3 * See http://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
4 * Server support for kernels:
5 2.6.32-220.4.2.el6 (RHEL6)
6 * Client support for unpatched kernels:
7 2.6.32-220.4.2.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 --------------------------------------------------------------------------------
20 04-30-2012 Whamcloud, Inc.
22 * See http://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
23 * Server support for kernels:
24 2.6.32-220.4.2.el6 (RHEL6)
25 * Client support for unpatched kernels:
26 2.6.32-220.4.2.el6 (RHEL6)
28 2.6.18-238.19.1.el5 (RHEL5)
29 2.6.16.60-0.69.1 (SLES10)
30 * Recommended e2fsprogs version: 1.41.90.wc4
31 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
32 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
33 stack overflow. For more information, please refer to bugzilla 17630.
34 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
35 the re-exporting nodes to be mounted with "32bitapi" mount option
37 Severity : enhancement
39 Description: Update RHEL6.2 kernel to 2.6.32-220.4.2.el6
41 Severity : enhancement
43 Description: Update RHEL6.2 kernel to 2.6.32-220.4.1.el6
45 Severity : enhancement
47 Description: Update RHEL6.2 kernel to 2.6.32-220.el6
49 Severity : enhancement
51 Description: Update RHEL6.1 kernel to 2.6.32-131.17.1.el6
53 --------------------------------------------------------------------------------
55 09-30-2011 Whamcloud, Inc.
57 * Server support for kernels:
58 2.6.18-238.19.1.el5 (RHEL5)
59 2.6.32-131.6.1.el6 (RHEL6)
60 * Client support for unpatched kernels:
61 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
62 2.6.18-238.19.1.el5 (RHEL5)
63 2.6.32-131.6.1.el6 (RHEL6)
64 2.6.29.4-167.fc11 (FC11)
65 2.6.16.60-0.69.1 (SLES10)
67 * Recommended e2fsprogs version: 1.41.90.wc3
68 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
69 Lustre filesystem with 4K stack may cause a stack overflow. For more
70 information, please refer to bugzilla 17630.
71 * NFSv4 reexport requires Lustre client on the re-exporting nodes to be
72 mounted with "32bitapi" mount option
74 Severity : enhancement
75 Description: Update RHEL6.1 kernel to 2.6.32-131.6.1.el6
77 Severity : enhancement
78 Description: Update RHEL5.6 kernel to 2.6.18-238.19.1.el5
80 Severity : enhancement
81 Description: Update RHEL5.6 kernel to 2.6.18-238.9.1.el5
84 Frequency : only with SLES10
86 Description: Use OFED "KMP" provided by Novell
87 Details : SLES10 SP3 ships with OFED in a separate "KMP" package.
88 Lustre is now built against this package. That means you need to
89 install the ofed-kmp package from Novell for the patchless client.
90 Note that the ofed-kmp that Novell ships may not exactly match the
91 kernel version but should still be compatible.
93 Severity : enhancement
95 Description: Update RHEL5.5 kernel to 2.6.18-194.17.1.el5
96 Update OEL5.5 kernel to 2.6.18-194.17.1.0.1.el5.
98 --------------------------------------------------------------------------------
100 2010-07-15 Oracle, Inc.
102 * Server support for kernels:
103 2.6.18-164.11.1.el5 (RHEL 5)
104 2.6.18-164.11.1.0.1.el5 (OEL 5)
105 * Client support for unpatched kernels:
106 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
107 2.6.18-164.11.1.el5 (RHEL 5),
108 2.6.18-164.11.1.0.1.el5 (OEL 5)
109 2.6.16.60-0.42.8 (SLES 10),
111 2.6.29.4-167.fc11 (FC11)
112 * Recommended e2fsprogs version: 1.41.10-sun2
114 Severity : enhancement
116 Description: Distribute lfs_migrate script with lustre
117 Details : The lfs_migrate mentioned in the manual is now included in the
118 lustre tree in lustre/scripts and will distributed in lustre RPMs.
120 Severity : enhancement
122 Description: Failover nid list added to import proc file.
124 Severity : enhancement
126 Description: "som_preview" mount option required to use SOM
127 Details : The clients will need to use som_preview mount option
131 Description: Lustre does not do 1MB IOs to HW RAID
132 Details : Bump MAX_PHYS/HW_SEGMENTS and SG_ALL to 256 in the RHEL5 kernel.
134 Severity : enhancement
136 Description: lfs quota output is not very convenient for awk/sed-parsing
137 Details : Some positions in lfs quota output table could be empty or
138 non-empty which made it hard to parse it with scripts, now
139 a dash is put instead of space where there is not supposed
144 Description: Update to OFED 1.5.1
145 Details : OFED 1.5.1 will be used in environments that don't use
146 vendor supplied OFED.
150 Description: Remove lnet. prefix from lctl params display
151 Details : Also renamed lnet.memused to lnet_memused to
156 Description: Use conf_param -d to remove permanant settings.
160 Description: Add list_param -R
161 Details : lctl list_param now supports recursive mode.
164 Bugzilla : 21829, 22312
165 Description: Fix llobdstat.
166 Details : 1. Now updating the write_bytes counter
167 2. Port script fixes from 1.8.
171 Description: decrease the usage of memory on clients.
172 Details : 1. On clients, recycle dentries and inodes unused.
173 2. Delete the code related to ll_deathrow(att 6215 in bug 1443). It
178 Description: llobdstat fix and enhancement
179 Details : add a counter to set a limit to how many samples will be returned
180 fix a wildcard in the path to limit to obdfilter stats only
184 Description: lfs check only allowed for root
185 Details : Code cleanups and added a regresion test of non-root use of
190 Description: MDS Oopses in ptlrpc_dispatch_difficult_reply()/__list_add()
191 Details : Fix hr->hr_index overload.
195 Description: Fix FIEMAP functionality.
199 Description: FC11 patchless client support added
203 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and
204 OEL5.4 kernel to 2.6.18-164.11.1.0.1.el5.
208 Description: New acc-sm script and various test-framework enhancments.
212 Description: Previously landed patch was reverted.
216 Description: Update SLES10 kernel to 2.6.16.60-0.42.8 (patchless)
220 Description: Add ldiskfs maxdirsize mount option.
224 Description: Fix mkfs.lustre support for 16TB LUNs.
228 Description: Print mmp_check_interval
229 Details : print mmp_check_interval and make it possible to abort
230 mount operation in case it takes too long.
234 Description: Add cascading_rw to lustre mpi tests.
238 Description: Store SOM data in EA.
239 Details : Also several SOM related cleanups.
243 Description: Add refcount to md_open_data
247 Description: Use refcount on lov_request_set to prevent memory corruption.
251 Description: Statahead: use parent dir's i_mutex for synchronization.
255 Description: loadgen improvements.
259 Description: Enable peer health detection in ksocklnd and ko2iblnd.
263 Description: Control DCACHE_LUSTRE_INVALID flag with MDS_INODE_LOOKCUP lock
267 Description: Fix race on obd_devcie:md_stats
271 Description: Fix SLES 11 client builds.
272 Details : Resolved uninitalized pointer gcc warnings.
276 Description: Allow non-root access for "lfs check".
277 Details : Added a check in obd_class_ioctl() for OBD_IOC_PING_TARGET.
281 Description: Clear imp_force_reconnect correctly in ptlrpc_connect_interpret()
283 Severity : enhancement
285 Description: Update kernel to RHEL5.4 2.6.18-164.6.1.el5 and
286 OEL5 2.6.18-164.6.1.0.1.el5(Both in-kernel OFED enabled).
290 Description: Sanity test_27p fails.
291 Details : OBD_FAIL_OST_ENOSPC has been moved to filter_statfs().
292 Sanity tests 27 were cleaned up and improved.
296 Description: Unify req format on client/servers
297 Details : Use new req_capsule API [almost] everywhere instead of old PTLRPC
298 buffers and swabbers approach..
303 Description: drop join file stuff.
305 Severity : enhancement
307 Description: Add 2.6.27 and SLES11 (patchless client) support
311 Description: Fix in ptlrpc_expire_one_request() to print the signed time
314 Severity : enhancement
316 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
319 Severity : enhancement
321 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
323 Severity : enhancement
325 Description: correctly handle big reply message.
326 Details : send LNet event if reply is bigger then buffer and adjust this buffer
331 Description: Drop unnecessary __GFP_NOMEMALLOC flag from filter_get_page()
333 Severity : enhancement
335 Description: Add support for OEL5.
339 Description: Conf-sanity.sh 50g test - deactivated OST should not cause a panic.
341 Severity : enhancement
343 Description: "weak-modules" support
344 Details : Implement "weak-modules" support which enables kernel modules
345 to be used with any kernel that implements the same kABI. In
346 order to achieve this modules are now installed in
347 /lib/modules/$(uname -r)/updates/kernel on all distributions.
351 Description: client could not reconnect to OST because of an active request.
352 Details : The client now retries to reconnect to the same server, if a
353 connect request failed with EBUSY or -EAGAIN. Test case 35b
354 conf-sanity.sh has been created.
358 Description: Don't allow make backward step on assignin osc next id.
359 Details : race between allocation next id and ll_sync thread can be cause
360 of set wrong osc next id and can be kill valid ost objects.
364 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
368 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
372 Description: Error handling in osc_statfs_interpret() has been improved.
373 Details : Check in osc_statfs_interpret() for EBADR.
377 Description: Do not send statfs() requests to OSTs disabled by administrator.
378 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
380 Severity : enhancement
382 Description: Added sanity-quota test to for lfs(1) to work with numeric uid/gid.
384 Severity : enhancement
386 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
388 Severity : enhancement
390 Description: Enhancement for lfs(1) command to use numeric uid/gid.
394 Description: Improved cleanup procedure for conf-sanity test_46a.
398 Description: Do not start bulk transfer, if client tried to reconnect.
399 Details : Do not start bulk transfer, if client tried to reconnect,
400 but failed due to active rpcs. Check for exp_abort_active_req
401 has been implemented.
405 Description: Prevent inconsistences between linux and lustre mount structures.
406 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
407 Make the sleep interruptible.
409 Severity : enhancement
411 Description: provide server to client comms path
412 Details : server to client communications path via new LDLM_SET_INFO rpc
416 Description: truncate starts GFP_FS allocation under transaction causing deadlock
417 Details : ldiskfs_truncate calls grab_cache_page which may start page
418 allocation under an open transaction. This may lead to
419 calling prune_icache with consequent lustre reentrance.
423 Description: send correctly lsm on open replay
424 Details : MDS is trust to LSM size on replay open, but client can set wrong size
429 Description: enable client interface failover
430 Details : When a child reconnects from another NID, properly update export
431 nid hash position and ldlm reverse import.
433 Severity : enhancement
435 Description: Communicate OST degraded/readonly state via statfs to MDS
436 Details : Flags in the statfs returned from OSTs indicate whether the
437 OST is in a degraded RAID state, or if the filesystem has
438 turned read-only after a filesystem error is detected.
443 Description: don't panic if EPROTO was hit when reading symlink
444 Details : correctly handling request reference in error cases.
446 Severity : enhancement
448 Description: Add LustreNetLink, a kernel-userspace communcation path.
450 Severity : enhancement
452 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
457 Description: access to llog context before init.
458 Details : move handling CATALOGS file at osc layer and forbid access to llog
462 Frequency : rare, only when client is writing to OST on local node
464 Description: Avoid deadlock for local client writes
465 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
466 memory freeing context. This allows OST threads to set the
467 PF_MEMALLOC flag on task structures in order to allocate memory
468 from reserved pools and complete IO.
469 Use GFP_HIGHUSER for OST allocations for non-local client writes,
470 so that the OST threads generate memory pressure and allow
471 inactive pages to be reclaimed.
473 Severity : enhancement
475 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
480 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
481 Details : move ldlm namespace creation in setup phase to avoid grab
482 _lprocfs_lock with cli_sem held.
486 Description: Temporarily disable grant shrink.
487 Details : Disable the feature for debugging.
490 Frequency : only when formatting test filesystems
492 Description: Unable to run several mkfs.lustre on loop devices at the same time
493 Details : mkfs.lustre returns error 256 on the concurrent loop devices
494 formatting. The solution is to proper handle the error.
496 Severity : enhancement
498 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
500 Severity : enhancement
502 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
507 Description: resolve race between obd_disconnect and class_disconnect_exports
508 Details : if obd_disconnect will be called to already disconnected export he
509 forget release one reference and osc module can't unloaded.
511 Severity : enhancement
513 Description: Allow tuning service thread via /proc
514 Details : For each service a new
515 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
516 created that can be used to set min/max thread counts, and get the
517 current number of running threads.
522 Descriptoin: don't return error if have partially created objects for file.
523 Details : lov_update_create_set uses set->set_success as index for created
524 objects, so if some requests will be failed, they will have hole
525 at end of array and qos_shrink_lsm for allocate correct lsm.
527 Severity : enhancement
529 Description: Update OFED support to 1.4.1
534 Description: Reduce small size read RPC
535 Details : Set read-ahead limit for every file and only do read-ahead when
536 available read-ahead pages are bigger than 1M to avoid small size
539 Severity : enhancement
541 Description: move AT tunable parameters for more consistent usage
542 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
544 Severity : enhancement
546 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
547 Details : allow skip disconnected ost for send statfs request and hide error
553 Description: fix racy locking of mballoc block bitmaps causing BUG
554 Details : The locking of the mballoc buddy bitmap and the in-memory
555 block bitmap was using two different spin locks in some
556 cases. This made it possible to incorrectly access the
557 mballoc bitmap while another process was modifying it,
558 causing a sanity assertion to fail. While no on-disk corruption
559 was reported, there was some risk of this happening.
562 Frequency : rare, on llog test 6
564 Descriptoin: don't allow connect to already connected import
565 Details : allowing connect to already connected import is hide connecting problem.
570 Descriptoin: allow kill process which wait statahead result
571 Details : for some reasons 'ls' can stick in waiting result from statahead,
572 in this case need way for kill this process.
574 Severity : enhancement
576 Description: Add state history info file, enhance import info file
577 Details : Track import connection state changes in a new osc/mdc proc file;
578 add overview-type data to the osc/mdc import proc file.
580 Severity : enhancement
582 Description: MDS create should not wait for statfs RPC while holding DLM lock.
584 Severity : enhancement
586 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
591 Descriptoin: OST grant too much space to client even there are not enough space.
592 Details : Client will shrink its grant cache to OST if there are no write
593 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
594 this grant cache if there are already not enough avaible space
595 (left_space < total_clients * 32M).
598 Frequency : start MDS on uncleanly shutdowned MDS device
600 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
601 Details : stay in waiting mds<>ost recovery finished produce random bugs
602 due race between two ll_sync thread for one lov target. send
603 ACTIVATE event only if connect realy finished and import have
607 Frequency : rare, connect and disconnect target at same time
609 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
610 Details : don't call obd_disconnect under lov_lock. this long time
611 operation and can block ptlrpcd which answer to connect request.
616 Descriptoin: don't lose wakeup for imp_recovery_waitq
617 Details : recover_import_no_retry or invalidate_import and import_close can
618 both sleep on imp_recovery_waitq, but we was send only one wakeup
622 Frequency : always with long access acl
624 Descriptoin: mds can't pack reply with long acl.
625 Details : mds don't control size of acl but they limited by reint/getattr
628 Severity : enhancement
630 Description: Update to SLES10 kernel-2.6.16.60-0.33.
632 Severity : enhancement
634 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
637 Frequency : start MDS on uncleanly shutdowned MDS device
639 Descriptoin: aborting recovery hang on MDS
640 Details : don't throttle destroy RPCs for the MDT.
643 Frequency : on remount
645 Description: external journal device not working after the remount
646 Details : clear dev_rdonly flag for external journal devices in
652 Description: shutdown vs evict race
653 Details : client_disconnect_export vs connect request race.
654 if client will evicted at this time - we start invalidate
655 thread without referece to import and import can be freed
659 Frequency : rare, need acl's on inode.
661 Description: client can't handle ost additional correctly
662 Details : if ost was added after client connected to mds client can have
663 hit lnet_try_match_md ... to big messages to wide striped files.
664 in this case need teach client to handle config events about add
665 lov target and update client max ea size at that event.
667 Severity : enhancement
669 Description: Changelogs
670 Details : Changelogs are a lightweight mechanism to track filesystem
671 metadata and namespace changes. The changelog is recorded
672 permanently on the MDTs, and is periodically "consumed" / purged
673 when records are no longer needed.
675 Severity : enhancement
677 Description: compact fld format with extents
678 Details : Store range of seq rather than every seq in FLD. Seq
679 controller update FLD rather than clients. In Case of CMD, mdt0
680 has FLD, all other metadata server act as non persistent proxy
681 for FLD queries and cache fld entries in fld cache.
686 Description: don't skip ost target if they assigned to file
687 Details : Drop slow OSCs if we can, but not for requested start idx.
688 This means "if OSC is slow and it is not the requested
689 start OST, then it can be skipped, otherwise skip it only
690 if it is inactive/recovering/out-of-space.
694 Description: more cleanup in mds_lov
695 Details : not send LOV EA under replay, we can't know about they size at this
696 time. Don't allow client connect to mds before any ost connected,
697 for avoid problems with LOV EA size and returning EIO to client.
699 Severity : enhancement
701 Description: Interoperability at server side (Disk interoperability)
703 Severity : enhancement
705 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
707 Severity : enhancement
709 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
711 Severity : enhancement
713 Description: New client IO stack (CLIO).
715 Severity : enhancement
717 Description: Commit on sharing. Eliminate inter-client dependencies between
718 uncommitted transactions by doing transaction commits.
719 Thereby clients may recovery independently.
722 Frequency : Create a symlink file with a very long name
724 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
725 Details : If there is no extra space in the request for early cancels,
726 ldlm_req_handles_avail() returns 0 instead of a negative value.
728 Severity : enhancement
730 Description: Add /proc entry for import status
731 Details : The mdc, osc, and mgc import directories now have
732 an import directory that contains useful import data for debugging
735 Severity : enhancement
737 Description: Re-disable certain /proc logging
738 Details : Enable and disable client's offset_stats, extents_stats and
739 extents_stats_per_process stats logging on the fly.
742 Frequency : Only on FC kernels 2.6.22+
744 Description: oops in statahead
745 Details : Do not drop reference count for the dentry from VFS when lookup,
746 VFS will do that by itself.
748 Severity : enhancement
750 Description: Generic /proc file permissions
751 Details : Set /Proc file permissions in a more generic way to enable non-
752 root users operate on some /proc files.
756 Description: Hitting mdc_commit_close() ASSERTION
757 Details : Properly handle request reference release in
758 ll_release_openhandle().
762 Description: quota recovery deadlock during mds failover
763 Details : This patch includes att18982, att18236, att18237 in bz14840.
765 1. fix osts hang when mds does failover with quotaon
766 2. prevent watchdog storm when osts threads wait for the
771 Frequency : only patchless client
772 Description: add workaround for race between add/remove dentry from hash
774 Severity : enhancement
776 Description: Allow OST glimpses to return PW locks
780 Description: LBUG when llog conf file is full
781 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
785 Description: Prevent import from entering FULL state when server in recovery
789 Description: service mount cannot take device name with ":"
790 Details : Only when device name contains ":/" will mount treat it as
796 Description: replace ptlrpcd with the statahead thread to interpret the async
797 statahead RPC callback
801 Frequency : on recovery
802 Description: I/O failures after umount during fail back
803 Details : if client reconnected to restarted server we need join to recovery
804 instead of find server handler is changed and process self
805 eviction with cancel all locks.
807 Severity : enhancement
809 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
811 Severity : enhancement
813 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
815 Severity : enhancement
817 Description: Upcall on Lustre log has been dumped
818 Details : Allow for a user mode script to be called once a Lustre log has
819 been dumped. It passes the filename of the dumped log to the
820 script, the location of the script can be specified via
821 /proc/sys/lnet/debug_log_upcall.
826 Description: avoid idr_remove called for id which is not allocated.
827 Details : Move assigment s_dev for clustered nfs to end of initialization,
828 to avoid problem with error handling.
833 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
834 Details : When connection is reused this not moved from CONN_UNUSED_HASH
835 into CONN_USED_HASH and this prodice warning when put connection
836 again in unused hash.
839 Severity : enhancement
841 Description: File striping can now be set to use an arbitrary pool of OSTs.
843 Severity : enhancement
845 Description: Export bytes_read/bytes_write count on OSC/OST.
849 Description: Early reply size mismatch, MGC loses connection
850 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
851 the connect flags are properly negotiated.
856 Description: quotas are not honored with O_DIRECT
857 Details : all writes with the flag O_DIRECT will use grants which leads to
858 this problem. Now using OBD_BRW_SYNC to guard this.
862 Description: add quota statistics
863 Details : 1. sort out quota proc entries and proc code.
864 2. add quota statistics
866 Severity : enhancement
868 Description: enable quota support for HEAD.
872 Description: Properly propagate oinfo flags from lov to osc for statfs
873 Details : restore missing copy oi_flags to lov requests.
875 Severity : enhancement
877 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
878 : l_getgroups(8), lst(8), routerstat(8)
879 Details : included man pages for llobdstat(8), llstat(8),
880 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
882 Severity : enhancement
884 Description: configure's --enable-quota should check the
885 : kernel .config for CONFIG_QUOTA
886 Details : configure is terminated if --enable-quota is passed but
887 : no quota support is in kernel
891 Description: Remove portals compatibility
892 Details : Remove portals compatibility, not interoperable with releases
897 Description: Resolve device initialization race
898 Details : Prevent proc handler from accessing devices added to the
899 obd_devs array but yet be intialized.
901 Severity : enhancement
903 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
905 Severity : enhancement
907 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
912 Description: Using wrong pointer in osc_brw_prep_request
913 Details : Access to array[-1] can produce panic if kernel compiled with
914 CONFIG_PAGE_ALLOC enabled
918 Description: Client runs out of low memory
919 Details : Consider only lowmem when counting initial number of llap pages
923 Description: *optional* service tags registration
924 Details : if the "service tags" package is installed on a Lustre node
925 When the filesystem is mounted, a local-node service tag will
926 be created. See http://inventory.sun.com/ for more information
927 about the Service Tags asset management system.
931 Description: Kernel BUG tries to release flock
932 Details : Lustre does not destroy flock lock before last reference goes
933 away. So always drop flock locks when client is evicted and
934 perform unlock regardless of successfulness of speaking to MDS.
938 Description: add refcount protection for osc callbacks, avoid panic on shutdown
942 Description: sanity test 65a fails if stripecount of -1 is set
943 Details : handle -1 striping on filesystem in ll_dirstripe_verify
948 Description: ASSERTION(CheckWriteback(page,cmd)) failed
949 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
952 Severity : enhancement
954 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
958 Description: do not process already freed flock
959 Details : flock can possibly be freed by another thread before it reaches
960 to ldlm_flock_completion_ast.
964 Description: LBUG during stress test
965 Details : Need properly lock accesses the flock deadlock detection list.
969 Description: oops in page fault handler
970 Details : kernel page fault handler can return two special 'pages' in error
971 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
975 Description: timeout with invalidate import.
976 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
977 should be handled by ptlrpcd. This produce long age waiting and
978 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
980 Severity : enhancement
982 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
986 Description: enable MGS and MDT services start separately
987 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
988 a co-located MGS without starting the MGS, which is a complement
989 to 'nosvc' mount option.
993 Frequency : after recovery
994 Description: precreate to many object's after del orphan.
995 Details : del orphan st in oscc last_id == next_id and this triger growing
996 count of precreated objects. Set flag LOW to skip increase count
997 of precreated objects.
1001 Frequency : rare, on clear nid stats
1002 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
1003 Details : when clean nid stats sometimes try destroy live entry,
1004 and this produce panic in free.
1008 Description: Stack overflow during MDS log replay
1009 ease stack pressure by using a thread dealing llog_process.
