sentinel meaning
4641 "unassigned handle" --- avoid it. Also coalesce two critical
4642 sections in this function into one.
4644 Severity : enhancement
4646 Description: allow liblustre clients to delegate truncate locking to OST
4647 Details : To avoid overhead of locking, liblustre client instructs OST to
4648 take extent lock in ost_punch() on client's behalf. New connection
4649 flag is added to handle backward compatibility.
4651 Severity : enhancement
4652 Bugzilla : 4928, 7341, 9758
4653 Description: allow number of OST service threads to be specified
4654 Details : a module parameter allows the number of OST service threads
4655 to be specified via "options ost ost_num_threads={N}" in the
4656 OSS's /etc/modules.conf or /etc/modprobe.conf.
4660 Bugzilla : 6146, 9635, 9895
4661 Description: servers crash with bad pointer in target_handle_connect()
4662 Details : In rare cases when a client is reconnecting it was possible that
4663 the connection request was the last reference for that export.
4664 We would temporarily drop the export reference and get a new
4665 one, but this may have been the last reference and the export
4666 was just destroyed. Get new reference before dropping old one.
4668 Severity : enhancement
4669 Frequency : if client is started with failover MDS
4671 Description: Allow multiple MDS hostnames in the mount command
4672 Details : Try to read the configuration from all specified MDS
4673 hostnames during a client mount in case the "primary"
4676 Severity : enhancement
4678 Description: Stop sending data to evicted clients as soon as possible.
4679 Details : Check if the client we are about to send or are sending data to
4680 was evicted already. (Check is done every second of waiting,
4681 for which l_wait_event interface was extended to allow checking
4682 of exit condition at specified intervals).
4685 Frequency : rare, normally only when NFS exporting is done from client
4687 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
4689 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
4692 Frequency : always, if srand() is called before liblustre initialization
4694 Description: Liblustre uses system PRNG disturbing its usage by user application
4695 Details : Introduce internal to lustre fast and high-quality PRNG for
4696 lustre usage and make liblustre and some other places in generic
4697 lustre code to use it.
4699 Severity : enhancement
4700 Bugzilla : 9477, 9557, 9870
4701 Description: Verify that the MDS configuration logs are updated when xml is
4702 Details : Check if the .xml configuration logs are newer than the config
4703 logs stored on the MDS and report an error if this is the case.
4704 Request --write-conf, or allow starting with --old_conf.
4706 Severity : enhancement
4708 Description: Handle symlinks in the path when checking if Lustre is mounted.
4709 Details : Resolve intermediate symlinks when checking if a client has
4710 mounted a filesystem to avoid duplicate client mounts.
4715 Description: lconf can hit an error exception but still return success.
4716 Details : The lconf command catches the Command error exception at the top
4717 level script context and will exit with the associated exit
4718 status, but doesn't ensure that this exit status is non-zero.
4723 Description: failure of ptlrpc thread startup can cause oops
4724 Details : Starting a ptlrpc service thread can fail if there are a large
4725 number of threads or the server memory is very fragmented.
4726 Handle this without oopsing.
4729 Frequency : always, only if liblustre and non-default acceptor port was used
4731 Description: liblustre cannot connect to servers with non-default acceptor port
4732 Details : tcpnal_set_default_params() was not called and was therefore
4733 ignoring the environment varaible TCPNAL_PORT, as well as other
4734 TCPNAL_ environment variables
4739 Description: two objects could be created on the same OST for a single file
4740 Details : If an OST is down, in some cases it was possible to create two
4741 objects on a single OST for a single file. No problems other
4742 than potential performance impact and spurious error messages.
4746 Bugzilla : 5681, 9562
4747 Description: Client may oops in ll_unhash_aliases
4748 Details : Client dcache may become inconsistent in race condition.
4749 In some cases "getcwd" can fail if the current directory is
4755 Description: Inode refcounting problems in NFS export code
4756 Details : link_raw functions used to call d_instantiate without obtaining
4757 extra inode reference first.
4761 Bugzilla : 9942, 9903
4762 Description: Referencing freed requests leading to crash, memleaks with NFS.
4763 Details : We used to require that call to ll_revalidate_it was always
4764 followed by ll_lookup_it. Also with revalidate_special() it is
4765 possible to call ll_revalidate_it() twice for the same dentry
4766 even if first occurence returned success. This fix changes semantic
4767 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
4768 reference on a request referred from the intent.
4769 ll_intent_release() then releases such a request.
4772 Frequency : rare, normally benchmark loads only
4774 Description: unlinked inodes were kept in memory on the client
4775 Details : If a client is repeatedly creating and unlinking files it
4776 can accumulate a lot of stale inodes in the inode slab cache.
4777 If there is no other client load running this can cause the
4778 client node to run out of memory. Instead flush old inodes
4779 from client cache that have the same inode number as a new inode.
4782 Frequency : SLES9 2.6.5 kernel and long filenames only
4783 Bugzilla : 9969, 10379
4784 Description: utime reports stale NFS file handle
4785 Details : SLES9 uses out-of-dentry names in some cases, which confused
4786 the lustre dentry revalidation. Change it to always use the
4790 Frequency : rare, unless heavy write-truncate concurrency is continuous
4791 Bugzilla : 4180, 6984, 7171, 9963, 9331
4792 Description: OST becomes very slow and/or deadlocked during object unlink
4793 Details : filter_destroy() was holding onto the parent directory lock
4794 while truncating+unlinking objects. For very large objects this
4795 may block other threads for a long time and slow overall OST
4796 responsiveness. It may also be possible to get a lock ordering
4797 deadlock in this case, or run out of journal credits because of
4798 the combined truncate+unlink. Solution is to do object truncate
4799 first in one transaction without parent lock, and then do the
4800 final unlink in a new transaction with the parent lock. This
4801 reduces the lock hold time dramatically.
4804 Frequency : rare, 2.4 kernels only
4806 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
4807 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
4808 before unmounting their filesystem in order to avoid deadlock.
4809 The kernel_locked() function in 2.4 kernels only checks whether
4810 the kernel lock is held, not whether it is this process that is
4811 holding it as 2.6 kernels do.
4816 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
4817 Details : The client ptlrpc code may be trying to reconnect to a down
4818 server before a previous connection attempt has timed out.
4819 Increase the reconnect interval to be longer than the connection
4820 timeout interval to avoid sending duplicate connections to
4824 Frequency : echo_client brw_test command
4826 Description: fix echo_client to work with OST preallocated code
4827 Details : OST preallocation code (5137) didn't take echo_client IO path
4828 into account: echo_client calls filter methods outside of any
4829 OST thread and, hence, there is no per-thread preallocated
4830 pages and buffers to use. Solution: hijack pga pages for IO. As
4831 a byproduct, this avoids unnecessary data copying.
4835 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
4836 Description: Client can oops in mdc_commit_close() after open replay
4837 Details : It was possible for the MDS to return an open request with no
4838 transaction number in mds_finish_transno() if the client was
4839 evicted, but without actually returning an error. Clients
4840 would later try to replay that open and may trip an assertion
4841 Simplify the client close codepath, and always return an error
4842 from the MDS in case the open is not successful.
4845 Frequency : rare, 2.6 OSTs only
4847 Description: OST may deadlock under high load on fragmented files
4848 Details : If there was a heavy load and highly-fragmented OST filesystems
4849 it was possible to have all the OST threads deadlock waiting on
4850 allocation of biovecs, because the biovecs were not released
4851 until the entire RPC IO was completed. Instead, release biovecs
4852 as soon as they are complete to ensure forward IO progress.
4854 Severity : enhancement
4856 Description: Support for specifying external journal device at mount
4857 Details : If an OST or MDS device is formatted with an external journal
4858 device, this device major/minor is stored in the ext3 superblock
4859 and may not be valid for failover. Allow detecting and
4860 specifying the external journal at mount time.
4865 Description: Mounting an MDS with pending unlinked files may cause oops
4866 Details : target_finish_recovery() calls mds_postrecov() which returned
4867 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
4868 considers this an error and immediately begins cleaning up the
4869 lov, just after starting the mds_lov process
4871 Severity : enhancement
4873 Description: Implement 'lfs df' to report actual free space on per-OST basis
4874 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
4875 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
4876 usage of objects; '-h' to report in human readable format.
