3 * See http://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
4 * Server support for kernels:
5 2.6.32-279.5.1.el6 (RHEL6)
6 * Client support for unpatched kernels:
7 2.6.32-279.5.1.el6 (RHEL6)
9 2.6.18-238.19.1.el5 (RHEL5)
10 2.6.16.60-0.69.1 (SLES10)
11 * Recommended e2fsprogs version: 1.41.90.wc4
12 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
13 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
14 stack overflow. For more information, please refer to bugzilla 17630.
15 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
16 the re-exporting nodes to be mounted with "32bitapi" mount option
18 Severity : enhancement
20 Description: Update RHEL6.3 kernel to 2.6.32-279.5.1.el6.
22 Severity : enhancement
24 Description: Update RHEL6.3 kernel to 2.6.32-279.2.1.el6.
26 Severity : enhancement
28 Description: Update RHEL6.2 kernel to 2.6.32-220.17.1.el6
30 Severity : enhancement
32 Description: Update RHEL6.2 kernel to 2.6.32-220.13.1.el6
34 Severity : enhancement
36 Description: Update RHEL6.2 kernel to 2.6.32-220.7.1.el6
38 --------------------------------------------------------------------------------
40 04-30-2012 Whamcloud, Inc.
42 * See http://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
43 * Server support for kernels:
44 2.6.32-220.4.2.el6 (RHEL6)
45 * Client support for unpatched kernels:
46 2.6.32-220.4.2.el6 (RHEL6)
48 2.6.18-238.19.1.el5 (RHEL5)
49 2.6.16.60-0.69.1 (SLES10)
50 * Recommended e2fsprogs version: 1.41.90.wc4
51 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
52 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
53 stack overflow. For more information, please refer to bugzilla 17630.
54 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
55 the re-exporting nodes to be mounted with "32bitapi" mount option
57 Severity : enhancement
59 Description: Update RHEL6.2 kernel to 2.6.32-220.4.2.el6
61 Severity : enhancement
63 Description: Update RHEL6.2 kernel to 2.6.32-220.4.1.el6
65 Severity : enhancement
67 Description: Update RHEL6.2 kernel to 2.6.32-220.el6
69 Severity : enhancement
71 Description: Update RHEL6.1 kernel to 2.6.32-131.17.1.el6
73 --------------------------------------------------------------------------------
75 09-30-2011 Whamcloud, Inc.
77 * Server support for kernels:
78 2.6.18-238.19.1.el5 (RHEL5)
79 2.6.32-131.6.1.el6 (RHEL6)
80 * Client support for unpatched kernels:
81 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
82 2.6.18-238.19.1.el5 (RHEL5)
83 2.6.32-131.6.1.el6 (RHEL6)
84 2.6.29.4-167.fc11 (FC11)
85 2.6.16.60-0.69.1 (SLES10)
87 * Recommended e2fsprogs version: 1.41.90.wc3
88 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
89 Lustre filesystem with 4K stack may cause a stack overflow. For more
90 information, please refer to bugzilla 17630.
91 * NFSv4 reexport requires Lustre client on the re-exporting nodes to be
92 mounted with "32bitapi" mount option
94 Severity : enhancement
95 Description: Update RHEL6.1 kernel to 2.6.32-131.6.1.el6
97 Severity : enhancement
98 Description: Update RHEL5.6 kernel to 2.6.18-238.19.1.el5
100 Severity : enhancement
101 Description: Update RHEL5.6 kernel to 2.6.18-238.9.1.el5
104 Frequency : only with SLES10
106 Description: Use OFED "KMP" provided by Novell
107 Details : SLES10 SP3 ships with OFED in a separate "KMP" package.
108 Lustre is now built against this package. That means you need to
109 install the ofed-kmp package from Novell for the patchless client.
110 Note that the ofed-kmp that Novell ships may not exactly match the
111 kernel version but should still be compatible.
113 Severity : enhancement
115 Description: Update RHEL5.5 kernel to 2.6.18-194.17.1.el5
116 Update OEL5.5 kernel to 2.6.18-194.17.1.0.1.el5.
118 --------------------------------------------------------------------------------
120 2010-07-15 Oracle, Inc.
122 * Server support for kernels:
123 2.6.18-164.11.1.el5 (RHEL 5)
124 2.6.18-164.11.1.0.1.el5 (OEL 5)
125 * Client support for unpatched kernels:
126 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
127 2.6.18-164.11.1.el5 (RHEL 5),
128 2.6.18-164.11.1.0.1.el5 (OEL 5)
129 2.6.16.60-0.42.8 (SLES 10),
131 2.6.29.4-167.fc11 (FC11)
132 * Recommended e2fsprogs version: 1.41.10-sun2
134 Severity : enhancement
136 Description: Distribute lfs_migrate script with lustre
137 Details : The lfs_migrate mentioned in the manual is now included in the
138 lustre tree in lustre/scripts and will distributed in lustre RPMs.
140 Severity : enhancement
142 Description: Failover nid list added to import proc file.
144 Severity : enhancement
146 Description: "som_preview" mount option required to use SOM
147 Details : The clients will need to use som_preview mount option
151 Description: Lustre does not do 1MB IOs to HW RAID
152 Details : Bump MAX_PHYS/HW_SEGMENTS and SG_ALL to 256 in the RHEL5 kernel.
154 Severity : enhancement
156 Description: lfs quota output is not very convenient for awk/sed-parsing
157 Details : Some positions in lfs quota output table could be empty or
158 non-empty which made it hard to parse it with scripts, now
159 a dash is put instead of space where there is not supposed
164 Description: Update to OFED 1.5.1
165 Details : OFED 1.5.1 will be used in environments that don't use
166 vendor supplied OFED.
170 Description: Remove lnet. prefix from lctl params display
171 Details : Also renamed lnet.memused to lnet_memused to
176 Description: Use conf_param -d to remove permanant settings.
180 Description: Add list_param -R
181 Details : lctl list_param now supports recursive mode.
184 Bugzilla : 21829, 22312
185 Description: Fix llobdstat.
186 Details : 1. Now updating the write_bytes counter
187 2. Port script fixes from 1.8.
191 Description: decrease the usage of memory on clients.
192 Details : 1. On clients, recycle dentries and inodes unused.
193 2. Delete the code related to ll_deathrow(att 6215 in bug 1443). It
198 Description: llobdstat fix and enhancement
199 Details : add a counter to set a limit to how many samples will be returned
200 fix a wildcard in the path to limit to obdfilter stats only
204 Description: lfs check only allowed for root
205 Details : Code cleanups and added a regresion test of non-root use of
210 Description: MDS Oopses in ptlrpc_dispatch_difficult_reply()/__list_add()
211 Details : Fix hr->hr_index overload.
215 Description: Fix FIEMAP functionality.
219 Description: FC11 patchless client support added
223 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and
224 OEL5.4 kernel to 2.6.18-164.11.1.0.1.el5.
228 Description: New acc-sm script and various test-framework enhancments.
232 Description: Previously landed patch was reverted.
236 Description: Update SLES10 kernel to 2.6.16.60-0.42.8 (patchless)
240 Description: Add ldiskfs maxdirsize mount option.
244 Description: Fix mkfs.lustre support for 16TB LUNs.
248 Description: Print mmp_check_interval
249 Details : print mmp_check_interval and make it possible to abort
250 mount operation in case it takes too long.
254 Description: Add cascading_rw to lustre mpi tests.
258 Description: Store SOM data in EA.
259 Details : Also several SOM related cleanups.
263 Description: Add refcount to md_open_data
267 Description: Use refcount on lov_request_set to prevent memory corruption.
271 Description: Statahead: use parent dir's i_mutex for synchronization.
275 Description: loadgen improvements.
279 Description: Enable peer health detection in ksocklnd and ko2iblnd.
283 Description: Control DCACHE_LUSTRE_INVALID flag with MDS_INODE_LOOKCUP lock
287 Description: Fix race on obd_devcie:md_stats
291 Description: Fix SLES 11 client builds.
292 Details : Resolved uninitalized pointer gcc warnings.
296 Description: Allow non-root access for "lfs check".
297 Details : Added a check in obd_class_ioctl() for OBD_IOC_PING_TARGET.
301 Description: Clear imp_force_reconnect correctly in ptlrpc_connect_interpret()
303 Severity : enhancement
305 Description: Update kernel to RHEL5.4 2.6.18-164.6.1.el5 and
306 OEL5 2.6.18-164.6.1.0.1.el5(Both in-kernel OFED enabled).
310 Description: Sanity test_27p fails.
311 Details : OBD_FAIL_OST_ENOSPC has been moved to filter_statfs().
312 Sanity tests 27 were cleaned up and improved.
316 Description: Unify req format on client/servers
317 Details : Use new req_capsule API [almost] everywhere instead of old PTLRPC
318 buffers and swabbers approach..
323 Description: drop join file stuff.
325 Severity : enhancement
327 Description: Add 2.6.27 and SLES11 (patchless client) support
331 Description: Fix in ptlrpc_expire_one_request() to print the signed time
334 Severity : enhancement
336 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
339 Severity : enhancement
341 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
343 Severity : enhancement
345 Description: correctly handle big reply message.
346 Details : send LNet event if reply is bigger then buffer and adjust this buffer
351 Description: Drop unnecessary __GFP_NOMEMALLOC flag from filter_get_page()
353 Severity : enhancement
355 Description: Add support for OEL5.
359 Description: Conf-sanity.sh 50g test - deactivated OST should not cause a panic.
361 Severity : enhancement
363 Description: "weak-modules" support
364 Details : Implement "weak-modules" support which enables kernel modules
365 to be used with any kernel that implements the same kABI. In
366 order to achieve this modules are now installed in
367 /lib/modules/$(uname -r)/updates/kernel on all distributions.
371 Description: client could not reconnect to OST because of an active request.
372 Details : The client now retries to reconnect to the same server, if a
373 connect request failed with EBUSY or -EAGAIN. Test case 35b
374 conf-sanity.sh has been created.
378 Description: Don't allow make backward step on assignin osc next id.
379 Details : race between allocation next id and ll_sync thread can be cause
380 of set wrong osc next id and can be kill valid ost objects.
384 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
388 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
392 Description: Error handling in osc_statfs_interpret() has been improved.
393 Details : Check in osc_statfs_interpret() for EBADR.
397 Description: Do not send statfs() requests to OSTs disabled by administrator.
398 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
400 Severity : enhancement
402 Description: Added sanity-quota test to for lfs(1) to work with numeric uid/gid.
404 Severity : enhancement
406 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
408 Severity : enhancement
410 Description: Enhancement for lfs(1) command to use numeric uid/gid.
414 Description: Improved cleanup procedure for conf-sanity test_46a.
418 Description: Do not start bulk transfer, if client tried to reconnect.
419 Details : Do not start bulk transfer, if client tried to reconnect,
420 but failed due to active rpcs. Check for exp_abort_active_req
421 has been implemented.
425 Description: Prevent inconsistences between linux and lustre mount structures.
426 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
427 Make the sleep interruptible.
429 Severity : enhancement
431 Description: provide server to client comms path
432 Details : server to client communications path via new LDLM_SET_INFO rpc
436 Description: truncate starts GFP_FS allocation under transaction causing deadlock
437 Details : ldiskfs_truncate calls grab_cache_page which may start page
438 allocation under an open transaction. This may lead to
439 calling prune_icache with consequent lustre reentrance.
443 Description: send correctly lsm on open replay
444 Details : MDS is trust to LSM size on replay open, but client can set wrong size
449 Description: enable client interface failover
450 Details : When a child reconnects from another NID, properly update export
451 nid hash position and ldlm reverse import.
453 Severity : enhancement
455 Description: Communicate OST degraded/readonly state via statfs to MDS
456 Details : Flags in the statfs returned from OSTs indicate whether the
457 OST is in a degraded RAID state, or if the filesystem has
458 turned read-only after a filesystem error is detected.
463 Description: don't panic if EPROTO was hit when reading symlink
464 Details : correctly handling request reference in error cases.
466 Severity : enhancement
468 Description: Add LustreNetLink, a kernel-userspace communcation path.
470 Severity : enhancement
472 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
477 Description: access to llog context before init.
478 Details : move handling CATALOGS file at osc layer and forbid access to llog
482 Frequency : rare, only when client is writing to OST on local node
484 Description: Avoid deadlock for local client writes
485 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
486 memory freeing context. This allows OST threads to set the
487 PF_MEMALLOC flag on task structures in order to allocate memory
488 from reserved pools and complete IO.
489 Use GFP_HIGHUSER for OST allocations for non-local client writes,
490 so that the OST threads generate memory pressure and allow
491 inactive pages to be reclaimed.
493 Severity : enhancement
495 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
500 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
501 Details : move ldlm namespace creation in setup phase to avoid grab
502 _lprocfs_lock with cli_sem held.
506 Description: Temporarily disable grant shrink.
507 Details : Disable the feature for debugging.
510 Frequency : only when formatting test filesystems
512 Description: Unable to run several mkfs.lustre on loop devices at the same time
513 Details : mkfs.lustre returns error 256 on the concurrent loop devices
514 formatting. The solution is to proper handle the error.
516 Severity : enhancement
518 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
520 Severity : enhancement
522 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
527 Description: resolve race between obd_disconnect and class_disconnect_exports
528 Details : if obd_disconnect will be called to already disconnected export he
529 forget release one reference and osc module can't unloaded.
531 Severity : enhancement
533 Description: Allow tuning service thread via /proc
534 Details : For each service a new
535 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
536 created that can be used to set min/max thread counts, and get the
537 current number of running threads.
542 Descriptoin: don't return error if have partially created objects for file.
543 Details : lov_update_create_set uses set->set_success as index for created
544 objects, so if some requests will be failed, they will have hole
545 at end of array and qos_shrink_lsm for allocate correct lsm.
547 Severity : enhancement
549 Description: Update OFED support to 1.4.1
554 Description: Reduce small size read RPC
555 Details : Set read-ahead limit for every file and only do read-ahead when
556 available read-ahead pages are bigger than 1M to avoid small size
559 Severity : enhancement
561 Description: move AT tunable parameters for more consistent usage
562 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
564 Severity : enhancement
566 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
567 Details : allow skip disconnected ost for send statfs request and hide error
573 Description: fix racy locking of mballoc block bitmaps causing BUG
574 Details : The locking of the mballoc buddy bitmap and the in-memory
575 block bitmap was using two different spin locks in some
576 cases. This made it possible to incorrectly access the
577 mballoc bitmap while another process was modifying it,
578 causing a sanity assertion to fail. While no on-disk corruption
579 was reported, there was some risk of this happening.
582 Frequency : rare, on llog test 6
584 Descriptoin: don't allow connect to already connected import
585 Details : allowing connect to already connected import is hide connecting problem.
590 Descriptoin: allow kill process which wait statahead result
591 Details : for some reasons 'ls' can stick in waiting result from statahead,
592 in this case need way for kill this process.
594 Severity : enhancement
596 Description: Add state history info file, enhance import info file
597 Details : Track import connection state changes in a new osc/mdc proc file;
598 add overview-type data to the osc/mdc import proc file.
600 Severity : enhancement
602 Description: MDS create should not wait for statfs RPC while holding DLM lock.
604 Severity : enhancement
606 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
611 Descriptoin: OST grant too much space to client even there are not enough space.
612 Details : Client will shrink its grant cache to OST if there are no write
613 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
614 this grant cache if there are already not enough avaible space
615 (left_space < total_clients * 32M).
618 Frequency : start MDS on uncleanly shutdowned MDS device
620 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
621 Details : stay in waiting mds<>ost recovery finished produce random bugs
622 due race between two ll_sync thread for one lov target. send
623 ACTIVATE event only if connect realy finished and import have
627 Frequency : rare, connect and disconnect target at same time
629 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
630 Details : don't call obd_disconnect under lov_lock. this long time
631 operation and can block ptlrpcd which answer to connect request.
636 Descriptoin: don't lose wakeup for imp_recovery_waitq
637 Details : recover_import_no_retry or invalidate_import and import_close can
638 both sleep on imp_recovery_waitq, but we was send only one wakeup
642 Frequency : always with long access acl
644 Descriptoin: mds can't pack reply with long acl.
645 Details : mds don't control size of acl but they limited by reint/getattr
648 Severity : enhancement
650 Description: Update to SLES10 kernel-2.6.16.60-0.33.
652 Severity : enhancement
654 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
657 Frequency : start MDS on uncleanly shutdowned MDS device
659 Descriptoin: aborting recovery hang on MDS
660 Details : don't throttle destroy RPCs for the MDT.
663 Frequency : on remount
665 Description: external journal device not working after the remount
666 Details : clear dev_rdonly flag for external journal devices in
672 Description: shutdown vs evict race
673 Details : client_disconnect_export vs connect request race.
674 if client will evicted at this time - we start invalidate
675 thread without referece to import and import can be freed
679 Frequency : rare, need acl's on inode.
681 Description: client can't handle ost additional correctly
682 Details : if ost was added after client connected to mds client can have
683 hit lnet_try_match_md ... to big messages to wide striped files.
684 in this case need teach client to handle config events about add
685 lov target and update client max ea size at that event.
687 Severity : enhancement
689 Description: Changelogs
690 Details : Changelogs are a lightweight mechanism to track filesystem
691 metadata and namespace changes. The changelog is recorded
692 permanently on the MDTs, and is periodically "consumed" / purged
693 when records are no longer needed.
695 Severity : enhancement
697 Description: compact fld format with extents
698 Details : Store range of seq rather than every seq in FLD. Seq
699 controller update FLD rather than clients. In Case of CMD, mdt0
700 has FLD, all other metadata server act as non persistent proxy
701 for FLD queries and cache fld entries in fld cache.
706 Description: don't skip ost target if they assigned to file
707 Details : Drop slow OSCs if we can, but not for requested start idx.
708 This means "if OSC is slow and it is not the requested
709 start OST, then it can be skipped, otherwise skip it only
710 if it is inactive/recovering/out-of-space.
714 Description: more cleanup in mds_lov
715 Details : not send LOV EA under replay, we can't know about they size at this
716 time. Don't allow client connect to mds before any ost connected,
717 for avoid problems with LOV EA size and returning EIO to client.
719 Severity : enhancement
721 Description: Interoperability at server side (Disk interoperability)
723 Severity : enhancement
725 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
727 Severity : enhancement
729 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
731 Severity : enhancement
733 Description: New client IO stack (CLIO).
735 Severity : enhancement
737 Description: Commit on sharing. Eliminate inter-client dependencies between
738 uncommitted transactions by doing transaction commits.
739 Thereby clients may recovery independently.
742 Frequency : Create a symlink file with a very long name
744 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
745 Details : If there is no extra space in the request for early cancels,
746 ldlm_req_handles_avail() returns 0 instead of a negative value.
748 Severity : enhancement
750 Description: Add /proc entry for import status
751 Details : The mdc, osc, and mgc import directories now have
752 an import directory that contains useful import data for debugging
755 Severity : enhancement
757 Description: Re-disable certain /proc logging
758 Details : Enable and disable client's offset_stats, extents_stats and
759 extents_stats_per_process stats logging on the fly.
762 Frequency : Only on FC kernels 2.6.22+
764 Description: oops in statahead
765 Details : Do not drop reference count for the dentry from VFS when lookup,
766 VFS will do that by itself.
768 Severity : enhancement
770 Description: Generic /proc file permissions
771 Details : Set /Proc file permissions in a more generic way to enable non-
772 root users operate on some /proc files.
776 Description: Hitting mdc_commit_close() ASSERTION
777 Details : Properly handle request reference release in
778 ll_release_openhandle().
782 Description: quota recovery deadlock during mds failover
783 Details : This patch includes att18982, att18236, att18237 in bz14840.
785 1. fix osts hang when mds does failover with quotaon
786 2. prevent watchdog storm when osts threads wait for the
791 Frequency : only patchless client
792 Description: add workaround for race between add/remove dentry from hash
794 Severity : enhancement
796 Description: Allow OST glimpses to return PW locks
800 Description: LBUG when llog conf file is full
801 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
805 Description: Prevent import from entering FULL state when server in recovery
809 Description: service mount cannot take device name with ":"
810 Details : Only when device name contains ":/" will mount treat it as
816 Description: replace ptlrpcd with the statahead thread to interpret the async
817 statahead RPC callback
821 Frequency : on recovery
822 Description: I/O failures after umount during fail back
823 Details : if client reconnected to restarted server we need join to recovery
824 instead of find server handler is changed and process self
825 eviction with cancel all locks.
827 Severity : enhancement
829 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
831 Severity : enhancement
833 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
835 Severity : enhancement
837 Description: Upcall on Lustre log has been dumped
838 Details : Allow for a user mode script to be called once a Lustre log has
839 been dumped. It passes the filename of the dumped log to the
840 script, the location of the script can be specified via
841 /proc/sys/lnet/debug_log_upcall.
846 Description: avoid idr_remove called for id which is not allocated.
847 Details : Move assigment s_dev for clustered nfs to end of initialization,
848 to avoid problem with error handling.
853 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
854 Details : When connection is reused this not moved from CONN_UNUSED_HASH
855 into CONN_USED_HASH and this prodice warning when put connection
856 again in unused hash.
859 Severity : enhancement
861 Description: File striping can now be set to use an arbitrary pool of OSTs.
863 Severity : enhancement
865 Description: Export bytes_read/bytes_write count on OSC/OST.
869 Description: Early reply size mismatch, MGC loses connection
870 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
871 the connect flags are properly negotiated.
876 Description: quotas are not honored with O_DIRECT
877 Details : all writes with the flag O_DIRECT will use grants which leads to
878 this problem. Now using OBD_BRW_SYNC to guard this.
882 Description: add quota statistics
883 Details : 1. sort out quota proc entries and proc code.
884 2. add quota statistics
886 Severity : enhancement
888 Description: enable quota support for HEAD.
892 Description: Properly propagate oinfo flags from lov to osc for statfs
893 Details : restore missing copy oi_flags to lov requests.
895 Severity : enhancement
897 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
898 : l_getgroups(8), lst(8), routerstat(8)
899 Details : included man pages for llobdstat(8), llstat(8),
900 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
902 Severity : enhancement
904 Description: configure's --enable-quota should check the
905 : kernel .config for CONFIG_QUOTA
906 Details : configure is terminated if --enable-quota is passed but
907 : no quota support is in kernel
911 Description: Remove portals compatibility
912 Details : Remove portals compatibility, not interoperable with releases
917 Description: Resolve device initialization race
918 Details : Prevent proc handler from accessing devices added to the
919 obd_devs array but yet be intialized.
921 Severity : enhancement
923 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
925 Severity : enhancement
927 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
932 Description: Using wrong pointer in osc_brw_prep_request
933 Details : Access to array[-1] can produce panic if kernel compiled with
934 CONFIG_PAGE_ALLOC enabled
938 Description: Client runs out of low memory
939 Details : Consider only lowmem when counting initial number of llap pages
943 Description: *optional* service tags registration
944 Details : if the "service tags" package is installed on a Lustre node
945 When the filesystem is mounted, a local-node service tag will
946 be created. See http://inventory.sun.com/ for more information
947 about the Service Tags asset management system.
951 Description: Kernel BUG tries to release flock
952 Details : Lustre does not destroy flock lock before last reference goes
953 away. So always drop flock locks when client is evicted and
954 perform unlock regardless of successfulness of speaking to MDS.
958 Description: add refcount protection for osc callbacks, avoid panic on shutdown
962 Description: sanity test 65a fails if stripecount of -1 is set
963 Details : handle -1 striping on filesystem in ll_dirstripe_verify
968 Description: ASSERTION(CheckWriteback(page,cmd)) failed
969 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
972 Severity : enhancement
974 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
978 Description: do not process already freed flock
979 Details : flock can possibly be freed by another thread before it reaches
980 to ldlm_flock_completion_ast.
