3 * See http://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
4 * Server support for kernels:
5 2.6.32-220.7.1.el6 (RHEL6)
6 * Client support for unpatched kernels:
7 2.6.32-220.7.1.el6 (RHEL6)
9 2.6.18-238.19.1.el5 (RHEL5)
10 2.6.16.60-0.69.1 (SLES10)
11 * Recommended e2fsprogs version: 1.41.90.wc4
12 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
13 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
14 stack overflow. For more information, please refer to bugzilla 17630.
15 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
16 the re-exporting nodes to be mounted with "32bitapi" mount option
18 Severity : enhancement
20 Description: Update RHEL6.2 kernel to 2.6.32-220.7.1.el6
22 --------------------------------------------------------------------------------
24 04-30-2012 Whamcloud, Inc.
26 * See http://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
27 * Server support for kernels:
28 2.6.32-220.4.2.el6 (RHEL6)
29 * Client support for unpatched kernels:
30 2.6.32-220.4.2.el6 (RHEL6)
32 2.6.18-238.19.1.el5 (RHEL5)
33 2.6.16.60-0.69.1 (SLES10)
34 * Recommended e2fsprogs version: 1.41.90.wc4
35 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
36 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
37 stack overflow. For more information, please refer to bugzilla 17630.
38 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
39 the re-exporting nodes to be mounted with "32bitapi" mount option
41 Severity : enhancement
43 Description: Update RHEL6.2 kernel to 2.6.32-220.4.2.el6
45 Severity : enhancement
47 Description: Update RHEL6.2 kernel to 2.6.32-220.4.1.el6
49 Severity : enhancement
51 Description: Update RHEL6.2 kernel to 2.6.32-220.el6
53 Severity : enhancement
55 Description: Update RHEL6.1 kernel to 2.6.32-131.17.1.el6
57 --------------------------------------------------------------------------------
59 09-30-2011 Whamcloud, Inc.
61 * Server support for kernels:
62 2.6.18-238.19.1.el5 (RHEL5)
63 2.6.32-131.6.1.el6 (RHEL6)
64 * Client support for unpatched kernels:
65 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
66 2.6.18-238.19.1.el5 (RHEL5)
67 2.6.32-131.6.1.el6 (RHEL6)
68 2.6.29.4-167.fc11 (FC11)
69 2.6.16.60-0.69.1 (SLES10)
71 * Recommended e2fsprogs version: 1.41.90.wc3
72 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
73 Lustre filesystem with 4K stack may cause a stack overflow. For more
74 information, please refer to bugzilla 17630.
75 * NFSv4 reexport requires Lustre client on the re-exporting nodes to be
76 mounted with "32bitapi" mount option
78 Severity : enhancement
79 Description: Update RHEL6.1 kernel to 2.6.32-131.6.1.el6
81 Severity : enhancement
82 Description: Update RHEL5.6 kernel to 2.6.18-238.19.1.el5
84 Severity : enhancement
85 Description: Update RHEL5.6 kernel to 2.6.18-238.9.1.el5
88 Frequency : only with SLES10
90 Description: Use OFED "KMP" provided by Novell
91 Details : SLES10 SP3 ships with OFED in a separate "KMP" package.
92 Lustre is now built against this package. That means you need to
93 install the ofed-kmp package from Novell for the patchless client.
94 Note that the ofed-kmp that Novell ships may not exactly match the
95 kernel version but should still be compatible.
97 Severity : enhancement
99 Description: Update RHEL5.5 kernel to 2.6.18-194.17.1.el5
100 Update OEL5.5 kernel to 2.6.18-194.17.1.0.1.el5.
102 --------------------------------------------------------------------------------
104 2010-07-15 Oracle, Inc.
106 * Server support for kernels:
107 2.6.18-164.11.1.el5 (RHEL 5)
108 2.6.18-164.11.1.0.1.el5 (OEL 5)
109 * Client support for unpatched kernels:
110 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
111 2.6.18-164.11.1.el5 (RHEL 5),
112 2.6.18-164.11.1.0.1.el5 (OEL 5)
113 2.6.16.60-0.42.8 (SLES 10),
115 2.6.29.4-167.fc11 (FC11)
116 * Recommended e2fsprogs version: 1.41.10-sun2
118 Severity : enhancement
120 Description: Distribute lfs_migrate script with lustre
121 Details : The lfs_migrate mentioned in the manual is now included in the
122 lustre tree in lustre/scripts and will distributed in lustre RPMs.
124 Severity : enhancement
126 Description: Failover nid list added to import proc file.
128 Severity : enhancement
130 Description: "som_preview" mount option required to use SOM
131 Details : The clients will need to use som_preview mount option
135 Description: Lustre does not do 1MB IOs to HW RAID
136 Details : Bump MAX_PHYS/HW_SEGMENTS and SG_ALL to 256 in the RHEL5 kernel.
138 Severity : enhancement
140 Description: lfs quota output is not very convenient for awk/sed-parsing
141 Details : Some positions in lfs quota output table could be empty or
142 non-empty which made it hard to parse it with scripts, now
143 a dash is put instead of space where there is not supposed
148 Description: Update to OFED 1.5.1
149 Details : OFED 1.5.1 will be used in environments that don't use
150 vendor supplied OFED.
154 Description: Remove lnet. prefix from lctl params display
155 Details : Also renamed lnet.memused to lnet_memused to
160 Description: Use conf_param -d to remove permanant settings.
164 Description: Add list_param -R
165 Details : lctl list_param now supports recursive mode.
168 Bugzilla : 21829, 22312
169 Description: Fix llobdstat.
170 Details : 1. Now updating the write_bytes counter
171 2. Port script fixes from 1.8.
175 Description: decrease the usage of memory on clients.
176 Details : 1. On clients, recycle dentries and inodes unused.
177 2. Delete the code related to ll_deathrow(att 6215 in bug 1443). It
182 Description: llobdstat fix and enhancement
183 Details : add a counter to set a limit to how many samples will be returned
184 fix a wildcard in the path to limit to obdfilter stats only
188 Description: lfs check only allowed for root
189 Details : Code cleanups and added a regresion test of non-root use of
194 Description: MDS Oopses in ptlrpc_dispatch_difficult_reply()/__list_add()
195 Details : Fix hr->hr_index overload.
199 Description: Fix FIEMAP functionality.
203 Description: FC11 patchless client support added
207 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and
208 OEL5.4 kernel to 2.6.18-164.11.1.0.1.el5.
212 Description: New acc-sm script and various test-framework enhancments.
216 Description: Previously landed patch was reverted.
220 Description: Update SLES10 kernel to 2.6.16.60-0.42.8 (patchless)
224 Description: Add ldiskfs maxdirsize mount option.
228 Description: Fix mkfs.lustre support for 16TB LUNs.
232 Description: Print mmp_check_interval
233 Details : print mmp_check_interval and make it possible to abort
234 mount operation in case it takes too long.
238 Description: Add cascading_rw to lustre mpi tests.
242 Description: Store SOM data in EA.
243 Details : Also several SOM related cleanups.
247 Description: Add refcount to md_open_data
251 Description: Use refcount on lov_request_set to prevent memory corruption.
255 Description: Statahead: use parent dir's i_mutex for synchronization.
259 Description: loadgen improvements.
263 Description: Enable peer health detection in ksocklnd and ko2iblnd.
267 Description: Control DCACHE_LUSTRE_INVALID flag with MDS_INODE_LOOKCUP lock
271 Description: Fix race on obd_devcie:md_stats
275 Description: Fix SLES 11 client builds.
276 Details : Resolved uninitalized pointer gcc warnings.
280 Description: Allow non-root access for "lfs check".
281 Details : Added a check in obd_class_ioctl() for OBD_IOC_PING_TARGET.
285 Description: Clear imp_force_reconnect correctly in ptlrpc_connect_interpret()
287 Severity : enhancement
289 Description: Update kernel to RHEL5.4 2.6.18-164.6.1.el5 and
290 OEL5 2.6.18-164.6.1.0.1.el5(Both in-kernel OFED enabled).
294 Description: Sanity test_27p fails.
295 Details : OBD_FAIL_OST_ENOSPC has been moved to filter_statfs().
296 Sanity tests 27 were cleaned up and improved.
300 Description: Unify req format on client/servers
301 Details : Use new req_capsule API [almost] everywhere instead of old PTLRPC
302 buffers and swabbers approach..
307 Description: drop join file stuff.
309 Severity : enhancement
311 Description: Add 2.6.27 and SLES11 (patchless client) support
315 Description: Fix in ptlrpc_expire_one_request() to print the signed time
318 Severity : enhancement
320 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
323 Severity : enhancement
325 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
327 Severity : enhancement
329 Description: correctly handle big reply message.
330 Details : send LNet event if reply is bigger then buffer and adjust this buffer
335 Description: Drop unnecessary __GFP_NOMEMALLOC flag from filter_get_page()
337 Severity : enhancement
339 Description: Add support for OEL5.
343 Description: Conf-sanity.sh 50g test - deactivated OST should not cause a panic.
345 Severity : enhancement
347 Description: "weak-modules" support
348 Details : Implement "weak-modules" support which enables kernel modules
349 to be used with any kernel that implements the same kABI. In
350 order to achieve this modules are now installed in
351 /lib/modules/$(uname -r)/updates/kernel on all distributions.
355 Description: client could not reconnect to OST because of an active request.
356 Details : The client now retries to reconnect to the same server, if a
357 connect request failed with EBUSY or -EAGAIN. Test case 35b
358 conf-sanity.sh has been created.
362 Description: Don't allow make backward step on assignin osc next id.
363 Details : race between allocation next id and ll_sync thread can be cause
364 of set wrong osc next id and can be kill valid ost objects.
368 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
372 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
376 Description: Error handling in osc_statfs_interpret() has been improved.
377 Details : Check in osc_statfs_interpret() for EBADR.
381 Description: Do not send statfs() requests to OSTs disabled by administrator.
382 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
384 Severity : enhancement
386 Description: Added sanity-quota test to for lfs(1) to work with numeric uid/gid.
388 Severity : enhancement
390 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
392 Severity : enhancement
394 Description: Enhancement for lfs(1) command to use numeric uid/gid.
398 Description: Improved cleanup procedure for conf-sanity test_46a.
402 Description: Do not start bulk transfer, if client tried to reconnect.
403 Details : Do not start bulk transfer, if client tried to reconnect,
404 but failed due to active rpcs. Check for exp_abort_active_req
405 has been implemented.
409 Description: Prevent inconsistences between linux and lustre mount structures.
410 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
411 Make the sleep interruptible.
413 Severity : enhancement
415 Description: provide server to client comms path
416 Details : server to client communications path via new LDLM_SET_INFO rpc
420 Description: truncate starts GFP_FS allocation under transaction causing deadlock
421 Details : ldiskfs_truncate calls grab_cache_page which may start page
422 allocation under an open transaction. This may lead to
423 calling prune_icache with consequent lustre reentrance.
427 Description: send correctly lsm on open replay
428 Details : MDS is trust to LSM size on replay open, but client can set wrong size
433 Description: enable client interface failover
434 Details : When a child reconnects from another NID, properly update export
435 nid hash position and ldlm reverse import.
437 Severity : enhancement
439 Description: Communicate OST degraded/readonly state via statfs to MDS
440 Details : Flags in the statfs returned from OSTs indicate whether the
441 OST is in a degraded RAID state, or if the filesystem has
442 turned read-only after a filesystem error is detected.
447 Description: don't panic if EPROTO was hit when reading symlink
448 Details : correctly handling request reference in error cases.
450 Severity : enhancement
452 Description: Add LustreNetLink, a kernel-userspace communcation path.
454 Severity : enhancement
456 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
461 Description: access to llog context before init.
462 Details : move handling CATALOGS file at osc layer and forbid access to llog
466 Frequency : rare, only when client is writing to OST on local node
468 Description: Avoid deadlock for local client writes
469 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
470 memory freeing context. This allows OST threads to set the
471 PF_MEMALLOC flag on task structures in order to allocate memory
472 from reserved pools and complete IO.
473 Use GFP_HIGHUSER for OST allocations for non-local client writes,
474 so that the OST threads generate memory pressure and allow
475 inactive pages to be reclaimed.
477 Severity : enhancement
479 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
484 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
485 Details : move ldlm namespace creation in setup phase to avoid grab
486 _lprocfs_lock with cli_sem held.
490 Description: Temporarily disable grant shrink.
491 Details : Disable the feature for debugging.
494 Frequency : only when formatting test filesystems
496 Description: Unable to run several mkfs.lustre on loop devices at the same time
497 Details : mkfs.lustre returns error 256 on the concurrent loop devices
498 formatting. The solution is to proper handle the error.
500 Severity : enhancement
502 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
504 Severity : enhancement
506 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
511 Description: resolve race between obd_disconnect and class_disconnect_exports
512 Details : if obd_disconnect will be called to already disconnected export he
513 forget release one reference and osc module can't unloaded.
515 Severity : enhancement
517 Description: Allow tuning service thread via /proc
518 Details : For each service a new
519 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
520 created that can be used to set min/max thread counts, and get the
521 current number of running threads.
526 Descriptoin: don't return error if have partially created objects for file.
527 Details : lov_update_create_set uses set->set_success as index for created
528 objects, so if some requests will be failed, they will have hole
529 at end of array and qos_shrink_lsm for allocate correct lsm.
531 Severity : enhancement
533 Description: Update OFED support to 1.4.1
538 Description: Reduce small size read RPC
539 Details : Set read-ahead limit for every file and only do read-ahead when
540 available read-ahead pages are bigger than 1M to avoid small size
543 Severity : enhancement
545 Description: move AT tunable parameters for more consistent usage
546 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
548 Severity : enhancement
550 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
551 Details : allow skip disconnected ost for send statfs request and hide error
557 Description: fix racy locking of mballoc block bitmaps causing BUG
558 Details : The locking of the mballoc buddy bitmap and the in-memory
559 block bitmap was using two different spin locks in some
560 cases. This made it possible to incorrectly access the
561 mballoc bitmap while another process was modifying it,
562 causing a sanity assertion to fail. While no on-disk corruption
563 was reported, there was some risk of this happening.
566 Frequency : rare, on llog test 6
568 Descriptoin: don't allow connect to already connected import
569 Details : allowing connect to already connected import is hide connecting problem.
574 Descriptoin: allow kill process which wait statahead result
575 Details : for some reasons 'ls' can stick in waiting result from statahead,
576 in this case need way for kill this process.
578 Severity : enhancement
580 Description: Add state history info file, enhance import info file
581 Details : Track import connection state changes in a new osc/mdc proc file;
582 add overview-type data to the osc/mdc import proc file.
584 Severity : enhancement
586 Description: MDS create should not wait for statfs RPC while holding DLM lock.
588 Severity : enhancement
590 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
595 Descriptoin: OST grant too much space to client even there are not enough space.
596 Details : Client will shrink its grant cache to OST if there are no write
597 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
598 this grant cache if there are already not enough avaible space
599 (left_space < total_clients * 32M).
602 Frequency : start MDS on uncleanly shutdowned MDS device
604 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
605 Details : stay in waiting mds<>ost recovery finished produce random bugs
606 due race between two ll_sync thread for one lov target. send
607 ACTIVATE event only if connect realy finished and import have
611 Frequency : rare, connect and disconnect target at same time
613 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
614 Details : don't call obd_disconnect under lov_lock. this long time
615 operation and can block ptlrpcd which answer to connect request.
620 Descriptoin: don't lose wakeup for imp_recovery_waitq
621 Details : recover_import_no_retry or invalidate_import and import_close can
622 both sleep on imp_recovery_waitq, but we was send only one wakeup
626 Frequency : always with long access acl
628 Descriptoin: mds can't pack reply with long acl.
629 Details : mds don't control size of acl but they limited by reint/getattr
632 Severity : enhancement
634 Description: Update to SLES10 kernel-2.6.16.60-0.33.
636 Severity : enhancement
638 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
641 Frequency : start MDS on uncleanly shutdowned MDS device
643 Descriptoin: aborting recovery hang on MDS
644 Details : don't throttle destroy RPCs for the MDT.
647 Frequency : on remount
649 Description: external journal device not working after the remount
650 Details : clear dev_rdonly flag for external journal devices in
656 Description: shutdown vs evict race
657 Details : client_disconnect_export vs connect request race.
658 if client will evicted at this time - we start invalidate
659 thread without referece to import and import can be freed
663 Frequency : rare, need acl's on inode.
665 Description: client can't handle ost additional correctly
666 Details : if ost was added after client connected to mds client can have
667 hit lnet_try_match_md ... to big messages to wide striped files.
668 in this case need teach client to handle config events about add
669 lov target and update client max ea size at that event.
671 Severity : enhancement
673 Description: Changelogs
674 Details : Changelogs are a lightweight mechanism to track filesystem
675 metadata and namespace changes. The changelog is recorded
676 permanently on the MDTs, and is periodically "consumed" / purged
677 when records are no longer needed.
679 Severity : enhancement
681 Description: compact fld format with extents
682 Details : Store range of seq rather than every seq in FLD. Seq
683 controller update FLD rather than clients. In Case of CMD, mdt0
684 has FLD, all other metadata server act as non persistent proxy
685 for FLD queries and cache fld entries in fld cache.
690 Description: don't skip ost target if they assigned to file
691 Details : Drop slow OSCs if we can, but not for requested start idx.
692 This means "if OSC is slow and it is not the requested
693 start OST, then it can be skipped, otherwise skip it only
694 if it is inactive/recovering/out-of-space.
698 Description: more cleanup in mds_lov
699 Details : not send LOV EA under replay, we can't know about they size at this
700 time. Don't allow client connect to mds before any ost connected,
701 for avoid problems with LOV EA size and returning EIO to client.
703 Severity : enhancement
705 Description: Interoperability at server side (Disk interoperability)
707 Severity : enhancement
709 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
711 Severity : enhancement
713 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
715 Severity : enhancement
717 Description: New client IO stack (CLIO).
719 Severity : enhancement
721 Description: Commit on sharing. Eliminate inter-client dependencies between
722 uncommitted transactions by doing transaction commits.
723 Thereby clients may recovery independently.
726 Frequency : Create a symlink file with a very long name
728 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
729 Details : If there is no extra space in the request for early cancels,
730 ldlm_req_handles_avail() returns 0 instead of a negative value.
732 Severity : enhancement
734 Description: Add /proc entry for import status
735 Details : The mdc, osc, and mgc import directories now have
736 an import directory that contains useful import data for debugging
739 Severity : enhancement
741 Description: Re-disable certain /proc logging
742 Details : Enable and disable client's offset_stats, extents_stats and
743 extents_stats_per_process stats logging on the fly.
746 Frequency : Only on FC kernels 2.6.22+
748 Description: oops in statahead
749 Details : Do not drop reference count for the dentry from VFS when lookup,
750 VFS will do that by itself.
752 Severity : enhancement
754 Description: Generic /proc file permissions
755 Details : Set /Proc file permissions in a more generic way to enable non-
756 root users operate on some /proc files.
760 Description: Hitting mdc_commit_close() ASSERTION
761 Details : Properly handle request reference release in
762 ll_release_openhandle().
766 Description: quota recovery deadlock during mds failover
767 Details : This patch includes att18982, att18236, att18237 in bz14840.
769 1. fix osts hang when mds does failover with quotaon
770 2. prevent watchdog storm when osts threads wait for the
775 Frequency : only patchless client
776 Description: add workaround for race between add/remove dentry from hash
778 Severity : enhancement
780 Description: Allow OST glimpses to return PW locks
784 Description: LBUG when llog conf file is full
785 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
789 Description: Prevent import from entering FULL state when server in recovery
793 Description: service mount cannot take device name with ":"
794 Details : Only when device name contains ":/" will mount treat it as
800 Description: replace ptlrpcd with the statahead thread to interpret the async
801 statahead RPC callback
805 Frequency : on recovery
806 Description: I/O failures after umount during fail back
807 Details : if client reconnected to restarted server we need join to recovery
808 instead of find server handler is changed and process self
809 eviction with cancel all locks.
811 Severity : enhancement
813 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
815 Severity : enhancement
817 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
819 Severity : enhancement
821 Description: Upcall on Lustre log has been dumped
822 Details : Allow for a user mode script to be called once a Lustre log has
823 been dumped. It passes the filename of the dumped log to the
824 script, the location of the script can be specified via
825 /proc/sys/lnet/debug_log_upcall.
830 Description: avoid idr_remove called for id which is not allocated.
831 Details : Move assigment s_dev for clustered nfs to end of initialization,
832 to avoid problem with error handling.
837 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
838 Details : When connection is reused this not moved from CONN_UNUSED_HASH
839 into CONN_USED_HASH and this prodice warning when put connection
840 again in unused hash.
843 Severity : enhancement
845 Description: File striping can now be set to use an arbitrary pool of OSTs.
847 Severity : enhancement
849 Description: Export bytes_read/bytes_write count on OSC/OST.
853 Description: Early reply size mismatch, MGC loses connection
854 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
855 the connect flags are properly negotiated.
860 Description: quotas are not honored with O_DIRECT
861 Details : all writes with the flag O_DIRECT will use grants which leads to
862 this problem. Now using OBD_BRW_SYNC to guard this.
866 Description: add quota statistics
867 Details : 1. sort out quota proc entries and proc code.
868 2. add quota statistics
870 Severity : enhancement
872 Description: enable quota support for HEAD.
876 Description: Properly propagate oinfo flags from lov to osc for statfs
877 Details : restore missing copy oi_flags to lov requests.
879 Severity : enhancement
881 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
882 : l_getgroups(8), lst(8), routerstat(8)
883 Details : included man pages for llobdstat(8), llstat(8),
884 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
886 Severity : enhancement
888 Description: configure's --enable-quota should check the
889 : kernel .config for CONFIG_QUOTA
890 Details : configure is terminated if --enable-quota is passed but
891 : no quota support is in kernel
895 Description: Remove portals compatibility
896 Details : Remove portals compatibility, not interoperable with releases
901 Description: Resolve device initialization race
902 Details : Prevent proc handler from accessing devices added to the
903 obd_devs array but yet be intialized.
905 Severity : enhancement
907 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
909 Severity : enhancement
911 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
916 Description: Using wrong pointer in osc_brw_prep_request
917 Details : Access to array[-1] can produce panic if kernel compiled with
918 CONFIG_PAGE_ALLOC enabled
922 Description: Client runs out of low memory
923 Details : Consider only lowmem when counting initial number of llap pages
927 Description: *optional* service tags registration
928 Details : if the "service tags" package is installed on a Lustre node
929 When the filesystem is mounted, a local-node service tag will
930 be created. See http://inventory.sun.com/ for more information
931 about the Service Tags asset management system.
935 Description: Kernel BUG tries to release flock
936 Details : Lustre does not destroy flock lock before last reference goes
937 away. So always drop flock locks when client is evicted and
938 perform unlock regardless of successfulness of speaking to MDS.
942 Description: add refcount protection for osc callbacks, avoid panic on shutdown
946 Description: sanity test 65a fails if stripecount of -1 is set
947 Details : handle -1 striping on filesystem in ll_dirstripe_verify
952 Description: ASSERTION(CheckWriteback(page,cmd)) failed
953 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
956 Severity : enhancement
958 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
962 Description: do not process already freed flock
963 Details : flock can possibly be freed by another thread before it reaches
964 to ldlm_flock_completion_ast.
968 Description: LBUG during stress test
969 Details : Need properly lock accesses the flock deadlock detection list.
973 Description: oops in page fault handler
974 Details : kernel page fault handler can return two special 'pages' in error
975 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
979 Description: timeout with invalidate import.
980 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
981 should be handled by ptlrpcd. This produce long age waiting and
982 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
984 Severity : enhancement
986 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
990 Description: enable MGS and MDT services start separately
991 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
992 a co-located MGS without starting the MGS, which is a complement
993 to 'nosvc' mount option.