4878 ------------------------------------------------------------------------------
4880 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
4887 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
4888 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
4889 at startup during orphan unlinking. Instead, skip buddy update
4890 before setup, as it will be regenerated after recovery is complete.
4895 Description: If an OST is inactive, its locks might reference stale inodes.
4896 Details : lov_change_cbdata() must iterate over all namespaces, even if
4897 they are inactive to clear inode references from the lock.
4899 Severity : enhancement
4900 Frequency : occasional, if non-standard max_dirty_mb used
4902 Description: Client will block write RPCs if not enough grant
4903 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
4904 then the client will block writes while waiting for another RPC
4905 to complete instead of consuming its dirty limit. With change
4906 we get improved performance when max_dirty_mb is small.
4908 Severity : enhancement
4909 Bugzilla : 3389, 6253
4910 Description: Add support for supplementary groups on the MDS.
4911 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
4912 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
4913 lookups for user supplementary groups into a cache.
4917 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
4918 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
4919 for liblustre O_EXCL is enforced by MDS.
4923 Description: readdir never returns NULL in liblustre.
4924 Details : Corrected llu_iop_getdirentries logic, to return offset of next
4925 dentry in struct dirent.
4929 Frequency : liblustre only, depends on application IO pattern
4930 Description: liblustre clients evicted if not contacting servers
4931 Details : Don't put liblustre clients into the ping_evictor list, so
4932 they will not be evicted by the pinger ever.
4934 Severity : enhancement
4936 Description: Add ability to evict clients by NID from MDS.
4937 Details : By echoing "nid:$NID" string into
4938 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
4939 $NID would be instantly evicted from this MDS and from all active
4940 OSTs connected to it.
4944 Description: Do not query file size twice, somewhat slowing stat(2) calls.
4945 Details : lookup_it_finish() used to query file size from OSTs that was not
4950 Description: service threads change working directory to that of init
4951 Details : Starting lustre service threads may pin the working directory
4952 of the parent thread, making that filesystem busy. Threads
4953 now change to the working directory of init to avoid this.
4957 Frequency : during shutdown only
4958 Description: shutdown with a failed MDS or OST can cause unmount to hang
4959 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
4964 Frequency : 2.6 only
4965 Description: chmod/chown may include an extra supplementary group
4966 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
4967 supplementary group and if none is specified this is used.
4970 Bugzilla : 5479 (6816)
4972 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
4973 Details : If lookup is in progress on a file that is unlinked we might try
4974 to revalidate the inode and fail in revalidate after lookup is
4975 complete and ll_file_open() enqueues the open again but
4976 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
4979 Frequency : always, if lconf --abort_recovery used
4981 Description: lconf --abort_recovery fails with 'Operation not supported'
4982 Details : lconf was attempting to abort recovery on the MDT device and not
4985 ------------------------------------------------------------------------------
4987 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
4992 Frequency : rare (only unsupported configurations with a node running as an
4994 Bugzilla : 6514, 5137
4995 Description: Mounting a Lustre file system on a node running as an OST could
4997 Details : OSTs now preallocates memory needed to write out data at
4998 startup, instead of when needed, to avoid having to
4999 allocate memory in possibly low memory situations.
5000 Specifically, if the file system is mounted on on OST,
5001 memory pressure could force it to try to write out data,
5002 which it needed to allocate memory to do. Due to the low
5003 memory, it would be unable to do so and the node would
5004 become unresponsive.
5006 Severity : enhancement
5008 Description: Addition of lconf --service command line option
5009 Details : lconf now accepts a '--service <arg>' option, which is
5010 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
5012 Severity : enhancement
5014 Description: Failover mode is now the default for OSTs.
5015 Details : By default, OSTs will now run in failover mode. To return to
5016 the old behaviour, add '--failout' to the lmc line for OSTs.
5018 Severity : enhancement
5020 Description: Health checks are now provided for MDS and OSTs
5021 Details : Additional detailed health check information on MSD and OSTs
5022 is now provided through the procfs health_check value.
5025 Frequency : occasional, depends on IO load
5027 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
5028 after numerous create/delete cycles
5029 Details : The ext3 inode allocation policy would not allocate new inodes
5030 very well on the OSTs because there are no new directories
5031 being created. Instead we look for groups with free space if
5032 the parent directories are nearly full.
5037 Description: Network or server problems during mount may cause partially
5038 mounted clients instead of returning an error.
5039 Details : The config llog parsing code may overwrite the error return
5040 code during mount error handling, returning success instead
5046 Description: MDS can fail to allocate large reply buffers
5047 Details : After long uptimes the MDS can fail to allocate large reply
5048 buffers (e.g. zconf client mount config records) due to memory
5049 fragmentation or consumption by the buffer cache. Preallocate
5050 some large reply buffers so that these replies can be sent even
5051 under memory pressure.
5055 Frequency : rare (liblustre)
5056 Description: fsx running with liblustre complained that using truncate() to
5057 extend the file doesn't work. This patch corrects that issue.
5058 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
5059 ATTR_SIZE and lsm use in llu_setattr_raw.
5063 Frequency : rare, only 2.6 kernels
5064 Description: Unusual file access patterns on the MDS may result in inode
5065 data being lost in very rare circumstances.
5066 Details : Bad interaction between the ea-in-inode patch and the "no-read"
5067 code in the 2.6 kernel caused the inode and/or EA data not to
5068 be read from disk, causing single-file corruption.
5072 Frequency : rare, only 2.6 filesystems using extents
5073 Description: Heavy concurrent write and delete load may cause data corruption.
5074 Details : It was possible under high-load situations to have an extent
5075 metadata block in the block device cache from a just-unlinked
5076 file overwrite a newly-allocated data block. We now unmap any
5077 metadata buffers that alias just-allocated data blocks.
5081 Frequency : filesystems with default stripe_count larger than 77
5082 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
5083 Details : lconf specifies an inode size of 4096 bytes when the default
5084 stripe_count is larger than 77. This conflicts with the default
5085 inode density of 1 per 4096 bytes. Allocate smaller inodes in
5086 this case to avoid pinning too much memory for large EAs.
5088 ------------------------------------------------------------------------------
5090 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
5095 Frequency : rare (extremely heavy IO load with hundreds of clients)
5097 Description: Client is evicted, gets IO error writing to file
5098 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
5099 caused clients to be evicted for AST timeouts. The fixes in
5100 bug 5192 mean we no longer need to have such short AST timeouts
5101 so ldlm_timeout has been increased.
5104 Frequency : occasional during --force or --failover shutdown under load
5105 Bugzilla : 5949, 4834
5106 Description: Server oops/LBUG if stopped with --force or --failover under load
5107 Details : a collection of import/export refcount and cleanup ordering
5108 issues fixed for safer force cleanup
5111 Frequency : only filesystems larger than 120 OSTs
5112 Bugzilla : 5990, 6223
5113 Description: lfs getstripe would oops on a very large filesystem
5114 Details : lov_getconfig used kfree on vmalloc'd memory
5117 Frequency : only filesystems exporting via NFS to Solaris 10 clients
5118 Bugzilla : 6242, 6243
5119 Description: reading from files that had been truncated to a non-zero size
5120 but never opened returned no data
5121 Details : ll_file_read() reads zeros from no-object files to EOF
5126 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
5127 Details : There's little harm in aborting MDS/OSS recovery and letting it
5128 try again, so I removed the LASSERT and return an error instead.
5130 Severity : enhancement
5132 Description: New debugging infrastructure for tracking down data corruption
5133 Details : The I/O checksum code was replaced to: (a) control it at runtime,
5134 (b) cover more of the client-side code path, and (c) try to narrow
5135 down where problems occurred
5139 Bugzilla : 3819, 4364, 4397, 6313
5140 Description: Racing close and eviction MDS could cause assertion in mds_close
5141 Details : It was possible to get multiple mfd references during close and
5142 client eviction, leading to one thread referencing a freed mfd.
5144 Severity: : enhancement
5145 Bugzilla : 3262, 6359
5146 Description: Attempts to reconnect to servers are now more aggressive.
5147 Details : This builds on the enhanced upcall-less recovery that was added
5148 in 1.4.2. When trying to reconnect to servers, clients will
5149 now try each server in the failover group every 10 seconds. By
5150 default, clients would previously try one server every 25 seconds.