984 Description: LBUG during stress test
985 Details : Need properly lock accesses the flock deadlock detection list.
989 Description: oops in page fault handler
990 Details : kernel page fault handler can return two special 'pages' in error
991 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
995 Description: timeout with invalidate import.
996 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
997 should be handled by ptlrpcd. This produce long age waiting and
998 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
1000 Severity : enhancement
1002 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
1006 Description: enable MGS and MDT services start separately
1007 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
1008 a co-located MGS without starting the MGS, which is a complement
1009 to 'nosvc' mount option.
1013 Frequency : after recovery
1014 Description: precreate to many object's after del orphan.
1015 Details : del orphan st in oscc last_id == next_id and this triger growing
1016 count of precreated objects. Set flag LOW to skip increase count
1017 of precreated objects.
1021 Frequency : rare, on clear nid stats
1022 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
1023 Details : when clean nid stats sometimes try destroy live entry,
1024 and this produce panic in free.
1028 Description: Stack overflow during MDS log replay
1029 ease stack pressure by using a thread dealing llog_process.
1033 Description: wait until IO finished before start new when do lock cancel.
1034 Details : VM protocol want old IO finished before start new, in this case
1035 need wait until PG_writeback is cleared until check dirty flag and
1036 call writepages in lock cancel callback.
1038 Severity : enhancement
1040 Description: using special macro for print time and cleanup in includes.
1044 Description: mds_mfd_close() ASSERTION(rc == 0)
1045 Details : In mds_mfd_close(), we need protect inode's writecount change
1046 within its orphan write semaphore to prevent possible races.
1050 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
1054 Frequency : rare, on shutdown ost
1055 Description: don't hit live lock with umount ost.
1056 Details : shrink_dcache_parent can be in long loop with destroy dentries,
1057 use shrink_dcache_sb instead.
1061 Description: don't panic with use echo client
1062 Details : echo client pass NULL as client nid pointer and this produce null
1063 pointer dereference.
1067 Description: fix build on ppc32
1068 Details : compile code with -m64 flag produce wrong object file for ppc32.
1072 Description: add message levels for liblustreapi
1076 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
1077 Details : Move the 'good_osts' check before the 'total_bavail' check. This
1078 will result in an -EAGAIN and in the exit call path we call
1079 alloc_rr() which will with increasing aggressiveness attempt to
1080 aquire precreated objects on the minimum number of required OSCs.
1084 Description: Use old size assignment to avoid deadlock
1085 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
1086 the scheduling while holding a spinlock. We do not need locking
1087 for size in ll_update_inode() because size is only updated from
1088 the MDS for directories or files without objects, so there is no
1089 other place to do the update, and concurrent access to such inodes
1090 are protected by the inode lock.
1094 Description: resolve "_IOWR redefined" build error on SLES10
1098 Description: dump the memory debugging after all modules are unloaded to
1099 suppress false negative in conf_sanity test 39
1101 Severity : enhancement
1103 Description: build kernel-ib packages for OFED 1.3 in our release cycle
1108 Description: fix SLES kernel versioning
1109 Details : the kernel version for our SLES 10 kernel did not include a "-"
1110 before the "smp" at the end. while this was not a problem in
1111 general, it did mean that software trying to use the kernel
1112 version to try to detect a vendor specific kernel would fail.
1113 this was most evident by the OFED build scripts.
1117 Description: Don't update lov_desc members until making sure they are valid
1118 Details : When updating lov_desc members via proc fs, need fix their
1119 validities before doing the real update.
1123 Description: don't put request into delay list while invalidate in flight.
1124 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
1125 import in flight. this produce timeout for invalidate and sometimes
1126 can cause stale data.
1130 Frequency : on ppc only
1131 Description: not convert ost objects for directory because it's not exist.
1132 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
1136 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
1138 Severity : enhancement
1140 Description: Update to support 2.6.22.14 vanilla kernel.
1144 Frequency : rare, on recovery
1145 Description: read procfs can produce deadlock in some situation
1146 Details : Holding lprocfs lock which send rpc can produce block for destroy
1147 obd objects and this also block reconnect with -EALREADY. This
1148 isn't fix all lprocfs bugs - but make it rare.
1150 Severity : enhancement
1152 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
1155 Frequency : frequent on X2 node
1157 Description: mdc_set_open_replay_data LBUG
1158 Details : Set replay data for requests that are eligible for replay.
1162 Description: lustre_mgs: operation 101 on unconnected MGS
1163 Details : When MGC is disconnected from MGS long enough, MGS will evict the
1164 MGC, and late on MGC cannot successfully connect to MGS and a lot
1165 of the error messages complaining that MGS is not connected.
1169 Frequency : on network error
1170 Description: panic with double free request if network error
1171 Details : mdc_finish_enqueue is finish request if any network error ocuring,
1172 but it's true only for synchronus enqueue, for async enqueue
1173 (via ptlrpcd) this incorrect and ptlrpcd want finish request
1176 Severity : enhancement
1178 Description: client-side metadata stat-ahead during readdir(directory readahead)
1179 Details : perform client-side metadata stat-ahead when the client detects
1180 readdir and sequential stat of dir entries therein
1183 Frequency : on start mds
1185 Description: Implement get_info(last_id) in obdfilter.
1188 Frequency : occasional
1190 Description: Correctly check stale fid, not start epoch if ost not support SOM
1191 Details : open with flag O_CREATE need set old fid in op_fid3 because
1192 op_fid2 was overwritten with new generated fid, but mds can answer
1193 with one of these two fids and both is not stale. Setattr
1194 incorrectly started an epoch and assume will be called
1195 done_writing, but without SOM done_writing ever being called.
1198 Frequency : rare, depends on device drivers and load
1200 Description: MDS or OSS nodes crash due to stack overflow
1201 Details : Code changes in 1.8.0 increased the stack usage of some functions.
1202 In some cases, in conjunction with device drivers that use a lot
1203 of stack the MDS (or possibly OSS) service threads could overflow
1204 the stack. One change which was identified to consume additional
1205 stack has been reworked to avoid the extra stack usage.
1208 Frequency : occasional
1210 Description: Do not fail import if osc_interpret_create gets -EAGAIN
1211 Details : If osc_interpret_create got -EAGAIN it immediately exits and
1212 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
1213 oscc_has_objects and see OSC has no objests and call
1214 oscc_internal_create to resend create request.
1216 Severity : enhancement
1218 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
1220 Severity : enhancement
1222 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
1225 Frequency : very rare
1227 Description: Fix replay if there is an un-replied request and open
1228 Details : In some cases, older replay request will revert the
1229 mcd->mcd_last_xid on MDS which is used to record the client's
1230 latest sent request.
1232 Severity : enhancement
1234 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
1236 Severity : enhancement
1238 Description: Add rhel5 support to HEAD.
1240 Serverity : enhancement
1242 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
1247 Description: Don't allow skipping OSTs if index has been specified.
1248 Details : Don't allow skipping OSTs if index has been specified, make
1249 locking in internal create lots better.
1253 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
1254 Details : access to bitfield in structure is always rounded to long
1255 and this produce problem with not atomic change any bit.
1259 Description: Lustre make rpms failed.
1260 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
1261 builds Lustre rpms from tarball.
1264 Frequency : rare on shutdown ost
1266 Description: If llog cancel was not send before clean_exports phase, this can
1267 produce deadlock in llog code.
1268 Details : If llog thread has last reference to obd and call class_import_put
1269 this produce deadlock because llog_cleanup_commit_master wait when
1270 last llog_commit_thread exited, but this never success because was
1271 called from llog_commit_thread.
1275 Description: allow userland application know is lost one of stripes.
1276 Details : fill lvb_blocks with error code on ost and return it to
1277 application if error flag found.
1281 Description: NULL lov_tgts causing MDS oops
1282 Details : more safe checks for NULL lov_tgts for avoid oops.
1284 Severity : enhancement
1286 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
1290 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
1291 Details : If an OST is down the MDS will hang indefinitely in
1292 obd_statfs_rqset() waiting for the statfs data. While for
1293 MDS QOS usage of statfs, it should not stuck in waiting.
1295 Severity : enhancement
1297 Description: remote_acl support
1298 Details : Support ACL-based permission check for remote user.
1299 Support setfacl/getfacl for remote user with the utils
1300 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
1301 the system "{s,g}etfacl" utils.
1303 Severity : enhancement
1305 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
1307 Severity : enhancement
1309 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
1313 Description: make mgs_setparam() handle fsname containing dash
1314 Details : fsname containing a dash does not work with lctl conf_param
1316 Severity : enhancement
1318 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
1320 Severity : enhancement
1322 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
1326 Frequency : rare, at shutdown
1327 Description: access already free / zero obd_namespace.
1328 Details : if client_disconnect_export was called without force flag set,
1329 and exist connect request in flight, this can produce access to
1330 NULL pointer (or already free pointer) when connect_interpret
1331 store ocd flags in obd_namespace.
1335 Frequency : only at startup
1336 Description: not alloc memory with spinlock held.
1337 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
1338 if any spinlock held.
1340 Severity : enhancement
1342 Description: make lustre randomly fail allocating memory
1343 Details : Make lustre randomly failed allocating memory for testing purpose.
1345 Severity : enhancement
1347 Description: lost problems with lov objid file
1348 Details : Fixes some scability and access to not inited memory problems
1349 in work with lov objdid file.
1354 Description: lfs find does not continue on file error
1355 Details : Continue other files processing when a file/dir is absent.
1359 Description: Inconsistent usage of lustre_pack_reply()
1360 Details : Standardize the usage of lustre_pack_reply() such that it
1361 always generate a CERROR on failure.
1364 Frequency : occasional
1366 Description: MDS hang or stay in waiting lock
1367 Details : If client receive lock with CBPENDING flag ldlm need send lock
1368 cancel as separate rpc, to avoid situation when cancel request
1369 can't processed due all i/o threads stay in wait lock.
1373 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1377 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1381 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1385 Description: Update to RHEL4 latest kernel.
1387 Severity : enhancement
1389 Description: Build SLES10 patchless client fails
1390 Details : The configure was broken by run ./configure with
1391 --with-linux-obj=.... argument for patchless client. When the
1392 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1393 can't search header adequately. Use absolute path such as
1394 -I($LINUX)/include instead.
1398 Description: interrupt oig_wait produce painc on resend.
1399 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1400 private set, and this produce situation when rq_ptlrpcd_data not
1401 copyed to new allocated request and triggered LBUG on assert
1402 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1403 ptlrpcd set if request is changed and can be safety changed to use
1408 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1409 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1410 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1411 these macros and replace them with their difinitions instead.
1415 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1416 Details : Update which_patch & target file for SLES10 latest kernel.
1419 Description: Remove client patches from SLES 10 kernel.
1420 Details : This causes SLES 10 clients to behave as patchless clients
1421 even on a Lustre-patched (server) kernel.
1423 Severity : enhancement
1425 Description: self-adjustable client's lru lists
1426 Details : use adaptive algorithm for managing client cached locks lru
1427 lists according to current server load, other client's work
1428 pattern, memory activities, etc. Both, server and client
1429 side namespaces provide number of proc tunables for controlling
1432 Severity : enhancement
1434 Description: light-weight GSS support
1435 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
1436 and reduce performance overhead.
1438 Severity : enhancement
1440 Description: Linux keyring support
1441 Details : Support using service of Linux keyring for Lustre GSS internal
1442 context refresh/cache mechanism.
1446 Description: Fix errors in lfs documentation
1447 Details : Fixes man pages
1451 Description: don't use GFP_* in generic Lustre code.
1452 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
1456 Description: obdclass is limited by single OBD_ALLOC(idarray)
1457 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
1461 Description: warnings with build patchless client with vanila 2.6.19 and up
1462 Details : change old ctl_table style and replace ctl_table/ctl_table_header
1463 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
1467 Frequency : Only for SLES
1468 Description: sanity_quota fail test_1
1469 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
1470 file, which makes TSTID[2] inunique.
1474 Frequency : Only for SLES9
1475 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
1476 Details : Update target/ChangeLog/which_patch .
1480 Frequency : Only for test_10 in sanity-quota.sh
1481 Description: a bug in quota test code
1482 Details : checking a test flag in a wrong place
1486 Frequency : Only for test_13 in sanity-quota.sh
1487 Description: a bug in quota test script
1488 Details : shouldn't check overfull allocated quota
1492 Description: jbd statistics
1493 Details : Port older jbd statistics patch for sles10
1495 Severity : enhancement
1497 Description: Add printing inode info into message about error in writepage.
1502 Description: reply_lock_interpret crash due to race with it and lock cancel.
1503 Details : Do not replay locks that are being cancelled. Do not reference
1504 locks by their address during replay, just by their handle.
1508 Frequency : When flocks are used.
1509 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
1510 Details : Flock locks might destroy just granted lock if it could be merged
1511 with another existing flock, this is done in completion handler,
1512 so teach ldlm_cli_enquque_fini that this is a valid case for
1518 Description: Oops in read and write path when failing to allocate lock.
1519 Details : Check if lock allocation failed and return error back.
1523 Description: lstripe command fails for valid OST index
1524 Details : The stripe offset is compared to lov->desc.ld_tgt_count
1525 instead of lov->desc.ld_active_tgt_count.
1529 Description: sanity.sh failed test 103
1530 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
1535 Description: df doesn't work properly if diskfs blocksize != 4K
1536 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
1540 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
1541 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
1543 Severity : enhancement
1545 Description: Adaptive timeouts
1546 Details : RPC timeouts adapt to changing server load and network
1547 conditions to reduce resend attempts and improve recovery time.
1551 Description: llapi_file_create() does not allow some changes
1552 Details : add llapi_file_open() that allows specifying the mode and
1553 open flags, and also returns an open file handle.
1557 Description: merge and cleanup kernel patches.
1558 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
1562 Description: Add journal checksum support.(Kernel part)
1563 Details : The journal checksum feature adds two new flags i.e
1564 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
1565 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
1566 indicates that the commit block contains the checksum for
1567 the blocks described by the descriptor blocks. Now commit
1568 record can be sent to disk without waiting for descriptor
1569 blocks to be written to disk. This behavior is controlled
1570 using JBD2_FEATURE_ASYNC_COMMIT flag.
1574 Description: OSS needs mutliple precreate threads
1575 Details : Add ability to start more than one create thread per OSS.
1579 Description: Sanity.sh test_65k failed: setstripe should have succeeded
1580 Details : execute lfs setstripe on client
1584 Description: mds_obd_create error creating tmp object
1585 Details : When the user sets quota on root, llog will be affected and can't
1586 create files and write files.
1589 Frequency : Always on ia64 patchless client, and possibly others.
1591 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
1592 Details : This allows the patchless client to be loaded on architectures
1593 without this export.
1597 Description: RedHat Update kernel for RHEL5
1598 Details : Add the kernel config file for RHEL5.
1602 Description: RedHat Update kernel for RHEL5
1603 Details : Modify the kernel config file more closer RHEL5.
1607 Description: Build failure against Centos5 (RHEL5)
1608 Details : Define PAGE_SIZE when it isn't present.
1612 Description: test_7 on sanity-quota.sh can't be used on separate machines
1613 Details : add facet to handle it
1617 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
1618 Details : with NFS, the anon dentry's parent was set to itself in
1619 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
1620 corresponding dentry other than use rec->ur_name.
1622 Severity : enhancement
1624 Description: lfs setstripe enhancement
1625 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
1629 Description: enable data checksumming by default
1630 Details : enable checksum by default, allow --disable-checksum
1631 configure option and "-o nochecksum" mount option. Checksums
1632 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
1636 Description: lustre support for RHEL5
1637 Details : Add support for RHEL5.
1641 Description: Client eviction due to failover config
1642 Details : after a connection loss, the lustre client should attempt to
1643 reconnect to the last active server first before trying the
1644 other potential connections.
1648 Description: when mds and osts use different quota unit(32bit and 64bit),
1649 quota will be released repeatly.
1650 Details : void sending multiple quota reqs to mds, which will keep the
1651 status between the reqs.
1655 Description: osts not allocated evenly to files
1656 Details : change the condition to increase offset_idx
1659 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
1661 Description: Data corruption for OSTs that are formatted larger than 2TB
1663 Details : When generating the bio request for lustre file writes the
1664 sector number would overflow a temporary variable before being
1665 used for the IO. The data reads correctly from Lustre (which
1666 will overflow in a similar manner) but other file data or
1667 filesystem metadata may be corrupted in some cases.
1671 Description: Tune the kernel for good SCSI performance.
1672 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
1673 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
1678 Description: rewrite ext2-derived code in obdclass/uuid.c
1679 Details : rewrite inherited code (uuid parsing code from ext2 utils)
1680 from scratch preserving functionality.
1684 Description: "lfs find -obd UUID" prints directories
1685 Details : "lfs find -obd UUID" will return all directory names instead
1686 of just file names. It is incorrect because the directories
1687 do not reside on the OSTs.
1691 Description: use i_size_read and i_size_write in 2.6 port
1692 Details : replace inode->i_size access with i_size_read/write()
1695 Frequency : when removing large files
1697 Description: scheduling issue during removal of large Lustre files
1698 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
1699 It causes scheduling issues when removing large files (17TB in the
1703 Frequency : only with liblustre clients on XT3
1705 Description: evictions taking too long
1706 Details : allow llrd to evict clients directly on OSTs
1709 Frequency : only on ppc
1711 Description: /proc/fs/lustre/devices broken on ppc
1712 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
1713 We should make sure the type of 'index' is loff_t and then cast
1714 explicitly as needed below. Do not assign an explicitly cast
1718 Frequency : only for rhel5
1720 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
1721 Details : Modify the target file & which_kernel.
1723 Severity : enhancement
1725 Description: omit set fsid for export NFS
1726 Details : fix set/restore device id for avoid EMFILE error and mark lustre
1727 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
1731 Frequency : Always, for kernels after 2.6.16
1732 Description: Fix warning idr_remove called for id=.. which is not allocated.
1733 Details : Last kernels save old s_dev before kill super and not allow
1734 to restore from callback - restore it before call kill_anon_super.
1738 Description: buffer overruns could theoretically occur
1739 Details : llapi_semantic_traverse() modifies the "path" argument by
1740 appending values to the end of the origin string, and a
1741 overrun may occur. Adding buffer overrun check in liblustreapi.
1745 Description: Fix error on 'ls .' at the top of the Lustre mount.
1746 Details : Don't revalidate dentry if it is a root dentry.
1750 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
1751 Details : Modify vm-tunables-rhel4.patch.
1755 Description: Kernel config for 2.6.18-vanilla.
1756 Details : Modify targets/2.6-vanilla.target.in.
1757 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
1758 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
1759 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
1760 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
1764 Description: improve handling recoverable errors
1765 Details : if request processig with error which can be recoverable on server
1766 request should be resend, otherwise page released from cache and
1771 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
1772 Details : Modify target file & which_patch.
1773 A flaw was found in the IA32 system call emulation provided
1774 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
1775 value could be stored in the %RAX register, which could trigger an
1776 out-of-bounds system call table access. An untrusted local user
1777 could exploit this flaw to run code in the kernel
1778 (ie a root privilege escalation). (CVE-2007-4573).
1782 Description: O_DIRECT bypasses client statistics.
1783 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
1784 read_ahead_stats were updated. copied the stats section from
1785 osc_send_oap_rpc() into async_internal().
1789 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
1793 Description: change order of libsysio includes
1794 Details : '#include sysio.h' should always come before '#include xtio.h'
1798 Description: handle "serious error: objid * already exists" more gracefully
1799 Details : If LAST_ID value on disk is smaller than the objects existing in
1800 the O/0/d* directories, it indicates disk corruption and causes an
1801 LBUG(). If the object is 0-length, then we should use the existing
1802 object. This will help to avoid a full fsck in most cases.
1804 Severity : enhancement
1806 Description: adapt the lustre_config script to support the upgrade case
1807 Details : Add "-u" option for lustre_config script to support upgrading 1.4
1808 server targets to 1.6 in parallel.
1812 Description: To avoid grant space > avaible space when the disk is almost
1813 full. Without this patch you might see the error "grant XXXX >
1814 available" or some LBUG about grant, when the disk is almost
1816 Details : In filter_check_grant, for non_grant cache write, we should
1817 check the left space by if (*left > ungranted + bytes), instead
1818 of (*left > ungranted), because only we are sure the left space
1819 is enough for another "bytes", then the ungrant space should be
1820 increase. In client, we should update cl_avail_grant only there
1821 is OBD_MD_FLGRANT in the reply.
1825 Description: Update RHEL 4 kernel to fix local root privilege escalation.
1826 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
1827 described in CVE-2007-4573. This problem could allow untrusted
1828 local users to gain root access.
1831 Frequency : when using O_DIRECT and quotas
1833 Description: Incorrect file ownership on O_DIRECT output files
1834 Details : block usage reported by 'lfs quota' does not take into account
1835 files that have been written with O_DIRECT.
1840 Description: touch file failed when fs is not full
1841 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
1842 otherwise we can get ENOSP while fs is not full.
1846 Description: parallel lock callbacks
1847 Details : Instead of sending blocking and completion callbacks as separated
1848 requests, adding them to a set and sending in parallel.
1851 Frequency : only for Cray XT3
1852 Bugzilla : 12829/13455
1853 Description: Changing primary group doesn't change the group lustre assigns to
1855 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
1856 overridden with the primary group provided by the group upcall,
1857 whereas the supplied fsgid can be trusted if it is in the list of
1858 supplementary groups returned by the group upcall.
1860 Severity : enhancement
1862 Description: Allow masking D_WARNING, D_ERROR messages from console
1863 Details : Console messages can now be disabled via lnet.printk.
1867 Description: User code with malformed file open parameter crashes client node
1868 Details : Before packing join_file req, all the related reference should be
1869 checked carefully in case some malformed flags cause fake
1870 join_file req on client.
1874 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
1875 Details : ldlm_completion_ast() assumes that a lock is granted when the req
1876 mode is equal to the granted mode. However, it should also check
1877 that LDLM_FL_CP_REQD is not set.
1881 Description: Heavy nfs access might result in deadlocks
1882 Details : After ELC code landed, it is now improper to enqueue any mds
1883 locks under och_sem, because enqueue might want to decide to
1884 cancel open locks for same inode we are holding och_sem for.
1888 Description: Client eviction while running blogbench
1889 Details : A lot of unlink operations with concurrent I/O can lead to a
1890 deadlock causing evictions. To address the problem, the number of
1891 oustanding OST_DESTROY requests is now throttled to
1892 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
1893 callbacks are processed in priority.
1897 Description: enable ACLs on MDS by default
1898 Details : ACLs must be enabled on MDS by default.
1901 Frequency : PPC/PPC64 only
1903 Description: conflicts between asm-ppc64/types.h and lustre_types.h
1904 Details : fix duplicated definitions between asm-ppc64/types.h and
1905 lustre_types.h on PPC.
1908 Frequency : PPC/PPC64 only
1910 Description: asm-ppc/segment.h does not exist
1911 Details : fix compile issue on PPC.
1915 Description: better handle error messages in extents code
1918 Frequency : RHEL4 only
1920 Description: mkfs is very slow on IA64/RHEL4
1921 Details : A performance regression has been discovered in the MPT Fusion
1922 driver between versions 3.02.73rh and 3.02.99.00rh. As a
1923 consequence, we have downgraded the MPT Fusion driver in the RHEL4
1924 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
1926 Severity : enhancement
1928 Description: SNMP support enhancement
1929 Details : Adding total number of sampled request for an MDS node in snmp
1932 Severity : enhancement
1934 Description: Optimize ldlm waiting list processing for PR extent locks
1935 Details : When processing waiting list for read extent lock and meeting read
1936 lock that is same or wider to it that is not contended, skip
1937 processing rest of the list and immediatelly return current
1938 status of conflictness, since we are guaranteed there are no
1939 conflicting locks in the rest of the list.
1943 Description: Time out and refuse to reconnect
1944 Details : When the failover node is the primary node, it is possible
1945 to have two identical connections in imp_conn_list. We must
1946 compare not conn's pointers but NIDs, otherwise we can defeat
1947 connection throttling.