997 Frequency : after recovery
998 Description: precreate to many object's after del orphan.
999 Details : del orphan st in oscc last_id == next_id and this triger growing
1000 count of precreated objects. Set flag LOW to skip increase count
1001 of precreated objects.
1005 Frequency : rare, on clear nid stats
1006 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
1007 Details : when clean nid stats sometimes try destroy live entry,
1008 and this produce panic in free.
1012 Description: Stack overflow during MDS log replay
1013 ease stack pressure by using a thread dealing llog_process.
1017 Description: wait until IO finished before start new when do lock cancel.
1018 Details : VM protocol want old IO finished before start new, in this case
1019 need wait until PG_writeback is cleared until check dirty flag and
1020 call writepages in lock cancel callback.
1022 Severity : enhancement
1024 Description: using special macro for print time and cleanup in includes.
1028 Description: mds_mfd_close() ASSERTION(rc == 0)
1029 Details : In mds_mfd_close(), we need protect inode's writecount change
1030 within its orphan write semaphore to prevent possible races.
1034 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
1038 Frequency : rare, on shutdown ost
1039 Description: don't hit live lock with umount ost.
1040 Details : shrink_dcache_parent can be in long loop with destroy dentries,
1041 use shrink_dcache_sb instead.
1045 Description: don't panic with use echo client
1046 Details : echo client pass NULL as client nid pointer and this produce null
1047 pointer dereference.
1051 Description: fix build on ppc32
1052 Details : compile code with -m64 flag produce wrong object file for ppc32.
1056 Description: add message levels for liblustreapi
1060 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
1061 Details : Move the 'good_osts' check before the 'total_bavail' check. This
1062 will result in an -EAGAIN and in the exit call path we call
1063 alloc_rr() which will with increasing aggressiveness attempt to
1064 aquire precreated objects on the minimum number of required OSCs.
1068 Description: Use old size assignment to avoid deadlock
1069 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
1070 the scheduling while holding a spinlock. We do not need locking
1071 for size in ll_update_inode() because size is only updated from
1072 the MDS for directories or files without objects, so there is no
1073 other place to do the update, and concurrent access to such inodes
1074 are protected by the inode lock.
1078 Description: resolve "_IOWR redefined" build error on SLES10
1082 Description: dump the memory debugging after all modules are unloaded to
1083 suppress false negative in conf_sanity test 39
1085 Severity : enhancement
1087 Description: build kernel-ib packages for OFED 1.3 in our release cycle
1092 Description: fix SLES kernel versioning
1093 Details : the kernel version for our SLES 10 kernel did not include a "-"
1094 before the "smp" at the end. while this was not a problem in
1095 general, it did mean that software trying to use the kernel
1096 version to try to detect a vendor specific kernel would fail.
1097 this was most evident by the OFED build scripts.
1101 Description: Don't update lov_desc members until making sure they are valid
1102 Details : When updating lov_desc members via proc fs, need fix their
1103 validities before doing the real update.
1107 Description: don't put request into delay list while invalidate in flight.
1108 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
1109 import in flight. this produce timeout for invalidate and sometimes
1110 can cause stale data.
1114 Frequency : on ppc only
1115 Description: not convert ost objects for directory because it's not exist.
1116 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
1120 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
1122 Severity : enhancement
1124 Description: Update to support 2.6.22.14 vanilla kernel.
1128 Frequency : rare, on recovery
1129 Description: read procfs can produce deadlock in some situation
1130 Details : Holding lprocfs lock which send rpc can produce block for destroy
1131 obd objects and this also block reconnect with -EALREADY. This
1132 isn't fix all lprocfs bugs - but make it rare.
1134 Severity : enhancement
1136 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
1139 Frequency : frequent on X2 node
1141 Description: mdc_set_open_replay_data LBUG
1142 Details : Set replay data for requests that are eligible for replay.
1146 Description: lustre_mgs: operation 101 on unconnected MGS
1147 Details : When MGC is disconnected from MGS long enough, MGS will evict the
1148 MGC, and late on MGC cannot successfully connect to MGS and a lot
1149 of the error messages complaining that MGS is not connected.
1153 Frequency : on network error
1154 Description: panic with double free request if network error
1155 Details : mdc_finish_enqueue is finish request if any network error ocuring,
1156 but it's true only for synchronus enqueue, for async enqueue
1157 (via ptlrpcd) this incorrect and ptlrpcd want finish request
1160 Severity : enhancement
1162 Description: client-side metadata stat-ahead during readdir(directory readahead)
1163 Details : perform client-side metadata stat-ahead when the client detects
1164 readdir and sequential stat of dir entries therein
1167 Frequency : on start mds
1169 Description: Implement get_info(last_id) in obdfilter.
1172 Frequency : occasional
1174 Description: Correctly check stale fid, not start epoch if ost not support SOM
1175 Details : open with flag O_CREATE need set old fid in op_fid3 because
1176 op_fid2 was overwritten with new generated fid, but mds can answer
1177 with one of these two fids and both is not stale. Setattr
1178 incorrectly started an epoch and assume will be called
1179 done_writing, but without SOM done_writing ever being called.
1182 Frequency : rare, depends on device drivers and load
1184 Description: MDS or OSS nodes crash due to stack overflow
1185 Details : Code changes in 1.8.0 increased the stack usage of some functions.
1186 In some cases, in conjunction with device drivers that use a lot
1187 of stack the MDS (or possibly OSS) service threads could overflow
1188 the stack. One change which was identified to consume additional
1189 stack has been reworked to avoid the extra stack usage.
1192 Frequency : occasional
1194 Description: Do not fail import if osc_interpret_create gets -EAGAIN
1195 Details : If osc_interpret_create got -EAGAIN it immediately exits and
1196 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
1197 oscc_has_objects and see OSC has no objests and call
1198 oscc_internal_create to resend create request.
1200 Severity : enhancement
1202 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
1204 Severity : enhancement
1206 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
1209 Frequency : very rare
1211 Description: Fix replay if there is an un-replied request and open
1212 Details : In some cases, older replay request will revert the
1213 mcd->mcd_last_xid on MDS which is used to record the client's
1214 latest sent request.
1216 Severity : enhancement
1218 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
1220 Severity : enhancement
1222 Description: Add rhel5 support to HEAD.
1224 Serverity : enhancement
1226 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
1231 Description: Don't allow skipping OSTs if index has been specified.
1232 Details : Don't allow skipping OSTs if index has been specified, make
1233 locking in internal create lots better.
1237 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
1238 Details : access to bitfield in structure is always rounded to long
1239 and this produce problem with not atomic change any bit.
1243 Description: Lustre make rpms failed.
1244 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
1245 builds Lustre rpms from tarball.
1248 Frequency : rare on shutdown ost
1250 Description: If llog cancel was not send before clean_exports phase, this can
1251 produce deadlock in llog code.
1252 Details : If llog thread has last reference to obd and call class_import_put
1253 this produce deadlock because llog_cleanup_commit_master wait when
1254 last llog_commit_thread exited, but this never success because was
1255 called from llog_commit_thread.
1259 Description: allow userland application know is lost one of stripes.
1260 Details : fill lvb_blocks with error code on ost and return it to
1261 application if error flag found.
1265 Description: NULL lov_tgts causing MDS oops
1266 Details : more safe checks for NULL lov_tgts for avoid oops.
1268 Severity : enhancement
1270 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
1274 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
1275 Details : If an OST is down the MDS will hang indefinitely in
1276 obd_statfs_rqset() waiting for the statfs data. While for
1277 MDS QOS usage of statfs, it should not stuck in waiting.
1279 Severity : enhancement
1281 Description: remote_acl support
1282 Details : Support ACL-based permission check for remote user.
1283 Support setfacl/getfacl for remote user with the utils
1284 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
1285 the system "{s,g}etfacl" utils.
1287 Severity : enhancement
1289 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
1291 Severity : enhancement
1293 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
1297 Description: make mgs_setparam() handle fsname containing dash
1298 Details : fsname containing a dash does not work with lctl conf_param
1300 Severity : enhancement
1302 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
1304 Severity : enhancement
1306 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
1310 Frequency : rare, at shutdown
1311 Description: access already free / zero obd_namespace.
1312 Details : if client_disconnect_export was called without force flag set,
1313 and exist connect request in flight, this can produce access to
1314 NULL pointer (or already free pointer) when connect_interpret
1315 store ocd flags in obd_namespace.
1319 Frequency : only at startup
1320 Description: not alloc memory with spinlock held.
1321 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
1322 if any spinlock held.
1324 Severity : enhancement
1326 Description: make lustre randomly fail allocating memory
1327 Details : Make lustre randomly failed allocating memory for testing purpose.
1329 Severity : enhancement
1331 Description: lost problems with lov objid file
1332 Details : Fixes some scability and access to not inited memory problems
1333 in work with lov objdid file.
1338 Description: lfs find does not continue on file error
1339 Details : Continue other files processing when a file/dir is absent.
1343 Description: Inconsistent usage of lustre_pack_reply()
1344 Details : Standardize the usage of lustre_pack_reply() such that it
1345 always generate a CERROR on failure.
1348 Frequency : occasional
1350 Description: MDS hang or stay in waiting lock
1351 Details : If client receive lock with CBPENDING flag ldlm need send lock
1352 cancel as separate rpc, to avoid situation when cancel request
1353 can't processed due all i/o threads stay in wait lock.
1357 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1361 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1365 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1369 Description: Update to RHEL4 latest kernel.
1371 Severity : enhancement
1373 Description: Build SLES10 patchless client fails
1374 Details : The configure was broken by run ./configure with
1375 --with-linux-obj=.... argument for patchless client. When the
1376 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1377 can't search header adequately. Use absolute path such as
1378 -I($LINUX)/include instead.
1382 Description: interrupt oig_wait produce painc on resend.
1383 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1384 private set, and this produce situation when rq_ptlrpcd_data not
1385 copyed to new allocated request and triggered LBUG on assert
1386 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1387 ptlrpcd set if request is changed and can be safety changed to use
1392 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1393 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1394 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1395 these macros and replace them with their difinitions instead.
1399 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1400 Details : Update which_patch & target file for SLES10 latest kernel.
1403 Description: Remove client patches from SLES 10 kernel.
1404 Details : This causes SLES 10 clients to behave as patchless clients
1405 even on a Lustre-patched (server) kernel.
1407 Severity : enhancement
1409 Description: self-adjustable client's lru lists
1410 Details : use adaptive algorithm for managing client cached locks lru
1411 lists according to current server load, other client's work
1412 pattern, memory activities, etc. Both, server and client
1413 side namespaces provide number of proc tunables for controlling
1416 Severity : enhancement
1418 Description: light-weight GSS support
1419 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
1420 and reduce performance overhead.
1422 Severity : enhancement
1424 Description: Linux keyring support
1425 Details : Support using service of Linux keyring for Lustre GSS internal
1426 context refresh/cache mechanism.
1430 Description: Fix errors in lfs documentation
1431 Details : Fixes man pages
1435 Description: don't use GFP_* in generic Lustre code.
1436 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
1440 Description: obdclass is limited by single OBD_ALLOC(idarray)
1441 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
1445 Description: warnings with build patchless client with vanila 2.6.19 and up
1446 Details : change old ctl_table style and replace ctl_table/ctl_table_header
1447 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
1451 Frequency : Only for SLES
1452 Description: sanity_quota fail test_1
1453 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
1454 file, which makes TSTID[2] inunique.
1458 Frequency : Only for SLES9
1459 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
1460 Details : Update target/ChangeLog/which_patch .
1464 Frequency : Only for test_10 in sanity-quota.sh
1465 Description: a bug in quota test code
1466 Details : checking a test flag in a wrong place
1470 Frequency : Only for test_13 in sanity-quota.sh
1471 Description: a bug in quota test script
1472 Details : shouldn't check overfull allocated quota
1476 Description: jbd statistics
1477 Details : Port older jbd statistics patch for sles10
1479 Severity : enhancement
1481 Description: Add printing inode info into message about error in writepage.
1486 Description: reply_lock_interpret crash due to race with it and lock cancel.
1487 Details : Do not replay locks that are being cancelled. Do not reference
1488 locks by their address during replay, just by their handle.
1492 Frequency : When flocks are used.
1493 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
1494 Details : Flock locks might destroy just granted lock if it could be merged
1495 with another existing flock, this is done in completion handler,
1496 so teach ldlm_cli_enquque_fini that this is a valid case for
1502 Description: Oops in read and write path when failing to allocate lock.
1503 Details : Check if lock allocation failed and return error back.
1507 Description: lstripe command fails for valid OST index
1508 Details : The stripe offset is compared to lov->desc.ld_tgt_count
1509 instead of lov->desc.ld_active_tgt_count.
1513 Description: sanity.sh failed test 103
1514 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
1519 Description: df doesn't work properly if diskfs blocksize != 4K
1520 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
1524 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
1525 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
1527 Severity : enhancement
1529 Description: Adaptive timeouts
1530 Details : RPC timeouts adapt to changing server load and network
1531 conditions to reduce resend attempts and improve recovery time.
1535 Description: llapi_file_create() does not allow some changes
1536 Details : add llapi_file_open() that allows specifying the mode and
1537 open flags, and also returns an open file handle.
1541 Description: merge and cleanup kernel patches.
1542 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
1546 Description: Add journal checksum support.(Kernel part)
1547 Details : The journal checksum feature adds two new flags i.e
1548 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
1549 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
1550 indicates that the commit block contains the checksum for
1551 the blocks described by the descriptor blocks. Now commit
1552 record can be sent to disk without waiting for descriptor
1553 blocks to be written to disk. This behavior is controlled
1554 using JBD2_FEATURE_ASYNC_COMMIT flag.
1558 Description: OSS needs mutliple precreate threads
1559 Details : Add ability to start more than one create thread per OSS.
1563 Description: Sanity.sh test_65k failed: setstripe should have succeeded
1564 Details : execute lfs setstripe on client
1568 Description: mds_obd_create error creating tmp object
1569 Details : When the user sets quota on root, llog will be affected and can't
1570 create files and write files.
1573 Frequency : Always on ia64 patchless client, and possibly others.
1575 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
1576 Details : This allows the patchless client to be loaded on architectures
1577 without this export.
1581 Description: RedHat Update kernel for RHEL5
1582 Details : Add the kernel config file for RHEL5.
1586 Description: RedHat Update kernel for RHEL5
1587 Details : Modify the kernel config file more closer RHEL5.
1591 Description: Build failure against Centos5 (RHEL5)
1592 Details : Define PAGE_SIZE when it isn't present.
1596 Description: test_7 on sanity-quota.sh can't be used on separate machines
1597 Details : add facet to handle it
1601 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
1602 Details : with NFS, the anon dentry's parent was set to itself in
1603 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
1604 corresponding dentry other than use rec->ur_name.
1606 Severity : enhancement
1608 Description: lfs setstripe enhancement
1609 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
1613 Description: enable data checksumming by default
1614 Details : enable checksum by default, allow --disable-checksum
1615 configure option and "-o nochecksum" mount option. Checksums
1616 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
1620 Description: lustre support for RHEL5
1621 Details : Add support for RHEL5.
1625 Description: Client eviction due to failover config
1626 Details : after a connection loss, the lustre client should attempt to
1627 reconnect to the last active server first before trying the
1628 other potential connections.
1632 Description: when mds and osts use different quota unit(32bit and 64bit),
1633 quota will be released repeatly.
1634 Details : void sending multiple quota reqs to mds, which will keep the
1635 status between the reqs.
1639 Description: osts not allocated evenly to files
1640 Details : change the condition to increase offset_idx
1643 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
1645 Description: Data corruption for OSTs that are formatted larger than 2TB
1647 Details : When generating the bio request for lustre file writes the
1648 sector number would overflow a temporary variable before being
1649 used for the IO. The data reads correctly from Lustre (which
1650 will overflow in a similar manner) but other file data or
1651 filesystem metadata may be corrupted in some cases.
1655 Description: Tune the kernel for good SCSI performance.
1656 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
1657 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
1662 Description: rewrite ext2-derived code in obdclass/uuid.c
1663 Details : rewrite inherited code (uuid parsing code from ext2 utils)
1664 from scratch preserving functionality.
1668 Description: "lfs find -obd UUID" prints directories
1669 Details : "lfs find -obd UUID" will return all directory names instead
1670 of just file names. It is incorrect because the directories
1671 do not reside on the OSTs.
1675 Description: use i_size_read and i_size_write in 2.6 port
1676 Details : replace inode->i_size access with i_size_read/write()
1679 Frequency : when removing large files
1681 Description: scheduling issue during removal of large Lustre files
1682 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
1683 It causes scheduling issues when removing large files (17TB in the
1687 Frequency : only with liblustre clients on XT3
1689 Description: evictions taking too long
1690 Details : allow llrd to evict clients directly on OSTs
1693 Frequency : only on ppc
1695 Description: /proc/fs/lustre/devices broken on ppc
1696 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
1697 We should make sure the type of 'index' is loff_t and then cast
1698 explicitly as needed below. Do not assign an explicitly cast
1702 Frequency : only for rhel5
1704 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
1705 Details : Modify the target file & which_kernel.
1707 Severity : enhancement
1709 Description: omit set fsid for export NFS
1710 Details : fix set/restore device id for avoid EMFILE error and mark lustre
1711 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
1715 Frequency : Always, for kernels after 2.6.16
1716 Description: Fix warning idr_remove called for id=.. which is not allocated.
1717 Details : Last kernels save old s_dev before kill super and not allow
1718 to restore from callback - restore it before call kill_anon_super.
1722 Description: buffer overruns could theoretically occur
1723 Details : llapi_semantic_traverse() modifies the "path" argument by
1724 appending values to the end of the origin string, and a
1725 overrun may occur. Adding buffer overrun check in liblustreapi.
1729 Description: Fix error on 'ls .' at the top of the Lustre mount.
1730 Details : Don't revalidate dentry if it is a root dentry.
1734 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
1735 Details : Modify vm-tunables-rhel4.patch.
1739 Description: Kernel config for 2.6.18-vanilla.
1740 Details : Modify targets/2.6-vanilla.target.in.
1741 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
1742 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
1743 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
1744 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
1748 Description: improve handling recoverable errors
1749 Details : if request processig with error which can be recoverable on server
1750 request should be resend, otherwise page released from cache and
1755 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
1756 Details : Modify target file & which_patch.
1757 A flaw was found in the IA32 system call emulation provided
1758 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
1759 value could be stored in the %RAX register, which could trigger an
1760 out-of-bounds system call table access. An untrusted local user
1761 could exploit this flaw to run code in the kernel
1762 (ie a root privilege escalation). (CVE-2007-4573).
1766 Description: O_DIRECT bypasses client statistics.
1767 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
1768 read_ahead_stats were updated. copied the stats section from
1769 osc_send_oap_rpc() into async_internal().
1773 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
1777 Description: change order of libsysio includes
1778 Details : '#include sysio.h' should always come before '#include xtio.h'
1782 Description: handle "serious error: objid * already exists" more gracefully
1783 Details : If LAST_ID value on disk is smaller than the objects existing in
1784 the O/0/d* directories, it indicates disk corruption and causes an
1785 LBUG(). If the object is 0-length, then we should use the existing
1786 object. This will help to avoid a full fsck in most cases.
1788 Severity : enhancement
1790 Description: adapt the lustre_config script to support the upgrade case
1791 Details : Add "-u" option for lustre_config script to support upgrading 1.4
1792 server targets to 1.6 in parallel.
1796 Description: To avoid grant space > avaible space when the disk is almost
1797 full. Without this patch you might see the error "grant XXXX >
1798 available" or some LBUG about grant, when the disk is almost
1800 Details : In filter_check_grant, for non_grant cache write, we should
1801 check the left space by if (*left > ungranted + bytes), instead
1802 of (*left > ungranted), because only we are sure the left space
1803 is enough for another "bytes", then the ungrant space should be
1804 increase. In client, we should update cl_avail_grant only there
1805 is OBD_MD_FLGRANT in the reply.
1809 Description: Update RHEL 4 kernel to fix local root privilege escalation.
1810 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
1811 described in CVE-2007-4573. This problem could allow untrusted
1812 local users to gain root access.
1815 Frequency : when using O_DIRECT and quotas
1817 Description: Incorrect file ownership on O_DIRECT output files
1818 Details : block usage reported by 'lfs quota' does not take into account
1819 files that have been written with O_DIRECT.
1824 Description: touch file failed when fs is not full
1825 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
1826 otherwise we can get ENOSP while fs is not full.
1830 Description: parallel lock callbacks
1831 Details : Instead of sending blocking and completion callbacks as separated
1832 requests, adding them to a set and sending in parallel.
1835 Frequency : only for Cray XT3
1836 Bugzilla : 12829/13455
1837 Description: Changing primary group doesn't change the group lustre assigns to
1839 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
1840 overridden with the primary group provided by the group upcall,
1841 whereas the supplied fsgid can be trusted if it is in the list of
1842 supplementary groups returned by the group upcall.
1844 Severity : enhancement
1846 Description: Allow masking D_WARNING, D_ERROR messages from console
1847 Details : Console messages can now be disabled via lnet.printk.
1851 Description: User code with malformed file open parameter crashes client node
1852 Details : Before packing join_file req, all the related reference should be
1853 checked carefully in case some malformed flags cause fake
1854 join_file req on client.
1858 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
1859 Details : ldlm_completion_ast() assumes that a lock is granted when the req
1860 mode is equal to the granted mode. However, it should also check
1861 that LDLM_FL_CP_REQD is not set.
1865 Description: Heavy nfs access might result in deadlocks
1866 Details : After ELC code landed, it is now improper to enqueue any mds
1867 locks under och_sem, because enqueue might want to decide to
1868 cancel open locks for same inode we are holding och_sem for.
1872 Description: Client eviction while running blogbench
1873 Details : A lot of unlink operations with concurrent I/O can lead to a
1874 deadlock causing evictions. To address the problem, the number of
1875 oustanding OST_DESTROY requests is now throttled to
1876 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
1877 callbacks are processed in priority.
1881 Description: enable ACLs on MDS by default
1882 Details : ACLs must be enabled on MDS by default.
1885 Frequency : PPC/PPC64 only
1887 Description: conflicts between asm-ppc64/types.h and lustre_types.h
1888 Details : fix duplicated definitions between asm-ppc64/types.h and
1889 lustre_types.h on PPC.
1892 Frequency : PPC/PPC64 only
1894 Description: asm-ppc/segment.h does not exist
1895 Details : fix compile issue on PPC.
1899 Description: better handle error messages in extents code
1902 Frequency : RHEL4 only
1904 Description: mkfs is very slow on IA64/RHEL4
1905 Details : A performance regression has been discovered in the MPT Fusion
1906 driver between versions 3.02.73rh and 3.02.99.00rh. As a
1907 consequence, we have downgraded the MPT Fusion driver in the RHEL4
1908 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
1910 Severity : enhancement
1912 Description: SNMP support enhancement
1913 Details : Adding total number of sampled request for an MDS node in snmp
1916 Severity : enhancement
1918 Description: Optimize ldlm waiting list processing for PR extent locks
1919 Details : When processing waiting list for read extent lock and meeting read
1920 lock that is same or wider to it that is not contended, skip
1921 processing rest of the list and immediatelly return current
1922 status of conflictness, since we are guaranteed there are no
1923 conflicting locks in the rest of the list.
1927 Description: Time out and refuse to reconnect
1928 Details : When the failover node is the primary node, it is possible
1929 to have two identical connections in imp_conn_list. We must
1930 compare not conn's pointers but NIDs, otherwise we can defeat
1931 connection throttling.
1935 Description: port llog fixes from b1_6 into HEAD
1936 Details : Port llog reference couting and some llog cleanups from b1_6
1937 (bug 10800) into HEAD, for protect from panic and access to already
1938 free llog structures.
1942 Description: Detect stride IO mode in read-ahead
1943 Details : When a client does stride read, read-ahead should detect that and
1944 read-ahead pages according to the detected stride pattern.