5155 Description: After recovery, certain operations trigger a failed
5156 assertion on a client.
5157 Details : Failing over an mds, using lconf -d --failover, while a
5158 client was doing a readdir() call would cause the client to
5159 LBUG after recovery completed and the readdir() was resent.
5161 Severity : enhancement
5163 Description: Default groups are now added by lconf
5164 Details : You can now run lconf --group <servicename> without having to
5165 manually add groups with lmc.
5168 Frequency : occasional
5170 Description: Nodes with an elan id of 0 trigger a failed assertion
5173 Frequency : always when accessing e.g. tty/console device nodes
5175 Description: tty and some other devices nodes cannot be used on lustre
5176 Details : file's private_data field is used by device data and lustre
5177 values in there got lost. New field was added to struct file to
5178 store fs-specific private data.
5181 Frequency : when exporting Lustre via NFS
5183 Description: NFSD failed occasionally when looking up a path component
5184 Details : NFSD is looking up ".." which was broken in ext3 directories
5185 that had grown large enough to become hashed.
5188 Frequency : Clusters with multiple interfaces not on the same subnet
5190 Description: Nodes will repeatedly try to reconnect to an interface which it
5191 cannot reach and report an error to the log.
5192 Details : Extra peer list entries will be created by lconf with some peers
5193 unreachable. lconf now validates the peer before adding it.
5196 Frequency : Only if a default stripe is set on the filesystem root.
5198 Description: Setting a default stripe on the filesystem root prevented the
5199 filesystem from being remounted.
5200 Details : The client was sending extra request flags in the root getattr
5201 request and did not allocate a reply buffer for the dir EA.
5204 Frequency : occasional, higher if lots of files are accessed by one client
5205 Bugzilla : 6159, 6097
5206 Description: Client trips assertion regarding lsm mismatch/magic
5207 Details : While revalidating inodes the VFS looks up inodes with ifind()
5208 and in rare cases can find an inode that is being freed.
5209 The ll_test_inode() code will free the lsm during ifind()
5210 when it finds an existing inode and then the VFS later attaches
5211 this free lsm to a new inode.
5215 Bugzilla : 6422, 7030
5216 Description: MDS deadlock between mkdir and client eviction
5217 Details : Creating a new file via mkdir or mknod (starting a transaction
5218 and getting the ns lock) can deadlock with client eviction
5219 (gets ns lock and trying to finish a synchronous transaction).
5222 Frequency : occasional
5223 Description: While starting a server, the fsfilt_ext3 module could not be
5225 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
5226 kernels. Previously, lconf would still use the ext3 name
5227 when trying to load modules. Now, it will correctly use
5228 ext3 on 2.4 and ldiskfs on 2.6.
5230 Severity : enhancement
5231 Description: The default stripe count has been changed to 1
5232 Details : The interpretation of the default stripe count (0, to lfs
5233 or lmc) has been changed to mean striping across a single
5234 OST, rather than all available. For general usage we have
5235 found a stripe count of 1 or 2 works best.
5237 Severity : enhancement
5238 Description: Add support for compiling against Cray portals.
5239 Details : Conditional compiling for some areas that are different
5243 Frequency : occasional
5244 Bugzilla : 6409, 6834
5245 Description: Creating files with an explicit stripe count may lead to
5246 a failed assertion on the MDS
5247 Details : If some OSTs are full or unavailable, creating files may
5248 trigger a failed assertion on the MDS. Now, Lustre will
5249 try to use other servers or return an error to the
5253 Frequency : occasional
5255 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
5256 caused lock timeout during readahead (since 1.4.2).
5257 Details : Processes doing readahead might match a lock that hasn't been
5258 granted yet if there are overlapping and conflicting lock
5259 requests. The readahead process waits on ungranted lock
5260 (original lock is CBPENDING), while OST waits for that process
5261 to cancel CBPENDING read lock and eventually evicts client.
5263 Severity : enhancement
5265 Description: Initial enabling of flock support for clients
5266 Details : Implements fcntl advisory locking and file status functions.
5267 This feature is provided as an optional mount flag (default
5268 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
5269 locking are implemented yet, and those that are are not guaranteed
5270 to be completely correct in production environments.
5271 mount -t lustre -o [flock|noflock] ...
5274 Frequency : occasional
5276 Description: OSTs running 2.4 kernels but with extents enabled might trip an
5277 assertion in the ext3 JBD (journaling) layer.
5278 Details : The b_committed_data struct is protected by the big kernel lock
5279 in 2.4 kernels, serializing journal_commit_transaction() and
5280 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
5281 access to this struct. In 2.6 kernels there is finer grained
5282 locking to improve SMP performance of the JBD layer.
5286 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
5291 Description: merge and cleanup kernel patches.
5294 -----------------------------------------------------------------------------
5296 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
5298 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
5299 versions of Lustre. Please update all servers and clients to
5300 version 1.4.2 or later at the same time. You must also run
5301 "lconf --write-conf {config}.xml" on the MDS while it is stopped
5302 to update the configuration logs.
5304 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
5305 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
5306 - swab llog records of type '0' so we get proper header size/idx (5861)
5307 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
5308 - fix rename of one directory over another leaking an inode (5953)
5309 - avoid SetPageDirty on 2.6 (5981)
5310 - don't re-add just-being-destroyed locks to the waiting list (5653)
5311 - when creating new directories, inherit the parent's custom
5312 striping settings if present parent (3048)
5313 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
5314 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
5315 - don't hold client locks on temporary worklist from l_lru (5666)
5316 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
5317 - automatically evict dead clients (5921)
5318 - Update file size properly in create+truncate+fstat case (6196)
5319 - Do not unhash mountpoint dentries, do not allow removal of
5321 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
5322 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
5323 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
5324 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
5325 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
5326 - Correctly handle reads of files with no objects (6243)
5327 - lctl recover will also mark a device active if deactivate used (5933)
5329 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
5330 - allow --write-conf on an MDS with different nettype than client (5619)
5331 - don't write config llogs to MDS for mounts not from that MDS (5617)
5332 - lconf should create multiple TCP connections from a client (5201)
5333 - init scripts are now turned off by default; run chkconfig --on
5334 lustre and chkconfig --on lustrefs to use them
5335 - upcalls are no longer needed for clients to recover to failover
5337 - add --abort-recovery option to lconf to abort recovery on device
5339 - add support for an arbitrary number of OSTs (3026)
5340 - Quota support protocol changes.
