3 * See http://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
4 * Server support for kernels:
5 2.6.32-358.23.2.el6 (RHEL6)
6 * Client support for unpatched kernels:
7 2.6.32-358.23.2.el6 (RHEL6)
9 2.6.18-348.3.1.el5 (RHEL5)
10 2.6.16.60-0.69.1 (SLES10)
12 * Recommended e2fsprogs version: 1.42.7.wc1
13 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
14 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
15 stack overflow. For more information, please refer to bugzilla 17630.
16 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
17 the re-exporting nodes to be mounted with "32bitapi" mount option
19 --------------------------------------------------------------------------------
21 10-31-2013 Intel Corporation
23 * See http://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
24 * Server support for kernels:
25 2.6.32-358.18.1.el6 (RHEL6)
26 * Client support for unpatched kernels:
27 2.6.32-358.18.1.el6 (RHEL6)
29 2.6.18-348.3.1.el5 (RHEL5)
30 2.6.16.60-0.69.1 (SLES10)
32 * Recommended e2fsprogs version: 1.42.7.wc1
33 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
34 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
35 stack overflow. For more information, please refer to bugzilla 17630.
36 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
37 the re-exporting nodes to be mounted with "32bitapi" mount option
39 --------------------------------------------------------------------------------
41 05-30-2013 Intel Corporation
43 * See http://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
44 * Server support for kernels:
45 2.6.32-358.6.1.el6 (RHEL6)
46 * Client support for unpatched kernels:
47 2.6.32-358.6.1.el6 (RHEL6)
49 2.6.18-348.3.1.el5 (RHEL5)
50 2.6.16.60-0.69.1 (SLES10)
52 * Recommended e2fsprogs version: 1.42.7.wc1
53 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
54 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
55 stack overflow. For more information, please refer to bugzilla 17630.
56 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
57 the re-exporting nodes to be mounted with "32bitapi" mount option
59 --------------------------------------------------------------------------------
61 9-30-2012 Whamcloud, Inc.
63 * See http://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
64 * Server support for kernels:
65 2.6.32-279.5.1.el6 (RHEL6)
66 * Client support for unpatched kernels:
67 2.6.32-279.5.1.el6 (RHEL6)
69 2.6.18-238.19.1.el5 (RHEL5)
70 2.6.16.60-0.69.1 (SLES10)
71 * Recommended e2fsprogs version: 1.41.90.wc4
72 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
73 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
74 stack overflow. For more information, please refer to bugzilla 17630.
75 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
76 the re-exporting nodes to be mounted with "32bitapi" mount option
78 Severity : enhancement
80 Description: Update RHEL6.3 kernel to 2.6.32-279.5.1.el6.
82 Severity : enhancement
84 Description: Update RHEL6.3 kernel to 2.6.32-279.2.1.el6.
86 Severity : enhancement
88 Description: Update RHEL6.2 kernel to 2.6.32-220.17.1.el6
90 Severity : enhancement
92 Description: Update RHEL6.2 kernel to 2.6.32-220.13.1.el6
94 Severity : enhancement
96 Description: Update RHEL6.2 kernel to 2.6.32-220.7.1.el6
98 --------------------------------------------------------------------------------
100 04-30-2012 Whamcloud, Inc.
102 * See http://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
103 * Server support for kernels:
104 2.6.32-220.4.2.el6 (RHEL6)
105 * Client support for unpatched kernels:
106 2.6.32-220.4.2.el6 (RHEL6)
108 2.6.18-238.19.1.el5 (RHEL5)
109 2.6.16.60-0.69.1 (SLES10)
110 * Recommended e2fsprogs version: 1.41.90.wc4
111 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
112 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
113 stack overflow. For more information, please refer to bugzilla 17630.
114 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
115 the re-exporting nodes to be mounted with "32bitapi" mount option
117 Severity : enhancement
119 Description: Update RHEL6.2 kernel to 2.6.32-220.4.2.el6
121 Severity : enhancement
123 Description: Update RHEL6.2 kernel to 2.6.32-220.4.1.el6
125 Severity : enhancement
127 Description: Update RHEL6.2 kernel to 2.6.32-220.el6
129 Severity : enhancement
131 Description: Update RHEL6.1 kernel to 2.6.32-131.17.1.el6
133 --------------------------------------------------------------------------------
135 09-30-2011 Whamcloud, Inc.
137 * Server support for kernels:
138 2.6.18-238.19.1.el5 (RHEL5)
139 2.6.32-131.6.1.el6 (RHEL6)
140 * Client support for unpatched kernels:
141 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
142 2.6.18-238.19.1.el5 (RHEL5)
143 2.6.32-131.6.1.el6 (RHEL6)
144 2.6.29.4-167.fc11 (FC11)
145 2.6.16.60-0.69.1 (SLES10)
147 * Recommended e2fsprogs version: 1.41.90.wc3
148 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
149 Lustre filesystem with 4K stack may cause a stack overflow. For more
150 information, please refer to bugzilla 17630.
151 * NFSv4 reexport requires Lustre client on the re-exporting nodes to be
152 mounted with "32bitapi" mount option
154 Severity : enhancement
155 Description: Update RHEL6.1 kernel to 2.6.32-131.6.1.el6
157 Severity : enhancement
158 Description: Update RHEL5.6 kernel to 2.6.18-238.19.1.el5
160 Severity : enhancement
161 Description: Update RHEL5.6 kernel to 2.6.18-238.9.1.el5
164 Frequency : only with SLES10
166 Description: Use OFED "KMP" provided by Novell
167 Details : SLES10 SP3 ships with OFED in a separate "KMP" package.
168 Lustre is now built against this package. That means you need to
169 install the ofed-kmp package from Novell for the patchless client.
170 Note that the ofed-kmp that Novell ships may not exactly match the
171 kernel version but should still be compatible.
173 Severity : enhancement
175 Description: Update RHEL5.5 kernel to 2.6.18-194.17.1.el5
176 Update OEL5.5 kernel to 2.6.18-194.17.1.0.1.el5.
178 --------------------------------------------------------------------------------
180 2010-07-15 Oracle, Inc.
182 * Server support for kernels:
183 2.6.18-164.11.1.el5 (RHEL 5)
184 2.6.18-164.11.1.0.1.el5 (OEL 5)
185 * Client support for unpatched kernels:
186 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
187 2.6.18-164.11.1.el5 (RHEL 5),
188 2.6.18-164.11.1.0.1.el5 (OEL 5)
189 2.6.16.60-0.42.8 (SLES 10),
191 2.6.29.4-167.fc11 (FC11)
192 * Recommended e2fsprogs version: 1.41.10-sun2
194 Severity : enhancement
196 Description: Distribute lfs_migrate script with lustre
197 Details : The lfs_migrate mentioned in the manual is now included in the
198 lustre tree in lustre/scripts and will distributed in lustre RPMs.
200 Severity : enhancement
202 Description: Failover nid list added to import proc file.
204 Severity : enhancement
206 Description: "som_preview" mount option required to use SOM
207 Details : The clients will need to use som_preview mount option
211 Description: Lustre does not do 1MB IOs to HW RAID
212 Details : Bump MAX_PHYS/HW_SEGMENTS and SG_ALL to 256 in the RHEL5 kernel.
214 Severity : enhancement
216 Description: lfs quota output is not very convenient for awk/sed-parsing
217 Details : Some positions in lfs quota output table could be empty or
218 non-empty which made it hard to parse it with scripts, now
219 a dash is put instead of space where there is not supposed
224 Description: Update to OFED 1.5.1
225 Details : OFED 1.5.1 will be used in environments that don't use
226 vendor supplied OFED.
230 Description: Remove lnet. prefix from lctl params display
231 Details : Also renamed lnet.memused to lnet_memused to
236 Description: Use conf_param -d to remove permanant settings.
240 Description: Add list_param -R
241 Details : lctl list_param now supports recursive mode.
244 Bugzilla : 21829, 22312
245 Description: Fix llobdstat.
246 Details : 1. Now updating the write_bytes counter
247 2. Port script fixes from 1.8.
251 Description: decrease the usage of memory on clients.
252 Details : 1. On clients, recycle dentries and inodes unused.
253 2. Delete the code related to ll_deathrow(att 6215 in bug 1443). It
258 Description: llobdstat fix and enhancement
259 Details : add a counter to set a limit to how many samples will be returned
260 fix a wildcard in the path to limit to obdfilter stats only
264 Description: lfs check only allowed for root
265 Details : Code cleanups and added a regresion test of non-root use of
270 Description: MDS Oopses in ptlrpc_dispatch_difficult_reply()/__list_add()
271 Details : Fix hr->hr_index overload.
275 Description: Fix FIEMAP functionality.
279 Description: FC11 patchless client support added
283 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and
284 OEL5.4 kernel to 2.6.18-164.11.1.0.1.el5.
288 Description: New acc-sm script and various test-framework enhancments.
292 Description: Previously landed patch was reverted.
296 Description: Update SLES10 kernel to 2.6.16.60-0.42.8 (patchless)
300 Description: Add ldiskfs maxdirsize mount option.
304 Description: Fix mkfs.lustre support for 16TB LUNs.
308 Description: Print mmp_check_interval
309 Details : print mmp_check_interval and make it possible to abort
310 mount operation in case it takes too long.
314 Description: Add cascading_rw to lustre mpi tests.
318 Description: Store SOM data in EA.
319 Details : Also several SOM related cleanups.
323 Description: Add refcount to md_open_data
327 Description: Use refcount on lov_request_set to prevent memory corruption.
331 Description: Statahead: use parent dir's i_mutex for synchronization.
335 Description: loadgen improvements.
339 Description: Enable peer health detection in ksocklnd and ko2iblnd.
343 Description: Control DCACHE_LUSTRE_INVALID flag with MDS_INODE_LOOKCUP lock
347 Description: Fix race on obd_devcie:md_stats
351 Description: Fix SLES 11 client builds.
352 Details : Resolved uninitalized pointer gcc warnings.
356 Description: Allow non-root access for "lfs check".
357 Details : Added a check in obd_class_ioctl() for OBD_IOC_PING_TARGET.
361 Description: Clear imp_force_reconnect correctly in ptlrpc_connect_interpret()
363 Severity : enhancement
365 Description: Update kernel to RHEL5.4 2.6.18-164.6.1.el5 and
366 OEL5 2.6.18-164.6.1.0.1.el5(Both in-kernel OFED enabled).
370 Description: Sanity test_27p fails.
371 Details : OBD_FAIL_OST_ENOSPC has been moved to filter_statfs().
372 Sanity tests 27 were cleaned up and improved.
376 Description: Unify req format on client/servers
377 Details : Use new req_capsule API [almost] everywhere instead of old PTLRPC
378 buffers and swabbers approach..
383 Description: drop join file stuff.
385 Severity : enhancement
387 Description: Add 2.6.27 and SLES11 (patchless client) support
391 Description: Fix in ptlrpc_expire_one_request() to print the signed time
394 Severity : enhancement
396 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
399 Severity : enhancement
401 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
403 Severity : enhancement
405 Description: correctly handle big reply message.
406 Details : send LNet event if reply is bigger then buffer and adjust this buffer
411 Description: Drop unnecessary __GFP_NOMEMALLOC flag from filter_get_page()
413 Severity : enhancement
415 Description: Add support for OEL5.
419 Description: Conf-sanity.sh 50g test - deactivated OST should not cause a panic.
421 Severity : enhancement
423 Description: "weak-modules" support
424 Details : Implement "weak-modules" support which enables kernel modules
425 to be used with any kernel that implements the same kABI. In
426 order to achieve this modules are now installed in
427 /lib/modules/$(uname -r)/updates/kernel on all distributions.
431 Description: client could not reconnect to OST because of an active request.
432 Details : The client now retries to reconnect to the same server, if a
433 connect request failed with EBUSY or -EAGAIN. Test case 35b
434 conf-sanity.sh has been created.
438 Description: Don't allow make backward step on assignin osc next id.
439 Details : race between allocation next id and ll_sync thread can be cause
440 of set wrong osc next id and can be kill valid ost objects.