1948 Description: data checksumming impacts single node performance
1949 Details : add support for several checksum algorithm. Currently, only CRC32
1950 and Adler-32 are supported. The checksum type can be changed on
1951 the fly via /proc/fs/lustre/osc/*/checksum_type.
1955 Description: use adler32 for page checksums
1956 Details : when available, use the Adler-32 algorithm instead of CRC32 for
1961 Description: build for x2 fails
1962 Details : fix compile issue on Cray systems.
1966 Description: Properly match for duplicate locks
1967 Details : Due to different lock order from skiplists code, we need to
1968 traverse entire list for now
1971 Frequency : only on PPC/SLES10
1973 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
1974 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
1975 be defined. Add a hack in mkfs_lustre.c to work around this compile
1980 Description: LASSERT on MDS when client holding flock lock dies
1981 Details : ldlm pool logic depends on number of granted locks equal to
1982 number of released locks which is not true for flock locks, so
1983 just exclude such locks from consideration.
1987 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
1988 Details : Use fsfilt_sync() for both the whole filesystem sync and
1989 individual file sync to eliminate dangerous inode locking
1990 with I_LOCK that can lead to a deadlock.
1994 Description: performance in 1.6.3
1995 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
1996 sure that 1MB requests are not fragmented by the block layer.
1998 Severity : enhancement
2000 Description: organize the server-side client stats on per-nid basis
2001 Details : Change the structure of stats under obdfilter and mds to
2009 The "uuid"s file would list the uuids of _active_ exports.
2010 And the clear entry is to clear all stats and stale nids.
2012 Severity : enhancement
2014 Description: eliminate client locks in face of contention
2015 Details : file contention detection and lockless i/o implementation
2016 for contended files.
2020 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
2023 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
2025 Description: MDS may lose file striping (and hence file data) in some cases
2026 Details : If there are additional extended attributes stored on the MDS,
2027 in particular ACLs, SELinux, or user attributes (if user_xattr
2028 is specified for the client mount options) then there is a risk
2029 of attribute loss. Additionally, the Lustre file striping
2030 needs to be larger than default (e.g. striped over all OSTs),
2031 and an additional attribute must be stored initially in the
2032 inode and then increase in size enough to be moved to the
2033 external attribute block (e.g. ACL growing in size) for file
2038 Description: skiplist implementation simplification
2039 Details : skiplists are used to group compatible locks on granted list
2040 that was implemented as tracking first and last lock of each
2041 lock group the patch changes that to using doubly linked lists
2045 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
2046 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
2047 because LASSERT on both the data supplied by a client, and the
2048 data on disk is dangerous and incorrect.
2050 Severity : enhancement
2052 Description: Slow truncate/writes to huge files at high offsets.
2053 Details : Directly associate cached pages to lock that protect those pages,
2054 this allows us to quickly find what pages to write and remove
2055 once lock callback is received.
2059 Description: more ldlm soft lockups
2060 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
2061 starve other threads from the resource lock for a long time in
2062 case of long waiting queue, so change the debug level from
2063 D_OTHER to the less frequently used D_INFO.
2065 Severity : enhancement
2067 Description: add -gid, -group, -uid, -user options to lfs find
2071 Description: Hung threads in invalidate_inode_pages2_range
2072 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
2073 once one is completed. As a result, some RPCs are stuck in the
2074 queue and are never sent.
2078 Description: filter threads hungs on waiting journal commit
2079 Details : Cleanup filter group llog code, then only filter group llog will
2080 be only created in the MDS/OST syncing process.
2084 Description: Procfs and llog threads access destoryed import sometimes.
2085 Details : Sync the import destoryed process with procfs and llog threads by
2086 the import refcount and semaphore.
2088 Severity : enhancement
2090 Description: openlock cache of b1_6 port to HEAD
2095 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
2096 Details : If insertion of an extent fails, then discard the inode
2097 preallocation and free data blocks else it can lead to duplicate
2102 Description: don't always update ctime in ext3_xattr_set_handle()
2103 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
2104 In some cases the ctime should not be updated, for example for
2105 2.0->1.8 compatibility it is necessary to delete an xattr and it
2106 should not update the ctime.
2110 Bugzilla : 15713/16362
2111 Description: Assertion in iopen_connect_dentry in 1.6.3
2112 Details : looking up an inode via iopen with the wrong generation number can
2113 populate the dcache with a disconneced dentry while the inode
2114 number is in the process of being reallocated. This causes an
2115 assertion failure in iopen since the inode's dentry list contains
2116 both a connected and disconnected dentry.
2120 Description: assertion failure in ldlm_handle2lock()
2121 Details : fix a race between class_handle_unhash() and class_handle2object()
2122 introduced in lustre 1.6.5 by bug 13622.
2127 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
2128 Details : remove the limit of 256 scsi disks in the sd_iostat patch
2133 Description: oops in sd_iostats_seq_show()
2134 Details : unloading/reloading the scsi low level driver triggers a kernel
2135 bug when trying to access the sd iostat file.
2140 Description: Kernel panics during QLogic driver reload
2141 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
2142 patch, causing memory corruption.
2147 Description: journal_dev option does not work in b1_6
2148 Details : pass mount option during pre-mount.
2150 Severity : enhancement
2152 Description: Add a FIEMAP(FIle Extent MAP) ioctl
2153 Details : FIEMAP ioctl will allow an application to efficiently fetch the
2154 extent information of a file. It can be used to map logical blocks
2155 in a file to physical blocks in the block device.
2159 Description: LDLM soft lockups - improvement
2160 Details : It is be possible to send the lock handle along with each read
2161 or write request because the client is already doing a lock match
2162 itself so there isn't any reason the OST should have to re-do that
2166 Frequency : only on Cray X2
2168 Description: X2 build failures
2169 Details : fix build failures on Cray X2.
2173 Description: xid & resent requests
2174 Details : Initialize RPC XID from clock at startup (randomly if clock is
2177 Severity : enhancement
2179 Description: Add lustre_start utility to start or stop multiple Lustre servers
2184 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
2185 Details : In case of memory pressure, list_del() can be called twice on
2186 req->rq_history_list, causing a kernel oops.
2190 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
2191 Details : fix stack overflow in the distributed lock manager by defering
2192 export eviction after a failed AST to the elt thread instead of
2193 handling it in the dlm interpret routine.
2197 Description: Convert some comments to new format.
2198 Details : Update documenting comments to match doxygen conventions.
2202 Description: Grammar fixes.
2203 Details : A couple of trivial spelling fixes.
2207 Description: OSD_COUNTERS-mandatory
2208 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
2209 to simplify the code.
2213 Description: simplify cmm_device freeing logic.
2214 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
2218 Description: Add lockdep support to dt_object_operations locking interface.
2219 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
2220 parameter indicating lock ordering. Update mdd code to use new
2225 Description: Introduce failloc constants for lockless IO tests.
2226 Details : Add two new failloc constants to test lockless IO. Only one of
2227 them in implemented---another is checked in yet to be landed
2232 Description: Add lockdep support for inode mutex.
2233 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
2234 in the situations where Lustre has to lock more than one inode
2239 Description: Add optional invariants checking support.
2240 Details : Add new LINVRNT() macro, optional on new --enable-invariants
2241 configure switch. This macro is to be used for consistency and
2242 sanity checks that are too expensive to be left in `production'
2247 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
2248 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
2253 Description: Add lockdep support for ldlm_lock and ldlm_resource.
2254 Details : Use spin_lock_nested() in (the only) situation where more than
2255 one ldlm_lock is locked simultaneously. Also, fix possible
2256 dead-lock in ldlm_lock_change_resource() by enforcing particular
2261 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
2262 Details : Instead of specifying each ldlm_lock call-back through separate
2263 parameter, wrap them into struct ldlm_callback_suite.
2267 Description: Kill join_lru obd method and its callers.
2268 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
2269 in memory---a requirement implemented through ->o_join_lru() obd
2270 method in HEAD. Remove this method and its users.
2274 Description: Add asynchronous ldlm ENQUEUE completion handler.
2275 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
2276 case that stock ldlm_completion_ast() cannot handle as it waits
2277 until lock is granted. Introduce new ldlm_completion_ast_async()
2278 for this. Also comment ldlm_completion_ast().
2282 Description: ldlm_error <-> errno conversion.
2283 Details : Add functions to map (rather arbitrary) between LDLM error codes
2284 and standard errno values. CLIO needs this to prevent LDLM specific
2285 constants from escaping ldlm and osc.
2289 Description: Kill unused ldlm_handle2lock_ns() function.
2290 Details : Kill unused ldlm_handle2lock_ns() function.
2294 Description: Add lu_ref support to ldlm_lock
2295 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
2296 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
2297 and ldlm_resource. LDLM interface has to be changed, because code that
2298 releases a reference on a lock, has to "know" what reference this is.
2299 In the most frequent case
2301 lock = ldlm_handle2lock(handle);
2303 LDLM_LOCK_PUT(lock);
2305 no changes are required. When any other reference (received _not_ from
2306 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
2307 instead of LDLM_LOCK_PUT().
2309 Arguably, changes are pervasive, and interface requires some discipline
2310 for proper use. On the other hand, it was very instrumental in finding
2311 a few leaked lock references.
2315 Description: Add ldlm_lock_addref_try().
2316 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
2317 attempts to addref a lock that might be being canceled
2322 Description: Add ldlm_weigh_callback().
2323 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
2324 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
2325 hard-coded `number of pages' logic.
2329 Description: Add lockdep annotations to llog code.
2330 Details : Use appropriately tagged _nested() locking calls in the places
2331 where llog takes more than one ->lgh_lock lock.
2335 Description: Add loi_kms_set().
2336 Details : Wrap kms updates into a helper function.
2340 Description: Constify instances of struct lsm_operations.
2341 Details : Constify instances of struct lsm_operations.
2345 Description: lu_conf support.
2346 Details : On a server, a file system object is uniquely identified
2347 by a fid, which is sufficient to locate and load all object
2348 state (inode). On a client, on the other hand, more data are
2349 necessary instantiate an object. Change lu_object_find() and
2350 friends to take additional `lu_conf' argument describing object.
2351 Typically this includes layout information.
2355 Description: lu_context fixes.
2356 Details : Introduce new lu_context functions that are needed on the client
2357 side, where some system threads (ptlrpcd) are shared by multiple
2358 modules, and so cannot be stopped during module shutdown.
2362 Description: Add start and stop methods to lu_device_type_operations.
2363 Details : Introduce two new methods in lu_device_type_operations, that are
2364 invoked when first instance of a given type is created and last one
2365 is destroyed respectively. This is need by CLIO.
2369 Description: Add lu_ref support to struct lu_device.
2370 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
2371 track leaked references.
2375 Description: Introduce lu_kmem_descr.
2376 Details : lu_kmem_descr and its companion interface allow to create
2377 and destroy a number of kmem caches at once.
2381 Description: Fix lu_object finalization race.
2382 Details : Fix a race between lu_object_find() finding an object and its
2383 concurrent finalization. This race is (most likely) not possible
2384 on the server, but might happen on the client.
2388 Description: Introduce lu_ref interface.
2389 Details : lu_ref is a debugging module allowing to track references to
2390 a given object. It is quite cpu expensive, and has to be
2391 explicitly enabled with --enable-lu_ref. See usage description
2396 Description: Factor lu_site procfs stats into a separate function.
2397 Details : Separate lu_site stats printing code into a separate function
2398 that can be reused on a client.
2402 Description: Constify instances of struct {lu,dt,md}_device_operations.
2403 Details : Constify instances of struct {lu,dt,md}_device_operations.
2407 Description: Introduce struct md_site and move meta-data specific parts of
2408 struct lu_site here.
2409 Details : Move md-specific fields out of struct lu_site into special struct
2410 md_site, so that lu_site can be used on a client.
2414 Description: Kill mdd_lov_destroy().
2415 Details : Remove unused mdd code.
2419 Description: Add st_block checking to multistat.c.
2420 Details : Add st_block checking to multistat.c.
2424 Description: Add lu_ref support to struct obd_device.
2425 Details : Add lu_ref tracking to obd_device.
2429 Description: Kill obd_set_fail_loc().
2430 Details : Remove unused code.
2434 Description: Add special type for ptlrpc_request interpret functions.
2435 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
2436 there. Actual usage will be in CLIO.
2440 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
2441 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
2442 doesn't work with lockdep.
2446 Description: Add rwv.c test program.
2447 Details : New testing program exercising readv(2) and writev(2) (Qian).
2451 Description: Add sendfile.c test program.
2452 Details : New testing program exercising sendfile(2) (Jay).
2456 Description: Ratelimit a message that can be very frequent.
2457 Details : Ratelimit a memory allocation failure message that can
2462 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2463 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2467 Description: Liblustre build fixes.
2468 Details : Liblustre build fixes.
2472 Description: libcfs: add cfs_{need,cond}_resched() interface.
2473 Details : libcfs: add cfs_{need,cond}_resched() definition and
2474 implementations for Linux, NT, and liblustre.
2476 Severity : enhancement
2478 Description: More exported tunables for mballoc
2479 Details : Add support for tunable preallocation window and new tunables for
2480 large/small requests
2484 Description: Detect corruption of block bitmap and checking for preallocations
2485 Details : Checks validity of on-disk block bitmap. Also it does better
2486 checking of number of applied preallocations. When corruption is
2487 found, it turns filesystem readonly to prevent further corruptions.
2491 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
2492 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
2493 window to 0. If it does hit cache miss, and read-pattern is still
2494 stride-io mode, does not reset the stride window, but also does not
2495 increase the stride window length in this case.
2499 Frequency : only for big-endian servers
2500 Description: Check if system is big-endian while mounting fs with extents feature
2501 Details : Mounting a filesystem with extents feature will fail on big-endian
2502 systems since ext3-based ldiskfs is not supported on big-endian
2503 systems. This can be over-riden with "bigendian_extents" mount option.
2505 Severity : enhancement
2507 Description: The root squash functionality
2508 Details : A security feature, which is to prevent users from being able
2509 to mount lustre on their desktop, run as root, and delete
2510 all of the files in the filesystem. The goal is accomplished by
2511 remapping user id (UID) and group id (GID) of the root user to
2512 a UID and GID specified by the system administartor via Lustre
2513 configuration management server (MGS). The functionality also
2514 allows to specify sets of clients for which the remapping does
2519 Description: Excessive recovery window
2520 Details : With AT enabled, the recovery window can be excessively long (6000+
2521 seconds). To address this problem, we no longer use
2522 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
2523 timeout no longer depends on the service time, it is set to
2524 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
2525 time via pb_service_time.
2529 Description: Watchdog triggered on MDS failover
2530 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
2531 the OSTs know that the MDS "UUID" can be reused for the same export
2532 from a different NID, so we do not need to wait for the export to be
2536 Frequency : rare, only if using MMP with Linux RAID
2538 Description: MMP doesn't work with Linux RAID
2539 Details : While using HA for Lustre servers with Linux RAID, it is possible
2540 that MMP will not detect multiple mounts. To make this work we
2541 need to unplug the device queue in RAID when the MMP block is being
2542 written. Also while reading the MMP block, we should read it from
2543 disk and not the cached one.
2545 Severity : enhancement
2547 Description: open file using fid
2548 Details : A file can be opened using just its fid, like
2549 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
2552 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
2555 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
2556 Details : add proper locking for creation of jbd_4k slab cache
2560 Description: MMP check in ext3_remount() fails without displaying any error
2561 Details : When multiple mount protection fails during remount, proper error
2564 Severity : enhancement
2566 Description: Allow stripe size to be up to 4G-64k
2567 Details : Fix math logic to allow large stripe sizes.
2571 Description: add check for >8TB ldiskfs filesystems
2572 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
2573 Don't allow >8TB ldiskfs filesystems to be mounted without
2574 force_over_8tb mount option
2579 Descriptoin: OSS complains "too many blocks" and crashes
2580 Details : In filter_commitrw_write():
2581 1. lquota_chkquota();
2582 2. obdo_from_inode();
2583 3. lquota_pending_commit();
2584 "2." has changed uid and gid in this case, then caused
2585 lquota_chkquota() and lquota_pending_commit() to use
2586 different uid and gid. That is the root of the bug.
2590 Descriptoin: getstripe doesn't report stripe count = -1 sanely
2591 Details : Detect and print out the 'stripe count = -1' case properly.
2595 Description: Out or order replies might be lost on replay
2596 Details : In ptlrpc_retain_replayable_request if we cannot find retained
2597 request with tid smaller then one currently being added, add it
2598 to the start, not end of the list.
2600 --------------------------------------------------------------------------------
2602 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
2604 * Support for kernels:
2605 2.6.5-7.283 (SLES 9),
2606 2.6.9-55.EL (RHEL 4),
2607 2.6.16.46-0.14 (SLES 10),
2608 2.6.18.8 vanilla (kernel.org)
2609 * Client support for unpatched kernels:
2610 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2611 2.6.16 - 2.6.22 vanilla (kernel.org)
2612 * Due to recently discovered recovery problems, we do not recommend
2613 using patchless RHEL 4 clients with this or any earlier release.
2614 * Recommended e2fsprogs version: 1.39.cfs8
2615 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2616 * Starting with this release, the ldiskfs backing filesystem required
2617 by Lustre is now in its own package, lustre-ldiskfs. This package
2618 should be installed. It is versioned separately from Lustre and
2619 may be released separately in future.
2623 Description: block reactivating mgc import until all deactivates complete
2624 Details : Fix race when failing back MDT/MGS to itself (testing)
2627 Frequency : at statup only
2629 Description: mds_lov_synchronize race leads to various problems
2630 Details : simultaneous MDT->OST connections at startup can cause the
2631 sync to abort, leaving the OSC in a bad state.
2633 Severity : enhancement
2635 Description: add optional extra BUILD_VERSION info
2636 Details : add a new environment variable (namely LUSTRE_VERS) which allows
2637 to override the lustre version.
2640 Frequency : 2.6.18 servers only
2642 Description: ll_kern_mount() doesn't release the module reference
2643 Details : The ldiskfs module reference count never drops down to 0
2644 because ll_kern_mount() doesn't release the module reference.
2649 Description: server LBUG when using old ost_num_threads parameter
2650 Details : Accept the old ost_num_threads parameter but warn that it
2651 is deprecated, and fix an off-by-one error that caused an LBUG.
2656 Description: Transient SCSI error results in persistent IO issue
2657 Details : iobuf->dr_error is not reinitialized to 0 between two
2661 Frequency : sometimes when underlying device returns I/O errors
2663 Description: OSTs not going read-only during write failures
2664 Details : OSTs are not remounted read-only when the journal commit threads
2665 get I/O errors because fsfilt_ext3 calls journal_start/stop()
2666 instead of the ext3 wrappers.
2670 Description: poor connect scaling with increasing client count
2671 Details : Don't run filter_grant_sanity_check for more than 100 exports
2672 to improve scaling for large numbers of clients.
2675 Frequency : SLES10 only
2677 Description: sanity-quota.sh quotacheck failed: rc = -22
2678 Details : Quotas cannot be enabled on SLES10.
2681 Frequency : liblustre clients only
2683 Description: getdirentries does not give error when run on compute nodes
2684 Details : getdirentries does not fail when the size specified as an argument
2685 is too small to contain at least one entry
2687 Severity : enhancement
2689 Description: Add LNET router traceability for debug purposes
2690 Details : If a checksum failure occurs with a router as part of the
2691 IO path, the NID of the last router that forwarded the bulk data
2692 is printed so it can be identified.
2697 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
2698 Details : Due to a race condition, liblustre clients were occasionally
2699 evicted incorrectly.
2701 Severity : enhancement
2703 Description: lfs setstripe use optional parameters instead of postional
2706 Severity : enhancement
2708 Description: Nanosecond timestamp support for ldiskfs
2709 Details : The on-disk ldiskfs filesystem has added support for nanosecond
2710 resolution timestamps. There is not yet support for this at
2711 the Lustre filesystem level.
2714 Frequency : during server recovery
2716 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
2717 Details : request with rq_no_resend flag not awake l_wait_event if they get
2721 Frequency : nfs export on patchless client
2723 Description: connectathon hang when test nfs export over patchless client
2724 Details : Disconnected dentry cannot be found with lookup, so we do not need
2725 to unhash it or make it invalid
2728 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
2729 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
2730 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
2735 Description: open req refcounting wrong on reconnect
2736 Details : If reconnect happened between getting open reply from server and
2737 call to mdc_set_replay_data in ll_file_open, we will schedule
2738 replay for unreferenced request that we are about to free.
2739 Subsequent close will crash in variety of ways.
2740 Check that request is still eligible for replay in
2741 mdc_set_replay_data().
2746 Description: disable writes to filesystem when reading health_check file
2747 Details : the default for reading the health_check proc file has changed
2748 to NOT do a journal transaction and write to disk, because this
2749 can cause reads of the /proc file to hang and block HA state
2750 checking on a healthy but otherwise heavily loaded system. It
2751 is possible to return to the previous behaviour during configure
2752 with --enable-health-write.
2754 Severity : enhancement
2756 Description: 64-bit inode version
2757 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
2758 to the inode. This will be required for version-based recovery.
2763 Description: MDS fails to start if a duplicate client export is detected
2764 Details : in some rare cases it was possible for a client to connect to
2765 an MDS multiple times. Upon recovery the MDS would detect this
2766 and fail during startup. Handle this more gracefully.
2768 Severity : enhancement
2770 Description: Add -o localflock option to simulate old noflock
2772 Details : This will achieve local-only flock/fcntl locks
2778 Description: log_commit_thread vs filter_destroy race leads to crash
2779 Details : Take import reference before releasing llog record semaphore
2784 Description: Wrong request locking in request set processing
2785 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
2786 imp_delayed_list, in this place should be used imp_lock.
2789 Frequency : when reconnection
2791 Description: Grant Leak when osc reconnect to OST
2792 Details : When osc reconnect ost, OST(filter) should check whether it
2793 should grant more space to client by comparing fed_grant and
2794 cl_avail_grant, and return the granted space to client instead
2795 of "new granted" space, because client will call osc_init_grant
2796 to update the client grant space info.
2799 Frequency : when client reconnect to OST
2801 Description: Grant Leak when osc do resend and replay bulk write
2802 Details : When osc reconnect to OST, OST(filter)should clear grant info of
2803 bulk write request, because the grant info while be sync between
2804 OSC and OST when reconnect, and we should ignore the grant info
2805 these of resend/replay write req.
2810 Description: Grant space more than avaiable left space sometimes.
2811 Details : When then OST is about to be full, if two bulk writing from
2812 different clients came to OST. Accord the avaliable space of the
2813 OST, the first req should be permitted, and the second one
2814 should be denied by ENOSPC. But if the seconde arrived before
2815 the first one is commited. The OST might wrongly permit second
2816 writing, which will cause grant space > avaiable space.
2819 Frequency : when client is evicted
2821 Description: Grant might be wrongly erased when osc is evicted by OST
2822 Details : when the import is evicted by server, it will fork another
2823 thread ptlrpc_invalidate_import_thread to invalidate the
2824 import, where the grant will be set to 0. While the original
2825 thread will update the grant it got when connecting. So if
2826 the former happened latter, the grant will be wrongly errased
2827 because of this race.
2832 Description: Checking Stale with correct fid
2833 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
2834 to check whether it is stale, because sometimes, we want the
2835 enqueue happened anyway, and op_data.fid2 will not be initialized.
2837 Severity : enhancement
2839 Description: update patchless client
2840 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
2843 Frequency : only with 2.4 kernel
2845 Description: random memory corruption
2846 Details : size of struct ll_inode_info is to big for union inode.u and this
2847 can be cause of random memory corruption.
2852 Description: Memory leak in recovery
2853 Details : Lov_mds_md was not free in an error handler in mds_create_object.
2854 It should also check obd_fail before fsfilt_start, otherwise if
2855 fsfilt_start return -EROFS,(failover mds during mds recovery).
2856 then the req will return with repmsg->transno = 0 and rc = EROFS.
2857 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
2858 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
2859 be freed no matter whether fsfilt_commit success or not.