5341 - forward compatibility changes to wire structs (6007)
5342 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5343 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5344 - improved "lustre status" script
5345 - initialize blocksize for non-regular files (6062)
5346 - added --disable-server and --disable-client configure options (5782)
5347 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5348 - Vanilla 2.4.29 support
5349 - increase maximum number of obd devices to 520 (6242)
5350 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5351 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5352 SLES 9 2.6.5 kernels
5353 - limit stripes per file to 160 (the maximum EA size) (6093)
5355 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5358 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5359 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5360 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5361 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5362 - small ext3 extents cleanups and fixes (5733)
5363 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5364 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5365 - avoid needless client->OST connect, fix handle mismatch (5317)
5366 - fix DLM error path that led to out-of-sync client, long delays (5779)
5367 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5368 - fix several locking issues related to i_size (5492,5624,5654,5672)
5369 - don't move pending lock onto export if it is already evicted (5683)
5370 - fix kernel oops when creating .foo in unlinked directory (5548)
5371 - fix deadlock in obdfilter statistics vs. object create (5811)
5372 - use time_{before,after} to avoid timer jiffies wrap (5882)
5373 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5374 - Do not leak request if server was not able to process it (5154)
5375 - If mds_open unable to find parent dir, make that negative lookup(5154)
5376 - don't create new directories with extent-mapping (5909, 5936)
5378 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5379 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5380 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5381 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5382 - support for RHEL4 kernel on i686 (5773)
5383 - provide error messages when incompatible logs are encountered (5898)
5385 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5386 * version 1.4.0.10 (1.4.1 release candidate 1)
5388 - don't keep a lock reference when lock is not granted (4238)
5389 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5390 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5391 - drop import inflight refcount on signal_completed_replay error (5255)
5392 - unlock page after async write error during send (3677)
5393 - handle missing objects in filter_preprw_read properly (5265)
5394 - no transno return for symlink open, don't save no-trasno open (3440)
5395 - don't try to complete elan receive that already failed (4012)
5396 - free RPC server reply state on error (5406)
5397 - clean up thread from ptlrpc_start_thread() on error (5160)
5398 - readahead could read extra page into cache that wasn't ejected (5388)
5399 - prevent races in class_attach/setup/cleanup/detach (5260)
5400 - don't dereference de->d_inode after l_dput of de (5458)
5401 - use "int" for stripe value returned from lock_to_stripe (5544)
5402 - mballoc allocation and error-checking fixes in 2.6 (5504)
5403 - block device patches to fix I/O request sizes in 2.6 (5482)
5404 - look up hostnames for IB nals (5602)
5405 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5406 - don't start multiple acceptors for the same port (5277)
5407 - fix incorrect LASSERT in mds_getattr_name (5635)
5408 - export a proc file for general "ping" checking (5628)
5409 - fix "lfs check" to not block when the MDS is down (5628)
5411 - service request history (4965)
5412 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5413 - create an "evict_client" /proc entry on OSTs, like the MDS has
5414 - fix mount usage message, return errors per mount(8) (5168)
5415 - change grep [] to grep "[]" in tests so they work in more UMLs
5416 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5417 - remove ancient LOV_MAGIC_V0 EA support (5047)
5418 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5419 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5420 - allow readahead allocations to fail when low on memory (5383)
5421 - mmap locking landed again, after considerable improvement (2828)
5422 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
5424 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
5427 - send OST transaction number in read/write reply to free req (4966)
5428 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5429 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5430 - the watchdog thread now runs as interruptible (5246)
5431 - flock/lockf fixes (but it's still disabled, pending 5135)
5432 - don't use EXT3 constants in llite code (5094)
5433 - memory shortage at startup could cause assertion (5176)
5435 - reorganization of lov code
5436 - single portals codebase
5438 - add extents/mballoc support (5025)
5439 - direct I/O reads in the obdfilter (4048)
5440 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
5442 tbd Cluster File Systems, Inc. <info@clusterfs.com>
5445 - send OST transaction number in read/write reply to free req (4966)
5446 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5447 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5448 - the watchdog thread now runs as interruptible (5246)
5449 - handle missing objects in filter_preprw_read properly (5265)
5450 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5451 - drop import inflight refcount on signal_completed_replay error (5255)
5452 - unlock page after async write error during send (3677)
5453 - return original error code on reconstructed replies (3761)
5454 - no transno return for symlink open, don't save no-trasno open (3440)
5456 - add pid to ldlm debugging output (4922)
5457 - bump the watchdog timeouts -- we can't handle 30sec yet
5458 - extra debugging for orphan dentry/inode bug (5259)
5460 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
5463 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
5464 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
5465 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
5466 - fix NULL dereference in /proc/sys/portals/routes (4827)
5467 - allow failed mdc_close() operations to be interrupted (4561)
5468 - stop precreate on OST before MDS would time out on it (4778)
5469 - don't send partial-page writes before EOF from client (4410)
5470 - discard client grant for sub-page writes on large-page clients (4520)
5471 - don't free dentries not owned by NFS code, check generation (4806)
5472 - fix lsm leak if mds_create_objects() fails (4801)
5473 - limit debug_daemon file size, always print CERROR messages (4789)
5474 - use transno after validating reply (3892)
5475 - process timed out requests if import state changes (3754)
5476 - update mtime on OST during writes, return in glimpse (4829)
5477 - add mkfsoptions to LDAP (4679)
5478 - use ->max_readahead method instead of zapping global ra (5039)
5479 - don't interrupt __l_wait_event() during strace
5481 - add software watchdogs to catch hung threads quickly (4941)
5482 - make lustrefs init script start after nfs is mounted
5483 - fix CWARN/ERROR duplication (4930)
5484 - return async write errors to application if possible (2248)
5485 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
5486 - print NAL number in %x format (4645)
5487 - update barely-supported suse-2.4.21-171 series (4842)
5488 - support for sles 9 %post scripts
5489 - support for building 2.6 kernel-source packages
5490 - support for sles km_* packages
5492 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
5495 - ignore -ENOENT errors in osc_destroy (3639)
5496 - notify osc create thread that OSC is being cleaned up (4600)
5497 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
5498 - reconstruct ost_handle() like mds_handle() (4657)
5499 - create a new thread to do import eviction to avoid deadlock (3969)
5500 - let lconf resolve symlinked-to devices (4629)
5501 - don't unlink "objects" from directory with default EA (4554)
5502 - hold socknal file ref over connect in case target is down (4394)
5503 - allow more than 32000 subdirectories in a single directory (3244)
5504 - fix blocks count for O_DIRECT writes (3751)
5505 - OST returns ENOSPC from object create when no space left (4539)
5506 - don't send truncate RPC if file size isn't changing (4410)
5507 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
5508 - bind to privileged port in socknal and tcpnal (3689)
5510 - rate limit CERROR/CWARN console message to avoid overload (4519)
5511 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
5512 - basic mmap support (3918)
5513 - kernel patch series update from b1_4 (4711)
5515 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
5518 - avoid crash during MDS cleanup with OST shut down (2775)
5519 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
5520 - don't use bad inodes on the MDS (3744)
5521 - dynamic object preallocation to improve recovery speed (4236)
5522 - don't hold spinlock over lock dumping or change debug flags (4401)
5523 - don't zero obd_dev when it is force cleaned (3651)
5524 - print grants to console if they go negative (4431)
5525 - "lctl deactivate" will stop automatic recovery attempts (3406)
5526 - look for existing locks in ldlm_handle_enqueue() (3764)
5527 - don't resolve lock handle twice in recovery avoiding race (4401)
5528 - revalidate should check working dir is a directory (4134)
5530 - don't always mark "slow" obdfilter messages as errors (4418)
5532 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
5535 - don't close LustreDB during write_conf until it is done (3860)