1013 Description: wait until IO finished before start new when do lock cancel.
1014 Details : VM protocol want old IO finished before start new, in this case
1015 need wait until PG_writeback is cleared until check dirty flag and
1016 call writepages in lock cancel callback.
1018 Severity : enhancement
1020 Description: using special macro for print time and cleanup in includes.
1024 Description: mds_mfd_close() ASSERTION(rc == 0)
1025 Details : In mds_mfd_close(), we need protect inode's writecount change
1026 within its orphan write semaphore to prevent possible races.
1030 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
1034 Frequency : rare, on shutdown ost
1035 Description: don't hit live lock with umount ost.
1036 Details : shrink_dcache_parent can be in long loop with destroy dentries,
1037 use shrink_dcache_sb instead.
1041 Description: don't panic with use echo client
1042 Details : echo client pass NULL as client nid pointer and this produce null
1043 pointer dereference.
1047 Description: fix build on ppc32
1048 Details : compile code with -m64 flag produce wrong object file for ppc32.
1052 Description: add message levels for liblustreapi
1056 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
1057 Details : Move the 'good_osts' check before the 'total_bavail' check. This
1058 will result in an -EAGAIN and in the exit call path we call
1059 alloc_rr() which will with increasing aggressiveness attempt to
1060 aquire precreated objects on the minimum number of required OSCs.
1064 Description: Use old size assignment to avoid deadlock
1065 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
1066 the scheduling while holding a spinlock. We do not need locking
1067 for size in ll_update_inode() because size is only updated from
1068 the MDS for directories or files without objects, so there is no
1069 other place to do the update, and concurrent access to such inodes
1070 are protected by the inode lock.
1074 Description: resolve "_IOWR redefined" build error on SLES10
1078 Description: dump the memory debugging after all modules are unloaded to
1079 suppress false negative in conf_sanity test 39
1081 Severity : enhancement
1083 Description: build kernel-ib packages for OFED 1.3 in our release cycle
1088 Description: fix SLES kernel versioning
1089 Details : the kernel version for our SLES 10 kernel did not include a "-"
1090 before the "smp" at the end. while this was not a problem in
1091 general, it did mean that software trying to use the kernel
1092 version to try to detect a vendor specific kernel would fail.
1093 this was most evident by the OFED build scripts.
1097 Description: Don't update lov_desc members until making sure they are valid
1098 Details : When updating lov_desc members via proc fs, need fix their
1099 validities before doing the real update.
1103 Description: don't put request into delay list while invalidate in flight.
1104 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
1105 import in flight. this produce timeout for invalidate and sometimes
1106 can cause stale data.
1110 Frequency : on ppc only
1111 Description: not convert ost objects for directory because it's not exist.
1112 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
1116 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
1118 Severity : enhancement
1120 Description: Update to support 2.6.22.14 vanilla kernel.
1124 Frequency : rare, on recovery
1125 Description: read procfs can produce deadlock in some situation
1126 Details : Holding lprocfs lock which send rpc can produce block for destroy
1127 obd objects and this also block reconnect with -EALREADY. This
1128 isn't fix all lprocfs bugs - but make it rare.
1130 Severity : enhancement
1132 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
1135 Frequency : frequent on X2 node
1137 Description: mdc_set_open_replay_data LBUG
1138 Details : Set replay data for requests that are eligible for replay.
1142 Description: lustre_mgs: operation 101 on unconnected MGS
1143 Details : When MGC is disconnected from MGS long enough, MGS will evict the
1144 MGC, and late on MGC cannot successfully connect to MGS and a lot
1145 of the error messages complaining that MGS is not connected.
1149 Frequency : on network error
1150 Description: panic with double free request if network error
1151 Details : mdc_finish_enqueue is finish request if any network error ocuring,
1152 but it's true only for synchronus enqueue, for async enqueue
1153 (via ptlrpcd) this incorrect and ptlrpcd want finish request
1156 Severity : enhancement
1158 Description: client-side metadata stat-ahead during readdir(directory readahead)
1159 Details : perform client-side metadata stat-ahead when the client detects
1160 readdir and sequential stat of dir entries therein
1163 Frequency : on start mds
1165 Description: Implement get_info(last_id) in obdfilter.
1168 Frequency : occasional
1170 Description: Correctly check stale fid, not start epoch if ost not support SOM
1171 Details : open with flag O_CREATE need set old fid in op_fid3 because
1172 op_fid2 was overwritten with new generated fid, but mds can answer
1173 with one of these two fids and both is not stale. Setattr
1174 incorrectly started an epoch and assume will be called
1175 done_writing, but without SOM done_writing ever being called.
1178 Frequency : rare, depends on device drivers and load
1180 Description: MDS or OSS nodes crash due to stack overflow
1181 Details : Code changes in 1.8.0 increased the stack usage of some functions.
1182 In some cases, in conjunction with device drivers that use a lot
1183 of stack the MDS (or possibly OSS) service threads could overflow
1184 the stack. One change which was identified to consume additional
1185 stack has been reworked to avoid the extra stack usage.
1188 Frequency : occasional
1190 Description: Do not fail import if osc_interpret_create gets -EAGAIN
1191 Details : If osc_interpret_create got -EAGAIN it immediately exits and
1192 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
1193 oscc_has_objects and see OSC has no objests and call
1194 oscc_internal_create to resend create request.
1196 Severity : enhancement
1198 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
1200 Severity : enhancement
1202 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
1205 Frequency : very rare
1207 Description: Fix replay if there is an un-replied request and open
1208 Details : In some cases, older replay request will revert the
1209 mcd->mcd_last_xid on MDS which is used to record the client's
1210 latest sent request.
1212 Severity : enhancement
1214 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
1216 Severity : enhancement
1218 Description: Add rhel5 support to HEAD.
1220 Serverity : enhancement
1222 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
1227 Description: Don't allow skipping OSTs if index has been specified.
1228 Details : Don't allow skipping OSTs if index has been specified, make
1229 locking in internal create lots better.
1233 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
1234 Details : access to bitfield in structure is always rounded to long
1235 and this produce problem with not atomic change any bit.
1239 Description: Lustre make rpms failed.
1240 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
1241 builds Lustre rpms from tarball.
1244 Frequency : rare on shutdown ost
1246 Description: If llog cancel was not send before clean_exports phase, this can
1247 produce deadlock in llog code.
1248 Details : If llog thread has last reference to obd and call class_import_put
1249 this produce deadlock because llog_cleanup_commit_master wait when
1250 last llog_commit_thread exited, but this never success because was
1251 called from llog_commit_thread.
1255 Description: allow userland application know is lost one of stripes.
1256 Details : fill lvb_blocks with error code on ost and return it to
1257 application if error flag found.
1261 Description: NULL lov_tgts causing MDS oops
1262 Details : more safe checks for NULL lov_tgts for avoid oops.
1264 Severity : enhancement
1266 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
1270 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
1271 Details : If an OST is down the MDS will hang indefinitely in
1272 obd_statfs_rqset() waiting for the statfs data. While for
1273 MDS QOS usage of statfs, it should not stuck in waiting.
1275 Severity : enhancement
1277 Description: remote_acl support
1278 Details : Support ACL-based permission check for remote user.
1279 Support setfacl/getfacl for remote user with the utils
1280 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
1281 the system "{s,g}etfacl" utils.
1283 Severity : enhancement
1285 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
1287 Severity : enhancement
1289 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
1293 Description: make mgs_setparam() handle fsname containing dash
1294 Details : fsname containing a dash does not work with lctl conf_param
1296 Severity : enhancement
1298 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
1300 Severity : enhancement
1302 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
1306 Frequency : rare, at shutdown
1307 Description: access already free / zero obd_namespace.
1308 Details : if client_disconnect_export was called without force flag set,
1309 and exist connect request in flight, this can produce access to
1310 NULL pointer (or already free pointer) when connect_interpret
1311 store ocd flags in obd_namespace.
1315 Frequency : only at startup
1316 Description: not alloc memory with spinlock held.
1317 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
1318 if any spinlock held.
1320 Severity : enhancement
1322 Description: make lustre randomly fail allocating memory
1323 Details : Make lustre randomly failed allocating memory for testing purpose.
1325 Severity : enhancement
1327 Description: lost problems with lov objid file
1328 Details : Fixes some scability and access to not inited memory problems
1329 in work with lov objdid file.
1334 Description: lfs find does not continue on file error
1335 Details : Continue other files processing when a file/dir is absent.
1339 Description: Inconsistent usage of lustre_pack_reply()
1340 Details : Standardize the usage of lustre_pack_reply() such that it
1341 always generate a CERROR on failure.
1344 Frequency : occasional
1346 Description: MDS hang or stay in waiting lock
1347 Details : If client receive lock with CBPENDING flag ldlm need send lock
1348 cancel as separate rpc, to avoid situation when cancel request
1349 can't processed due all i/o threads stay in wait lock.
1353 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1357 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1361 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1365 Description: Update to RHEL4 latest kernel.
1367 Severity : enhancement
1369 Description: Build SLES10 patchless client fails
1370 Details : The configure was broken by run ./configure with
1371 --with-linux-obj=.... argument for patchless client. When the
1372 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1373 can't search header adequately. Use absolute path such as
1374 -I($LINUX)/include instead.
1378 Description: interrupt oig_wait produce painc on resend.
1379 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1380 private set, and this produce situation when rq_ptlrpcd_data not
1381 copyed to new allocated request and triggered LBUG on assert
1382 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1383 ptlrpcd set if request is changed and can be safety changed to use
1388 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1389 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1390 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1391 these macros and replace them with their difinitions instead.
1395 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1396 Details : Update which_patch & target file for SLES10 latest kernel.
1399 Description: Remove client patches from SLES 10 kernel.
1400 Details : This causes SLES 10 clients to behave as patchless clients
1401 even on a Lustre-patched (server) kernel.
1403 Severity : enhancement
1405 Description: self-adjustable client's lru lists
1406 Details : use adaptive algorithm for managing client cached locks lru
1407 lists according to current server load, other client's work
1408 pattern, memory activities, etc. Both, server and client
1409 side namespaces provide number of proc tunables for controlling
1412 Severity : enhancement
1414 Description: light-weight GSS support
1415 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
1416 and reduce performance overhead.
1418 Severity : enhancement
1420 Description: Linux keyring support
1421 Details : Support using service of Linux keyring for Lustre GSS internal
1422 context refresh/cache mechanism.
1426 Description: Fix errors in lfs documentation
1427 Details : Fixes man pages
1431 Description: don't use GFP_* in generic Lustre code.
1432 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
1436 Description: obdclass is limited by single OBD_ALLOC(idarray)
1437 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
1441 Description: warnings with build patchless client with vanila 2.6.19 and up
1442 Details : change old ctl_table style and replace ctl_table/ctl_table_header
1443 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
1447 Frequency : Only for SLES
1448 Description: sanity_quota fail test_1
1449 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
1450 file, which makes TSTID[2] inunique.
1454 Frequency : Only for SLES9
1455 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
1456 Details : Update target/ChangeLog/which_patch .
1460 Frequency : Only for test_10 in sanity-quota.sh
1461 Description: a bug in quota test code
1462 Details : checking a test flag in a wrong place
1466 Frequency : Only for test_13 in sanity-quota.sh
1467 Description: a bug in quota test script
1468 Details : shouldn't check overfull allocated quota
1472 Description: jbd statistics
1473 Details : Port older jbd statistics patch for sles10
1475 Severity : enhancement
1477 Description: Add printing inode info into message about error in writepage.
1482 Description: reply_lock_interpret crash due to race with it and lock cancel.
1483 Details : Do not replay locks that are being cancelled. Do not reference
1484 locks by their address during replay, just by their handle.
1488 Frequency : When flocks are used.
1489 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
1490 Details : Flock locks might destroy just granted lock if it could be merged
1491 with another existing flock, this is done in completion handler,
1492 so teach ldlm_cli_enquque_fini that this is a valid case for
1498 Description: Oops in read and write path when failing to allocate lock.
1499 Details : Check if lock allocation failed and return error back.
1503 Description: lstripe command fails for valid OST index
1504 Details : The stripe offset is compared to lov->desc.ld_tgt_count
1505 instead of lov->desc.ld_active_tgt_count.
1509 Description: sanity.sh failed test 103
1510 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
1515 Description: df doesn't work properly if diskfs blocksize != 4K
1516 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
1520 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
1521 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
1523 Severity : enhancement
1525 Description: Adaptive timeouts
1526 Details : RPC timeouts adapt to changing server load and network
1527 conditions to reduce resend attempts and improve recovery time.
1531 Description: llapi_file_create() does not allow some changes
1532 Details : add llapi_file_open() that allows specifying the mode and
1533 open flags, and also returns an open file handle.
1537 Description: merge and cleanup kernel patches.
1538 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
1542 Description: Add journal checksum support.(Kernel part)
1543 Details : The journal checksum feature adds two new flags i.e
1544 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
1545 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
1546 indicates that the commit block contains the checksum for
1547 the blocks described by the descriptor blocks. Now commit
1548 record can be sent to disk without waiting for descriptor
1549 blocks to be written to disk. This behavior is controlled
1550 using JBD2_FEATURE_ASYNC_COMMIT flag.
1554 Description: OSS needs mutliple precreate threads
1555 Details : Add ability to start more than one create thread per OSS.
1559 Description: Sanity.sh test_65k failed: setstripe should have succeeded
1560 Details : execute lfs setstripe on client
1564 Description: mds_obd_create error creating tmp object
1565 Details : When the user sets quota on root, llog will be affected and can't
1566 create files and write files.
1569 Frequency : Always on ia64 patchless client, and possibly others.
1571 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
1572 Details : This allows the patchless client to be loaded on architectures
1573 without this export.
1577 Description: RedHat Update kernel for RHEL5
1578 Details : Add the kernel config file for RHEL5.
1582 Description: RedHat Update kernel for RHEL5
1583 Details : Modify the kernel config file more closer RHEL5.
1587 Description: Build failure against Centos5 (RHEL5)
1588 Details : Define PAGE_SIZE when it isn't present.
1592 Description: test_7 on sanity-quota.sh can't be used on separate machines
1593 Details : add facet to handle it
1597 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
1598 Details : with NFS, the anon dentry's parent was set to itself in
1599 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
1600 corresponding dentry other than use rec->ur_name.
1602 Severity : enhancement
1604 Description: lfs setstripe enhancement
1605 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
1609 Description: enable data checksumming by default
1610 Details : enable checksum by default, allow --disable-checksum
1611 configure option and "-o nochecksum" mount option. Checksums
1612 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
1616 Description: lustre support for RHEL5
1617 Details : Add support for RHEL5.
1621 Description: Client eviction due to failover config
1622 Details : after a connection loss, the lustre client should attempt to
1623 reconnect to the last active server first before trying the
1624 other potential connections.
1628 Description: when mds and osts use different quota unit(32bit and 64bit),
1629 quota will be released repeatly.
1630 Details : void sending multiple quota reqs to mds, which will keep the
1631 status between the reqs.
1635 Description: osts not allocated evenly to files
1636 Details : change the condition to increase offset_idx
1639 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
1641 Description: Data corruption for OSTs that are formatted larger than 2TB
1643 Details : When generating the bio request for lustre file writes the
1644 sector number would overflow a temporary variable before being
1645 used for the IO. The data reads correctly from Lustre (which
1646 will overflow in a similar manner) but other file data or
1647 filesystem metadata may be corrupted in some cases.
1651 Description: Tune the kernel for good SCSI performance.
1652 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
1653 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
1658 Description: rewrite ext2-derived code in obdclass/uuid.c
1659 Details : rewrite inherited code (uuid parsing code from ext2 utils)
1660 from scratch preserving functionality.
1664 Description: "lfs find -obd UUID" prints directories
1665 Details : "lfs find -obd UUID" will return all directory names instead
1666 of just file names. It is incorrect because the directories
1667 do not reside on the OSTs.
1671 Description: use i_size_read and i_size_write in 2.6 port
1672 Details : replace inode->i_size access with i_size_read/write()
1675 Frequency : when removing large files
1677 Description: scheduling issue during removal of large Lustre files
1678 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
1679 It causes scheduling issues when removing large files (17TB in the
1683 Frequency : only with liblustre clients on XT3
1685 Description: evictions taking too long
1686 Details : allow llrd to evict clients directly on OSTs
1689 Frequency : only on ppc
1691 Description: /proc/fs/lustre/devices broken on ppc
1692 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
1693 We should make sure the type of 'index' is loff_t and then cast
1694 explicitly as needed below. Do not assign an explicitly cast
1698 Frequency : only for rhel5
1700 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
1701 Details : Modify the target file & which_kernel.
1703 Severity : enhancement
1705 Description: omit set fsid for export NFS
1706 Details : fix set/restore device id for avoid EMFILE error and mark lustre
1707 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
1711 Frequency : Always, for kernels after 2.6.16
1712 Description: Fix warning idr_remove called for id=.. which is not allocated.
1713 Details : Last kernels save old s_dev before kill super and not allow
1714 to restore from callback - restore it before call kill_anon_super.
1718 Description: buffer overruns could theoretically occur
1719 Details : llapi_semantic_traverse() modifies the "path" argument by
1720 appending values to the end of the origin string, and a
1721 overrun may occur. Adding buffer overrun check in liblustreapi.
1725 Description: Fix error on 'ls .' at the top of the Lustre mount.
1726 Details : Don't revalidate dentry if it is a root dentry.
1730 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
1731 Details : Modify vm-tunables-rhel4.patch.
1735 Description: Kernel config for 2.6.18-vanilla.
1736 Details : Modify targets/2.6-vanilla.target.in.
1737 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
1738 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
1739 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
1740 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
1744 Description: improve handling recoverable errors
1745 Details : if request processig with error which can be recoverable on server
1746 request should be resend, otherwise page released from cache and
1751 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
1752 Details : Modify target file & which_patch.
1753 A flaw was found in the IA32 system call emulation provided
1754 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
1755 value could be stored in the %RAX register, which could trigger an
1756 out-of-bounds system call table access. An untrusted local user
1757 could exploit this flaw to run code in the kernel
1758 (ie a root privilege escalation). (CVE-2007-4573).
1762 Description: O_DIRECT bypasses client statistics.
1763 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
1764 read_ahead_stats were updated. copied the stats section from
1765 osc_send_oap_rpc() into async_internal().
1769 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
1773 Description: change order of libsysio includes
1774 Details : '#include sysio.h' should always come before '#include xtio.h'
1778 Description: handle "serious error: objid * already exists" more gracefully
1779 Details : If LAST_ID value on disk is smaller than the objects existing in
1780 the O/0/d* directories, it indicates disk corruption and causes an
1781 LBUG(). If the object is 0-length, then we should use the existing
1782 object. This will help to avoid a full fsck in most cases.
1784 Severity : enhancement
1786 Description: adapt the lustre_config script to support the upgrade case
1787 Details : Add "-u" option for lustre_config script to support upgrading 1.4
1788 server targets to 1.6 in parallel.
1792 Description: To avoid grant space > avaible space when the disk is almost
1793 full. Without this patch you might see the error "grant XXXX >
1794 available" or some LBUG about grant, when the disk is almost
1796 Details : In filter_check_grant, for non_grant cache write, we should
1797 check the left space by if (*left > ungranted + bytes), instead
1798 of (*left > ungranted), because only we are sure the left space
1799 is enough for another "bytes", then the ungrant space should be
1800 increase. In client, we should update cl_avail_grant only there
1801 is OBD_MD_FLGRANT in the reply.
1805 Description: Update RHEL 4 kernel to fix local root privilege escalation.
1806 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
1807 described in CVE-2007-4573. This problem could allow untrusted
1808 local users to gain root access.
1811 Frequency : when using O_DIRECT and quotas
1813 Description: Incorrect file ownership on O_DIRECT output files
1814 Details : block usage reported by 'lfs quota' does not take into account
1815 files that have been written with O_DIRECT.
1820 Description: touch file failed when fs is not full
1821 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
1822 otherwise we can get ENOSP while fs is not full.
1826 Description: parallel lock callbacks
1827 Details : Instead of sending blocking and completion callbacks as separated
1828 requests, adding them to a set and sending in parallel.
1831 Frequency : only for Cray XT3
1832 Bugzilla : 12829/13455
1833 Description: Changing primary group doesn't change the group lustre assigns to
1835 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
1836 overridden with the primary group provided by the group upcall,
1837 whereas the supplied fsgid can be trusted if it is in the list of
1838 supplementary groups returned by the group upcall.
1840 Severity : enhancement
1842 Description: Allow masking D_WARNING, D_ERROR messages from console
1843 Details : Console messages can now be disabled via lnet.printk.
1847 Description: User code with malformed file open parameter crashes client node
1848 Details : Before packing join_file req, all the related reference should be
1849 checked carefully in case some malformed flags cause fake
1850 join_file req on client.
1854 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
1855 Details : ldlm_completion_ast() assumes that a lock is granted when the req
1856 mode is equal to the granted mode. However, it should also check
1857 that LDLM_FL_CP_REQD is not set.
1861 Description: Heavy nfs access might result in deadlocks
1862 Details : After ELC code landed, it is now improper to enqueue any mds
1863 locks under och_sem, because enqueue might want to decide to
1864 cancel open locks for same inode we are holding och_sem for.
1868 Description: Client eviction while running blogbench
1869 Details : A lot of unlink operations with concurrent I/O can lead to a
1870 deadlock causing evictions. To address the problem, the number of
1871 oustanding OST_DESTROY requests is now throttled to
1872 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
1873 callbacks are processed in priority.
1877 Description: enable ACLs on MDS by default
1878 Details : ACLs must be enabled on MDS by default.
1881 Frequency : PPC/PPC64 only
1883 Description: conflicts between asm-ppc64/types.h and lustre_types.h
1884 Details : fix duplicated definitions between asm-ppc64/types.h and
1885 lustre_types.h on PPC.
1888 Frequency : PPC/PPC64 only
1890 Description: asm-ppc/segment.h does not exist
1891 Details : fix compile issue on PPC.
1895 Description: better handle error messages in extents code
1898 Frequency : RHEL4 only
1900 Description: mkfs is very slow on IA64/RHEL4
1901 Details : A performance regression has been discovered in the MPT Fusion
1902 driver between versions 3.02.73rh and 3.02.99.00rh. As a
1903 consequence, we have downgraded the MPT Fusion driver in the RHEL4
1904 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
1906 Severity : enhancement
1908 Description: SNMP support enhancement
1909 Details : Adding total number of sampled request for an MDS node in snmp
1912 Severity : enhancement
1914 Description: Optimize ldlm waiting list processing for PR extent locks
1915 Details : When processing waiting list for read extent lock and meeting read
1916 lock that is same or wider to it that is not contended, skip
1917 processing rest of the list and immediatelly return current
1918 status of conflictness, since we are guaranteed there are no
1919 conflicting locks in the rest of the list.
1923 Description: Time out and refuse to reconnect
1924 Details : When the failover node is the primary node, it is possible
1925 to have two identical connections in imp_conn_list. We must
1926 compare not conn's pointers but NIDs, otherwise we can defeat
1927 connection throttling.
1931 Description: port llog fixes from b1_6 into HEAD
1932 Details : Port llog reference couting and some llog cleanups from b1_6
1933 (bug 10800) into HEAD, for protect from panic and access to already
1934 free llog structures.
1938 Description: Detect stride IO mode in read-ahead
1939 Details : When a client does stride read, read-ahead should detect that and
1940 read-ahead pages according to the detected stride pattern.
1944 Description: data checksumming impacts single node performance
1945 Details : add support for several checksum algorithm. Currently, only CRC32
1946 and Adler-32 are supported. The checksum type can be changed on
1947 the fly via /proc/fs/lustre/osc/*/checksum_type.