2862 Frequency : only with huge count clients
2864 Description: Prevents from taking the superblock lock in llap_from_page for
2866 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
2867 for llap_from_page call in ll_removepage prevents from taking the
2868 superblock lock for a soon died page.
2873 Description: Not check open intent error before release open handle
2874 Details : in some rare cases, the open intent error is not checked before
2875 release open handle, which may cause
2876 ASSERTION(open_req->rq_transno != 0), because it tries to release
2877 the failed open handle.
2882 Description: Set cat log bitmap only after create log success.
2883 Details : in some rare cases, the cat log bitmap is set too early. and it
2884 should be set only after create log success.
2888 Description: Accessing a block bevice can re-enable I/O when Lustre is
2889 tearing down a device.
2890 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
2895 Description: service threads may hog cpus when there are a lot of requests
2896 Details : Insert cond_resched to give other threads a chance to use some CPU
2901 Description: the cat log was not initialized in recovery
2902 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
2903 unlink log on mds will not be initialized until mds post
2904 recovery. And also in mds post recovery, the unlink log will
2905 initialization will be done asynchronausly, so there will be race
2906 between add unlink log and unlink log initialization.
2910 Description: brw_stats were being printed incorrectly
2911 Details : brw_stats were being printed as log2 but all of them were not
2912 recorded as log2. Also remove some code duplication arising from
2913 filter_tally_{read,write}.
2917 Frequency : rare, only in recovery.
2918 Description: ASSERTION(req->rq_type != LI_POISON) failed
2919 Details : imp_lock should be held while iterating over imp_sending_list for
2920 prevent destroy request after get timeout in ptlrpc_queue_wait.
2924 Description: replay-single.sh test 52 fails
2925 Details : A lock's skiplist need to be cleanup when it being unlinked
2926 from its resource list.
2930 Description: Short directio read returns full requested size rather than
2932 Details : Direct I/O operations should return actual amount of bytes
2933 transferred rather than requested size.
2935 Severity : enhancement
2937 Description: metadata RPC reduction (e.g. for rm performance)
2938 Details : decrease the amount of synchronous RPC between clients and servers
2939 by canceling conflicing lock before the operation on the client
2940 and packing thier handles into the main operation RPC to server.
2942 Severity : enhancement
2944 Description: Async OSC create to avoid the blocking unnecessarily.
2945 Details : If a OST has no remain object, system will block on the creating
2946 when need to create a new object on this OST. Now, ways use
2947 pre-created objects when available, instead of blocking on an
2948 empty osc while others are not empty. If we must block, we block
2949 for the shortest possible period of time.
2953 Description: improve handling recoverable errors
2954 Details : if request processig with error which can be recoverable on server
2955 request should be resend, otherwise page released from cache and
2958 Severity : enhancement
2960 Description: refine locking for avoid write wrong info into lov_objid file
2961 Details : fix possible races with add new target and write/update data in
2964 --------------------------------------------------------------------------------
2966 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
2971 Frequency : on some architectures
2973 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
2974 Details : Uninitialized flags sometimes cause configuration commands to
2978 Frequency : patchless clients only
2980 Description: missing __iget() symbol export
2981 Details : The __iget() symbol export is missing. To avoid the need for
2982 this on patchless clients the deathrow inode reaper is turned
2983 off, and we depend on the VM to clean up old inodes. This
2984 dependency was during via the fix for bug 12181.
2986 --------------------------------------------------------------------------------
2988 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
2990 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
2991 INTEROPERATE with older versions automatically. In many cases a
2992 special upgrade step is needed. Please read the
2993 user documentation before upgrading any part of a live system.
2994 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
2995 version will not interoperate with 1.6 betas before beta5 (1.5.95).
2996 * WARNING: Lustre configuration and startup changes are required with
2997 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
3001 Severity : enhancement
3003 Description: MountConf
3004 Details : Lustre configuration is now managed via mkfs and mount
3005 commands instead of lmc and lconf. New obd types (MGS, MGC)
3006 are added for dynamic configuration management. See
3007 https://mail.clusterfs.com/wikis/lustre/MountConf for
3010 Severity : enhancement
3012 Description: dynamic OST addition
3013 Details : OSTs can now be added to a live filesystem
3015 Severity : enhancement
3017 Description: startup order invariance
3018 Details : MDTs and OSTs can be started in any order. Clients only
3019 require the MDT to complete startup.
3021 Severity : enhancement
3023 Description: parallel, asynchronous orphan cleanup
3024 Details : orphan cleanup is now performed in separate threads for each
3025 OST, allowing parallel non-blocking operation.
3027 Severity : enhancement
3029 Description: optimized stripe assignment
3030 Details : stripe assignments are now made based on ost space available,
3031 ost previous usage, and OSS previous usage, in order to try
3032 to optimize storage space and networking resources.
3034 Severity : enhancement
3036 Description: Permanently set tunables
3037 Details : All writable /proc/fs/lustre tunables can now be permanently
3038 set on a per-server basis, at mkfs time or on a live system.
3040 Severity : enhancement
3042 Description: Lustre message v2
3043 Details : Add lustre message format v2.
3045 Severity : enhancement
3047 Description: client OST exclusion list
3048 Details : Clients can be started with a list of OSTs that should be
3049 declared "inactive" for known non-responsive OSTs.
3053 Description: SPEC SFS validation failure on NFS v2 over lustre.
3054 Details : Changes the blocksize for regular files to be 2x RPC size,
3055 and not depend on stripe size.
3057 Severity : enhancement
3059 Description: Multiple MD RPCs in flight.
3060 Details : Further unserialise some read-only MDS RPCs - learn about intents.
3061 To avoid overly-overloading MDS, introduce a limit on number of
3062 MDS RPCs in flight for a single client and add /proc controls
3063 to adjust this limit.
3065 Severity : enhancement
3067 Description: client read/write statistics
3068 Details : Add client read/write call usage stats for performance
3069 analysis of user processes.
3070 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
3071 file access. extents_stats shows chunk size distribution.
3072 extents_stats_per_process show chunk size distribution per
3075 Severity : enhancement
3077 Description: mds statistics
3078 Details : Add detailed mds operations statistics in
3079 /proc/fs/lustre/mds/*/stats.
3083 Description: Failure of copying files with lustre special EAs.
3084 Details : Client side always return success for setxattr call for lustre
3085 special xattr (currently only "trusted.lov").
3090 Description: Refcount LNET uuids
3091 Details : The global LNET uuid list grew linearly with every startup;
3092 refcount repeated list entries instead of always adding to
3095 Severity : enhancement
3097 Description: Dynamic service threads
3098 Details : Within a small range, start extra service threads
3099 automatically when the request queue builds up.
3102 Frequency : mixed-endian client/server environments
3104 Description: mixed-endian crashes
3105 Details : The new msg_v2 system had some failures in mixed-endian
3108 Severity : enhancement
3110 Description: Easy OST removal
3111 Details : OSTs can be permanently deactivated with e.g. 'lctl
3112 conf_param lustre-OST0001.osc.active=0'
3114 Severity : enhancement
3116 Description: MGS proc entries
3117 Details : Added basic proc entries for the MGS showing what filesystems
3120 Severity : enhancement
3122 Description: provide MGS failover
3123 Details : Added config lock reacquisition after MGS server failover.
3125 Severity : enhancement
3127 Description: add Linux 2.4 support
3128 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
3132 Description: a large application tries to do I/O to the same resource and dies
3133 in the middle of it.
3134 Details : Check the req->rq_arrival time after the call to
3135 ost_brw_lock_get(), but before we do anything about
3136 processing it & sending the BULK transfer request. This
3137 should help move old stale pending locks off the queue as
3138 quickly as obd_timeout.
3141 Frequency : when an incorrect nid is specified during startup
3143 Description: ptlrpc connect to non-existant node causes kernel crash
3144 Details : LNET can't be re-entered from an event callback, which
3145 happened when we expire a message after the export has been
3146 cleaned up. Instead, hand the zombie cleanup off to another
3149 Severity : enhancement
3151 Description: plain/inodebits lock performance improvement
3152 Details : Grouping plain/inodebits in granted list by their request modes
3153 and bits policy, thus improving the performance of search through
3157 Frequency : only if OST filesystem is corrupted
3159 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3160 Details : for a short time RPCs with bulk IO are in the replay list,
3161 but replay of bulk IOs is unimplemented. If the OST filesystem
3162 is corrupted due to disk cache incoherency and then replay is
3163 started it is possible to trip an assertion. Avoid putting
3164 committed RPCs into the replay list at all to avoid this issue.
3167 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3169 Description: System hang on startup
3170 Details : This bug allowed the liblustre (e.g. catamount) client to
3171 return to the app before handling all startup RPCs. This
3172 could leave the node unresponsive to lustre network traffic
3173 and manifested as a server ptllnd timeout.
3175 Severity : enhancement
3177 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3178 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
3179 boolean to control whether to print peer debug info when a
3180 client's RPC times out.
3183 Frequency : only for kernels with patches from Lustre below 1.4.3
3185 Description: Remove old rdonly API
3186 Details : Remove old rdonly API which unused from at least lustre 1.4.3
3189 Frequency : only for devices with external journals
3191 Description: Set external device read-only also
3192 Details : During a commanded failover stop, we set the disk device
3193 read-only while the server shuts down. We now also set any
3194 external journal device read-only at the same time.
3197 Frequency : when upgrading from 1.4 while trying to change parameters
3199 Description: The wrong (new) MDC name was used when setting parameters for
3200 upgraded MDT's. Also allows changing of OSC (and MDC)
3201 parameters if --writeconf is specified at tunefs upgrade time.
3204 Frequency : when setting specific ost indicies
3206 Description: QOS code breaks on skipped indicies
3207 Details : Add checks for missing OST indicies in the QOS code, so OSTs
3208 created with --index need not be sequential.
3210 Severity : enhancement
3212 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3213 Details : The uninit_groups feature works in conjunction with the kernel
3214 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3215 up the pass1 processing of e2fsck. This is a read-only feature
3216 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3217 mount filesystems that have had this feature enabled.
3219 Severity : enhancement
3221 Description: Improve multi-block allocation algorithm to avoid fragmentation
3222 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3223 allocation locality and avoid filesystem fragmentation.
3225 ------------------------------------------------------------------------------
3227 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
3229 * Support for kernels:
3230 2.4.21-47.0.1.EL (RHEL 3)
3231 2.6.5-7.283 (SLES 9)
3232 2.6.9-42.0.10.EL (RHEL 4)
3233 2.6.12.6 vanilla (kernel.org)
3234 2.6.16.27-0.9 (SLES 10)
3235 * Recommended e2fsprogs version: 1.39.cfs5
3237 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3241 Frequency : occasional, depends on client load and configuration
3242 Bugzilla : 12181, 12203
3243 Description: data loss for recently-modified files
3245 Details : In some cases it is possible that recently written or created
3246 files may not be written to disk in a timely manner (this should
3247 normally be within 30s unless client IO load is very high).
3248 The problem appears as zero-length files or files that are a
3249 multiple of 1MB in size after a client crash or client eviction
3250 that are missing data at the end of the file.
3252 This problem is more likely to be hit on clients where files are
3253 repeatedly created and unlinked in the same directory, clients
3254 have a large amount of RAM, have many CPUs, the filesystem has
3255 many OSTs, the clients are rebooted frequently, and/or the files
3256 are not accessed by other nodes after being written.
3258 The presence of the problem can be detected by looking at
3259 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
3260 smaller than the second (nr_unused) then dirty files will not
3261 be flushed automatically to disk. "sync; sleep 10" should be
3262 run several times on the node before unmounting it to update
3263 Lustre (this is also safe to run on nodes without this problem).
3265 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
3266 that can cause this same problem, so customers using that kernel
3267 also need to update the kernel in addition to Lustre. In order
3268 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
3270 It is normal that files written just before a client crash (less
3271 than 30s) may not yet have been flushed to disk, even for local
3275 Frequency : frequent on thin XT3 nodes
3277 Description: UUID collision on thin XT3 Linux nodes
3278 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
3279 randomly, since we relied on an insufficiently-seeded PRNG.
3284 Description: OSS hangs after "All ost request buffers busy"
3285 Details : A deadlock between quota and journal operations caused OSS
3286 hangs after printing "All ost request buffers busy."
3289 Frequency : always on liblustre builds
3291 Description: Cleanup compiler warnings on liblustre
3294 Frequency : always on liblustre builds on XT3
3296 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
3300 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
3301 > 32000 subdirectories
3302 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
3303 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
3304 subdirectory count crosses 32000. This will aid e2fsck to
3305 correctly handle more than 32000 subdirectories.
3308 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3310 Description: System hang on startup
3311 Details : This bug allowed the liblustre (e.g. catamount) client to
3312 return to the app before handling all startup RPCs. This
3313 could leave the node unresponsive to lustre network traffic
3314 and manifested as a server ptllnd timeout.
3316 Severity : enhancement
3318 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3319 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
3320 boolean to control whether to print peer debug info when a
3321 client's RPC times out.
3326 Description: make O_SYNC working on 2.6 kernels
3327 Details : 2.6 kernels use different method for mark pages for write,
3328 so need add a code to lustre for O_SYNC work.
3333 Description: Failure to close file and release space on NFS
3334 Details : Put inode details into lock acquired in ll_intent_file_open.
3335 Use mdc_intent_lock in ll_intent_open to properly
3336 detect all kind of errors unhandled by mdc_enqueue.
3341 Description: proc file read during shutdown sometimes raced obd removal,
3343 Details : Add lock to prevent obd access after proc file removal.
3346 Frequency : Only for files larger than 4GB on 32-bit clients.
3348 Description: improperly doing page alignment of locks
3349 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
3350 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3355 Description: RPCs being resent when they shouldn't be
3356 Details : Some RPCs that should not be resent are being resent. This
3357 can cause inconsistencies in the RPC state machine. Do not
3358 resend such requests.
3361 Frequency : rare, only with NFS export
3363 Description: Crash on NFS re-export node
3364 Details : under very unusual load conditions an assertion is hit in
3365 ll_intent_file_open()
3368 Frequency : only if OST filesystem is corrupted
3370 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3371 Details : for a short time RPCs with bulk IO are in the replay list,
3372 but replay of bulk IOs is unimplemented. If the OST filesystem
3373 is corrupted due to disk cache incoherency and then replay is
3374 started it is possible to trip an assertion. Avoid putting
3375 committed RPCs into the replay list at all to avoid this issue.
3380 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3381 Details : Large single O_DIRECT read and write calls can fail to allocate
3382 a sufficiently large buffer to process the request. In case of
3383 allocation failure the allocation is retried with a smaller
3384 buffer and broken into smaller requests.
3386 Severity : enhancement
3388 Description: Add -o localflock option to simulate old noflock behaviour.
3389 Details : This will achieve local-only flock/fcntl locks coherentness.
3394 Description: versioning check is incomplete
3395 Details : Checking the version difference of client vs. server, report
3396 error if the gap is too big.
3401 Description: add support PG_writeback bit
3402 Details : add support for PG_writeback bit for Lustre, for more carefull
3403 work with page cache in 2.6 kernel. This also fix some deadlocks
3404 and remove hack for work O_SYNC with 2.6 kernel.
3406 Severity : enhancement
3408 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3409 Details : The uninit_groups feature works in conjunction with the kernel
3410 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3411 up the pass1 processing of e2fsck. This is a read-only feature
3412 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3413 mount filesystems that have had this feature enabled.
3415 Severity : enhancement
3417 Description: Improve multi-block allocation algorithm to avoid fragmentation
3418 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3419 allocation locality and avoid filesystem fragmentation.
3421 ------------------------------------------------------------------------------
3423 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3425 * Support for kernels:
3426 2.6.9-42.0.3.EL (RHEL 4)
3427 2.6.5-7.276 (SLES 9)
3428 2.4.21-47.0.1.EL (RHEL 3)
3429 2.6.12.6 vanilla (kernel.org)
3430 2.6.16.21-0.8 (SLES10)
3431 * Recommended e2fsprogs version: 1.39.cfs2-0
3433 * The backwards-compatible /proc/sys/portals symlink has been removed
3434 in this release. Before upgrading, please ensure that you change
3435 any configuration scripts or /etc/sysctl.conf files that access
3436 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3437 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3438 in advance of the upgrade on any system running Lustre 1.4.6 or
3439 newer, since /proc/sys/lnet was added in that version.
3440 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3444 Frequency : only when quota is used
3446 Description: avoid scanning export list for quota master
3447 Details : Change the algorithms to avoid scanning export list in order
3448 to improve the efficiency.
3451 Frequency : MDS failover only, very rarely
3453 Description: "went back in time" messages on mds failover
3454 Details : The greatest transno may be lost when the current operation
3455 finishes with an error (transno==0) and the client's last_rcvd
3456 record is over-written. Save the greatest transno in the
3457 mds_last_transno for this case.
3460 Frequency : always for specific kernels and striping counts
3462 Description: client may get "Matching packet too big" without ACL support
3463 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
3464 when trying to access files in certain configurations. The
3465 clients should in fact be denied when mounting because they do
3466 not understand ACLs.
3469 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
3471 Description: many clients connecting with IO in progress causes connect timeouts
3472 Details : Avoid synchronous journal commits to avoid delays caused by many
3473 clients connecting/disconnecting when bulk IO is in progress.
3474 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
3475 OST_IO_PORTAL to avoid delays behind potentially many pending
3479 Frequency : occasionally with multiple writers to a single file
3481 Description: shared writes to file may result in wrong size reported by stat()
3482 Details : Allow growing of kms when extent lock is cancelled
3485 Frequency : always with random mmap IO to multi-striped file
3487 Description: mmap write might be lost if we are writing to a 'hole' in stripe
3488 Details : Only if the hole is at the end of OST object so that kms is too
3489 small. Fix is to increase kms accordingly in ll_nopage.
3492 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
3494 Description: writes to a missing object would leak memory on the OST
3495 Details : If there is an inconsistency between the MDS and OST filesystems,
3496 such that the MDS references an object that doesn't exist, writes
3497 to that object will leak memory due to incorrect cleanup in the
3498 error handling path, eventually running out of memory on the OST.
3503 Description: Creating too long symlink causes lustre errors
3504 Details : Check symlink and name lengths before sending requests to MDS.
3507 Frequency : only if flock is enabled (not on by default)
3509 Description: posix locks not released on fd closure on 2.6.9+
3510 Details : We failed to add posix locks to list of inode locks on 2.6.9+
3511 kernels, this caused such locks not to be released on fd close and
3512 then assertions on fs unmount about still used locks.
3515 Frequency : MDS failover only, very rarely
3517 Description: clients may get ASSERTION(granted_lock != NULL)
3518 Details : When request was taking a long time, and a client was resending
3519 a getattr by name lock request. The were multiple lock requests
3520 with the same client lock handle and
3521 mds_getattr_name->fixup_handle_for_resent_request found one of the
3522 lock handles but later failed with ASSERTION(granted_lock != NULL).
3527 Description: handle->h_buffer_credits > 0, assertion failure
3528 Details : h_buffer_credits is zero after truncate, causing assertion
3529 failure. This patch extends the transaction or creates a new
3533 Frequency : NFS re-export or patchless client
3534 Bugzilla : 11179, 10796
3535 Description: Crash on NFS re-export node (__d_move)
3536 Details : We do not want to hash the dentry if we don't have a lock.
3537 But if this dentry is later used in d_move, we'd hit uninitialised
3538 list head d_hash, so we just do this to init d_hash field but
3539 leave dentry unhashed.
3542 Frequency : NFS re-export or patchless client
3544 Description: NFS exports has problem with symbolic link
3545 Details : lustre client didn't properly install dentry when re-exported
3546 to NFS or running patchless client.
3549 Frequency : NFS re-export or patchless client
3551 Description: Various nfs/patchless fixes.
3552 Details : fixes reuse disconected alias for lookup process - this fixes
3553 warning "find_exported_dentry: npd != pd",
3554 fix permission error with open files at nfs.
3555 fix apply umask when do revalidate.
3558 Frequency : occasional
3560 Description: Crash on NFS re-export node
3561 Details : calling clear_page() on the wrong pointer triggered oops in
3562 generic_mapping_read().
3565 Frequency : rarely, using O_DIRECT IO
3567 Description: unaligned directio crashes client with LASSERT
3568 Details : check for unaligned buffers before trying any requests.
3571 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
3573 Description: stale data returned from RAID cache
3574 Details : If only a small amount of IO is done to the RAID device before
3575 reading it again it is possible to get stale data from the RAID
3576 cache instead of reading it from disk.
3579 Frequency : always for sles10 kernel
3581 Description: sles10 support
3582 Details : ll_follow_link: compile fixes and using of nd_set_link
3583 under newer kernels.
3586 Frequency : depends on arch, kernel and compiler version, always on sles10
3589 Description: recursive or deep enough symlinks cause stack overflow
3590 Details : getting rid of large stack-allocated variable in
3594 Frequency : depends on hardware
3596 Description: lustre write performance loss in the SLES10 kernel
3597 Details : the performance loss is caused by using of write barriers in the
3598 ext3 code. The SLES10 kernel turns barrier support on by
3599 default. The fix is to undo that change for ldiskfs.
3601 ------------------------------------------------------------------------------
3603 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
3605 * Support for kernels:
3606 2.6.9-42.0.3EL (RHEL 4)
3607 2.6.5-7.276 (SLES 9)
3608 2.4.21-47.0.1.EL (RHEL 3)
3609 2.6.12.6 vanilla (kernel.org)
3613 Frequency : quota enabled and large files being deleted
3615 Description: releasing more than 4GB of quota at once hangs OST
3616 Details : If a user deletes more than 4GB of files on a single OST it
3617 will cause the OST to spin in an infinite loop. Release
3618 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3623 Description: statfs data retrieved from /proc may be stale or zero
3624 Details : When reading per-device statfs data from /proc, in the
3625 {kbytes,files}_{total,free,avail} files, it may appear
3626 as zero or be out of date.
3629 Frequency : systems with MD RAID1 external journal devices
3631 Description: lconf's call to blkid is confused by RAID1 journal devices
3632 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3633 of returning all block devices that match the journal UUID.
3636 Frequency : always, for aggregate stripe size over 4GB
3638 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
3639 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
3640 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
3643 Frequency : always if "lfs find" used on a local file/directory
3645 Description: "lfs find" segfaults if used on a local file/directory
3646 Details : The case where a directory component was not specified wasn't
3647 handled correctly. Handle this properly.
3650 Frequency : always on ppc64
3652 Description: the write to an ext3 filesystem mounted with mballoc got stuck
3653 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3654 perform endianness conversion.
3657 Frequency : rarely (truncate to non-zero file size after write under load)
3658 Bugzilla : 10730, 10687
3659 Description: Files padded with zeros to next 4K multiple
3660 Details : With filesystems mounted using the "extents" option (2.6 kernels)
3661 it is possible that files that are truncated to a non-zero size
3662 immediately after being written are filled with zero bytes beyond
3663 the truncated size. No file data is lost.
3665 Severity : enhancement
3667 Description: Allow recovery/failover for liblustre clients.
3668 Details : liblustre clients were unaware of failover configurations until
3671 Severity : enhancement
3673 Description: user file locks should fail when not mounting with flock option
3674 Details : Set up an error-returning stub in ll_file_operations.lock field
3675 to prevent incorrect behaviour when client is mounted without
3676 flock option. Also, set up properly f_op->flock field for
3680 Frequency : always on ia64
3682 Description: "lfs df" loops on printing out MDS statfs information
3683 Details : The obd_ioctl_data was not initialized and in some systems
3684 this caused a failure during the ioctl that did not return
3685 an error. Initialize the struct and return an error on failure.
3688 Frequency : SLES 9 only
3690 Description: Error of copying files with lustre special EAs as root
3691 Details : Client side always return success for setxattr call for lustre
3692 special xattr (currently only "trusted.lov").