5536 - fix typo in lconf for_each_profile (3821)
5537 - allow dumping logs from multiple threads at one time (3820)
5538 - don't allow multiple threads in OSC recovery (3812)
5539 - fix debug_size parameters (3864)
5540 - fix mds_postrecov to initialize import for llog ctxt (3121)
5541 - replace config semaphore with spinlock (3306)
5542 - be sure to send a reply for a CANCEL rpc with bad export (3863)
5543 - don't allow enqueue to complete on a destroyed export (3822)
5544 - down write_lock before checking llog header bitmap (3825)
5545 - recover from lock replay timeout (3764)
5546 - up llog sem before sending rpc (3652)
5547 - reduce ns lock hold times when setting kms (3267)
5548 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
5549 - fix NULL deref and obd_dev leak on setup error (3312)
5550 - replace some LBUG about llog ops with error handling (3841)
5551 - don't match INVALID dentries from d_lookup and spin (3784)
5552 - hold dcache_lock while marking dentries INVALID and hashing (4255)
5553 - fix invalid assertion in ptlrpc_set_wait (3880)
5555 - add libwrap support for the TCP acceptor (3996)
5556 - add /proc/sys/portals/routes for non-root route listing (3994)
5557 - allow setting MDS UUID in .xml (2580)
5558 - print the stack of a process that LBUGs (4228)
5560 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
5563 - don't cleanup request in ll_file_open() on failed MDS open (3430)
5564 - make sure to unset replay flag from failed open requests (3440)
5565 - if default stripe count is 0, use OST count for inode size (3636)
5566 - update parent mtime/ctime on client for create/unlink (2611)
5567 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
5568 - free recovery state on server during a forced cleanup (3571)
5569 - unregister_reply for resent reqs (3063)
5570 - loop back devices mounting and status check on 2.6 (3563)
5571 - fix resource-creation race that can provoke i_size == 0 (3513)
5572 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
5573 - more debugging for page-accounting assertion (3746)
5574 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
5575 - avoid deadlock after precreation failure (3758)
5576 - fix race and lock order deadlock in orphan handling (3450, 3750)
5577 - add validity checks when grabbing inodes from l_ast_data (3599)
5579 - add /proc/.../recovery_status to obdfilter (3428)
5580 - lightweight CDEBUG infrastructure, debug daemon (3668)
5581 - change default OSC RPC parameters to be better on small clusters
5582 - turn off OST read cache for files smaller than 32MB
5583 - install man pages and include them in rpms (3100)
5584 - add new init script for (un)mounting lustre filesystems (2593)
5585 - run chkconfig in %post for init scripts (3701)
5586 - drop scimac NAL (unmaintained)
5588 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
5591 - clean kiobufs before and after use (3485)
5592 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
5593 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
5594 - change init script to fail more clearly if not run as root (1528)
5595 - allow clients to reconnect during replay (1742)
5596 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
5597 - don't do DNS lookups on NIDs too small for IP addresses (3442)
5598 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
5599 - fix cond_resched (3554)
5600 - only evict unfinished clients after recovery (3515)
5601 - allow bulk resend, prevent data loss (3570)
5602 - dynamic ptlrpc request buffer allocation (2102)
5603 - don't allow unlinking open directory if it isn't empty (2904)
5604 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
5605 - remove extraneous obd dereference causing LASSERT failure (3334)
5606 - don't use get_cycles() when creating temp. files on the mds (3156)
5607 - hold i_sem when setting i_size in ll_extent_lock() (3564)
5608 - handle EEXIST for set-stripe, set proper directory name (3336)
5610 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
5611 - fix ksocknal_fmb_callback() error messages (2918)
5613 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
5616 - don't copy lvb into (possibly NULL) reply on error (2983)
5617 - don't deref dentry after dput, don't free lvb on error (2922)
5618 - use the kms to determine writeback rpc length (2947)
5619 - increment oti_logcookies when osc is inactive (2948)
5620 - update client's i_blocks count via lvb messages (2543)
5621 - handle intent open/close of special files properly (1557)
5622 - mount MDS with errors=remount-ro, like obdfilter (2009)
5623 - initialize lock handle to avoid ASSERT on error cleanup (3057)
5624 - don't use cancelling-locks' kms values (2947)
5625 - use highest lock extent for kms, not last one (2925)
5626 - don't dereference ERR_PTR() dentry in error handling path (3107)
5627 - fix thread race in portals_debug_dumplog() (3122)
5628 - create lprocfs device entries at setup instead of at attach (1519)
5629 - common AST error handler, don't evict client on completion race (3145)
5630 - zero nameidata in detach_mnt in 2.6 (3118)
5631 - verify d_inode after revalidate_special is valid in 2.6 (3116)
5632 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
5633 - initialize RPC timeout timer earlier for 2.6 (3219)
5634 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
5635 - print nal/nid for unknown nid (3258)
5636 - additional checks for oscc recovery before doing precreate (3284)
5637 - fix ll_extent_lock() error return code for 64-bit systems (3043)
5638 - don't crash in mdc_close for bad permissions on open (3285)
5639 - zero i_rdev for non-device files (3147)
5640 - clear page->private before handing to FS, better assertion (3119)
5641 - tune the read pipeline (3236)
5642 - fix incorrect decref of invalidated dentry (2350)
5643 - provide read-ahead stats and refine rpc in flight stats (3328)
5644 - don't hold journal transaction open across create RPC (3313)
5645 - update atime on MDS at close time (3265)
5646 - close LDAP connection when recovering to avoid server load (3315)
5647 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
5648 - don't leak open file on MDS after open resend (3325)
5649 - serialize filter_precreate and filter_destroy_precreated (3329)
5650 - loop device shouldn't call sync_dev() for nul device (3092)
5651 - clear page cache after eviction (2766)
5652 - resynchronize MDS->OST in background (2824)
5653 - refuse to mount the same filesystem twice on same mountpoint (3394)
5654 - allow llmount to create routes for mounting behind routers (3320)
5655 - push lock cancellation to blocking thread for glimpse ASTs (3409)
5656 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
5657 - fix rare problem with rename on htree directories (3417)
5659 - allow default OST striping configuration per directory (1414)
5660 - fix compilation for qswnal for 2.6 kernels (3125)
5661 - increase maximum number of MDS request buffers for large systems
5662 - change liblustreapi to be useful for external progs like lfsck (3098)
5663 - increase local configuration timeout for slow disks (3353)
5664 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
5666 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5669 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
5670 - don't overwrite extent policy data in reply if lock was blocked (2901)
5671 - drop filter export grants atomically with removal from device (2663)
5672 - del obd_self_export from work_list in class_disconnect_exports (2908)
5673 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
5674 - swab reply message in mdc_close, other PPC fixes (2464)
5675 - fix destroying of named logs (2325)
5676 - overwrite old logs when running lconf --write_conf (2264)
5677 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
5678 - fix race in target_handle_connect (2898)
5679 - mds_reint_create() should take same inode create lock (2926)
5680 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
5681 - don't close files for self_export to avoid uninitialized obd (2936)
5682 - allow MDS with the same name as client node (2939)
5683 - hold dentry reference for closed log files for unlink (2325)
5684 - reserve space for all logs during transactions (2059)
5685 - don't evict page beyond end of stripe extent (2925)
5686 - don't oops on a deleted current working directory (2399)
5687 - handle hard links to targets without a parent properly (2517)
5688 - don't dereference NULL lock when racing during eviction (2867)
5689 - don't grow lock extents when lots of conflicting locks (2919)
5691 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
5694 - account for cache space usage on clients to avoid data loss (974)
5695 - lfsck support in lustre kernel code (2349)
5696 - reduce journal credits needed for BRW writes (2370)
5697 - orphan handling to avoid losing space on client/server crashes
5698 - ptlrpcd can be blocked, stopping ALL progress (2477)
5699 - use lock value blocks to assist in proper KMS, faster stat (1021)
5700 - takes i_sem instead of DLM locks internally on obdfilter (2720)
5701 - recovery for initial connections (2355)
5702 - fixes for mds_cleanup_orphans (1934)
5703 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
5704 - block all file creations until orphan recovery completes (1901)
5705 - client remove rq_connection from request struct (2423)
5706 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
5707 - recovery timer race (2670)
5708 - mdc_close recovey bug (2532)
5709 - ptlrpc cleanup bug (2710)
5710 - mds timeout on local locks (2588)
5711 - namespace lock held during RPCs (2431)
5712 - handle interrupted sync write properly (2503)
5713 - don't try to handle a message that hasn't been replied to (2699)
5714 - client assert failure during cleanup after abort recovery (2701)
5715 - leak mdc device after failed mount (2712)
5716 - ptlrpc_check_set allows timedout requests to complete (2714)