1951 Description: use adler32 for page checksums
1952 Details : when available, use the Adler-32 algorithm instead of CRC32 for
1957 Description: build for x2 fails
1958 Details : fix compile issue on Cray systems.
1962 Description: Properly match for duplicate locks
1963 Details : Due to different lock order from skiplists code, we need to
1964 traverse entire list for now
1967 Frequency : only on PPC/SLES10
1969 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
1970 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
1971 be defined. Add a hack in mkfs_lustre.c to work around this compile
1976 Description: LASSERT on MDS when client holding flock lock dies
1977 Details : ldlm pool logic depends on number of granted locks equal to
1978 number of released locks which is not true for flock locks, so
1979 just exclude such locks from consideration.
1983 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
1984 Details : Use fsfilt_sync() for both the whole filesystem sync and
1985 individual file sync to eliminate dangerous inode locking
1986 with I_LOCK that can lead to a deadlock.
1990 Description: performance in 1.6.3
1991 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
1992 sure that 1MB requests are not fragmented by the block layer.
1994 Severity : enhancement
1996 Description: organize the server-side client stats on per-nid basis
1997 Details : Change the structure of stats under obdfilter and mds to
2005 The "uuid"s file would list the uuids of _active_ exports.
2006 And the clear entry is to clear all stats and stale nids.
2008 Severity : enhancement
2010 Description: eliminate client locks in face of contention
2011 Details : file contention detection and lockless i/o implementation
2012 for contended files.
2016 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
2019 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
2021 Description: MDS may lose file striping (and hence file data) in some cases
2022 Details : If there are additional extended attributes stored on the MDS,
2023 in particular ACLs, SELinux, or user attributes (if user_xattr
2024 is specified for the client mount options) then there is a risk
2025 of attribute loss. Additionally, the Lustre file striping
2026 needs to be larger than default (e.g. striped over all OSTs),
2027 and an additional attribute must be stored initially in the
2028 inode and then increase in size enough to be moved to the
2029 external attribute block (e.g. ACL growing in size) for file
2034 Description: skiplist implementation simplification
2035 Details : skiplists are used to group compatible locks on granted list
2036 that was implemented as tracking first and last lock of each
2037 lock group the patch changes that to using doubly linked lists
2041 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
2042 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
2043 because LASSERT on both the data supplied by a client, and the
2044 data on disk is dangerous and incorrect.
2046 Severity : enhancement
2048 Description: Slow truncate/writes to huge files at high offsets.
2049 Details : Directly associate cached pages to lock that protect those pages,
2050 this allows us to quickly find what pages to write and remove
2051 once lock callback is received.
2055 Description: more ldlm soft lockups
2056 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
2057 starve other threads from the resource lock for a long time in
2058 case of long waiting queue, so change the debug level from
2059 D_OTHER to the less frequently used D_INFO.
2061 Severity : enhancement
2063 Description: add -gid, -group, -uid, -user options to lfs find
2067 Description: Hung threads in invalidate_inode_pages2_range
2068 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
2069 once one is completed. As a result, some RPCs are stuck in the
2070 queue and are never sent.
2074 Description: filter threads hungs on waiting journal commit
2075 Details : Cleanup filter group llog code, then only filter group llog will
2076 be only created in the MDS/OST syncing process.
2080 Description: Procfs and llog threads access destoryed import sometimes.
2081 Details : Sync the import destoryed process with procfs and llog threads by
2082 the import refcount and semaphore.
2084 Severity : enhancement
2086 Description: openlock cache of b1_6 port to HEAD
2091 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
2092 Details : If insertion of an extent fails, then discard the inode
2093 preallocation and free data blocks else it can lead to duplicate
2098 Description: don't always update ctime in ext3_xattr_set_handle()
2099 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
2100 In some cases the ctime should not be updated, for example for
2101 2.0->1.8 compatibility it is necessary to delete an xattr and it
2102 should not update the ctime.
2106 Bugzilla : 15713/16362
2107 Description: Assertion in iopen_connect_dentry in 1.6.3
2108 Details : looking up an inode via iopen with the wrong generation number can
2109 populate the dcache with a disconneced dentry while the inode
2110 number is in the process of being reallocated. This causes an
2111 assertion failure in iopen since the inode's dentry list contains
2112 both a connected and disconnected dentry.
2116 Description: assertion failure in ldlm_handle2lock()
2117 Details : fix a race between class_handle_unhash() and class_handle2object()
2118 introduced in lustre 1.6.5 by bug 13622.
2123 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
2124 Details : remove the limit of 256 scsi disks in the sd_iostat patch
2129 Description: oops in sd_iostats_seq_show()
2130 Details : unloading/reloading the scsi low level driver triggers a kernel
2131 bug when trying to access the sd iostat file.
2136 Description: Kernel panics during QLogic driver reload
2137 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
2138 patch, causing memory corruption.
2143 Description: journal_dev option does not work in b1_6
2144 Details : pass mount option during pre-mount.
2146 Severity : enhancement
2148 Description: Add a FIEMAP(FIle Extent MAP) ioctl
2149 Details : FIEMAP ioctl will allow an application to efficiently fetch the
2150 extent information of a file. It can be used to map logical blocks
2151 in a file to physical blocks in the block device.
2155 Description: LDLM soft lockups - improvement
2156 Details : It is be possible to send the lock handle along with each read
2157 or write request because the client is already doing a lock match
2158 itself so there isn't any reason the OST should have to re-do that
2162 Frequency : only on Cray X2
2164 Description: X2 build failures
2165 Details : fix build failures on Cray X2.
2169 Description: xid & resent requests
2170 Details : Initialize RPC XID from clock at startup (randomly if clock is
2173 Severity : enhancement
2175 Description: Add lustre_start utility to start or stop multiple Lustre servers
2180 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
2181 Details : In case of memory pressure, list_del() can be called twice on
2182 req->rq_history_list, causing a kernel oops.
2186 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
2187 Details : fix stack overflow in the distributed lock manager by defering
2188 export eviction after a failed AST to the elt thread instead of
2189 handling it in the dlm interpret routine.
2193 Description: Convert some comments to new format.
2194 Details : Update documenting comments to match doxygen conventions.
2198 Description: Grammar fixes.
2199 Details : A couple of trivial spelling fixes.
2203 Description: OSD_COUNTERS-mandatory
2204 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
2205 to simplify the code.
2209 Description: simplify cmm_device freeing logic.
2210 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
2214 Description: Add lockdep support to dt_object_operations locking interface.
2215 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
2216 parameter indicating lock ordering. Update mdd code to use new
2221 Description: Introduce failloc constants for lockless IO tests.
2222 Details : Add two new failloc constants to test lockless IO. Only one of
2223 them in implemented---another is checked in yet to be landed
2228 Description: Add lockdep support for inode mutex.
2229 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
2230 in the situations where Lustre has to lock more than one inode
2235 Description: Add optional invariants checking support.
2236 Details : Add new LINVRNT() macro, optional on new --enable-invariants
2237 configure switch. This macro is to be used for consistency and
2238 sanity checks that are too expensive to be left in `production'
2243 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
2244 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
2249 Description: Add lockdep support for ldlm_lock and ldlm_resource.
2250 Details : Use spin_lock_nested() in (the only) situation where more than
2251 one ldlm_lock is locked simultaneously. Also, fix possible
2252 dead-lock in ldlm_lock_change_resource() by enforcing particular
2257 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
2258 Details : Instead of specifying each ldlm_lock call-back through separate
2259 parameter, wrap them into struct ldlm_callback_suite.
2263 Description: Kill join_lru obd method and its callers.
2264 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
2265 in memory---a requirement implemented through ->o_join_lru() obd
2266 method in HEAD. Remove this method and its users.
2270 Description: Add asynchronous ldlm ENQUEUE completion handler.
2271 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
2272 case that stock ldlm_completion_ast() cannot handle as it waits
2273 until lock is granted. Introduce new ldlm_completion_ast_async()
2274 for this. Also comment ldlm_completion_ast().
2278 Description: ldlm_error <-> errno conversion.
2279 Details : Add functions to map (rather arbitrary) between LDLM error codes
2280 and standard errno values. CLIO needs this to prevent LDLM specific
2281 constants from escaping ldlm and osc.
2285 Description: Kill unused ldlm_handle2lock_ns() function.
2286 Details : Kill unused ldlm_handle2lock_ns() function.
2290 Description: Add lu_ref support to ldlm_lock
2291 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
2292 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
2293 and ldlm_resource. LDLM interface has to be changed, because code that
2294 releases a reference on a lock, has to "know" what reference this is.
2295 In the most frequent case
2297 lock = ldlm_handle2lock(handle);
2299 LDLM_LOCK_PUT(lock);
2301 no changes are required. When any other reference (received _not_ from
2302 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
2303 instead of LDLM_LOCK_PUT().
2305 Arguably, changes are pervasive, and interface requires some discipline
2306 for proper use. On the other hand, it was very instrumental in finding
2307 a few leaked lock references.
2311 Description: Add ldlm_lock_addref_try().
2312 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
2313 attempts to addref a lock that might be being canceled
2318 Description: Add ldlm_weigh_callback().
2319 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
2320 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
2321 hard-coded `number of pages' logic.
2325 Description: Add lockdep annotations to llog code.
2326 Details : Use appropriately tagged _nested() locking calls in the places
2327 where llog takes more than one ->lgh_lock lock.
2331 Description: Add loi_kms_set().
2332 Details : Wrap kms updates into a helper function.
2336 Description: Constify instances of struct lsm_operations.
2337 Details : Constify instances of struct lsm_operations.
2341 Description: lu_conf support.
2342 Details : On a server, a file system object is uniquely identified
2343 by a fid, which is sufficient to locate and load all object
2344 state (inode). On a client, on the other hand, more data are
2345 necessary instantiate an object. Change lu_object_find() and
2346 friends to take additional `lu_conf' argument describing object.
2347 Typically this includes layout information.
2351 Description: lu_context fixes.
2352 Details : Introduce new lu_context functions that are needed on the client
2353 side, where some system threads (ptlrpcd) are shared by multiple
2354 modules, and so cannot be stopped during module shutdown.
2358 Description: Add start and stop methods to lu_device_type_operations.
2359 Details : Introduce two new methods in lu_device_type_operations, that are
2360 invoked when first instance of a given type is created and last one
2361 is destroyed respectively. This is need by CLIO.
2365 Description: Add lu_ref support to struct lu_device.
2366 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
2367 track leaked references.
2371 Description: Introduce lu_kmem_descr.
2372 Details : lu_kmem_descr and its companion interface allow to create
2373 and destroy a number of kmem caches at once.
2377 Description: Fix lu_object finalization race.
2378 Details : Fix a race between lu_object_find() finding an object and its
2379 concurrent finalization. This race is (most likely) not possible
2380 on the server, but might happen on the client.
2384 Description: Introduce lu_ref interface.
2385 Details : lu_ref is a debugging module allowing to track references to
2386 a given object. It is quite cpu expensive, and has to be
2387 explicitly enabled with --enable-lu_ref. See usage description
2392 Description: Factor lu_site procfs stats into a separate function.
2393 Details : Separate lu_site stats printing code into a separate function
2394 that can be reused on a client.
2398 Description: Constify instances of struct {lu,dt,md}_device_operations.
2399 Details : Constify instances of struct {lu,dt,md}_device_operations.
2403 Description: Introduce struct md_site and move meta-data specific parts of
2404 struct lu_site here.
2405 Details : Move md-specific fields out of struct lu_site into special struct
2406 md_site, so that lu_site can be used on a client.
2410 Description: Kill mdd_lov_destroy().
2411 Details : Remove unused mdd code.
2415 Description: Add st_block checking to multistat.c.
2416 Details : Add st_block checking to multistat.c.
2420 Description: Add lu_ref support to struct obd_device.
2421 Details : Add lu_ref tracking to obd_device.
2425 Description: Kill obd_set_fail_loc().
2426 Details : Remove unused code.
2430 Description: Add special type for ptlrpc_request interpret functions.
2431 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
2432 there. Actual usage will be in CLIO.
2436 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
2437 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
2438 doesn't work with lockdep.
2442 Description: Add rwv.c test program.
2443 Details : New testing program exercising readv(2) and writev(2) (Qian).
2447 Description: Add sendfile.c test program.
2448 Details : New testing program exercising sendfile(2) (Jay).
2452 Description: Ratelimit a message that can be very frequent.
2453 Details : Ratelimit a memory allocation failure message that can
2458 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2459 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2463 Description: Liblustre build fixes.
2464 Details : Liblustre build fixes.
2468 Description: libcfs: add cfs_{need,cond}_resched() interface.
2469 Details : libcfs: add cfs_{need,cond}_resched() definition and
2470 implementations for Linux, NT, and liblustre.
2472 Severity : enhancement
2474 Description: More exported tunables for mballoc
2475 Details : Add support for tunable preallocation window and new tunables for
2476 large/small requests
2480 Description: Detect corruption of block bitmap and checking for preallocations
2481 Details : Checks validity of on-disk block bitmap. Also it does better
2482 checking of number of applied preallocations. When corruption is
2483 found, it turns filesystem readonly to prevent further corruptions.
2487 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
2488 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
2489 window to 0. If it does hit cache miss, and read-pattern is still
2490 stride-io mode, does not reset the stride window, but also does not
2491 increase the stride window length in this case.
2495 Frequency : only for big-endian servers
2496 Description: Check if system is big-endian while mounting fs with extents feature
2497 Details : Mounting a filesystem with extents feature will fail on big-endian
2498 systems since ext3-based ldiskfs is not supported on big-endian
2499 systems. This can be over-riden with "bigendian_extents" mount option.
2501 Severity : enhancement
2503 Description: The root squash functionality
2504 Details : A security feature, which is to prevent users from being able
2505 to mount lustre on their desktop, run as root, and delete
2506 all of the files in the filesystem. The goal is accomplished by
2507 remapping user id (UID) and group id (GID) of the root user to
2508 a UID and GID specified by the system administartor via Lustre
2509 configuration management server (MGS). The functionality also
2510 allows to specify sets of clients for which the remapping does
2515 Description: Excessive recovery window
2516 Details : With AT enabled, the recovery window can be excessively long (6000+
2517 seconds). To address this problem, we no longer use
2518 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
2519 timeout no longer depends on the service time, it is set to
2520 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
2521 time via pb_service_time.
2525 Description: Watchdog triggered on MDS failover
2526 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
2527 the OSTs know that the MDS "UUID" can be reused for the same export
2528 from a different NID, so we do not need to wait for the export to be
2532 Frequency : rare, only if using MMP with Linux RAID
2534 Description: MMP doesn't work with Linux RAID
2535 Details : While using HA for Lustre servers with Linux RAID, it is possible
2536 that MMP will not detect multiple mounts. To make this work we
2537 need to unplug the device queue in RAID when the MMP block is being
2538 written. Also while reading the MMP block, we should read it from
2539 disk and not the cached one.
2541 Severity : enhancement
2543 Description: open file using fid
2544 Details : A file can be opened using just its fid, like
2545 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
2548 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
2551 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
2552 Details : add proper locking for creation of jbd_4k slab cache
2556 Description: MMP check in ext3_remount() fails without displaying any error
2557 Details : When multiple mount protection fails during remount, proper error
2560 Severity : enhancement
2562 Description: Allow stripe size to be up to 4G-64k
2563 Details : Fix math logic to allow large stripe sizes.
2567 Description: add check for >8TB ldiskfs filesystems
2568 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
2569 Don't allow >8TB ldiskfs filesystems to be mounted without
2570 force_over_8tb mount option
2575 Descriptoin: OSS complains "too many blocks" and crashes
2576 Details : In filter_commitrw_write():
2577 1. lquota_chkquota();
2578 2. obdo_from_inode();
2579 3. lquota_pending_commit();
2580 "2." has changed uid and gid in this case, then caused
2581 lquota_chkquota() and lquota_pending_commit() to use
2582 different uid and gid. That is the root of the bug.
2586 Descriptoin: getstripe doesn't report stripe count = -1 sanely
2587 Details : Detect and print out the 'stripe count = -1' case properly.
2591 Description: Out or order replies might be lost on replay
2592 Details : In ptlrpc_retain_replayable_request if we cannot find retained
2593 request with tid smaller then one currently being added, add it
2594 to the start, not end of the list.
2596 --------------------------------------------------------------------------------
2598 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
2600 * Support for kernels:
2601 2.6.5-7.283 (SLES 9),
2602 2.6.9-55.EL (RHEL 4),
2603 2.6.16.46-0.14 (SLES 10),
2604 2.6.18.8 vanilla (kernel.org)
2605 * Client support for unpatched kernels:
2606 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2607 2.6.16 - 2.6.22 vanilla (kernel.org)
2608 * Due to recently discovered recovery problems, we do not recommend
2609 using patchless RHEL 4 clients with this or any earlier release.
2610 * Recommended e2fsprogs version: 1.39.cfs8
2611 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2612 * Starting with this release, the ldiskfs backing filesystem required
2613 by Lustre is now in its own package, lustre-ldiskfs. This package
2614 should be installed. It is versioned separately from Lustre and
2615 may be released separately in future.
2619 Description: block reactivating mgc import until all deactivates complete
2620 Details : Fix race when failing back MDT/MGS to itself (testing)
2623 Frequency : at statup only
2625 Description: mds_lov_synchronize race leads to various problems
2626 Details : simultaneous MDT->OST connections at startup can cause the
2627 sync to abort, leaving the OSC in a bad state.
2629 Severity : enhancement
2631 Description: add optional extra BUILD_VERSION info
2632 Details : add a new environment variable (namely LUSTRE_VERS) which allows
2633 to override the lustre version.
2636 Frequency : 2.6.18 servers only
2638 Description: ll_kern_mount() doesn't release the module reference
2639 Details : The ldiskfs module reference count never drops down to 0
2640 because ll_kern_mount() doesn't release the module reference.
2645 Description: server LBUG when using old ost_num_threads parameter
2646 Details : Accept the old ost_num_threads parameter but warn that it
2647 is deprecated, and fix an off-by-one error that caused an LBUG.
2652 Description: Transient SCSI error results in persistent IO issue
2653 Details : iobuf->dr_error is not reinitialized to 0 between two
2657 Frequency : sometimes when underlying device returns I/O errors
2659 Description: OSTs not going read-only during write failures
2660 Details : OSTs are not remounted read-only when the journal commit threads
2661 get I/O errors because fsfilt_ext3 calls journal_start/stop()
2662 instead of the ext3 wrappers.
2666 Description: poor connect scaling with increasing client count
2667 Details : Don't run filter_grant_sanity_check for more than 100 exports
2668 to improve scaling for large numbers of clients.
2671 Frequency : SLES10 only
2673 Description: sanity-quota.sh quotacheck failed: rc = -22
2674 Details : Quotas cannot be enabled on SLES10.
2677 Frequency : liblustre clients only
2679 Description: getdirentries does not give error when run on compute nodes
2680 Details : getdirentries does not fail when the size specified as an argument
2681 is too small to contain at least one entry
2683 Severity : enhancement
2685 Description: Add LNET router traceability for debug purposes
2686 Details : If a checksum failure occurs with a router as part of the
2687 IO path, the NID of the last router that forwarded the bulk data
2688 is printed so it can be identified.
2693 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
2694 Details : Due to a race condition, liblustre clients were occasionally
2695 evicted incorrectly.
2697 Severity : enhancement
2699 Description: lfs setstripe use optional parameters instead of postional
2702 Severity : enhancement
2704 Description: Nanosecond timestamp support for ldiskfs
2705 Details : The on-disk ldiskfs filesystem has added support for nanosecond
2706 resolution timestamps. There is not yet support for this at
2707 the Lustre filesystem level.
2710 Frequency : during server recovery
2712 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
2713 Details : request with rq_no_resend flag not awake l_wait_event if they get
2717 Frequency : nfs export on patchless client
2719 Description: connectathon hang when test nfs export over patchless client
2720 Details : Disconnected dentry cannot be found with lookup, so we do not need
2721 to unhash it or make it invalid
2724 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
2725 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
2726 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
2731 Description: open req refcounting wrong on reconnect
2732 Details : If reconnect happened between getting open reply from server and
2733 call to mdc_set_replay_data in ll_file_open, we will schedule
2734 replay for unreferenced request that we are about to free.
2735 Subsequent close will crash in variety of ways.
2736 Check that request is still eligible for replay in
2737 mdc_set_replay_data().
2742 Description: disable writes to filesystem when reading health_check file
2743 Details : the default for reading the health_check proc file has changed
2744 to NOT do a journal transaction and write to disk, because this
2745 can cause reads of the /proc file to hang and block HA state
2746 checking on a healthy but otherwise heavily loaded system. It
2747 is possible to return to the previous behaviour during configure
2748 with --enable-health-write.
2750 Severity : enhancement
2752 Description: 64-bit inode version
2753 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
2754 to the inode. This will be required for version-based recovery.
2759 Description: MDS fails to start if a duplicate client export is detected
2760 Details : in some rare cases it was possible for a client to connect to
2761 an MDS multiple times. Upon recovery the MDS would detect this
2762 and fail during startup. Handle this more gracefully.
2764 Severity : enhancement
2766 Description: Add -o localflock option to simulate old noflock
2768 Details : This will achieve local-only flock/fcntl locks
2774 Description: log_commit_thread vs filter_destroy race leads to crash
2775 Details : Take import reference before releasing llog record semaphore
2780 Description: Wrong request locking in request set processing
2781 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
2782 imp_delayed_list, in this place should be used imp_lock.
2785 Frequency : when reconnection
2787 Description: Grant Leak when osc reconnect to OST
2788 Details : When osc reconnect ost, OST(filter) should check whether it
2789 should grant more space to client by comparing fed_grant and
2790 cl_avail_grant, and return the granted space to client instead
2791 of "new granted" space, because client will call osc_init_grant
2792 to update the client grant space info.
2795 Frequency : when client reconnect to OST
2797 Description: Grant Leak when osc do resend and replay bulk write
2798 Details : When osc reconnect to OST, OST(filter)should clear grant info of
2799 bulk write request, because the grant info while be sync between
2800 OSC and OST when reconnect, and we should ignore the grant info
2801 these of resend/replay write req.
2806 Description: Grant space more than avaiable left space sometimes.
2807 Details : When then OST is about to be full, if two bulk writing from
2808 different clients came to OST. Accord the avaliable space of the
2809 OST, the first req should be permitted, and the second one
2810 should be denied by ENOSPC. But if the seconde arrived before
2811 the first one is commited. The OST might wrongly permit second
2812 writing, which will cause grant space > avaiable space.
2815 Frequency : when client is evicted
2817 Description: Grant might be wrongly erased when osc is evicted by OST
2818 Details : when the import is evicted by server, it will fork another
2819 thread ptlrpc_invalidate_import_thread to invalidate the
2820 import, where the grant will be set to 0. While the original
2821 thread will update the grant it got when connecting. So if
2822 the former happened latter, the grant will be wrongly errased
2823 because of this race.
2828 Description: Checking Stale with correct fid
2829 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
2830 to check whether it is stale, because sometimes, we want the
2831 enqueue happened anyway, and op_data.fid2 will not be initialized.
2833 Severity : enhancement
2835 Description: update patchless client
2836 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
2839 Frequency : only with 2.4 kernel
2841 Description: random memory corruption
2842 Details : size of struct ll_inode_info is to big for union inode.u and this
2843 can be cause of random memory corruption.
2848 Description: Memory leak in recovery
2849 Details : Lov_mds_md was not free in an error handler in mds_create_object.
2850 It should also check obd_fail before fsfilt_start, otherwise if
2851 fsfilt_start return -EROFS,(failover mds during mds recovery).
2852 then the req will return with repmsg->transno = 0 and rc = EROFS.
2853 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
2854 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
2855 be freed no matter whether fsfilt_commit success or not.
2858 Frequency : only with huge count clients
2860 Description: Prevents from taking the superblock lock in llap_from_page for
2862 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
2863 for llap_from_page call in ll_removepage prevents from taking the
2864 superblock lock for a soon died page.