3695 Frequency : rarely on clusters with both ia64+i386 clients
3697 Description: ia64+i686 clients doing shared IO on the same file may LBUG
3698 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
3699 clients are doing concurrent writes to the same file it is
3700 possible that the ia64 clients may LASSERT because the OST
3701 extent locks are not PAGE_SIZE aligned. Ensure that grown
3702 locks are always aligned on the request boundary.
3705 Frequency : specific use, occasional
3707 Description: Overwriting in use executable truncates on-disk binary image
3708 Details : If one node attempts to overwrite an executable in use by
3709 another node, we now correctly return ETXTBSY instead of
3710 truncating the file.
3712 Severity : enhancement
3714 Description: Async OSC create to avoid the blocking unnecessarily.
3715 Details : If a OST has no remain object, system will block on the creating
3716 when need to create a new object on this OST. Now, ways use
3717 pre-created objects when available, instead of blocking on an
3718 empty osc while others are not empty. If we must block, we block
3719 for the shortest possible period of time.
3724 Description: chmod on Lustre root is propagated to other clients
3725 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
3726 invalid by the follow_mount time.
3729 Frequency : liblustre clients only
3731 Description: Race in 'instant cancel' lock handling could lead to such locks
3732 never to be granted in case of SMP MDS
3733 Details : Do not destroy not yet granted but cbpending locks in
3737 Frequency : replay/resend of open
3739 Description: non null lock assetion failure in mds_intent_policy
3740 Details : Trying to replay/resend lockless open requests resulted in
3741 mds_open() returning 0 with no lock. Now it sets a flag if
3742 a lock is going to be returned.
3744 Severity : enhancement
3746 Description: Checksum enhancements
3747 Details : New checksum enhancements allow for resending RPCs that failed
3750 Severity : enhancement
3752 Description: Tunables on number of dirty pages in cacche
3753 Details : Allow to set limit on number of dirty pages cached.
3758 Description: client crash on unmount - lock still has references
3759 Details : In some error handling cases it was possible to leak a lock
3760 reference on a client while accessing a file. This was not
3761 harmful to the client during operation, but would cause the
3762 client to crash when the filesystem is unmounted.
3765 Frequency : specific case, rare
3767 Description: ETXTBSY on mds though file not in use by client
3768 Details : ETXTBSY is no longer incorrectly returned when attempting to
3769 chmod or chown a directory that the user previously tried to
3770 execute or a currently-executing binary.
3773 Frequency : extremely rare except on liblustre-based clients
3775 Description: Lustre space not freed when files are deleted
3776 Details : Clean up open-unlinked files after client eviction. Previously
3777 the unlink was skipped and the files remained as orphans.
3782 Description: OST failure "would be an LBUG" in waiting_locks_callback()
3783 Details : In some cases it was possible to send a blocking callback to
3784 a client doing a glimpse, even though that client didn't get
3785 a lock granted. When the glimpse lock is cancelled on the OST
3786 the freed lock is left on the waiting list and corrupted the list.
3789 Frequency : all core dumps
3791 Description: Broke core dumps to lustre
3792 Details : Negative dentry may be unhashed if parent does not have UPDATE
3793 lock, but some callers, e.g. do_coredump, expect dentry to be
3794 hashed after successful create, hash it in ll_create_it.
3796 ------------------------------------------------------------------------------
3798 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
3800 * Support for kernels:
3801 2.6.9-42.0.2.EL (RHEL 4)
3802 2.6.5-7.276 (SLES 9)
3803 2.4.21-40.EL (RHEL 3)
3804 2.6.12.6 vanilla (kernel.org)
3808 Frequency : always on RHEL 3
3810 Description: Number of open files grows over time
3811 Details : The number of open files grows over time, whether or not
3812 Lustre is started. This was due to a filp leak introduced
3813 by one of our kernel patches.
3815 ------------------------------------------------------------------------------
3817 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
3819 * Support for kernels:
3820 2.6.9-42.EL (RHEL 4)
3821 2.6.5-7.276 (SLES 9)
3822 2.4.21-40.EL (RHEL 3)
3823 2.6.12.6 vanilla (kernel.org)
3828 Bugzilla : 5719, 9635, 9792, 9684
3829 Description: OST (or MDS) trips assertions in (re)connection under heavy load
3830 Details : If a server is under heavy load and cannot reply to new
3831 connection requests before the client resends the (re)connect,
3832 the connection handling code can behave badly if two service
3833 threads are concurrently handing separate (re)connections from
3834 the same client. Add better locking to the connection handling
3835 code, and ensure that only a single connection will be processed
3836 for a given client UUID, even if the lock is dropped.
3838 Severity : enhancement
3840 Description: add TCP zero-copy support to kernel
3841 Details : Add support to the kernel TCP stack to allow zero-copy bulk
3842 sends if the hardware supports scatter-gather and checksumming.
3843 This allows socklnd to do client-write and server-read more
3844 efficiently and reduce CPU utilization from skbuf copying.
3847 Frequency : only if NFS exporting from client
3849 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
3850 Details : If setting attributes on a file created under NFS that had
3851 never been opened it would be possible to oops the client
3852 if the file had no objects.
3855 Frequency : always for liblustre
3857 Description: liblustre client does MDS+OSTs setattr RPC for each write
3858 Details : When doing a write from a liblustre client, the client
3859 incorrectly issued an RPC to the MDS and each OST the file was
3860 striped over in order to update the timestamps. When writing
3861 with small chunks and many clients this could overwhelm the MDS
3862 with RPCs. In all cases it would slow down the write because
3863 these RPCs are unnecessary.
3865 Severity : enhancement
3867 Description: allow number of MDS service threads to be changed at module load
3868 Details : It is now possible to change the number of MDS service threads
3869 running. Adding "options mds mds_num_threads={N}" to the MDS's
3870 /etc/modprobe.conf will set the number of threads for the next
3871 time Lustre is restarted (assuming the "mds" module is also
3872 reloaded at that time). The default number of threads will
3873 stay the same, 32 for most systems.
3878 Description: OST crash if filesystem is unformatted or corrupt
3879 Details : If an OST is started on a device that has never been formatted
3880 or if the filesystem is corrupt and cannot even mount then the
3881 error handling cleanup routines would dereference a NULL pointer.
3886 Description: NULL pointer deref in llap_from_page.
3887 Details : get_cache_page_nowait can return a page with NULL (or otherwise
3888 incorrect) mapping if the page was truncated/reclaimed while it was
3889 searched for. Check for this condition and skip such pages when
3890 doing readahead. Introduce extra check to llap_from_page() to
3891 verify page->mapping->host is non-NULL (so page is not anonymous).
3894 Frequency : Sometimes when using sys_sendfile
3896 Description: "page not covered by a lock" warnings from ll_readpage
3897 Details : sendfile called ll_readpage without right page locks present.
3898 Now we introduced ll_file_sendfile that does necessary locking
3899 around call to generic_file_sendfile() much like we do in
3903 Frequency : with certain MDS communication failures at client mount time
3905 Description: NULL pointer deref after failed client mount
3906 Details : a client connection request may delayed by the network layer
3907 and not be sent until after the PTLRPC layer has timed out the
3908 request. If the client fails the mount immediately it will try
3909 to clean up before the network times out the request. Add a
3910 reference from the request import to the obd device and delay
3911 the cleanup until the network drops the request.
3914 Frequency : occasionally during client (re)connect
3916 Description: assertion failure during client (re)connect
3917 Details : processing a client connection request may be delayed by the
3918 client or server longer than the client connect timeout. This
3919 causes the client to resend the connection request. If the
3920 original connection request is replied in this interval, the
3921 client may trip an assertion failure in ptlrpc_connect_interpret()
3922 which thought it would be the only running connect process.
3925 Frequency : only with obd_echo servers and clients that are rebooted
3927 Description: kernel BUG accessing uninitialized data structure
3928 Details : When running an obd_echo server it did not start the ping_evictor
3929 thread, and when a client was evicted an uninitialized data
3930 structure was accessed. Start the ping_evictor in the RPC
3931 service startup instead of the OBD startup.
3933 Severity : enhancement
3934 Bugzilla : 10193 (patchless)
3935 Description: Remove dependency on various unexported kernel interfaces.
3936 Details : No longer need reparent_to_init, exit_mm, exit_files,
3937 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
3940 Frequency : rare (only users of deprecated and unsupported LDAP config)
3942 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
3943 Details : LDAP apparently contains 'lustreName' attributes instead of
3944 'name'. A simple remapping of the name is sufficient.
3947 Frequency : rare (only with non-default dump_on_timeout debug enabled)
3949 Description: waiting_locks_callback trips kernel BUG if client is evicted
3950 Details : Running with the dump_on_timeout debug flag turned on makes
3951 it possible that the waiting_locks_callback() can try to dump
3952 the Lustre kernel debug logs from an interrupt handler. Defer
3953 this log dumping to the expired_lock_main() thread.
3955 Severity : enhancement
3957 Description: Support NFS exporting on 2.6 kernels.
3958 Details : Implement non-rawops metadata methods for NFS server to use without
3959 changing NFS server code.
3962 Frequency : very rare (synthetic metadata workload only)
3964 Description: two racing renames might cause an MDS thread to deadlock
3965 Details : Running the "racer" program may cause one MDS thread to rename
3966 a file from being the source of a rename to being the target of
3967 a rename at exactly the same time that another thread is doing
3968 so, and the second thread has already enqueued these locks after
3969 doing a lookup of the target and is trying to relock them in
3970 order. Ensure that we don't try to re-lock the same resource.
3973 Frequency : only very large systems with liblustre clients
3975 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
3976 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
3977 all OSTs in parallel. This allows the eviction of stale liblustre
3978 clients to proceed much faster than if they were done in series,
3979 and also offers similar improvements for other set_info RPCs.
3984 Description: excessive CPU usage during initial read phase on client
3985 Details : During the initial read phase on a client, it would agressively
3986 retry readahead on the file, consuming too much CPU and impacting
3987 performance (since 1.4.5.8). Improve the readahead algorithm
3988 to avoid this, and also improve some other common cases (read
3989 of small files in particular, where "small" is files smaller than
3990 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
3995 Description: MDS crash when receiving packet with unknown intent.
3996 Details : Do not LBUG in unknown intent case, just return -EFAULT
3998 Severity : enhancement
3999 Bugzilla : 9293, 9385
4000 Description: MDS RPCs are serialised on client. This is unnecessary for some.
4001 Details : Do not serialize getattr (non-intent version) and statfs.
4004 Frequency : occasional, when OST network is overloaded/intermittent
4006 Description: client evicted by OST after bulk IO timeout
4007 Details : If a client sends a bulk IO request (read or write) the OST
4008 may evict the client if it is unresposive to its data GET/PUT
4009 request. This is incorrect if the network is overloaded (takes
4010 too long to transfer the RPC data) or dropped the OST GET/PUT
4011 request. There is no need to evict the client at all, since
4012 the pinger and/or lock callbacks will handle this, and the
4013 client can restart the bulk request.
4016 Frequency : Always when mmapping file with no objects
4018 Description: client crashes when mmapping file with no objects
4019 Details : Check that we actually have objects in a file before doing any
4020 operations on objects in ll_vm_open, ll_vm_close and
4026 Description: Request leak when working with deleted CWD
4027 Details : Introduce advanced request refcount tracking for requests
4028 referenced from lustre intent.
4030 Severity : Enhancement
4032 Description: Cache open file handles on client.
4033 Details : MDS now will return special lock along with openhandle, if
4034 requested and client is allowed to hold openhandle, even if unused,
4035 until such a lock is revoked. Helps NFS a lot, since NFS is opening
4036 closing files for every read/write openration.
4038 Severity : Enhancement
4040 Description: Cache open negative dentries on client when possible.
4041 Details : Guard negative dentries with UPDATE lock on parent dir, drop
4042 negative dentries on lock revocation.
4047 Description: Remounting a client read-only wasn't possible with a zconf mount
4048 Details : It wasn't possible to remount a client read-only with llmount.
4050 Severity : enhancement
4051 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
4052 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
4053 MPICH is included to add Lustre-specific ADIO interfaces.
4054 This is based closely on the UFS ADIO layer and only differs
4055 in file creation, in order to allow the OST striping to be set.
4056 This is user-contributed code and not supported by CFS.
4061 Description: extended inode attributes (immutable, append-only) work improperly
4062 when 2.4 and 2.6 kernels are used on client/server or vice versa
4063 Details : Introduce kernel-independent values for these flags.
4065 Severity : enhancement
4068 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
4069 Details : Many of the /proc/ tunables can only be tuned at a megabyte
4070 granularity. Now, Fractional MB granularity is be supported,
4071 this is very useful for low memory system.
4073 Severity : enhancement
4075 Description: Getattr by fid
4076 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
4077 locks, avoids extra getattr rpc requests to MDS, allows '/' to
4078 have locks and avoids getattr rpc requests for it on every stat.
4081 Frequency : Always, for filesystems larger than 2TB
4083 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4084 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
4085 filesystems larger than 2TB to be mounted with mballoc enabled.
4088 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
4090 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4091 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
4092 filesystems larger than 2TB is mounted then the kernel will
4093 silently corrupt the start of the filesystem. CONFIG_LBD is
4094 enabled for all CFS-supported kernels, but the possibility of
4095 this happening with a modified kernel config exists.
4097 Severity : enhancement
4099 Description: add client O_DIRECT support for 2.6 kernels
4100 Details : It is now possible to do O_DIRECT reads and writes to files
4101 in the Lustre client mountpoint on 2.6 kernel clients.
4103 Severity : enhancement
4105 Description: parallel glimpse, setattr, statfs, punch, destroy requests
4106 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
4107 parallel, not waiting for response from every OST before sending
4108 a rpc to the next OST.
4113 Description: setattr vs write race when updating file timestamps
4114 Details : Client processes that update a file timestamp into the past
4115 right after writing to the file (e.g. tar) it is possible that
4116 the updated file modification time can be reset to the current
4117 time due to a race between processing the setattr and write RPC.
4119 Severity : enhancement
4121 Description: Bring 'lfs find' closer in line with regular Linux find.
4122 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
4123 -print0 options and obtains all the needed info through the lustre
4126 Severity : enhancement
4128 Description: support up to 1024 configured devices on one node
4129 Details : change obd_dev array from statically allocated to dynamically
4130 allocated structs as they are first used to reduce memory usage
4135 Description: Flush dirty partially truncated pages during truncate
4136 Details : Immediatelly flush partially truncated pages in filter_setattr,
4137 this way we completely avoid having any pages in page cache on OST
4138 and can retire ugly workarounds during writes to flush such pages.
4143 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
4144 Details : protect inode from truncation within vfs_unlink() context
4145 just take a reference before calling vfs_unlink() and release it
4146 when parent's i_sem is free.
4151 Description: last_id value checked outside lock on OST caused LASSERT failure
4152 Details : If there were multiple MDS->OST object precreate requests in
4153 flight, it was possible that the OST's last object id was checked
4154 outside a lock and incorrectly tripped an assertion. Move checks
4155 inside locks, and discard old precreate requests.
4158 Frequency : always, if extents are used on OSTs
4160 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
4161 Details : OSTs using the extents format would not zero the high 16 bits of
4162 the index physical block number. This is not a problem for any
4163 OST filesystems smaller than 16TB, and no kernels support ext3
4164 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
4165 new/modified files) and can be fixed for existing filesystems
4166 with e2fsprogs-1.39-cfs1.
4171 Description: import connection selection may be incorrect if timer wraps
4172 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
4173 import connections to be ignored if the 32-bit jiffies counter
4174 wraps. Use a 64-bit jiffies counter.
4177 Frequency : very large clusters immediately after boot
4179 Description: LNET request buffers exhausted under heavy short-term load
4180 Details : If a large number of client requests are generated on a service
4181 that has previously never seen so many requests it is possible
4182 that the request buffer growth cannot keep up with the spike in
4183 demand. Instead of dropping incoming requests, they are held in
4184 the LND until the RPC service can accept more requests.
4187 Frequency : Sometimes during replay
4189 Description: Assertion failure in ll_local_open after replay.
4190 Details : If replay happened on an open request reply before we were able
4191 to set replay handler, reply will become not swabbed tripping the
4192 assertion in ll_local_open. Now we set the handler right after
4193 recognising of open request
4196 Frequency : very rare
4198 Description: kernel reports "badness in vsnprintf"
4199 Details : Reading from the "recovery_status" /proc file in small chunks
4200 may cause a negative length in lprocfs_obd_rd_recovery_status()
4201 call to vsnprintf() (which is otherwise harmless). Exit early
4202 if there is no more space in the output buffer.
4204 Severity : enhancement
4206 Description: clear OBD RPC statistics by writing to them
4207 Details : It is now possible to clear the OBD RPC statistics by writing
4208 to the "stats" file.
4213 Description: Client mtime is not the same on different clients after utimes
4214 Details : In some cases, the client was using the utimes() syscall on
4215 a file cached on another node. The clients now validate the
4216 ctime from the MDS + OSTs to determine which one is right.
4221 Description: Inability to activate failout mode
4222 Details : lconf script incorrectly assumed that in pythong string's numeric
4223 value is used in comparisons.
4226 Frequency : always with multiple stripes per file
4228 Description: Inefficient object allocation for mutli-stripe files
4229 Details : When selecting which OSTs to stripe files over, for files with
4230 a stripe count that divides evenly into the number of OSTs,
4231 the MDS is always picking the same starting OST for each file.
4232 Return the OST selection heuristic to the original design.
4237 Description: mount failures may take full timeout to return an error
4238 Details : Under some heavy load conditions it is possible that a
4239 failed mount can wait for the full obd_timeout interval,
4240 possibly several minutes, before reporting an error.
4241 Instead return an error as soon as the status is known.
4243 Frequency : quota enabled and large files being deleted
4245 Description: releasing more than 4GB of quota at once hangs OST
4246 Details : If a user deletes more than 4GB of files on a single OST it
4247 will cause the OST to spin in an infinite loop. Release
4248 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4253 Description: statfs data retrieved from /proc may be stale or zero
4254 Details : When reading per-device statfs data from /proc, in the
4255 {kbytes,files}_{total,free,avail} files, it may appear
4256 as zero or be out of date.
4259 Frequency : systems with MD RAID1 external journal devices
4261 Description: lconf's call to blkid is confused by RAID1 journal devices
4262 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4263 of returning all block devices that match the journal UUID.
4266 Frequency : always, for aggregate stripe size over 4GB
4268 Description: assertion fails when trying to use 4GB stripe size
4269 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
4270 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
4273 Frequency : always on ppc64
4275 Description: the first write on an ext3 filesystem with mballoc got stuck
4276 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4277 perform endianness conversion.
4279 ------------------------------------------------------------------------------
4281 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4283 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
4284 INTEROPERATE with older versions automatically. Please read the
4285 user documentation before upgrading any part of a live system.
4286 * WARNING: Lustre networking configuration changes are required with
4287 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
4290 * Support for kernels:
4291 2.6.9-22.0.2.EL (RHEL 4)
4292 2.6.5-7.244 (SLES 9)
4293 2.6.12.6 vanilla (kernel.org)
4296 Severity : enhancement
4297 Bugzilla : 7981/8208
4298 Description: Introduced Lustre Networking (LNET)
4299 Details : LNET is new networking infrastructure for Lustre, it includes
4300 a reorganized network configuration mode (see the user
4301 documentation for full details) as well as support for routing
4302 between different network fabrics. Lustre Networking Devices
4303 (LNDs) for the supported network fabrics have also been
4304 created for this new infrastructure.
4306 Severity : enhancement
4307 Description: Introduced Access control lists
4308 Details : clients can set ACLs on files and directories in order to have
4309 more fine-grained permissions than the standard Unix UGO+RWX.
4310 The MDS must be started with the "-o acl" mount option.
4312 Severity : enhancement
4313 Description: Introduced filesystem quotas
4314 Details : Administrators may now establish per-user quotas on the
4317 Severity : enhancement
4319 Description: Configuration change for the XT3
4320 The PTLLND is now used to run Lustre over Portals on the XT3
4321 The configure option(s) --with-cray-portals are no longer used.
4322 Rather --with-portals=<path-to-portals-includes> is used to
4323 enable building on the XT3. In addition to enable XT3 specific
4324 features the option --enable-cray-xt3 must be used.
4329 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
4330 Details : A race between allocating a new llcd and re-getting the llcd_lock
4331 allowed another thread to grab newly-allocated llcd.
4333 Severity : enhancement
4335 Description: 2.6 OST async journal commit and locking fix to improve performance
4336 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
4337 2.6 kernels are now async as they already were in 2.4 kernels so
4338 that they can commit concurrently with the network bulk transfer.
4339 For block-allocated files the filter allocation semaphore is held
4340 to avoid filesystem fragmentation during allocation. BKL lock
4341 removed for 2.6 xattr operations where it is no longer needed.
4346 Description: lconf incorrectly determined whether two IP networks could talk
4347 Details : In some more complicated routing and multiple-network
4348 configurations, lconf will avoid trying to make a network
4349 connection to a disjoint part of the IP space. It was doing the
4350 math incorrectly for one set of cases.
4355 Description: Fix for potential infinite loop processing records in an llog.
4356 Details : If an llog record is corrupted/zeroed, it is possible to loop
4357 forever in llog_process(). Validate the llog record length
4358 and skip the remainder of the block on error.
4361 Frequency : occasional (liblustre only)
4363 Description: liblustre could not open files whose last component is a symlink
4364 Details : sysio_path_walk() would incorrectly pass the open intent to
4365 intermediate path components.
4368 Frequency : rare (liblustre only with non-standard tuning)
4369 Bugzilla : 7201 (7350)
4370 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4371 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4372 after it has been released, only do this for referenced locks
4374 Severity : enhancement
4376 Description: specifying an (invalid) directory default stripe_size of -1
4377 would reset the directory default striping
4378 Details : stripe_size -1 was used internally to signal directory stripe
4379 removal, now use "all default" to signal dir stripe removal
4380 as a directory striping of "all default" is not useful
4383 Frequency : common for large clusters running liblustre clients
4385 Description: doing an ls when liblustre clients are running is slow
4386 Details : sending a glimpse AST to a liblustre client waits for every AST
4387 to time out, as liblustre clients will not respond. Since they
4388 cannot cache data we refresh the OST lock LVB from disk instead.
4390 Severity : enhancement
4392 Description: doing an ls at the same time as file IO can be slow
4393 Details : enqueue and other "small" requests can be blocked behind many
4394 large IO requests. Create a new OST IO portal for non-IO
4395 requests so they can be processed faster.
4398 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4400 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4401 created file with mode 000
4402 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4403 allow an ftruncate() as a non-root user to a file with mode 000.
4404 Lustre now respects this flag to disable mode checks when
4405 truncating a file owned by the user
4408 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4411 Description: Revoking locks from clients that went dead or catatonic might take
4413 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4414 cancellation of such locks instant on servers without waiting for
4415 any reply from clients. Clients drops these locks when cancel
4416 notification from server is received without replying.
4419 Frequency : liblustre-only, when liblustre client dies or becomes busy
4421 Description: Doing ls on Linux clients can take a long time with active
4423 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4424 granting such locks to it in the first place if possible. Locks
4425 are taken by proxy on the OST during the read or write and
4426 dropped immediately afterward. Add connect flags handling, do
4427 not grant locks to liblustre clients for glimpse ASTs.
4429 Severity : enhancement
4431 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4432 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4433 sequential reads and behaves badly with applications doing random
4434 reads. Improve it by reading ahead at least read region, and
4435 avoiding excessive large RPC for small reads.
4437 Severity : enhancement
4439 Description: Creating more than 1000 files for a single job may cause a load
4440 imbalance on the OSTs if there are also a large number of OSTs.
4441 Details : qos_prep_create() uses an OST index reseed value that is an
4442 even multiple of the number of available OSTs so that if the
4443 reseed happens in the middle of the object allocation it will
4444 still utilize the OSTs as uniformly as possible.