1951 Description: port llog fixes from b1_6 into HEAD
1952 Details : Port llog reference couting and some llog cleanups from b1_6
1953 (bug 10800) into HEAD, for protect from panic and access to already
1954 free llog structures.
1958 Description: Detect stride IO mode in read-ahead
1959 Details : When a client does stride read, read-ahead should detect that and
1960 read-ahead pages according to the detected stride pattern.
1964 Description: data checksumming impacts single node performance
1965 Details : add support for several checksum algorithm. Currently, only CRC32
1966 and Adler-32 are supported. The checksum type can be changed on
1967 the fly via /proc/fs/lustre/osc/*/checksum_type.
1971 Description: use adler32 for page checksums
1972 Details : when available, use the Adler-32 algorithm instead of CRC32 for
1977 Description: build for x2 fails
1978 Details : fix compile issue on Cray systems.
1982 Description: Properly match for duplicate locks
1983 Details : Due to different lock order from skiplists code, we need to
1984 traverse entire list for now
1987 Frequency : only on PPC/SLES10
1989 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
1990 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
1991 be defined. Add a hack in mkfs_lustre.c to work around this compile
1996 Description: LASSERT on MDS when client holding flock lock dies
1997 Details : ldlm pool logic depends on number of granted locks equal to
1998 number of released locks which is not true for flock locks, so
1999 just exclude such locks from consideration.
2003 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
2004 Details : Use fsfilt_sync() for both the whole filesystem sync and
2005 individual file sync to eliminate dangerous inode locking
2006 with I_LOCK that can lead to a deadlock.
2010 Description: performance in 1.6.3
2011 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
2012 sure that 1MB requests are not fragmented by the block layer.
2014 Severity : enhancement
2016 Description: organize the server-side client stats on per-nid basis
2017 Details : Change the structure of stats under obdfilter and mds to
2025 The "uuid"s file would list the uuids of _active_ exports.
2026 And the clear entry is to clear all stats and stale nids.
2028 Severity : enhancement
2030 Description: eliminate client locks in face of contention
2031 Details : file contention detection and lockless i/o implementation
2032 for contended files.
2036 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
2039 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
2041 Description: MDS may lose file striping (and hence file data) in some cases
2042 Details : If there are additional extended attributes stored on the MDS,
2043 in particular ACLs, SELinux, or user attributes (if user_xattr
2044 is specified for the client mount options) then there is a risk
2045 of attribute loss. Additionally, the Lustre file striping
2046 needs to be larger than default (e.g. striped over all OSTs),
2047 and an additional attribute must be stored initially in the
2048 inode and then increase in size enough to be moved to the
2049 external attribute block (e.g. ACL growing in size) for file
2054 Description: skiplist implementation simplification
2055 Details : skiplists are used to group compatible locks on granted list
2056 that was implemented as tracking first and last lock of each
2057 lock group the patch changes that to using doubly linked lists
2061 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
2062 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
2063 because LASSERT on both the data supplied by a client, and the
2064 data on disk is dangerous and incorrect.
2066 Severity : enhancement
2068 Description: Slow truncate/writes to huge files at high offsets.
2069 Details : Directly associate cached pages to lock that protect those pages,
2070 this allows us to quickly find what pages to write and remove
2071 once lock callback is received.
2075 Description: more ldlm soft lockups
2076 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
2077 starve other threads from the resource lock for a long time in
2078 case of long waiting queue, so change the debug level from
2079 D_OTHER to the less frequently used D_INFO.
2081 Severity : enhancement
2083 Description: add -gid, -group, -uid, -user options to lfs find
2087 Description: Hung threads in invalidate_inode_pages2_range
2088 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
2089 once one is completed. As a result, some RPCs are stuck in the
2090 queue and are never sent.
2094 Description: filter threads hungs on waiting journal commit
2095 Details : Cleanup filter group llog code, then only filter group llog will
2096 be only created in the MDS/OST syncing process.
2100 Description: Procfs and llog threads access destoryed import sometimes.
2101 Details : Sync the import destoryed process with procfs and llog threads by
2102 the import refcount and semaphore.
2104 Severity : enhancement
2106 Description: openlock cache of b1_6 port to HEAD
2111 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
2112 Details : If insertion of an extent fails, then discard the inode
2113 preallocation and free data blocks else it can lead to duplicate
2118 Description: don't always update ctime in ext3_xattr_set_handle()
2119 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
2120 In some cases the ctime should not be updated, for example for
2121 2.0->1.8 compatibility it is necessary to delete an xattr and it
2122 should not update the ctime.
2126 Bugzilla : 15713/16362
2127 Description: Assertion in iopen_connect_dentry in 1.6.3
2128 Details : looking up an inode via iopen with the wrong generation number can
2129 populate the dcache with a disconneced dentry while the inode
2130 number is in the process of being reallocated. This causes an
2131 assertion failure in iopen since the inode's dentry list contains
2132 both a connected and disconnected dentry.
2136 Description: assertion failure in ldlm_handle2lock()
2137 Details : fix a race between class_handle_unhash() and class_handle2object()
2138 introduced in lustre 1.6.5 by bug 13622.
2143 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
2144 Details : remove the limit of 256 scsi disks in the sd_iostat patch
2149 Description: oops in sd_iostats_seq_show()
2150 Details : unloading/reloading the scsi low level driver triggers a kernel
2151 bug when trying to access the sd iostat file.
2156 Description: Kernel panics during QLogic driver reload
2157 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
2158 patch, causing memory corruption.
2163 Description: journal_dev option does not work in b1_6
2164 Details : pass mount option during pre-mount.
2166 Severity : enhancement
2168 Description: Add a FIEMAP(FIle Extent MAP) ioctl
2169 Details : FIEMAP ioctl will allow an application to efficiently fetch the
2170 extent information of a file. It can be used to map logical blocks
2171 in a file to physical blocks in the block device.
2175 Description: LDLM soft lockups - improvement
2176 Details : It is be possible to send the lock handle along with each read
2177 or write request because the client is already doing a lock match
2178 itself so there isn't any reason the OST should have to re-do that
2182 Frequency : only on Cray X2
2184 Description: X2 build failures
2185 Details : fix build failures on Cray X2.
2189 Description: xid & resent requests
2190 Details : Initialize RPC XID from clock at startup (randomly if clock is
2193 Severity : enhancement
2195 Description: Add lustre_start utility to start or stop multiple Lustre servers
2200 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
2201 Details : In case of memory pressure, list_del() can be called twice on
2202 req->rq_history_list, causing a kernel oops.
2206 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
2207 Details : fix stack overflow in the distributed lock manager by defering
2208 export eviction after a failed AST to the elt thread instead of
2209 handling it in the dlm interpret routine.
2213 Description: Convert some comments to new format.
2214 Details : Update documenting comments to match doxygen conventions.
2218 Description: Grammar fixes.
2219 Details : A couple of trivial spelling fixes.
2223 Description: OSD_COUNTERS-mandatory
2224 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
2225 to simplify the code.
2229 Description: simplify cmm_device freeing logic.
2230 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
2234 Description: Add lockdep support to dt_object_operations locking interface.
2235 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
2236 parameter indicating lock ordering. Update mdd code to use new
2241 Description: Introduce failloc constants for lockless IO tests.
2242 Details : Add two new failloc constants to test lockless IO. Only one of
2243 them in implemented---another is checked in yet to be landed
2248 Description: Add lockdep support for inode mutex.
2249 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
2250 in the situations where Lustre has to lock more than one inode
2255 Description: Add optional invariants checking support.
2256 Details : Add new LINVRNT() macro, optional on new --enable-invariants
2257 configure switch. This macro is to be used for consistency and
2258 sanity checks that are too expensive to be left in `production'
2263 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
2264 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
2269 Description: Add lockdep support for ldlm_lock and ldlm_resource.
2270 Details : Use spin_lock_nested() in (the only) situation where more than
2271 one ldlm_lock is locked simultaneously. Also, fix possible
2272 dead-lock in ldlm_lock_change_resource() by enforcing particular
2277 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
2278 Details : Instead of specifying each ldlm_lock call-back through separate
2279 parameter, wrap them into struct ldlm_callback_suite.
2283 Description: Kill join_lru obd method and its callers.
2284 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
2285 in memory---a requirement implemented through ->o_join_lru() obd
2286 method in HEAD. Remove this method and its users.
2290 Description: Add asynchronous ldlm ENQUEUE completion handler.
2291 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
2292 case that stock ldlm_completion_ast() cannot handle as it waits
2293 until lock is granted. Introduce new ldlm_completion_ast_async()
2294 for this. Also comment ldlm_completion_ast().
2298 Description: ldlm_error <-> errno conversion.
2299 Details : Add functions to map (rather arbitrary) between LDLM error codes
2300 and standard errno values. CLIO needs this to prevent LDLM specific
2301 constants from escaping ldlm and osc.
2305 Description: Kill unused ldlm_handle2lock_ns() function.
2306 Details : Kill unused ldlm_handle2lock_ns() function.
2310 Description: Add lu_ref support to ldlm_lock
2311 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
2312 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
2313 and ldlm_resource. LDLM interface has to be changed, because code that
2314 releases a reference on a lock, has to "know" what reference this is.
2315 In the most frequent case
2317 lock = ldlm_handle2lock(handle);
2319 LDLM_LOCK_PUT(lock);
2321 no changes are required. When any other reference (received _not_ from
2322 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
2323 instead of LDLM_LOCK_PUT().
2325 Arguably, changes are pervasive, and interface requires some discipline
2326 for proper use. On the other hand, it was very instrumental in finding
2327 a few leaked lock references.
2331 Description: Add ldlm_lock_addref_try().
2332 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
2333 attempts to addref a lock that might be being canceled
2338 Description: Add ldlm_weigh_callback().
2339 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
2340 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
2341 hard-coded `number of pages' logic.
2345 Description: Add lockdep annotations to llog code.
2346 Details : Use appropriately tagged _nested() locking calls in the places
2347 where llog takes more than one ->lgh_lock lock.
2351 Description: Add loi_kms_set().
2352 Details : Wrap kms updates into a helper function.
2356 Description: Constify instances of struct lsm_operations.
2357 Details : Constify instances of struct lsm_operations.
2361 Description: lu_conf support.
2362 Details : On a server, a file system object is uniquely identified
2363 by a fid, which is sufficient to locate and load all object
2364 state (inode). On a client, on the other hand, more data are
2365 necessary instantiate an object. Change lu_object_find() and
2366 friends to take additional `lu_conf' argument describing object.
2367 Typically this includes layout information.
2371 Description: lu_context fixes.
2372 Details : Introduce new lu_context functions that are needed on the client
2373 side, where some system threads (ptlrpcd) are shared by multiple
2374 modules, and so cannot be stopped during module shutdown.
2378 Description: Add start and stop methods to lu_device_type_operations.
2379 Details : Introduce two new methods in lu_device_type_operations, that are
2380 invoked when first instance of a given type is created and last one
2381 is destroyed respectively. This is need by CLIO.
2385 Description: Add lu_ref support to struct lu_device.
2386 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
2387 track leaked references.
2391 Description: Introduce lu_kmem_descr.
2392 Details : lu_kmem_descr and its companion interface allow to create
2393 and destroy a number of kmem caches at once.
2397 Description: Fix lu_object finalization race.
2398 Details : Fix a race between lu_object_find() finding an object and its
2399 concurrent finalization. This race is (most likely) not possible
2400 on the server, but might happen on the client.
2404 Description: Introduce lu_ref interface.
2405 Details : lu_ref is a debugging module allowing to track references to
2406 a given object. It is quite cpu expensive, and has to be
2407 explicitly enabled with --enable-lu_ref. See usage description
2412 Description: Factor lu_site procfs stats into a separate function.
2413 Details : Separate lu_site stats printing code into a separate function
2414 that can be reused on a client.
2418 Description: Constify instances of struct {lu,dt,md}_device_operations.
2419 Details : Constify instances of struct {lu,dt,md}_device_operations.
2423 Description: Introduce struct md_site and move meta-data specific parts of
2424 struct lu_site here.
2425 Details : Move md-specific fields out of struct lu_site into special struct
2426 md_site, so that lu_site can be used on a client.
2430 Description: Kill mdd_lov_destroy().
2431 Details : Remove unused mdd code.
2435 Description: Add st_block checking to multistat.c.
2436 Details : Add st_block checking to multistat.c.
2440 Description: Add lu_ref support to struct obd_device.
2441 Details : Add lu_ref tracking to obd_device.
2445 Description: Kill obd_set_fail_loc().
2446 Details : Remove unused code.
2450 Description: Add special type for ptlrpc_request interpret functions.
2451 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
2452 there. Actual usage will be in CLIO.
2456 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
2457 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
2458 doesn't work with lockdep.
2462 Description: Add rwv.c test program.
2463 Details : New testing program exercising readv(2) and writev(2) (Qian).
2467 Description: Add sendfile.c test program.
2468 Details : New testing program exercising sendfile(2) (Jay).
2472 Description: Ratelimit a message that can be very frequent.
2473 Details : Ratelimit a memory allocation failure message that can
2478 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2479 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2483 Description: Liblustre build fixes.
2484 Details : Liblustre build fixes.
2488 Description: libcfs: add cfs_{need,cond}_resched() interface.
2489 Details : libcfs: add cfs_{need,cond}_resched() definition and
2490 implementations for Linux, NT, and liblustre.
2492 Severity : enhancement
2494 Description: More exported tunables for mballoc
2495 Details : Add support for tunable preallocation window and new tunables for
2496 large/small requests
2500 Description: Detect corruption of block bitmap and checking for preallocations
2501 Details : Checks validity of on-disk block bitmap. Also it does better
2502 checking of number of applied preallocations. When corruption is
2503 found, it turns filesystem readonly to prevent further corruptions.
2507 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
2508 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
2509 window to 0. If it does hit cache miss, and read-pattern is still
2510 stride-io mode, does not reset the stride window, but also does not
2511 increase the stride window length in this case.
2515 Frequency : only for big-endian servers
2516 Description: Check if system is big-endian while mounting fs with extents feature
2517 Details : Mounting a filesystem with extents feature will fail on big-endian
2518 systems since ext3-based ldiskfs is not supported on big-endian
2519 systems. This can be over-riden with "bigendian_extents" mount option.
2521 Severity : enhancement
2523 Description: The root squash functionality
2524 Details : A security feature, which is to prevent users from being able
2525 to mount lustre on their desktop, run as root, and delete
2526 all of the files in the filesystem. The goal is accomplished by
2527 remapping user id (UID) and group id (GID) of the root user to
2528 a UID and GID specified by the system administartor via Lustre
2529 configuration management server (MGS). The functionality also
2530 allows to specify sets of clients for which the remapping does
2535 Description: Excessive recovery window
2536 Details : With AT enabled, the recovery window can be excessively long (6000+
2537 seconds). To address this problem, we no longer use
2538 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
2539 timeout no longer depends on the service time, it is set to
2540 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
2541 time via pb_service_time.
2545 Description: Watchdog triggered on MDS failover
2546 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
2547 the OSTs know that the MDS "UUID" can be reused for the same export
2548 from a different NID, so we do not need to wait for the export to be
2552 Frequency : rare, only if using MMP with Linux RAID
2554 Description: MMP doesn't work with Linux RAID
2555 Details : While using HA for Lustre servers with Linux RAID, it is possible
2556 that MMP will not detect multiple mounts. To make this work we
2557 need to unplug the device queue in RAID when the MMP block is being
2558 written. Also while reading the MMP block, we should read it from
2559 disk and not the cached one.
2561 Severity : enhancement
2563 Description: open file using fid
2564 Details : A file can be opened using just its fid, like
2565 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
2568 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
2571 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
2572 Details : add proper locking for creation of jbd_4k slab cache
2576 Description: MMP check in ext3_remount() fails without displaying any error
2577 Details : When multiple mount protection fails during remount, proper error
2580 Severity : enhancement
2582 Description: Allow stripe size to be up to 4G-64k
2583 Details : Fix math logic to allow large stripe sizes.
2587 Description: add check for >8TB ldiskfs filesystems
2588 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
2589 Don't allow >8TB ldiskfs filesystems to be mounted without
2590 force_over_8tb mount option
2595 Descriptoin: OSS complains "too many blocks" and crashes
2596 Details : In filter_commitrw_write():
2597 1. lquota_chkquota();
2598 2. obdo_from_inode();
2599 3. lquota_pending_commit();
2600 "2." has changed uid and gid in this case, then caused
2601 lquota_chkquota() and lquota_pending_commit() to use
2602 different uid and gid. That is the root of the bug.
2606 Descriptoin: getstripe doesn't report stripe count = -1 sanely
2607 Details : Detect and print out the 'stripe count = -1' case properly.
2611 Description: Out or order replies might be lost on replay
2612 Details : In ptlrpc_retain_replayable_request if we cannot find retained
2613 request with tid smaller then one currently being added, add it
2614 to the start, not end of the list.
2616 --------------------------------------------------------------------------------
2618 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
2620 * Support for kernels:
2621 2.6.5-7.283 (SLES 9),
2622 2.6.9-55.EL (RHEL 4),
2623 2.6.16.46-0.14 (SLES 10),
2624 2.6.18.8 vanilla (kernel.org)
2625 * Client support for unpatched kernels:
2626 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2627 2.6.16 - 2.6.22 vanilla (kernel.org)
2628 * Due to recently discovered recovery problems, we do not recommend
2629 using patchless RHEL 4 clients with this or any earlier release.
2630 * Recommended e2fsprogs version: 1.39.cfs8
2631 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2632 * Starting with this release, the ldiskfs backing filesystem required
2633 by Lustre is now in its own package, lustre-ldiskfs. This package
2634 should be installed. It is versioned separately from Lustre and
2635 may be released separately in future.
2639 Description: block reactivating mgc import until all deactivates complete
2640 Details : Fix race when failing back MDT/MGS to itself (testing)
2643 Frequency : at statup only
2645 Description: mds_lov_synchronize race leads to various problems
2646 Details : simultaneous MDT->OST connections at startup can cause the
2647 sync to abort, leaving the OSC in a bad state.
2649 Severity : enhancement
2651 Description: add optional extra BUILD_VERSION info
2652 Details : add a new environment variable (namely LUSTRE_VERS) which allows
2653 to override the lustre version.
2656 Frequency : 2.6.18 servers only
2658 Description: ll_kern_mount() doesn't release the module reference
2659 Details : The ldiskfs module reference count never drops down to 0
2660 because ll_kern_mount() doesn't release the module reference.
2665 Description: server LBUG when using old ost_num_threads parameter
2666 Details : Accept the old ost_num_threads parameter but warn that it
2667 is deprecated, and fix an off-by-one error that caused an LBUG.
2672 Description: Transient SCSI error results in persistent IO issue
2673 Details : iobuf->dr_error is not reinitialized to 0 between two
2677 Frequency : sometimes when underlying device returns I/O errors
2679 Description: OSTs not going read-only during write failures
2680 Details : OSTs are not remounted read-only when the journal commit threads
2681 get I/O errors because fsfilt_ext3 calls journal_start/stop()
2682 instead of the ext3 wrappers.
2686 Description: poor connect scaling with increasing client count
2687 Details : Don't run filter_grant_sanity_check for more than 100 exports
2688 to improve scaling for large numbers of clients.
2691 Frequency : SLES10 only
2693 Description: sanity-quota.sh quotacheck failed: rc = -22
2694 Details : Quotas cannot be enabled on SLES10.
2697 Frequency : liblustre clients only
2699 Description: getdirentries does not give error when run on compute nodes
2700 Details : getdirentries does not fail when the size specified as an argument
2701 is too small to contain at least one entry
2703 Severity : enhancement
2705 Description: Add LNET router traceability for debug purposes
2706 Details : If a checksum failure occurs with a router as part of the
2707 IO path, the NID of the last router that forwarded the bulk data
2708 is printed so it can be identified.
2713 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
2714 Details : Due to a race condition, liblustre clients were occasionally
2715 evicted incorrectly.
2717 Severity : enhancement
2719 Description: lfs setstripe use optional parameters instead of postional
2722 Severity : enhancement
2724 Description: Nanosecond timestamp support for ldiskfs
2725 Details : The on-disk ldiskfs filesystem has added support for nanosecond
2726 resolution timestamps. There is not yet support for this at
2727 the Lustre filesystem level.
2730 Frequency : during server recovery
2732 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
2733 Details : request with rq_no_resend flag not awake l_wait_event if they get
2737 Frequency : nfs export on patchless client
2739 Description: connectathon hang when test nfs export over patchless client
2740 Details : Disconnected dentry cannot be found with lookup, so we do not need
2741 to unhash it or make it invalid
2744 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
2745 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
2746 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
2751 Description: open req refcounting wrong on reconnect
2752 Details : If reconnect happened between getting open reply from server and
2753 call to mdc_set_replay_data in ll_file_open, we will schedule
2754 replay for unreferenced request that we are about to free.
2755 Subsequent close will crash in variety of ways.
2756 Check that request is still eligible for replay in
2757 mdc_set_replay_data().
2762 Description: disable writes to filesystem when reading health_check file
2763 Details : the default for reading the health_check proc file has changed
2764 to NOT do a journal transaction and write to disk, because this
2765 can cause reads of the /proc file to hang and block HA state
2766 checking on a healthy but otherwise heavily loaded system. It
2767 is possible to return to the previous behaviour during configure
2768 with --enable-health-write.
2770 Severity : enhancement
2772 Description: 64-bit inode version
2773 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
2774 to the inode. This will be required for version-based recovery.
2779 Description: MDS fails to start if a duplicate client export is detected
2780 Details : in some rare cases it was possible for a client to connect to
2781 an MDS multiple times. Upon recovery the MDS would detect this
2782 and fail during startup. Handle this more gracefully.
2784 Severity : enhancement
2786 Description: Add -o localflock option to simulate old noflock
2788 Details : This will achieve local-only flock/fcntl locks
2794 Description: log_commit_thread vs filter_destroy race leads to crash
2795 Details : Take import reference before releasing llog record semaphore
2800 Description: Wrong request locking in request set processing
2801 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
2802 imp_delayed_list, in this place should be used imp_lock.
2805 Frequency : when reconnection
2807 Description: Grant Leak when osc reconnect to OST
2808 Details : When osc reconnect ost, OST(filter) should check whether it
2809 should grant more space to client by comparing fed_grant and
2810 cl_avail_grant, and return the granted space to client instead
2811 of "new granted" space, because client will call osc_init_grant
2812 to update the client grant space info.
2815 Frequency : when client reconnect to OST
2817 Description: Grant Leak when osc do resend and replay bulk write
2818 Details : When osc reconnect to OST, OST(filter)should clear grant info of
2819 bulk write request, because the grant info while be sync between
2820 OSC and OST when reconnect, and we should ignore the grant info
2821 these of resend/replay write req.
2826 Description: Grant space more than avaiable left space sometimes.
2827 Details : When then OST is about to be full, if two bulk writing from
2828 different clients came to OST. Accord the avaliable space of the
2829 OST, the first req should be permitted, and the second one
2830 should be denied by ENOSPC. But if the seconde arrived before
2831 the first one is commited. The OST might wrongly permit second
2832 writing, which will cause grant space > avaiable space.
2835 Frequency : when client is evicted
2837 Description: Grant might be wrongly erased when osc is evicted by OST
2838 Details : when the import is evicted by server, it will fork another
2839 thread ptlrpc_invalidate_import_thread to invalidate the
2840 import, where the grant will be set to 0. While the original
2841 thread will update the grant it got when connecting. So if
2842 the former happened latter, the grant will be wrongly errased
2843 because of this race.
2848 Description: Checking Stale with correct fid
2849 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
2850 to check whether it is stale, because sometimes, we want the
2851 enqueue happened anyway, and op_data.fid2 will not be initialized.
2853 Severity : enhancement
2855 Description: update patchless client
2856 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
2859 Frequency : only with 2.4 kernel
2861 Description: random memory corruption
2862 Details : size of struct ll_inode_info is to big for union inode.u and this
2863 can be cause of random memory corruption.