2869 Description: Not check open intent error before release open handle
2870 Details : in some rare cases, the open intent error is not checked before
2871 release open handle, which may cause
2872 ASSERTION(open_req->rq_transno != 0), because it tries to release
2873 the failed open handle.
2878 Description: Set cat log bitmap only after create log success.
2879 Details : in some rare cases, the cat log bitmap is set too early. and it
2880 should be set only after create log success.
2884 Description: Accessing a block bevice can re-enable I/O when Lustre is
2885 tearing down a device.
2886 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
2891 Description: service threads may hog cpus when there are a lot of requests
2892 Details : Insert cond_resched to give other threads a chance to use some CPU
2897 Description: the cat log was not initialized in recovery
2898 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
2899 unlink log on mds will not be initialized until mds post
2900 recovery. And also in mds post recovery, the unlink log will
2901 initialization will be done asynchronausly, so there will be race
2902 between add unlink log and unlink log initialization.
2906 Description: brw_stats were being printed incorrectly
2907 Details : brw_stats were being printed as log2 but all of them were not
2908 recorded as log2. Also remove some code duplication arising from
2909 filter_tally_{read,write}.
2913 Frequency : rare, only in recovery.
2914 Description: ASSERTION(req->rq_type != LI_POISON) failed
2915 Details : imp_lock should be held while iterating over imp_sending_list for
2916 prevent destroy request after get timeout in ptlrpc_queue_wait.
2920 Description: replay-single.sh test 52 fails
2921 Details : A lock's skiplist need to be cleanup when it being unlinked
2922 from its resource list.
2926 Description: Short directio read returns full requested size rather than
2928 Details : Direct I/O operations should return actual amount of bytes
2929 transferred rather than requested size.
2931 Severity : enhancement
2933 Description: metadata RPC reduction (e.g. for rm performance)
2934 Details : decrease the amount of synchronous RPC between clients and servers
2935 by canceling conflicing lock before the operation on the client
2936 and packing thier handles into the main operation RPC to server.
2938 Severity : enhancement
2940 Description: Async OSC create to avoid the blocking unnecessarily.
2941 Details : If a OST has no remain object, system will block on the creating
2942 when need to create a new object on this OST. Now, ways use
2943 pre-created objects when available, instead of blocking on an
2944 empty osc while others are not empty. If we must block, we block
2945 for the shortest possible period of time.
2949 Description: improve handling recoverable errors
2950 Details : if request processig with error which can be recoverable on server
2951 request should be resend, otherwise page released from cache and
2954 Severity : enhancement
2956 Description: refine locking for avoid write wrong info into lov_objid file
2957 Details : fix possible races with add new target and write/update data in
2960 --------------------------------------------------------------------------------
2962 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
2967 Frequency : on some architectures
2969 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
2970 Details : Uninitialized flags sometimes cause configuration commands to
2974 Frequency : patchless clients only
2976 Description: missing __iget() symbol export
2977 Details : The __iget() symbol export is missing. To avoid the need for
2978 this on patchless clients the deathrow inode reaper is turned
2979 off, and we depend on the VM to clean up old inodes. This
2980 dependency was during via the fix for bug 12181.
2982 --------------------------------------------------------------------------------
2984 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
2986 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
2987 INTEROPERATE with older versions automatically. In many cases a
2988 special upgrade step is needed. Please read the
2989 user documentation before upgrading any part of a live system.
2990 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
2991 version will not interoperate with 1.6 betas before beta5 (1.5.95).
2992 * WARNING: Lustre configuration and startup changes are required with
2993 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
2997 Severity : enhancement
2999 Description: MountConf
3000 Details : Lustre configuration is now managed via mkfs and mount
3001 commands instead of lmc and lconf. New obd types (MGS, MGC)
3002 are added for dynamic configuration management. See
3003 https://mail.clusterfs.com/wikis/lustre/MountConf for
3006 Severity : enhancement
3008 Description: dynamic OST addition
3009 Details : OSTs can now be added to a live filesystem
3011 Severity : enhancement
3013 Description: startup order invariance
3014 Details : MDTs and OSTs can be started in any order. Clients only
3015 require the MDT to complete startup.
3017 Severity : enhancement
3019 Description: parallel, asynchronous orphan cleanup
3020 Details : orphan cleanup is now performed in separate threads for each
3021 OST, allowing parallel non-blocking operation.
3023 Severity : enhancement
3025 Description: optimized stripe assignment
3026 Details : stripe assignments are now made based on ost space available,
3027 ost previous usage, and OSS previous usage, in order to try
3028 to optimize storage space and networking resources.
3030 Severity : enhancement
3032 Description: Permanently set tunables
3033 Details : All writable /proc/fs/lustre tunables can now be permanently
3034 set on a per-server basis, at mkfs time or on a live system.
3036 Severity : enhancement
3038 Description: Lustre message v2
3039 Details : Add lustre message format v2.
3041 Severity : enhancement
3043 Description: client OST exclusion list
3044 Details : Clients can be started with a list of OSTs that should be
3045 declared "inactive" for known non-responsive OSTs.
3049 Description: SPEC SFS validation failure on NFS v2 over lustre.
3050 Details : Changes the blocksize for regular files to be 2x RPC size,
3051 and not depend on stripe size.
3053 Severity : enhancement
3055 Description: Multiple MD RPCs in flight.
3056 Details : Further unserialise some read-only MDS RPCs - learn about intents.
3057 To avoid overly-overloading MDS, introduce a limit on number of
3058 MDS RPCs in flight for a single client and add /proc controls
3059 to adjust this limit.
3061 Severity : enhancement
3063 Description: client read/write statistics
3064 Details : Add client read/write call usage stats for performance
3065 analysis of user processes.
3066 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
3067 file access. extents_stats shows chunk size distribution.
3068 extents_stats_per_process show chunk size distribution per
3071 Severity : enhancement
3073 Description: mds statistics
3074 Details : Add detailed mds operations statistics in
3075 /proc/fs/lustre/mds/*/stats.
3079 Description: Failure of copying files with lustre special EAs.
3080 Details : Client side always return success for setxattr call for lustre
3081 special xattr (currently only "trusted.lov").
3086 Description: Refcount LNET uuids
3087 Details : The global LNET uuid list grew linearly with every startup;
3088 refcount repeated list entries instead of always adding to
3091 Severity : enhancement
3093 Description: Dynamic service threads
3094 Details : Within a small range, start extra service threads
3095 automatically when the request queue builds up.
3098 Frequency : mixed-endian client/server environments
3100 Description: mixed-endian crashes
3101 Details : The new msg_v2 system had some failures in mixed-endian
3104 Severity : enhancement
3106 Description: Easy OST removal
3107 Details : OSTs can be permanently deactivated with e.g. 'lctl
3108 conf_param lustre-OST0001.osc.active=0'
3110 Severity : enhancement
3112 Description: MGS proc entries
3113 Details : Added basic proc entries for the MGS showing what filesystems
3116 Severity : enhancement
3118 Description: provide MGS failover
3119 Details : Added config lock reacquisition after MGS server failover.
3121 Severity : enhancement
3123 Description: add Linux 2.4 support
3124 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
3128 Description: a large application tries to do I/O to the same resource and dies
3129 in the middle of it.
3130 Details : Check the req->rq_arrival time after the call to
3131 ost_brw_lock_get(), but before we do anything about
3132 processing it & sending the BULK transfer request. This
3133 should help move old stale pending locks off the queue as
3134 quickly as obd_timeout.
3137 Frequency : when an incorrect nid is specified during startup
3139 Description: ptlrpc connect to non-existant node causes kernel crash
3140 Details : LNET can't be re-entered from an event callback, which
3141 happened when we expire a message after the export has been
3142 cleaned up. Instead, hand the zombie cleanup off to another
3145 Severity : enhancement
3147 Description: plain/inodebits lock performance improvement
3148 Details : Grouping plain/inodebits in granted list by their request modes
3149 and bits policy, thus improving the performance of search through
3153 Frequency : only if OST filesystem is corrupted
3155 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3156 Details : for a short time RPCs with bulk IO are in the replay list,
3157 but replay of bulk IOs is unimplemented. If the OST filesystem
3158 is corrupted due to disk cache incoherency and then replay is
3159 started it is possible to trip an assertion. Avoid putting
3160 committed RPCs into the replay list at all to avoid this issue.
3163 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3165 Description: System hang on startup
3166 Details : This bug allowed the liblustre (e.g. catamount) client to
3167 return to the app before handling all startup RPCs. This
3168 could leave the node unresponsive to lustre network traffic
3169 and manifested as a server ptllnd timeout.
3171 Severity : enhancement
3173 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3174 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
3175 boolean to control whether to print peer debug info when a
3176 client's RPC times out.
3179 Frequency : only for kernels with patches from Lustre below 1.4.3
3181 Description: Remove old rdonly API
3182 Details : Remove old rdonly API which unused from at least lustre 1.4.3
3185 Frequency : only for devices with external journals
3187 Description: Set external device read-only also
3188 Details : During a commanded failover stop, we set the disk device
3189 read-only while the server shuts down. We now also set any
3190 external journal device read-only at the same time.
3193 Frequency : when upgrading from 1.4 while trying to change parameters
3195 Description: The wrong (new) MDC name was used when setting parameters for
3196 upgraded MDT's. Also allows changing of OSC (and MDC)
3197 parameters if --writeconf is specified at tunefs upgrade time.
3200 Frequency : when setting specific ost indicies
3202 Description: QOS code breaks on skipped indicies
3203 Details : Add checks for missing OST indicies in the QOS code, so OSTs
3204 created with --index need not be sequential.
3206 Severity : enhancement
3208 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3209 Details : The uninit_groups feature works in conjunction with the kernel
3210 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3211 up the pass1 processing of e2fsck. This is a read-only feature
3212 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3213 mount filesystems that have had this feature enabled.
3215 Severity : enhancement
3217 Description: Improve multi-block allocation algorithm to avoid fragmentation
3218 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3219 allocation locality and avoid filesystem fragmentation.
3221 ------------------------------------------------------------------------------
3223 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
3225 * Support for kernels:
3226 2.4.21-47.0.1.EL (RHEL 3)
3227 2.6.5-7.283 (SLES 9)
3228 2.6.9-42.0.10.EL (RHEL 4)
3229 2.6.12.6 vanilla (kernel.org)
3230 2.6.16.27-0.9 (SLES 10)
3231 * Recommended e2fsprogs version: 1.39.cfs5
3233 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3237 Frequency : occasional, depends on client load and configuration
3238 Bugzilla : 12181, 12203
3239 Description: data loss for recently-modified files
3241 Details : In some cases it is possible that recently written or created
3242 files may not be written to disk in a timely manner (this should
3243 normally be within 30s unless client IO load is very high).
3244 The problem appears as zero-length files or files that are a
3245 multiple of 1MB in size after a client crash or client eviction
3246 that are missing data at the end of the file.
3248 This problem is more likely to be hit on clients where files are
3249 repeatedly created and unlinked in the same directory, clients
3250 have a large amount of RAM, have many CPUs, the filesystem has
3251 many OSTs, the clients are rebooted frequently, and/or the files
3252 are not accessed by other nodes after being written.
3254 The presence of the problem can be detected by looking at
3255 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
3256 smaller than the second (nr_unused) then dirty files will not
3257 be flushed automatically to disk. "sync; sleep 10" should be
3258 run several times on the node before unmounting it to update
3259 Lustre (this is also safe to run on nodes without this problem).
3261 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
3262 that can cause this same problem, so customers using that kernel
3263 also need to update the kernel in addition to Lustre. In order
3264 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
3266 It is normal that files written just before a client crash (less
3267 than 30s) may not yet have been flushed to disk, even for local
3271 Frequency : frequent on thin XT3 nodes
3273 Description: UUID collision on thin XT3 Linux nodes
3274 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
3275 randomly, since we relied on an insufficiently-seeded PRNG.
3280 Description: OSS hangs after "All ost request buffers busy"
3281 Details : A deadlock between quota and journal operations caused OSS
3282 hangs after printing "All ost request buffers busy."
3285 Frequency : always on liblustre builds
3287 Description: Cleanup compiler warnings on liblustre
3290 Frequency : always on liblustre builds on XT3
3292 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
3296 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
3297 > 32000 subdirectories
3298 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
3299 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
3300 subdirectory count crosses 32000. This will aid e2fsck to
3301 correctly handle more than 32000 subdirectories.
3304 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3306 Description: System hang on startup
3307 Details : This bug allowed the liblustre (e.g. catamount) client to
3308 return to the app before handling all startup RPCs. This
3309 could leave the node unresponsive to lustre network traffic
3310 and manifested as a server ptllnd timeout.
3312 Severity : enhancement
3314 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3315 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
3316 boolean to control whether to print peer debug info when a
3317 client's RPC times out.
3322 Description: make O_SYNC working on 2.6 kernels
3323 Details : 2.6 kernels use different method for mark pages for write,
3324 so need add a code to lustre for O_SYNC work.
3329 Description: Failure to close file and release space on NFS
3330 Details : Put inode details into lock acquired in ll_intent_file_open.
3331 Use mdc_intent_lock in ll_intent_open to properly
3332 detect all kind of errors unhandled by mdc_enqueue.
3337 Description: proc file read during shutdown sometimes raced obd removal,
3339 Details : Add lock to prevent obd access after proc file removal.
3342 Frequency : Only for files larger than 4GB on 32-bit clients.
3344 Description: improperly doing page alignment of locks
3345 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
3346 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3351 Description: RPCs being resent when they shouldn't be
3352 Details : Some RPCs that should not be resent are being resent. This
3353 can cause inconsistencies in the RPC state machine. Do not
3354 resend such requests.
3357 Frequency : rare, only with NFS export
3359 Description: Crash on NFS re-export node
3360 Details : under very unusual load conditions an assertion is hit in
3361 ll_intent_file_open()
3364 Frequency : only if OST filesystem is corrupted
3366 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3367 Details : for a short time RPCs with bulk IO are in the replay list,
3368 but replay of bulk IOs is unimplemented. If the OST filesystem
3369 is corrupted due to disk cache incoherency and then replay is
3370 started it is possible to trip an assertion. Avoid putting
3371 committed RPCs into the replay list at all to avoid this issue.
3376 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3377 Details : Large single O_DIRECT read and write calls can fail to allocate
3378 a sufficiently large buffer to process the request. In case of
3379 allocation failure the allocation is retried with a smaller
3380 buffer and broken into smaller requests.
3382 Severity : enhancement
3384 Description: Add -o localflock option to simulate old noflock behaviour.
3385 Details : This will achieve local-only flock/fcntl locks coherentness.
3390 Description: versioning check is incomplete
3391 Details : Checking the version difference of client vs. server, report
3392 error if the gap is too big.
3397 Description: add support PG_writeback bit
3398 Details : add support for PG_writeback bit for Lustre, for more carefull
3399 work with page cache in 2.6 kernel. This also fix some deadlocks
3400 and remove hack for work O_SYNC with 2.6 kernel.
3402 Severity : enhancement
3404 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3405 Details : The uninit_groups feature works in conjunction with the kernel
3406 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3407 up the pass1 processing of e2fsck. This is a read-only feature
3408 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3409 mount filesystems that have had this feature enabled.
3411 Severity : enhancement
3413 Description: Improve multi-block allocation algorithm to avoid fragmentation
3414 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3415 allocation locality and avoid filesystem fragmentation.
3417 ------------------------------------------------------------------------------
3419 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3421 * Support for kernels:
3422 2.6.9-42.0.3.EL (RHEL 4)
3423 2.6.5-7.276 (SLES 9)
3424 2.4.21-47.0.1.EL (RHEL 3)
3425 2.6.12.6 vanilla (kernel.org)
3426 2.6.16.21-0.8 (SLES10)
3427 * Recommended e2fsprogs version: 1.39.cfs2-0
3429 * The backwards-compatible /proc/sys/portals symlink has been removed
3430 in this release. Before upgrading, please ensure that you change
3431 any configuration scripts or /etc/sysctl.conf files that access
3432 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3433 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3434 in advance of the upgrade on any system running Lustre 1.4.6 or
3435 newer, since /proc/sys/lnet was added in that version.
3436 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3440 Frequency : only when quota is used
3442 Description: avoid scanning export list for quota master
3443 Details : Change the algorithms to avoid scanning export list in order
3444 to improve the efficiency.
3447 Frequency : MDS failover only, very rarely
3449 Description: "went back in time" messages on mds failover
3450 Details : The greatest transno may be lost when the current operation
3451 finishes with an error (transno==0) and the client's last_rcvd
3452 record is over-written. Save the greatest transno in the
3453 mds_last_transno for this case.
3456 Frequency : always for specific kernels and striping counts
3458 Description: client may get "Matching packet too big" without ACL support
3459 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
3460 when trying to access files in certain configurations. The
3461 clients should in fact be denied when mounting because they do
3462 not understand ACLs.
3465 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
3467 Description: many clients connecting with IO in progress causes connect timeouts
3468 Details : Avoid synchronous journal commits to avoid delays caused by many
3469 clients connecting/disconnecting when bulk IO is in progress.
3470 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
3471 OST_IO_PORTAL to avoid delays behind potentially many pending
3475 Frequency : occasionally with multiple writers to a single file
3477 Description: shared writes to file may result in wrong size reported by stat()
3478 Details : Allow growing of kms when extent lock is cancelled
3481 Frequency : always with random mmap IO to multi-striped file
3483 Description: mmap write might be lost if we are writing to a 'hole' in stripe
3484 Details : Only if the hole is at the end of OST object so that kms is too
3485 small. Fix is to increase kms accordingly in ll_nopage.
3488 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
3490 Description: writes to a missing object would leak memory on the OST
3491 Details : If there is an inconsistency between the MDS and OST filesystems,
3492 such that the MDS references an object that doesn't exist, writes
3493 to that object will leak memory due to incorrect cleanup in the
3494 error handling path, eventually running out of memory on the OST.
3499 Description: Creating too long symlink causes lustre errors
3500 Details : Check symlink and name lengths before sending requests to MDS.
3503 Frequency : only if flock is enabled (not on by default)
3505 Description: posix locks not released on fd closure on 2.6.9+
3506 Details : We failed to add posix locks to list of inode locks on 2.6.9+
3507 kernels, this caused such locks not to be released on fd close and
3508 then assertions on fs unmount about still used locks.
3511 Frequency : MDS failover only, very rarely
3513 Description: clients may get ASSERTION(granted_lock != NULL)
3514 Details : When request was taking a long time, and a client was resending
3515 a getattr by name lock request. The were multiple lock requests
3516 with the same client lock handle and
3517 mds_getattr_name->fixup_handle_for_resent_request found one of the
3518 lock handles but later failed with ASSERTION(granted_lock != NULL).
3523 Description: handle->h_buffer_credits > 0, assertion failure
3524 Details : h_buffer_credits is zero after truncate, causing assertion
3525 failure. This patch extends the transaction or creates a new
3529 Frequency : NFS re-export or patchless client
3530 Bugzilla : 11179, 10796
3531 Description: Crash on NFS re-export node (__d_move)
3532 Details : We do not want to hash the dentry if we don't have a lock.
3533 But if this dentry is later used in d_move, we'd hit uninitialised
3534 list head d_hash, so we just do this to init d_hash field but
3535 leave dentry unhashed.
3538 Frequency : NFS re-export or patchless client
3540 Description: NFS exports has problem with symbolic link
3541 Details : lustre client didn't properly install dentry when re-exported
3542 to NFS or running patchless client.
3545 Frequency : NFS re-export or patchless client
3547 Description: Various nfs/patchless fixes.
3548 Details : fixes reuse disconected alias for lookup process - this fixes
3549 warning "find_exported_dentry: npd != pd",
3550 fix permission error with open files at nfs.
3551 fix apply umask when do revalidate.
3554 Frequency : occasional
3556 Description: Crash on NFS re-export node
3557 Details : calling clear_page() on the wrong pointer triggered oops in
3558 generic_mapping_read().
3561 Frequency : rarely, using O_DIRECT IO
3563 Description: unaligned directio crashes client with LASSERT
3564 Details : check for unaligned buffers before trying any requests.
3567 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
3569 Description: stale data returned from RAID cache
3570 Details : If only a small amount of IO is done to the RAID device before
3571 reading it again it is possible to get stale data from the RAID
3572 cache instead of reading it from disk.
3575 Frequency : always for sles10 kernel
3577 Description: sles10 support
3578 Details : ll_follow_link: compile fixes and using of nd_set_link
3579 under newer kernels.
3582 Frequency : depends on arch, kernel and compiler version, always on sles10
3585 Description: recursive or deep enough symlinks cause stack overflow
3586 Details : getting rid of large stack-allocated variable in
3590 Frequency : depends on hardware
3592 Description: lustre write performance loss in the SLES10 kernel
3593 Details : the performance loss is caused by using of write barriers in the
3594 ext3 code. The SLES10 kernel turns barrier support on by
3595 default. The fix is to undo that change for ldiskfs.
3597 ------------------------------------------------------------------------------
3599 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
3601 * Support for kernels:
3602 2.6.9-42.0.3EL (RHEL 4)
3603 2.6.5-7.276 (SLES 9)
3604 2.4.21-47.0.1.EL (RHEL 3)
3605 2.6.12.6 vanilla (kernel.org)
3609 Frequency : quota enabled and large files being deleted
3611 Description: releasing more than 4GB of quota at once hangs OST
3612 Details : If a user deletes more than 4GB of files on a single OST it
3613 will cause the OST to spin in an infinite loop. Release
3614 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3619 Description: statfs data retrieved from /proc may be stale or zero
3620 Details : When reading per-device statfs data from /proc, in the
3621 {kbytes,files}_{total,free,avail} files, it may appear
3622 as zero or be out of date.
3625 Frequency : systems with MD RAID1 external journal devices
3627 Description: lconf's call to blkid is confused by RAID1 journal devices
3628 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3629 of returning all block devices that match the journal UUID.
3632 Frequency : always, for aggregate stripe size over 4GB
3634 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
3635 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
3636 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
3639 Frequency : always if "lfs find" used on a local file/directory
3641 Description: "lfs find" segfaults if used on a local file/directory
3642 Details : The case where a directory component was not specified wasn't
3643 handled correctly. Handle this properly.
3646 Frequency : always on ppc64
3648 Description: the write to an ext3 filesystem mounted with mballoc got stuck
3649 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3650 perform endianness conversion.
3653 Frequency : rarely (truncate to non-zero file size after write under load)
3654 Bugzilla : 10730, 10687
3655 Description: Files padded with zeros to next 4K multiple
3656 Details : With filesystems mounted using the "extents" option (2.6 kernels)
3657 it is possible that files that are truncated to a non-zero size
3658 immediately after being written are filled with zero bytes beyond
3659 the truncated size. No file data is lost.
3661 Severity : enhancement
3663 Description: Allow recovery/failover for liblustre clients.
3664 Details : liblustre clients were unaware of failover configurations until
3667 Severity : enhancement
3669 Description: user file locks should fail when not mounting with flock option
3670 Details : Set up an error-returning stub in ll_file_operations.lock field
3671 to prevent incorrect behaviour when client is mounted without
3672 flock option. Also, set up properly f_op->flock field for
3676 Frequency : always on ia64
3678 Description: "lfs df" loops on printing out MDS statfs information
3679 Details : The obd_ioctl_data was not initialized and in some systems
3680 this caused a failure during the ioctl that did not return
3681 an error. Initialize the struct and return an error on failure.
3684 Frequency : SLES 9 only
3686 Description: Error of copying files with lustre special EAs as root
3687 Details : Client side always return success for setxattr call for lustre
3688 special xattr (currently only "trusted.lov").
3691 Frequency : rarely on clusters with both ia64+i386 clients
3693 Description: ia64+i686 clients doing shared IO on the same file may LBUG
3694 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
3695 clients are doing concurrent writes to the same file it is
3696 possible that the ia64 clients may LASSERT because the OST
3697 extent locks are not PAGE_SIZE aligned. Ensure that grown
3698 locks are always aligned on the request boundary.