4449 Description: OST or MDS may oops in ping_evictor_main()
4450 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
4451 but doesn't restart at beginning of obd_exports_timed list
4454 Severity : enhancement
4456 Description: improve by-nid export eviction on the MDS and OST
4457 Details : allow multiple exports with the same NID to be evicted at one
4458 time without re-searching the exports list.
4461 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
4463 Description: MDS may oops in groups_free()
4464 Details : in rare race conditions a newly allocated group_info struct is
4465 freed again, and this can be NULL. The 2.4 compatibility code
4466 for groups_free() checked for a NULL pointer, but 2.6 did not.
4469 Frequency : common for liblustre clients doing little filesystem IO
4470 Bugzilla : 9352, 7313
4471 Description: server may evict liblustre clients accessing contended locks
4472 Details : if a client is granted a lock or receives a completion AST
4473 with a blocking AST already set it would not reply to the AST
4474 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
4477 Frequency : lfs setstripe, only systems with more than 160 OSTs
4479 Description: unable to set striping with a starting offset beyond OST 160
4480 Details : llapi_create_file() incorrectly limited the starting stripe
4481 index to the maximum single-file stripe count.
4484 Frequency : LDAP users only
4486 Description: lconf did not handle in-kernel recovery with LDAP properly
4487 Details : lconf/LustreDB get_refs() is searching the wrong namespace
4489 Severity : enhancement
4491 Description: bind OST threads to NUMA nodes to improve performance
4492 Details : all OST threads are uniformly bound to CPUs on a single NUMA
4493 node and do their allocations there to localize memory access
4495 Severity : enhancement
4497 Description: llmount can determine client NID directly from Myrinet (GM)
4498 Details : the client NID code from gmnalnid was moved directly into
4499 llmount, removing the need to use this or specifying the
4500 client NID explicitly when mounting GM clients with zeroconf
4503 Frequency : if client is started with down MDS
4505 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
4506 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
4507 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
4508 ptlrpc_import_delayed_req() because we didn't check if the
4509 request was interrupted, and we also didn't break out of the
4510 event loop if there was no timeout
4515 Description: data loss during non-page-aligned writes to a single file from
4516 both multiple nodes and multiple threads on one node at same time
4517 Details : updates to KMS and lsm weren't protected by common lock. Resulting
4518 inconsistency led to false short-reads, that were cached and later
4519 used by ->prepare_write() to fill in partially written page,
4520 leading to data loss.
4523 Frequency : always, if lconf --abort_recovery used
4525 Description: lconf --abort_recovery fails with 'Operation not supported'
4526 Details : lconf was attempting to abort recovery on the MDT device and not
4529 Severity : enhancement
4531 Description: remove cleanup logs
4532 Details : replace lconf-generated cleanup logs with lustre internal
4533 cleanup routines. Eliminates the need for client-cleanup and
4536 Severity : enhancement
4538 Description: add support for EAs (user and system) on lustre filesystems
4539 Details : it is now possible to store extended attributes in the Lustre
4540 client filesystem, and with the user_xattr mount option it
4541 is possible to allow users to store EAs on their files also
4543 Severity : enhancement
4545 Description: Add possibility (config option) to show minimal available OST free
4547 Details : When compiled with --enable-mindf configure option, statfs(2)
4548 (and so, df) will return least minimal free space available from
4549 all OSTs as amount of free space on FS, instead of summary of
4550 free spaces of all OSTs.
4552 Severity : enhancement
4554 Description: do not expand extent locks acquired on OST-side
4555 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
4556 by server: they are not cached anyway.
4559 Frequency : when mmap is used/binaries executed from Lustre
4561 Description: Unmmap pages before throwing them away from read cache.
4562 Details : llap_shrink cache now attempts to unmap pages before discarding
4563 them (if unmapping failed - do not discard). SLES9 kernel has
4564 extra checks that trigger if this unmapping is not done first.
4569 Description: lconf didn't resolve symlinks before checking to see whether a
4570 given mountpoint was already in use
4573 Frequency : when migrating failover services
4574 Bugzilla : 6395, 9514
4575 Description: When migrating a subset of services from a node (e.g. failback
4576 from a failover service node) the remaining services would
4577 time out and evict clients.
4578 Details : lconf --force (implied by --failover) sets the global obd_timeout
4579 to 5 seconds in order to quickly disconnect, but this caused
4580 other RPCs to time out too quickly. Do not change the global
4581 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
4583 Severity : enhancement
4584 Frequency : if MDS is started with down OST
4585 Bugzilla : 9439,5706
4586 Description: Allow startup/shutdown of an MDS without depending on the
4587 availability of the OSTs.
4588 Details : Asynchronously call mds_lov_synchronize during MDS startup.
4589 Add appropriate locking and lov-osc refcounts for safe
4590 cleaning. Add osc abort_inflight calls in case the
4591 synchronize never started.
4594 Frequency : occasional (Cray XT3 only)
4596 Description: root not authorized to access files in CRAY_PORTALS environment
4597 Details : The client process capabilities were not honoured on the MDS in
4598 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
4599 been accessed by an authorized user then root was able to access
4600 the file on the local client also. The root user capabilities
4601 are now allowed on the MDS, as this environment has secure UID.
4604 Frequency : occasional
4606 Description: ldiskfs "too long searching" message happens too often
4607 Details : A debugging message (otherwise harmless) prints too often on
4608 the OST console. This has been reduced to only happen when
4609 there are fragmentation problems on the filesystem.
4614 Description: Division by zero in statfs when all OSCs are inactive
4615 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
4616 lov_statfs divides by value returned by lov_get_stripecnt().
4620 Bugzilla : 9489, 3273
4621 Description: First write from each client to each OST was only 4kB in size,
4622 to initialize client writeback cache, which caused sub-optimal
4623 RPCs and poor layout on disk for the first writen file.
4624 Details : Clients now request an initial cache grant at (re)connect time
4625 and so that they can start streaming writes to the cache right
4626 away and always do full-sized RPCs if there is enough data.
4627 If the OST is rebooted the client also re-establishes its grant
4628 so that client cached writes will be honoured under the grant.
4633 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
4634 Details : Now I/O RPCs go to different portal number and (presumably) fast
4635 lock requests (and glimses) and other RPCs get their own service
4636 threads pool that should be able to service those RPCs
4639 Severity : enhancement
4641 Description: Ability to exchange lustre version between client and servers and
4642 issue warnings at client side if client is too old. Also for
4643 liblustre clients there is ability to refuse connection of too old
4645 Details : New 'version' field is added to connect data structure that is
4646 filled with version info. That info is later checked by server and
4650 Frequency : rare, liblustre only.
4651 Bugzilla : 9296, 9581
4652 Description: Two simultaneous writes from liblustre at offset within same page
4653 might proceed at the same time overwriting eachother with stale
4655 Details : I/O lock withing llu_file_prwv was released too early, before data
4656 actually was hitting the wire. Extended lock-holding time until
4657 server acknowledges receiving data.
4660 Frequency : extremely rare. Never observed in practice.
4662 Description: avoid generating lustre_handle cookie of 0.
4663 Details : class_handle_hash() generates handle cookies by incrementing
4664 global counter, and can hit 0 occasionaly (this is unlikely, but
4665 not impossible, because initial value of cookie counter is
4666 selected randonly). Value of 0 is used as a sentinel meaning
4667 "unassigned handle" --- avoid it. Also coalesce two critical
4668 sections in this function into one.
4670 Severity : enhancement
4672 Description: allow liblustre clients to delegate truncate locking to OST
4673 Details : To avoid overhead of locking, liblustre client instructs OST to
4674 take extent lock in ost_punch() on client's behalf. New connection
4675 flag is added to handle backward compatibility.
4677 Severity : enhancement
4678 Bugzilla : 4928, 7341, 9758
4679 Description: allow number of OST service threads to be specified
4680 Details : a module parameter allows the number of OST service threads
4681 to be specified via "options ost ost_num_threads={N}" in the
4682 OSS's /etc/modules.conf or /etc/modprobe.conf.
4686 Bugzilla : 6146, 9635, 9895
4687 Description: servers crash with bad pointer in target_handle_connect()
4688 Details : In rare cases when a client is reconnecting it was possible that
4689 the connection request was the last reference for that export.
4690 We would temporarily drop the export reference and get a new
4691 one, but this may have been the last reference and the export
4692 was just destroyed. Get new reference before dropping old one.
4694 Severity : enhancement
4695 Frequency : if client is started with failover MDS
4697 Description: Allow multiple MDS hostnames in the mount command
4698 Details : Try to read the configuration from all specified MDS
4699 hostnames during a client mount in case the "primary"
4702 Severity : enhancement
4704 Description: Stop sending data to evicted clients as soon as possible.
4705 Details : Check if the client we are about to send or are sending data to
4706 was evicted already. (Check is done every second of waiting,
4707 for which l_wait_event interface was extended to allow checking
4708 of exit condition at specified intervals).
4711 Frequency : rare, normally only when NFS exporting is done from client
4713 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
4715 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
4718 Frequency : always, if srand() is called before liblustre initialization
4720 Description: Liblustre uses system PRNG disturbing its usage by user application
4721 Details : Introduce internal to lustre fast and high-quality PRNG for
4722 lustre usage and make liblustre and some other places in generic
4723 lustre code to use it.
4725 Severity : enhancement
4726 Bugzilla : 9477, 9557, 9870
4727 Description: Verify that the MDS configuration logs are updated when xml is
4728 Details : Check if the .xml configuration logs are newer than the config
4729 logs stored on the MDS and report an error if this is the case.
4730 Request --write-conf, or allow starting with --old_conf.
4732 Severity : enhancement
4734 Description: Handle symlinks in the path when checking if Lustre is mounted.
4735 Details : Resolve intermediate symlinks when checking if a client has
4736 mounted a filesystem to avoid duplicate client mounts.
4741 Description: lconf can hit an error exception but still return success.
4742 Details : The lconf command catches the Command error exception at the top
4743 level script context and will exit with the associated exit
4744 status, but doesn't ensure that this exit status is non-zero.
4749 Description: failure of ptlrpc thread startup can cause oops
4750 Details : Starting a ptlrpc service thread can fail if there are a large
4751 number of threads or the server memory is very fragmented.
4752 Handle this without oopsing.
4755 Frequency : always, only if liblustre and non-default acceptor port was used
4757 Description: liblustre cannot connect to servers with non-default acceptor port
4758 Details : tcpnal_set_default_params() was not called and was therefore
4759 ignoring the environment varaible TCPNAL_PORT, as well as other
4760 TCPNAL_ environment variables
4765 Description: two objects could be created on the same OST for a single file
4766 Details : If an OST is down, in some cases it was possible to create two
4767 objects on a single OST for a single file. No problems other
4768 than potential performance impact and spurious error messages.
4772 Bugzilla : 5681, 9562
4773 Description: Client may oops in ll_unhash_aliases
4774 Details : Client dcache may become inconsistent in race condition.
4775 In some cases "getcwd" can fail if the current directory is
4781 Description: Inode refcounting problems in NFS export code
4782 Details : link_raw functions used to call d_instantiate without obtaining
4783 extra inode reference first.
4787 Bugzilla : 9942, 9903
4788 Description: Referencing freed requests leading to crash, memleaks with NFS.
4789 Details : We used to require that call to ll_revalidate_it was always
4790 followed by ll_lookup_it. Also with revalidate_special() it is
4791 possible to call ll_revalidate_it() twice for the same dentry
4792 even if first occurence returned success. This fix changes semantic
4793 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
4794 reference on a request referred from the intent.
4795 ll_intent_release() then releases such a request.
4798 Frequency : rare, normally benchmark loads only
4800 Description: unlinked inodes were kept in memory on the client
4801 Details : If a client is repeatedly creating and unlinking files it
4802 can accumulate a lot of stale inodes in the inode slab cache.
4803 If there is no other client load running this can cause the
4804 client node to run out of memory. Instead flush old inodes
4805 from client cache that have the same inode number as a new inode.
4808 Frequency : SLES9 2.6.5 kernel and long filenames only
4809 Bugzilla : 9969, 10379
4810 Description: utime reports stale NFS file handle
4811 Details : SLES9 uses out-of-dentry names in some cases, which confused
4812 the lustre dentry revalidation. Change it to always use the
4816 Frequency : rare, unless heavy write-truncate concurrency is continuous
4817 Bugzilla : 4180, 6984, 7171, 9963, 9331
4818 Description: OST becomes very slow and/or deadlocked during object unlink
4819 Details : filter_destroy() was holding onto the parent directory lock
4820 while truncating+unlinking objects. For very large objects this
4821 may block other threads for a long time and slow overall OST
4822 responsiveness. It may also be possible to get a lock ordering
4823 deadlock in this case, or run out of journal credits because of
4824 the combined truncate+unlink. Solution is to do object truncate
4825 first in one transaction without parent lock, and then do the
4826 final unlink in a new transaction with the parent lock. This
4827 reduces the lock hold time dramatically.
4830 Frequency : rare, 2.4 kernels only
4832 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
4833 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
4834 before unmounting their filesystem in order to avoid deadlock.
4835 The kernel_locked() function in 2.4 kernels only checks whether
4836 the kernel lock is held, not whether it is this process that is
4837 holding it as 2.6 kernels do.
4842 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
4843 Details : The client ptlrpc code may be trying to reconnect to a down
4844 server before a previous connection attempt has timed out.
4845 Increase the reconnect interval to be longer than the connection
4846 timeout interval to avoid sending duplicate connections to
4850 Frequency : echo_client brw_test command
4852 Description: fix echo_client to work with OST preallocated code
4853 Details : OST preallocation code (5137) didn't take echo_client IO path
4854 into account: echo_client calls filter methods outside of any
4855 OST thread and, hence, there is no per-thread preallocated
4856 pages and buffers to use. Solution: hijack pga pages for IO. As
4857 a byproduct, this avoids unnecessary data copying.
4861 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
4862 Description: Client can oops in mdc_commit_close() after open replay
4863 Details : It was possible for the MDS to return an open request with no
4864 transaction number in mds_finish_transno() if the client was
4865 evicted, but without actually returning an error. Clients
4866 would later try to replay that open and may trip an assertion
4867 Simplify the client close codepath, and always return an error
4868 from the MDS in case the open is not successful.
4871 Frequency : rare, 2.6 OSTs only
4873 Description: OST may deadlock under high load on fragmented files
4874 Details : If there was a heavy load and highly-fragmented OST filesystems
4875 it was possible to have all the OST threads deadlock waiting on
4876 allocation of biovecs, because the biovecs were not released
4877 until the entire RPC IO was completed. Instead, release biovecs
4878 as soon as they are complete to ensure forward IO progress.
4880 Severity : enhancement
4882 Description: Support for specifying external journal device at mount
4883 Details : If an OST or MDS device is formatted with an external journal
4884 device, this device major/minor is stored in the ext3 superblock
4885 and may not be valid for failover. Allow detecting and
4886 specifying the external journal at mount time.
4891 Description: Mounting an MDS with pending unlinked files may cause oops
4892 Details : target_finish_recovery() calls mds_postrecov() which returned
4893 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
4894 considers this an error and immediately begins cleaning up the
4895 lov, just after starting the mds_lov process
4897 Severity : enhancement
4899 Description: Implement 'lfs df' to report actual free space on per-OST basis
4900 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
4901 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
4902 usage of objects; '-h' to report in human readable format.
4904 ------------------------------------------------------------------------------
4906 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
4913 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
4914 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
4915 at startup during orphan unlinking. Instead, skip buddy update
4916 before setup, as it will be regenerated after recovery is complete.
4921 Description: If an OST is inactive, its locks might reference stale inodes.
4922 Details : lov_change_cbdata() must iterate over all namespaces, even if
4923 they are inactive to clear inode references from the lock.
4925 Severity : enhancement
4926 Frequency : occasional, if non-standard max_dirty_mb used
4928 Description: Client will block write RPCs if not enough grant
4929 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
4930 then the client will block writes while waiting for another RPC
4931 to complete instead of consuming its dirty limit. With change
4932 we get improved performance when max_dirty_mb is small.
4934 Severity : enhancement
4935 Bugzilla : 3389, 6253
4936 Description: Add support for supplementary groups on the MDS.
4937 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
4938 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
4939 lookups for user supplementary groups into a cache.
4943 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
4944 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
4945 for liblustre O_EXCL is enforced by MDS.
4949 Description: readdir never returns NULL in liblustre.
4950 Details : Corrected llu_iop_getdirentries logic, to return offset of next
4951 dentry in struct dirent.
4955 Frequency : liblustre only, depends on application IO pattern
4956 Description: liblustre clients evicted if not contacting servers
4957 Details : Don't put liblustre clients into the ping_evictor list, so
4958 they will not be evicted by the pinger ever.
4960 Severity : enhancement
4962 Description: Add ability to evict clients by NID from MDS.
4963 Details : By echoing "nid:$NID" string into
4964 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
4965 $NID would be instantly evicted from this MDS and from all active
4966 OSTs connected to it.
4970 Description: Do not query file size twice, somewhat slowing stat(2) calls.
4971 Details : lookup_it_finish() used to query file size from OSTs that was not
4976 Description: service threads change working directory to that of init
4977 Details : Starting lustre service threads may pin the working directory
4978 of the parent thread, making that filesystem busy. Threads
4979 now change to the working directory of init to avoid this.
4983 Frequency : during shutdown only
4984 Description: shutdown with a failed MDS or OST can cause unmount to hang
4985 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
4990 Frequency : 2.6 only
4991 Description: chmod/chown may include an extra supplementary group
4992 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
4993 supplementary group and if none is specified this is used.
4996 Bugzilla : 5479 (6816)
4998 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
4999 Details : If lookup is in progress on a file that is unlinked we might try
5000 to revalidate the inode and fail in revalidate after lookup is
5001 complete and ll_file_open() enqueues the open again but
5002 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
5005 Frequency : always, if lconf --abort_recovery used
5007 Description: lconf --abort_recovery fails with 'Operation not supported'
5008 Details : lconf was attempting to abort recovery on the MDT device and not
5011 ------------------------------------------------------------------------------
5013 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
5018 Frequency : rare (only unsupported configurations with a node running as an
5020 Bugzilla : 6514, 5137
5021 Description: Mounting a Lustre file system on a node running as an OST could
5023 Details : OSTs now preallocates memory needed to write out data at
5024 startup, instead of when needed, to avoid having to
5025 allocate memory in possibly low memory situations.
5026 Specifically, if the file system is mounted on on OST,
5027 memory pressure could force it to try to write out data,
5028 which it needed to allocate memory to do. Due to the low
5029 memory, it would be unable to do so and the node would
5030 become unresponsive.
5032 Severity : enhancement
5034 Description: Addition of lconf --service command line option
5035 Details : lconf now accepts a '--service <arg>' option, which is
5036 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
5038 Severity : enhancement
5040 Description: Failover mode is now the default for OSTs.
5041 Details : By default, OSTs will now run in failover mode. To return to
5042 the old behaviour, add '--failout' to the lmc line for OSTs.
5044 Severity : enhancement
5046 Description: Health checks are now provided for MDS and OSTs
5047 Details : Additional detailed health check information on MSD and OSTs
5048 is now provided through the procfs health_check value.
5051 Frequency : occasional, depends on IO load
5053 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
5054 after numerous create/delete cycles
5055 Details : The ext3 inode allocation policy would not allocate new inodes
5056 very well on the OSTs because there are no new directories
5057 being created. Instead we look for groups with free space if
5058 the parent directories are nearly full.
5063 Description: Network or server problems during mount may cause partially
5064 mounted clients instead of returning an error.
5065 Details : The config llog parsing code may overwrite the error return
5066 code during mount error handling, returning success instead
5072 Description: MDS can fail to allocate large reply buffers
5073 Details : After long uptimes the MDS can fail to allocate large reply
5074 buffers (e.g. zconf client mount config records) due to memory
5075 fragmentation or consumption by the buffer cache. Preallocate
5076 some large reply buffers so that these replies can be sent even
5077 under memory pressure.
5081 Frequency : rare (liblustre)
5082 Description: fsx running with liblustre complained that using truncate() to
5083 extend the file doesn't work. This patch corrects that issue.
5084 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
5085 ATTR_SIZE and lsm use in llu_setattr_raw.
5089 Frequency : rare, only 2.6 kernels
5090 Description: Unusual file access patterns on the MDS may result in inode
5091 data being lost in very rare circumstances.
5092 Details : Bad interaction between the ea-in-inode patch and the "no-read"
5093 code in the 2.6 kernel caused the inode and/or EA data not to
5094 be read from disk, causing single-file corruption.
5098 Frequency : rare, only 2.6 filesystems using extents
5099 Description: Heavy concurrent write and delete load may cause data corruption.
5100 Details : It was possible under high-load situations to have an extent
5101 metadata block in the block device cache from a just-unlinked
5102 file overwrite a newly-allocated data block. We now unmap any
5103 metadata buffers that alias just-allocated data blocks.
5107 Frequency : filesystems with default stripe_count larger than 77
5108 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
5109 Details : lconf specifies an inode size of 4096 bytes when the default
5110 stripe_count is larger than 77. This conflicts with the default
5111 inode density of 1 per 4096 bytes. Allocate smaller inodes in
5112 this case to avoid pinning too much memory for large EAs.
5114 ------------------------------------------------------------------------------
5116 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
5121 Frequency : rare (extremely heavy IO load with hundreds of clients)
5123 Description: Client is evicted, gets IO error writing to file
5124 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
5125 caused clients to be evicted for AST timeouts. The fixes in
5126 bug 5192 mean we no longer need to have such short AST timeouts
5127 so ldlm_timeout has been increased.
5130 Frequency : occasional during --force or --failover shutdown under load
5131 Bugzilla : 5949, 4834
5132 Description: Server oops/LBUG if stopped with --force or --failover under load
5133 Details : a collection of import/export refcount and cleanup ordering
5134 issues fixed for safer force cleanup
5137 Frequency : only filesystems larger than 120 OSTs
5138 Bugzilla : 5990, 6223
5139 Description: lfs getstripe would oops on a very large filesystem
5140 Details : lov_getconfig used kfree on vmalloc'd memory
5143 Frequency : only filesystems exporting via NFS to Solaris 10 clients
5144 Bugzilla : 6242, 6243
5145 Description: reading from files that had been truncated to a non-zero size
5146 but never opened returned no data
5147 Details : ll_file_read() reads zeros from no-object files to EOF
5152 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
5153 Details : There's little harm in aborting MDS/OSS recovery and letting it
5154 try again, so I removed the LASSERT and return an error instead.
5156 Severity : enhancement
5158 Description: New debugging infrastructure for tracking down data corruption
5159 Details : The I/O checksum code was replaced to: (a) control it at runtime,
5160 (b) cover more of the client-side code path, and (c) try to narrow
5161 down where problems occurred
5165 Bugzilla : 3819, 4364, 4397, 6313
5166 Description: Racing close and eviction MDS could cause assertion in mds_close
5167 Details : It was possible to get multiple mfd references during close and
5168 client eviction, leading to one thread referencing a freed mfd.
5170 Severity: : enhancement
5171 Bugzilla : 3262, 6359
5172 Description: Attempts to reconnect to servers are now more aggressive.
5173 Details : This builds on the enhanced upcall-less recovery that was added
5174 in 1.4.2. When trying to reconnect to servers, clients will
5175 now try each server in the failover group every 10 seconds. By
5176 default, clients would previously try one server every 25 seconds.
5181 Description: After recovery, certain operations trigger a failed
5182 assertion on a client.
5183 Details : Failing over an mds, using lconf -d --failover, while a
5184 client was doing a readdir() call would cause the client to
5185 LBUG after recovery completed and the readdir() was resent.
5187 Severity : enhancement
5189 Description: Default groups are now added by lconf
5190 Details : You can now run lconf --group <servicename> without having to
5191 manually add groups with lmc.
5194 Frequency : occasional
5196 Description: Nodes with an elan id of 0 trigger a failed assertion
5199 Frequency : always when accessing e.g. tty/console device nodes
5201 Description: tty and some other devices nodes cannot be used on lustre
5202 Details : file's private_data field is used by device data and lustre
5203 values in there got lost. New field was added to struct file to
5204 store fs-specific private data.