2868 Description: Memory leak in recovery
2869 Details : Lov_mds_md was not free in an error handler in mds_create_object.
2870 It should also check obd_fail before fsfilt_start, otherwise if
2871 fsfilt_start return -EROFS,(failover mds during mds recovery).
2872 then the req will return with repmsg->transno = 0 and rc = EROFS.
2873 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
2874 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
2875 be freed no matter whether fsfilt_commit success or not.
2878 Frequency : only with huge count clients
2880 Description: Prevents from taking the superblock lock in llap_from_page for
2882 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
2883 for llap_from_page call in ll_removepage prevents from taking the
2884 superblock lock for a soon died page.
2889 Description: Not check open intent error before release open handle
2890 Details : in some rare cases, the open intent error is not checked before
2891 release open handle, which may cause
2892 ASSERTION(open_req->rq_transno != 0), because it tries to release
2893 the failed open handle.
2898 Description: Set cat log bitmap only after create log success.
2899 Details : in some rare cases, the cat log bitmap is set too early. and it
2900 should be set only after create log success.
2904 Description: Accessing a block bevice can re-enable I/O when Lustre is
2905 tearing down a device.
2906 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
2911 Description: service threads may hog cpus when there are a lot of requests
2912 Details : Insert cond_resched to give other threads a chance to use some CPU
2917 Description: the cat log was not initialized in recovery
2918 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
2919 unlink log on mds will not be initialized until mds post
2920 recovery. And also in mds post recovery, the unlink log will
2921 initialization will be done asynchronausly, so there will be race
2922 between add unlink log and unlink log initialization.
2926 Description: brw_stats were being printed incorrectly
2927 Details : brw_stats were being printed as log2 but all of them were not
2928 recorded as log2. Also remove some code duplication arising from
2929 filter_tally_{read,write}.
2933 Frequency : rare, only in recovery.
2934 Description: ASSERTION(req->rq_type != LI_POISON) failed
2935 Details : imp_lock should be held while iterating over imp_sending_list for
2936 prevent destroy request after get timeout in ptlrpc_queue_wait.
2940 Description: replay-single.sh test 52 fails
2941 Details : A lock's skiplist need to be cleanup when it being unlinked
2942 from its resource list.
2946 Description: Short directio read returns full requested size rather than
2948 Details : Direct I/O operations should return actual amount of bytes
2949 transferred rather than requested size.
2951 Severity : enhancement
2953 Description: metadata RPC reduction (e.g. for rm performance)
2954 Details : decrease the amount of synchronous RPC between clients and servers
2955 by canceling conflicing lock before the operation on the client
2956 and packing thier handles into the main operation RPC to server.
2958 Severity : enhancement
2960 Description: Async OSC create to avoid the blocking unnecessarily.
2961 Details : If a OST has no remain object, system will block on the creating
2962 when need to create a new object on this OST. Now, ways use
2963 pre-created objects when available, instead of blocking on an
2964 empty osc while others are not empty. If we must block, we block
2965 for the shortest possible period of time.
2969 Description: improve handling recoverable errors
2970 Details : if request processig with error which can be recoverable on server
2971 request should be resend, otherwise page released from cache and
2974 Severity : enhancement
2976 Description: refine locking for avoid write wrong info into lov_objid file
2977 Details : fix possible races with add new target and write/update data in
2980 --------------------------------------------------------------------------------
2982 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
2987 Frequency : on some architectures
2989 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
2990 Details : Uninitialized flags sometimes cause configuration commands to
2994 Frequency : patchless clients only
2996 Description: missing __iget() symbol export
2997 Details : The __iget() symbol export is missing. To avoid the need for
2998 this on patchless clients the deathrow inode reaper is turned
2999 off, and we depend on the VM to clean up old inodes. This
3000 dependency was during via the fix for bug 12181.
3002 --------------------------------------------------------------------------------
3004 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
3006 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
3007 INTEROPERATE with older versions automatically. In many cases a
3008 special upgrade step is needed. Please read the
3009 user documentation before upgrading any part of a live system.
3010 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
3011 version will not interoperate with 1.6 betas before beta5 (1.5.95).
3012 * WARNING: Lustre configuration and startup changes are required with
3013 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
3017 Severity : enhancement
3019 Description: MountConf
3020 Details : Lustre configuration is now managed via mkfs and mount
3021 commands instead of lmc and lconf. New obd types (MGS, MGC)
3022 are added for dynamic configuration management. See
3023 https://mail.clusterfs.com/wikis/lustre/MountConf for
3026 Severity : enhancement
3028 Description: dynamic OST addition
3029 Details : OSTs can now be added to a live filesystem
3031 Severity : enhancement
3033 Description: startup order invariance
3034 Details : MDTs and OSTs can be started in any order. Clients only
3035 require the MDT to complete startup.
3037 Severity : enhancement
3039 Description: parallel, asynchronous orphan cleanup
3040 Details : orphan cleanup is now performed in separate threads for each
3041 OST, allowing parallel non-blocking operation.
3043 Severity : enhancement
3045 Description: optimized stripe assignment
3046 Details : stripe assignments are now made based on ost space available,
3047 ost previous usage, and OSS previous usage, in order to try
3048 to optimize storage space and networking resources.
3050 Severity : enhancement
3052 Description: Permanently set tunables
3053 Details : All writable /proc/fs/lustre tunables can now be permanently
3054 set on a per-server basis, at mkfs time or on a live system.
3056 Severity : enhancement
3058 Description: Lustre message v2
3059 Details : Add lustre message format v2.
3061 Severity : enhancement
3063 Description: client OST exclusion list
3064 Details : Clients can be started with a list of OSTs that should be
3065 declared "inactive" for known non-responsive OSTs.
3069 Description: SPEC SFS validation failure on NFS v2 over lustre.
3070 Details : Changes the blocksize for regular files to be 2x RPC size,
3071 and not depend on stripe size.
3073 Severity : enhancement
3075 Description: Multiple MD RPCs in flight.
3076 Details : Further unserialise some read-only MDS RPCs - learn about intents.
3077 To avoid overly-overloading MDS, introduce a limit on number of
3078 MDS RPCs in flight for a single client and add /proc controls
3079 to adjust this limit.
3081 Severity : enhancement
3083 Description: client read/write statistics
3084 Details : Add client read/write call usage stats for performance
3085 analysis of user processes.
3086 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
3087 file access. extents_stats shows chunk size distribution.
3088 extents_stats_per_process show chunk size distribution per
3091 Severity : enhancement
3093 Description: mds statistics
3094 Details : Add detailed mds operations statistics in
3095 /proc/fs/lustre/mds/*/stats.
3099 Description: Failure of copying files with lustre special EAs.
3100 Details : Client side always return success for setxattr call for lustre
3101 special xattr (currently only "trusted.lov").
3106 Description: Refcount LNET uuids
3107 Details : The global LNET uuid list grew linearly with every startup;
3108 refcount repeated list entries instead of always adding to
3111 Severity : enhancement
3113 Description: Dynamic service threads
3114 Details : Within a small range, start extra service threads
3115 automatically when the request queue builds up.
3118 Frequency : mixed-endian client/server environments
3120 Description: mixed-endian crashes
3121 Details : The new msg_v2 system had some failures in mixed-endian
3124 Severity : enhancement
3126 Description: Easy OST removal
3127 Details : OSTs can be permanently deactivated with e.g. 'lctl
3128 conf_param lustre-OST0001.osc.active=0'
3130 Severity : enhancement
3132 Description: MGS proc entries
3133 Details : Added basic proc entries for the MGS showing what filesystems
3136 Severity : enhancement
3138 Description: provide MGS failover
3139 Details : Added config lock reacquisition after MGS server failover.
3141 Severity : enhancement
3143 Description: add Linux 2.4 support
3144 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
3148 Description: a large application tries to do I/O to the same resource and dies
3149 in the middle of it.
3150 Details : Check the req->rq_arrival time after the call to
3151 ost_brw_lock_get(), but before we do anything about
3152 processing it & sending the BULK transfer request. This
3153 should help move old stale pending locks off the queue as
3154 quickly as obd_timeout.
3157 Frequency : when an incorrect nid is specified during startup
3159 Description: ptlrpc connect to non-existant node causes kernel crash
3160 Details : LNET can't be re-entered from an event callback, which
3161 happened when we expire a message after the export has been
3162 cleaned up. Instead, hand the zombie cleanup off to another
3165 Severity : enhancement
3167 Description: plain/inodebits lock performance improvement
3168 Details : Grouping plain/inodebits in granted list by their request modes
3169 and bits policy, thus improving the performance of search through
3173 Frequency : only if OST filesystem is corrupted
3175 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3176 Details : for a short time RPCs with bulk IO are in the replay list,
3177 but replay of bulk IOs is unimplemented. If the OST filesystem
3178 is corrupted due to disk cache incoherency and then replay is
3179 started it is possible to trip an assertion. Avoid putting
3180 committed RPCs into the replay list at all to avoid this issue.
3183 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3185 Description: System hang on startup
3186 Details : This bug allowed the liblustre (e.g. catamount) client to
3187 return to the app before handling all startup RPCs. This
3188 could leave the node unresponsive to lustre network traffic
3189 and manifested as a server ptllnd timeout.
3191 Severity : enhancement
3193 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3194 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
3195 boolean to control whether to print peer debug info when a
3196 client's RPC times out.
3199 Frequency : only for kernels with patches from Lustre below 1.4.3
3201 Description: Remove old rdonly API
3202 Details : Remove old rdonly API which unused from at least lustre 1.4.3
3205 Frequency : only for devices with external journals
3207 Description: Set external device read-only also
3208 Details : During a commanded failover stop, we set the disk device
3209 read-only while the server shuts down. We now also set any
3210 external journal device read-only at the same time.
3213 Frequency : when upgrading from 1.4 while trying to change parameters
3215 Description: The wrong (new) MDC name was used when setting parameters for
3216 upgraded MDT's. Also allows changing of OSC (and MDC)
3217 parameters if --writeconf is specified at tunefs upgrade time.
3220 Frequency : when setting specific ost indicies
3222 Description: QOS code breaks on skipped indicies
3223 Details : Add checks for missing OST indicies in the QOS code, so OSTs
3224 created with --index need not be sequential.
3226 Severity : enhancement
3228 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3229 Details : The uninit_groups feature works in conjunction with the kernel
3230 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3231 up the pass1 processing of e2fsck. This is a read-only feature
3232 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3233 mount filesystems that have had this feature enabled.
3235 Severity : enhancement
3237 Description: Improve multi-block allocation algorithm to avoid fragmentation
3238 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3239 allocation locality and avoid filesystem fragmentation.
3241 ------------------------------------------------------------------------------
3243 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
3245 * Support for kernels:
3246 2.4.21-47.0.1.EL (RHEL 3)
3247 2.6.5-7.283 (SLES 9)
3248 2.6.9-42.0.10.EL (RHEL 4)
3249 2.6.12.6 vanilla (kernel.org)
3250 2.6.16.27-0.9 (SLES 10)
3251 * Recommended e2fsprogs version: 1.39.cfs5
3253 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3257 Frequency : occasional, depends on client load and configuration
3258 Bugzilla : 12181, 12203
3259 Description: data loss for recently-modified files
3261 Details : In some cases it is possible that recently written or created
3262 files may not be written to disk in a timely manner (this should
3263 normally be within 30s unless client IO load is very high).
3264 The problem appears as zero-length files or files that are a
3265 multiple of 1MB in size after a client crash or client eviction
3266 that are missing data at the end of the file.
3268 This problem is more likely to be hit on clients where files are
3269 repeatedly created and unlinked in the same directory, clients
3270 have a large amount of RAM, have many CPUs, the filesystem has
3271 many OSTs, the clients are rebooted frequently, and/or the files
3272 are not accessed by other nodes after being written.
3274 The presence of the problem can be detected by looking at
3275 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
3276 smaller than the second (nr_unused) then dirty files will not
3277 be flushed automatically to disk. "sync; sleep 10" should be
3278 run several times on the node before unmounting it to update
3279 Lustre (this is also safe to run on nodes without this problem).
3281 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
3282 that can cause this same problem, so customers using that kernel
3283 also need to update the kernel in addition to Lustre. In order
3284 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
3286 It is normal that files written just before a client crash (less
3287 than 30s) may not yet have been flushed to disk, even for local
3291 Frequency : frequent on thin XT3 nodes
3293 Description: UUID collision on thin XT3 Linux nodes
3294 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
3295 randomly, since we relied on an insufficiently-seeded PRNG.
3300 Description: OSS hangs after "All ost request buffers busy"
3301 Details : A deadlock between quota and journal operations caused OSS
3302 hangs after printing "All ost request buffers busy."
3305 Frequency : always on liblustre builds
3307 Description: Cleanup compiler warnings on liblustre
3310 Frequency : always on liblustre builds on XT3
3312 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
3316 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
3317 > 32000 subdirectories
3318 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
3319 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
3320 subdirectory count crosses 32000. This will aid e2fsck to
3321 correctly handle more than 32000 subdirectories.
3324 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3326 Description: System hang on startup
3327 Details : This bug allowed the liblustre (e.g. catamount) client to
3328 return to the app before handling all startup RPCs. This
3329 could leave the node unresponsive to lustre network traffic
3330 and manifested as a server ptllnd timeout.
3332 Severity : enhancement
3334 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3335 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
3336 boolean to control whether to print peer debug info when a
3337 client's RPC times out.
3342 Description: make O_SYNC working on 2.6 kernels
3343 Details : 2.6 kernels use different method for mark pages for write,
3344 so need add a code to lustre for O_SYNC work.
3349 Description: Failure to close file and release space on NFS
3350 Details : Put inode details into lock acquired in ll_intent_file_open.
3351 Use mdc_intent_lock in ll_intent_open to properly
3352 detect all kind of errors unhandled by mdc_enqueue.
3357 Description: proc file read during shutdown sometimes raced obd removal,
3359 Details : Add lock to prevent obd access after proc file removal.
3362 Frequency : Only for files larger than 4GB on 32-bit clients.
3364 Description: improperly doing page alignment of locks
3365 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
3366 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3371 Description: RPCs being resent when they shouldn't be
3372 Details : Some RPCs that should not be resent are being resent. This
3373 can cause inconsistencies in the RPC state machine. Do not
3374 resend such requests.
3377 Frequency : rare, only with NFS export
3379 Description: Crash on NFS re-export node
3380 Details : under very unusual load conditions an assertion is hit in
3381 ll_intent_file_open()
3384 Frequency : only if OST filesystem is corrupted
3386 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3387 Details : for a short time RPCs with bulk IO are in the replay list,
3388 but replay of bulk IOs is unimplemented. If the OST filesystem
3389 is corrupted due to disk cache incoherency and then replay is
3390 started it is possible to trip an assertion. Avoid putting
3391 committed RPCs into the replay list at all to avoid this issue.
3396 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3397 Details : Large single O_DIRECT read and write calls can fail to allocate
3398 a sufficiently large buffer to process the request. In case of
3399 allocation failure the allocation is retried with a smaller
3400 buffer and broken into smaller requests.
3402 Severity : enhancement
3404 Description: Add -o localflock option to simulate old noflock behaviour.
3405 Details : This will achieve local-only flock/fcntl locks coherentness.
3410 Description: versioning check is incomplete
3411 Details : Checking the version difference of client vs. server, report
3412 error if the gap is too big.
3417 Description: add support PG_writeback bit
3418 Details : add support for PG_writeback bit for Lustre, for more carefull
3419 work with page cache in 2.6 kernel. This also fix some deadlocks
3420 and remove hack for work O_SYNC with 2.6 kernel.
3422 Severity : enhancement
3424 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3425 Details : The uninit_groups feature works in conjunction with the kernel
3426 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3427 up the pass1 processing of e2fsck. This is a read-only feature
3428 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3429 mount filesystems that have had this feature enabled.
3431 Severity : enhancement
3433 Description: Improve multi-block allocation algorithm to avoid fragmentation
3434 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3435 allocation locality and avoid filesystem fragmentation.
3437 ------------------------------------------------------------------------------
3439 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3441 * Support for kernels:
3442 2.6.9-42.0.3.EL (RHEL 4)
3443 2.6.5-7.276 (SLES 9)
3444 2.4.21-47.0.1.EL (RHEL 3)
3445 2.6.12.6 vanilla (kernel.org)
3446 2.6.16.21-0.8 (SLES10)
3447 * Recommended e2fsprogs version: 1.39.cfs2-0
3449 * The backwards-compatible /proc/sys/portals symlink has been removed
3450 in this release. Before upgrading, please ensure that you change
3451 any configuration scripts or /etc/sysctl.conf files that access
3452 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3453 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3454 in advance of the upgrade on any system running Lustre 1.4.6 or
3455 newer, since /proc/sys/lnet was added in that version.
3456 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3460 Frequency : only when quota is used
3462 Description: avoid scanning export list for quota master
3463 Details : Change the algorithms to avoid scanning export list in order
3464 to improve the efficiency.
3467 Frequency : MDS failover only, very rarely
3469 Description: "went back in time" messages on mds failover
3470 Details : The greatest transno may be lost when the current operation
3471 finishes with an error (transno==0) and the client's last_rcvd
3472 record is over-written. Save the greatest transno in the
3473 mds_last_transno for this case.
3476 Frequency : always for specific kernels and striping counts
3478 Description: client may get "Matching packet too big" without ACL support
3479 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
3480 when trying to access files in certain configurations. The
3481 clients should in fact be denied when mounting because they do
3482 not understand ACLs.
3485 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
3487 Description: many clients connecting with IO in progress causes connect timeouts
3488 Details : Avoid synchronous journal commits to avoid delays caused by many
3489 clients connecting/disconnecting when bulk IO is in progress.
3490 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
3491 OST_IO_PORTAL to avoid delays behind potentially many pending
3495 Frequency : occasionally with multiple writers to a single file
3497 Description: shared writes to file may result in wrong size reported by stat()
3498 Details : Allow growing of kms when extent lock is cancelled
3501 Frequency : always with random mmap IO to multi-striped file
3503 Description: mmap write might be lost if we are writing to a 'hole' in stripe
3504 Details : Only if the hole is at the end of OST object so that kms is too
3505 small. Fix is to increase kms accordingly in ll_nopage.
3508 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
3510 Description: writes to a missing object would leak memory on the OST
3511 Details : If there is an inconsistency between the MDS and OST filesystems,
3512 such that the MDS references an object that doesn't exist, writes
3513 to that object will leak memory due to incorrect cleanup in the
3514 error handling path, eventually running out of memory on the OST.
3519 Description: Creating too long symlink causes lustre errors
3520 Details : Check symlink and name lengths before sending requests to MDS.
3523 Frequency : only if flock is enabled (not on by default)
3525 Description: posix locks not released on fd closure on 2.6.9+
3526 Details : We failed to add posix locks to list of inode locks on 2.6.9+
3527 kernels, this caused such locks not to be released on fd close and
3528 then assertions on fs unmount about still used locks.
3531 Frequency : MDS failover only, very rarely
3533 Description: clients may get ASSERTION(granted_lock != NULL)
3534 Details : When request was taking a long time, and a client was resending
3535 a getattr by name lock request. The were multiple lock requests
3536 with the same client lock handle and
3537 mds_getattr_name->fixup_handle_for_resent_request found one of the
3538 lock handles but later failed with ASSERTION(granted_lock != NULL).
3543 Description: handle->h_buffer_credits > 0, assertion failure
3544 Details : h_buffer_credits is zero after truncate, causing assertion
3545 failure. This patch extends the transaction or creates a new
3549 Frequency : NFS re-export or patchless client
3550 Bugzilla : 11179, 10796
3551 Description: Crash on NFS re-export node (__d_move)
3552 Details : We do not want to hash the dentry if we don't have a lock.
3553 But if this dentry is later used in d_move, we'd hit uninitialised
3554 list head d_hash, so we just do this to init d_hash field but
3555 leave dentry unhashed.
3558 Frequency : NFS re-export or patchless client
3560 Description: NFS exports has problem with symbolic link
3561 Details : lustre client didn't properly install dentry when re-exported
3562 to NFS or running patchless client.
3565 Frequency : NFS re-export or patchless client
3567 Description: Various nfs/patchless fixes.
3568 Details : fixes reuse disconected alias for lookup process - this fixes
3569 warning "find_exported_dentry: npd != pd",
3570 fix permission error with open files at nfs.
3571 fix apply umask when do revalidate.
3574 Frequency : occasional
3576 Description: Crash on NFS re-export node
3577 Details : calling clear_page() on the wrong pointer triggered oops in
3578 generic_mapping_read().
3581 Frequency : rarely, using O_DIRECT IO
3583 Description: unaligned directio crashes client with LASSERT
3584 Details : check for unaligned buffers before trying any requests.
3587 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
3589 Description: stale data returned from RAID cache
3590 Details : If only a small amount of IO is done to the RAID device before
3591 reading it again it is possible to get stale data from the RAID
3592 cache instead of reading it from disk.
3595 Frequency : always for sles10 kernel
3597 Description: sles10 support
3598 Details : ll_follow_link: compile fixes and using of nd_set_link
3599 under newer kernels.
3602 Frequency : depends on arch, kernel and compiler version, always on sles10
3605 Description: recursive or deep enough symlinks cause stack overflow
3606 Details : getting rid of large stack-allocated variable in
3610 Frequency : depends on hardware
3612 Description: lustre write performance loss in the SLES10 kernel
3613 Details : the performance loss is caused by using of write barriers in the
3614 ext3 code. The SLES10 kernel turns barrier support on by
3615 default. The fix is to undo that change for ldiskfs.
3617 ------------------------------------------------------------------------------
3619 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
3621 * Support for kernels:
3622 2.6.9-42.0.3EL (RHEL 4)
3623 2.6.5-7.276 (SLES 9)
3624 2.4.21-47.0.1.EL (RHEL 3)
3625 2.6.12.6 vanilla (kernel.org)
3629 Frequency : quota enabled and large files being deleted
3631 Description: releasing more than 4GB of quota at once hangs OST
3632 Details : If a user deletes more than 4GB of files on a single OST it
3633 will cause the OST to spin in an infinite loop. Release
3634 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3639 Description: statfs data retrieved from /proc may be stale or zero
3640 Details : When reading per-device statfs data from /proc, in the
3641 {kbytes,files}_{total,free,avail} files, it may appear
3642 as zero or be out of date.
3645 Frequency : systems with MD RAID1 external journal devices
3647 Description: lconf's call to blkid is confused by RAID1 journal devices
3648 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3649 of returning all block devices that match the journal UUID.
3652 Frequency : always, for aggregate stripe size over 4GB
3654 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
3655 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
3656 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
3659 Frequency : always if "lfs find" used on a local file/directory
3661 Description: "lfs find" segfaults if used on a local file/directory
3662 Details : The case where a directory component was not specified wasn't
3663 handled correctly. Handle this properly.
3666 Frequency : always on ppc64
3668 Description: the write to an ext3 filesystem mounted with mballoc got stuck
3669 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3670 perform endianness conversion.
3673 Frequency : rarely (truncate to non-zero file size after write under load)
3674 Bugzilla : 10730, 10687
3675 Description: Files padded with zeros to next 4K multiple
3676 Details : With filesystems mounted using the "extents" option (2.6 kernels)
3677 it is possible that files that are truncated to a non-zero size
3678 immediately after being written are filled with zero bytes beyond
3679 the truncated size. No file data is lost.
3681 Severity : enhancement
3683 Description: Allow recovery/failover for liblustre clients.
3684 Details : liblustre clients were unaware of failover configurations until
3687 Severity : enhancement
3689 Description: user file locks should fail when not mounting with flock option
3690 Details : Set up an error-returning stub in ll_file_operations.lock field
3691 to prevent incorrect behaviour when client is mounted without
3692 flock option. Also, set up properly f_op->flock field for
3696 Frequency : always on ia64
3698 Description: "lfs df" loops on printing out MDS statfs information
3699 Details : The obd_ioctl_data was not initialized and in some systems
3700 this caused a failure during the ioctl that did not return
3701 an error. Initialize the struct and return an error on failure.