444 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
448 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
452 Description: Error handling in osc_statfs_interpret() has been improved.
453 Details : Check in osc_statfs_interpret() for EBADR.
457 Description: Do not send statfs() requests to OSTs disabled by administrator.
458 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
460 Severity : enhancement
462 Description: Added sanity-quota test to for lfs(1) to work with numeric uid/gid.
464 Severity : enhancement
466 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
468 Severity : enhancement
470 Description: Enhancement for lfs(1) command to use numeric uid/gid.
474 Description: Improved cleanup procedure for conf-sanity test_46a.
478 Description: Do not start bulk transfer, if client tried to reconnect.
479 Details : Do not start bulk transfer, if client tried to reconnect,
480 but failed due to active rpcs. Check for exp_abort_active_req
481 has been implemented.
485 Description: Prevent inconsistences between linux and lustre mount structures.
486 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
487 Make the sleep interruptible.
489 Severity : enhancement
491 Description: provide server to client comms path
492 Details : server to client communications path via new LDLM_SET_INFO rpc
496 Description: truncate starts GFP_FS allocation under transaction causing deadlock
497 Details : ldiskfs_truncate calls grab_cache_page which may start page
498 allocation under an open transaction. This may lead to
499 calling prune_icache with consequent lustre reentrance.
503 Description: send correctly lsm on open replay
504 Details : MDS is trust to LSM size on replay open, but client can set wrong size
509 Description: enable client interface failover
510 Details : When a child reconnects from another NID, properly update export
511 nid hash position and ldlm reverse import.
513 Severity : enhancement
515 Description: Communicate OST degraded/readonly state via statfs to MDS
516 Details : Flags in the statfs returned from OSTs indicate whether the
517 OST is in a degraded RAID state, or if the filesystem has
518 turned read-only after a filesystem error is detected.
523 Description: don't panic if EPROTO was hit when reading symlink
524 Details : correctly handling request reference in error cases.
526 Severity : enhancement
528 Description: Add LustreNetLink, a kernel-userspace communcation path.
530 Severity : enhancement
532 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
537 Description: access to llog context before init.
538 Details : move handling CATALOGS file at osc layer and forbid access to llog
542 Frequency : rare, only when client is writing to OST on local node
544 Description: Avoid deadlock for local client writes
545 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
546 memory freeing context. This allows OST threads to set the
547 PF_MEMALLOC flag on task structures in order to allocate memory
548 from reserved pools and complete IO.
549 Use GFP_HIGHUSER for OST allocations for non-local client writes,
550 so that the OST threads generate memory pressure and allow
551 inactive pages to be reclaimed.
553 Severity : enhancement
555 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
560 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
561 Details : move ldlm namespace creation in setup phase to avoid grab
562 _lprocfs_lock with cli_sem held.
566 Description: Temporarily disable grant shrink.
567 Details : Disable the feature for debugging.
570 Frequency : only when formatting test filesystems
572 Description: Unable to run several mkfs.lustre on loop devices at the same time
573 Details : mkfs.lustre returns error 256 on the concurrent loop devices
574 formatting. The solution is to proper handle the error.
576 Severity : enhancement
578 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
580 Severity : enhancement
582 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
587 Description: resolve race between obd_disconnect and class_disconnect_exports
588 Details : if obd_disconnect will be called to already disconnected export he
589 forget release one reference and osc module can't unloaded.
591 Severity : enhancement
593 Description: Allow tuning service thread via /proc
594 Details : For each service a new
595 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
596 created that can be used to set min/max thread counts, and get the
597 current number of running threads.
602 Descriptoin: don't return error if have partially created objects for file.
603 Details : lov_update_create_set uses set->set_success as index for created
604 objects, so if some requests will be failed, they will have hole
605 at end of array and qos_shrink_lsm for allocate correct lsm.
607 Severity : enhancement
609 Description: Update OFED support to 1.4.1
614 Description: Reduce small size read RPC
615 Details : Set read-ahead limit for every file and only do read-ahead when
616 available read-ahead pages are bigger than 1M to avoid small size
619 Severity : enhancement
621 Description: move AT tunable parameters for more consistent usage
622 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
624 Severity : enhancement
626 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
627 Details : allow skip disconnected ost for send statfs request and hide error
633 Description: fix racy locking of mballoc block bitmaps causing BUG
634 Details : The locking of the mballoc buddy bitmap and the in-memory
635 block bitmap was using two different spin locks in some
636 cases. This made it possible to incorrectly access the
637 mballoc bitmap while another process was modifying it,
638 causing a sanity assertion to fail. While no on-disk corruption
639 was reported, there was some risk of this happening.
642 Frequency : rare, on llog test 6
644 Descriptoin: don't allow connect to already connected import
645 Details : allowing connect to already connected import is hide connecting problem.
650 Descriptoin: allow kill process which wait statahead result
651 Details : for some reasons 'ls' can stick in waiting result from statahead,
652 in this case need way for kill this process.
654 Severity : enhancement
656 Description: Add state history info file, enhance import info file
657 Details : Track import connection state changes in a new osc/mdc proc file;
658 add overview-type data to the osc/mdc import proc file.
660 Severity : enhancement
662 Description: MDS create should not wait for statfs RPC while holding DLM lock.
664 Severity : enhancement
666 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
671 Descriptoin: OST grant too much space to client even there are not enough space.
672 Details : Client will shrink its grant cache to OST if there are no write
673 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
674 this grant cache if there are already not enough avaible space
675 (left_space < total_clients * 32M).
678 Frequency : start MDS on uncleanly shutdowned MDS device
680 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
681 Details : stay in waiting mds<>ost recovery finished produce random bugs
682 due race between two ll_sync thread for one lov target. send
683 ACTIVATE event only if connect realy finished and import have
687 Frequency : rare, connect and disconnect target at same time
689 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
690 Details : don't call obd_disconnect under lov_lock. this long time
691 operation and can block ptlrpcd which answer to connect request.
696 Descriptoin: don't lose wakeup for imp_recovery_waitq
697 Details : recover_import_no_retry or invalidate_import and import_close can
698 both sleep on imp_recovery_waitq, but we was send only one wakeup
702 Frequency : always with long access acl
704 Descriptoin: mds can't pack reply with long acl.
705 Details : mds don't control size of acl but they limited by reint/getattr
708 Severity : enhancement
710 Description: Update to SLES10 kernel-2.6.16.60-0.33.
712 Severity : enhancement
714 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
717 Frequency : start MDS on uncleanly shutdowned MDS device
719 Descriptoin: aborting recovery hang on MDS
720 Details : don't throttle destroy RPCs for the MDT.
723 Frequency : on remount
725 Description: external journal device not working after the remount
726 Details : clear dev_rdonly flag for external journal devices in
732 Description: shutdown vs evict race
733 Details : client_disconnect_export vs connect request race.
734 if client will evicted at this time - we start invalidate
735 thread without referece to import and import can be freed
739 Frequency : rare, need acl's on inode.
741 Description: client can't handle ost additional correctly
742 Details : if ost was added after client connected to mds client can have
743 hit lnet_try_match_md ... to big messages to wide striped files.
744 in this case need teach client to handle config events about add
745 lov target and update client max ea size at that event.
747 Severity : enhancement
749 Description: Changelogs
750 Details : Changelogs are a lightweight mechanism to track filesystem
751 metadata and namespace changes. The changelog is recorded
752 permanently on the MDTs, and is periodically "consumed" / purged
753 when records are no longer needed.
755 Severity : enhancement
757 Description: compact fld format with extents
758 Details : Store range of seq rather than every seq in FLD. Seq
759 controller update FLD rather than clients. In Case of CMD, mdt0
760 has FLD, all other metadata server act as non persistent proxy
761 for FLD queries and cache fld entries in fld cache.
766 Description: don't skip ost target if they assigned to file
767 Details : Drop slow OSCs if we can, but not for requested start idx.
768 This means "if OSC is slow and it is not the requested
769 start OST, then it can be skipped, otherwise skip it only
770 if it is inactive/recovering/out-of-space.
774 Description: more cleanup in mds_lov
775 Details : not send LOV EA under replay, we can't know about they size at this
776 time. Don't allow client connect to mds before any ost connected,
777 for avoid problems with LOV EA size and returning EIO to client.
779 Severity : enhancement
781 Description: Interoperability at server side (Disk interoperability)
783 Severity : enhancement
785 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
787 Severity : enhancement
789 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
791 Severity : enhancement
793 Description: New client IO stack (CLIO).
795 Severity : enhancement
797 Description: Commit on sharing. Eliminate inter-client dependencies between
798 uncommitted transactions by doing transaction commits.
799 Thereby clients may recovery independently.
802 Frequency : Create a symlink file with a very long name
804 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
805 Details : If there is no extra space in the request for early cancels,
806 ldlm_req_handles_avail() returns 0 instead of a negative value.
808 Severity : enhancement
810 Description: Add /proc entry for import status
811 Details : The mdc, osc, and mgc import directories now have
812 an import directory that contains useful import data for debugging
815 Severity : enhancement
817 Description: Re-disable certain /proc logging
818 Details : Enable and disable client's offset_stats, extents_stats and
819 extents_stats_per_process stats logging on the fly.
822 Frequency : Only on FC kernels 2.6.22+
824 Description: oops in statahead
825 Details : Do not drop reference count for the dentry from VFS when lookup,
826 VFS will do that by itself.
828 Severity : enhancement
830 Description: Generic /proc file permissions
831 Details : Set /Proc file permissions in a more generic way to enable non-
832 root users operate on some /proc files.
836 Description: Hitting mdc_commit_close() ASSERTION
837 Details : Properly handle request reference release in
838 ll_release_openhandle().
842 Description: quota recovery deadlock during mds failover
843 Details : This patch includes att18982, att18236, att18237 in bz14840.
845 1. fix osts hang when mds does failover with quotaon
846 2. prevent watchdog storm when osts threads wait for the
851 Frequency : only patchless client
852 Description: add workaround for race between add/remove dentry from hash
854 Severity : enhancement
856 Description: Allow OST glimpses to return PW locks
860 Description: LBUG when llog conf file is full
861 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
865 Description: Prevent import from entering FULL state when server in recovery
869 Description: service mount cannot take device name with ":"
870 Details : Only when device name contains ":/" will mount treat it as
876 Description: replace ptlrpcd with the statahead thread to interpret the async
877 statahead RPC callback
881 Frequency : on recovery
882 Description: I/O failures after umount during fail back
883 Details : if client reconnected to restarted server we need join to recovery
884 instead of find server handler is changed and process self
885 eviction with cancel all locks.
887 Severity : enhancement
889 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
891 Severity : enhancement
893 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
895 Severity : enhancement
897 Description: Upcall on Lustre log has been dumped
898 Details : Allow for a user mode script to be called once a Lustre log has
899 been dumped. It passes the filename of the dumped log to the
900 script, the location of the script can be specified via
901 /proc/sys/lnet/debug_log_upcall.
906 Description: avoid idr_remove called for id which is not allocated.
907 Details : Move assigment s_dev for clustered nfs to end of initialization,
908 to avoid problem with error handling.
913 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
914 Details : When connection is reused this not moved from CONN_UNUSED_HASH
915 into CONN_USED_HASH and this prodice warning when put connection
916 again in unused hash.
919 Severity : enhancement
921 Description: File striping can now be set to use an arbitrary pool of OSTs.
923 Severity : enhancement
925 Description: Export bytes_read/bytes_write count on OSC/OST.
929 Description: Early reply size mismatch, MGC loses connection
930 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
931 the connect flags are properly negotiated.
936 Description: quotas are not honored with O_DIRECT
937 Details : all writes with the flag O_DIRECT will use grants which leads to
938 this problem. Now using OBD_BRW_SYNC to guard this.
942 Description: add quota statistics
943 Details : 1. sort out quota proc entries and proc code.
944 2. add quota statistics
946 Severity : enhancement
948 Description: enable quota support for HEAD.
952 Description: Properly propagate oinfo flags from lov to osc for statfs
953 Details : restore missing copy oi_flags to lov requests.
955 Severity : enhancement
957 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
958 : l_getgroups(8), lst(8), routerstat(8)
959 Details : included man pages for llobdstat(8), llstat(8),
960 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
962 Severity : enhancement
964 Description: configure's --enable-quota should check the
965 : kernel .config for CONFIG_QUOTA
966 Details : configure is terminated if --enable-quota is passed but
967 : no quota support is in kernel
971 Description: Remove portals compatibility
972 Details : Remove portals compatibility, not interoperable with releases
977 Description: Resolve device initialization race
978 Details : Prevent proc handler from accessing devices added to the
979 obd_devs array but yet be intialized.