5717 - wait for inflight reqs when ptlrpcd finishes (2710)
5718 - make sure unregistered services are removed from the srv_list
5719 - reset bulk XID's when resending them (caught by 1138 test)
5720 - unregister_bulk after timeout
5721 - fix lconf error (2694)
5722 - handle write after unfinished setstripe, stripe-only getstripe (2388)
5723 - readahead locks pages, leaves pending causing memory pressure (2673)
5724 - increase OST request buffers to 4096 on large machines (2729)
5725 - fix up permission of existing directories in simple_mkdir (2661)
5726 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
5727 - don't assign transno to errored transactions (2742)
5728 - don't delete objects on OST if given a bogus objid from MDS (2751)
5729 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
5730 - if rq_no_resend, then timeout request after recovery (2432)
5731 - fix MDS llog_logid record size, 64-bit array alignment (2733)
5732 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
5733 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
5734 - MDS recovery shouldn't delete objects that it has given out (2730)
5735 - if enqueue arrives after completion, don't clobber LVB (2819)
5736 - don't unlock pages twice when trigger_group_io returns error (2814)
5737 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
5738 - don't write pages to disk if there was an error (1450)
5739 - don't ping imports that have recovery disabled (2676)
5740 - take buffered bytes into account when balancing socknal conn (2817)
5741 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
5742 - reconnect unlink llog connection after MDS reconnects to OST (2816)
5743 - remove little-endian swabbing of llog records (1987)
5744 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
5745 - retry reposting request buffers if they fail (1191)
5746 - grow extent at grant time to avoid granting a revoked lock (2809)
5747 - lock revoke doesn't evict page if covered by a second lock (2765)
5748 - disable VM readahead to avoid reading outside lock extents (2805)
5750 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
5751 - updated kernel patches for hp-2.4.20 kernel (2681)
5753 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
5756 - fix truncated write corruption (2366)
5757 - fix for failed assertion in iopen_connect_dentry (1792,2517)
5759 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
5760 - ensure len(uuid) < 37 in lmc (1171)
5761 - fix ia64 OOPS in llog_test (2255)
5762 - zero end of page at obdfilter for partial page writes (2648)
5763 - don't leave stale dentries around after renames (bug 2428)
5764 - fix timeouts when evicting a client with a single lock held (2642)
5765 - set deadline for the initial HELLO message to drain (2634)
5766 - print out dotted-quad IP addresses in the socknal (2302)
5768 - additional debugging for MDS client eviction problem (2443)
5769 - fix mkfsoptions support for osts (2603, 2604)
5771 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
5774 - add series for the vanilla 2.6.0 kernel
5775 - add series for the vanilla 2.4.24 kernel
5776 - add series for a cray x86/64 UL kernel drop
5777 - fix xattr patches for the vanilla 2.4.19 series
5779 - generate true UUIDs in lmc (1171)
5780 - have portals stack dumping break in UML (2466)
5781 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
5782 - allocate with _NOFS in ldlm to avoid deadlock (1933)
5783 - wake callback waiting threads on client eviction (2460)
5784 - Add --ptldebug and --subsystem to lmc (1719)
5785 - update assertion to allow safe interrupt allocation
5786 - set rq_no_resend for cancel requests (2432)
5787 - recalculate ptlrpcd timeout after resend (2494)
5788 - call vfs_rmdir when removing pending directories (2368)
5789 - fix renaming a file to itself (2429)
5790 - lmc creates a default one-stripe lov (2454)
5791 - expand procfs space to handle large clusters (2326)
5792 - increase UML stack to avoid overflow
5793 - update lconf's list of debug and subsystem masks
5794 - fix lfs find --obd (2510)
5795 - /proc tunable for disabling filter read caching (2591)
5796 - stop rpm packages from altering slapd.conf (2301)
5797 - disable nagle in the socknal under 0conf (2578)
5798 - choose mds inode size based on stripe count (2572)
5799 - fix kernel-source rpm problems (2516)
5801 - add --disable-doc to avoid pdf generation (2421)
5802 - update documentation, tests, type-os, comments
5803 - avoid format warnings on ia64
5804 - remove the TOE NAL
5805 - tiny code cleanups by removing unused fields
5807 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
5810 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
5811 - remove the most blatant lies from BUILDING file (2371)
5812 - change default debug level to reasonable production setting
5813 - reduce client side cache size to reduce cache flush time
5814 - reduce max RPCs in flight to avoid unnecessary file fragmentation
5815 - make TCP zerocopy and pinger support enabled by default (2476)
5816 - sync writes completed after process exits caused crashes (2319)
5817 - maintain correct mount count on the MDS (2356)
5818 - backout 1557, because 2316 wasn't really fixed
5819 - better file I/O statistics gathering in /proc
5820 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
5821 - another kernel patch to fix zero-copy TCP function export
5822 - don't take duplicate lock when processing re-sent getattr (2420)
5823 - lctl uses obd_self_export instead of creating new conn (2353)
5824 - MDS/OST recovery case which requires object creation asserted (2425)
5825 - move lfs from /usr/sbin to /usr/bin in packages
5826 - fix race between mds_client_add and mds_client_free (2417)
5827 - use kmalloc instead of slabs in portals (2430)
5828 - don't create duplicate records when a failover MDS is present (2442)
5829 - remove unnecessary mount age check (2332)
5830 - don't remove directory inodes from locks prematurely (2451)
5831 - don't break if MDS service name is the same as hostname (2103)
5832 - fix races in client write RPC generation when cache full (2482)
5834 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
5837 - remove now-unused request->rq_obd (278)
5838 - if an allocation fails, print out how much memory we've used (1933)
5839 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
5840 - add the "configurable stack size" patch to most series files (1256)
5841 - ability to write large log records, for 100+ OST configs (2306)
5842 - fix NULL deref when filter_prep fails (2314)
5843 - fix operator precedence error in filter_sync
5844 - dynamic allocation of socknal TX descriptors (2315)
5845 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
5846 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
5847 - ensure that CWARN messages in Portals always get to the syslog
5848 - __init/__exit are not for prototype decls (ldlm_init/exit)
5849 - x86-64 compile warning fixes
5850 - fix gateway LMC keyword conflict (2318)
5851 - fix MDS lock inversions in getattr/reint paths (1844)
5852 - fix a rare lock re-ordering bug, which caused deadlock (2322)
5853 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
5854 - DLM race condition prevented some lock evictions (2328)
5855 - ENOMEM detection and retry on socknal sends (2230)
5856 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
5857 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
5858 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
5859 - better error messages when a client is rejected during recovery (1505)
5860 - avoid cancelling locks which were never granted, after failure (2330)
5861 - fix i_sem/journal inversion in mds_client_add (2333)
5862 - fix truncate/getattr lock cycle deadlock (2334)
5863 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
5864 - fix two rare exit paths which could leak an l_lock() ref (2321)
5865 - fencepost error in MDS/OST orphan recovery (2226)
5866 - make log record alignment 8 bytes (1988)
5867 - lstripe now fails when requested offset > ost_count (2237)
5868 - ensure that all kernel series have a complete list.h (1607)
5869 - fix crashes in special-file operations (2316)
5870 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
5872 - allow configurable automake binary, for testing new versions
5873 - small update to the lfs documentation
5875 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
5877 * fix negative export reference count in fsfilt_sync (2312)
5879 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
5880 * release candidate 0.9.1
5882 - orphans are moved into the PENDING directory for possible recovery
5883 - replayed opens now open by fid for orphan/rename safety (1042)
5884 - last close of an orphan inode generates a transno (683)
5885 - chdir() and mount() now pin the directory entry (1020)
5886 - avoid CERROR in normal ll_setattr_raw() error case (1500)
5887 - discard very old requests without processing them (1502)
5888 - remove some common, well-understood CERRORs (1505)
5889 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
5890 - clear "grant" flags in OST replies until OST grant code lands (1644)
5891 - fix read performance by not clobbering i_blksize on client (1598)
5892 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
5893 - make LRU size a /proc tunable, clears locks when reduced (707)
5894 - fix some lprocfs rot that prevented ptlbd from loading (1732)
5895 - server locks take references on exports now (1558)
5896 - build fixes for 2.4.20-rh trees (1663)
5897 - return an error from lov_create if all OSCs are inactive (1751)
5898 - fix import levels when a reconnect happens without a timeout (1597)
5899 - exit early from mds_open if we get a lookup error (1749)
5900 - partial page read at EOF wouldn't wait for disk before sending (1642)
5901 - avoid NULL deref in obdfilter when reading page past EOF (1592)
5902 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