3704 Frequency : SLES 9 only
3706 Description: Error of copying files with lustre special EAs as root
3707 Details : Client side always return success for setxattr call for lustre
3708 special xattr (currently only "trusted.lov").
3711 Frequency : rarely on clusters with both ia64+i386 clients
3713 Description: ia64+i686 clients doing shared IO on the same file may LBUG
3714 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
3715 clients are doing concurrent writes to the same file it is
3716 possible that the ia64 clients may LASSERT because the OST
3717 extent locks are not PAGE_SIZE aligned. Ensure that grown
3718 locks are always aligned on the request boundary.
3721 Frequency : specific use, occasional
3723 Description: Overwriting in use executable truncates on-disk binary image
3724 Details : If one node attempts to overwrite an executable in use by
3725 another node, we now correctly return ETXTBSY instead of
3726 truncating the file.
3728 Severity : enhancement
3730 Description: Async OSC create to avoid the blocking unnecessarily.
3731 Details : If a OST has no remain object, system will block on the creating
3732 when need to create a new object on this OST. Now, ways use
3733 pre-created objects when available, instead of blocking on an
3734 empty osc while others are not empty. If we must block, we block
3735 for the shortest possible period of time.
3740 Description: chmod on Lustre root is propagated to other clients
3741 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
3742 invalid by the follow_mount time.
3745 Frequency : liblustre clients only
3747 Description: Race in 'instant cancel' lock handling could lead to such locks
3748 never to be granted in case of SMP MDS
3749 Details : Do not destroy not yet granted but cbpending locks in
3753 Frequency : replay/resend of open
3755 Description: non null lock assetion failure in mds_intent_policy
3756 Details : Trying to replay/resend lockless open requests resulted in
3757 mds_open() returning 0 with no lock. Now it sets a flag if
3758 a lock is going to be returned.
3760 Severity : enhancement
3762 Description: Checksum enhancements
3763 Details : New checksum enhancements allow for resending RPCs that failed
3766 Severity : enhancement
3768 Description: Tunables on number of dirty pages in cacche
3769 Details : Allow to set limit on number of dirty pages cached.
3774 Description: client crash on unmount - lock still has references
3775 Details : In some error handling cases it was possible to leak a lock
3776 reference on a client while accessing a file. This was not
3777 harmful to the client during operation, but would cause the
3778 client to crash when the filesystem is unmounted.
3781 Frequency : specific case, rare
3783 Description: ETXTBSY on mds though file not in use by client
3784 Details : ETXTBSY is no longer incorrectly returned when attempting to
3785 chmod or chown a directory that the user previously tried to
3786 execute or a currently-executing binary.
3789 Frequency : extremely rare except on liblustre-based clients
3791 Description: Lustre space not freed when files are deleted
3792 Details : Clean up open-unlinked files after client eviction. Previously
3793 the unlink was skipped and the files remained as orphans.
3798 Description: OST failure "would be an LBUG" in waiting_locks_callback()
3799 Details : In some cases it was possible to send a blocking callback to
3800 a client doing a glimpse, even though that client didn't get
3801 a lock granted. When the glimpse lock is cancelled on the OST
3802 the freed lock is left on the waiting list and corrupted the list.
3805 Frequency : all core dumps
3807 Description: Broke core dumps to lustre
3808 Details : Negative dentry may be unhashed if parent does not have UPDATE
3809 lock, but some callers, e.g. do_coredump, expect dentry to be
3810 hashed after successful create, hash it in ll_create_it.
3812 ------------------------------------------------------------------------------
3814 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
3816 * Support for kernels:
3817 2.6.9-42.0.2.EL (RHEL 4)
3818 2.6.5-7.276 (SLES 9)
3819 2.4.21-40.EL (RHEL 3)
3820 2.6.12.6 vanilla (kernel.org)
3824 Frequency : always on RHEL 3
3826 Description: Number of open files grows over time
3827 Details : The number of open files grows over time, whether or not
3828 Lustre is started. This was due to a filp leak introduced
3829 by one of our kernel patches.
3831 ------------------------------------------------------------------------------
3833 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
3835 * Support for kernels:
3836 2.6.9-42.EL (RHEL 4)
3837 2.6.5-7.276 (SLES 9)
3838 2.4.21-40.EL (RHEL 3)
3839 2.6.12.6 vanilla (kernel.org)
3844 Bugzilla : 5719, 9635, 9792, 9684
3845 Description: OST (or MDS) trips assertions in (re)connection under heavy load
3846 Details : If a server is under heavy load and cannot reply to new
3847 connection requests before the client resends the (re)connect,
3848 the connection handling code can behave badly if two service
3849 threads are concurrently handing separate (re)connections from
3850 the same client. Add better locking to the connection handling
3851 code, and ensure that only a single connection will be processed
3852 for a given client UUID, even if the lock is dropped.
3854 Severity : enhancement
3856 Description: add TCP zero-copy support to kernel
3857 Details : Add support to the kernel TCP stack to allow zero-copy bulk
3858 sends if the hardware supports scatter-gather and checksumming.
3859 This allows socklnd to do client-write and server-read more
3860 efficiently and reduce CPU utilization from skbuf copying.
3863 Frequency : only if NFS exporting from client
3865 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
3866 Details : If setting attributes on a file created under NFS that had
3867 never been opened it would be possible to oops the client
3868 if the file had no objects.
3871 Frequency : always for liblustre
3873 Description: liblustre client does MDS+OSTs setattr RPC for each write
3874 Details : When doing a write from a liblustre client, the client
3875 incorrectly issued an RPC to the MDS and each OST the file was
3876 striped over in order to update the timestamps. When writing
3877 with small chunks and many clients this could overwhelm the MDS
3878 with RPCs. In all cases it would slow down the write because
3879 these RPCs are unnecessary.
3881 Severity : enhancement
3883 Description: allow number of MDS service threads to be changed at module load
3884 Details : It is now possible to change the number of MDS service threads
3885 running. Adding "options mds mds_num_threads={N}" to the MDS's
3886 /etc/modprobe.conf will set the number of threads for the next
3887 time Lustre is restarted (assuming the "mds" module is also
3888 reloaded at that time). The default number of threads will
3889 stay the same, 32 for most systems.
3894 Description: OST crash if filesystem is unformatted or corrupt
3895 Details : If an OST is started on a device that has never been formatted
3896 or if the filesystem is corrupt and cannot even mount then the
3897 error handling cleanup routines would dereference a NULL pointer.
3902 Description: NULL pointer deref in llap_from_page.
3903 Details : get_cache_page_nowait can return a page with NULL (or otherwise
3904 incorrect) mapping if the page was truncated/reclaimed while it was
3905 searched for. Check for this condition and skip such pages when
3906 doing readahead. Introduce extra check to llap_from_page() to
3907 verify page->mapping->host is non-NULL (so page is not anonymous).
3910 Frequency : Sometimes when using sys_sendfile
3912 Description: "page not covered by a lock" warnings from ll_readpage
3913 Details : sendfile called ll_readpage without right page locks present.
3914 Now we introduced ll_file_sendfile that does necessary locking
3915 around call to generic_file_sendfile() much like we do in
3919 Frequency : with certain MDS communication failures at client mount time
3921 Description: NULL pointer deref after failed client mount
3922 Details : a client connection request may delayed by the network layer
3923 and not be sent until after the PTLRPC layer has timed out the
3924 request. If the client fails the mount immediately it will try
3925 to clean up before the network times out the request. Add a
3926 reference from the request import to the obd device and delay
3927 the cleanup until the network drops the request.
3930 Frequency : occasionally during client (re)connect
3932 Description: assertion failure during client (re)connect
3933 Details : processing a client connection request may be delayed by the
3934 client or server longer than the client connect timeout. This
3935 causes the client to resend the connection request. If the
3936 original connection request is replied in this interval, the
3937 client may trip an assertion failure in ptlrpc_connect_interpret()
3938 which thought it would be the only running connect process.
3941 Frequency : only with obd_echo servers and clients that are rebooted
3943 Description: kernel BUG accessing uninitialized data structure
3944 Details : When running an obd_echo server it did not start the ping_evictor
3945 thread, and when a client was evicted an uninitialized data
3946 structure was accessed. Start the ping_evictor in the RPC
3947 service startup instead of the OBD startup.
3949 Severity : enhancement
3950 Bugzilla : 10193 (patchless)
3951 Description: Remove dependency on various unexported kernel interfaces.
3952 Details : No longer need reparent_to_init, exit_mm, exit_files,
3953 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
3956 Frequency : rare (only users of deprecated and unsupported LDAP config)
3958 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
3959 Details : LDAP apparently contains 'lustreName' attributes instead of
3960 'name'. A simple remapping of the name is sufficient.
3963 Frequency : rare (only with non-default dump_on_timeout debug enabled)
3965 Description: waiting_locks_callback trips kernel BUG if client is evicted
3966 Details : Running with the dump_on_timeout debug flag turned on makes
3967 it possible that the waiting_locks_callback() can try to dump
3968 the Lustre kernel debug logs from an interrupt handler. Defer
3969 this log dumping to the expired_lock_main() thread.
3971 Severity : enhancement
3973 Description: Support NFS exporting on 2.6 kernels.
3974 Details : Implement non-rawops metadata methods for NFS server to use without
3975 changing NFS server code.
3978 Frequency : very rare (synthetic metadata workload only)
3980 Description: two racing renames might cause an MDS thread to deadlock
3981 Details : Running the "racer" program may cause one MDS thread to rename
3982 a file from being the source of a rename to being the target of
3983 a rename at exactly the same time that another thread is doing
3984 so, and the second thread has already enqueued these locks after
3985 doing a lookup of the target and is trying to relock them in
3986 order. Ensure that we don't try to re-lock the same resource.
3989 Frequency : only very large systems with liblustre clients
3991 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
3992 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
3993 all OSTs in parallel. This allows the eviction of stale liblustre
3994 clients to proceed much faster than if they were done in series,
3995 and also offers similar improvements for other set_info RPCs.
4000 Description: excessive CPU usage during initial read phase on client
4001 Details : During the initial read phase on a client, it would agressively
4002 retry readahead on the file, consuming too much CPU and impacting
4003 performance (since 1.4.5.8). Improve the readahead algorithm
4004 to avoid this, and also improve some other common cases (read
4005 of small files in particular, where "small" is files smaller than
4006 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
4011 Description: MDS crash when receiving packet with unknown intent.
4012 Details : Do not LBUG in unknown intent case, just return -EFAULT
4014 Severity : enhancement
4015 Bugzilla : 9293, 9385
4016 Description: MDS RPCs are serialised on client. This is unnecessary for some.
4017 Details : Do not serialize getattr (non-intent version) and statfs.
4020 Frequency : occasional, when OST network is overloaded/intermittent
4022 Description: client evicted by OST after bulk IO timeout
4023 Details : If a client sends a bulk IO request (read or write) the OST
4024 may evict the client if it is unresposive to its data GET/PUT
4025 request. This is incorrect if the network is overloaded (takes
4026 too long to transfer the RPC data) or dropped the OST GET/PUT
4027 request. There is no need to evict the client at all, since
4028 the pinger and/or lock callbacks will handle this, and the
4029 client can restart the bulk request.
4032 Frequency : Always when mmapping file with no objects
4034 Description: client crashes when mmapping file with no objects
4035 Details : Check that we actually have objects in a file before doing any
4036 operations on objects in ll_vm_open, ll_vm_close and
4042 Description: Request leak when working with deleted CWD
4043 Details : Introduce advanced request refcount tracking for requests
4044 referenced from lustre intent.
4046 Severity : Enhancement
4048 Description: Cache open file handles on client.
4049 Details : MDS now will return special lock along with openhandle, if
4050 requested and client is allowed to hold openhandle, even if unused,
4051 until such a lock is revoked. Helps NFS a lot, since NFS is opening
4052 closing files for every read/write openration.
4054 Severity : Enhancement
4056 Description: Cache open negative dentries on client when possible.
4057 Details : Guard negative dentries with UPDATE lock on parent dir, drop
4058 negative dentries on lock revocation.
4063 Description: Remounting a client read-only wasn't possible with a zconf mount
4064 Details : It wasn't possible to remount a client read-only with llmount.
4066 Severity : enhancement
4067 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
4068 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
4069 MPICH is included to add Lustre-specific ADIO interfaces.
4070 This is based closely on the UFS ADIO layer and only differs
4071 in file creation, in order to allow the OST striping to be set.
4072 This is user-contributed code and not supported by CFS.
4077 Description: extended inode attributes (immutable, append-only) work improperly
4078 when 2.4 and 2.6 kernels are used on client/server or vice versa
4079 Details : Introduce kernel-independent values for these flags.
4081 Severity : enhancement
4084 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
4085 Details : Many of the /proc/ tunables can only be tuned at a megabyte
4086 granularity. Now, Fractional MB granularity is be supported,
4087 this is very useful for low memory system.
4089 Severity : enhancement
4091 Description: Getattr by fid
4092 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
4093 locks, avoids extra getattr rpc requests to MDS, allows '/' to
4094 have locks and avoids getattr rpc requests for it on every stat.
4097 Frequency : Always, for filesystems larger than 2TB
4099 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4100 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
4101 filesystems larger than 2TB to be mounted with mballoc enabled.
4104 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
4106 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4107 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
4108 filesystems larger than 2TB is mounted then the kernel will
4109 silently corrupt the start of the filesystem. CONFIG_LBD is
4110 enabled for all CFS-supported kernels, but the possibility of
4111 this happening with a modified kernel config exists.
4113 Severity : enhancement
4115 Description: add client O_DIRECT support for 2.6 kernels
4116 Details : It is now possible to do O_DIRECT reads and writes to files
4117 in the Lustre client mountpoint on 2.6 kernel clients.
4119 Severity : enhancement
4121 Description: parallel glimpse, setattr, statfs, punch, destroy requests
4122 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
4123 parallel, not waiting for response from every OST before sending
4124 a rpc to the next OST.
4129 Description: setattr vs write race when updating file timestamps
4130 Details : Client processes that update a file timestamp into the past
4131 right after writing to the file (e.g. tar) it is possible that
4132 the updated file modification time can be reset to the current
4133 time due to a race between processing the setattr and write RPC.
4135 Severity : enhancement
4137 Description: Bring 'lfs find' closer in line with regular Linux find.
4138 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
4139 -print0 options and obtains all the needed info through the lustre
4142 Severity : enhancement
4144 Description: support up to 1024 configured devices on one node
4145 Details : change obd_dev array from statically allocated to dynamically
4146 allocated structs as they are first used to reduce memory usage
4151 Description: Flush dirty partially truncated pages during truncate
4152 Details : Immediatelly flush partially truncated pages in filter_setattr,
4153 this way we completely avoid having any pages in page cache on OST
4154 and can retire ugly workarounds during writes to flush such pages.
4159 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
4160 Details : protect inode from truncation within vfs_unlink() context
4161 just take a reference before calling vfs_unlink() and release it
4162 when parent's i_sem is free.
4167 Description: last_id value checked outside lock on OST caused LASSERT failure
4168 Details : If there were multiple MDS->OST object precreate requests in
4169 flight, it was possible that the OST's last object id was checked
4170 outside a lock and incorrectly tripped an assertion. Move checks
4171 inside locks, and discard old precreate requests.
4174 Frequency : always, if extents are used on OSTs
4176 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
4177 Details : OSTs using the extents format would not zero the high 16 bits of
4178 the index physical block number. This is not a problem for any
4179 OST filesystems smaller than 16TB, and no kernels support ext3
4180 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
4181 new/modified files) and can be fixed for existing filesystems
4182 with e2fsprogs-1.39-cfs1.
4187 Description: import connection selection may be incorrect if timer wraps
4188 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
4189 import connections to be ignored if the 32-bit jiffies counter
4190 wraps. Use a 64-bit jiffies counter.
4193 Frequency : very large clusters immediately after boot
4195 Description: LNET request buffers exhausted under heavy short-term load
4196 Details : If a large number of client requests are generated on a service
4197 that has previously never seen so many requests it is possible
4198 that the request buffer growth cannot keep up with the spike in
4199 demand. Instead of dropping incoming requests, they are held in
4200 the LND until the RPC service can accept more requests.
4203 Frequency : Sometimes during replay
4205 Description: Assertion failure in ll_local_open after replay.
4206 Details : If replay happened on an open request reply before we were able
4207 to set replay handler, reply will become not swabbed tripping the
4208 assertion in ll_local_open. Now we set the handler right after
4209 recognising of open request
4212 Frequency : very rare
4214 Description: kernel reports "badness in vsnprintf"
4215 Details : Reading from the "recovery_status" /proc file in small chunks
4216 may cause a negative length in lprocfs_obd_rd_recovery_status()
4217 call to vsnprintf() (which is otherwise harmless). Exit early
4218 if there is no more space in the output buffer.
4220 Severity : enhancement
4222 Description: clear OBD RPC statistics by writing to them
4223 Details : It is now possible to clear the OBD RPC statistics by writing
4224 to the "stats" file.
4229 Description: Client mtime is not the same on different clients after utimes
4230 Details : In some cases, the client was using the utimes() syscall on
4231 a file cached on another node. The clients now validate the
4232 ctime from the MDS + OSTs to determine which one is right.
4237 Description: Inability to activate failout mode
4238 Details : lconf script incorrectly assumed that in pythong string's numeric
4239 value is used in comparisons.
4242 Frequency : always with multiple stripes per file
4244 Description: Inefficient object allocation for mutli-stripe files
4245 Details : When selecting which OSTs to stripe files over, for files with
4246 a stripe count that divides evenly into the number of OSTs,
4247 the MDS is always picking the same starting OST for each file.
4248 Return the OST selection heuristic to the original design.
4253 Description: mount failures may take full timeout to return an error
4254 Details : Under some heavy load conditions it is possible that a
4255 failed mount can wait for the full obd_timeout interval,
4256 possibly several minutes, before reporting an error.
4257 Instead return an error as soon as the status is known.
4259 Frequency : quota enabled and large files being deleted
4261 Description: releasing more than 4GB of quota at once hangs OST
4262 Details : If a user deletes more than 4GB of files on a single OST it
4263 will cause the OST to spin in an infinite loop. Release
4264 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4269 Description: statfs data retrieved from /proc may be stale or zero
4270 Details : When reading per-device statfs data from /proc, in the
4271 {kbytes,files}_{total,free,avail} files, it may appear
4272 as zero or be out of date.
4275 Frequency : systems with MD RAID1 external journal devices
4277 Description: lconf's call to blkid is confused by RAID1 journal devices
4278 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4279 of returning all block devices that match the journal UUID.
4282 Frequency : always, for aggregate stripe size over 4GB
4284 Description: assertion fails when trying to use 4GB stripe size
4285 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
4286 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
4289 Frequency : always on ppc64
4291 Description: the first write on an ext3 filesystem with mballoc got stuck
4292 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4293 perform endianness conversion.
4295 ------------------------------------------------------------------------------
4297 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4299 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
4300 INTEROPERATE with older versions automatically. Please read the
4301 user documentation before upgrading any part of a live system.
4302 * WARNING: Lustre networking configuration changes are required with
4303 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
4306 * Support for kernels:
4307 2.6.9-22.0.2.EL (RHEL 4)
4308 2.6.5-7.244 (SLES 9)
4309 2.6.12.6 vanilla (kernel.org)
4312 Severity : enhancement
4313 Bugzilla : 7981/8208
4314 Description: Introduced Lustre Networking (LNET)
4315 Details : LNET is new networking infrastructure for Lustre, it includes
4316 a reorganized network configuration mode (see the user
4317 documentation for full details) as well as support for routing
4318 between different network fabrics. Lustre Networking Devices
4319 (LNDs) for the supported network fabrics have also been
4320 created for this new infrastructure.
4322 Severity : enhancement
4323 Description: Introduced Access control lists
4324 Details : clients can set ACLs on files and directories in order to have
4325 more fine-grained permissions than the standard Unix UGO+RWX.
4326 The MDS must be started with the "-o acl" mount option.
4328 Severity : enhancement
4329 Description: Introduced filesystem quotas
4330 Details : Administrators may now establish per-user quotas on the
4333 Severity : enhancement
4335 Description: Configuration change for the XT3
4336 The PTLLND is now used to run Lustre over Portals on the XT3
4337 The configure option(s) --with-cray-portals are no longer used.
4338 Rather --with-portals=<path-to-portals-includes> is used to
4339 enable building on the XT3. In addition to enable XT3 specific
4340 features the option --enable-cray-xt3 must be used.
4345 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
4346 Details : A race between allocating a new llcd and re-getting the llcd_lock
4347 allowed another thread to grab newly-allocated llcd.
4349 Severity : enhancement
4351 Description: 2.6 OST async journal commit and locking fix to improve performance
4352 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
4353 2.6 kernels are now async as they already were in 2.4 kernels so
4354 that they can commit concurrently with the network bulk transfer.
4355 For block-allocated files the filter allocation semaphore is held
4356 to avoid filesystem fragmentation during allocation. BKL lock
4357 removed for 2.6 xattr operations where it is no longer needed.
4362 Description: lconf incorrectly determined whether two IP networks could talk
4363 Details : In some more complicated routing and multiple-network
4364 configurations, lconf will avoid trying to make a network
4365 connection to a disjoint part of the IP space. It was doing the
4366 math incorrectly for one set of cases.
4371 Description: Fix for potential infinite loop processing records in an llog.
4372 Details : If an llog record is corrupted/zeroed, it is possible to loop
4373 forever in llog_process(). Validate the llog record length
4374 and skip the remainder of the block on error.
4377 Frequency : occasional (liblustre only)
4379 Description: liblustre could not open files whose last component is a symlink
4380 Details : sysio_path_walk() would incorrectly pass the open intent to
4381 intermediate path components.
4384 Frequency : rare (liblustre only with non-standard tuning)
4385 Bugzilla : 7201 (7350)
4386 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4387 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4388 after it has been released, only do this for referenced locks
4390 Severity : enhancement
4392 Description: specifying an (invalid) directory default stripe_size of -1
4393 would reset the directory default striping
4394 Details : stripe_size -1 was used internally to signal directory stripe
4395 removal, now use "all default" to signal dir stripe removal
4396 as a directory striping of "all default" is not useful
4399 Frequency : common for large clusters running liblustre clients
4401 Description: doing an ls when liblustre clients are running is slow
4402 Details : sending a glimpse AST to a liblustre client waits for every AST
4403 to time out, as liblustre clients will not respond. Since they
4404 cannot cache data we refresh the OST lock LVB from disk instead.
4406 Severity : enhancement
4408 Description: doing an ls at the same time as file IO can be slow
4409 Details : enqueue and other "small" requests can be blocked behind many
4410 large IO requests. Create a new OST IO portal for non-IO
4411 requests so they can be processed faster.
4414 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4416 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4417 created file with mode 000
4418 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4419 allow an ftruncate() as a non-root user to a file with mode 000.
4420 Lustre now respects this flag to disable mode checks when
4421 truncating a file owned by the user
4424 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4427 Description: Revoking locks from clients that went dead or catatonic might take
4429 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4430 cancellation of such locks instant on servers without waiting for
4431 any reply from clients. Clients drops these locks when cancel
4432 notification from server is received without replying.
4435 Frequency : liblustre-only, when liblustre client dies or becomes busy
4437 Description: Doing ls on Linux clients can take a long time with active
4439 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4440 granting such locks to it in the first place if possible. Locks
4441 are taken by proxy on the OST during the read or write and
4442 dropped immediately afterward. Add connect flags handling, do
4443 not grant locks to liblustre clients for glimpse ASTs.
4445 Severity : enhancement
4447 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4448 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4449 sequential reads and behaves badly with applications doing random
4450 reads. Improve it by reading ahead at least read region, and
4451 avoiding excessive large RPC for small reads.
4453 Severity : enhancement
4455 Description: Creating more than 1000 files for a single job may cause a load
4456 imbalance on the OSTs if there are also a large number of OSTs.
4457 Details : qos_prep_create() uses an OST index reseed value that is an
4458 even multiple of the number of available OSTs so that if the
4459 reseed happens in the middle of the object allocation it will
4460 still utilize the OSTs as uniformly as possible.