5207 Frequency : when exporting Lustre via NFS
5209 Description: NFSD failed occasionally when looking up a path component
5210 Details : NFSD is looking up ".." which was broken in ext3 directories
5211 that had grown large enough to become hashed.
5214 Frequency : Clusters with multiple interfaces not on the same subnet
5216 Description: Nodes will repeatedly try to reconnect to an interface which it
5217 cannot reach and report an error to the log.
5218 Details : Extra peer list entries will be created by lconf with some peers
5219 unreachable. lconf now validates the peer before adding it.
5222 Frequency : Only if a default stripe is set on the filesystem root.
5224 Description: Setting a default stripe on the filesystem root prevented the
5225 filesystem from being remounted.
5226 Details : The client was sending extra request flags in the root getattr
5227 request and did not allocate a reply buffer for the dir EA.
5230 Frequency : occasional, higher if lots of files are accessed by one client
5231 Bugzilla : 6159, 6097
5232 Description: Client trips assertion regarding lsm mismatch/magic
5233 Details : While revalidating inodes the VFS looks up inodes with ifind()
5234 and in rare cases can find an inode that is being freed.
5235 The ll_test_inode() code will free the lsm during ifind()
5236 when it finds an existing inode and then the VFS later attaches
5237 this free lsm to a new inode.
5241 Bugzilla : 6422, 7030
5242 Description: MDS deadlock between mkdir and client eviction
5243 Details : Creating a new file via mkdir or mknod (starting a transaction
5244 and getting the ns lock) can deadlock with client eviction
5245 (gets ns lock and trying to finish a synchronous transaction).
5248 Frequency : occasional
5249 Description: While starting a server, the fsfilt_ext3 module could not be
5251 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
5252 kernels. Previously, lconf would still use the ext3 name
5253 when trying to load modules. Now, it will correctly use
5254 ext3 on 2.4 and ldiskfs on 2.6.
5256 Severity : enhancement
5257 Description: The default stripe count has been changed to 1
5258 Details : The interpretation of the default stripe count (0, to lfs
5259 or lmc) has been changed to mean striping across a single
5260 OST, rather than all available. For general usage we have
5261 found a stripe count of 1 or 2 works best.
5263 Severity : enhancement
5264 Description: Add support for compiling against Cray portals.
5265 Details : Conditional compiling for some areas that are different
5269 Frequency : occasional
5270 Bugzilla : 6409, 6834
5271 Description: Creating files with an explicit stripe count may lead to
5272 a failed assertion on the MDS
5273 Details : If some OSTs are full or unavailable, creating files may
5274 trigger a failed assertion on the MDS. Now, Lustre will
5275 try to use other servers or return an error to the
5279 Frequency : occasional
5281 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
5282 caused lock timeout during readahead (since 1.4.2).
5283 Details : Processes doing readahead might match a lock that hasn't been
5284 granted yet if there are overlapping and conflicting lock
5285 requests. The readahead process waits on ungranted lock
5286 (original lock is CBPENDING), while OST waits for that process
5287 to cancel CBPENDING read lock and eventually evicts client.
5289 Severity : enhancement
5291 Description: Initial enabling of flock support for clients
5292 Details : Implements fcntl advisory locking and file status functions.
5293 This feature is provided as an optional mount flag (default
5294 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
5295 locking are implemented yet, and those that are are not guaranteed
5296 to be completely correct in production environments.
5297 mount -t lustre -o [flock|noflock] ...
5300 Frequency : occasional
5302 Description: OSTs running 2.4 kernels but with extents enabled might trip an
5303 assertion in the ext3 JBD (journaling) layer.
5304 Details : The b_committed_data struct is protected by the big kernel lock
5305 in 2.4 kernels, serializing journal_commit_transaction() and
5306 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
5307 access to this struct. In 2.6 kernels there is finer grained
5308 locking to improve SMP performance of the JBD layer.
5312 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
5317 Description: merge and cleanup kernel patches.
5320 -----------------------------------------------------------------------------
5322 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
5324 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
5325 versions of Lustre. Please update all servers and clients to
5326 version 1.4.2 or later at the same time. You must also run
5327 "lconf --write-conf {config}.xml" on the MDS while it is stopped
5328 to update the configuration logs.
5330 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
5331 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
5332 - swab llog records of type '0' so we get proper header size/idx (5861)
5333 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
5334 - fix rename of one directory over another leaking an inode (5953)
5335 - avoid SetPageDirty on 2.6 (5981)
5336 - don't re-add just-being-destroyed locks to the waiting list (5653)
5337 - when creating new directories, inherit the parent's custom
5338 striping settings if present parent (3048)
5339 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
5340 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
5341 - don't hold client locks on temporary worklist from l_lru (5666)
5342 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
5343 - automatically evict dead clients (5921)
5344 - Update file size properly in create+truncate+fstat case (6196)
5345 - Do not unhash mountpoint dentries, do not allow removal of
5347 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
5348 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
5349 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
5350 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
5351 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
5352 - Correctly handle reads of files with no objects (6243)
5353 - lctl recover will also mark a device active if deactivate used (5933)
5355 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
5356 - allow --write-conf on an MDS with different nettype than client (5619)
5357 - don't write config llogs to MDS for mounts not from that MDS (5617)
5358 - lconf should create multiple TCP connections from a client (5201)
5359 - init scripts are now turned off by default; run chkconfig --on
5360 lustre and chkconfig --on lustrefs to use them
5361 - upcalls are no longer needed for clients to recover to failover
5363 - add --abort-recovery option to lconf to abort recovery on device
5365 - add support for an arbitrary number of OSTs (3026)
5366 - Quota support protocol changes.
5367 - forward compatibility changes to wire structs (6007)
5368 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5369 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5370 - improved "lustre status" script
5371 - initialize blocksize for non-regular files (6062)
5372 - added --disable-server and --disable-client configure options (5782)
5373 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5374 - Vanilla 2.4.29 support
5375 - increase maximum number of obd devices to 520 (6242)
5376 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5377 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5378 SLES 9 2.6.5 kernels
5379 - limit stripes per file to 160 (the maximum EA size) (6093)
5381 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5384 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5385 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5386 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5387 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5388 - small ext3 extents cleanups and fixes (5733)
5389 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5390 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5391 - avoid needless client->OST connect, fix handle mismatch (5317)
5392 - fix DLM error path that led to out-of-sync client, long delays (5779)
5393 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5394 - fix several locking issues related to i_size (5492,5624,5654,5672)
5395 - don't move pending lock onto export if it is already evicted (5683)
5396 - fix kernel oops when creating .foo in unlinked directory (5548)
5397 - fix deadlock in obdfilter statistics vs. object create (5811)
5398 - use time_{before,after} to avoid timer jiffies wrap (5882)
5399 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5400 - Do not leak request if server was not able to process it (5154)
5401 - If mds_open unable to find parent dir, make that negative lookup(5154)
5402 - don't create new directories with extent-mapping (5909, 5936)
5404 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5405 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5406 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5407 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5408 - support for RHEL4 kernel on i686 (5773)
5409 - provide error messages when incompatible logs are encountered (5898)
5411 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5412 * version 1.4.0.10 (1.4.1 release candidate 1)
5414 - don't keep a lock reference when lock is not granted (4238)
5415 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5416 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5417 - drop import inflight refcount on signal_completed_replay error (5255)
5418 - unlock page after async write error during send (3677)
5419 - handle missing objects in filter_preprw_read properly (5265)
5420 - no transno return for symlink open, don't save no-trasno open (3440)
5421 - don't try to complete elan receive that already failed (4012)
5422 - free RPC server reply state on error (5406)
5423 - clean up thread from ptlrpc_start_thread() on error (5160)
5424 - readahead could read extra page into cache that wasn't ejected (5388)
5425 - prevent races in class_attach/setup/cleanup/detach (5260)
5426 - don't dereference de->d_inode after l_dput of de (5458)
5427 - use "int" for stripe value returned from lock_to_stripe (5544)
5428 - mballoc allocation and error-checking fixes in 2.6 (5504)
5429 - block device patches to fix I/O request sizes in 2.6 (5482)
5430 - look up hostnames for IB nals (5602)
5431 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5432 - don't start multiple acceptors for the same port (5277)
5433 - fix incorrect LASSERT in mds_getattr_name (5635)
5434 - export a proc file for general "ping" checking (5628)
5435 - fix "lfs check" to not block when the MDS is down (5628)
5437 - service request history (4965)
5438 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5439 - create an "evict_client" /proc entry on OSTs, like the MDS has
5440 - fix mount usage message, return errors per mount(8) (5168)
5441 - change grep [] to grep "[]" in tests so they work in more UMLs
5442 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5443 - remove ancient LOV_MAGIC_V0 EA support (5047)
5444 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5445 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5446 - allow readahead allocations to fail when low on memory (5383)
5447 - mmap locking landed again, after considerable improvement (2828)
5448 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
5450 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
5453 - send OST transaction number in read/write reply to free req (4966)
5454 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5455 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5456 - the watchdog thread now runs as interruptible (5246)
5457 - flock/lockf fixes (but it's still disabled, pending 5135)
5458 - don't use EXT3 constants in llite code (5094)
5459 - memory shortage at startup could cause assertion (5176)
5461 - reorganization of lov code
5462 - single portals codebase
5464 - add extents/mballoc support (5025)
5465 - direct I/O reads in the obdfilter (4048)
5466 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
5468 tbd Cluster File Systems, Inc. <info@clusterfs.com>
5471 - send OST transaction number in read/write reply to free req (4966)
5472 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5473 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5474 - the watchdog thread now runs as interruptible (5246)
5475 - handle missing objects in filter_preprw_read properly (5265)
5476 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5477 - drop import inflight refcount on signal_completed_replay error (5255)
5478 - unlock page after async write error during send (3677)
5479 - return original error code on reconstructed replies (3761)
5480 - no transno return for symlink open, don't save no-trasno open (3440)
5482 - add pid to ldlm debugging output (4922)
5483 - bump the watchdog timeouts -- we can't handle 30sec yet
5484 - extra debugging for orphan dentry/inode bug (5259)
5486 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
5489 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
5490 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
5491 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
5492 - fix NULL dereference in /proc/sys/portals/routes (4827)
5493 - allow failed mdc_close() operations to be interrupted (4561)
5494 - stop precreate on OST before MDS would time out on it (4778)
5495 - don't send partial-page writes before EOF from client (4410)
5496 - discard client grant for sub-page writes on large-page clients (4520)
5497 - don't free dentries not owned by NFS code, check generation (4806)
5498 - fix lsm leak if mds_create_objects() fails (4801)
5499 - limit debug_daemon file size, always print CERROR messages (4789)
5500 - use transno after validating reply (3892)
5501 - process timed out requests if import state changes (3754)
5502 - update mtime on OST during writes, return in glimpse (4829)
5503 - add mkfsoptions to LDAP (4679)
5504 - use ->max_readahead method instead of zapping global ra (5039)
5505 - don't interrupt __l_wait_event() during strace
5507 - add software watchdogs to catch hung threads quickly (4941)
5508 - make lustrefs init script start after nfs is mounted
5509 - fix CWARN/ERROR duplication (4930)
5510 - return async write errors to application if possible (2248)
5511 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
5512 - print NAL number in %x format (4645)
5513 - update barely-supported suse-2.4.21-171 series (4842)
5514 - support for sles 9 %post scripts
5515 - support for building 2.6 kernel-source packages
5516 - support for sles km_* packages
5518 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
5521 - ignore -ENOENT errors in osc_destroy (3639)
5522 - notify osc create thread that OSC is being cleaned up (4600)
5523 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
5524 - reconstruct ost_handle() like mds_handle() (4657)
5525 - create a new thread to do import eviction to avoid deadlock (3969)
5526 - let lconf resolve symlinked-to devices (4629)
5527 - don't unlink "objects" from directory with default EA (4554)
5528 - hold socknal file ref over connect in case target is down (4394)
5529 - allow more than 32000 subdirectories in a single directory (3244)
5530 - fix blocks count for O_DIRECT writes (3751)
5531 - OST returns ENOSPC from object create when no space left (4539)
5532 - don't send truncate RPC if file size isn't changing (4410)
5533 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
5534 - bind to privileged port in socknal and tcpnal (3689)
5536 - rate limit CERROR/CWARN console message to avoid overload (4519)
5537 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
5538 - basic mmap support (3918)
5539 - kernel patch series update from b1_4 (4711)
5541 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
5544 - avoid crash during MDS cleanup with OST shut down (2775)
5545 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
5546 - don't use bad inodes on the MDS (3744)
5547 - dynamic object preallocation to improve recovery speed (4236)
5548 - don't hold spinlock over lock dumping or change debug flags (4401)
5549 - don't zero obd_dev when it is force cleaned (3651)
5550 - print grants to console if they go negative (4431)
5551 - "lctl deactivate" will stop automatic recovery attempts (3406)
5552 - look for existing locks in ldlm_handle_enqueue() (3764)
5553 - don't resolve lock handle twice in recovery avoiding race (4401)
5554 - revalidate should check working dir is a directory (4134)
5556 - don't always mark "slow" obdfilter messages as errors (4418)
5558 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
5561 - don't close LustreDB during write_conf until it is done (3860)
5562 - fix typo in lconf for_each_profile (3821)
5563 - allow dumping logs from multiple threads at one time (3820)
5564 - don't allow multiple threads in OSC recovery (3812)
5565 - fix debug_size parameters (3864)
5566 - fix mds_postrecov to initialize import for llog ctxt (3121)
5567 - replace config semaphore with spinlock (3306)
5568 - be sure to send a reply for a CANCEL rpc with bad export (3863)
5569 - don't allow enqueue to complete on a destroyed export (3822)
5570 - down write_lock before checking llog header bitmap (3825)
5571 - recover from lock replay timeout (3764)
5572 - up llog sem before sending rpc (3652)
5573 - reduce ns lock hold times when setting kms (3267)
5574 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
5575 - fix NULL deref and obd_dev leak on setup error (3312)
5576 - replace some LBUG about llog ops with error handling (3841)
5577 - don't match INVALID dentries from d_lookup and spin (3784)
5578 - hold dcache_lock while marking dentries INVALID and hashing (4255)
5579 - fix invalid assertion in ptlrpc_set_wait (3880)
5581 - add libwrap support for the TCP acceptor (3996)
5582 - add /proc/sys/portals/routes for non-root route listing (3994)
5583 - allow setting MDS UUID in .xml (2580)
5584 - print the stack of a process that LBUGs (4228)
5586 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
5589 - don't cleanup request in ll_file_open() on failed MDS open (3430)
5590 - make sure to unset replay flag from failed open requests (3440)
5591 - if default stripe count is 0, use OST count for inode size (3636)
5592 - update parent mtime/ctime on client for create/unlink (2611)
5593 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
5594 - free recovery state on server during a forced cleanup (3571)
5595 - unregister_reply for resent reqs (3063)
5596 - loop back devices mounting and status check on 2.6 (3563)
5597 - fix resource-creation race that can provoke i_size == 0 (3513)
5598 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
5599 - more debugging for page-accounting assertion (3746)
5600 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
5601 - avoid deadlock after precreation failure (3758)
5602 - fix race and lock order deadlock in orphan handling (3450, 3750)
5603 - add validity checks when grabbing inodes from l_ast_data (3599)
5605 - add /proc/.../recovery_status to obdfilter (3428)
5606 - lightweight CDEBUG infrastructure, debug daemon (3668)
5607 - change default OSC RPC parameters to be better on small clusters
5608 - turn off OST read cache for files smaller than 32MB
5609 - install man pages and include them in rpms (3100)
5610 - add new init script for (un)mounting lustre filesystems (2593)
5611 - run chkconfig in %post for init scripts (3701)
5612 - drop scimac NAL (unmaintained)
5614 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
5617 - clean kiobufs before and after use (3485)
5618 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
5619 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
5620 - change init script to fail more clearly if not run as root (1528)
5621 - allow clients to reconnect during replay (1742)
5622 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
5623 - don't do DNS lookups on NIDs too small for IP addresses (3442)
5624 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
5625 - fix cond_resched (3554)
5626 - only evict unfinished clients after recovery (3515)
5627 - allow bulk resend, prevent data loss (3570)
5628 - dynamic ptlrpc request buffer allocation (2102)
5629 - don't allow unlinking open directory if it isn't empty (2904)
5630 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
5631 - remove extraneous obd dereference causing LASSERT failure (3334)
5632 - don't use get_cycles() when creating temp. files on the mds (3156)
5633 - hold i_sem when setting i_size in ll_extent_lock() (3564)
5634 - handle EEXIST for set-stripe, set proper directory name (3336)
5636 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
5637 - fix ksocknal_fmb_callback() error messages (2918)
5639 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
5642 - don't copy lvb into (possibly NULL) reply on error (2983)
5643 - don't deref dentry after dput, don't free lvb on error (2922)
5644 - use the kms to determine writeback rpc length (2947)
5645 - increment oti_logcookies when osc is inactive (2948)
5646 - update client's i_blocks count via lvb messages (2543)
5647 - handle intent open/close of special files properly (1557)
5648 - mount MDS with errors=remount-ro, like obdfilter (2009)
5649 - initialize lock handle to avoid ASSERT on error cleanup (3057)
5650 - don't use cancelling-locks' kms values (2947)
5651 - use highest lock extent for kms, not last one (2925)
5652 - don't dereference ERR_PTR() dentry in error handling path (3107)
5653 - fix thread race in portals_debug_dumplog() (3122)
5654 - create lprocfs device entries at setup instead of at attach (1519)
5655 - common AST error handler, don't evict client on completion race (3145)
5656 - zero nameidata in detach_mnt in 2.6 (3118)
5657 - verify d_inode after revalidate_special is valid in 2.6 (3116)
5658 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
5659 - initialize RPC timeout timer earlier for 2.6 (3219)
5660 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
5661 - print nal/nid for unknown nid (3258)
5662 - additional checks for oscc recovery before doing precreate (3284)
5663 - fix ll_extent_lock() error return code for 64-bit systems (3043)
5664 - don't crash in mdc_close for bad permissions on open (3285)
5665 - zero i_rdev for non-device files (3147)
5666 - clear page->private before handing to FS, better assertion (3119)
5667 - tune the read pipeline (3236)
5668 - fix incorrect decref of invalidated dentry (2350)
5669 - provide read-ahead stats and refine rpc in flight stats (3328)
5670 - don't hold journal transaction open across create RPC (3313)
5671 - update atime on MDS at close time (3265)
5672 - close LDAP connection when recovering to avoid server load (3315)
5673 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
5674 - don't leak open file on MDS after open resend (3325)
5675 - serialize filter_precreate and filter_destroy_precreated (3329)
5676 - loop device shouldn't call sync_dev() for nul device (3092)
5677 - clear page cache after eviction (2766)
5678 - resynchronize MDS->OST in background (2824)
5679 - refuse to mount the same filesystem twice on same mountpoint (3394)
5680 - allow llmount to create routes for mounting behind routers (3320)
5681 - push lock cancellation to blocking thread for glimpse ASTs (3409)
5682 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
5683 - fix rare problem with rename on htree directories (3417)
5685 - allow default OST striping configuration per directory (1414)
5686 - fix compilation for qswnal for 2.6 kernels (3125)
5687 - increase maximum number of MDS request buffers for large systems
5688 - change liblustreapi to be useful for external progs like lfsck (3098)
5689 - increase local configuration timeout for slow disks (3353)
5690 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
5692 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5695 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
5696 - don't overwrite extent policy data in reply if lock was blocked (2901)
5697 - drop filter export grants atomically with removal from device (2663)
5698 - del obd_self_export from work_list in class_disconnect_exports (2908)
5699 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
5700 - swab reply message in mdc_close, other PPC fixes (2464)
5701 - fix destroying of named logs (2325)
5702 - overwrite old logs when running lconf --write_conf (2264)
5703 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
5704 - fix race in target_handle_connect (2898)
5705 - mds_reint_create() should take same inode create lock (2926)
5706 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
5707 - don't close files for self_export to avoid uninitialized obd (2936)
5708 - allow MDS with the same name as client node (2939)
5709 - hold dentry reference for closed log files for unlink (2325)
5710 - reserve space for all logs during transactions (2059)
5711 - don't evict page beyond end of stripe extent (2925)
5712 - don't oops on a deleted current working directory (2399)
5713 - handle hard links to targets without a parent properly (2517)
5714 - don't dereference NULL lock when racing during eviction (2867)
5715 - don't grow lock extents when lots of conflicting locks (2919)
5717 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
5720 - account for cache space usage on clients to avoid data loss (974)
5721 - lfsck support in lustre kernel code (2349)
5722 - reduce journal credits needed for BRW writes (2370)
5723 - orphan handling to avoid losing space on client/server crashes
5724 - ptlrpcd can be blocked, stopping ALL progress (2477)
5725 - use lock value blocks to assist in proper KMS, faster stat (1021)
5726 - takes i_sem instead of DLM locks internally on obdfilter (2720)
5727 - recovery for initial connections (2355)
5728 - fixes for mds_cleanup_orphans (1934)
5729 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
5730 - block all file creations until orphan recovery completes (1901)
5731 - client remove rq_connection from request struct (2423)
5732 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
5733 - recovery timer race (2670)
5734 - mdc_close recovey bug (2532)
5735 - ptlrpc cleanup bug (2710)
5736 - mds timeout on local locks (2588)
5737 - namespace lock held during RPCs (2431)
5738 - handle interrupted sync write properly (2503)
5739 - don't try to handle a message that hasn't been replied to (2699)
5740 - client assert failure during cleanup after abort recovery (2701)
5741 - leak mdc device after failed mount (2712)
5742 - ptlrpc_check_set allows timedout requests to complete (2714)
5743 - wait for inflight reqs when ptlrpcd finishes (2710)
5744 - make sure unregistered services are removed from the srv_list
5745 - reset bulk XID's when resending them (caught by 1138 test)
5746 - unregister_bulk after timeout
5747 - fix lconf error (2694)
5748 - handle write after unfinished setstripe, stripe-only getstripe (2388)
5749 - readahead locks pages, leaves pending causing memory pressure (2673)
5750 - increase OST request buffers to 4096 on large machines (2729)
5751 - fix up permission of existing directories in simple_mkdir (2661)
5752 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
5753 - don't assign transno to errored transactions (2742)
5754 - don't delete objects on OST if given a bogus objid from MDS (2751)
5755 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
5756 - if rq_no_resend, then timeout request after recovery (2432)
5757 - fix MDS llog_logid record size, 64-bit array alignment (2733)
5758 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
5759 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
5760 - MDS recovery shouldn't delete objects that it has given out (2730)
5761 - if enqueue arrives after completion, don't clobber LVB (2819)
5762 - don't unlock pages twice when trigger_group_io returns error (2814)
5763 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
5764 - don't write pages to disk if there was an error (1450)
5765 - don't ping imports that have recovery disabled (2676)
5766 - take buffered bytes into account when balancing socknal conn (2817)