3701 Frequency : specific use, occasional
3703 Description: Overwriting in use executable truncates on-disk binary image
3704 Details : If one node attempts to overwrite an executable in use by
3705 another node, we now correctly return ETXTBSY instead of
3706 truncating the file.
3708 Severity : enhancement
3710 Description: Async OSC create to avoid the blocking unnecessarily.
3711 Details : If a OST has no remain object, system will block on the creating
3712 when need to create a new object on this OST. Now, ways use
3713 pre-created objects when available, instead of blocking on an
3714 empty osc while others are not empty. If we must block, we block
3715 for the shortest possible period of time.
3720 Description: chmod on Lustre root is propagated to other clients
3721 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
3722 invalid by the follow_mount time.
3725 Frequency : liblustre clients only
3727 Description: Race in 'instant cancel' lock handling could lead to such locks
3728 never to be granted in case of SMP MDS
3729 Details : Do not destroy not yet granted but cbpending locks in
3733 Frequency : replay/resend of open
3735 Description: non null lock assetion failure in mds_intent_policy
3736 Details : Trying to replay/resend lockless open requests resulted in
3737 mds_open() returning 0 with no lock. Now it sets a flag if
3738 a lock is going to be returned.
3740 Severity : enhancement
3742 Description: Checksum enhancements
3743 Details : New checksum enhancements allow for resending RPCs that failed
3746 Severity : enhancement
3748 Description: Tunables on number of dirty pages in cacche
3749 Details : Allow to set limit on number of dirty pages cached.
3754 Description: client crash on unmount - lock still has references
3755 Details : In some error handling cases it was possible to leak a lock
3756 reference on a client while accessing a file. This was not
3757 harmful to the client during operation, but would cause the
3758 client to crash when the filesystem is unmounted.
3761 Frequency : specific case, rare
3763 Description: ETXTBSY on mds though file not in use by client
3764 Details : ETXTBSY is no longer incorrectly returned when attempting to
3765 chmod or chown a directory that the user previously tried to
3766 execute or a currently-executing binary.
3769 Frequency : extremely rare except on liblustre-based clients
3771 Description: Lustre space not freed when files are deleted
3772 Details : Clean up open-unlinked files after client eviction. Previously
3773 the unlink was skipped and the files remained as orphans.
3778 Description: OST failure "would be an LBUG" in waiting_locks_callback()
3779 Details : In some cases it was possible to send a blocking callback to
3780 a client doing a glimpse, even though that client didn't get
3781 a lock granted. When the glimpse lock is cancelled on the OST
3782 the freed lock is left on the waiting list and corrupted the list.
3785 Frequency : all core dumps
3787 Description: Broke core dumps to lustre
3788 Details : Negative dentry may be unhashed if parent does not have UPDATE
3789 lock, but some callers, e.g. do_coredump, expect dentry to be
3790 hashed after successful create, hash it in ll_create_it.
3792 ------------------------------------------------------------------------------
3794 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
3796 * Support for kernels:
3797 2.6.9-42.0.2.EL (RHEL 4)
3798 2.6.5-7.276 (SLES 9)
3799 2.4.21-40.EL (RHEL 3)
3800 2.6.12.6 vanilla (kernel.org)
3804 Frequency : always on RHEL 3
3806 Description: Number of open files grows over time
3807 Details : The number of open files grows over time, whether or not
3808 Lustre is started. This was due to a filp leak introduced
3809 by one of our kernel patches.
3811 ------------------------------------------------------------------------------
3813 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
3815 * Support for kernels:
3816 2.6.9-42.EL (RHEL 4)
3817 2.6.5-7.276 (SLES 9)
3818 2.4.21-40.EL (RHEL 3)
3819 2.6.12.6 vanilla (kernel.org)
3824 Bugzilla : 5719, 9635, 9792, 9684
3825 Description: OST (or MDS) trips assertions in (re)connection under heavy load
3826 Details : If a server is under heavy load and cannot reply to new
3827 connection requests before the client resends the (re)connect,
3828 the connection handling code can behave badly if two service
3829 threads are concurrently handing separate (re)connections from
3830 the same client. Add better locking to the connection handling
3831 code, and ensure that only a single connection will be processed
3832 for a given client UUID, even if the lock is dropped.
3834 Severity : enhancement
3836 Description: add TCP zero-copy support to kernel
3837 Details : Add support to the kernel TCP stack to allow zero-copy bulk
3838 sends if the hardware supports scatter-gather and checksumming.
3839 This allows socklnd to do client-write and server-read more
3840 efficiently and reduce CPU utilization from skbuf copying.
3843 Frequency : only if NFS exporting from client
3845 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
3846 Details : If setting attributes on a file created under NFS that had
3847 never been opened it would be possible to oops the client
3848 if the file had no objects.
3851 Frequency : always for liblustre
3853 Description: liblustre client does MDS+OSTs setattr RPC for each write
3854 Details : When doing a write from a liblustre client, the client
3855 incorrectly issued an RPC to the MDS and each OST the file was
3856 striped over in order to update the timestamps. When writing
3857 with small chunks and many clients this could overwhelm the MDS
3858 with RPCs. In all cases it would slow down the write because
3859 these RPCs are unnecessary.
3861 Severity : enhancement
3863 Description: allow number of MDS service threads to be changed at module load
3864 Details : It is now possible to change the number of MDS service threads
3865 running. Adding "options mds mds_num_threads={N}" to the MDS's
3866 /etc/modprobe.conf will set the number of threads for the next
3867 time Lustre is restarted (assuming the "mds" module is also
3868 reloaded at that time). The default number of threads will
3869 stay the same, 32 for most systems.
3874 Description: OST crash if filesystem is unformatted or corrupt
3875 Details : If an OST is started on a device that has never been formatted
3876 or if the filesystem is corrupt and cannot even mount then the
3877 error handling cleanup routines would dereference a NULL pointer.
3882 Description: NULL pointer deref in llap_from_page.
3883 Details : get_cache_page_nowait can return a page with NULL (or otherwise
3884 incorrect) mapping if the page was truncated/reclaimed while it was
3885 searched for. Check for this condition and skip such pages when
3886 doing readahead. Introduce extra check to llap_from_page() to
3887 verify page->mapping->host is non-NULL (so page is not anonymous).
3890 Frequency : Sometimes when using sys_sendfile
3892 Description: "page not covered by a lock" warnings from ll_readpage
3893 Details : sendfile called ll_readpage without right page locks present.
3894 Now we introduced ll_file_sendfile that does necessary locking
3895 around call to generic_file_sendfile() much like we do in
3899 Frequency : with certain MDS communication failures at client mount time
3901 Description: NULL pointer deref after failed client mount
3902 Details : a client connection request may delayed by the network layer
3903 and not be sent until after the PTLRPC layer has timed out the
3904 request. If the client fails the mount immediately it will try
3905 to clean up before the network times out the request. Add a
3906 reference from the request import to the obd device and delay
3907 the cleanup until the network drops the request.
3910 Frequency : occasionally during client (re)connect
3912 Description: assertion failure during client (re)connect
3913 Details : processing a client connection request may be delayed by the
3914 client or server longer than the client connect timeout. This
3915 causes the client to resend the connection request. If the
3916 original connection request is replied in this interval, the
3917 client may trip an assertion failure in ptlrpc_connect_interpret()
3918 which thought it would be the only running connect process.
3921 Frequency : only with obd_echo servers and clients that are rebooted
3923 Description: kernel BUG accessing uninitialized data structure
3924 Details : When running an obd_echo server it did not start the ping_evictor
3925 thread, and when a client was evicted an uninitialized data
3926 structure was accessed. Start the ping_evictor in the RPC
3927 service startup instead of the OBD startup.
3929 Severity : enhancement
3930 Bugzilla : 10193 (patchless)
3931 Description: Remove dependency on various unexported kernel interfaces.
3932 Details : No longer need reparent_to_init, exit_mm, exit_files,
3933 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
3936 Frequency : rare (only users of deprecated and unsupported LDAP config)
3938 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
3939 Details : LDAP apparently contains 'lustreName' attributes instead of
3940 'name'. A simple remapping of the name is sufficient.
3943 Frequency : rare (only with non-default dump_on_timeout debug enabled)
3945 Description: waiting_locks_callback trips kernel BUG if client is evicted
3946 Details : Running with the dump_on_timeout debug flag turned on makes
3947 it possible that the waiting_locks_callback() can try to dump
3948 the Lustre kernel debug logs from an interrupt handler. Defer
3949 this log dumping to the expired_lock_main() thread.
3951 Severity : enhancement
3953 Description: Support NFS exporting on 2.6 kernels.
3954 Details : Implement non-rawops metadata methods for NFS server to use without
3955 changing NFS server code.
3958 Frequency : very rare (synthetic metadata workload only)
3960 Description: two racing renames might cause an MDS thread to deadlock
3961 Details : Running the "racer" program may cause one MDS thread to rename
3962 a file from being the source of a rename to being the target of
3963 a rename at exactly the same time that another thread is doing
3964 so, and the second thread has already enqueued these locks after
3965 doing a lookup of the target and is trying to relock them in
3966 order. Ensure that we don't try to re-lock the same resource.
3969 Frequency : only very large systems with liblustre clients
3971 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
3972 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
3973 all OSTs in parallel. This allows the eviction of stale liblustre
3974 clients to proceed much faster than if they were done in series,
3975 and also offers similar improvements for other set_info RPCs.
3980 Description: excessive CPU usage during initial read phase on client
3981 Details : During the initial read phase on a client, it would agressively
3982 retry readahead on the file, consuming too much CPU and impacting
3983 performance (since 1.4.5.8). Improve the readahead algorithm
3984 to avoid this, and also improve some other common cases (read
3985 of small files in particular, where "small" is files smaller than
3986 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
3991 Description: MDS crash when receiving packet with unknown intent.
3992 Details : Do not LBUG in unknown intent case, just return -EFAULT
3994 Severity : enhancement
3995 Bugzilla : 9293, 9385
3996 Description: MDS RPCs are serialised on client. This is unnecessary for some.
3997 Details : Do not serialize getattr (non-intent version) and statfs.
4000 Frequency : occasional, when OST network is overloaded/intermittent
4002 Description: client evicted by OST after bulk IO timeout
4003 Details : If a client sends a bulk IO request (read or write) the OST
4004 may evict the client if it is unresposive to its data GET/PUT
4005 request. This is incorrect if the network is overloaded (takes
4006 too long to transfer the RPC data) or dropped the OST GET/PUT
4007 request. There is no need to evict the client at all, since
4008 the pinger and/or lock callbacks will handle this, and the
4009 client can restart the bulk request.
4012 Frequency : Always when mmapping file with no objects
4014 Description: client crashes when mmapping file with no objects
4015 Details : Check that we actually have objects in a file before doing any
4016 operations on objects in ll_vm_open, ll_vm_close and
4022 Description: Request leak when working with deleted CWD
4023 Details : Introduce advanced request refcount tracking for requests
4024 referenced from lustre intent.
4026 Severity : Enhancement
4028 Description: Cache open file handles on client.
4029 Details : MDS now will return special lock along with openhandle, if
4030 requested and client is allowed to hold openhandle, even if unused,
4031 until such a lock is revoked. Helps NFS a lot, since NFS is opening
4032 closing files for every read/write openration.
4034 Severity : Enhancement
4036 Description: Cache open negative dentries on client when possible.
4037 Details : Guard negative dentries with UPDATE lock on parent dir, drop
4038 negative dentries on lock revocation.
4043 Description: Remounting a client read-only wasn't possible with a zconf mount
4044 Details : It wasn't possible to remount a client read-only with llmount.
4046 Severity : enhancement
4047 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
4048 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
4049 MPICH is included to add Lustre-specific ADIO interfaces.
4050 This is based closely on the UFS ADIO layer and only differs
4051 in file creation, in order to allow the OST striping to be set.
4052 This is user-contributed code and not supported by CFS.
4057 Description: extended inode attributes (immutable, append-only) work improperly
4058 when 2.4 and 2.6 kernels are used on client/server or vice versa
4059 Details : Introduce kernel-independent values for these flags.
4061 Severity : enhancement
4064 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
4065 Details : Many of the /proc/ tunables can only be tuned at a megabyte
4066 granularity. Now, Fractional MB granularity is be supported,
4067 this is very useful for low memory system.
4069 Severity : enhancement
4071 Description: Getattr by fid
4072 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
4073 locks, avoids extra getattr rpc requests to MDS, allows '/' to
4074 have locks and avoids getattr rpc requests for it on every stat.
4077 Frequency : Always, for filesystems larger than 2TB
4079 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4080 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
4081 filesystems larger than 2TB to be mounted with mballoc enabled.
4084 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
4086 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4087 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
4088 filesystems larger than 2TB is mounted then the kernel will
4089 silently corrupt the start of the filesystem. CONFIG_LBD is
4090 enabled for all CFS-supported kernels, but the possibility of
4091 this happening with a modified kernel config exists.
4093 Severity : enhancement
4095 Description: add client O_DIRECT support for 2.6 kernels
4096 Details : It is now possible to do O_DIRECT reads and writes to files
4097 in the Lustre client mountpoint on 2.6 kernel clients.
4099 Severity : enhancement
4101 Description: parallel glimpse, setattr, statfs, punch, destroy requests
4102 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
4103 parallel, not waiting for response from every OST before sending
4104 a rpc to the next OST.
4109 Description: setattr vs write race when updating file timestamps
4110 Details : Client processes that update a file timestamp into the past
4111 right after writing to the file (e.g. tar) it is possible that
4112 the updated file modification time can be reset to the current
4113 time due to a race between processing the setattr and write RPC.
4115 Severity : enhancement
4117 Description: Bring 'lfs find' closer in line with regular Linux find.
4118 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
4119 -print0 options and obtains all the needed info through the lustre
4122 Severity : enhancement
4124 Description: support up to 1024 configured devices on one node
4125 Details : change obd_dev array from statically allocated to dynamically
4126 allocated structs as they are first used to reduce memory usage
4131 Description: Flush dirty partially truncated pages during truncate
4132 Details : Immediatelly flush partially truncated pages in filter_setattr,
4133 this way we completely avoid having any pages in page cache on OST
4134 and can retire ugly workarounds during writes to flush such pages.
4139 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
4140 Details : protect inode from truncation within vfs_unlink() context
4141 just take a reference before calling vfs_unlink() and release it
4142 when parent's i_sem is free.
4147 Description: last_id value checked outside lock on OST caused LASSERT failure
4148 Details : If there were multiple MDS->OST object precreate requests in
4149 flight, it was possible that the OST's last object id was checked
4150 outside a lock and incorrectly tripped an assertion. Move checks
4151 inside locks, and discard old precreate requests.
4154 Frequency : always, if extents are used on OSTs
4156 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
4157 Details : OSTs using the extents format would not zero the high 16 bits of
4158 the index physical block number. This is not a problem for any
4159 OST filesystems smaller than 16TB, and no kernels support ext3
4160 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
4161 new/modified files) and can be fixed for existing filesystems
4162 with e2fsprogs-1.39-cfs1.
4167 Description: import connection selection may be incorrect if timer wraps
4168 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
4169 import connections to be ignored if the 32-bit jiffies counter
4170 wraps. Use a 64-bit jiffies counter.
4173 Frequency : very large clusters immediately after boot
4175 Description: LNET request buffers exhausted under heavy short-term load
4176 Details : If a large number of client requests are generated on a service
4177 that has previously never seen so many requests it is possible
4178 that the request buffer growth cannot keep up with the spike in
4179 demand. Instead of dropping incoming requests, they are held in
4180 the LND until the RPC service can accept more requests.
4183 Frequency : Sometimes during replay
4185 Description: Assertion failure in ll_local_open after replay.
4186 Details : If replay happened on an open request reply before we were able
4187 to set replay handler, reply will become not swabbed tripping the
4188 assertion in ll_local_open. Now we set the handler right after
4189 recognising of open request
4192 Frequency : very rare
4194 Description: kernel reports "badness in vsnprintf"
4195 Details : Reading from the "recovery_status" /proc file in small chunks
4196 may cause a negative length in lprocfs_obd_rd_recovery_status()
4197 call to vsnprintf() (which is otherwise harmless). Exit early
4198 if there is no more space in the output buffer.
4200 Severity : enhancement
4202 Description: clear OBD RPC statistics by writing to them
4203 Details : It is now possible to clear the OBD RPC statistics by writing
4204 to the "stats" file.
4209 Description: Client mtime is not the same on different clients after utimes
4210 Details : In some cases, the client was using the utimes() syscall on
4211 a file cached on another node. The clients now validate the
4212 ctime from the MDS + OSTs to determine which one is right.
4217 Description: Inability to activate failout mode
4218 Details : lconf script incorrectly assumed that in pythong string's numeric
4219 value is used in comparisons.
4222 Frequency : always with multiple stripes per file
4224 Description: Inefficient object allocation for mutli-stripe files
4225 Details : When selecting which OSTs to stripe files over, for files with
4226 a stripe count that divides evenly into the number of OSTs,
4227 the MDS is always picking the same starting OST for each file.
4228 Return the OST selection heuristic to the original design.
4233 Description: mount failures may take full timeout to return an error
4234 Details : Under some heavy load conditions it is possible that a
4235 failed mount can wait for the full obd_timeout interval,
4236 possibly several minutes, before reporting an error.
4237 Instead return an error as soon as the status is known.
4239 Frequency : quota enabled and large files being deleted
4241 Description: releasing more than 4GB of quota at once hangs OST
4242 Details : If a user deletes more than 4GB of files on a single OST it
4243 will cause the OST to spin in an infinite loop. Release
4244 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4249 Description: statfs data retrieved from /proc may be stale or zero
4250 Details : When reading per-device statfs data from /proc, in the
4251 {kbytes,files}_{total,free,avail} files, it may appear
4252 as zero or be out of date.
4255 Frequency : systems with MD RAID1 external journal devices
4257 Description: lconf's call to blkid is confused by RAID1 journal devices
4258 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4259 of returning all block devices that match the journal UUID.
4262 Frequency : always, for aggregate stripe size over 4GB
4264 Description: assertion fails when trying to use 4GB stripe size
4265 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
4266 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
4269 Frequency : always on ppc64
4271 Description: the first write on an ext3 filesystem with mballoc got stuck
4272 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4273 perform endianness conversion.
4275 ------------------------------------------------------------------------------
4277 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4279 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
4280 INTEROPERATE with older versions automatically. Please read the
4281 user documentation before upgrading any part of a live system.
4282 * WARNING: Lustre networking configuration changes are required with
4283 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
4286 * Support for kernels:
4287 2.6.9-22.0.2.EL (RHEL 4)
4288 2.6.5-7.244 (SLES 9)
4289 2.6.12.6 vanilla (kernel.org)
4292 Severity : enhancement
4293 Bugzilla : 7981/8208
4294 Description: Introduced Lustre Networking (LNET)
4295 Details : LNET is new networking infrastructure for Lustre, it includes
4296 a reorganized network configuration mode (see the user
4297 documentation for full details) as well as support for routing
4298 between different network fabrics. Lustre Networking Devices
4299 (LNDs) for the supported network fabrics have also been
4300 created for this new infrastructure.
4302 Severity : enhancement
4303 Description: Introduced Access control lists
4304 Details : clients can set ACLs on files and directories in order to have
4305 more fine-grained permissions than the standard Unix UGO+RWX.
4306 The MDS must be started with the "-o acl" mount option.
4308 Severity : enhancement
4309 Description: Introduced filesystem quotas
4310 Details : Administrators may now establish per-user quotas on the
4313 Severity : enhancement
4315 Description: Configuration change for the XT3
4316 The PTLLND is now used to run Lustre over Portals on the XT3
4317 The configure option(s) --with-cray-portals are no longer used.
4318 Rather --with-portals=<path-to-portals-includes> is used to
4319 enable building on the XT3. In addition to enable XT3 specific
4320 features the option --enable-cray-xt3 must be used.
4325 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
4326 Details : A race between allocating a new llcd and re-getting the llcd_lock
4327 allowed another thread to grab newly-allocated llcd.
4329 Severity : enhancement
4331 Description: 2.6 OST async journal commit and locking fix to improve performance
4332 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
4333 2.6 kernels are now async as they already were in 2.4 kernels so
4334 that they can commit concurrently with the network bulk transfer.
4335 For block-allocated files the filter allocation semaphore is held
4336 to avoid filesystem fragmentation during allocation. BKL lock
4337 removed for 2.6 xattr operations where it is no longer needed.
4342 Description: lconf incorrectly determined whether two IP networks could talk
4343 Details : In some more complicated routing and multiple-network
4344 configurations, lconf will avoid trying to make a network
4345 connection to a disjoint part of the IP space. It was doing the
4346 math incorrectly for one set of cases.
4351 Description: Fix for potential infinite loop processing records in an llog.
4352 Details : If an llog record is corrupted/zeroed, it is possible to loop
4353 forever in llog_process(). Validate the llog record length
4354 and skip the remainder of the block on error.
4357 Frequency : occasional (liblustre only)
4359 Description: liblustre could not open files whose last component is a symlink
4360 Details : sysio_path_walk() would incorrectly pass the open intent to
4361 intermediate path components.
4364 Frequency : rare (liblustre only with non-standard tuning)
4365 Bugzilla : 7201 (7350)
4366 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4367 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4368 after it has been released, only do this for referenced locks
4370 Severity : enhancement
4372 Description: specifying an (invalid) directory default stripe_size of -1
4373 would reset the directory default striping
4374 Details : stripe_size -1 was used internally to signal directory stripe
4375 removal, now use "all default" to signal dir stripe removal
4376 as a directory striping of "all default" is not useful
4379 Frequency : common for large clusters running liblustre clients
4381 Description: doing an ls when liblustre clients are running is slow
4382 Details : sending a glimpse AST to a liblustre client waits for every AST
4383 to time out, as liblustre clients will not respond. Since they
4384 cannot cache data we refresh the OST lock LVB from disk instead.
4386 Severity : enhancement
4388 Description: doing an ls at the same time as file IO can be slow
4389 Details : enqueue and other "small" requests can be blocked behind many
4390 large IO requests. Create a new OST IO portal for non-IO
4391 requests so they can be processed faster.
4394 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4396 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4397 created file with mode 000
4398 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4399 allow an ftruncate() as a non-root user to a file with mode 000.
4400 Lustre now respects this flag to disable mode checks when
4401 truncating a file owned by the user
4404 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4407 Description: Revoking locks from clients that went dead or catatonic might take
4409 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4410 cancellation of such locks instant on servers without waiting for
4411 any reply from clients. Clients drops these locks when cancel
4412 notification from server is received without replying.
4415 Frequency : liblustre-only, when liblustre client dies or becomes busy
4417 Description: Doing ls on Linux clients can take a long time with active
4419 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4420 granting such locks to it in the first place if possible. Locks
4421 are taken by proxy on the OST during the read or write and
4422 dropped immediately afterward. Add connect flags handling, do
4423 not grant locks to liblustre clients for glimpse ASTs.
4425 Severity : enhancement
4427 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4428 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4429 sequential reads and behaves badly with applications doing random
4430 reads. Improve it by reading ahead at least read region, and
4431 avoiding excessive large RPC for small reads.
4433 Severity : enhancement
4435 Description: Creating more than 1000 files for a single job may cause a load
4436 imbalance on the OSTs if there are also a large number of OSTs.
4437 Details : qos_prep_create() uses an OST index reseed value that is an
4438 even multiple of the number of available OSTs so that if the
4439 reseed happens in the middle of the object allocation it will
4440 still utilize the OSTs as uniformly as possible.
4445 Description: OST or MDS may oops in ping_evictor_main()
4446 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
4447 but doesn't restart at beginning of obd_exports_timed list
4450 Severity : enhancement
4452 Description: improve by-nid export eviction on the MDS and OST
4453 Details : allow multiple exports with the same NID to be evicted at one
4454 time without re-searching the exports list.