4465 Description: OST or MDS may oops in ping_evictor_main()
4466 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
4467 but doesn't restart at beginning of obd_exports_timed list
4470 Severity : enhancement
4472 Description: improve by-nid export eviction on the MDS and OST
4473 Details : allow multiple exports with the same NID to be evicted at one
4474 time without re-searching the exports list.
4477 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
4479 Description: MDS may oops in groups_free()
4480 Details : in rare race conditions a newly allocated group_info struct is
4481 freed again, and this can be NULL. The 2.4 compatibility code
4482 for groups_free() checked for a NULL pointer, but 2.6 did not.
4485 Frequency : common for liblustre clients doing little filesystem IO
4486 Bugzilla : 9352, 7313
4487 Description: server may evict liblustre clients accessing contended locks
4488 Details : if a client is granted a lock or receives a completion AST
4489 with a blocking AST already set it would not reply to the AST
4490 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
4493 Frequency : lfs setstripe, only systems with more than 160 OSTs
4495 Description: unable to set striping with a starting offset beyond OST 160
4496 Details : llapi_create_file() incorrectly limited the starting stripe
4497 index to the maximum single-file stripe count.
4500 Frequency : LDAP users only
4502 Description: lconf did not handle in-kernel recovery with LDAP properly
4503 Details : lconf/LustreDB get_refs() is searching the wrong namespace
4505 Severity : enhancement
4507 Description: bind OST threads to NUMA nodes to improve performance
4508 Details : all OST threads are uniformly bound to CPUs on a single NUMA
4509 node and do their allocations there to localize memory access
4511 Severity : enhancement
4513 Description: llmount can determine client NID directly from Myrinet (GM)
4514 Details : the client NID code from gmnalnid was moved directly into
4515 llmount, removing the need to use this or specifying the
4516 client NID explicitly when mounting GM clients with zeroconf
4519 Frequency : if client is started with down MDS
4521 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
4522 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
4523 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
4524 ptlrpc_import_delayed_req() because we didn't check if the
4525 request was interrupted, and we also didn't break out of the
4526 event loop if there was no timeout
4531 Description: data loss during non-page-aligned writes to a single file from
4532 both multiple nodes and multiple threads on one node at same time
4533 Details : updates to KMS and lsm weren't protected by common lock. Resulting
4534 inconsistency led to false short-reads, that were cached and later
4535 used by ->prepare_write() to fill in partially written page,
4536 leading to data loss.
4539 Frequency : always, if lconf --abort_recovery used
4541 Description: lconf --abort_recovery fails with 'Operation not supported'
4542 Details : lconf was attempting to abort recovery on the MDT device and not
4545 Severity : enhancement
4547 Description: remove cleanup logs
4548 Details : replace lconf-generated cleanup logs with lustre internal
4549 cleanup routines. Eliminates the need for client-cleanup and
4552 Severity : enhancement
4554 Description: add support for EAs (user and system) on lustre filesystems
4555 Details : it is now possible to store extended attributes in the Lustre
4556 client filesystem, and with the user_xattr mount option it
4557 is possible to allow users to store EAs on their files also
4559 Severity : enhancement
4561 Description: Add possibility (config option) to show minimal available OST free
4563 Details : When compiled with --enable-mindf configure option, statfs(2)
4564 (and so, df) will return least minimal free space available from
4565 all OSTs as amount of free space on FS, instead of summary of
4566 free spaces of all OSTs.
4568 Severity : enhancement
4570 Description: do not expand extent locks acquired on OST-side
4571 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
4572 by server: they are not cached anyway.
4575 Frequency : when mmap is used/binaries executed from Lustre
4577 Description: Unmmap pages before throwing them away from read cache.
4578 Details : llap_shrink cache now attempts to unmap pages before discarding
4579 them (if unmapping failed - do not discard). SLES9 kernel has
4580 extra checks that trigger if this unmapping is not done first.
4585 Description: lconf didn't resolve symlinks before checking to see whether a
4586 given mountpoint was already in use
4589 Frequency : when migrating failover services
4590 Bugzilla : 6395, 9514
4591 Description: When migrating a subset of services from a node (e.g. failback
4592 from a failover service node) the remaining services would
4593 time out and evict clients.
4594 Details : lconf --force (implied by --failover) sets the global obd_timeout
4595 to 5 seconds in order to quickly disconnect, but this caused
4596 other RPCs to time out too quickly. Do not change the global
4597 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
4599 Severity : enhancement
4600 Frequency : if MDS is started with down OST
4601 Bugzilla : 9439,5706
4602 Description: Allow startup/shutdown of an MDS without depending on the
4603 availability of the OSTs.
4604 Details : Asynchronously call mds_lov_synchronize during MDS startup.
4605 Add appropriate locking and lov-osc refcounts for safe
4606 cleaning. Add osc abort_inflight calls in case the
4607 synchronize never started.
4610 Frequency : occasional (Cray XT3 only)
4612 Description: root not authorized to access files in CRAY_PORTALS environment
4613 Details : The client process capabilities were not honoured on the MDS in
4614 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
4615 been accessed by an authorized user then root was able to access
4616 the file on the local client also. The root user capabilities
4617 are now allowed on the MDS, as this environment has secure UID.
4620 Frequency : occasional
4622 Description: ldiskfs "too long searching" message happens too often
4623 Details : A debugging message (otherwise harmless) prints too often on
4624 the OST console. This has been reduced to only happen when
4625 there are fragmentation problems on the filesystem.
4630 Description: Division by zero in statfs when all OSCs are inactive
4631 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
4632 lov_statfs divides by value returned by lov_get_stripecnt().
4636 Bugzilla : 9489, 3273
4637 Description: First write from each client to each OST was only 4kB in size,
4638 to initialize client writeback cache, which caused sub-optimal
4639 RPCs and poor layout on disk for the first writen file.
4640 Details : Clients now request an initial cache grant at (re)connect time
4641 and so that they can start streaming writes to the cache right
4642 away and always do full-sized RPCs if there is enough data.
4643 If the OST is rebooted the client also re-establishes its grant
4644 so that client cached writes will be honoured under the grant.
4649 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
4650 Details : Now I/O RPCs go to different portal number and (presumably) fast
4651 lock requests (and glimses) and other RPCs get their own service
4652 threads pool that should be able to service those RPCs
4655 Severity : enhancement
4657 Description: Ability to exchange lustre version between client and servers and
4658 issue warnings at client side if client is too old. Also for
4659 liblustre clients there is ability to refuse connection of too old
4661 Details : New 'version' field is added to connect data structure that is
4662 filled with version info. That info is later checked by server and
4666 Frequency : rare, liblustre only.
4667 Bugzilla : 9296, 9581
4668 Description: Two simultaneous writes from liblustre at offset within same page
4669 might proceed at the same time overwriting eachother with stale
4671 Details : I/O lock withing llu_file_prwv was released too early, before data
4672 actually was hitting the wire. Extended lock-holding time until
4673 server acknowledges receiving data.
4676 Frequency : extremely rare. Never observed in practice.
4678 Description: avoid generating lustre_handle cookie of 0.
4679 Details : class_handle_hash() generates handle cookies by incrementing
4680 global counter, and can hit 0 occasionaly (this is unlikely, but
4681 not impossible, because initial value of cookie counter is
4682 selected randonly). Value of 0 is used as a sentinel meaning
4683 "unassigned handle" --- avoid it. Also coalesce two critical
4684 sections in this function into one.
4686 Severity : enhancement
4688 Description: allow liblustre clients to delegate truncate locking to OST
4689 Details : To avoid overhead of locking, liblustre client instructs OST to
4690 take extent lock in ost_punch() on client's behalf. New connection
4691 flag is added to handle backward compatibility.
4693 Severity : enhancement
4694 Bugzilla : 4928, 7341, 9758
4695 Description: allow number of OST service threads to be specified
4696 Details : a module parameter allows the number of OST service threads
4697 to be specified via "options ost ost_num_threads={N}" in the
4698 OSS's /etc/modules.conf or /etc/modprobe.conf.
4702 Bugzilla : 6146, 9635, 9895
4703 Description: servers crash with bad pointer in target_handle_connect()
4704 Details : In rare cases when a client is reconnecting it was possible that
4705 the connection request was the last reference for that export.
4706 We would temporarily drop the export reference and get a new
4707 one, but this may have been the last reference and the export
4708 was just destroyed. Get new reference before dropping old one.
4710 Severity : enhancement
4711 Frequency : if client is started with failover MDS
4713 Description: Allow multiple MDS hostnames in the mount command
4714 Details : Try to read the configuration from all specified MDS
4715 hostnames during a client mount in case the "primary"
4718 Severity : enhancement
4720 Description: Stop sending data to evicted clients as soon as possible.
4721 Details : Check if the client we are about to send or are sending data to
4722 was evicted already. (Check is done every second of waiting,
4723 for which l_wait_event interface was extended to allow checking
4724 of exit condition at specified intervals).
4727 Frequency : rare, normally only when NFS exporting is done from client
4729 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
4731 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
4734 Frequency : always, if srand() is called before liblustre initialization
4736 Description: Liblustre uses system PRNG disturbing its usage by user application
4737 Details : Introduce internal to lustre fast and high-quality PRNG for
4738 lustre usage and make liblustre and some other places in generic
4739 lustre code to use it.
4741 Severity : enhancement
4742 Bugzilla : 9477, 9557, 9870
4743 Description: Verify that the MDS configuration logs are updated when xml is
4744 Details : Check if the .xml configuration logs are newer than the config
4745 logs stored on the MDS and report an error if this is the case.
4746 Request --write-conf, or allow starting with --old_conf.
4748 Severity : enhancement
4750 Description: Handle symlinks in the path when checking if Lustre is mounted.
4751 Details : Resolve intermediate symlinks when checking if a client has
4752 mounted a filesystem to avoid duplicate client mounts.
4757 Description: lconf can hit an error exception but still return success.
4758 Details : The lconf command catches the Command error exception at the top
4759 level script context and will exit with the associated exit
4760 status, but doesn't ensure that this exit status is non-zero.
4765 Description: failure of ptlrpc thread startup can cause oops
4766 Details : Starting a ptlrpc service thread can fail if there are a large
4767 number of threads or the server memory is very fragmented.
4768 Handle this without oopsing.
4771 Frequency : always, only if liblustre and non-default acceptor port was used
4773 Description: liblustre cannot connect to servers with non-default acceptor port
4774 Details : tcpnal_set_default_params() was not called and was therefore
4775 ignoring the environment varaible TCPNAL_PORT, as well as other
4776 TCPNAL_ environment variables
4781 Description: two objects could be created on the same OST for a single file
4782 Details : If an OST is down, in some cases it was possible to create two
4783 objects on a single OST for a single file. No problems other
4784 than potential performance impact and spurious error messages.
4788 Bugzilla : 5681, 9562
4789 Description: Client may oops in ll_unhash_aliases
4790 Details : Client dcache may become inconsistent in race condition.
4791 In some cases "getcwd" can fail if the current directory is
4797 Description: Inode refcounting problems in NFS export code
4798 Details : link_raw functions used to call d_instantiate without obtaining
4799 extra inode reference first.
4803 Bugzilla : 9942, 9903
4804 Description: Referencing freed requests leading to crash, memleaks with NFS.
4805 Details : We used to require that call to ll_revalidate_it was always
4806 followed by ll_lookup_it. Also with revalidate_special() it is
4807 possible to call ll_revalidate_it() twice for the same dentry
4808 even if first occurence returned success. This fix changes semantic
4809 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
4810 reference on a request referred from the intent.
4811 ll_intent_release() then releases such a request.
4814 Frequency : rare, normally benchmark loads only
4816 Description: unlinked inodes were kept in memory on the client
4817 Details : If a client is repeatedly creating and unlinking files it
4818 can accumulate a lot of stale inodes in the inode slab cache.
4819 If there is no other client load running this can cause the
4820 client node to run out of memory. Instead flush old inodes
4821 from client cache that have the same inode number as a new inode.
4824 Frequency : SLES9 2.6.5 kernel and long filenames only
4825 Bugzilla : 9969, 10379
4826 Description: utime reports stale NFS file handle
4827 Details : SLES9 uses out-of-dentry names in some cases, which confused
4828 the lustre dentry revalidation. Change it to always use the
4832 Frequency : rare, unless heavy write-truncate concurrency is continuous
4833 Bugzilla : 4180, 6984, 7171, 9963, 9331
4834 Description: OST becomes very slow and/or deadlocked during object unlink
4835 Details : filter_destroy() was holding onto the parent directory lock
4836 while truncating+unlinking objects. For very large objects this
4837 may block other threads for a long time and slow overall OST
4838 responsiveness. It may also be possible to get a lock ordering
4839 deadlock in this case, or run out of journal credits because of
4840 the combined truncate+unlink. Solution is to do object truncate
4841 first in one transaction without parent lock, and then do the
4842 final unlink in a new transaction with the parent lock. This
4843 reduces the lock hold time dramatically.
4846 Frequency : rare, 2.4 kernels only
4848 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
4849 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
4850 before unmounting their filesystem in order to avoid deadlock.
4851 The kernel_locked() function in 2.4 kernels only checks whether
4852 the kernel lock is held, not whether it is this process that is
4853 holding it as 2.6 kernels do.
4858 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
4859 Details : The client ptlrpc code may be trying to reconnect to a down
4860 server before a previous connection attempt has timed out.
4861 Increase the reconnect interval to be longer than the connection
4862 timeout interval to avoid sending duplicate connections to
4866 Frequency : echo_client brw_test command
4868 Description: fix echo_client to work with OST preallocated code
4869 Details : OST preallocation code (5137) didn't take echo_client IO path
4870 into account: echo_client calls filter methods outside of any
4871 OST thread and, hence, there is no per-thread preallocated
4872 pages and buffers to use. Solution: hijack pga pages for IO. As
4873 a byproduct, this avoids unnecessary data copying.
4877 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
4878 Description: Client can oops in mdc_commit_close() after open replay
4879 Details : It was possible for the MDS to return an open request with no
4880 transaction number in mds_finish_transno() if the client was
4881 evicted, but without actually returning an error. Clients
4882 would later try to replay that open and may trip an assertion
4883 Simplify the client close codepath, and always return an error
4884 from the MDS in case the open is not successful.
4887 Frequency : rare, 2.6 OSTs only
4889 Description: OST may deadlock under high load on fragmented files
4890 Details : If there was a heavy load and highly-fragmented OST filesystems
4891 it was possible to have all the OST threads deadlock waiting on
4892 allocation of biovecs, because the biovecs were not released
4893 until the entire RPC IO was completed. Instead, release biovecs
4894 as soon as they are complete to ensure forward IO progress.
4896 Severity : enhancement
4898 Description: Support for specifying external journal device at mount
4899 Details : If an OST or MDS device is formatted with an external journal
4900 device, this device major/minor is stored in the ext3 superblock
4901 and may not be valid for failover. Allow detecting and
4902 specifying the external journal at mount time.
4907 Description: Mounting an MDS with pending unlinked files may cause oops
4908 Details : target_finish_recovery() calls mds_postrecov() which returned
4909 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
4910 considers this an error and immediately begins cleaning up the
4911 lov, just after starting the mds_lov process
4913 Severity : enhancement
4915 Description: Implement 'lfs df' to report actual free space on per-OST basis
4916 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
4917 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
4918 usage of objects; '-h' to report in human readable format.
4920 ------------------------------------------------------------------------------
4922 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
4929 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
4930 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
4931 at startup during orphan unlinking. Instead, skip buddy update
4932 before setup, as it will be regenerated after recovery is complete.
4937 Description: If an OST is inactive, its locks might reference stale inodes.
4938 Details : lov_change_cbdata() must iterate over all namespaces, even if
4939 they are inactive to clear inode references from the lock.
4941 Severity : enhancement
4942 Frequency : occasional, if non-standard max_dirty_mb used
4944 Description: Client will block write RPCs if not enough grant
4945 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
4946 then the client will block writes while waiting for another RPC
4947 to complete instead of consuming its dirty limit. With change
4948 we get improved performance when max_dirty_mb is small.
4950 Severity : enhancement
4951 Bugzilla : 3389, 6253
4952 Description: Add support for supplementary groups on the MDS.
4953 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
4954 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
4955 lookups for user supplementary groups into a cache.
4959 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
4960 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
4961 for liblustre O_EXCL is enforced by MDS.
4965 Description: readdir never returns NULL in liblustre.
4966 Details : Corrected llu_iop_getdirentries logic, to return offset of next
4967 dentry in struct dirent.
4971 Frequency : liblustre only, depends on application IO pattern
4972 Description: liblustre clients evicted if not contacting servers
4973 Details : Don't put liblustre clients into the ping_evictor list, so
4974 they will not be evicted by the pinger ever.
4976 Severity : enhancement
4978 Description: Add ability to evict clients by NID from MDS.
4979 Details : By echoing "nid:$NID" string into
4980 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
4981 $NID would be instantly evicted from this MDS and from all active
4982 OSTs connected to it.
4986 Description: Do not query file size twice, somewhat slowing stat(2) calls.
4987 Details : lookup_it_finish() used to query file size from OSTs that was not
4992 Description: service threads change working directory to that of init
4993 Details : Starting lustre service threads may pin the working directory
4994 of the parent thread, making that filesystem busy. Threads
4995 now change to the working directory of init to avoid this.
4999 Frequency : during shutdown only
5000 Description: shutdown with a failed MDS or OST can cause unmount to hang
5001 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
5006 Frequency : 2.6 only
5007 Description: chmod/chown may include an extra supplementary group
5008 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
5009 supplementary group and if none is specified this is used.
5012 Bugzilla : 5479 (6816)
5014 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
5015 Details : If lookup is in progress on a file that is unlinked we might try
5016 to revalidate the inode and fail in revalidate after lookup is
5017 complete and ll_file_open() enqueues the open again but
5018 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
5021 Frequency : always, if lconf --abort_recovery used
5023 Description: lconf --abort_recovery fails with 'Operation not supported'
5024 Details : lconf was attempting to abort recovery on the MDT device and not
5027 ------------------------------------------------------------------------------
5029 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
5034 Frequency : rare (only unsupported configurations with a node running as an
5036 Bugzilla : 6514, 5137
5037 Description: Mounting a Lustre file system on a node running as an OST could
5039 Details : OSTs now preallocates memory needed to write out data at
5040 startup, instead of when needed, to avoid having to
5041 allocate memory in possibly low memory situations.
5042 Specifically, if the file system is mounted on on OST,
5043 memory pressure could force it to try to write out data,
5044 which it needed to allocate memory to do. Due to the low
5045 memory, it would be unable to do so and the node would
5046 become unresponsive.
5048 Severity : enhancement
5050 Description: Addition of lconf --service command line option
5051 Details : lconf now accepts a '--service <arg>' option, which is
5052 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
5054 Severity : enhancement
5056 Description: Failover mode is now the default for OSTs.
5057 Details : By default, OSTs will now run in failover mode. To return to
5058 the old behaviour, add '--failout' to the lmc line for OSTs.
5060 Severity : enhancement
5062 Description: Health checks are now provided for MDS and OSTs
5063 Details : Additional detailed health check information on MSD and OSTs
5064 is now provided through the procfs health_check value.
5067 Frequency : occasional, depends on IO load
5069 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
5070 after numerous create/delete cycles
5071 Details : The ext3 inode allocation policy would not allocate new inodes
5072 very well on the OSTs because there are no new directories
5073 being created. Instead we look for groups with free space if
5074 the parent directories are nearly full.
5079 Description: Network or server problems during mount may cause partially
5080 mounted clients instead of returning an error.
5081 Details : The config llog parsing code may overwrite the error return
5082 code during mount error handling, returning success instead
5088 Description: MDS can fail to allocate large reply buffers
5089 Details : After long uptimes the MDS can fail to allocate large reply
5090 buffers (e.g. zconf client mount config records) due to memory
5091 fragmentation or consumption by the buffer cache. Preallocate
5092 some large reply buffers so that these replies can be sent even
5093 under memory pressure.
5097 Frequency : rare (liblustre)
5098 Description: fsx running with liblustre complained that using truncate() to
5099 extend the file doesn't work. This patch corrects that issue.
5100 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
5101 ATTR_SIZE and lsm use in llu_setattr_raw.
5105 Frequency : rare, only 2.6 kernels
5106 Description: Unusual file access patterns on the MDS may result in inode
5107 data being lost in very rare circumstances.
5108 Details : Bad interaction between the ea-in-inode patch and the "no-read"
5109 code in the 2.6 kernel caused the inode and/or EA data not to
5110 be read from disk, causing single-file corruption.
5114 Frequency : rare, only 2.6 filesystems using extents
5115 Description: Heavy concurrent write and delete load may cause data corruption.
5116 Details : It was possible under high-load situations to have an extent
5117 metadata block in the block device cache from a just-unlinked
5118 file overwrite a newly-allocated data block. We now unmap any
5119 metadata buffers that alias just-allocated data blocks.
5123 Frequency : filesystems with default stripe_count larger than 77
5124 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
5125 Details : lconf specifies an inode size of 4096 bytes when the default
5126 stripe_count is larger than 77. This conflicts with the default
5127 inode density of 1 per 4096 bytes. Allocate smaller inodes in
5128 this case to avoid pinning too much memory for large EAs.
5130 ------------------------------------------------------------------------------
5132 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
5137 Frequency : rare (extremely heavy IO load with hundreds of clients)
5139 Description: Client is evicted, gets IO error writing to file
5140 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
5141 caused clients to be evicted for AST timeouts. The fixes in
5142 bug 5192 mean we no longer need to have such short AST timeouts
5143 so ldlm_timeout has been increased.
5146 Frequency : occasional during --force or --failover shutdown under load
5147 Bugzilla : 5949, 4834
5148 Description: Server oops/LBUG if stopped with --force or --failover under load
5149 Details : a collection of import/export refcount and cleanup ordering
5150 issues fixed for safer force cleanup
5153 Frequency : only filesystems larger than 120 OSTs
5154 Bugzilla : 5990, 6223
5155 Description: lfs getstripe would oops on a very large filesystem
5156 Details : lov_getconfig used kfree on vmalloc'd memory
5159 Frequency : only filesystems exporting via NFS to Solaris 10 clients
5160 Bugzilla : 6242, 6243
5161 Description: reading from files that had been truncated to a non-zero size
5162 but never opened returned no data
5163 Details : ll_file_read() reads zeros from no-object files to EOF
5168 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
5169 Details : There's little harm in aborting MDS/OSS recovery and letting it
5170 try again, so I removed the LASSERT and return an error instead.
5172 Severity : enhancement
5174 Description: New debugging infrastructure for tracking down data corruption
5175 Details : The I/O checksum code was replaced to: (a) control it at runtime,
5176 (b) cover more of the client-side code path, and (c) try to narrow
5177 down where problems occurred
5181 Bugzilla : 3819, 4364, 4397, 6313
5182 Description: Racing close and eviction MDS could cause assertion in mds_close
5183 Details : It was possible to get multiple mfd references during close and
5184 client eviction, leading to one thread referencing a freed mfd.
5186 Severity: : enhancement
5187 Bugzilla : 3262, 6359
5188 Description: Attempts to reconnect to servers are now more aggressive.
5189 Details : This builds on the enhanced upcall-less recovery that was added
5190 in 1.4.2. When trying to reconnect to servers, clients will
5191 now try each server in the failover group every 10 seconds. By
5192 default, clients would previously try one server every 25 seconds.
5197 Description: After recovery, certain operations trigger a failed
5198 assertion on a client.
5199 Details : Failing over an mds, using lconf -d --failover, while a
5200 client was doing a readdir() call would cause the client to
5201 LBUG after recovery completed and the readdir() was resent.
5203 Severity : enhancement
5205 Description: Default groups are now added by lconf
5206 Details : You can now run lconf --group <servicename> without having to
5207 manually add groups with lmc.