5903 - fix LBUG in ll_it_open_error with rc = -2 (1861)
5904 - write/truncate lock inversion (1639)
5905 - Don't auto-load obdclass, portals modules during cleanup (1495)
5906 - fix timestamps from jumping to "now" (1763)
5907 - extra journal assertions (1648)
5908 - add an extra multiunlink test (1771)
5909 - fix read_record/write_record API (1776)
5910 - fix leak of offset_extent, possible incorrect i_size later (1772)
5911 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
5912 - Debugging for the kqswnal_get_idle_tx problems (1820)
5913 - Allow recovery to be attempted multiple times (1536)
5914 - Write out MDS last_rcvd file after it is first created (1600)
5915 - Fix tx_descriptor leak in failed transmit situations (1827)
5916 - ext3 journaling fixes for assertion failure after IO error (1871)
5917 - class_export_put() on freed export after completion AST error (1896)
5918 - Fix revalidate looping in VFS (1322)
5919 - Don't access a freed export during MDS_REINT timeout (1521)
5920 - Add open-unlink recovery support on the MDS (1673,1764)
5921 - Return an error if no MDS data was read from last_rcvd (1946)
5922 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
5923 - Don't setup a disk device that doesn't match exported UUID (317)
5924 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
5925 - avoid committing NULL handle in force close
5926 - local.sh is now a one-stripe LOV configuration
5927 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
5928 - fix ext3 htree duplicate directory entry corruption (1516)
5929 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
5930 - update to new LOV EA format (2097)
5931 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
5933 2003-06-15 Phil Schwan <phil@clusterfs.com>
5936 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
5937 - per-import/export recovery handling (958, 931, 959)
5938 - multiple last-rcvd slots, for serving multiple FSes (949)
5939 - connections are again shared between multiple imp/exports (963, 964)
5940 - "umount -f" would hang if any requests needed to be sent (393, 978)
5941 - avoid pinning large req buffer by copying for queued messages (989)
5942 - add "uuid" to "lctl device" command to help upcalls (991)
5943 - "open" RPCs with transnos would confuse recovery counters (1037)
5944 - do proper endian conversion of all wire messages (288, 340, 891)
5945 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
5946 - call waiting locks callback from LDLM recovery thread (1127, 1151)
5947 - fix ptlrpc_connection leak in target_handle_connect (1174)
5948 - fix import refcounting bug in OST and MDS cleanup (1134)
5949 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
5950 - fix very unlikely obd_types race condition (501)
5951 - remove osc_open hack for echo_client (1187)
5952 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
5953 - a failure in read_inode2 leads to deadlock (1139)
5954 - cancel ack-locks as soon as transaction is committed (1072)
5955 - fix major leaks and crashes in the bulk I/O path (937, 1057)
5956 - make sure to commitrw after any preprw to avoid deadlock (1162)
5957 - failing to execute a file in a lustre FS would lock inode (1203)
5958 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
5959 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
5960 - obd_uuid2tgt would walk off the end of the list (1255)
5961 - on IA64 the osc would give portals incorrect bulk size (1258)
5962 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
5963 - fix lock inversion caused by new llite matching code (1282)
5964 - limit the number of dirty pages on a client to 10MB (1286)
5965 - timed out locks were not being corrected cancelled (1289)
5966 - fix O_DIRECT above 4GB on IA-32 (1292)
5967 * major user-visible changes
5968 - fail out/fail over policy now controlled by the upcall (993)
5970 - add OBD_PING to check server availability and failure (954)
5971 - lustre messages are now sent in sending host order (288, 340, 891)
5972 - add eadatalen to MDS getattr reply (340)
5973 - OST read replies may contain second buffer, with per-page status (593)
5975 2003-03-11 Phil Schwan <phil@clusterfs.com>
5978 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
5979 - failed open()s could cause deadlock; fixed (867, 869)
5980 - stop cancelling OST locks when files are closed (481)
5981 - overlapping XID spaces caused network corruption (851, 853)
5982 - fix unsafe fsfilt counter arithmetic; change to atomic_t
5983 - setattr_raw added, to do single-RPC, server-side setattrs
5984 - lmc/lconf syntax change for OST UUIDs
5985 - fix crashy race condition between ptlrpc_free_req and osc_close
5986 - don't use request in mdc_enqueue if we hit a timeout (889)
5987 - don't set the inode i_size for regular files from the MDS (896)
5988 - handle out of order completion AST (842)
5989 - don't LBUG if a lock request times out after receiving AST (913)
5990 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
5991 - if a bad lock AST arrives, send an error instead of dropping entirely
5992 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
5993 - fix leak in bulk IO when only partially completed (899, 900, 926)
5994 - fix O_DIRECT for ia64 (55)
5995 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
5996 - C-z after timeout could hang a process forever; fixed (977)
5998 - client-side I/O cache (678, 924, 929, 941, 970)
6000 - READPAGE and SETATTRs which don't take server-side locks get
6003 2003-02-11 Phil Schwan <phil@clusterfs.com>
6006 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
6007 - Fix fsfilt_extN_readpage() to read a full page of directory
6008 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
6009 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
6010 - Fix imbalanced LOV object allocation and out-of-bound access (469)
6011 - Most intent operations were removed, in favour of a new RPC mode
6012 that does a single RPC to the server and bypasses most of the VFS
6013 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
6014 - Aggressively cancel local locks on DLM servers
6015 - mds_reint_unlink sends EA to the client if it's the last nlink.
6016 client uses that EA to unlink OST objects.
6017 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
6018 - recursive symlinks were fixed (439)
6019 - fixed NULL deref in DEBUG_REQ
6020 - filter_update_lastobjid no longer calls sync, which annoyed extN
6021 - fixed multi-client small-writes to a single file problem (445)
6022 - fixed mtime updates during file writes (607)
6023 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
6024 - fixed bug in obd_brw_read/write() (under guise of testing 367)
6025 - fixed Linux OST size reporting problem (444, 656)
6026 - OST now updates object mtime with writes or setattr (607, 619)
6027 - client verifies file size before zeroing page past EOF (445)
6028 - OST now writes last allocated objid to disk with allocation (108)
6029 - LOV on echo now works (409)
6031 - mds_reint_unlink sends a new buffer, with the EA included. this
6032 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
6033 set if a regular file was being unlinked, and it was the last link
6034 - use PtlGet from the target for bulk writes (315)
6035 - OST now updates object mtime with writes or setattr (607, 619)
6036 - LDLM now has a grant-time callback to revalidate locked items, if
6038 - Many MDS operations were reorganized to combat race conditions
6040 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
6041 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
6043 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
6046 - Fully reactivate OST imports after reconnection (512, others)
6047 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
6048 - More graceful error handling for truncating on dead OST (515)
6049 - Don't error out unless we're actually accessing dead stripes (474)
6050 - Fix garbage sizes when stripes are missing (410)
6051 - LRU counters were broken, causing constant lock purge (433, 432)
6052 - garbage on read from stripes with failed OSTs (441)
6053 - mark OSCs as active before reconnecting during recovery (438)
6054 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
6055 - lfind did not preserve OST order in output (443)
6056 - symlinks cause hung clients, incorrect data (439)
6057 - stop dereferencing request after dropping refcount (457)
6058 - don't LASSERT(spin_is_locked) on non-SMP (455)
6059 - fixes for many rename() bugs
6060 - fstat didn't correctly synchronize attributes (399)
6061 - server must handle lock cancellation during blocking AST prep (487)
6062 - bulk descriptors were free()d too soon (511)
6063 - fix paths in lconf, which would load incorrect modules (451, 507)
6064 - fix confusing lconf 'host not found' error message (386)
6065 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
6066 - fix race condition in mdc_blocking_ast() for inode access (526)
6067 - fix lov_unpackmd() unpacking wrong number of stripes (537)
6068 - fix lov_set_osc_active() marking wrong OSC inactive (440)
6069 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
6070 - fix multiple writes of stripe MD to MDS (358, maybe 519)
6071 - fix lstripe in several ways (kernel side) (527)
6072 - fix request leak in ldlm_cli_enqueue (262)
6073 - incorrect OSC was marked inactive after OST failure
6074 - call mds_fs_cleanup before unmounting filesystem (524)
6075 - fix races between taking ns_lock and ldlm_lock_change_resource
6076 - fix races updating LOV export open file list
6077 - fix lov_enqueue error path, avoid decref-ing bad lock handle
6078 - fix recovery NULL deref in ldlm_cli_cancel_unused
6079 - fix some DLM races by using new hash table for lock handles (419)
6080 - permit the client to specify desired inodes, at replay
6081 - duplicate requests when we queue them for replay reintegration
6082 - fix last_rcvd offset calculation