981 Severity : enhancement
983 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
985 Severity : enhancement
987 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
992 Description: Using wrong pointer in osc_brw_prep_request
993 Details : Access to array[-1] can produce panic if kernel compiled with
994 CONFIG_PAGE_ALLOC enabled
998 Description: Client runs out of low memory
999 Details : Consider only lowmem when counting initial number of llap pages
1003 Description: *optional* service tags registration
1004 Details : if the "service tags" package is installed on a Lustre node
1005 When the filesystem is mounted, a local-node service tag will
1006 be created. See http://inventory.sun.com/ for more information
1007 about the Service Tags asset management system.
1011 Description: Kernel BUG tries to release flock
1012 Details : Lustre does not destroy flock lock before last reference goes
1013 away. So always drop flock locks when client is evicted and
1014 perform unlock regardless of successfulness of speaking to MDS.
1018 Description: add refcount protection for osc callbacks, avoid panic on shutdown
1022 Description: sanity test 65a fails if stripecount of -1 is set
1023 Details : handle -1 striping on filesystem in ll_dirstripe_verify
1028 Description: ASSERTION(CheckWriteback(page,cmd)) failed
1029 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
1032 Severity : enhancement
1034 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
1038 Description: do not process already freed flock
1039 Details : flock can possibly be freed by another thread before it reaches
1040 to ldlm_flock_completion_ast.
1044 Description: LBUG during stress test
1045 Details : Need properly lock accesses the flock deadlock detection list.
1049 Description: oops in page fault handler
1050 Details : kernel page fault handler can return two special 'pages' in error
1051 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
1055 Description: timeout with invalidate import.
1056 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
1057 should be handled by ptlrpcd. This produce long age waiting and
1058 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
1060 Severity : enhancement
1062 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
1066 Description: enable MGS and MDT services start separately
1067 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
1068 a co-located MGS without starting the MGS, which is a complement
1069 to 'nosvc' mount option.
1073 Frequency : after recovery
1074 Description: precreate to many object's after del orphan.
1075 Details : del orphan st in oscc last_id == next_id and this triger growing
1076 count of precreated objects. Set flag LOW to skip increase count
1077 of precreated objects.
1081 Frequency : rare, on clear nid stats
1082 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
1083 Details : when clean nid stats sometimes try destroy live entry,
1084 and this produce panic in free.
1088 Description: Stack overflow during MDS log replay
1089 ease stack pressure by using a thread dealing llog_process.
1093 Description: wait until IO finished before start new when do lock cancel.
1094 Details : VM protocol want old IO finished before start new, in this case
1095 need wait until PG_writeback is cleared until check dirty flag and
1096 call writepages in lock cancel callback.
1098 Severity : enhancement
1100 Description: using special macro for print time and cleanup in includes.
1104 Description: mds_mfd_close() ASSERTION(rc == 0)
1105 Details : In mds_mfd_close(), we need protect inode's writecount change
1106 within its orphan write semaphore to prevent possible races.
1110 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
1114 Frequency : rare, on shutdown ost
1115 Description: don't hit live lock with umount ost.
1116 Details : shrink_dcache_parent can be in long loop with destroy dentries,
1117 use shrink_dcache_sb instead.
1121 Description: don't panic with use echo client
1122 Details : echo client pass NULL as client nid pointer and this produce null
1123 pointer dereference.
1127 Description: fix build on ppc32
1128 Details : compile code with -m64 flag produce wrong object file for ppc32.
1132 Description: add message levels for liblustreapi
1136 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
1137 Details : Move the 'good_osts' check before the 'total_bavail' check. This
1138 will result in an -EAGAIN and in the exit call path we call
1139 alloc_rr() which will with increasing aggressiveness attempt to
1140 aquire precreated objects on the minimum number of required OSCs.
1144 Description: Use old size assignment to avoid deadlock
1145 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
1146 the scheduling while holding a spinlock. We do not need locking
1147 for size in ll_update_inode() because size is only updated from
1148 the MDS for directories or files without objects, so there is no
1149 other place to do the update, and concurrent access to such inodes
1150 are protected by the inode lock.
1154 Description: resolve "_IOWR redefined" build error on SLES10
1158 Description: dump the memory debugging after all modules are unloaded to
1159 suppress false negative in conf_sanity test 39
1161 Severity : enhancement
1163 Description: build kernel-ib packages for OFED 1.3 in our release cycle
1168 Description: fix SLES kernel versioning
1169 Details : the kernel version for our SLES 10 kernel did not include a "-"
1170 before the "smp" at the end. while this was not a problem in
1171 general, it did mean that software trying to use the kernel
1172 version to try to detect a vendor specific kernel would fail.
1173 this was most evident by the OFED build scripts.
1177 Description: Don't update lov_desc members until making sure they are valid
1178 Details : When updating lov_desc members via proc fs, need fix their
1179 validities before doing the real update.
1183 Description: don't put request into delay list while invalidate in flight.
1184 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
1185 import in flight. this produce timeout for invalidate and sometimes
1186 can cause stale data.
1190 Frequency : on ppc only
1191 Description: not convert ost objects for directory because it's not exist.
1192 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
1196 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
1198 Severity : enhancement
1200 Description: Update to support 2.6.22.14 vanilla kernel.
1204 Frequency : rare, on recovery
1205 Description: read procfs can produce deadlock in some situation
1206 Details : Holding lprocfs lock which send rpc can produce block for destroy
1207 obd objects and this also block reconnect with -EALREADY. This
1208 isn't fix all lprocfs bugs - but make it rare.
1210 Severity : enhancement
1212 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
1215 Frequency : frequent on X2 node
1217 Description: mdc_set_open_replay_data LBUG
1218 Details : Set replay data for requests that are eligible for replay.
1222 Description: lustre_mgs: operation 101 on unconnected MGS
1223 Details : When MGC is disconnected from MGS long enough, MGS will evict the
1224 MGC, and late on MGC cannot successfully connect to MGS and a lot
1225 of the error messages complaining that MGS is not connected.
1229 Frequency : on network error
1230 Description: panic with double free request if network error
1231 Details : mdc_finish_enqueue is finish request if any network error ocuring,
1232 but it's true only for synchronus enqueue, for async enqueue
1233 (via ptlrpcd) this incorrect and ptlrpcd want finish request
1236 Severity : enhancement
1238 Description: client-side metadata stat-ahead during readdir(directory readahead)
1239 Details : perform client-side metadata stat-ahead when the client detects
1240 readdir and sequential stat of dir entries therein
1243 Frequency : on start mds
1245 Description: Implement get_info(last_id) in obdfilter.
1248 Frequency : occasional
1250 Description: Correctly check stale fid, not start epoch if ost not support SOM
1251 Details : open with flag O_CREATE need set old fid in op_fid3 because
1252 op_fid2 was overwritten with new generated fid, but mds can answer
1253 with one of these two fids and both is not stale. Setattr
1254 incorrectly started an epoch and assume will be called
1255 done_writing, but without SOM done_writing ever being called.
1258 Frequency : rare, depends on device drivers and load
1260 Description: MDS or OSS nodes crash due to stack overflow
1261 Details : Code changes in 1.8.0 increased the stack usage of some functions.
1262 In some cases, in conjunction with device drivers that use a lot
1263 of stack the MDS (or possibly OSS) service threads could overflow
1264 the stack. One change which was identified to consume additional
1265 stack has been reworked to avoid the extra stack usage.
1268 Frequency : occasional
1270 Description: Do not fail import if osc_interpret_create gets -EAGAIN
1271 Details : If osc_interpret_create got -EAGAIN it immediately exits and
1272 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
1273 oscc_has_objects and see OSC has no objests and call
1274 oscc_internal_create to resend create request.
1276 Severity : enhancement
1278 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
1280 Severity : enhancement
1282 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
1285 Frequency : very rare
1287 Description: Fix replay if there is an un-replied request and open
1288 Details : In some cases, older replay request will revert the
1289 mcd->mcd_last_xid on MDS which is used to record the client's
1290 latest sent request.
1292 Severity : enhancement
1294 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
1296 Severity : enhancement
1298 Description: Add rhel5 support to HEAD.
1300 Serverity : enhancement
1302 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
1307 Description: Don't allow skipping OSTs if index has been specified.
1308 Details : Don't allow skipping OSTs if index has been specified, make
1309 locking in internal create lots better.
1313 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
1314 Details : access to bitfield in structure is always rounded to long
1315 and this produce problem with not atomic change any bit.
1319 Description: Lustre make rpms failed.
1320 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
1321 builds Lustre rpms from tarball.
1324 Frequency : rare on shutdown ost
1326 Description: If llog cancel was not send before clean_exports phase, this can
1327 produce deadlock in llog code.
1328 Details : If llog thread has last reference to obd and call class_import_put
1329 this produce deadlock because llog_cleanup_commit_master wait when
1330 last llog_commit_thread exited, but this never success because was
1331 called from llog_commit_thread.
1335 Description: allow userland application know is lost one of stripes.
1336 Details : fill lvb_blocks with error code on ost and return it to
1337 application if error flag found.
1341 Description: NULL lov_tgts causing MDS oops
1342 Details : more safe checks for NULL lov_tgts for avoid oops.
1344 Severity : enhancement
1346 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
1350 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
1351 Details : If an OST is down the MDS will hang indefinitely in
1352 obd_statfs_rqset() waiting for the statfs data. While for
1353 MDS QOS usage of statfs, it should not stuck in waiting.
1355 Severity : enhancement
1357 Description: remote_acl support
1358 Details : Support ACL-based permission check for remote user.
1359 Support setfacl/getfacl for remote user with the utils
1360 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
1361 the system "{s,g}etfacl" utils.
1363 Severity : enhancement
1365 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
1367 Severity : enhancement
1369 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
1373 Description: make mgs_setparam() handle fsname containing dash
1374 Details : fsname containing a dash does not work with lctl conf_param
1376 Severity : enhancement
1378 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
1380 Severity : enhancement
1382 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
1386 Frequency : rare, at shutdown
1387 Description: access already free / zero obd_namespace.
1388 Details : if client_disconnect_export was called without force flag set,
1389 and exist connect request in flight, this can produce access to
1390 NULL pointer (or already free pointer) when connect_interpret
1391 store ocd flags in obd_namespace.
1395 Frequency : only at startup
1396 Description: not alloc memory with spinlock held.
1397 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
1398 if any spinlock held.
1400 Severity : enhancement
1402 Description: make lustre randomly fail allocating memory
1403 Details : Make lustre randomly failed allocating memory for testing purpose.
1405 Severity : enhancement
1407 Description: lost problems with lov objid file
1408 Details : Fixes some scability and access to not inited memory problems
1409 in work with lov objdid file.
1414 Description: lfs find does not continue on file error
1415 Details : Continue other files processing when a file/dir is absent.
1419 Description: Inconsistent usage of lustre_pack_reply()
1420 Details : Standardize the usage of lustre_pack_reply() such that it
1421 always generate a CERROR on failure.
1424 Frequency : occasional
1426 Description: MDS hang or stay in waiting lock
1427 Details : If client receive lock with CBPENDING flag ldlm need send lock
1428 cancel as separate rpc, to avoid situation when cancel request
1429 can't processed due all i/o threads stay in wait lock.
1433 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1437 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1441 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1445 Description: Update to RHEL4 latest kernel.
1447 Severity : enhancement
1449 Description: Build SLES10 patchless client fails
1450 Details : The configure was broken by run ./configure with
1451 --with-linux-obj=.... argument for patchless client. When the
1452 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1453 can't search header adequately. Use absolute path such as
1454 -I($LINUX)/include instead.
1458 Description: interrupt oig_wait produce painc on resend.
1459 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1460 private set, and this produce situation when rq_ptlrpcd_data not
1461 copyed to new allocated request and triggered LBUG on assert
1462 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1463 ptlrpcd set if request is changed and can be safety changed to use
1468 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1469 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1470 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1471 these macros and replace them with their difinitions instead.