5210 Frequency : occasional
5212 Description: Nodes with an elan id of 0 trigger a failed assertion
5215 Frequency : always when accessing e.g. tty/console device nodes
5217 Description: tty and some other devices nodes cannot be used on lustre
5218 Details : file's private_data field is used by device data and lustre
5219 values in there got lost. New field was added to struct file to
5220 store fs-specific private data.
5223 Frequency : when exporting Lustre via NFS
5225 Description: NFSD failed occasionally when looking up a path component
5226 Details : NFSD is looking up ".." which was broken in ext3 directories
5227 that had grown large enough to become hashed.
5230 Frequency : Clusters with multiple interfaces not on the same subnet
5232 Description: Nodes will repeatedly try to reconnect to an interface which it
5233 cannot reach and report an error to the log.
5234 Details : Extra peer list entries will be created by lconf with some peers
5235 unreachable. lconf now validates the peer before adding it.
5238 Frequency : Only if a default stripe is set on the filesystem root.
5240 Description: Setting a default stripe on the filesystem root prevented the
5241 filesystem from being remounted.
5242 Details : The client was sending extra request flags in the root getattr
5243 request and did not allocate a reply buffer for the dir EA.
5246 Frequency : occasional, higher if lots of files are accessed by one client
5247 Bugzilla : 6159, 6097
5248 Description: Client trips assertion regarding lsm mismatch/magic
5249 Details : While revalidating inodes the VFS looks up inodes with ifind()
5250 and in rare cases can find an inode that is being freed.
5251 The ll_test_inode() code will free the lsm during ifind()
5252 when it finds an existing inode and then the VFS later attaches
5253 this free lsm to a new inode.
5257 Bugzilla : 6422, 7030
5258 Description: MDS deadlock between mkdir and client eviction
5259 Details : Creating a new file via mkdir or mknod (starting a transaction
5260 and getting the ns lock) can deadlock with client eviction
5261 (gets ns lock and trying to finish a synchronous transaction).
5264 Frequency : occasional
5265 Description: While starting a server, the fsfilt_ext3 module could not be
5267 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
5268 kernels. Previously, lconf would still use the ext3 name
5269 when trying to load modules. Now, it will correctly use
5270 ext3 on 2.4 and ldiskfs on 2.6.
5272 Severity : enhancement
5273 Description: The default stripe count has been changed to 1
5274 Details : The interpretation of the default stripe count (0, to lfs
5275 or lmc) has been changed to mean striping across a single
5276 OST, rather than all available. For general usage we have
5277 found a stripe count of 1 or 2 works best.
5279 Severity : enhancement
5280 Description: Add support for compiling against Cray portals.
5281 Details : Conditional compiling for some areas that are different
5285 Frequency : occasional
5286 Bugzilla : 6409, 6834
5287 Description: Creating files with an explicit stripe count may lead to
5288 a failed assertion on the MDS
5289 Details : If some OSTs are full or unavailable, creating files may
5290 trigger a failed assertion on the MDS. Now, Lustre will
5291 try to use other servers or return an error to the
5295 Frequency : occasional
5297 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
5298 caused lock timeout during readahead (since 1.4.2).
5299 Details : Processes doing readahead might match a lock that hasn't been
5300 granted yet if there are overlapping and conflicting lock
5301 requests. The readahead process waits on ungranted lock
5302 (original lock is CBPENDING), while OST waits for that process
5303 to cancel CBPENDING read lock and eventually evicts client.
5305 Severity : enhancement
5307 Description: Initial enabling of flock support for clients
5308 Details : Implements fcntl advisory locking and file status functions.
5309 This feature is provided as an optional mount flag (default
5310 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
5311 locking are implemented yet, and those that are are not guaranteed
5312 to be completely correct in production environments.
5313 mount -t lustre -o [flock|noflock] ...
5316 Frequency : occasional
5318 Description: OSTs running 2.4 kernels but with extents enabled might trip an
5319 assertion in the ext3 JBD (journaling) layer.
5320 Details : The b_committed_data struct is protected by the big kernel lock
5321 in 2.4 kernels, serializing journal_commit_transaction() and
5322 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
5323 access to this struct. In 2.6 kernels there is finer grained
5324 locking to improve SMP performance of the JBD layer.
5328 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
5333 Description: merge and cleanup kernel patches.
5336 -----------------------------------------------------------------------------
5338 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
5340 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
5341 versions of Lustre. Please update all servers and clients to
5342 version 1.4.2 or later at the same time. You must also run
5343 "lconf --write-conf {config}.xml" on the MDS while it is stopped
5344 to update the configuration logs.
5346 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
5347 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
5348 - swab llog records of type '0' so we get proper header size/idx (5861)
5349 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
5350 - fix rename of one directory over another leaking an inode (5953)
5351 - avoid SetPageDirty on 2.6 (5981)
5352 - don't re-add just-being-destroyed locks to the waiting list (5653)
5353 - when creating new directories, inherit the parent's custom
5354 striping settings if present parent (3048)
5355 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
5356 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
5357 - don't hold client locks on temporary worklist from l_lru (5666)
5358 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
5359 - automatically evict dead clients (5921)
5360 - Update file size properly in create+truncate+fstat case (6196)
5361 - Do not unhash mountpoint dentries, do not allow removal of
5363 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
5364 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
5365 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
5366 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
5367 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
5368 - Correctly handle reads of files with no objects (6243)
5369 - lctl recover will also mark a device active if deactivate used (5933)
5371 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
5372 - allow --write-conf on an MDS with different nettype than client (5619)
5373 - don't write config llogs to MDS for mounts not from that MDS (5617)
5374 - lconf should create multiple TCP connections from a client (5201)
5375 - init scripts are now turned off by default; run chkconfig --on
5376 lustre and chkconfig --on lustrefs to use them
5377 - upcalls are no longer needed for clients to recover to failover
5379 - add --abort-recovery option to lconf to abort recovery on device
5381 - add support for an arbitrary number of OSTs (3026)
5382 - Quota support protocol changes.
5383 - forward compatibility changes to wire structs (6007)
5384 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5385 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5386 - improved "lustre status" script
5387 - initialize blocksize for non-regular files (6062)
5388 - added --disable-server and --disable-client configure options (5782)
5389 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5390 - Vanilla 2.4.29 support
5391 - increase maximum number of obd devices to 520 (6242)
5392 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5393 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5394 SLES 9 2.6.5 kernels
5395 - limit stripes per file to 160 (the maximum EA size) (6093)
5397 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5400 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5401 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5402 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5403 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5404 - small ext3 extents cleanups and fixes (5733)
5405 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5406 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5407 - avoid needless client->OST connect, fix handle mismatch (5317)
5408 - fix DLM error path that led to out-of-sync client, long delays (5779)
5409 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5410 - fix several locking issues related to i_size (5492,5624,5654,5672)
5411 - don't move pending lock onto export if it is already evicted (5683)
5412 - fix kernel oops when creating .foo in unlinked directory (5548)
5413 - fix deadlock in obdfilter statistics vs. object create (5811)
5414 - use time_{before,after} to avoid timer jiffies wrap (5882)
5415 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5416 - Do not leak request if server was not able to process it (5154)
5417 - If mds_open unable to find parent dir, make that negative lookup(5154)
5418 - don't create new directories with extent-mapping (5909, 5936)
5420 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5421 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5422 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5423 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5424 - support for RHEL4 kernel on i686 (5773)
5425 - provide error messages when incompatible logs are encountered (5898)
5427 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5428 * version 1.4.0.10 (1.4.1 release candidate 1)
5430 - don't keep a lock reference when lock is not granted (4238)
5431 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5432 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5433 - drop import inflight refcount on signal_completed_replay error (5255)
5434 - unlock page after async write error during send (3677)
5435 - handle missing objects in filter_preprw_read properly (5265)
5436 - no transno return for symlink open, don't save no-trasno open (3440)
5437 - don't try to complete elan receive that already failed (4012)
5438 - free RPC server reply state on error (5406)
5439 - clean up thread from ptlrpc_start_thread() on error (5160)
5440 - readahead could read extra page into cache that wasn't ejected (5388)
5441 - prevent races in class_attach/setup/cleanup/detach (5260)
5442 - don't dereference de->d_inode after l_dput of de (5458)
5443 - use "int" for stripe value returned from lock_to_stripe (5544)
5444 - mballoc allocation and error-checking fixes in 2.6 (5504)
5445 - block device patches to fix I/O request sizes in 2.6 (5482)
5446 - look up hostnames for IB nals (5602)
5447 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5448 - don't start multiple acceptors for the same port (5277)
5449 - fix incorrect LASSERT in mds_getattr_name (5635)
5450 - export a proc file for general "ping" checking (5628)
5451 - fix "lfs check" to not block when the MDS is down (5628)
5453 - service request history (4965)
5454 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5455 - create an "evict_client" /proc entry on OSTs, like the MDS has
5456 - fix mount usage message, return errors per mount(8) (5168)
5457 - change grep [] to grep "[]" in tests so they work in more UMLs
5458 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5459 - remove ancient LOV_MAGIC_V0 EA support (5047)
5460 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5461 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5462 - allow readahead allocations to fail when low on memory (5383)
5463 - mmap locking landed again, after considerable improvement (2828)
5464 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
5466 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
5469 - send OST transaction number in read/write reply to free req (4966)
5470 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5471 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5472 - the watchdog thread now runs as interruptible (5246)
5473 - flock/lockf fixes (but it's still disabled, pending 5135)
5474 - don't use EXT3 constants in llite code (5094)
5475 - memory shortage at startup could cause assertion (5176)
5477 - reorganization of lov code
5478 - single portals codebase
5480 - add extents/mballoc support (5025)
5481 - direct I/O reads in the obdfilter (4048)
5482 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
5484 tbd Cluster File Systems, Inc. <info@clusterfs.com>
5487 - send OST transaction number in read/write reply to free req (4966)
5488 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5489 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5490 - the watchdog thread now runs as interruptible (5246)
5491 - handle missing objects in filter_preprw_read properly (5265)
5492 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5493 - drop import inflight refcount on signal_completed_replay error (5255)
5494 - unlock page after async write error during send (3677)
5495 - return original error code on reconstructed replies (3761)
5496 - no transno return for symlink open, don't save no-trasno open (3440)
5498 - add pid to ldlm debugging output (4922)
5499 - bump the watchdog timeouts -- we can't handle 30sec yet
5500 - extra debugging for orphan dentry/inode bug (5259)
5502 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
5505 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
5506 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
5507 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
5508 - fix NULL dereference in /proc/sys/portals/routes (4827)
5509 - allow failed mdc_close() operations to be interrupted (4561)
5510 - stop precreate on OST before MDS would time out on it (4778)
5511 - don't send partial-page writes before EOF from client (4410)
5512 - discard client grant for sub-page writes on large-page clients (4520)
5513 - don't free dentries not owned by NFS code, check generation (4806)
5514 - fix lsm leak if mds_create_objects() fails (4801)
5515 - limit debug_daemon file size, always print CERROR messages (4789)
5516 - use transno after validating reply (3892)
5517 - process timed out requests if import state changes (3754)
5518 - update mtime on OST during writes, return in glimpse (4829)
5519 - add mkfsoptions to LDAP (4679)
5520 - use ->max_readahead method instead of zapping global ra (5039)
5521 - don't interrupt __l_wait_event() during strace
5523 - add software watchdogs to catch hung threads quickly (4941)
5524 - make lustrefs init script start after nfs is mounted
5525 - fix CWARN/ERROR duplication (4930)
5526 - return async write errors to application if possible (2248)
5527 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
5528 - print NAL number in %x format (4645)
5529 - update barely-supported suse-2.4.21-171 series (4842)
5530 - support for sles 9 %post scripts
5531 - support for building 2.6 kernel-source packages
5532 - support for sles km_* packages
5534 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
5537 - ignore -ENOENT errors in osc_destroy (3639)
5538 - notify osc create thread that OSC is being cleaned up (4600)
5539 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
5540 - reconstruct ost_handle() like mds_handle() (4657)
5541 - create a new thread to do import eviction to avoid deadlock (3969)
5542 - let lconf resolve symlinked-to devices (4629)
5543 - don't unlink "objects" from directory with default EA (4554)
5544 - hold socknal file ref over connect in case target is down (4394)
5545 - allow more than 32000 subdirectories in a single directory (3244)
5546 - fix blocks count for O_DIRECT writes (3751)
5547 - OST returns ENOSPC from object create when no space left (4539)
5548 - don't send truncate RPC if file size isn't changing (4410)
5549 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
5550 - bind to privileged port in socknal and tcpnal (3689)
5552 - rate limit CERROR/CWARN console message to avoid overload (4519)
5553 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
5554 - basic mmap support (3918)
5555 - kernel patch series update from b1_4 (4711)
5557 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
5560 - avoid crash during MDS cleanup with OST shut down (2775)
5561 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
5562 - don't use bad inodes on the MDS (3744)
5563 - dynamic object preallocation to improve recovery speed (4236)
5564 - don't hold spinlock over lock dumping or change debug flags (4401)
5565 - don't zero obd_dev when it is force cleaned (3651)
5566 - print grants to console if they go negative (4431)
5567 - "lctl deactivate" will stop automatic recovery attempts (3406)
5568 - look for existing locks in ldlm_handle_enqueue() (3764)
5569 - don't resolve lock handle twice in recovery avoiding race (4401)
5570 - revalidate should check working dir is a directory (4134)
5572 - don't always mark "slow" obdfilter messages as errors (4418)
5574 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
5577 - don't close LustreDB during write_conf until it is done (3860)
5578 - fix typo in lconf for_each_profile (3821)
5579 - allow dumping logs from multiple threads at one time (3820)
5580 - don't allow multiple threads in OSC recovery (3812)
5581 - fix debug_size parameters (3864)
5582 - fix mds_postrecov to initialize import for llog ctxt (3121)
5583 - replace config semaphore with spinlock (3306)
5584 - be sure to send a reply for a CANCEL rpc with bad export (3863)
5585 - don't allow enqueue to complete on a destroyed export (3822)
5586 - down write_lock before checking llog header bitmap (3825)
5587 - recover from lock replay timeout (3764)
5588 - up llog sem before sending rpc (3652)
5589 - reduce ns lock hold times when setting kms (3267)
5590 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
5591 - fix NULL deref and obd_dev leak on setup error (3312)
5592 - replace some LBUG about llog ops with error handling (3841)
5593 - don't match INVALID dentries from d_lookup and spin (3784)
5594 - hold dcache_lock while marking dentries INVALID and hashing (4255)
5595 - fix invalid assertion in ptlrpc_set_wait (3880)
5597 - add libwrap support for the TCP acceptor (3996)
5598 - add /proc/sys/portals/routes for non-root route listing (3994)
5599 - allow setting MDS UUID in .xml (2580)
5600 - print the stack of a process that LBUGs (4228)
5602 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
5605 - don't cleanup request in ll_file_open() on failed MDS open (3430)
5606 - make sure to unset replay flag from failed open requests (3440)
5607 - if default stripe count is 0, use OST count for inode size (3636)
5608 - update parent mtime/ctime on client for create/unlink (2611)
5609 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
5610 - free recovery state on server during a forced cleanup (3571)
5611 - unregister_reply for resent reqs (3063)
5612 - loop back devices mounting and status check on 2.6 (3563)
5613 - fix resource-creation race that can provoke i_size == 0 (3513)
5614 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
5615 - more debugging for page-accounting assertion (3746)
5616 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
5617 - avoid deadlock after precreation failure (3758)
5618 - fix race and lock order deadlock in orphan handling (3450, 3750)
5619 - add validity checks when grabbing inodes from l_ast_data (3599)
5621 - add /proc/.../recovery_status to obdfilter (3428)
5622 - lightweight CDEBUG infrastructure, debug daemon (3668)
5623 - change default OSC RPC parameters to be better on small clusters
5624 - turn off OST read cache for files smaller than 32MB
5625 - install man pages and include them in rpms (3100)
5626 - add new init script for (un)mounting lustre filesystems (2593)
5627 - run chkconfig in %post for init scripts (3701)
5628 - drop scimac NAL (unmaintained)
5630 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
5633 - clean kiobufs before and after use (3485)
5634 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
5635 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
5636 - change init script to fail more clearly if not run as root (1528)
5637 - allow clients to reconnect during replay (1742)
5638 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
5639 - don't do DNS lookups on NIDs too small for IP addresses (3442)
5640 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
5641 - fix cond_resched (3554)
5642 - only evict unfinished clients after recovery (3515)
5643 - allow bulk resend, prevent data loss (3570)
5644 - dynamic ptlrpc request buffer allocation (2102)
5645 - don't allow unlinking open directory if it isn't empty (2904)
5646 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
5647 - remove extraneous obd dereference causing LASSERT failure (3334)
5648 - don't use get_cycles() when creating temp. files on the mds (3156)
5649 - hold i_sem when setting i_size in ll_extent_lock() (3564)
5650 - handle EEXIST for set-stripe, set proper directory name (3336)
5652 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
5653 - fix ksocknal_fmb_callback() error messages (2918)
5655 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
5658 - don't copy lvb into (possibly NULL) reply on error (2983)
5659 - don't deref dentry after dput, don't free lvb on error (2922)
5660 - use the kms to determine writeback rpc length (2947)
5661 - increment oti_logcookies when osc is inactive (2948)
5662 - update client's i_blocks count via lvb messages (2543)
5663 - handle intent open/close of special files properly (1557)
5664 - mount MDS with errors=remount-ro, like obdfilter (2009)
5665 - initialize lock handle to avoid ASSERT on error cleanup (3057)
5666 - don't use cancelling-locks' kms values (2947)
5667 - use highest lock extent for kms, not last one (2925)
5668 - don't dereference ERR_PTR() dentry in error handling path (3107)
5669 - fix thread race in portals_debug_dumplog() (3122)
5670 - create lprocfs device entries at setup instead of at attach (1519)
5671 - common AST error handler, don't evict client on completion race (3145)
5672 - zero nameidata in detach_mnt in 2.6 (3118)
5673 - verify d_inode after revalidate_special is valid in 2.6 (3116)
5674 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
5675 - initialize RPC timeout timer earlier for 2.6 (3219)
5676 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
5677 - print nal/nid for unknown nid (3258)
5678 - additional checks for oscc recovery before doing precreate (3284)
5679 - fix ll_extent_lock() error return code for 64-bit systems (3043)
5680 - don't crash in mdc_close for bad permissions on open (3285)
5681 - zero i_rdev for non-device files (3147)
5682 - clear page->private before handing to FS, better assertion (3119)
5683 - tune the read pipeline (3236)
5684 - fix incorrect decref of invalidated dentry (2350)
5685 - provide read-ahead stats and refine rpc in flight stats (3328)
5686 - don't hold journal transaction open across create RPC (3313)
5687 - update atime on MDS at close time (3265)
5688 - close LDAP connection when recovering to avoid server load (3315)
5689 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
5690 - don't leak open file on MDS after open resend (3325)
5691 - serialize filter_precreate and filter_destroy_precreated (3329)
5692 - loop device shouldn't call sync_dev() for nul device (3092)
5693 - clear page cache after eviction (2766)
5694 - resynchronize MDS->OST in background (2824)
5695 - refuse to mount the same filesystem twice on same mountpoint (3394)
5696 - allow llmount to create routes for mounting behind routers (3320)
5697 - push lock cancellation to blocking thread for glimpse ASTs (3409)
5698 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
5699 - fix rare problem with rename on htree directories (3417)
5701 - allow default OST striping configuration per directory (1414)
5702 - fix compilation for qswnal for 2.6 kernels (3125)
5703 - increase maximum number of MDS request buffers for large systems
5704 - change liblustreapi to be useful for external progs like lfsck (3098)
5705 - increase local configuration timeout for slow disks (3353)
5706 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
5708 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5711 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
5712 - don't overwrite extent policy data in reply if lock was blocked (2901)
5713 - drop filter export grants atomically with removal from device (2663)
5714 - del obd_self_export from work_list in class_disconnect_exports (2908)
5715 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
5716 - swab reply message in mdc_close, other PPC fixes (2464)
5717 - fix destroying of named logs (2325)
5718 - overwrite old logs when running lconf --write_conf (2264)
5719 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
5720 - fix race in target_handle_connect (2898)
5721 - mds_reint_create() should take same inode create lock (2926)
5722 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
5723 - don't close files for self_export to avoid uninitialized obd (2936)
5724 - allow MDS with the same name as client node (2939)
5725 - hold dentry reference for closed log files for unlink (2325)
5726 - reserve space for all logs during transactions (2059)
5727 - don't evict page beyond end of stripe extent (2925)
5728 - don't oops on a deleted current working directory (2399)
5729 - handle hard links to targets without a parent properly (2517)
5730 - don't dereference NULL lock when racing during eviction (2867)
5731 - don't grow lock extents when lots of conflicting locks (2919)
5733 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
5736 - account for cache space usage on clients to avoid data loss (974)
5737 - lfsck support in lustre kernel code (2349)
5738 - reduce journal credits needed for BRW writes (2370)
5739 - orphan handling to avoid losing space on client/server crashes
5740 - ptlrpcd can be blocked, stopping ALL progress (2477)
5741 - use lock value blocks to assist in proper KMS, faster stat (1021)
5742 - takes i_sem instead of DLM locks internally on obdfilter (2720)
5743 - recovery for initial connections (2355)
5744 - fixes for mds_cleanup_orphans (1934)
5745 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
5746 - block all file creations until orphan recovery completes (1901)
5747 - client remove rq_connection from request struct (2423)
5748 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
5749 - recovery timer race (2670)
5750 - mdc_close recovey bug (2532)
5751 - ptlrpc cleanup bug (2710)
5752 - mds timeout on local locks (2588)
5753 - namespace lock held during RPCs (2431)
5754 - handle interrupted sync write properly (2503)
5755 - don't try to handle a message that hasn't been replied to (2699)
5756 - client assert failure during cleanup after abort recovery (2701)
5757 - leak mdc device after failed mount (2712)
5758 - ptlrpc_check_set allows timedout requests to complete (2714)
5759 - wait for inflight reqs when ptlrpcd finishes (2710)
5760 - make sure unregistered services are removed from the srv_list
5761 - reset bulk XID's when resending them (caught by 1138 test)
5762 - unregister_bulk after timeout
5763 - fix lconf error (2694)
5764 - handle write after unfinished setstripe, stripe-only getstripe (2388)
5765 - readahead locks pages, leaves pending causing memory pressure (2673)
5766 - increase OST request buffers to 4096 on large machines (2729)
5767 - fix up permission of existing directories in simple_mkdir (2661)
5768 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
5769 - don't assign transno to errored transactions (2742)
5770 - don't delete objects on OST if given a bogus objid from MDS (2751)
5771 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
5772 - if rq_no_resend, then timeout request after recovery (2432)
5773 - fix MDS llog_logid record size, 64-bit array alignment (2733)
5774 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
5775 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
5776 - MDS recovery shouldn't delete objects that it has given out (2730)
5777 - if enqueue arrives after completion, don't clobber LVB (2819)
5778 - don't unlock pages twice when trigger_group_io returns error (2814)