4457 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
4459 Description: MDS may oops in groups_free()
4460 Details : in rare race conditions a newly allocated group_info struct is
4461 freed again, and this can be NULL. The 2.4 compatibility code
4462 for groups_free() checked for a NULL pointer, but 2.6 did not.
4465 Frequency : common for liblustre clients doing little filesystem IO
4466 Bugzilla : 9352, 7313
4467 Description: server may evict liblustre clients accessing contended locks
4468 Details : if a client is granted a lock or receives a completion AST
4469 with a blocking AST already set it would not reply to the AST
4470 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
4473 Frequency : lfs setstripe, only systems with more than 160 OSTs
4475 Description: unable to set striping with a starting offset beyond OST 160
4476 Details : llapi_create_file() incorrectly limited the starting stripe
4477 index to the maximum single-file stripe count.
4480 Frequency : LDAP users only
4482 Description: lconf did not handle in-kernel recovery with LDAP properly
4483 Details : lconf/LustreDB get_refs() is searching the wrong namespace
4485 Severity : enhancement
4487 Description: bind OST threads to NUMA nodes to improve performance
4488 Details : all OST threads are uniformly bound to CPUs on a single NUMA
4489 node and do their allocations there to localize memory access
4491 Severity : enhancement
4493 Description: llmount can determine client NID directly from Myrinet (GM)
4494 Details : the client NID code from gmnalnid was moved directly into
4495 llmount, removing the need to use this or specifying the
4496 client NID explicitly when mounting GM clients with zeroconf
4499 Frequency : if client is started with down MDS
4501 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
4502 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
4503 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
4504 ptlrpc_import_delayed_req() because we didn't check if the
4505 request was interrupted, and we also didn't break out of the
4506 event loop if there was no timeout
4511 Description: data loss during non-page-aligned writes to a single file from
4512 both multiple nodes and multiple threads on one node at same time
4513 Details : updates to KMS and lsm weren't protected by common lock. Resulting
4514 inconsistency led to false short-reads, that were cached and later
4515 used by ->prepare_write() to fill in partially written page,
4516 leading to data loss.
4519 Frequency : always, if lconf --abort_recovery used
4521 Description: lconf --abort_recovery fails with 'Operation not supported'
4522 Details : lconf was attempting to abort recovery on the MDT device and not
4525 Severity : enhancement
4527 Description: remove cleanup logs
4528 Details : replace lconf-generated cleanup logs with lustre internal
4529 cleanup routines. Eliminates the need for client-cleanup and
4532 Severity : enhancement
4534 Description: add support for EAs (user and system) on lustre filesystems
4535 Details : it is now possible to store extended attributes in the Lustre
4536 client filesystem, and with the user_xattr mount option it
4537 is possible to allow users to store EAs on their files also
4539 Severity : enhancement
4541 Description: Add possibility (config option) to show minimal available OST free
4543 Details : When compiled with --enable-mindf configure option, statfs(2)
4544 (and so, df) will return least minimal free space available from
4545 all OSTs as amount of free space on FS, instead of summary of
4546 free spaces of all OSTs.
4548 Severity : enhancement
4550 Description: do not expand extent locks acquired on OST-side
4551 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
4552 by server: they are not cached anyway.
4555 Frequency : when mmap is used/binaries executed from Lustre
4557 Description: Unmmap pages before throwing them away from read cache.
4558 Details : llap_shrink cache now attempts to unmap pages before discarding
4559 them (if unmapping failed - do not discard). SLES9 kernel has
4560 extra checks that trigger if this unmapping is not done first.
4565 Description: lconf didn't resolve symlinks before checking to see whether a
4566 given mountpoint was already in use
4569 Frequency : when migrating failover services
4570 Bugzilla : 6395, 9514
4571 Description: When migrating a subset of services from a node (e.g. failback
4572 from a failover service node) the remaining services would
4573 time out and evict clients.
4574 Details : lconf --force (implied by --failover) sets the global obd_timeout
4575 to 5 seconds in order to quickly disconnect, but this caused
4576 other RPCs to time out too quickly. Do not change the global
4577 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
4579 Severity : enhancement
4580 Frequency : if MDS is started with down OST
4581 Bugzilla : 9439,5706
4582 Description: Allow startup/shutdown of an MDS without depending on the
4583 availability of the OSTs.
4584 Details : Asynchronously call mds_lov_synchronize during MDS startup.
4585 Add appropriate locking and lov-osc refcounts for safe
4586 cleaning. Add osc abort_inflight calls in case the
4587 synchronize never started.
4590 Frequency : occasional (Cray XT3 only)
4592 Description: root not authorized to access files in CRAY_PORTALS environment
4593 Details : The client process capabilities were not honoured on the MDS in
4594 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
4595 been accessed by an authorized user then root was able to access
4596 the file on the local client also. The root user capabilities
4597 are now allowed on the MDS, as this environment has secure UID.
4600 Frequency : occasional
4602 Description: ldiskfs "too long searching" message happens too often
4603 Details : A debugging message (otherwise harmless) prints too often on
4604 the OST console. This has been reduced to only happen when
4605 there are fragmentation problems on the filesystem.
4610 Description: Division by zero in statfs when all OSCs are inactive
4611 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
4612 lov_statfs divides by value returned by lov_get_stripecnt().
4616 Bugzilla : 9489, 3273
4617 Description: First write from each client to each OST was only 4kB in size,
4618 to initialize client writeback cache, which caused sub-optimal
4619 RPCs and poor layout on disk for the first writen file.
4620 Details : Clients now request an initial cache grant at (re)connect time
4621 and so that they can start streaming writes to the cache right
4622 away and always do full-sized RPCs if there is enough data.
4623 If the OST is rebooted the client also re-establishes its grant
4624 so that client cached writes will be honoured under the grant.
4629 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
4630 Details : Now I/O RPCs go to different portal number and (presumably) fast
4631 lock requests (and glimses) and other RPCs get their own service
4632 threads pool that should be able to service those RPCs
4635 Severity : enhancement
4637 Description: Ability to exchange lustre version between client and servers and
4638 issue warnings at client side if client is too old. Also for
4639 liblustre clients there is ability to refuse connection of too old
4641 Details : New 'version' field is added to connect data structure that is
4642 filled with version info. That info is later checked by server and
4646 Frequency : rare, liblustre only.
4647 Bugzilla : 9296, 9581
4648 Description: Two simultaneous writes from liblustre at offset within same page
4649 might proceed at the same time overwriting eachother with stale
4651 Details : I/O lock withing llu_file_prwv was released too early, before data
4652 actually was hitting the wire. Extended lock-holding time until
4653 server acknowledges receiving data.
4656 Frequency : extremely rare. Never observed in practice.
4658 Description: avoid generating lustre_handle cookie of 0.
4659 Details : class_handle_hash() generates handle cookies by incrementing
4660 global counter, and can hit 0 occasionaly (this is unlikely, but
4661 not impossible, because initial value of cookie counter is
4662 selected randonly). Value of 0 is used as a sentinel meaning
4663 "unassigned handle" --- avoid it. Also coalesce two critical
4664 sections in this function into one.
4666 Severity : enhancement
4668 Description: allow liblustre clients to delegate truncate locking to OST
4669 Details : To avoid overhead of locking, liblustre client instructs OST to
4670 take extent lock in ost_punch() on client's behalf. New connection
4671 flag is added to handle backward compatibility.
4673 Severity : enhancement
4674 Bugzilla : 4928, 7341, 9758
4675 Description: allow number of OST service threads to be specified
4676 Details : a module parameter allows the number of OST service threads
4677 to be specified via "options ost ost_num_threads={N}" in the
4678 OSS's /etc/modules.conf or /etc/modprobe.conf.
4682 Bugzilla : 6146, 9635, 9895
4683 Description: servers crash with bad pointer in target_handle_connect()
4684 Details : In rare cases when a client is reconnecting it was possible that
4685 the connection request was the last reference for that export.
4686 We would temporarily drop the export reference and get a new
4687 one, but this may have been the last reference and the export
4688 was just destroyed. Get new reference before dropping old one.
4690 Severity : enhancement
4691 Frequency : if client is started with failover MDS
4693 Description: Allow multiple MDS hostnames in the mount command
4694 Details : Try to read the configuration from all specified MDS
4695 hostnames during a client mount in case the "primary"
4698 Severity : enhancement
4700 Description: Stop sending data to evicted clients as soon as possible.
4701 Details : Check if the client we are about to send or are sending data to
4702 was evicted already. (Check is done every second of waiting,
4703 for which l_wait_event interface was extended to allow checking
4704 of exit condition at specified intervals).
4707 Frequency : rare, normally only when NFS exporting is done from client
4709 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
4711 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
4714 Frequency : always, if srand() is called before liblustre initialization
4716 Description: Liblustre uses system PRNG disturbing its usage by user application
4717 Details : Introduce internal to lustre fast and high-quality PRNG for
4718 lustre usage and make liblustre and some other places in generic
4719 lustre code to use it.
4721 Severity : enhancement
4722 Bugzilla : 9477, 9557, 9870
4723 Description: Verify that the MDS configuration logs are updated when xml is
4724 Details : Check if the .xml configuration logs are newer than the config
4725 logs stored on the MDS and report an error if this is the case.
4726 Request --write-conf, or allow starting with --old_conf.
4728 Severity : enhancement
4730 Description: Handle symlinks in the path when checking if Lustre is mounted.
4731 Details : Resolve intermediate symlinks when checking if a client has
4732 mounted a filesystem to avoid duplicate client mounts.
4737 Description: lconf can hit an error exception but still return success.
4738 Details : The lconf command catches the Command error exception at the top
4739 level script context and will exit with the associated exit
4740 status, but doesn't ensure that this exit status is non-zero.
4745 Description: failure of ptlrpc thread startup can cause oops
4746 Details : Starting a ptlrpc service thread can fail if there are a large
4747 number of threads or the server memory is very fragmented.
4748 Handle this without oopsing.
4751 Frequency : always, only if liblustre and non-default acceptor port was used
4753 Description: liblustre cannot connect to servers with non-default acceptor port
4754 Details : tcpnal_set_default_params() was not called and was therefore
4755 ignoring the environment varaible TCPNAL_PORT, as well as other
4756 TCPNAL_ environment variables
4761 Description: two objects could be created on the same OST for a single file
4762 Details : If an OST is down, in some cases it was possible to create two
4763 objects on a single OST for a single file. No problems other
4764 than potential performance impact and spurious error messages.
4768 Bugzilla : 5681, 9562
4769 Description: Client may oops in ll_unhash_aliases
4770 Details : Client dcache may become inconsistent in race condition.
4771 In some cases "getcwd" can fail if the current directory is
4777 Description: Inode refcounting problems in NFS export code
4778 Details : link_raw functions used to call d_instantiate without obtaining
4779 extra inode reference first.
4783 Bugzilla : 9942, 9903
4784 Description: Referencing freed requests leading to crash, memleaks with NFS.
4785 Details : We used to require that call to ll_revalidate_it was always
4786 followed by ll_lookup_it. Also with revalidate_special() it is
4787 possible to call ll_revalidate_it() twice for the same dentry
4788 even if first occurence returned success. This fix changes semantic
4789 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
4790 reference on a request referred from the intent.
4791 ll_intent_release() then releases such a request.
4794 Frequency : rare, normally benchmark loads only
4796 Description: unlinked inodes were kept in memory on the client
4797 Details : If a client is repeatedly creating and unlinking files it
4798 can accumulate a lot of stale inodes in the inode slab cache.
4799 If there is no other client load running this can cause the
4800 client node to run out of memory. Instead flush old inodes
4801 from client cache that have the same inode number as a new inode.
4804 Frequency : SLES9 2.6.5 kernel and long filenames only
4805 Bugzilla : 9969, 10379
4806 Description: utime reports stale NFS file handle
4807 Details : SLES9 uses out-of-dentry names in some cases, which confused
4808 the lustre dentry revalidation. Change it to always use the
4812 Frequency : rare, unless heavy write-truncate concurrency is continuous
4813 Bugzilla : 4180, 6984, 7171, 9963, 9331
4814 Description: OST becomes very slow and/or deadlocked during object unlink
4815 Details : filter_destroy() was holding onto the parent directory lock
4816 while truncating+unlinking objects. For very large objects this
4817 may block other threads for a long time and slow overall OST
4818 responsiveness. It may also be possible to get a lock ordering
4819 deadlock in this case, or run out of journal credits because of
4820 the combined truncate+unlink. Solution is to do object truncate
4821 first in one transaction without parent lock, and then do the
4822 final unlink in a new transaction with the parent lock. This
4823 reduces the lock hold time dramatically.
4826 Frequency : rare, 2.4 kernels only
4828 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
4829 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
4830 before unmounting their filesystem in order to avoid deadlock.
4831 The kernel_locked() function in 2.4 kernels only checks whether
4832 the kernel lock is held, not whether it is this process that is
4833 holding it as 2.6 kernels do.
4838 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
4839 Details : The client ptlrpc code may be trying to reconnect to a down
4840 server before a previous connection attempt has timed out.
4841 Increase the reconnect interval to be longer than the connection
4842 timeout interval to avoid sending duplicate connections to
4846 Frequency : echo_client brw_test command
4848 Description: fix echo_client to work with OST preallocated code
4849 Details : OST preallocation code (5137) didn't take echo_client IO path
4850 into account: echo_client calls filter methods outside of any
4851 OST thread and, hence, there is no per-thread preallocated
4852 pages and buffers to use. Solution: hijack pga pages for IO. As
4853 a byproduct, this avoids unnecessary data copying.
4857 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
4858 Description: Client can oops in mdc_commit_close() after open replay
4859 Details : It was possible for the MDS to return an open request with no
4860 transaction number in mds_finish_transno() if the client was
4861 evicted, but without actually returning an error. Clients
4862 would later try to replay that open and may trip an assertion
4863 Simplify the client close codepath, and always return an error
4864 from the MDS in case the open is not successful.
4867 Frequency : rare, 2.6 OSTs only
4869 Description: OST may deadlock under high load on fragmented files
4870 Details : If there was a heavy load and highly-fragmented OST filesystems
4871 it was possible to have all the OST threads deadlock waiting on
4872 allocation of biovecs, because the biovecs were not released
4873 until the entire RPC IO was completed. Instead, release biovecs
4874 as soon as they are complete to ensure forward IO progress.
4876 Severity : enhancement
4878 Description: Support for specifying external journal device at mount
4879 Details : If an OST or MDS device is formatted with an external journal
4880 device, this device major/minor is stored in the ext3 superblock
4881 and may not be valid for failover. Allow detecting and
4882 specifying the external journal at mount time.
4887 Description: Mounting an MDS with pending unlinked files may cause oops
4888 Details : target_finish_recovery() calls mds_postrecov() which returned
4889 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
4890 considers this an error and immediately begins cleaning up the
4891 lov, just after starting the mds_lov process
4893 Severity : enhancement
4895 Description: Implement 'lfs df' to report actual free space on per-OST basis
4896 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
4897 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
4898 usage of objects; '-h' to report in human readable format.
4900 ------------------------------------------------------------------------------
4902 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
4909 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
4910 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
4911 at startup during orphan unlinking. Instead, skip buddy update
4912 before setup, as it will be regenerated after recovery is complete.
4917 Description: If an OST is inactive, its locks might reference stale inodes.
4918 Details : lov_change_cbdata() must iterate over all namespaces, even if
4919 they are inactive to clear inode references from the lock.
4921 Severity : enhancement
4922 Frequency : occasional, if non-standard max_dirty_mb used
4924 Description: Client will block write RPCs if not enough grant
4925 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
4926 then the client will block writes while waiting for another RPC
4927 to complete instead of consuming its dirty limit. With change
4928 we get improved performance when max_dirty_mb is small.
4930 Severity : enhancement
4931 Bugzilla : 3389, 6253
4932 Description: Add support for supplementary groups on the MDS.
4933 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
4934 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
4935 lookups for user supplementary groups into a cache.
4939 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
4940 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
4941 for liblustre O_EXCL is enforced by MDS.
4945 Description: readdir never returns NULL in liblustre.
4946 Details : Corrected llu_iop_getdirentries logic, to return offset of next
4947 dentry in struct dirent.
4951 Frequency : liblustre only, depends on application IO pattern
4952 Description: liblustre clients evicted if not contacting servers
4953 Details : Don't put liblustre clients into the ping_evictor list, so
4954 they will not be evicted by the pinger ever.
4956 Severity : enhancement
4958 Description: Add ability to evict clients by NID from MDS.
4959 Details : By echoing "nid:$NID" string into
4960 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
4961 $NID would be instantly evicted from this MDS and from all active
4962 OSTs connected to it.
4966 Description: Do not query file size twice, somewhat slowing stat(2) calls.
4967 Details : lookup_it_finish() used to query file size from OSTs that was not
4972 Description: service threads change working directory to that of init
4973 Details : Starting lustre service threads may pin the working directory
4974 of the parent thread, making that filesystem busy. Threads
4975 now change to the working directory of init to avoid this.
4979 Frequency : during shutdown only
4980 Description: shutdown with a failed MDS or OST can cause unmount to hang
4981 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
4986 Frequency : 2.6 only
4987 Description: chmod/chown may include an extra supplementary group
4988 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
4989 supplementary group and if none is specified this is used.
4992 Bugzilla : 5479 (6816)
4994 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
4995 Details : If lookup is in progress on a file that is unlinked we might try
4996 to revalidate the inode and fail in revalidate after lookup is
4997 complete and ll_file_open() enqueues the open again but
4998 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
5001 Frequency : always, if lconf --abort_recovery used
5003 Description: lconf --abort_recovery fails with 'Operation not supported'
5004 Details : lconf was attempting to abort recovery on the MDT device and not
5007 ------------------------------------------------------------------------------
5009 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
5014 Frequency : rare (only unsupported configurations with a node running as an
5016 Bugzilla : 6514, 5137
5017 Description: Mounting a Lustre file system on a node running as an OST could
5019 Details : OSTs now preallocates memory needed to write out data at
5020 startup, instead of when needed, to avoid having to
5021 allocate memory in possibly low memory situations.
5022 Specifically, if the file system is mounted on on OST,
5023 memory pressure could force it to try to write out data,
5024 which it needed to allocate memory to do. Due to the low
5025 memory, it would be unable to do so and the node would
5026 become unresponsive.
5028 Severity : enhancement
5030 Description: Addition of lconf --service command line option
5031 Details : lconf now accepts a '--service <arg>' option, which is
5032 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
5034 Severity : enhancement
5036 Description: Failover mode is now the default for OSTs.
5037 Details : By default, OSTs will now run in failover mode. To return to
5038 the old behaviour, add '--failout' to the lmc line for OSTs.
5040 Severity : enhancement
5042 Description: Health checks are now provided for MDS and OSTs
5043 Details : Additional detailed health check information on MSD and OSTs
5044 is now provided through the procfs health_check value.
5047 Frequency : occasional, depends on IO load
5049 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
5050 after numerous create/delete cycles
5051 Details : The ext3 inode allocation policy would not allocate new inodes
5052 very well on the OSTs because there are no new directories
5053 being created. Instead we look for groups with free space if
5054 the parent directories are nearly full.
5059 Description: Network or server problems during mount may cause partially
5060 mounted clients instead of returning an error.
5061 Details : The config llog parsing code may overwrite the error return
5062 code during mount error handling, returning success instead
5068 Description: MDS can fail to allocate large reply buffers
5069 Details : After long uptimes the MDS can fail to allocate large reply
5070 buffers (e.g. zconf client mount config records) due to memory
5071 fragmentation or consumption by the buffer cache. Preallocate
5072 some large reply buffers so that these replies can be sent even
5073 under memory pressure.
5077 Frequency : rare (liblustre)
5078 Description: fsx running with liblustre complained that using truncate() to
5079 extend the file doesn't work. This patch corrects that issue.
5080 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
5081 ATTR_SIZE and lsm use in llu_setattr_raw.
5085 Frequency : rare, only 2.6 kernels
5086 Description: Unusual file access patterns on the MDS may result in inode
5087 data being lost in very rare circumstances.
5088 Details : Bad interaction between the ea-in-inode patch and the "no-read"
5089 code in the 2.6 kernel caused the inode and/or EA data not to
5090 be read from disk, causing single-file corruption.
5094 Frequency : rare, only 2.6 filesystems using extents
5095 Description: Heavy concurrent write and delete load may cause data corruption.
5096 Details : It was possible under high-load situations to have an extent
5097 metadata block in the block device cache from a just-unlinked
5098 file overwrite a newly-allocated data block. We now unmap any
5099 metadata buffers that alias just-allocated data blocks.
5103 Frequency : filesystems with default stripe_count larger than 77
5104 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
5105 Details : lconf specifies an inode size of 4096 bytes when the default
5106 stripe_count is larger than 77. This conflicts with the default
5107 inode density of 1 per 4096 bytes. Allocate smaller inodes in
5108 this case to avoid pinning too much memory for large EAs.
5110 ------------------------------------------------------------------------------
5112 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
5117 Frequency : rare (extremely heavy IO load with hundreds of clients)
5119 Description: Client is evicted, gets IO error writing to file
5120 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
5121 caused clients to be evicted for AST timeouts. The fixes in
5122 bug 5192 mean we no longer need to have such short AST timeouts
5123 so ldlm_timeout has been increased.
5126 Frequency : occasional during --force or --failover shutdown under load
5127 Bugzilla : 5949, 4834
5128 Description: Server oops/LBUG if stopped with --force or --failover under load
5129 Details : a collection of import/export refcount and cleanup ordering
5130 issues fixed for safer force cleanup
5133 Frequency : only filesystems larger than 120 OSTs
5134 Bugzilla : 5990, 6223
5135 Description: lfs getstripe would oops on a very large filesystem
5136 Details : lov_getconfig used kfree on vmalloc'd memory
5139 Frequency : only filesystems exporting via NFS to Solaris 10 clients
5140 Bugzilla : 6242, 6243
5141 Description: reading from files that had been truncated to a non-zero size
5142 but never opened returned no data
5143 Details : ll_file_read() reads zeros from no-object files to EOF
5148 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
5149 Details : There's little harm in aborting MDS/OSS recovery and letting it
5150 try again, so I removed the LASSERT and return an error instead.
5152 Severity : enhancement
5154 Description: New debugging infrastructure for tracking down data corruption
5155 Details : The I/O checksum code was replaced to: (a) control it at runtime,
5156 (b) cover more of the client-side code path, and (c) try to narrow
5157 down where problems occurred
5161 Bugzilla : 3819, 4364, 4397, 6313
5162 Description: Racing close and eviction MDS could cause assertion in mds_close
5163 Details : It was possible to get multiple mfd references during close and
5164 client eviction, leading to one thread referencing a freed mfd.
5166 Severity: : enhancement
5167 Bugzilla : 3262, 6359
5168 Description: Attempts to reconnect to servers are now more aggressive.
5169 Details : This builds on the enhanced upcall-less recovery that was added
5170 in 1.4.2. When trying to reconnect to servers, clients will
5171 now try each server in the failover group every 10 seconds. By
5172 default, clients would previously try one server every 25 seconds.
5177 Description: After recovery, certain operations trigger a failed
5178 assertion on a client.
5179 Details : Failing over an mds, using lconf -d --failover, while a
5180 client was doing a readdir() call would cause the client to
5181 LBUG after recovery completed and the readdir() was resent.
5183 Severity : enhancement
5185 Description: Default groups are now added by lconf
5186 Details : You can now run lconf --group <servicename> without having to
5187 manually add groups with lmc.
5190 Frequency : occasional
5192 Description: Nodes with an elan id of 0 trigger a failed assertion
5195 Frequency : always when accessing e.g. tty/console device nodes
5197 Description: tty and some other devices nodes cannot be used on lustre
5198 Details : file's private_data field is used by device data and lustre
5199 values in there got lost. New field was added to struct file to
5200 store fs-specific private data.