1475 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1476 Details : Update which_patch & target file for SLES10 latest kernel.
1479 Description: Remove client patches from SLES 10 kernel.
1480 Details : This causes SLES 10 clients to behave as patchless clients
1481 even on a Lustre-patched (server) kernel.
1483 Severity : enhancement
1485 Description: self-adjustable client's lru lists
1486 Details : use adaptive algorithm for managing client cached locks lru
1487 lists according to current server load, other client's work
1488 pattern, memory activities, etc. Both, server and client
1489 side namespaces provide number of proc tunables for controlling
1492 Severity : enhancement
1494 Description: light-weight GSS support
1495 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
1496 and reduce performance overhead.
1498 Severity : enhancement
1500 Description: Linux keyring support
1501 Details : Support using service of Linux keyring for Lustre GSS internal
1502 context refresh/cache mechanism.
1506 Description: Fix errors in lfs documentation
1507 Details : Fixes man pages
1511 Description: don't use GFP_* in generic Lustre code.
1512 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
1516 Description: obdclass is limited by single OBD_ALLOC(idarray)
1517 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
1521 Description: warnings with build patchless client with vanila 2.6.19 and up
1522 Details : change old ctl_table style and replace ctl_table/ctl_table_header
1523 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
1527 Frequency : Only for SLES
1528 Description: sanity_quota fail test_1
1529 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
1530 file, which makes TSTID[2] inunique.
1534 Frequency : Only for SLES9
1535 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
1536 Details : Update target/ChangeLog/which_patch .
1540 Frequency : Only for test_10 in sanity-quota.sh
1541 Description: a bug in quota test code
1542 Details : checking a test flag in a wrong place
1546 Frequency : Only for test_13 in sanity-quota.sh
1547 Description: a bug in quota test script
1548 Details : shouldn't check overfull allocated quota
1552 Description: jbd statistics
1553 Details : Port older jbd statistics patch for sles10
1555 Severity : enhancement
1557 Description: Add printing inode info into message about error in writepage.
1562 Description: reply_lock_interpret crash due to race with it and lock cancel.
1563 Details : Do not replay locks that are being cancelled. Do not reference
1564 locks by their address during replay, just by their handle.
1568 Frequency : When flocks are used.
1569 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
1570 Details : Flock locks might destroy just granted lock if it could be merged
1571 with another existing flock, this is done in completion handler,
1572 so teach ldlm_cli_enquque_fini that this is a valid case for
1578 Description: Oops in read and write path when failing to allocate lock.
1579 Details : Check if lock allocation failed and return error back.
1583 Description: lstripe command fails for valid OST index
1584 Details : The stripe offset is compared to lov->desc.ld_tgt_count
1585 instead of lov->desc.ld_active_tgt_count.
1589 Description: sanity.sh failed test 103
1590 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
1595 Description: df doesn't work properly if diskfs blocksize != 4K
1596 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
1600 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
1601 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
1603 Severity : enhancement
1605 Description: Adaptive timeouts
1606 Details : RPC timeouts adapt to changing server load and network
1607 conditions to reduce resend attempts and improve recovery time.
1611 Description: llapi_file_create() does not allow some changes
1612 Details : add llapi_file_open() that allows specifying the mode and
1613 open flags, and also returns an open file handle.
1617 Description: merge and cleanup kernel patches.
1618 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
1622 Description: Add journal checksum support.(Kernel part)
1623 Details : The journal checksum feature adds two new flags i.e
1624 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
1625 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
1626 indicates that the commit block contains the checksum for
1627 the blocks described by the descriptor blocks. Now commit
1628 record can be sent to disk without waiting for descriptor
1629 blocks to be written to disk. This behavior is controlled
1630 using JBD2_FEATURE_ASYNC_COMMIT flag.
1634 Description: OSS needs mutliple precreate threads
1635 Details : Add ability to start more than one create thread per OSS.
1639 Description: Sanity.sh test_65k failed: setstripe should have succeeded
1640 Details : execute lfs setstripe on client
1644 Description: mds_obd_create error creating tmp object
1645 Details : When the user sets quota on root, llog will be affected and can't
1646 create files and write files.
1649 Frequency : Always on ia64 patchless client, and possibly others.
1651 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
1652 Details : This allows the patchless client to be loaded on architectures
1653 without this export.
1657 Description: RedHat Update kernel for RHEL5
1658 Details : Add the kernel config file for RHEL5.
1662 Description: RedHat Update kernel for RHEL5
1663 Details : Modify the kernel config file more closer RHEL5.
1667 Description: Build failure against Centos5 (RHEL5)
1668 Details : Define PAGE_SIZE when it isn't present.
1672 Description: test_7 on sanity-quota.sh can't be used on separate machines
1673 Details : add facet to handle it
1677 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
1678 Details : with NFS, the anon dentry's parent was set to itself in
1679 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
1680 corresponding dentry other than use rec->ur_name.
1682 Severity : enhancement
1684 Description: lfs setstripe enhancement
1685 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
1689 Description: enable data checksumming by default
1690 Details : enable checksum by default, allow --disable-checksum
1691 configure option and "-o nochecksum" mount option. Checksums
1692 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
1696 Description: lustre support for RHEL5
1697 Details : Add support for RHEL5.
1701 Description: Client eviction due to failover config
1702 Details : after a connection loss, the lustre client should attempt to
1703 reconnect to the last active server first before trying the
1704 other potential connections.
1708 Description: when mds and osts use different quota unit(32bit and 64bit),
1709 quota will be released repeatly.
1710 Details : void sending multiple quota reqs to mds, which will keep the
1711 status between the reqs.
1715 Description: osts not allocated evenly to files
1716 Details : change the condition to increase offset_idx
1719 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
1721 Description: Data corruption for OSTs that are formatted larger than 2TB
1723 Details : When generating the bio request for lustre file writes the
1724 sector number would overflow a temporary variable before being
1725 used for the IO. The data reads correctly from Lustre (which
1726 will overflow in a similar manner) but other file data or
1727 filesystem metadata may be corrupted in some cases.
1731 Description: Tune the kernel for good SCSI performance.
1732 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
1733 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
1738 Description: rewrite ext2-derived code in obdclass/uuid.c
1739 Details : rewrite inherited code (uuid parsing code from ext2 utils)
1740 from scratch preserving functionality.
1744 Description: "lfs find -obd UUID" prints directories
1745 Details : "lfs find -obd UUID" will return all directory names instead
1746 of just file names. It is incorrect because the directories
1747 do not reside on the OSTs.
1751 Description: use i_size_read and i_size_write in 2.6 port
1752 Details : replace inode->i_size access with i_size_read/write()
1755 Frequency : when removing large files
1757 Description: scheduling issue during removal of large Lustre files
1758 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
1759 It causes scheduling issues when removing large files (17TB in the
1763 Frequency : only with liblustre clients on XT3
1765 Description: evictions taking too long
1766 Details : allow llrd to evict clients directly on OSTs
1769 Frequency : only on ppc
1771 Description: /proc/fs/lustre/devices broken on ppc
1772 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
1773 We should make sure the type of 'index' is loff_t and then cast
1774 explicitly as needed below. Do not assign an explicitly cast
1778 Frequency : only for rhel5
1780 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
1781 Details : Modify the target file & which_kernel.
1783 Severity : enhancement
1785 Description: omit set fsid for export NFS
1786 Details : fix set/restore device id for avoid EMFILE error and mark lustre
1787 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
1791 Frequency : Always, for kernels after 2.6.16
1792 Description: Fix warning idr_remove called for id=.. which is not allocated.
1793 Details : Last kernels save old s_dev before kill super and not allow
1794 to restore from callback - restore it before call kill_anon_super.
1798 Description: buffer overruns could theoretically occur
1799 Details : llapi_semantic_traverse() modifies the "path" argument by
1800 appending values to the end of the origin string, and a
1801 overrun may occur. Adding buffer overrun check in liblustreapi.
1805 Description: Fix error on 'ls .' at the top of the Lustre mount.
1806 Details : Don't revalidate dentry if it is a root dentry.
1810 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
1811 Details : Modify vm-tunables-rhel4.patch.
1815 Description: Kernel config for 2.6.18-vanilla.
1816 Details : Modify targets/2.6-vanilla.target.in.
1817 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
1818 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
1819 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
1820 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
1824 Description: improve handling recoverable errors
1825 Details : if request processig with error which can be recoverable on server
1826 request should be resend, otherwise page released from cache and
1831 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
1832 Details : Modify target file & which_patch.
1833 A flaw was found in the IA32 system call emulation provided
1834 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
1835 value could be stored in the %RAX register, which could trigger an
1836 out-of-bounds system call table access. An untrusted local user
1837 could exploit this flaw to run code in the kernel
1838 (ie a root privilege escalation). (CVE-2007-4573).
1842 Description: O_DIRECT bypasses client statistics.
1843 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
1844 read_ahead_stats were updated. copied the stats section from
1845 osc_send_oap_rpc() into async_internal().
1849 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
1853 Description: change order of libsysio includes
1854 Details : '#include sysio.h' should always come before '#include xtio.h'
1858 Description: handle "serious error: objid * already exists" more gracefully
1859 Details : If LAST_ID value on disk is smaller than the objects existing in
1860 the O/0/d* directories, it indicates disk corruption and causes an
1861 LBUG(). If the object is 0-length, then we should use the existing
1862 object. This will help to avoid a full fsck in most cases.
1864 Severity : enhancement
1866 Description: adapt the lustre_config script to support the upgrade case
1867 Details : Add "-u" option for lustre_config script to support upgrading 1.4
1868 server targets to 1.6 in parallel.
1872 Description: To avoid grant space > avaible space when the disk is almost
1873 full. Without this patch you might see the error "grant XXXX >
1874 available" or some LBUG about grant, when the disk is almost
1876 Details : In filter_check_grant, for non_grant cache write, we should
1877 check the left space by if (*left > ungranted + bytes), instead
1878 of (*left > ungranted), because only we are sure the left space
1879 is enough for another "bytes", then the ungrant space should be
1880 increase. In client, we should update cl_avail_grant only there
1881 is OBD_MD_FLGRANT in the reply.
1885 Description: Update RHEL 4 kernel to fix local root privilege escalation.
1886 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
1887 described in CVE-2007-4573. This problem could allow untrusted
1888 local users to gain root access.
1891 Frequency : when using O_DIRECT and quotas
1893 Description: Incorrect file ownership on O_DIRECT output files
1894 Details : block usage reported by 'lfs quota' does not take into account
1895 files that have been written with O_DIRECT.
1900 Description: touch file failed when fs is not full
1901 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
1902 otherwise we can get ENOSP while fs is not full.
1906 Description: parallel lock callbacks
1907 Details : Instead of sending blocking and completion callbacks as separated
1908 requests, adding them to a set and sending in parallel.
1911 Frequency : only for Cray XT3
1912 Bugzilla : 12829/13455
1913 Description: Changing primary group doesn't change the group lustre assigns to
1915 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
1916 overridden with the primary group provided by the group upcall,
1917 whereas the supplied fsgid can be trusted if it is in the list of
1918 supplementary groups returned by the group upcall.
1920 Severity : enhancement
1922 Description: Allow masking D_WARNING, D_ERROR messages from console
1923 Details : Console messages can now be disabled via lnet.printk.
1927 Description: User code with malformed file open parameter crashes client node
1928 Details : Before packing join_file req, all the related reference should be
1929 checked carefully in case some malformed flags cause fake
1930 join_file req on client.
1934 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
1935 Details : ldlm_completion_ast() assumes that a lock is granted when the req
1936 mode is equal to the granted mode. However, it should also check
1937 that LDLM_FL_CP_REQD is not set.
1941 Description: Heavy nfs access might result in deadlocks
1942 Details : After ELC code landed, it is now improper to enqueue any mds
1943 locks under och_sem, because enqueue might want to decide to
1944 cancel open locks for same inode we are holding och_sem for.