5767 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
5768 - reconnect unlink llog connection after MDS reconnects to OST (2816)
5769 - remove little-endian swabbing of llog records (1987)
5770 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
5771 - retry reposting request buffers if they fail (1191)
5772 - grow extent at grant time to avoid granting a revoked lock (2809)
5773 - lock revoke doesn't evict page if covered by a second lock (2765)
5774 - disable VM readahead to avoid reading outside lock extents (2805)
5776 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
5777 - updated kernel patches for hp-2.4.20 kernel (2681)
5779 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
5782 - fix truncated write corruption (2366)
5783 - fix for failed assertion in iopen_connect_dentry (1792,2517)
5785 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
5786 - ensure len(uuid) < 37 in lmc (1171)
5787 - fix ia64 OOPS in llog_test (2255)
5788 - zero end of page at obdfilter for partial page writes (2648)
5789 - don't leave stale dentries around after renames (bug 2428)
5790 - fix timeouts when evicting a client with a single lock held (2642)
5791 - set deadline for the initial HELLO message to drain (2634)
5792 - print out dotted-quad IP addresses in the socknal (2302)
5794 - additional debugging for MDS client eviction problem (2443)
5795 - fix mkfsoptions support for osts (2603, 2604)
5797 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
5800 - add series for the vanilla 2.6.0 kernel
5801 - add series for the vanilla 2.4.24 kernel
5802 - add series for a cray x86/64 UL kernel drop
5803 - fix xattr patches for the vanilla 2.4.19 series
5805 - generate true UUIDs in lmc (1171)
5806 - have portals stack dumping break in UML (2466)
5807 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
5808 - allocate with _NOFS in ldlm to avoid deadlock (1933)
5809 - wake callback waiting threads on client eviction (2460)
5810 - Add --ptldebug and --subsystem to lmc (1719)
5811 - update assertion to allow safe interrupt allocation
5812 - set rq_no_resend for cancel requests (2432)
5813 - recalculate ptlrpcd timeout after resend (2494)
5814 - call vfs_rmdir when removing pending directories (2368)
5815 - fix renaming a file to itself (2429)
5816 - lmc creates a default one-stripe lov (2454)
5817 - expand procfs space to handle large clusters (2326)
5818 - increase UML stack to avoid overflow
5819 - update lconf's list of debug and subsystem masks
5820 - fix lfs find --obd (2510)
5821 - /proc tunable for disabling filter read caching (2591)
5822 - stop rpm packages from altering slapd.conf (2301)
5823 - disable nagle in the socknal under 0conf (2578)
5824 - choose mds inode size based on stripe count (2572)
5825 - fix kernel-source rpm problems (2516)
5827 - add --disable-doc to avoid pdf generation (2421)
5828 - update documentation, tests, type-os, comments
5829 - avoid format warnings on ia64
5830 - remove the TOE NAL
5831 - tiny code cleanups by removing unused fields
5833 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
5836 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
5837 - remove the most blatant lies from BUILDING file (2371)
5838 - change default debug level to reasonable production setting
5839 - reduce client side cache size to reduce cache flush time
5840 - reduce max RPCs in flight to avoid unnecessary file fragmentation
5841 - make TCP zerocopy and pinger support enabled by default (2476)
5842 - sync writes completed after process exits caused crashes (2319)
5843 - maintain correct mount count on the MDS (2356)
5844 - backout 1557, because 2316 wasn't really fixed
5845 - better file I/O statistics gathering in /proc
5846 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
5847 - another kernel patch to fix zero-copy TCP function export
5848 - don't take duplicate lock when processing re-sent getattr (2420)
5849 - lctl uses obd_self_export instead of creating new conn (2353)
5850 - MDS/OST recovery case which requires object creation asserted (2425)
5851 - move lfs from /usr/sbin to /usr/bin in packages
5852 - fix race between mds_client_add and mds_client_free (2417)
5853 - use kmalloc instead of slabs in portals (2430)
5854 - don't create duplicate records when a failover MDS is present (2442)
5855 - remove unnecessary mount age check (2332)
5856 - don't remove directory inodes from locks prematurely (2451)
5857 - don't break if MDS service name is the same as hostname (2103)
5858 - fix races in client write RPC generation when cache full (2482)
5860 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
5863 - remove now-unused request->rq_obd (278)
5864 - if an allocation fails, print out how much memory we've used (1933)
5865 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
5866 - add the "configurable stack size" patch to most series files (1256)
5867 - ability to write large log records, for 100+ OST configs (2306)
5868 - fix NULL deref when filter_prep fails (2314)
5869 - fix operator precedence error in filter_sync
5870 - dynamic allocation of socknal TX descriptors (2315)
5871 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
5872 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
5873 - ensure that CWARN messages in Portals always get to the syslog
5874 - __init/__exit are not for prototype decls (ldlm_init/exit)
5875 - x86-64 compile warning fixes
5876 - fix gateway LMC keyword conflict (2318)
5877 - fix MDS lock inversions in getattr/reint paths (1844)
5878 - fix a rare lock re-ordering bug, which caused deadlock (2322)
5879 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
5880 - DLM race condition prevented some lock evictions (2328)
5881 - ENOMEM detection and retry on socknal sends (2230)
5882 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
5883 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
5884 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
5885 - better error messages when a client is rejected during recovery (1505)
5886 - avoid cancelling locks which were never granted, after failure (2330)
5887 - fix i_sem/journal inversion in mds_client_add (2333)
5888 - fix truncate/getattr lock cycle deadlock (2334)
5889 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
5890 - fix two rare exit paths which could leak an l_lock() ref (2321)
5891 - fencepost error in MDS/OST orphan recovery (2226)
5892 - make log record alignment 8 bytes (1988)
5893 - lstripe now fails when requested offset > ost_count (2237)
5894 - ensure that all kernel series have a complete list.h (1607)
5895 - fix crashes in special-file operations (2316)
5896 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
5898 - allow configurable automake binary, for testing new versions
5899 - small update to the lfs documentation
5901 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
5903 * fix negative export reference count in fsfilt_sync (2312)
5905 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
5906 * release candidate 0.9.1
5908 - orphans are moved into the PENDING directory for possible recovery
5909 - replayed opens now open by fid for orphan/rename safety (1042)
5910 - last close of an orphan inode generates a transno (683)
5911 - chdir() and mount() now pin the directory entry (1020)
5912 - avoid CERROR in normal ll_setattr_raw() error case (1500)
5913 - discard very old requests without processing them (1502)
5914 - remove some common, well-understood CERRORs (1505)
5915 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
5916 - clear "grant" flags in OST replies until OST grant code lands (1644)
5917 - fix read performance by not clobbering i_blksize on client (1598)
5918 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
5919 - make LRU size a /proc tunable, clears locks when reduced (707)
5920 - fix some lprocfs rot that prevented ptlbd from loading (1732)
5921 - server locks take references on exports now (1558)
5922 - build fixes for 2.4.20-rh trees (1663)
5923 - return an error from lov_create if all OSCs are inactive (1751)
5924 - fix import levels when a reconnect happens without a timeout (1597)
5925 - exit early from mds_open if we get a lookup error (1749)
5926 - partial page read at EOF wouldn't wait for disk before sending (1642)
5927 - avoid NULL deref in obdfilter when reading page past EOF (1592)
5928 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
5929 - fix LBUG in ll_it_open_error with rc = -2 (1861)
5930 - write/truncate lock inversion (1639)
5931 - Don't auto-load obdclass, portals modules during cleanup (1495)
5932 - fix timestamps from jumping to "now" (1763)
5933 - extra journal assertions (1648)
5934 - add an extra multiunlink test (1771)
5935 - fix read_record/write_record API (1776)
5936 - fix leak of offset_extent, possible incorrect i_size later (1772)
5937 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
5938 - Debugging for the kqswnal_get_idle_tx problems (1820)
5939 - Allow recovery to be attempted multiple times (1536)
5940 - Write out MDS last_rcvd file after it is first created (1600)
5941 - Fix tx_descriptor leak in failed transmit situations (1827)
5942 - ext3 journaling fixes for assertion failure after IO error (1871)
5943 - class_export_put() on freed export after completion AST error (1896)
5944 - Fix revalidate looping in VFS (1322)
5945 - Don't access a freed export during MDS_REINT timeout (1521)
5946 - Add open-unlink recovery support on the MDS (1673,1764)
5947 - Return an error if no MDS data was read from last_rcvd (1946)
5948 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
5949 - Don't setup a disk device that doesn't match exported UUID (317)
5950 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
5951 - avoid committing NULL handle in force close
5952 - local.sh is now a one-stripe LOV configuration
5953 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
5954 - fix ext3 htree duplicate directory entry corruption (1516)
5955 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
5956 - update to new LOV EA format (2097)
5957 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
5959 2003-06-15 Phil Schwan <phil@clusterfs.com>
5962 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
5963 - per-import/export recovery handling (958, 931, 959)
5964 - multiple last-rcvd slots, for serving multiple FSes (949)
5965 - connections are again shared between multiple imp/exports (963, 964)
5966 - "umount -f" would hang if any requests needed to be sent (393, 978)
5967 - avoid pinning large req buffer by copying for queued messages (989)
5968 - add "uuid" to "lctl device" command to help upcalls (991)
5969 - "open" RPCs with transnos would confuse recovery counters (1037)
5970 - do proper endian conversion of all wire messages (288, 340, 891)
5971 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
5972 - call waiting locks callback from LDLM recovery thread (1127, 1151)
5973 - fix ptlrpc_connection leak in target_handle_connect (1174)
5974 - fix import refcounting bug in OST and MDS cleanup (1134)
5975 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
5976 - fix very unlikely obd_types race condition (501)
5977 - remove osc_open hack for echo_client (1187)
5978 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
5979 - a failure in read_inode2 leads to deadlock (1139)
5980 - cancel ack-locks as soon as transaction is committed (1072)
5981 - fix major leaks and crashes in the bulk I/O path (937, 1057)
5982 - make sure to commitrw after any preprw to avoid deadlock (1162)
5983 - failing to execute a file in a lustre FS would lock inode (1203)
5984 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
5985 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
5986 - obd_uuid2tgt would walk off the end of the list (1255)
5987 - on IA64 the osc would give portals incorrect bulk size (1258)
5988 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
5989 - fix lock inversion caused by new llite matching code (1282)
5990 - limit the number of dirty pages on a client to 10MB (1286)
5991 - timed out locks were not being corrected cancelled (1289)
5992 - fix O_DIRECT above 4GB on IA-32 (1292)
5993 * major user-visible changes
5994 - fail out/fail over policy now controlled by the upcall (993)
5996 - add OBD_PING to check server availability and failure (954)
5997 - lustre messages are now sent in sending host order (288, 340, 891)
5998 - add eadatalen to MDS getattr reply (340)
5999 - OST read replies may contain second buffer, with per-page status (593)
6001 2003-03-11 Phil Schwan <phil@clusterfs.com>
6004 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
6005 - failed open()s could cause deadlock; fixed (867, 869)
6006 - stop cancelling OST locks when files are closed (481)
6007 - overlapping XID spaces caused network corruption (851, 853)
6008 - fix unsafe fsfilt counter arithmetic; change to atomic_t
6009 - setattr_raw added, to do single-RPC, server-side setattrs
6010 - lmc/lconf syntax change for OST UUIDs
6011 - fix crashy race condition between ptlrpc_free_req and osc_close
6012 - don't use request in mdc_enqueue if we hit a timeout (889)
6013 - don't set the inode i_size for regular files from the MDS (896)
6014 - handle out of order completion AST (842)
6015 - don't LBUG if a lock request times out after receiving AST (913)
6016 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
6017 - if a bad lock AST arrives, send an error instead of dropping entirely
6018 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
6019 - fix leak in bulk IO when only partially completed (899, 900, 926)
6020 - fix O_DIRECT for ia64 (55)
6021 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
6022 - C-z after timeout could hang a process forever; fixed (977)
6024 - client-side I/O cache (678, 924, 929, 941, 970)
6026 - READPAGE and SETATTRs which don't take server-side locks get
6029 2003-02-11 Phil Schwan <phil@clusterfs.com>
6032 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
6033 - Fix fsfilt_extN_readpage() to read a full page of directory
6034 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
6035 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
6036 - Fix imbalanced LOV object allocation and out-of-bound access (469)
6037 - Most intent operations were removed, in favour of a new RPC mode
6038 that does a single RPC to the server and bypasses most of the VFS
6039 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
6040 - Aggressively cancel local locks on DLM servers
6041 - mds_reint_unlink sends EA to the client if it's the last nlink.
6042 client uses that EA to unlink OST objects.
6043 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
6044 - recursive symlinks were fixed (439)
6045 - fixed NULL deref in DEBUG_REQ
6046 - filter_update_lastobjid no longer calls sync, which annoyed extN
6047 - fixed multi-client small-writes to a single file problem (445)
6048 - fixed mtime updates during file writes (607)
6049 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
6050 - fixed bug in obd_brw_read/write() (under guise of testing 367)
6051 - fixed Linux OST size reporting problem (444, 656)
6052 - OST now updates object mtime with writes or setattr (607, 619)
6053 - client verifies file size before zeroing page past EOF (445)
6054 - OST now writes last allocated objid to disk with allocation (108)
6055 - LOV on echo now works (409)
6057 - mds_reint_unlink sends a new buffer, with the EA included. this
6058 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
6059 set if a regular file was being unlinked, and it was the last link
6060 - use PtlGet from the target for bulk writes (315)
6061 - OST now updates object mtime with writes or setattr (607, 619)
6062 - LDLM now has a grant-time callback to revalidate locked items, if
6064 - Many MDS operations were reorganized to combat race conditions
6066 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
6067 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
6069 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
6072 - Fully reactivate OST imports after reconnection (512, others)
6073 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
6074 - More graceful error handling for truncating on dead OST (515)
6075 - Don't error out unless we're actually accessing dead stripes (474)
6076 - Fix garbage sizes when stripes are missing (410)
6077 - LRU counters were broken, causing constant lock purge (433, 432)
6078 - garbage on read from stripes with failed OSTs (441)
6079 - mark OSCs as active before reconnecting during recovery (438)
6080 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
6081 - lfind did not preserve OST order in output (443)
6082 - symlinks cause hung clients, incorrect data (439)
6083 - stop dereferencing request after dropping refcount (457)
6084 - don't LASSERT(spin_is_locked) on non-SMP (455)
6085 - fixes for many rename() bugs
6086 - fstat didn't correctly synchronize attributes (399)
6087 - server must handle lock cancellation during blocking AST prep (487)
6088 - bulk descriptors were free()d too soon (511)
6089 - fix paths in lconf, which would load incorrect modules (451, 507)
6090 - fix confusing lconf 'host not found' error message (386)
6091 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
6092 - fix race condition in mdc_blocking_ast() for inode access (526)
6093 - fix lov_unpackmd() unpacking wrong number of stripes (537)
6094 - fix lov_set_osc_active() marking wrong OSC inactive (440)
6095 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
6096 - fix multiple writes of stripe MD to MDS (358, maybe 519)
6097 - fix lstripe in several ways (kernel side) (527)
6098 - fix request leak in ldlm_cli_enqueue (262)
6099 - incorrect OSC was marked inactive after OST failure
6100 - call mds_fs_cleanup before unmounting filesystem (524)
6101 - fix races between taking ns_lock and ldlm_lock_change_resource
6102 - fix races updating LOV export open file list
6103 - fix lov_enqueue error path, avoid decref-ing bad lock handle
6104 - fix recovery NULL deref in ldlm_cli_cancel_unused
6105 - fix some DLM races by using new hash table for lock handles (419)
6106 - permit the client to specify desired inodes, at replay
6107 - duplicate requests when we queue them for replay reintegration
6108 - fix last_rcvd offset calculation
6109 - sync after each recovered transaction, so we always make progress
6110 - never, not always, ERESTART requests without transnos
6111 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
6113 - skip replay if the MDS says that the client is already connected
6114 - don't check for a recovery-enabled export to match lctl's UUID
6115 - don't INC_USE_COUNT for phantom exports
6116 - don't crash when cleaning up phantom exports (567)
6117 - don't double-finish or set replay data for errored mdc_open requests
6118 - abort requests when they time out, so we don't get old replies
6119 - send/receive replies for AST messages again
6120 - if the client says that it doesn't have the lock, cancel it on the
6122 - if we timeout during I/O, don't try to cancel an in-use lock; instead
6123 mark it as destroyed, it will all work out when decref is called
6124 - fix module use counts (22, 581)
6126 - ASTs now expect a reply (server cancels lock on error reply)
6128 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
6131 - fix many simultaneous client startup (392)
6132 - fix dentry->d_it clobbering
6133 - credentials weren't being shipped for readdir/getattr operations
6134 - remove invalid assertions triggered during some concurrent MD
6136 - proper Lustre versions added (336, 389)
6137 - fix memory leak for create error case (398)
6138 - fix LOV locking bug that would get cli/srv out of sync
6139 - fix echo client over LOV (409)
6140 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
6141 - fix double-O_EXCL intent crash (424)
6142 - avoid sending multiple lock CANCELs (352)
6144 - MDS can do multi-client recovery (modulo bugs in new code)
6146 - many updates, edits, cleanups
6148 2002-11-18 Phil Schwan <phil@clusterfs.com>
6151 - fix null d_it dereference (346)
6152 - fix full OST/dbench hang (333)
6153 - fix permission problem with file removal (286)
6154 - fix removal of OSCs from LOV when they fail
6155 - fix NULL deref during bulk timeout (214)
6156 - fix problems related to multiple filesystems on one MDS (241)
6157 - fixed serious subtle metadata locking bugs
6158 - free locks on clients when inodes are removed due to memory
6160 - fix inode pointer in lock data (285)
6161 - partial support for multiple MDS on a single host (241)
6162 - data locks weren't cancelled at clear_inode time (290, 311)
6163 - intent locks could lead to unbounded lock growth (205)
6164 - added a maximum lock count, an LRU list, and a flusher
6165 - fix multiple rename (365)
6166 - properly abstracted the echo client
6167 - OSC locked 1 byte too many; fixed
6168 - rewrote brw callback code:
6169 - fixed recovery bugs related to LOVs (306)
6170 - fixed too-many-pages-in-one-write crash (191)
6171 - fixed (again) crash in sync_io_timeout (214)
6172 - probably fixed callback-related race (385)
6174 - Add capability to MDS protocol
6175 - LDLM cancellations and callbacks on different portals
6177 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
6180 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
6181 - defer open object destruction to close time (601981, 138)
6182 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
6183 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
6184 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
6185 - don't call obd_setattr in ll_file_release for destroyed objects
6186 * protocol change to lustre_msg: move |version| and add |flags|
6187 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
6188 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
6189 * added replay of create, unlink, link and rename operations during
6190 MDS failover; recovery should be much more robust now
6191 * remove failed OSCs from LOVs (only lov_create uses this so far)
6192 * the lustre-HOWTO was brought (more) up to date (582544)
6194 2002-10-23 Phil Schwan <phil@clusterfs.com>
6197 - in-use dentries weren't being reused properly (617851)
6198 - prevent multiple LDLM setup (599178)
6199 - fix LOV size calculations for truncate (617853)
6200 - fix client handling of MDS intent errors (POSIX)
6201 - fix permission bug in lovstripe.c test (624321)
6202 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
6203 - truncate past end of file could corrupt data
6204 - proper cleanup after timeouts, crashes, etc (592524, 550815)
6205 - a race in recovery could return ETIMEDOUT to apps (623947)
6206 - building outside the source directory was fixed
6207 * the lustre-HOWTO was brought (more) up to date (582544)
6208 * major progress was made on recovery functionality
6210 2002-10-10 Phil Schwan <phil@clusterfs.com>
6213 - recovery deadlock fix
6214 - rm -rf causes LBUG fix (617817)
6215 - file open by multiple tasks fix (618962)
6216 - directory permissions bugs (602707 and 620007)
6217 - journal_stop fixed with locking (611313)
6218 - O_APPEND failures resolved (618273, perhaps 614459)
6219 - lconf PATH fix (619770)
6220 - IA64 build fix (621450)
6221 - RPC buffer sizes scale with amount of memory
6223 2002-10-01 Phil Schwan <phil@clusterfs.com>
6226 - locks would be cancelled without throwing away data pages,
6227 resulting in inconsistent data (605627)
6228 - inode attributes were not always being refreshed (605627, 612449)
6229 - lconf now continues to cleanup after lctl reports an error
6230 - MDS now enforces user permissions (602707)
6231 - lprocfs cleanup fixed, but not yet enabled (614157)
6232 - fixed infinite server hang, should a client not respond to an AST
6233 - avoid going into recovery if user calls readlink() with a buffer
6234 that's too small (613941)
6235 - AST RPCs no longer require replies (614867) -- this may be changed
6236 - don't crash server if client sends an IOV that's too big (611336)
6237 - fixed lock conversion deadlock (611892)
6238 - fixed the following of symlinks (614622)
6239 * recovery: the server can remove locks from a client that dies, other
6240 clients can make progress
6241 * more extN patch fixes
6242 * compile-time configurable ptlrpc buffer allocations
6244 - collaborative read cache document
6245 - Lustre Lite Performance CDR document-in-progress
6247 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6250 - fix typo in patch-2.4.18
6252 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6255 - clear ptlrpc request each time in handle_incoming_request()
6256 - unlink of files now destroys the object on the OST
6258 2002-09-19 Peter Braam <braam@clusterfs.com>
6260 * add hard link support
6261 * change obdfile creation method
6262 * kernel patch changed
6264 2002-09-19 Peter Braam <braam@clusterfs.com>
6267 - stack overflow bug in extN fixed
6269 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
6271 * documentation updates
6272 - add man pages for config tools
6273 - update tests/README to describe testing with new config tools
6274 - finish metadata API descriptions
6275 * bug fixes and cleanups
6276 - statfs workaround for 16TB limit
6277 - LOV stripe allocation improved, can stripe on subset of OSTs
6278 - LOV file size/IO offset was wrong for files > 4GB in size
6279 - object EA data was being dropped, caused files to be unreadable
6280 - memory overflow with non-LOV OST caused memory corruption
6281 - fixed regression tests to work with new config tools, obdfilter
6282 - fixed bug when directory size became larger than 1 block
6283 - fixed bug (for single client case) when PWD was deleted
6284 - invalidate local directory pages when doing intent-based ops
6285 - avoid LDLM oops when lock callback contained bad data
6287 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
6289 * documentation updates
6290 * bug fixes and cleanups
6291 - configuration tools
6294 - 64-bit compile warnings
6295 - 64-bit internal statfs data
6297 * test_brw on persistent OST devices
6299 * lprocfs (disabled)
6301 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
6303 * documentation updates
6304 * bug fixes and cleanups
6305 * configuration tools
6307 2002-08-30 Peter J. Braam <braam@clusterfs.com>
6310 * many small fixes to 0_5_4
6311 * io/network handling
6312 * thinkos in MDS operations
6314 2002-08-24 Peter J. Braam <braam@clusterfs.com>
6317 * crucial basic fixes to 0.5.3
6318 * IOR, Iozone work over Elan
6321 2002-08-07 Phil Schwan <phil@clusterfs.com>
6322 * version 0_5_3, our first alpha
6323 * we use the new Portals iovs
6324 * documentation updates
6325 * bug fixes and cleanups
6326 * small changes in the DLM wire protocol
6328 2002-07-25 Peter J. Braam <braam@clusterfs.com>
6329 * version 0_5_1 with some initial stability,
6330 * locking on MD and file I/O.
6331 * documentation updates
6332 * several bug fixes since 0.5.0
6333 * small changes in wire protocol
6335 2002-07-18 Phil Schwan <phil@clusterfs.com>
6337 * delivered as Lustre Light Alpha
6338 * fixed a crash after handling invalid MDS requests
6339 * fixed directory pages for architectures with non-4k pages sizes
6341 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
6342 * release version v0_4_4
6343 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
6345 * Fixes a number of interruption problems with OST operations.
6346 * Update documentation for portals header changes
6347 * Move all wire protocol structs/defines to lustre_idl.h
6348 * Fixes symlink length bug.
6349 * Add tcpdump to repository.
6351 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
6352 * release version v0_4_3
6353 * Fixes statfs for inodes on extN.
6354 * Fixes bug in runtests which would delete /etc/hosts.
6355 * Use 64-bit object IDs wherever possible (not into VFS though)
6356 Remove ost_get_info, which is unused by lustre, and out of date.
6358 2002-07-03 Peter Braam <braam@clusterfs.com>
6359 * release version v0_4_2 Fixes a lookup error (type not passed)
6360 * move forward to head of Portals
6361 * move forward to latest Lustre kernel
6363 2002-06-25 Peter Braam <braam@clusterfs.com>
6364 * release version v0_4_1. Hopefully stable on single node use.
git://git.whamcloud.com / fs / lustre-release.git / blob