5203 Frequency : when exporting Lustre via NFS
5205 Description: NFSD failed occasionally when looking up a path component
5206 Details : NFSD is looking up ".." which was broken in ext3 directories
5207 that had grown large enough to become hashed.
5210 Frequency : Clusters with multiple interfaces not on the same subnet
5212 Description: Nodes will repeatedly try to reconnect to an interface which it
5213 cannot reach and report an error to the log.
5214 Details : Extra peer list entries will be created by lconf with some peers
5215 unreachable. lconf now validates the peer before adding it.
5218 Frequency : Only if a default stripe is set on the filesystem root.
5220 Description: Setting a default stripe on the filesystem root prevented the
5221 filesystem from being remounted.
5222 Details : The client was sending extra request flags in the root getattr
5223 request and did not allocate a reply buffer for the dir EA.
5226 Frequency : occasional, higher if lots of files are accessed by one client
5227 Bugzilla : 6159, 6097
5228 Description: Client trips assertion regarding lsm mismatch/magic
5229 Details : While revalidating inodes the VFS looks up inodes with ifind()
5230 and in rare cases can find an inode that is being freed.
5231 The ll_test_inode() code will free the lsm during ifind()
5232 when it finds an existing inode and then the VFS later attaches
5233 this free lsm to a new inode.
5237 Bugzilla : 6422, 7030
5238 Description: MDS deadlock between mkdir and client eviction
5239 Details : Creating a new file via mkdir or mknod (starting a transaction
5240 and getting the ns lock) can deadlock with client eviction
5241 (gets ns lock and trying to finish a synchronous transaction).
5244 Frequency : occasional
5245 Description: While starting a server, the fsfilt_ext3 module could not be
5247 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
5248 kernels. Previously, lconf would still use the ext3 name
5249 when trying to load modules. Now, it will correctly use
5250 ext3 on 2.4 and ldiskfs on 2.6.
5252 Severity : enhancement
5253 Description: The default stripe count has been changed to 1
5254 Details : The interpretation of the default stripe count (0, to lfs
5255 or lmc) has been changed to mean striping across a single
5256 OST, rather than all available. For general usage we have
5257 found a stripe count of 1 or 2 works best.
5259 Severity : enhancement
5260 Description: Add support for compiling against Cray portals.
5261 Details : Conditional compiling for some areas that are different
5265 Frequency : occasional
5266 Bugzilla : 6409, 6834
5267 Description: Creating files with an explicit stripe count may lead to
5268 a failed assertion on the MDS
5269 Details : If some OSTs are full or unavailable, creating files may
5270 trigger a failed assertion on the MDS. Now, Lustre will
5271 try to use other servers or return an error to the
5275 Frequency : occasional
5277 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
5278 caused lock timeout during readahead (since 1.4.2).
5279 Details : Processes doing readahead might match a lock that hasn't been
5280 granted yet if there are overlapping and conflicting lock
5281 requests. The readahead process waits on ungranted lock
5282 (original lock is CBPENDING), while OST waits for that process
5283 to cancel CBPENDING read lock and eventually evicts client.
5285 Severity : enhancement
5287 Description: Initial enabling of flock support for clients
5288 Details : Implements fcntl advisory locking and file status functions.
5289 This feature is provided as an optional mount flag (default
5290 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
5291 locking are implemented yet, and those that are are not guaranteed
5292 to be completely correct in production environments.
5293 mount -t lustre -o [flock|noflock] ...
5296 Frequency : occasional
5298 Description: OSTs running 2.4 kernels but with extents enabled might trip an
5299 assertion in the ext3 JBD (journaling) layer.
5300 Details : The b_committed_data struct is protected by the big kernel lock
5301 in 2.4 kernels, serializing journal_commit_transaction() and
5302 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
5303 access to this struct. In 2.6 kernels there is finer grained
5304 locking to improve SMP performance of the JBD layer.
5308 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
5313 Description: merge and cleanup kernel patches.
5316 -----------------------------------------------------------------------------
5318 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
5320 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
5321 versions of Lustre. Please update all servers and clients to
5322 version 1.4.2 or later at the same time. You must also run
5323 "lconf --write-conf {config}.xml" on the MDS while it is stopped
5324 to update the configuration logs.
5326 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
5327 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
5328 - swab llog records of type '0' so we get proper header size/idx (5861)
5329 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
5330 - fix rename of one directory over another leaking an inode (5953)
5331 - avoid SetPageDirty on 2.6 (5981)
5332 - don't re-add just-being-destroyed locks to the waiting list (5653)
5333 - when creating new directories, inherit the parent's custom
5334 striping settings if present parent (3048)
5335 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
5336 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
5337 - don't hold client locks on temporary worklist from l_lru (5666)
5338 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
5339 - automatically evict dead clients (5921)
5340 - Update file size properly in create+truncate+fstat case (6196)
5341 - Do not unhash mountpoint dentries, do not allow removal of
5343 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
5344 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
5345 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
5346 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
5347 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
5348 - Correctly handle reads of files with no objects (6243)
5349 - lctl recover will also mark a device active if deactivate used (5933)
5351 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
5352 - allow --write-conf on an MDS with different nettype than client (5619)
5353 - don't write config llogs to MDS for mounts not from that MDS (5617)
5354 - lconf should create multiple TCP connections from a client (5201)
5355 - init scripts are now turned off by default; run chkconfig --on
5356 lustre and chkconfig --on lustrefs to use them
5357 - upcalls are no longer needed for clients to recover to failover
5359 - add --abort-recovery option to lconf to abort recovery on device
5361 - add support for an arbitrary number of OSTs (3026)
5362 - Quota support protocol changes.
5363 - forward compatibility changes to wire structs (6007)
5364 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5365 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5366 - improved "lustre status" script
5367 - initialize blocksize for non-regular files (6062)
5368 - added --disable-server and --disable-client configure options (5782)
5369 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5370 - Vanilla 2.4.29 support
5371 - increase maximum number of obd devices to 520 (6242)
5372 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5373 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5374 SLES 9 2.6.5 kernels
5375 - limit stripes per file to 160 (the maximum EA size) (6093)
5377 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5380 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5381 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5382 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5383 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5384 - small ext3 extents cleanups and fixes (5733)
5385 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5386 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5387 - avoid needless client->OST connect, fix handle mismatch (5317)
5388 - fix DLM error path that led to out-of-sync client, long delays (5779)
5389 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5390 - fix several locking issues related to i_size (5492,5624,5654,5672)
5391 - don't move pending lock onto export if it is already evicted (5683)
5392 - fix kernel oops when creating .foo in unlinked directory (5548)
5393 - fix deadlock in obdfilter statistics vs. object create (5811)
5394 - use time_{before,after} to avoid timer jiffies wrap (5882)
5395 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5396 - Do not leak request if server was not able to process it (5154)
5397 - If mds_open unable to find parent dir, make that negative lookup(5154)
5398 - don't create new directories with extent-mapping (5909, 5936)
5400 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5401 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5402 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5403 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5404 - support for RHEL4 kernel on i686 (5773)
5405 - provide error messages when incompatible logs are encountered (5898)
5407 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5408 * version 1.4.0.10 (1.4.1 release candidate 1)
5410 - don't keep a lock reference when lock is not granted (4238)
5411 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5412 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5413 - drop import inflight refcount on signal_completed_replay error (5255)
5414 - unlock page after async write error during send (3677)
5415 - handle missing objects in filter_preprw_read properly (5265)
5416 - no transno return for symlink open, don't save no-trasno open (3440)
5417 - don't try to complete elan receive that already failed (4012)
5418 - free RPC server reply state on error (5406)
5419 - clean up thread from ptlrpc_start_thread() on error (5160)
5420 - readahead could read extra page into cache that wasn't ejected (5388)
5421 - prevent races in class_attach/setup/cleanup/detach (5260)
5422 - don't dereference de->d_inode after l_dput of de (5458)
5423 - use "int" for stripe value returned from lock_to_stripe (5544)
5424 - mballoc allocation and error-checking fixes in 2.6 (5504)
5425 - block device patches to fix I/O request sizes in 2.6 (5482)
5426 - look up hostnames for IB nals (5602)
5427 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5428 - don't start multiple acceptors for the same port (5277)
5429 - fix incorrect LASSERT in mds_getattr_name (5635)
5430 - export a proc file for general "ping" checking (5628)
5431 - fix "lfs check" to not block when the MDS is down (5628)
5433 - service request history (4965)
5434 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5435 - create an "evict_client" /proc entry on OSTs, like the MDS has
5436 - fix mount usage message, return errors per mount(8) (5168)
5437 - change grep [] to grep "[]" in tests so they work in more UMLs
5438 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5439 - remove ancient LOV_MAGIC_V0 EA support (5047)
5440 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5441 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5442 - allow readahead allocations to fail when low on memory (5383)
5443 - mmap locking landed again, after considerable improvement (2828)
5444 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
5446 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
5449 - send OST transaction number in read/write reply to free req (4966)
5450 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5451 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5452 - the watchdog thread now runs as interruptible (5246)
5453 - flock/lockf fixes (but it's still disabled, pending 5135)
5454 - don't use EXT3 constants in llite code (5094)
5455 - memory shortage at startup could cause assertion (5176)
5457 - reorganization of lov code
5458 - single portals codebase
5460 - add extents/mballoc support (5025)
5461 - direct I/O reads in the obdfilter (4048)
5462 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
5464 tbd Cluster File Systems, Inc. <info@clusterfs.com>
5467 - send OST transaction number in read/write reply to free req (4966)
5468 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5469 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5470 - the watchdog thread now runs as interruptible (5246)
5471 - handle missing objects in filter_preprw_read properly (5265)
5472 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5473 - drop import inflight refcount on signal_completed_replay error (5255)
5474 - unlock page after async write error during send (3677)
5475 - return original error code on reconstructed replies (3761)
5476 - no transno return for symlink open, don't save no-trasno open (3440)
5478 - add pid to ldlm debugging output (4922)
5479 - bump the watchdog timeouts -- we can't handle 30sec yet
5480 - extra debugging for orphan dentry/inode bug (5259)
5482 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
5485 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
5486 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
5487 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
5488 - fix NULL dereference in /proc/sys/portals/routes (4827)
5489 - allow failed mdc_close() operations to be interrupted (4561)
5490 - stop precreate on OST before MDS would time out on it (4778)
5491 - don't send partial-page writes before EOF from client (4410)
5492 - discard client grant for sub-page writes on large-page clients (4520)
5493 - don't free dentries not owned by NFS code, check generation (4806)
5494 - fix lsm leak if mds_create_objects() fails (4801)
5495 - limit debug_daemon file size, always print CERROR messages (4789)
5496 - use transno after validating reply (3892)
5497 - process timed out requests if import state changes (3754)
5498 - update mtime on OST during writes, return in glimpse (4829)
5499 - add mkfsoptions to LDAP (4679)
5500 - use ->max_readahead method instead of zapping global ra (5039)
5501 - don't interrupt __l_wait_event() during strace
5503 - add software watchdogs to catch hung threads quickly (4941)
5504 - make lustrefs init script start after nfs is mounted
5505 - fix CWARN/ERROR duplication (4930)
5506 - return async write errors to application if possible (2248)
5507 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
5508 - print NAL number in %x format (4645)
5509 - update barely-supported suse-2.4.21-171 series (4842)
5510 - support for sles 9 %post scripts
5511 - support for building 2.6 kernel-source packages
5512 - support for sles km_* packages
5514 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
5517 - ignore -ENOENT errors in osc_destroy (3639)
5518 - notify osc create thread that OSC is being cleaned up (4600)
5519 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
5520 - reconstruct ost_handle() like mds_handle() (4657)
5521 - create a new thread to do import eviction to avoid deadlock (3969)
5522 - let lconf resolve symlinked-to devices (4629)
5523 - don't unlink "objects" from directory with default EA (4554)
5524 - hold socknal file ref over connect in case target is down (4394)
5525 - allow more than 32000 subdirectories in a single directory (3244)
5526 - fix blocks count for O_DIRECT writes (3751)
5527 - OST returns ENOSPC from object create when no space left (4539)
5528 - don't send truncate RPC if file size isn't changing (4410)
5529 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
5530 - bind to privileged port in socknal and tcpnal (3689)
5532 - rate limit CERROR/CWARN console message to avoid overload (4519)
5533 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
5534 - basic mmap support (3918)
5535 - kernel patch series update from b1_4 (4711)
5537 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
5540 - avoid crash during MDS cleanup with OST shut down (2775)
5541 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
5542 - don't use bad inodes on the MDS (3744)
5543 - dynamic object preallocation to improve recovery speed (4236)
5544 - don't hold spinlock over lock dumping or change debug flags (4401)
5545 - don't zero obd_dev when it is force cleaned (3651)
5546 - print grants to console if they go negative (4431)
5547 - "lctl deactivate" will stop automatic recovery attempts (3406)
5548 - look for existing locks in ldlm_handle_enqueue() (3764)
5549 - don't resolve lock handle twice in recovery avoiding race (4401)
5550 - revalidate should check working dir is a directory (4134)
5552 - don't always mark "slow" obdfilter messages as errors (4418)
5554 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
5557 - don't close LustreDB during write_conf until it is done (3860)
5558 - fix typo in lconf for_each_profile (3821)
5559 - allow dumping logs from multiple threads at one time (3820)
5560 - don't allow multiple threads in OSC recovery (3812)
5561 - fix debug_size parameters (3864)
5562 - fix mds_postrecov to initialize import for llog ctxt (3121)
5563 - replace config semaphore with spinlock (3306)
5564 - be sure to send a reply for a CANCEL rpc with bad export (3863)
5565 - don't allow enqueue to complete on a destroyed export (3822)
5566 - down write_lock before checking llog header bitmap (3825)
5567 - recover from lock replay timeout (3764)
5568 - up llog sem before sending rpc (3652)
5569 - reduce ns lock hold times when setting kms (3267)
5570 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
5571 - fix NULL deref and obd_dev leak on setup error (3312)
5572 - replace some LBUG about llog ops with error handling (3841)
5573 - don't match INVALID dentries from d_lookup and spin (3784)
5574 - hold dcache_lock while marking dentries INVALID and hashing (4255)
5575 - fix invalid assertion in ptlrpc_set_wait (3880)
5577 - add libwrap support for the TCP acceptor (3996)
5578 - add /proc/sys/portals/routes for non-root route listing (3994)
5579 - allow setting MDS UUID in .xml (2580)
5580 - print the stack of a process that LBUGs (4228)
5582 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
5585 - don't cleanup request in ll_file_open() on failed MDS open (3430)
5586 - make sure to unset replay flag from failed open requests (3440)
5587 - if default stripe count is 0, use OST count for inode size (3636)
5588 - update parent mtime/ctime on client for create/unlink (2611)
5589 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
5590 - free recovery state on server during a forced cleanup (3571)
5591 - unregister_reply for resent reqs (3063)
5592 - loop back devices mounting and status check on 2.6 (3563)
5593 - fix resource-creation race that can provoke i_size == 0 (3513)
5594 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
5595 - more debugging for page-accounting assertion (3746)
5596 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
5597 - avoid deadlock after precreation failure (3758)
5598 - fix race and lock order deadlock in orphan handling (3450, 3750)
5599 - add validity checks when grabbing inodes from l_ast_data (3599)
5601 - add /proc/.../recovery_status to obdfilter (3428)
5602 - lightweight CDEBUG infrastructure, debug daemon (3668)
5603 - change default OSC RPC parameters to be better on small clusters
5604 - turn off OST read cache for files smaller than 32MB
5605 - install man pages and include them in rpms (3100)
5606 - add new init script for (un)mounting lustre filesystems (2593)
5607 - run chkconfig in %post for init scripts (3701)
5608 - drop scimac NAL (unmaintained)
5610 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
5613 - clean kiobufs before and after use (3485)
5614 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
5615 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
5616 - change init script to fail more clearly if not run as root (1528)
5617 - allow clients to reconnect during replay (1742)
5618 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
5619 - don't do DNS lookups on NIDs too small for IP addresses (3442)
5620 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
5621 - fix cond_resched (3554)
5622 - only evict unfinished clients after recovery (3515)
5623 - allow bulk resend, prevent data loss (3570)
5624 - dynamic ptlrpc request buffer allocation (2102)
5625 - don't allow unlinking open directory if it isn't empty (2904)
5626 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
5627 - remove extraneous obd dereference causing LASSERT failure (3334)
5628 - don't use get_cycles() when creating temp. files on the mds (3156)
5629 - hold i_sem when setting i_size in ll_extent_lock() (3564)
5630 - handle EEXIST for set-stripe, set proper directory name (3336)
5632 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
5633 - fix ksocknal_fmb_callback() error messages (2918)
5635 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
5638 - don't copy lvb into (possibly NULL) reply on error (2983)
5639 - don't deref dentry after dput, don't free lvb on error (2922)
5640 - use the kms to determine writeback rpc length (2947)
5641 - increment oti_logcookies when osc is inactive (2948)
5642 - update client's i_blocks count via lvb messages (2543)
5643 - handle intent open/close of special files properly (1557)
5644 - mount MDS with errors=remount-ro, like obdfilter (2009)
5645 - initialize lock handle to avoid ASSERT on error cleanup (3057)
5646 - don't use cancelling-locks' kms values (2947)
5647 - use highest lock extent for kms, not last one (2925)
5648 - don't dereference ERR_PTR() dentry in error handling path (3107)
5649 - fix thread race in portals_debug_dumplog() (3122)
5650 - create lprocfs device entries at setup instead of at attach (1519)
5651 - common AST error handler, don't evict client on completion race (3145)
5652 - zero nameidata in detach_mnt in 2.6 (3118)
5653 - verify d_inode after revalidate_special is valid in 2.6 (3116)
5654 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
5655 - initialize RPC timeout timer earlier for 2.6 (3219)
5656 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
5657 - print nal/nid for unknown nid (3258)
5658 - additional checks for oscc recovery before doing precreate (3284)
5659 - fix ll_extent_lock() error return code for 64-bit systems (3043)
5660 - don't crash in mdc_close for bad permissions on open (3285)
5661 - zero i_rdev for non-device files (3147)
5662 - clear page->private before handing to FS, better assertion (3119)
5663 - tune the read pipeline (3236)
5664 - fix incorrect decref of invalidated dentry (2350)
5665 - provide read-ahead stats and refine rpc in flight stats (3328)
5666 - don't hold journal transaction open across create RPC (3313)
5667 - update atime on MDS at close time (3265)
5668 - close LDAP connection when recovering to avoid server load (3315)
5669 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
5670 - don't leak open file on MDS after open resend (3325)
5671 - serialize filter_precreate and filter_destroy_precreated (3329)
5672 - loop device shouldn't call sync_dev() for nul device (3092)
5673 - clear page cache after eviction (2766)
5674 - resynchronize MDS->OST in background (2824)
5675 - refuse to mount the same filesystem twice on same mountpoint (3394)
5676 - allow llmount to create routes for mounting behind routers (3320)
5677 - push lock cancellation to blocking thread for glimpse ASTs (3409)
5678 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
5679 - fix rare problem with rename on htree directories (3417)
5681 - allow default OST striping configuration per directory (1414)
5682 - fix compilation for qswnal for 2.6 kernels (3125)
5683 - increase maximum number of MDS request buffers for large systems
5684 - change liblustreapi to be useful for external progs like lfsck (3098)
5685 - increase local configuration timeout for slow disks (3353)
5686 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
5688 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5691 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
5692 - don't overwrite extent policy data in reply if lock was blocked (2901)
5693 - drop filter export grants atomically with removal from device (2663)
5694 - del obd_self_export from work_list in class_disconnect_exports (2908)
5695 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
5696 - swab reply message in mdc_close, other PPC fixes (2464)
5697 - fix destroying of named logs (2325)
5698 - overwrite old logs when running lconf --write_conf (2264)
5699 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
5700 - fix race in target_handle_connect (2898)
5701 - mds_reint_create() should take same inode create lock (2926)
5702 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
5703 - don't close files for self_export to avoid uninitialized obd (2936)
5704 - allow MDS with the same name as client node (2939)
5705 - hold dentry reference for closed log files for unlink (2325)
5706 - reserve space for all logs during transactions (2059)
5707 - don't evict page beyond end of stripe extent (2925)
5708 - don't oops on a deleted current working directory (2399)
5709 - handle hard links to targets without a parent properly (2517)
5710 - don't dereference NULL lock when racing during eviction (2867)
5711 - don't grow lock extents when lots of conflicting locks (2919)
5713 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
5716 - account for cache space usage on clients to avoid data loss (974)
5717 - lfsck support in lustre kernel code (2349)
5718 - reduce journal credits needed for BRW writes (2370)
5719 - orphan handling to avoid losing space on client/server crashes
5720 - ptlrpcd can be blocked, stopping ALL progress (2477)
5721 - use lock value blocks to assist in proper KMS, faster stat (1021)
5722 - takes i_sem instead of DLM locks internally on obdfilter (2720)
5723 - recovery for initial connections (2355)
5724 - fixes for mds_cleanup_orphans (1934)
5725 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
5726 - block all file creations until orphan recovery completes (1901)
5727 - client remove rq_connection from request struct (2423)
5728 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
5729 - recovery timer race (2670)
5730 - mdc_close recovey bug (2532)
5731 - ptlrpc cleanup bug (2710)
5732 - mds timeout on local locks (2588)
5733 - namespace lock held during RPCs (2431)
5734 - handle interrupted sync write properly (2503)
5735 - don't try to handle a message that hasn't been replied to (2699)
5736 - client assert failure during cleanup after abort recovery (2701)
5737 - leak mdc device after failed mount (2712)
5738 - ptlrpc_check_set allows timedout requests to complete (2714)
5739 - wait for inflight reqs when ptlrpcd finishes (2710)
5740 - make sure unregistered services are removed from the srv_list
5741 - reset bulk XID's when resending them (caught by 1138 test)
5742 - unregister_bulk after timeout
5743 - fix lconf error (2694)
5744 - handle write after unfinished setstripe, stripe-only getstripe (2388)
5745 - readahead locks pages, leaves pending causing memory pressure (2673)
5746 - increase OST request buffers to 4096 on large machines (2729)
5747 - fix up permission of existing directories in simple_mkdir (2661)
5748 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
5749 - don't assign transno to errored transactions (2742)
5750 - don't delete objects on OST if given a bogus objid from MDS (2751)
5751 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
5752 - if rq_no_resend, then timeout request after recovery (2432)
5753 - fix MDS llog_logid record size, 64-bit array alignment (2733)
5754 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
5755 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
5756 - MDS recovery shouldn't delete objects that it has given out (2730)
5757 - if enqueue arrives after completion, don't clobber LVB (2819)
5758 - don't unlock pages twice when trigger_group_io returns error (2814)
5759 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
5760 - don't write pages to disk if there was an error (1450)
5761 - don't ping imports that have recovery disabled (2676)