1948 Description: Client eviction while running blogbench
1949 Details : A lot of unlink operations with concurrent I/O can lead to a
1950 deadlock causing evictions. To address the problem, the number of
1951 oustanding OST_DESTROY requests is now throttled to
1952 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
1953 callbacks are processed in priority.
1957 Description: enable ACLs on MDS by default
1958 Details : ACLs must be enabled on MDS by default.
1961 Frequency : PPC/PPC64 only
1963 Description: conflicts between asm-ppc64/types.h and lustre_types.h
1964 Details : fix duplicated definitions between asm-ppc64/types.h and
1965 lustre_types.h on PPC.
1968 Frequency : PPC/PPC64 only
1970 Description: asm-ppc/segment.h does not exist
1971 Details : fix compile issue on PPC.
1975 Description: better handle error messages in extents code
1978 Frequency : RHEL4 only
1980 Description: mkfs is very slow on IA64/RHEL4
1981 Details : A performance regression has been discovered in the MPT Fusion
1982 driver between versions 3.02.73rh and 3.02.99.00rh. As a
1983 consequence, we have downgraded the MPT Fusion driver in the RHEL4
1984 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
1986 Severity : enhancement
1988 Description: SNMP support enhancement
1989 Details : Adding total number of sampled request for an MDS node in snmp
1992 Severity : enhancement
1994 Description: Optimize ldlm waiting list processing for PR extent locks
1995 Details : When processing waiting list for read extent lock and meeting read
1996 lock that is same or wider to it that is not contended, skip
1997 processing rest of the list and immediatelly return current
1998 status of conflictness, since we are guaranteed there are no
1999 conflicting locks in the rest of the list.
2003 Description: Time out and refuse to reconnect
2004 Details : When the failover node is the primary node, it is possible
2005 to have two identical connections in imp_conn_list. We must
2006 compare not conn's pointers but NIDs, otherwise we can defeat
2007 connection throttling.
2011 Description: port llog fixes from b1_6 into HEAD
2012 Details : Port llog reference couting and some llog cleanups from b1_6
2013 (bug 10800) into HEAD, for protect from panic and access to already
2014 free llog structures.
2018 Description: Detect stride IO mode in read-ahead
2019 Details : When a client does stride read, read-ahead should detect that and
2020 read-ahead pages according to the detected stride pattern.
2024 Description: data checksumming impacts single node performance
2025 Details : add support for several checksum algorithm. Currently, only CRC32
2026 and Adler-32 are supported. The checksum type can be changed on
2027 the fly via /proc/fs/lustre/osc/*/checksum_type.
2031 Description: use adler32 for page checksums
2032 Details : when available, use the Adler-32 algorithm instead of CRC32 for
2037 Description: build for x2 fails
2038 Details : fix compile issue on Cray systems.
2042 Description: Properly match for duplicate locks
2043 Details : Due to different lock order from skiplists code, we need to
2044 traverse entire list for now
2047 Frequency : only on PPC/SLES10
2049 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
2050 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
2051 be defined. Add a hack in mkfs_lustre.c to work around this compile
2056 Description: LASSERT on MDS when client holding flock lock dies
2057 Details : ldlm pool logic depends on number of granted locks equal to
2058 number of released locks which is not true for flock locks, so
2059 just exclude such locks from consideration.
2063 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
2064 Details : Use fsfilt_sync() for both the whole filesystem sync and
2065 individual file sync to eliminate dangerous inode locking
2066 with I_LOCK that can lead to a deadlock.
2070 Description: performance in 1.6.3
2071 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
2072 sure that 1MB requests are not fragmented by the block layer.
2074 Severity : enhancement
2076 Description: organize the server-side client stats on per-nid basis
2077 Details : Change the structure of stats under obdfilter and mds to
2085 The "uuid"s file would list the uuids of _active_ exports.
2086 And the clear entry is to clear all stats and stale nids.
2088 Severity : enhancement
2090 Description: eliminate client locks in face of contention
2091 Details : file contention detection and lockless i/o implementation
2092 for contended files.
2096 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
2099 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
2101 Description: MDS may lose file striping (and hence file data) in some cases
2102 Details : If there are additional extended attributes stored on the MDS,
2103 in particular ACLs, SELinux, or user attributes (if user_xattr
2104 is specified for the client mount options) then there is a risk
2105 of attribute loss. Additionally, the Lustre file striping
2106 needs to be larger than default (e.g. striped over all OSTs),
2107 and an additional attribute must be stored initially in the
2108 inode and then increase in size enough to be moved to the
2109 external attribute block (e.g. ACL growing in size) for file
2114 Description: skiplist implementation simplification
2115 Details : skiplists are used to group compatible locks on granted list
2116 that was implemented as tracking first and last lock of each
2117 lock group the patch changes that to using doubly linked lists
2121 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
2122 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
2123 because LASSERT on both the data supplied by a client, and the
2124 data on disk is dangerous and incorrect.
2126 Severity : enhancement
2128 Description: Slow truncate/writes to huge files at high offsets.
2129 Details : Directly associate cached pages to lock that protect those pages,
2130 this allows us to quickly find what pages to write and remove
2131 once lock callback is received.
2135 Description: more ldlm soft lockups
2136 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
2137 starve other threads from the resource lock for a long time in
2138 case of long waiting queue, so change the debug level from
2139 D_OTHER to the less frequently used D_INFO.
2141 Severity : enhancement
2143 Description: add -gid, -group, -uid, -user options to lfs find
2147 Description: Hung threads in invalidate_inode_pages2_range
2148 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
2149 once one is completed. As a result, some RPCs are stuck in the
2150 queue and are never sent.
2154 Description: filter threads hungs on waiting journal commit
2155 Details : Cleanup filter group llog code, then only filter group llog will
2156 be only created in the MDS/OST syncing process.
2160 Description: Procfs and llog threads access destoryed import sometimes.
2161 Details : Sync the import destoryed process with procfs and llog threads by
2162 the import refcount and semaphore.
2164 Severity : enhancement
2166 Description: openlock cache of b1_6 port to HEAD
2171 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
2172 Details : If insertion of an extent fails, then discard the inode
2173 preallocation and free data blocks else it can lead to duplicate
2178 Description: don't always update ctime in ext3_xattr_set_handle()
2179 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
2180 In some cases the ctime should not be updated, for example for
2181 2.0->1.8 compatibility it is necessary to delete an xattr and it
2182 should not update the ctime.
2186 Bugzilla : 15713/16362
2187 Description: Assertion in iopen_connect_dentry in 1.6.3
2188 Details : looking up an inode via iopen with the wrong generation number can
2189 populate the dcache with a disconneced dentry while the inode
2190 number is in the process of being reallocated. This causes an
2191 assertion failure in iopen since the inode's dentry list contains
2192 both a connected and disconnected dentry.
2196 Description: assertion failure in ldlm_handle2lock()
2197 Details : fix a race between class_handle_unhash() and class_handle2object()
2198 introduced in lustre 1.6.5 by bug 13622.
2203 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
2204 Details : remove the limit of 256 scsi disks in the sd_iostat patch
2209 Description: oops in sd_iostats_seq_show()
2210 Details : unloading/reloading the scsi low level driver triggers a kernel
2211 bug when trying to access the sd iostat file.
2216 Description: Kernel panics during QLogic driver reload
2217 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
2218 patch, causing memory corruption.
2223 Description: journal_dev option does not work in b1_6
2224 Details : pass mount option during pre-mount.
2226 Severity : enhancement
2228 Description: Add a FIEMAP(FIle Extent MAP) ioctl
2229 Details : FIEMAP ioctl will allow an application to efficiently fetch the
2230 extent information of a file. It can be used to map logical blocks
2231 in a file to physical blocks in the block device.
2235 Description: LDLM soft lockups - improvement
2236 Details : It is be possible to send the lock handle along with each read
2237 or write request because the client is already doing a lock match
2238 itself so there isn't any reason the OST should have to re-do that
2242 Frequency : only on Cray X2
2244 Description: X2 build failures
2245 Details : fix build failures on Cray X2.
2249 Description: xid & resent requests
2250 Details : Initialize RPC XID from clock at startup (randomly if clock is
2253 Severity : enhancement
2255 Description: Add lustre_start utility to start or stop multiple Lustre servers
2260 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
2261 Details : In case of memory pressure, list_del() can be called twice on
2262 req->rq_history_list, causing a kernel oops.
2266 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
2267 Details : fix stack overflow in the distributed lock manager by defering
2268 export eviction after a failed AST to the elt thread instead of
2269 handling it in the dlm interpret routine.
2273 Description: Convert some comments to new format.
2274 Details : Update documenting comments to match doxygen conventions.
2278 Description: Grammar fixes.
2279 Details : A couple of trivial spelling fixes.
2283 Description: OSD_COUNTERS-mandatory
2284 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
2285 to simplify the code.
2289 Description: simplify cmm_device freeing logic.
2290 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
2294 Description: Add lockdep support to dt_object_operations locking interface.
2295 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
2296 parameter indicating lock ordering. Update mdd code to use new
2301 Description: Introduce failloc constants for lockless IO tests.
2302 Details : Add two new failloc constants to test lockless IO. Only one of
2303 them in implemented---another is checked in yet to be landed
2308 Description: Add lockdep support for inode mutex.
2309 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
2310 in the situations where Lustre has to lock more than one inode
2315 Description: Add optional invariants checking support.
2316 Details : Add new LINVRNT() macro, optional on new --enable-invariants
2317 configure switch. This macro is to be used for consistency and
2318 sanity checks that are too expensive to be left in `production'
2323 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
2324 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
2329 Description: Add lockdep support for ldlm_lock and ldlm_resource.
2330 Details : Use spin_lock_nested() in (the only) situation where more than
2331 one ldlm_lock is locked simultaneously. Also, fix possible
2332 dead-lock in ldlm_lock_change_resource() by enforcing particular
2337 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
2338 Details : Instead of specifying each ldlm_lock call-back through separate
2339 parameter, wrap them into struct ldlm_callback_suite.
2343 Description: Kill join_lru obd method and its callers.
2344 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
2345 in memory---a requirement implemented through ->o_join_lru() obd
2346 method in HEAD. Remove this method and its users.
2350 Description: Add asynchronous ldlm ENQUEUE completion handler.
2351 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
2352 case that stock ldlm_completion_ast() cannot handle as it waits
2353 until lock is granted. Introduce new ldlm_completion_ast_async()
2354 for this. Also comment ldlm_completion_ast().
2358 Description: ldlm_error <-> errno conversion.
2359 Details : Add functions to map (rather arbitrary) between LDLM error codes
2360 and standard errno values. CLIO needs this to prevent LDLM specific
2361 constants from escaping ldlm and osc.
2365 Description: Kill unused ldlm_handle2lock_ns() function.
2366 Details : Kill unused ldlm_handle2lock_ns() function.
2370 Description: Add lu_ref support to ldlm_lock
2371 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
2372 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
2373 and ldlm_resource. LDLM interface has to be changed, because code that
2374 releases a reference on a lock, has to "know" what reference this is.
2375 In the most frequent case
2377 lock = ldlm_handle2lock(handle);
2379 LDLM_LOCK_PUT(lock);
2381 no changes are required. When any other reference (received _not_ from
2382 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
2383 instead of LDLM_LOCK_PUT().
2385 Arguably, changes are pervasive, and interface requires some discipline
2386 for proper use. On the other hand, it was very instrumental in finding
2387 a few leaked lock references.
2391 Description: Add ldlm_lock_addref_try().
2392 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
2393 attempts to addref a lock that might be being canceled
2398 Description: Add ldlm_weigh_callback().
2399 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
2400 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
2401 hard-coded `number of pages' logic.
2405 Description: Add lockdep annotations to llog code.
2406 Details : Use appropriately tagged _nested() locking calls in the places
2407 where llog takes more than one ->lgh_lock lock.
2411 Description: Add loi_kms_set().
2412 Details : Wrap kms updates into a helper function.
2416 Description: Constify instances of struct lsm_operations.
2417 Details : Constify instances of struct lsm_operations.
2421 Description: lu_conf support.
2422 Details : On a server, a file system object is uniquely identified
2423 by a fid, which is sufficient to locate and load all object
2424 state (inode). On a client, on the other hand, more data are
2425 necessary instantiate an object. Change lu_object_find() and
2426 friends to take additional `lu_conf' argument describing object.