5779 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
5780 - don't write pages to disk if there was an error (1450)
5781 - don't ping imports that have recovery disabled (2676)
5782 - take buffered bytes into account when balancing socknal conn (2817)
5783 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
5784 - reconnect unlink llog connection after MDS reconnects to OST (2816)
5785 - remove little-endian swabbing of llog records (1987)
5786 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
5787 - retry reposting request buffers if they fail (1191)
5788 - grow extent at grant time to avoid granting a revoked lock (2809)
5789 - lock revoke doesn't evict page if covered by a second lock (2765)
5790 - disable VM readahead to avoid reading outside lock extents (2805)
5792 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
5793 - updated kernel patches for hp-2.4.20 kernel (2681)
5795 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
5798 - fix truncated write corruption (2366)
5799 - fix for failed assertion in iopen_connect_dentry (1792,2517)
5801 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
5802 - ensure len(uuid) < 37 in lmc (1171)
5803 - fix ia64 OOPS in llog_test (2255)
5804 - zero end of page at obdfilter for partial page writes (2648)
5805 - don't leave stale dentries around after renames (bug 2428)
5806 - fix timeouts when evicting a client with a single lock held (2642)
5807 - set deadline for the initial HELLO message to drain (2634)
5808 - print out dotted-quad IP addresses in the socknal (2302)
5810 - additional debugging for MDS client eviction problem (2443)
5811 - fix mkfsoptions support for osts (2603, 2604)
5813 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
5816 - add series for the vanilla 2.6.0 kernel
5817 - add series for the vanilla 2.4.24 kernel
5818 - add series for a cray x86/64 UL kernel drop
5819 - fix xattr patches for the vanilla 2.4.19 series
5821 - generate true UUIDs in lmc (1171)
5822 - have portals stack dumping break in UML (2466)
5823 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
5824 - allocate with _NOFS in ldlm to avoid deadlock (1933)
5825 - wake callback waiting threads on client eviction (2460)
5826 - Add --ptldebug and --subsystem to lmc (1719)
5827 - update assertion to allow safe interrupt allocation
5828 - set rq_no_resend for cancel requests (2432)
5829 - recalculate ptlrpcd timeout after resend (2494)
5830 - call vfs_rmdir when removing pending directories (2368)
5831 - fix renaming a file to itself (2429)
5832 - lmc creates a default one-stripe lov (2454)
5833 - expand procfs space to handle large clusters (2326)
5834 - increase UML stack to avoid overflow
5835 - update lconf's list of debug and subsystem masks
5836 - fix lfs find --obd (2510)
5837 - /proc tunable for disabling filter read caching (2591)
5838 - stop rpm packages from altering slapd.conf (2301)
5839 - disable nagle in the socknal under 0conf (2578)
5840 - choose mds inode size based on stripe count (2572)
5841 - fix kernel-source rpm problems (2516)
5843 - add --disable-doc to avoid pdf generation (2421)
5844 - update documentation, tests, type-os, comments
5845 - avoid format warnings on ia64
5846 - remove the TOE NAL
5847 - tiny code cleanups by removing unused fields
5849 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
5852 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
5853 - remove the most blatant lies from BUILDING file (2371)
5854 - change default debug level to reasonable production setting
5855 - reduce client side cache size to reduce cache flush time
5856 - reduce max RPCs in flight to avoid unnecessary file fragmentation
5857 - make TCP zerocopy and pinger support enabled by default (2476)
5858 - sync writes completed after process exits caused crashes (2319)
5859 - maintain correct mount count on the MDS (2356)
5860 - backout 1557, because 2316 wasn't really fixed
5861 - better file I/O statistics gathering in /proc
5862 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
5863 - another kernel patch to fix zero-copy TCP function export
5864 - don't take duplicate lock when processing re-sent getattr (2420)
5865 - lctl uses obd_self_export instead of creating new conn (2353)
5866 - MDS/OST recovery case which requires object creation asserted (2425)
5867 - move lfs from /usr/sbin to /usr/bin in packages
5868 - fix race between mds_client_add and mds_client_free (2417)
5869 - use kmalloc instead of slabs in portals (2430)
5870 - don't create duplicate records when a failover MDS is present (2442)
5871 - remove unnecessary mount age check (2332)
5872 - don't remove directory inodes from locks prematurely (2451)
5873 - don't break if MDS service name is the same as hostname (2103)
5874 - fix races in client write RPC generation when cache full (2482)
5876 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
5879 - remove now-unused request->rq_obd (278)
5880 - if an allocation fails, print out how much memory we've used (1933)
5881 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
5882 - add the "configurable stack size" patch to most series files (1256)
5883 - ability to write large log records, for 100+ OST configs (2306)
5884 - fix NULL deref when filter_prep fails (2314)
5885 - fix operator precedence error in filter_sync
5886 - dynamic allocation of socknal TX descriptors (2315)
5887 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
5888 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
5889 - ensure that CWARN messages in Portals always get to the syslog
5890 - __init/__exit are not for prototype decls (ldlm_init/exit)
5891 - x86-64 compile warning fixes
5892 - fix gateway LMC keyword conflict (2318)
5893 - fix MDS lock inversions in getattr/reint paths (1844)
5894 - fix a rare lock re-ordering bug, which caused deadlock (2322)
5895 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
5896 - DLM race condition prevented some lock evictions (2328)
5897 - ENOMEM detection and retry on socknal sends (2230)
5898 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
5899 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
5900 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
5901 - better error messages when a client is rejected during recovery (1505)
5902 - avoid cancelling locks which were never granted, after failure (2330)
5903 - fix i_sem/journal inversion in mds_client_add (2333)
5904 - fix truncate/getattr lock cycle deadlock (2334)
5905 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
5906 - fix two rare exit paths which could leak an l_lock() ref (2321)
5907 - fencepost error in MDS/OST orphan recovery (2226)
5908 - make log record alignment 8 bytes (1988)
5909 - lstripe now fails when requested offset > ost_count (2237)
5910 - ensure that all kernel series have a complete list.h (1607)
5911 - fix crashes in special-file operations (2316)
5912 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
5914 - allow configurable automake binary, for testing new versions
5915 - small update to the lfs documentation
5917 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
5919 * fix negative export reference count in fsfilt_sync (2312)
5921 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
5922 * release candidate 0.9.1
5924 - orphans are moved into the PENDING directory for possible recovery
5925 - replayed opens now open by fid for orphan/rename safety (1042)
5926 - last close of an orphan inode generates a transno (683)
5927 - chdir() and mount() now pin the directory entry (1020)
5928 - avoid CERROR in normal ll_setattr_raw() error case (1500)
5929 - discard very old requests without processing them (1502)
5930 - remove some common, well-understood CERRORs (1505)
5931 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
5932 - clear "grant" flags in OST replies until OST grant code lands (1644)
5933 - fix read performance by not clobbering i_blksize on client (1598)
5934 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
5935 - make LRU size a /proc tunable, clears locks when reduced (707)
5936 - fix some lprocfs rot that prevented ptlbd from loading (1732)
5937 - server locks take references on exports now (1558)
5938 - build fixes for 2.4.20-rh trees (1663)
5939 - return an error from lov_create if all OSCs are inactive (1751)
5940 - fix import levels when a reconnect happens without a timeout (1597)
5941 - exit early from mds_open if we get a lookup error (1749)
5942 - partial page read at EOF wouldn't wait for disk before sending (1642)
5943 - avoid NULL deref in obdfilter when reading page past EOF (1592)
5944 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
5945 - fix LBUG in ll_it_open_error with rc = -2 (1861)
5946 - write/truncate lock inversion (1639)
5947 - Don't auto-load obdclass, portals modules during cleanup (1495)
5948 - fix timestamps from jumping to "now" (1763)
5949 - extra journal assertions (1648)
5950 - add an extra multiunlink test (1771)
5951 - fix read_record/write_record API (1776)
5952 - fix leak of offset_extent, possible incorrect i_size later (1772)
5953 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
5954 - Debugging for the kqswnal_get_idle_tx problems (1820)
5955 - Allow recovery to be attempted multiple times (1536)
5956 - Write out MDS last_rcvd file after it is first created (1600)
5957 - Fix tx_descriptor leak in failed transmit situations (1827)
5958 - ext3 journaling fixes for assertion failure after IO error (1871)
5959 - class_export_put() on freed export after completion AST error (1896)
5960 - Fix revalidate looping in VFS (1322)
5961 - Don't access a freed export during MDS_REINT timeout (1521)
5962 - Add open-unlink recovery support on the MDS (1673,1764)
5963 - Return an error if no MDS data was read from last_rcvd (1946)
5964 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
5965 - Don't setup a disk device that doesn't match exported UUID (317)
5966 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
5967 - avoid committing NULL handle in force close
5968 - local.sh is now a one-stripe LOV configuration
5969 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
5970 - fix ext3 htree duplicate directory entry corruption (1516)
5971 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
5972 - update to new LOV EA format (2097)
5973 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
5975 2003-06-15 Phil Schwan <phil@clusterfs.com>
5978 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
5979 - per-import/export recovery handling (958, 931, 959)
5980 - multiple last-rcvd slots, for serving multiple FSes (949)
5981 - connections are again shared between multiple imp/exports (963, 964)
5982 - "umount -f" would hang if any requests needed to be sent (393, 978)
5983 - avoid pinning large req buffer by copying for queued messages (989)
5984 - add "uuid" to "lctl device" command to help upcalls (991)
5985 - "open" RPCs with transnos would confuse recovery counters (1037)
5986 - do proper endian conversion of all wire messages (288, 340, 891)
5987 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
5988 - call waiting locks callback from LDLM recovery thread (1127, 1151)
5989 - fix ptlrpc_connection leak in target_handle_connect (1174)
5990 - fix import refcounting bug in OST and MDS cleanup (1134)
5991 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
5992 - fix very unlikely obd_types race condition (501)
5993 - remove osc_open hack for echo_client (1187)
5994 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
5995 - a failure in read_inode2 leads to deadlock (1139)
5996 - cancel ack-locks as soon as transaction is committed (1072)
5997 - fix major leaks and crashes in the bulk I/O path (937, 1057)
5998 - make sure to commitrw after any preprw to avoid deadlock (1162)
5999 - failing to execute a file in a lustre FS would lock inode (1203)
6000 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
6001 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
6002 - obd_uuid2tgt would walk off the end of the list (1255)
6003 - on IA64 the osc would give portals incorrect bulk size (1258)
6004 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
6005 - fix lock inversion caused by new llite matching code (1282)
6006 - limit the number of dirty pages on a client to 10MB (1286)
6007 - timed out locks were not being corrected cancelled (1289)
6008 - fix O_DIRECT above 4GB on IA-32 (1292)
6009 * major user-visible changes
6010 - fail out/fail over policy now controlled by the upcall (993)
6012 - add OBD_PING to check server availability and failure (954)
6013 - lustre messages are now sent in sending host order (288, 340, 891)
6014 - add eadatalen to MDS getattr reply (340)
6015 - OST read replies may contain second buffer, with per-page status (593)
6017 2003-03-11 Phil Schwan <phil@clusterfs.com>
6020 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
6021 - failed open()s could cause deadlock; fixed (867, 869)
6022 - stop cancelling OST locks when files are closed (481)
6023 - overlapping XID spaces caused network corruption (851, 853)
6024 - fix unsafe fsfilt counter arithmetic; change to atomic_t
6025 - setattr_raw added, to do single-RPC, server-side setattrs
6026 - lmc/lconf syntax change for OST UUIDs
6027 - fix crashy race condition between ptlrpc_free_req and osc_close
6028 - don't use request in mdc_enqueue if we hit a timeout (889)
6029 - don't set the inode i_size for regular files from the MDS (896)
6030 - handle out of order completion AST (842)
6031 - don't LBUG if a lock request times out after receiving AST (913)
6032 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
6033 - if a bad lock AST arrives, send an error instead of dropping entirely
6034 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
6035 - fix leak in bulk IO when only partially completed (899, 900, 926)
6036 - fix O_DIRECT for ia64 (55)
6037 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
6038 - C-z after timeout could hang a process forever; fixed (977)
6040 - client-side I/O cache (678, 924, 929, 941, 970)
6042 - READPAGE and SETATTRs which don't take server-side locks get
6045 2003-02-11 Phil Schwan <phil@clusterfs.com>
6048 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
6049 - Fix fsfilt_extN_readpage() to read a full page of directory
6050 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
6051 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
6052 - Fix imbalanced LOV object allocation and out-of-bound access (469)
6053 - Most intent operations were removed, in favour of a new RPC mode
6054 that does a single RPC to the server and bypasses most of the VFS
6055 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
6056 - Aggressively cancel local locks on DLM servers
6057 - mds_reint_unlink sends EA to the client if it's the last nlink.
6058 client uses that EA to unlink OST objects.
6059 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
6060 - recursive symlinks were fixed (439)
6061 - fixed NULL deref in DEBUG_REQ
6062 - filter_update_lastobjid no longer calls sync, which annoyed extN
6063 - fixed multi-client small-writes to a single file problem (445)
6064 - fixed mtime updates during file writes (607)
6065 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
6066 - fixed bug in obd_brw_read/write() (under guise of testing 367)
6067 - fixed Linux OST size reporting problem (444, 656)
6068 - OST now updates object mtime with writes or setattr (607, 619)
6069 - client verifies file size before zeroing page past EOF (445)
6070 - OST now writes last allocated objid to disk with allocation (108)
6071 - LOV on echo now works (409)
6073 - mds_reint_unlink sends a new buffer, with the EA included. this
6074 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
6075 set if a regular file was being unlinked, and it was the last link
6076 - use PtlGet from the target for bulk writes (315)
6077 - OST now updates object mtime with writes or setattr (607, 619)
6078 - LDLM now has a grant-time callback to revalidate locked items, if
6080 - Many MDS operations were reorganized to combat race conditions
6082 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
6083 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
6085 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
6088 - Fully reactivate OST imports after reconnection (512, others)
6089 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
6090 - More graceful error handling for truncating on dead OST (515)
6091 - Don't error out unless we're actually accessing dead stripes (474)
6092 - Fix garbage sizes when stripes are missing (410)
6093 - LRU counters were broken, causing constant lock purge (433, 432)
6094 - garbage on read from stripes with failed OSTs (441)
6095 - mark OSCs as active before reconnecting during recovery (438)
6096 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
6097 - lfind did not preserve OST order in output (443)
6098 - symlinks cause hung clients, incorrect data (439)
6099 - stop dereferencing request after dropping refcount (457)
6100 - don't LASSERT(spin_is_locked) on non-SMP (455)
6101 - fixes for many rename() bugs
6102 - fstat didn't correctly synchronize attributes (399)
6103 - server must handle lock cancellation during blocking AST prep (487)
6104 - bulk descriptors were free()d too soon (511)
6105 - fix paths in lconf, which would load incorrect modules (451, 507)
6106 - fix confusing lconf 'host not found' error message (386)
6107 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
6108 - fix race condition in mdc_blocking_ast() for inode access (526)
6109 - fix lov_unpackmd() unpacking wrong number of stripes (537)
6110 - fix lov_set_osc_active() marking wrong OSC inactive (440)
6111 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
6112 - fix multiple writes of stripe MD to MDS (358, maybe 519)
6113 - fix lstripe in several ways (kernel side) (527)
6114 - fix request leak in ldlm_cli_enqueue (262)
6115 - incorrect OSC was marked inactive after OST failure
6116 - call mds_fs_cleanup before unmounting filesystem (524)
6117 - fix races between taking ns_lock and ldlm_lock_change_resource
6118 - fix races updating LOV export open file list
6119 - fix lov_enqueue error path, avoid decref-ing bad lock handle
6120 - fix recovery NULL deref in ldlm_cli_cancel_unused
6121 - fix some DLM races by using new hash table for lock handles (419)
6122 - permit the client to specify desired inodes, at replay
6123 - duplicate requests when we queue them for replay reintegration
6124 - fix last_rcvd offset calculation
6125 - sync after each recovered transaction, so we always make progress
6126 - never, not always, ERESTART requests without transnos
6127 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
6129 - skip replay if the MDS says that the client is already connected
6130 - don't check for a recovery-enabled export to match lctl's UUID
6131 - don't INC_USE_COUNT for phantom exports
6132 - don't crash when cleaning up phantom exports (567)
6133 - don't double-finish or set replay data for errored mdc_open requests
6134 - abort requests when they time out, so we don't get old replies
6135 - send/receive replies for AST messages again
6136 - if the client says that it doesn't have the lock, cancel it on the
6138 - if we timeout during I/O, don't try to cancel an in-use lock; instead
6139 mark it as destroyed, it will all work out when decref is called
6140 - fix module use counts (22, 581)
6142 - ASTs now expect a reply (server cancels lock on error reply)
6144 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
6147 - fix many simultaneous client startup (392)
6148 - fix dentry->d_it clobbering
6149 - credentials weren't being shipped for readdir/getattr operations
6150 - remove invalid assertions triggered during some concurrent MD
6152 - proper Lustre versions added (336, 389)
6153 - fix memory leak for create error case (398)
6154 - fix LOV locking bug that would get cli/srv out of sync
6155 - fix echo client over LOV (409)
6156 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
6157 - fix double-O_EXCL intent crash (424)
6158 - avoid sending multiple lock CANCELs (352)
6160 - MDS can do multi-client recovery (modulo bugs in new code)
6162 - many updates, edits, cleanups
6164 2002-11-18 Phil Schwan <phil@clusterfs.com>
6167 - fix null d_it dereference (346)
6168 - fix full OST/dbench hang (333)
6169 - fix permission problem with file removal (286)
6170 - fix removal of OSCs from LOV when they fail
6171 - fix NULL deref during bulk timeout (214)
6172 - fix problems related to multiple filesystems on one MDS (241)
6173 - fixed serious subtle metadata locking bugs
6174 - free locks on clients when inodes are removed due to memory
6176 - fix inode pointer in lock data (285)
6177 - partial support for multiple MDS on a single host (241)
6178 - data locks weren't cancelled at clear_inode time (290, 311)
6179 - intent locks could lead to unbounded lock growth (205)
6180 - added a maximum lock count, an LRU list, and a flusher
6181 - fix multiple rename (365)
6182 - properly abstracted the echo client
6183 - OSC locked 1 byte too many; fixed
6184 - rewrote brw callback code:
6185 - fixed recovery bugs related to LOVs (306)
6186 - fixed too-many-pages-in-one-write crash (191)
6187 - fixed (again) crash in sync_io_timeout (214)
6188 - probably fixed callback-related race (385)
6190 - Add capability to MDS protocol
6191 - LDLM cancellations and callbacks on different portals
6193 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
6196 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
6197 - defer open object destruction to close time (601981, 138)
6198 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
6199 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
6200 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
6201 - don't call obd_setattr in ll_file_release for destroyed objects
6202 * protocol change to lustre_msg: move |version| and add |flags|
6203 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
6204 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
6205 * added replay of create, unlink, link and rename operations during
6206 MDS failover; recovery should be much more robust now
6207 * remove failed OSCs from LOVs (only lov_create uses this so far)
6208 * the lustre-HOWTO was brought (more) up to date (582544)
6210 2002-10-23 Phil Schwan <phil@clusterfs.com>
6213 - in-use dentries weren't being reused properly (617851)
6214 - prevent multiple LDLM setup (599178)
6215 - fix LOV size calculations for truncate (617853)
6216 - fix client handling of MDS intent errors (POSIX)
6217 - fix permission bug in lovstripe.c test (624321)
6218 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
6219 - truncate past end of file could corrupt data
6220 - proper cleanup after timeouts, crashes, etc (592524, 550815)
6221 - a race in recovery could return ETIMEDOUT to apps (623947)
6222 - building outside the source directory was fixed
6223 * the lustre-HOWTO was brought (more) up to date (582544)
6224 * major progress was made on recovery functionality
6226 2002-10-10 Phil Schwan <phil@clusterfs.com>
6229 - recovery deadlock fix
6230 - rm -rf causes LBUG fix (617817)
6231 - file open by multiple tasks fix (618962)
6232 - directory permissions bugs (602707 and 620007)
6233 - journal_stop fixed with locking (611313)
6234 - O_APPEND failures resolved (618273, perhaps 614459)
6235 - lconf PATH fix (619770)
6236 - IA64 build fix (621450)
6237 - RPC buffer sizes scale with amount of memory
6239 2002-10-01 Phil Schwan <phil@clusterfs.com>
6242 - locks would be cancelled without throwing away data pages,
6243 resulting in inconsistent data (605627)
6244 - inode attributes were not always being refreshed (605627, 612449)
6245 - lconf now continues to cleanup after lctl reports an error
6246 - MDS now enforces user permissions (602707)
6247 - lprocfs cleanup fixed, but not yet enabled (614157)
6248 - fixed infinite server hang, should a client not respond to an AST
6249 - avoid going into recovery if user calls readlink() with a buffer
6250 that's too small (613941)
6251 - AST RPCs no longer require replies (614867) -- this may be changed
6252 - don't crash server if client sends an IOV that's too big (611336)
6253 - fixed lock conversion deadlock (611892)
6254 - fixed the following of symlinks (614622)
6255 * recovery: the server can remove locks from a client that dies, other
6256 clients can make progress
6257 * more extN patch fixes
6258 * compile-time configurable ptlrpc buffer allocations
6260 - collaborative read cache document
6261 - Lustre Lite Performance CDR document-in-progress
6263 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6266 - fix typo in patch-2.4.18
6268 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6271 - clear ptlrpc request each time in handle_incoming_request()
6272 - unlink of files now destroys the object on the OST
6274 2002-09-19 Peter Braam <braam@clusterfs.com>
6276 * add hard link support
6277 * change obdfile creation method
6278 * kernel patch changed
6280 2002-09-19 Peter Braam <braam@clusterfs.com>
6283 - stack overflow bug in extN fixed
6285 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
6287 * documentation updates
6288 - add man pages for config tools
6289 - update tests/README to describe testing with new config tools
6290 - finish metadata API descriptions
6291 * bug fixes and cleanups
6292 - statfs workaround for 16TB limit
6293 - LOV stripe allocation improved, can stripe on subset of OSTs
6294 - LOV file size/IO offset was wrong for files > 4GB in size
6295 - object EA data was being dropped, caused files to be unreadable
6296 - memory overflow with non-LOV OST caused memory corruption
6297 - fixed regression tests to work with new config tools, obdfilter
6298 - fixed bug when directory size became larger than 1 block
6299 - fixed bug (for single client case) when PWD was deleted
6300 - invalidate local directory pages when doing intent-based ops
6301 - avoid LDLM oops when lock callback contained bad data
6303 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
6305 * documentation updates
6306 * bug fixes and cleanups
6307 - configuration tools
6310 - 64-bit compile warnings
6311 - 64-bit internal statfs data
6313 * test_brw on persistent OST devices
6315 * lprocfs (disabled)
6317 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
6319 * documentation updates
6320 * bug fixes and cleanups
6321 * configuration tools
6323 2002-08-30 Peter J. Braam <braam@clusterfs.com>
6326 * many small fixes to 0_5_4
6327 * io/network handling
6328 * thinkos in MDS operations
6330 2002-08-24 Peter J. Braam <braam@clusterfs.com>
6333 * crucial basic fixes to 0.5.3
6334 * IOR, Iozone work over Elan
6337 2002-08-07 Phil Schwan <phil@clusterfs.com>
6338 * version 0_5_3, our first alpha
6339 * we use the new Portals iovs
6340 * documentation updates
6341 * bug fixes and cleanups
6342 * small changes in the DLM wire protocol
6344 2002-07-25 Peter J. Braam <braam@clusterfs.com>
6345 * version 0_5_1 with some initial stability,
6346 * locking on MD and file I/O.
6347 * documentation updates
6348 * several bug fixes since 0.5.0
6349 * small changes in wire protocol
6351 2002-07-18 Phil Schwan <phil@clusterfs.com>
6353 * delivered as Lustre Light Alpha
6354 * fixed a crash after handling invalid MDS requests
6355 * fixed directory pages for architectures with non-4k pages sizes
6357 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
6358 * release version v0_4_4
6359 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
6361 * Fixes a number of interruption problems with OST operations.
6362 * Update documentation for portals header changes
6363 * Move all wire protocol structs/defines to lustre_idl.h
6364 * Fixes symlink length bug.
6365 * Add tcpdump to repository.
6367 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
6368 * release version v0_4_3
6369 * Fixes statfs for inodes on extN.
6370 * Fixes bug in runtests which would delete /etc/hosts.
6371 * Use 64-bit object IDs wherever possible (not into VFS though)
6372 Remove ost_get_info, which is unused by lustre, and out of date.
6374 2002-07-03 Peter Braam <braam@clusterfs.com>
6375 * release version v0_4_2 Fixes a lookup error (type not passed)
6376 * move forward to head of Portals
6377 * move forward to latest Lustre kernel
6379 2002-06-25 Peter Braam <braam@clusterfs.com>
6380 * release version v0_4_1. Hopefully stable on single node use.