5762 - take buffered bytes into account when balancing socknal conn (2817)
5763 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
5764 - reconnect unlink llog connection after MDS reconnects to OST (2816)
5765 - remove little-endian swabbing of llog records (1987)
5766 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
5767 - retry reposting request buffers if they fail (1191)
5768 - grow extent at grant time to avoid granting a revoked lock (2809)
5769 - lock revoke doesn't evict page if covered by a second lock (2765)
5770 - disable VM readahead to avoid reading outside lock extents (2805)
5772 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
5773 - updated kernel patches for hp-2.4.20 kernel (2681)
5775 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
5778 - fix truncated write corruption (2366)
5779 - fix for failed assertion in iopen_connect_dentry (1792,2517)
5781 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
5782 - ensure len(uuid) < 37 in lmc (1171)
5783 - fix ia64 OOPS in llog_test (2255)
5784 - zero end of page at obdfilter for partial page writes (2648)
5785 - don't leave stale dentries around after renames (bug 2428)
5786 - fix timeouts when evicting a client with a single lock held (2642)
5787 - set deadline for the initial HELLO message to drain (2634)
5788 - print out dotted-quad IP addresses in the socknal (2302)
5790 - additional debugging for MDS client eviction problem (2443)
5791 - fix mkfsoptions support for osts (2603, 2604)
5793 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
5796 - add series for the vanilla 2.6.0 kernel
5797 - add series for the vanilla 2.4.24 kernel
5798 - add series for a cray x86/64 UL kernel drop
5799 - fix xattr patches for the vanilla 2.4.19 series
5801 - generate true UUIDs in lmc (1171)
5802 - have portals stack dumping break in UML (2466)
5803 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
5804 - allocate with _NOFS in ldlm to avoid deadlock (1933)
5805 - wake callback waiting threads on client eviction (2460)
5806 - Add --ptldebug and --subsystem to lmc (1719)
5807 - update assertion to allow safe interrupt allocation
5808 - set rq_no_resend for cancel requests (2432)
5809 - recalculate ptlrpcd timeout after resend (2494)
5810 - call vfs_rmdir when removing pending directories (2368)
5811 - fix renaming a file to itself (2429)
5812 - lmc creates a default one-stripe lov (2454)
5813 - expand procfs space to handle large clusters (2326)
5814 - increase UML stack to avoid overflow
5815 - update lconf's list of debug and subsystem masks
5816 - fix lfs find --obd (2510)
5817 - /proc tunable for disabling filter read caching (2591)
5818 - stop rpm packages from altering slapd.conf (2301)
5819 - disable nagle in the socknal under 0conf (2578)
5820 - choose mds inode size based on stripe count (2572)
5821 - fix kernel-source rpm problems (2516)
5823 - add --disable-doc to avoid pdf generation (2421)
5824 - update documentation, tests, type-os, comments
5825 - avoid format warnings on ia64
5826 - remove the TOE NAL
5827 - tiny code cleanups by removing unused fields
5829 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
5832 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
5833 - remove the most blatant lies from BUILDING file (2371)
5834 - change default debug level to reasonable production setting
5835 - reduce client side cache size to reduce cache flush time
5836 - reduce max RPCs in flight to avoid unnecessary file fragmentation
5837 - make TCP zerocopy and pinger support enabled by default (2476)
5838 - sync writes completed after process exits caused crashes (2319)
5839 - maintain correct mount count on the MDS (2356)
5840 - backout 1557, because 2316 wasn't really fixed
5841 - better file I/O statistics gathering in /proc
5842 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
5843 - another kernel patch to fix zero-copy TCP function export
5844 - don't take duplicate lock when processing re-sent getattr (2420)
5845 - lctl uses obd_self_export instead of creating new conn (2353)
5846 - MDS/OST recovery case which requires object creation asserted (2425)
5847 - move lfs from /usr/sbin to /usr/bin in packages
5848 - fix race between mds_client_add and mds_client_free (2417)
5849 - use kmalloc instead of slabs in portals (2430)
5850 - don't create duplicate records when a failover MDS is present (2442)
5851 - remove unnecessary mount age check (2332)
5852 - don't remove directory inodes from locks prematurely (2451)
5853 - don't break if MDS service name is the same as hostname (2103)
5854 - fix races in client write RPC generation when cache full (2482)
5856 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
5859 - remove now-unused request->rq_obd (278)
5860 - if an allocation fails, print out how much memory we've used (1933)
5861 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
5862 - add the "configurable stack size" patch to most series files (1256)
5863 - ability to write large log records, for 100+ OST configs (2306)
5864 - fix NULL deref when filter_prep fails (2314)
5865 - fix operator precedence error in filter_sync
5866 - dynamic allocation of socknal TX descriptors (2315)
5867 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
5868 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
5869 - ensure that CWARN messages in Portals always get to the syslog
5870 - __init/__exit are not for prototype decls (ldlm_init/exit)
5871 - x86-64 compile warning fixes
5872 - fix gateway LMC keyword conflict (2318)
5873 - fix MDS lock inversions in getattr/reint paths (1844)
5874 - fix a rare lock re-ordering bug, which caused deadlock (2322)
5875 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
5876 - DLM race condition prevented some lock evictions (2328)
5877 - ENOMEM detection and retry on socknal sends (2230)
5878 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
5879 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
5880 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
5881 - better error messages when a client is rejected during recovery (1505)
5882 - avoid cancelling locks which were never granted, after failure (2330)
5883 - fix i_sem/journal inversion in mds_client_add (2333)
5884 - fix truncate/getattr lock cycle deadlock (2334)
5885 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
5886 - fix two rare exit paths which could leak an l_lock() ref (2321)
5887 - fencepost error in MDS/OST orphan recovery (2226)
5888 - make log record alignment 8 bytes (1988)
5889 - lstripe now fails when requested offset > ost_count (2237)
5890 - ensure that all kernel series have a complete list.h (1607)
5891 - fix crashes in special-file operations (2316)
5892 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
5894 - allow configurable automake binary, for testing new versions
5895 - small update to the lfs documentation
5897 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
5899 * fix negative export reference count in fsfilt_sync (2312)
5901 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
5902 * release candidate 0.9.1
5904 - orphans are moved into the PENDING directory for possible recovery
5905 - replayed opens now open by fid for orphan/rename safety (1042)
5906 - last close of an orphan inode generates a transno (683)
5907 - chdir() and mount() now pin the directory entry (1020)
5908 - avoid CERROR in normal ll_setattr_raw() error case (1500)
5909 - discard very old requests without processing them (1502)
5910 - remove some common, well-understood CERRORs (1505)
5911 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
5912 - clear "grant" flags in OST replies until OST grant code lands (1644)
5913 - fix read performance by not clobbering i_blksize on client (1598)
5914 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
5915 - make LRU size a /proc tunable, clears locks when reduced (707)
5916 - fix some lprocfs rot that prevented ptlbd from loading (1732)
5917 - server locks take references on exports now (1558)
5918 - build fixes for 2.4.20-rh trees (1663)
5919 - return an error from lov_create if all OSCs are inactive (1751)
5920 - fix import levels when a reconnect happens without a timeout (1597)
5921 - exit early from mds_open if we get a lookup error (1749)
5922 - partial page read at EOF wouldn't wait for disk before sending (1642)
5923 - avoid NULL deref in obdfilter when reading page past EOF (1592)
5924 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
5925 - fix LBUG in ll_it_open_error with rc = -2 (1861)
5926 - write/truncate lock inversion (1639)
5927 - Don't auto-load obdclass, portals modules during cleanup (1495)
5928 - fix timestamps from jumping to "now" (1763)
5929 - extra journal assertions (1648)
5930 - add an extra multiunlink test (1771)
5931 - fix read_record/write_record API (1776)
5932 - fix leak of offset_extent, possible incorrect i_size later (1772)
5933 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
5934 - Debugging for the kqswnal_get_idle_tx problems (1820)
5935 - Allow recovery to be attempted multiple times (1536)
5936 - Write out MDS last_rcvd file after it is first created (1600)
5937 - Fix tx_descriptor leak in failed transmit situations (1827)
5938 - ext3 journaling fixes for assertion failure after IO error (1871)
5939 - class_export_put() on freed export after completion AST error (1896)
5940 - Fix revalidate looping in VFS (1322)
5941 - Don't access a freed export during MDS_REINT timeout (1521)
5942 - Add open-unlink recovery support on the MDS (1673,1764)
5943 - Return an error if no MDS data was read from last_rcvd (1946)
5944 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
5945 - Don't setup a disk device that doesn't match exported UUID (317)
5946 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
5947 - avoid committing NULL handle in force close
5948 - local.sh is now a one-stripe LOV configuration
5949 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
5950 - fix ext3 htree duplicate directory entry corruption (1516)
5951 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
5952 - update to new LOV EA format (2097)
5953 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
5955 2003-06-15 Phil Schwan <phil@clusterfs.com>
5958 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
5959 - per-import/export recovery handling (958, 931, 959)
5960 - multiple last-rcvd slots, for serving multiple FSes (949)
5961 - connections are again shared between multiple imp/exports (963, 964)
5962 - "umount -f" would hang if any requests needed to be sent (393, 978)
5963 - avoid pinning large req buffer by copying for queued messages (989)
5964 - add "uuid" to "lctl device" command to help upcalls (991)
5965 - "open" RPCs with transnos would confuse recovery counters (1037)
5966 - do proper endian conversion of all wire messages (288, 340, 891)
5967 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
5968 - call waiting locks callback from LDLM recovery thread (1127, 1151)
5969 - fix ptlrpc_connection leak in target_handle_connect (1174)
5970 - fix import refcounting bug in OST and MDS cleanup (1134)
5971 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
5972 - fix very unlikely obd_types race condition (501)
5973 - remove osc_open hack for echo_client (1187)
5974 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
5975 - a failure in read_inode2 leads to deadlock (1139)
5976 - cancel ack-locks as soon as transaction is committed (1072)
5977 - fix major leaks and crashes in the bulk I/O path (937, 1057)
5978 - make sure to commitrw after any preprw to avoid deadlock (1162)
5979 - failing to execute a file in a lustre FS would lock inode (1203)
5980 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
5981 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
5982 - obd_uuid2tgt would walk off the end of the list (1255)
5983 - on IA64 the osc would give portals incorrect bulk size (1258)
5984 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
5985 - fix lock inversion caused by new llite matching code (1282)
5986 - limit the number of dirty pages on a client to 10MB (1286)
5987 - timed out locks were not being corrected cancelled (1289)
5988 - fix O_DIRECT above 4GB on IA-32 (1292)
5989 * major user-visible changes
5990 - fail out/fail over policy now controlled by the upcall (993)
5992 - add OBD_PING to check server availability and failure (954)
5993 - lustre messages are now sent in sending host order (288, 340, 891)
5994 - add eadatalen to MDS getattr reply (340)
5995 - OST read replies may contain second buffer, with per-page status (593)
5997 2003-03-11 Phil Schwan <phil@clusterfs.com>
6000 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
6001 - failed open()s could cause deadlock; fixed (867, 869)
6002 - stop cancelling OST locks when files are closed (481)
6003 - overlapping XID spaces caused network corruption (851, 853)
6004 - fix unsafe fsfilt counter arithmetic; change to atomic_t
6005 - setattr_raw added, to do single-RPC, server-side setattrs
6006 - lmc/lconf syntax change for OST UUIDs
6007 - fix crashy race condition between ptlrpc_free_req and osc_close
6008 - don't use request in mdc_enqueue if we hit a timeout (889)
6009 - don't set the inode i_size for regular files from the MDS (896)
6010 - handle out of order completion AST (842)
6011 - don't LBUG if a lock request times out after receiving AST (913)
6012 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
6013 - if a bad lock AST arrives, send an error instead of dropping entirely
6014 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
6015 - fix leak in bulk IO when only partially completed (899, 900, 926)
6016 - fix O_DIRECT for ia64 (55)
6017 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
6018 - C-z after timeout could hang a process forever; fixed (977)
6020 - client-side I/O cache (678, 924, 929, 941, 970)
6022 - READPAGE and SETATTRs which don't take server-side locks get
6025 2003-02-11 Phil Schwan <phil@clusterfs.com>
6028 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
6029 - Fix fsfilt_extN_readpage() to read a full page of directory
6030 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
6031 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
6032 - Fix imbalanced LOV object allocation and out-of-bound access (469)
6033 - Most intent operations were removed, in favour of a new RPC mode
6034 that does a single RPC to the server and bypasses most of the VFS
6035 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
6036 - Aggressively cancel local locks on DLM servers
6037 - mds_reint_unlink sends EA to the client if it's the last nlink.
6038 client uses that EA to unlink OST objects.
6039 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
6040 - recursive symlinks were fixed (439)
6041 - fixed NULL deref in DEBUG_REQ
6042 - filter_update_lastobjid no longer calls sync, which annoyed extN
6043 - fixed multi-client small-writes to a single file problem (445)
6044 - fixed mtime updates during file writes (607)
6045 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
6046 - fixed bug in obd_brw_read/write() (under guise of testing 367)
6047 - fixed Linux OST size reporting problem (444, 656)
6048 - OST now updates object mtime with writes or setattr (607, 619)
6049 - client verifies file size before zeroing page past EOF (445)
6050 - OST now writes last allocated objid to disk with allocation (108)
6051 - LOV on echo now works (409)
6053 - mds_reint_unlink sends a new buffer, with the EA included. this
6054 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
6055 set if a regular file was being unlinked, and it was the last link
6056 - use PtlGet from the target for bulk writes (315)
6057 - OST now updates object mtime with writes or setattr (607, 619)
6058 - LDLM now has a grant-time callback to revalidate locked items, if
6060 - Many MDS operations were reorganized to combat race conditions
6062 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
6063 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
6065 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
6068 - Fully reactivate OST imports after reconnection (512, others)
6069 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
6070 - More graceful error handling for truncating on dead OST (515)
6071 - Don't error out unless we're actually accessing dead stripes (474)
6072 - Fix garbage sizes when stripes are missing (410)
6073 - LRU counters were broken, causing constant lock purge (433, 432)
6074 - garbage on read from stripes with failed OSTs (441)
6075 - mark OSCs as active before reconnecting during recovery (438)
6076 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
6077 - lfind did not preserve OST order in output (443)
6078 - symlinks cause hung clients, incorrect data (439)
6079 - stop dereferencing request after dropping refcount (457)
6080 - don't LASSERT(spin_is_locked) on non-SMP (455)
6081 - fixes for many rename() bugs
6082 - fstat didn't correctly synchronize attributes (399)
6083 - server must handle lock cancellation during blocking AST prep (487)
6084 - bulk descriptors were free()d too soon (511)
6085 - fix paths in lconf, which would load incorrect modules (451, 507)
6086 - fix confusing lconf 'host not found' error message (386)
6087 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
6088 - fix race condition in mdc_blocking_ast() for inode access (526)
6089 - fix lov_unpackmd() unpacking wrong number of stripes (537)
6090 - fix lov_set_osc_active() marking wrong OSC inactive (440)
6091 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
6092 - fix multiple writes of stripe MD to MDS (358, maybe 519)
6093 - fix lstripe in several ways (kernel side) (527)
6094 - fix request leak in ldlm_cli_enqueue (262)
6095 - incorrect OSC was marked inactive after OST failure
6096 - call mds_fs_cleanup before unmounting filesystem (524)
6097 - fix races between taking ns_lock and ldlm_lock_change_resource
6098 - fix races updating LOV export open file list
6099 - fix lov_enqueue error path, avoid decref-ing bad lock handle
6100 - fix recovery NULL deref in ldlm_cli_cancel_unused
6101 - fix some DLM races by using new hash table for lock handles (419)
6102 - permit the client to specify desired inodes, at replay
6103 - duplicate requests when we queue them for replay reintegration
6104 - fix last_rcvd offset calculation
6105 - sync after each recovered transaction, so we always make progress
6106 - never, not always, ERESTART requests without transnos
6107 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
6109 - skip replay if the MDS says that the client is already connected
6110 - don't check for a recovery-enabled export to match lctl's UUID
6111 - don't INC_USE_COUNT for phantom exports
6112 - don't crash when cleaning up phantom exports (567)
6113 - don't double-finish or set replay data for errored mdc_open requests
6114 - abort requests when they time out, so we don't get old replies
6115 - send/receive replies for AST messages again
6116 - if the client says that it doesn't have the lock, cancel it on the
6118 - if we timeout during I/O, don't try to cancel an in-use lock; instead
6119 mark it as destroyed, it will all work out when decref is called
6120 - fix module use counts (22, 581)
6122 - ASTs now expect a reply (server cancels lock on error reply)
6124 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
6127 - fix many simultaneous client startup (392)
6128 - fix dentry->d_it clobbering
6129 - credentials weren't being shipped for readdir/getattr operations
6130 - remove invalid assertions triggered during some concurrent MD
6132 - proper Lustre versions added (336, 389)
6133 - fix memory leak for create error case (398)
6134 - fix LOV locking bug that would get cli/srv out of sync
6135 - fix echo client over LOV (409)
6136 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
6137 - fix double-O_EXCL intent crash (424)
6138 - avoid sending multiple lock CANCELs (352)
6140 - MDS can do multi-client recovery (modulo bugs in new code)
6142 - many updates, edits, cleanups
6144 2002-11-18 Phil Schwan <phil@clusterfs.com>
6147 - fix null d_it dereference (346)
6148 - fix full OST/dbench hang (333)
6149 - fix permission problem with file removal (286)
6150 - fix removal of OSCs from LOV when they fail
6151 - fix NULL deref during bulk timeout (214)
6152 - fix problems related to multiple filesystems on one MDS (241)
6153 - fixed serious subtle metadata locking bugs
6154 - free locks on clients when inodes are removed due to memory
6156 - fix inode pointer in lock data (285)
6157 - partial support for multiple MDS on a single host (241)
6158 - data locks weren't cancelled at clear_inode time (290, 311)
6159 - intent locks could lead to unbounded lock growth (205)
6160 - added a maximum lock count, an LRU list, and a flusher
6161 - fix multiple rename (365)
6162 - properly abstracted the echo client
6163 - OSC locked 1 byte too many; fixed
6164 - rewrote brw callback code:
6165 - fixed recovery bugs related to LOVs (306)
6166 - fixed too-many-pages-in-one-write crash (191)
6167 - fixed (again) crash in sync_io_timeout (214)
6168 - probably fixed callback-related race (385)
6170 - Add capability to MDS protocol
6171 - LDLM cancellations and callbacks on different portals
6173 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
6176 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
6177 - defer open object destruction to close time (601981, 138)
6178 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
6179 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
6180 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
6181 - don't call obd_setattr in ll_file_release for destroyed objects
6182 * protocol change to lustre_msg: move |version| and add |flags|
6183 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
6184 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
6185 * added replay of create, unlink, link and rename operations during
6186 MDS failover; recovery should be much more robust now
6187 * remove failed OSCs from LOVs (only lov_create uses this so far)
6188 * the lustre-HOWTO was brought (more) up to date (582544)
6190 2002-10-23 Phil Schwan <phil@clusterfs.com>
6193 - in-use dentries weren't being reused properly (617851)
6194 - prevent multiple LDLM setup (599178)
6195 - fix LOV size calculations for truncate (617853)
6196 - fix client handling of MDS intent errors (POSIX)
6197 - fix permission bug in lovstripe.c test (624321)
6198 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
6199 - truncate past end of file could corrupt data
6200 - proper cleanup after timeouts, crashes, etc (592524, 550815)
6201 - a race in recovery could return ETIMEDOUT to apps (623947)
6202 - building outside the source directory was fixed
6203 * the lustre-HOWTO was brought (more) up to date (582544)
6204 * major progress was made on recovery functionality
6206 2002-10-10 Phil Schwan <phil@clusterfs.com>
6209 - recovery deadlock fix
6210 - rm -rf causes LBUG fix (617817)
6211 - file open by multiple tasks fix (618962)
6212 - directory permissions bugs (602707 and 620007)
6213 - journal_stop fixed with locking (611313)
6214 - O_APPEND failures resolved (618273, perhaps 614459)
6215 - lconf PATH fix (619770)
6216 - IA64 build fix (621450)
6217 - RPC buffer sizes scale with amount of memory
6219 2002-10-01 Phil Schwan <phil@clusterfs.com>
6222 - locks would be cancelled without throwing away data pages,
6223 resulting in inconsistent data (605627)
6224 - inode attributes were not always being refreshed (605627, 612449)
6225 - lconf now continues to cleanup after lctl reports an error
6226 - MDS now enforces user permissions (602707)
6227 - lprocfs cleanup fixed, but not yet enabled (614157)
6228 - fixed infinite server hang, should a client not respond to an AST
6229 - avoid going into recovery if user calls readlink() with a buffer
6230 that's too small (613941)
6231 - AST RPCs no longer require replies (614867) -- this may be changed
6232 - don't crash server if client sends an IOV that's too big (611336)
6233 - fixed lock conversion deadlock (611892)
6234 - fixed the following of symlinks (614622)
6235 * recovery: the server can remove locks from a client that dies, other
6236 clients can make progress
6237 * more extN patch fixes
6238 * compile-time configurable ptlrpc buffer allocations
6240 - collaborative read cache document
6241 - Lustre Lite Performance CDR document-in-progress
6243 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6246 - fix typo in patch-2.4.18
6248 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6251 - clear ptlrpc request each time in handle_incoming_request()
6252 - unlink of files now destroys the object on the OST
6254 2002-09-19 Peter Braam <braam@clusterfs.com>
6256 * add hard link support
6257 * change obdfile creation method
6258 * kernel patch changed
6260 2002-09-19 Peter Braam <braam@clusterfs.com>
6263 - stack overflow bug in extN fixed
6265 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
6267 * documentation updates
6268 - add man pages for config tools
6269 - update tests/README to describe testing with new config tools
6270 - finish metadata API descriptions
6271 * bug fixes and cleanups
6272 - statfs workaround for 16TB limit
6273 - LOV stripe allocation improved, can stripe on subset of OSTs
6274 - LOV file size/IO offset was wrong for files > 4GB in size
6275 - object EA data was being dropped, caused files to be unreadable
6276 - memory overflow with non-LOV OST caused memory corruption
6277 - fixed regression tests to work with new config tools, obdfilter
6278 - fixed bug when directory size became larger than 1 block
6279 - fixed bug (for single client case) when PWD was deleted
6280 - invalidate local directory pages when doing intent-based ops
6281 - avoid LDLM oops when lock callback contained bad data
6283 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
6285 * documentation updates
6286 * bug fixes and cleanups
6287 - configuration tools
6290 - 64-bit compile warnings
6291 - 64-bit internal statfs data
6293 * test_brw on persistent OST devices
6295 * lprocfs (disabled)
6297 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
6299 * documentation updates
6300 * bug fixes and cleanups
6301 * configuration tools
6303 2002-08-30 Peter J. Braam <braam@clusterfs.com>
6306 * many small fixes to 0_5_4
6307 * io/network handling
6308 * thinkos in MDS operations
6310 2002-08-24 Peter J. Braam <braam@clusterfs.com>
6313 * crucial basic fixes to 0.5.3
6314 * IOR, Iozone work over Elan
6317 2002-08-07 Phil Schwan <phil@clusterfs.com>
6318 * version 0_5_3, our first alpha
6319 * we use the new Portals iovs
6320 * documentation updates
6321 * bug fixes and cleanups
6322 * small changes in the DLM wire protocol
6324 2002-07-25 Peter J. Braam <braam@clusterfs.com>
6325 * version 0_5_1 with some initial stability,
6326 * locking on MD and file I/O.
6327 * documentation updates
6328 * several bug fixes since 0.5.0
6329 * small changes in wire protocol
6331 2002-07-18 Phil Schwan <phil@clusterfs.com>
6333 * delivered as Lustre Light Alpha
6334 * fixed a crash after handling invalid MDS requests
6335 * fixed directory pages for architectures with non-4k pages sizes
6337 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
6338 * release version v0_4_4
6339 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
6341 * Fixes a number of interruption problems with OST operations.
6342 * Update documentation for portals header changes
6343 * Move all wire protocol structs/defines to lustre_idl.h
6344 * Fixes symlink length bug.
6345 * Add tcpdump to repository.
6347 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
6348 * release version v0_4_3
6349 * Fixes statfs for inodes on extN.
6350 * Fixes bug in runtests which would delete /etc/hosts.
6351 * Use 64-bit object IDs wherever possible (not into VFS though)
6352 Remove ost_get_info, which is unused by lustre, and out of date.
6354 2002-07-03 Peter Braam <braam@clusterfs.com>
6355 * release version v0_4_2 Fixes a lookup error (type not passed)
6356 * move forward to head of Portals
6357 * move forward to latest Lustre kernel
6359 2002-06-25 Peter Braam <braam@clusterfs.com>
6360 * release version v0_4_1. Hopefully stable on single node use.