2427 Typically this includes layout information.
2431 Description: lu_context fixes.
2432 Details : Introduce new lu_context functions that are needed on the client
2433 side, where some system threads (ptlrpcd) are shared by multiple
2434 modules, and so cannot be stopped during module shutdown.
2438 Description: Add start and stop methods to lu_device_type_operations.
2439 Details : Introduce two new methods in lu_device_type_operations, that are
2440 invoked when first instance of a given type is created and last one
2441 is destroyed respectively. This is need by CLIO.
2445 Description: Add lu_ref support to struct lu_device.
2446 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
2447 track leaked references.
2451 Description: Introduce lu_kmem_descr.
2452 Details : lu_kmem_descr and its companion interface allow to create
2453 and destroy a number of kmem caches at once.
2457 Description: Fix lu_object finalization race.
2458 Details : Fix a race between lu_object_find() finding an object and its
2459 concurrent finalization. This race is (most likely) not possible
2460 on the server, but might happen on the client.
2464 Description: Introduce lu_ref interface.
2465 Details : lu_ref is a debugging module allowing to track references to
2466 a given object. It is quite cpu expensive, and has to be
2467 explicitly enabled with --enable-lu_ref. See usage description
2472 Description: Factor lu_site procfs stats into a separate function.
2473 Details : Separate lu_site stats printing code into a separate function
2474 that can be reused on a client.
2478 Description: Constify instances of struct {lu,dt,md}_device_operations.
2479 Details : Constify instances of struct {lu,dt,md}_device_operations.
2483 Description: Introduce struct md_site and move meta-data specific parts of
2484 struct lu_site here.
2485 Details : Move md-specific fields out of struct lu_site into special struct
2486 md_site, so that lu_site can be used on a client.
2490 Description: Kill mdd_lov_destroy().
2491 Details : Remove unused mdd code.
2495 Description: Add st_block checking to multistat.c.
2496 Details : Add st_block checking to multistat.c.
2500 Description: Add lu_ref support to struct obd_device.
2501 Details : Add lu_ref tracking to obd_device.
2505 Description: Kill obd_set_fail_loc().
2506 Details : Remove unused code.
2510 Description: Add special type for ptlrpc_request interpret functions.
2511 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
2512 there. Actual usage will be in CLIO.
2516 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
2517 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
2518 doesn't work with lockdep.
2522 Description: Add rwv.c test program.
2523 Details : New testing program exercising readv(2) and writev(2) (Qian).
2527 Description: Add sendfile.c test program.
2528 Details : New testing program exercising sendfile(2) (Jay).
2532 Description: Ratelimit a message that can be very frequent.
2533 Details : Ratelimit a memory allocation failure message that can
2538 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2539 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2543 Description: Liblustre build fixes.
2544 Details : Liblustre build fixes.
2548 Description: libcfs: add cfs_{need,cond}_resched() interface.
2549 Details : libcfs: add cfs_{need,cond}_resched() definition and
2550 implementations for Linux, NT, and liblustre.
2552 Severity : enhancement
2554 Description: More exported tunables for mballoc
2555 Details : Add support for tunable preallocation window and new tunables for
2556 large/small requests
2560 Description: Detect corruption of block bitmap and checking for preallocations
2561 Details : Checks validity of on-disk block bitmap. Also it does better
2562 checking of number of applied preallocations. When corruption is
2563 found, it turns filesystem readonly to prevent further corruptions.
2567 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
2568 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
2569 window to 0. If it does hit cache miss, and read-pattern is still
2570 stride-io mode, does not reset the stride window, but also does not
2571 increase the stride window length in this case.
2575 Frequency : only for big-endian servers
2576 Description: Check if system is big-endian while mounting fs with extents feature
2577 Details : Mounting a filesystem with extents feature will fail on big-endian
2578 systems since ext3-based ldiskfs is not supported on big-endian
2579 systems. This can be over-riden with "bigendian_extents" mount option.
2581 Severity : enhancement
2583 Description: The root squash functionality
2584 Details : A security feature, which is to prevent users from being able
2585 to mount lustre on their desktop, run as root, and delete
2586 all of the files in the filesystem. The goal is accomplished by
2587 remapping user id (UID) and group id (GID) of the root user to
2588 a UID and GID specified by the system administartor via Lustre
2589 configuration management server (MGS). The functionality also
2590 allows to specify sets of clients for which the remapping does
2595 Description: Excessive recovery window
2596 Details : With AT enabled, the recovery window can be excessively long (6000+
2597 seconds). To address this problem, we no longer use
2598 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
2599 timeout no longer depends on the service time, it is set to
2600 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
2601 time via pb_service_time.
2605 Description: Watchdog triggered on MDS failover
2606 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
2607 the OSTs know that the MDS "UUID" can be reused for the same export
2608 from a different NID, so we do not need to wait for the export to be
2612 Frequency : rare, only if using MMP with Linux RAID
2614 Description: MMP doesn't work with Linux RAID
2615 Details : While using HA for Lustre servers with Linux RAID, it is possible
2616 that MMP will not detect multiple mounts. To make this work we
2617 need to unplug the device queue in RAID when the MMP block is being
2618 written. Also while reading the MMP block, we should read it from
2619 disk and not the cached one.
2621 Severity : enhancement
2623 Description: open file using fid
2624 Details : A file can be opened using just its fid, like
2625 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
2628 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
2631 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
2632 Details : add proper locking for creation of jbd_4k slab cache
2636 Description: MMP check in ext3_remount() fails without displaying any error
2637 Details : When multiple mount protection fails during remount, proper error
2640 Severity : enhancement
2642 Description: Allow stripe size to be up to 4G-64k
2643 Details : Fix math logic to allow large stripe sizes.
2647 Description: add check for >8TB ldiskfs filesystems
2648 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
2649 Don't allow >8TB ldiskfs filesystems to be mounted without
2650 force_over_8tb mount option
2655 Descriptoin: OSS complains "too many blocks" and crashes
2656 Details : In filter_commitrw_write():
2657 1. lquota_chkquota();
2658 2. obdo_from_inode();
2659 3. lquota_pending_commit();
2660 "2." has changed uid and gid in this case, then caused
2661 lquota_chkquota() and lquota_pending_commit() to use
2662 different uid and gid. That is the root of the bug.
2666 Descriptoin: getstripe doesn't report stripe count = -1 sanely
2667 Details : Detect and print out the 'stripe count = -1' case properly.
2671 Description: Out or order replies might be lost on replay
2672 Details : In ptlrpc_retain_replayable_request if we cannot find retained
2673 request with tid smaller then one currently being added, add it
2674 to the start, not end of the list.
2676 --------------------------------------------------------------------------------
2678 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
2680 * Support for kernels:
2681 2.6.5-7.283 (SLES 9),
2682 2.6.9-55.EL (RHEL 4),
2683 2.6.16.46-0.14 (SLES 10),
2684 2.6.18.8 vanilla (kernel.org)
2685 * Client support for unpatched kernels:
2686 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2687 2.6.16 - 2.6.22 vanilla (kernel.org)
2688 * Due to recently discovered recovery problems, we do not recommend
2689 using patchless RHEL 4 clients with this or any earlier release.
2690 * Recommended e2fsprogs version: 1.39.cfs8
2691 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2692 * Starting with this release, the ldiskfs backing filesystem required
2693 by Lustre is now in its own package, lustre-ldiskfs. This package
2694 should be installed. It is versioned separately from Lustre and
2695 may be released separately in future.
2699 Description: block reactivating mgc import until all deactivates complete
2700 Details : Fix race when failing back MDT/MGS to itself (testing)
2703 Frequency : at statup only
2705 Description: mds_lov_synchronize race leads to various problems
2706 Details : simultaneous MDT->OST connections at startup can cause the
2707 sync to abort, leaving the OSC in a bad state.
2709 Severity : enhancement
2711 Description: add optional extra BUILD_VERSION info
2712 Details : add a new environment variable (namely LUSTRE_VERS) which allows
2713 to override the lustre version.
2716 Frequency : 2.6.18 servers only
2718 Description: ll_kern_mount() doesn't release the module reference
2719 Details : The ldiskfs module reference count never drops down to 0
2720 because ll_kern_mount() doesn't release the module reference.
2725 Description: server LBUG when using old ost_num_threads parameter
2726 Details : Accept the old ost_num_threads parameter but warn that it
2727 is deprecated, and fix an off-by-one error that caused an LBUG.
2732 Description: Transient SCSI error results in persistent IO issue
2733 Details : iobuf->dr_error is not reinitialized to 0 between two
2737 Frequency : sometimes when underlying device returns I/O errors
2739 Description: OSTs not going read-only during write failures
2740 Details : OSTs are not remounted read-only when the journal commit threads
2741 get I/O errors because fsfilt_ext3 calls journal_start/stop()
2742 instead of the ext3 wrappers.
2746 Description: poor connect scaling with increasing client count
2747 Details : Don't run filter_grant_sanity_check for more than 100 exports
2748 to improve scaling for large numbers of clients.
2751 Frequency : SLES10 only
2753 Description: sanity-quota.sh quotacheck failed: rc = -22
2754 Details : Quotas cannot be enabled on SLES10.
2757 Frequency : liblustre clients only
2759 Description: getdirentries does not give error when run on compute nodes
2760 Details : getdirentries does not fail when the size specified as an argument
2761 is too small to contain at least one entry
2763 Severity : enhancement
2765 Description: Add LNET router traceability for debug purposes
2766 Details : If a checksum failure occurs with a router as part of the
2767 IO path, the NID of the last router that forwarded the bulk data
2768 is printed so it can be identified.
2773 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
2774 Details : Due to a race condition, liblustre clients were occasionally
2775 evicted incorrectly.
2777 Severity : enhancement
2779 Description: lfs setstripe use optional parameters instead of postional
2782 Severity : enhancement
2784 Description: Nanosecond timestamp support for ldiskfs
2785 Details : The on-disk ldiskfs filesystem has added support for nanosecond
2786 resolution timestamps. There is not yet support for this at
2787 the Lustre filesystem level.
2790 Frequency : during server recovery
2792 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
2793 Details : request with rq_no_resend flag not awake l_wait_event if they get
2797 Frequency : nfs export on patchless client
2799 Description: connectathon hang when test nfs export over patchless client
2800 Details : Disconnected dentry cannot be found with lookup, so we do not need
2801 to unhash it or make it invalid
2804 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
2805 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
2806 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
2811 Description: open req refcounting wrong on reconnect
2812 Details : If reconnect happened between getting open reply from server and
2813 call to mdc_set_replay_data in ll_file_open, we will schedule
2814 replay for unreferenced request that we are about to free.
2815 Subsequent close will crash in variety of ways.
2816 Check that request is still eligible for replay in
2817 mdc_set_replay_data().
2822 Description: disable writes to filesystem when reading health_check file
2823 Details : the default for reading the health_check proc file has changed
2824 to NOT do a journal transaction and write to disk, because this
2825 can cause reads of the /proc file to hang and block HA state
2826 checking on a healthy but otherwise heavily loaded system. It
2827 is possible to return to the previous behaviour during configure
2828 with --enable-health-write.
2830 Severity : enhancement
2832 Description: 64-bit inode version
2833 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
2834 to the inode. This will be required for version-based recovery.
2839 Description: MDS fails to start if a duplicate client export is detected
2840 Details : in some rare cases it was possible for a client to connect to
2841 an MDS multiple times. Upon recovery the MDS would detect this
2842 and fail during startup. Handle this more gracefully.
2844 Severity : enhancement
2846 Description: Add -o localflock option to simulate old noflock
2848 Details : This will achieve local-only flock/fcntl locks
2854 Description: log_commit_thread vs filter_destroy race leads to crash
2855 Details : Take import reference before releasing llog record semaphore
2860 Description: Wrong request locking in request set processing
2861 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
2862 imp_delayed_list, in this place should be used imp_lock.
2865 Frequency : when reconnection
2867 Description: Grant Leak when osc reconnect to OST
2868 Details : When osc reconnect ost, OST(filter) should check whether it
2869 should grant more space to client by comparing fed_grant and
2870 cl_avail_grant, and return the granted space to client instead
2871 of "new granted" space, because client will call osc_init_grant
2872 to update the client grant space info.