6083 - sync after each recovered transaction, so we always make progress
6084 - never, not always, ERESTART requests without transnos
6085 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
6087 - skip replay if the MDS says that the client is already connected
6088 - don't check for a recovery-enabled export to match lctl's UUID
6089 - don't INC_USE_COUNT for phantom exports
6090 - don't crash when cleaning up phantom exports (567)
6091 - don't double-finish or set replay data for errored mdc_open requests
6092 - abort requests when they time out, so we don't get old replies
6093 - send/receive replies for AST messages again
6094 - if the client says that it doesn't have the lock, cancel it on the
6096 - if we timeout during I/O, don't try to cancel an in-use lock; instead
6097 mark it as destroyed, it will all work out when decref is called
6098 - fix module use counts (22, 581)
6100 - ASTs now expect a reply (server cancels lock on error reply)
6102 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
6105 - fix many simultaneous client startup (392)
6106 - fix dentry->d_it clobbering
6107 - credentials weren't being shipped for readdir/getattr operations
6108 - remove invalid assertions triggered during some concurrent MD
6110 - proper Lustre versions added (336, 389)
6111 - fix memory leak for create error case (398)
6112 - fix LOV locking bug that would get cli/srv out of sync
6113 - fix echo client over LOV (409)
6114 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
6115 - fix double-O_EXCL intent crash (424)
6116 - avoid sending multiple lock CANCELs (352)
6118 - MDS can do multi-client recovery (modulo bugs in new code)
6120 - many updates, edits, cleanups
6122 2002-11-18 Phil Schwan <phil@clusterfs.com>
6125 - fix null d_it dereference (346)
6126 - fix full OST/dbench hang (333)
6127 - fix permission problem with file removal (286)
6128 - fix removal of OSCs from LOV when they fail
6129 - fix NULL deref during bulk timeout (214)
6130 - fix problems related to multiple filesystems on one MDS (241)
6131 - fixed serious subtle metadata locking bugs
6132 - free locks on clients when inodes are removed due to memory
6134 - fix inode pointer in lock data (285)
6135 - partial support for multiple MDS on a single host (241)
6136 - data locks weren't cancelled at clear_inode time (290, 311)
6137 - intent locks could lead to unbounded lock growth (205)
6138 - added a maximum lock count, an LRU list, and a flusher
6139 - fix multiple rename (365)
6140 - properly abstracted the echo client
6141 - OSC locked 1 byte too many; fixed
6142 - rewrote brw callback code:
6143 - fixed recovery bugs related to LOVs (306)
6144 - fixed too-many-pages-in-one-write crash (191)
6145 - fixed (again) crash in sync_io_timeout (214)
6146 - probably fixed callback-related race (385)
6148 - Add capability to MDS protocol
6149 - LDLM cancellations and callbacks on different portals
6151 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
6154 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
6155 - defer open object destruction to close time (601981, 138)
6156 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
6157 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
6158 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
6159 - don't call obd_setattr in ll_file_release for destroyed objects
6160 * protocol change to lustre_msg: move |version| and add |flags|
6161 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
6162 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
6163 * added replay of create, unlink, link and rename operations during
6164 MDS failover; recovery should be much more robust now
6165 * remove failed OSCs from LOVs (only lov_create uses this so far)
6166 * the lustre-HOWTO was brought (more) up to date (582544)
6168 2002-10-23 Phil Schwan <phil@clusterfs.com>
6171 - in-use dentries weren't being reused properly (617851)
6172 - prevent multiple LDLM setup (599178)
6173 - fix LOV size calculations for truncate (617853)
6174 - fix client handling of MDS intent errors (POSIX)
6175 - fix permission bug in lovstripe.c test (624321)
6176 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
6177 - truncate past end of file could corrupt data
6178 - proper cleanup after timeouts, crashes, etc (592524, 550815)
6179 - a race in recovery could return ETIMEDOUT to apps (623947)
6180 - building outside the source directory was fixed
6181 * the lustre-HOWTO was brought (more) up to date (582544)
6182 * major progress was made on recovery functionality
6184 2002-10-10 Phil Schwan <phil@clusterfs.com>
6187 - recovery deadlock fix
6188 - rm -rf causes LBUG fix (617817)
6189 - file open by multiple tasks fix (618962)
6190 - directory permissions bugs (602707 and 620007)
6191 - journal_stop fixed with locking (611313)
6192 - O_APPEND failures resolved (618273, perhaps 614459)
6193 - lconf PATH fix (619770)
6194 - IA64 build fix (621450)
6195 - RPC buffer sizes scale with amount of memory
6197 2002-10-01 Phil Schwan <phil@clusterfs.com>
6200 - locks would be cancelled without throwing away data pages,
6201 resulting in inconsistent data (605627)
6202 - inode attributes were not always being refreshed (605627, 612449)
6203 - lconf now continues to cleanup after lctl reports an error
6204 - MDS now enforces user permissions (602707)
6205 - lprocfs cleanup fixed, but not yet enabled (614157)
6206 - fixed infinite server hang, should a client not respond to an AST
6207 - avoid going into recovery if user calls readlink() with a buffer
6208 that's too small (613941)
6209 - AST RPCs no longer require replies (614867) -- this may be changed
6210 - don't crash server if client sends an IOV that's too big (611336)
6211 - fixed lock conversion deadlock (611892)
6212 - fixed the following of symlinks (614622)
6213 * recovery: the server can remove locks from a client that dies, other
6214 clients can make progress
6215 * more extN patch fixes
6216 * compile-time configurable ptlrpc buffer allocations
6218 - collaborative read cache document
6219 - Lustre Lite Performance CDR document-in-progress
6221 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6224 - fix typo in patch-2.4.18
6226 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6229 - clear ptlrpc request each time in handle_incoming_request()
6230 - unlink of files now destroys the object on the OST
6232 2002-09-19 Peter Braam <braam@clusterfs.com>
6234 * add hard link support
6235 * change obdfile creation method
6236 * kernel patch changed
6238 2002-09-19 Peter Braam <braam@clusterfs.com>
6241 - stack overflow bug in extN fixed
6243 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
6245 * documentation updates
6246 - add man pages for config tools
6247 - update tests/README to describe testing with new config tools
6248 - finish metadata API descriptions
6249 * bug fixes and cleanups
6250 - statfs workaround for 16TB limit
6251 - LOV stripe allocation improved, can stripe on subset of OSTs
6252 - LOV file size/IO offset was wrong for files > 4GB in size
6253 - object EA data was being dropped, caused files to be unreadable
6254 - memory overflow with non-LOV OST caused memory corruption
6255 - fixed regression tests to work with new config tools, obdfilter
6256 - fixed bug when directory size became larger than 1 block
6257 - fixed bug (for single client case) when PWD was deleted
6258 - invalidate local directory pages when doing intent-based ops
6259 - avoid LDLM oops when lock callback contained bad data
6261 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
6263 * documentation updates
6264 * bug fixes and cleanups
6265 - configuration tools
6268 - 64-bit compile warnings
6269 - 64-bit internal statfs data
6271 * test_brw on persistent OST devices
6273 * lprocfs (disabled)
6275 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
6277 * documentation updates
6278 * bug fixes and cleanups
6279 * configuration tools
6281 2002-08-30 Peter J. Braam <braam@clusterfs.com>
6284 * many small fixes to 0_5_4
6285 * io/network handling
6286 * thinkos in MDS operations
6288 2002-08-24 Peter J. Braam <braam@clusterfs.com>
6291 * crucial basic fixes to 0.5.3
6292 * IOR, Iozone work over Elan
6295 2002-08-07 Phil Schwan <phil@clusterfs.com>
6296 * version 0_5_3, our first alpha
6297 * we use the new Portals iovs
6298 * documentation updates
6299 * bug fixes and cleanups
6300 * small changes in the DLM wire protocol
6302 2002-07-25 Peter J. Braam <braam@clusterfs.com>
6303 * version 0_5_1 with some initial stability,
6304 * locking on MD and file I/O.
6305 * documentation updates
6306 * several bug fixes since 0.5.0
6307 * small changes in wire protocol
6309 2002-07-18 Phil Schwan <phil@clusterfs.com>
6311 * delivered as Lustre Light Alpha
6312 * fixed a crash after handling invalid MDS requests
6313 * fixed directory pages for architectures with non-4k pages sizes
6315 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
6316 * release version v0_4_4
6317 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
6319 * Fixes a number of interruption problems with OST operations.
6320 * Update documentation for portals header changes
6321 * Move all wire protocol structs/defines to lustre_idl.h
6322 * Fixes symlink length bug.
6323 * Add tcpdump to repository.
6325 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
6326 * release version v0_4_3
6327 * Fixes statfs for inodes on extN.
6328 * Fixes bug in runtests which would delete /etc/hosts.
6329 * Use 64-bit object IDs wherever possible (not into VFS though)
6330 Remove ost_get_info, which is unused by lustre, and out of date.
6332 2002-07-03 Peter Braam <braam@clusterfs.com>
6333 * release version v0_4_2 Fixes a lookup error (type not passed)
6334 * move forward to head of Portals
6335 * move forward to latest Lustre kernel
6337 2002-06-25 Peter Braam <braam@clusterfs.com>
6338 * release version v0_4_1. Hopefully stable on single node use.