2875 Frequency : when client reconnect to OST
2877 Description: Grant Leak when osc do resend and replay bulk write
2878 Details : When osc reconnect to OST, OST(filter)should clear grant info of
2879 bulk write request, because the grant info while be sync between
2880 OSC and OST when reconnect, and we should ignore the grant info
2881 these of resend/replay write req.
2886 Description: Grant space more than avaiable left space sometimes.
2887 Details : When then OST is about to be full, if two bulk writing from
2888 different clients came to OST. Accord the avaliable space of the
2889 OST, the first req should be permitted, and the second one
2890 should be denied by ENOSPC. But if the seconde arrived before
2891 the first one is commited. The OST might wrongly permit second
2892 writing, which will cause grant space > avaiable space.
2895 Frequency : when client is evicted
2897 Description: Grant might be wrongly erased when osc is evicted by OST
2898 Details : when the import is evicted by server, it will fork another
2899 thread ptlrpc_invalidate_import_thread to invalidate the
2900 import, where the grant will be set to 0. While the original
2901 thread will update the grant it got when connecting. So if
2902 the former happened latter, the grant will be wrongly errased
2903 because of this race.
2908 Description: Checking Stale with correct fid
2909 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
2910 to check whether it is stale, because sometimes, we want the
2911 enqueue happened anyway, and op_data.fid2 will not be initialized.
2913 Severity : enhancement
2915 Description: update patchless client
2916 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
2919 Frequency : only with 2.4 kernel
2921 Description: random memory corruption
2922 Details : size of struct ll_inode_info is to big for union inode.u and this
2923 can be cause of random memory corruption.
2928 Description: Memory leak in recovery
2929 Details : Lov_mds_md was not free in an error handler in mds_create_object.
2930 It should also check obd_fail before fsfilt_start, otherwise if
2931 fsfilt_start return -EROFS,(failover mds during mds recovery).
2932 then the req will return with repmsg->transno = 0 and rc = EROFS.
2933 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
2934 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
2935 be freed no matter whether fsfilt_commit success or not.
2938 Frequency : only with huge count clients
2940 Description: Prevents from taking the superblock lock in llap_from_page for
2942 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
2943 for llap_from_page call in ll_removepage prevents from taking the
2944 superblock lock for a soon died page.
2949 Description: Not check open intent error before release open handle
2950 Details : in some rare cases, the open intent error is not checked before
2951 release open handle, which may cause
2952 ASSERTION(open_req->rq_transno != 0), because it tries to release
2953 the failed open handle.
2958 Description: Set cat log bitmap only after create log success.
2959 Details : in some rare cases, the cat log bitmap is set too early. and it
2960 should be set only after create log success.
2964 Description: Accessing a block bevice can re-enable I/O when Lustre is
2965 tearing down a device.
2966 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
2971 Description: service threads may hog cpus when there are a lot of requests
2972 Details : Insert cond_resched to give other threads a chance to use some CPU
2977 Description: the cat log was not initialized in recovery
2978 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
2979 unlink log on mds will not be initialized until mds post
2980 recovery. And also in mds post recovery, the unlink log will
2981 initialization will be done asynchronausly, so there will be race
2982 between add unlink log and unlink log initialization.
2986 Description: brw_stats were being printed incorrectly
2987 Details : brw_stats were being printed as log2 but all of them were not
2988 recorded as log2. Also remove some code duplication arising from
2989 filter_tally_{read,write}.
2993 Frequency : rare, only in recovery.
2994 Description: ASSERTION(req->rq_type != LI_POISON) failed
2995 Details : imp_lock should be held while iterating over imp_sending_list for
2996 prevent destroy request after get timeout in ptlrpc_queue_wait.
3000 Description: replay-single.sh test 52 fails
3001 Details : A lock's skiplist need to be cleanup when it being unlinked
3002 from its resource list.
3006 Description: Short directio read returns full requested size rather than
3008 Details : Direct I/O operations should return actual amount of bytes
3009 transferred rather than requested size.
3011 Severity : enhancement
3013 Description: metadata RPC reduction (e.g. for rm performance)
3014 Details : decrease the amount of synchronous RPC between clients and servers
3015 by canceling conflicing lock before the operation on the client
3016 and packing thier handles into the main operation RPC to server.
3018 Severity : enhancement
3020 Description: Async OSC create to avoid the blocking unnecessarily.
3021 Details : If a OST has no remain object, system will block on the creating
3022 when need to create a new object on this OST. Now, ways use
3023 pre-created objects when available, instead of blocking on an
3024 empty osc while others are not empty. If we must block, we block
3025 for the shortest possible period of time.
3029 Description: improve handling recoverable errors
3030 Details : if request processig with error which can be recoverable on server
3031 request should be resend, otherwise page released from cache and
3034 Severity : enhancement
3036 Description: refine locking for avoid write wrong info into lov_objid file
3037 Details : fix possible races with add new target and write/update data in
3040 --------------------------------------------------------------------------------
3042 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
3047 Frequency : on some architectures
3049 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
3050 Details : Uninitialized flags sometimes cause configuration commands to
3054 Frequency : patchless clients only
3056 Description: missing __iget() symbol export
3057 Details : The __iget() symbol export is missing. To avoid the need for
3058 this on patchless clients the deathrow inode reaper is turned
3059 off, and we depend on the VM to clean up old inodes. This
3060 dependency was during via the fix for bug 12181.
3062 --------------------------------------------------------------------------------
3064 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
3066 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
3067 INTEROPERATE with older versions automatically. In many cases a
3068 special upgrade step is needed. Please read the
3069 user documentation before upgrading any part of a live system.
3070 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
3071 version will not interoperate with 1.6 betas before beta5 (1.5.95).
3072 * WARNING: Lustre configuration and startup changes are required with
3073 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
3077 Severity : enhancement
3079 Description: MountConf
3080 Details : Lustre configuration is now managed via mkfs and mount
3081 commands instead of lmc and lconf. New obd types (MGS, MGC)
3082 are added for dynamic configuration management. See
3083 https://mail.clusterfs.com/wikis/lustre/MountConf for
3086 Severity : enhancement
3088 Description: dynamic OST addition
3089 Details : OSTs can now be added to a live filesystem
3091 Severity : enhancement
3093 Description: startup order invariance
3094 Details : MDTs and OSTs can be started in any order. Clients only
3095 require the MDT to complete startup.
3097 Severity : enhancement
3099 Description: parallel, asynchronous orphan cleanup
3100 Details : orphan cleanup is now performed in separate threads for each
3101 OST, allowing parallel non-blocking operation.
3103 Severity : enhancement
3105 Description: optimized stripe assignment
3106 Details : stripe assignments are now made based on ost space available,
3107 ost previous usage, and OSS previous usage, in order to try
3108 to optimize storage space and networking resources.
3110 Severity : enhancement
3112 Description: Permanently set tunables
3113 Details : All writable /proc/fs/lustre tunables can now be permanently
3114 set on a per-server basis, at mkfs time or on a live system.
3116 Severity : enhancement
3118 Description: Lustre message v2
3119 Details : Add lustre message format v2.
3121 Severity : enhancement
3123 Description: client OST exclusion list
3124 Details : Clients can be started with a list of OSTs that should be
3125 declared "inactive" for known non-responsive OSTs.
3129 Description: SPEC SFS validation failure on NFS v2 over lustre.
3130 Details : Changes the blocksize for regular files to be 2x RPC size,
3131 and not depend on stripe size.
3133 Severity : enhancement
3135 Description: Multiple MD RPCs in flight.
3136 Details : Further unserialise some read-only MDS RPCs - learn about intents.
3137 To avoid overly-overloading MDS, introduce a limit on number of
3138 MDS RPCs in flight for a single client and add /proc controls
3139 to adjust this limit.
3141 Severity : enhancement
3143 Description: client read/write statistics
3144 Details : Add client read/write call usage stats for performance
3145 analysis of user processes.
3146 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
3147 file access. extents_stats shows chunk size distribution.
3148 extents_stats_per_process show chunk size distribution per
3151 Severity : enhancement
3153 Description: mds statistics
3154 Details : Add detailed mds operations statistics in
3155 /proc/fs/lustre/mds/*/stats.
3159 Description: Failure of copying files with lustre special EAs.
3160 Details : Client side always return success for setxattr call for lustre
3161 special xattr (currently only "trusted.lov").
3166 Description: Refcount LNET uuids
3167 Details : The global LNET uuid list grew linearly with every startup;
3168 refcount repeated list entries instead of always adding to
3171 Severity : enhancement
3173 Description: Dynamic service threads
3174 Details : Within a small range, start extra service threads
3175 automatically when the request queue builds up.
3178 Frequency : mixed-endian client/server environments
3180 Description: mixed-endian crashes
3181 Details : The new msg_v2 system had some failures in mixed-endian
3184 Severity : enhancement
3186 Description: Easy OST removal
3187 Details : OSTs can be permanently deactivated with e.g. 'lctl
3188 conf_param lustre-OST0001.osc.active=0'
3190 Severity : enhancement
3192 Description: MGS proc entries
3193 Details : Added basic proc entries for the MGS showing what filesystems
3196 Severity : enhancement
3198 Description: provide MGS failover
3199 Details : Added config lock reacquisition after MGS server failover.
3201 Severity : enhancement
3203 Description: add Linux 2.4 support
3204 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
3208 Description: a large application tries to do I/O to the same resource and dies
3209 in the middle of it.
3210 Details : Check the req->rq_arrival time after the call to
3211 ost_brw_lock_get(), but before we do anything about
3212 processing it & sending the BULK transfer request. This
3213 should help move old stale pending locks off the queue as
3214 quickly as obd_timeout.
3217 Frequency : when an incorrect nid is specified during startup
3219 Description: ptlrpc connect to non-existant node causes kernel crash
3220 Details : LNET can't be re-entered from an event callback, which
3221 happened when we expire a message after the export has been
3222 cleaned up. Instead, hand the zombie cleanup off to another
3225 Severity : enhancement
3227 Description: plain/inodebits lock performance improvement
3228 Details : Grouping plain/inodebits in granted list by their request modes
3229 and bits policy, thus improving the performance of search through
3233 Frequency : only if OST filesystem is corrupted
3235 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3236 Details : for a short time RPCs with bulk IO are in the replay list,
3237 but replay of bulk IOs is unimplemented. If the OST filesystem
3238 is corrupted due to disk cache incoherency and then replay is
3239 started it is possible to trip an assertion. Avoid putting
3240 committed RPCs into the replay list at all to avoid this issue.
3243 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3245 Description: System hang on startup
3246 Details : This bug allowed the liblustre (e.g. catamount) client to
3247 return to the app before handling all startup RPCs. This
3248 could leave the node unresponsive to lustre network traffic
3249 and manifested as a server ptllnd timeout.
3251 Severity : enhancement
3253 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3254 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
3255 boolean to control whether to print peer debug info when a
3256 client's RPC times out.
3259 Frequency : only for kernels with patches from Lustre below 1.4.3
3261 Description: Remove old rdonly API
3262 Details : Remove old rdonly API which unused from at least lustre 1.4.3
3265 Frequency : only for devices with external journals
3267 Description: Set external device read-only also
3268 Details : During a commanded failover stop, we set the disk device
3269 read-only while the server shuts down. We now also set any
3270 external journal device read-only at the same time.
3273 Frequency : when upgrading from 1.4 while trying to change parameters
3275 Description: The wrong (new) MDC name was used when setting parameters for
3276 upgraded MDT's. Also allows changing of OSC (and MDC)
3277 parameters if --writeconf is specified at tunefs upgrade time.
3280 Frequency : when setting specific ost indicies
3282 Description: QOS code breaks on skipped indicies
3283 Details : Add checks for missing OST indicies in the QOS code, so OSTs
3284 created with --index need not be sequential.
3286 Severity : enhancement
3288 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3289 Details : The uninit_groups feature works in conjunction with the kernel
3290 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3291 up the pass1 processing of e2fsck. This is a read-only feature
3292 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3293 mount filesystems that have had this feature enabled.
3295 Severity : enhancement
3297 Description: Improve multi-block allocation algorithm to avoid fragmentation
3298 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3299 allocation locality and avoid filesystem fragmentation.
3301 ------------------------------------------------------------------------------
3303 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
3305 * Support for kernels:
3306 2.4.21-47.0.1.EL (RHEL 3)
3307 2.6.5-7.283 (SLES 9)
3308 2.6.9-42.0.10.EL (RHEL 4)
3309 2.6.12.6 vanilla (kernel.org)
3310 2.6.16.27-0.9 (SLES 10)
3311 * Recommended e2fsprogs version: 1.39.cfs5
3313 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3317 Frequency : occasional, depends on client load and configuration
3318 Bugzilla : 12181, 12203
3319 Description: data loss for recently-modified files
3321 Details : In some cases it is possible that recently written or created
3322 files may not be written to disk in a timely manner (this should
3323 normally be within 30s unless client IO load is very high).
3324 The problem appears as zero-length files or files that are a
3325 multiple of 1MB in size after a client crash or client eviction
3326 that are missing data at the end of the file.
3328 This problem is more likely to be hit on clients where files are
3329 repeatedly created and unlinked in the same directory, clients
3330 have a large amount of RAM, have many CPUs, the filesystem has
3331 many OSTs, the clients are rebooted frequently, and/or the files
3332 are not accessed by other nodes after being written.
3334 The presence of the problem can be detected by looking at
3335 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
3336 smaller than the second (nr_unused) then dirty files will not
3337 be flushed automatically to disk. "sync; sleep 10" should be
3338 run several times on the node before unmounting it to update
3339 Lustre (this is also safe to run on nodes without this problem).
3341 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
3342 that can cause this same problem, so customers using that kernel
3343 also need to update the kernel in addition to Lustre. In order
3344 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
3346 It is normal that files written just before a client crash (less
3347 than 30s) may not yet have been flushed to disk, even for local
3351 Frequency : frequent on thin XT3 nodes
3353 Description: UUID collision on thin XT3 Linux nodes
3354 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
3355 randomly, since we relied on an insufficiently-seeded PRNG.
3360 Description: OSS hangs after "All ost request buffers busy"
3361 Details : A deadlock between quota and journal operations caused OSS
3362 hangs after printing "All ost request buffers busy."
3365 Frequency : always on liblustre builds
3367 Description: Cleanup compiler warnings on liblustre
3370 Frequency : always on liblustre builds on XT3
3372 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
3376 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
3377 > 32000 subdirectories
3378 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
3379 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
3380 subdirectory count crosses 32000. This will aid e2fsck to
3381 correctly handle more than 32000 subdirectories.
3384 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3386 Description: System hang on startup
3387 Details : This bug allowed the liblustre (e.g. catamount) client to
3388 return to the app before handling all startup RPCs. This
3389 could leave the node unresponsive to lustre network traffic
3390 and manifested as a server ptllnd timeout.
3392 Severity : enhancement
3394 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3395 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
3396 boolean to control whether to print peer debug info when a
3397 client's RPC times out.
3402 Description: make O_SYNC working on 2.6 kernels
3403 Details : 2.6 kernels use different method for mark pages for write,
3404 so need add a code to lustre for O_SYNC work.
3409 Description: Failure to close file and release space on NFS
3410 Details : Put inode details into lock acquired in ll_intent_file_open.
3411 Use mdc_intent_lock in ll_intent_open to properly
3412 detect all kind of errors unhandled by mdc_enqueue.
3417 Description: proc file read during shutdown sometimes raced obd removal,
3419 Details : Add lock to prevent obd access after proc file removal.
3422 Frequency : Only for files larger than 4GB on 32-bit clients.
3424 Description: improperly doing page alignment of locks
3425 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
3426 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3431 Description: RPCs being resent when they shouldn't be
3432 Details : Some RPCs that should not be resent are being resent. This
3433 can cause inconsistencies in the RPC state machine. Do not
3434 resend such requests.
3437 Frequency : rare, only with NFS export
3439 Description: Crash on NFS re-export node
3440 Details : under very unusual load conditions an assertion is hit in
3441 ll_intent_file_open()
3444 Frequency : only if OST filesystem is corrupted
3446 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3447 Details : for a short time RPCs with bulk IO are in the replay list,
3448 but replay of bulk IOs is unimplemented. If the OST filesystem
3449 is corrupted due to disk cache incoherency and then replay is
3450 started it is possible to trip an assertion. Avoid putting
3451 committed RPCs into the replay list at all to avoid this issue.
3456 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3457 Details : Large single O_DIRECT read and write calls can fail to allocate
3458 a sufficiently large buffer to process the request. In case of
3459 allocation failure the allocation is retried with a smaller
3460 buffer and broken into smaller requests.
3462 Severity : enhancement
3464 Description: Add -o localflock option to simulate old noflock behaviour.
3465 Details : This will achieve local-only flock/fcntl locks coherentness.
3470 Description: versioning check is incomplete
3471 Details : Checking the version difference of client vs. server, report
3472 error if the gap is too big.
3477 Description: add support PG_writeback bit
3478 Details : add support for PG_writeback bit for Lustre, for more carefull
3479 work with page cache in 2.6 kernel. This also fix some deadlocks
3480 and remove hack for work O_SYNC with 2.6 kernel.
3482 Severity : enhancement
3484 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3485 Details : The uninit_groups feature works in conjunction with the kernel
3486 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3487 up the pass1 processing of e2fsck. This is a read-only feature
3488 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3489 mount filesystems that have had this feature enabled.
3491 Severity : enhancement
3493 Description: Improve multi-block allocation algorithm to avoid fragmentation
3494 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3495 allocation locality and avoid filesystem fragmentation.
3497 ------------------------------------------------------------------------------
3499 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3501 * Support for kernels:
3502 2.6.9-42.0.3.EL (RHEL 4)
3503 2.6.5-7.276 (SLES 9)
3504 2.4.21-47.0.1.EL (RHEL 3)
3505 2.6.12.6 vanilla (kernel.org)
3506 2.6.16.21-0.8 (SLES10)
3507 * Recommended e2fsprogs version: 1.39.cfs2-0
3509 * The backwards-compatible /proc/sys/portals symlink has been removed
3510 in this release. Before upgrading, please ensure that you change
3511 any configuration scripts or /etc/sysctl.conf files that access
3512 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3513 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3514 in advance of the upgrade on any system running Lustre 1.4.6 or
3515 newer, since /proc/sys/lnet was added in that version.
3516 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3520 Frequency : only when quota is used
3522 Description: avoid scanning export list for quota master
3523 Details : Change the algorithms to avoid scanning export list in order
3524 to improve the efficiency.
3527 Frequency : MDS failover only, very rarely
3529 Description: "went back in time" messages on mds failover
3530 Details : The greatest transno may be lost when the current operation
3531 finishes with an error (transno==0) and the client's last_rcvd
3532 record is over-written. Save the greatest transno in the
3533 mds_last_transno for this case.
3536 Frequency : always for specific kernels and striping counts
3538 Description: client may get "Matching packet too big" without ACL support
3539 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
3540 when trying to access files in certain configurations. The
3541 clients should in fact be denied when mounting because they do
3542 not understand ACLs.
3545 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
3547 Description: many clients connecting with IO in progress causes connect timeouts
3548 Details : Avoid synchronous journal commits to avoid delays caused by many
3549 clients connecting/disconnecting when bulk IO is in progress.
3550 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
3551 OST_IO_PORTAL to avoid delays behind potentially many pending
3555 Frequency : occasionally with multiple writers to a single file
3557 Description: shared writes to file may result in wrong size reported by stat()
3558 Details : Allow growing of kms when extent lock is cancelled
3561 Frequency : always with random mmap IO to multi-striped file
3563 Description: mmap write might be lost if we are writing to a 'hole' in stripe
3564 Details : Only if the hole is at the end of OST object so that kms is too
3565 small. Fix is to increase kms accordingly in ll_nopage.
3568 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
3570 Description: writes to a missing object would leak memory on the OST
3571 Details : If there is an inconsistency between the MDS and OST filesystems,
3572 such that the MDS references an object that doesn't exist, writes
3573 to that object will leak memory due to incorrect cleanup in the
3574 error handling path, eventually running out of memory on the OST.
3579 Description: Creating too long symlink causes lustre errors
3580 Details : Check symlink and name lengths before sending requests to MDS.
3583 Frequency : only if flock is enabled (not on by default)
3585 Description: posix locks not released on fd closure on 2.6.9+
3586 Details : We failed to add posix locks to list of inode locks on 2.6.9+
3587 kernels, this caused such locks not to be released on fd close and
3588 then assertions on fs unmount about still used locks.
3591 Frequency : MDS failover only, very rarely
3593 Description: clients may get ASSERTION(granted_lock != NULL)
3594 Details : When request was taking a long time, and a client was resending
3595 a getattr by name lock request. The were multiple lock requests
3596 with the same client lock handle and
3597 mds_getattr_name->fixup_handle_for_resent_request found one of the
3598 lock handles but later failed with ASSERTION(granted_lock != NULL).
3603 Description: handle->h_buffer_credits > 0, assertion failure
3604 Details : h_buffer_credits is zero after truncate, causing assertion
3605 failure. This patch extends the transaction or creates a new
3609 Frequency : NFS re-export or patchless client
3610 Bugzilla : 11179, 10796
3611 Description: Crash on NFS re-export node (__d_move)
3612 Details : We do not want to hash the dentry if we don't have a lock.
3613 But if this dentry is later used in d_move, we'd hit uninitialised
3614 list head d_hash, so we just do this to init d_hash field but
3615 leave dentry unhashed.
3618 Frequency : NFS re-export or patchless client
3620 Description: NFS exports has problem with symbolic link
3621 Details : lustre client didn't properly install dentry when re-exported
3622 to NFS or running patchless client.
3625 Frequency : NFS re-export or patchless client
3627 Description: Various nfs/patchless fixes.
3628 Details : fixes reuse disconected alias for lookup process - this fixes
3629 warning "find_exported_dentry: npd != pd",
3630 fix permission error with open files at nfs.
3631 fix apply umask when do revalidate.
3634 Frequency : occasional
3636 Description: Crash on NFS re-export node
3637 Details : calling clear_page() on the wrong pointer triggered oops in
3638 generic_mapping_read().
3641 Frequency : rarely, using O_DIRECT IO
3643 Description: unaligned directio crashes client with LASSERT
3644 Details : check for unaligned buffers before trying any requests.
3647 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
3649 Description: stale data returned from RAID cache
3650 Details : If only a small amount of IO is done to the RAID device before
3651 reading it again it is possible to get stale data from the RAID
3652 cache instead of reading it from disk.
3655 Frequency : always for sles10 kernel
3657 Description: sles10 support
3658 Details : ll_follow_link: compile fixes and using of nd_set_link
3659 under newer kernels.
3662 Frequency : depends on arch, kernel and compiler version, always on sles10
3665 Description: recursive or deep enough symlinks cause stack overflow
3666 Details : getting rid of large stack-allocated variable in
3670 Frequency : depends on hardware
3672 Description: lustre write performance loss in the SLES10 kernel
3673 Details : the performance loss is caused by using of write barriers in the
3674 ext3 code. The SLES10 kernel turns barrier support on by
3675 default. The fix is to undo that change for ldiskfs.
3677 ------------------------------------------------------------------------------
3679 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
3681 * Support for kernels:
3682 2.6.9-42.0.3EL (RHEL 4)
3683 2.6.5-7.276 (SLES 9)
3684 2.4.21-47.0.1.EL (RHEL 3)
3685 2.6.12.6 vanilla (kernel.org)
3689 Frequency : quota enabled and large files being deleted
3691 Description: releasing more than 4GB of quota at once hangs OST
3692 Details : If a user deletes more than 4GB of files on a single OST it
3693 will cause the OST to spin in an infinite loop. Release
3694 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3699 Description: statfs data retrieved from /proc may be stale or zero
3700 Details : When reading per-device statfs data from /proc, in the
3701 {kbytes,files}_{total,free,avail} files, it may appear
3702 as zero or be out of date.
3705 Frequency : systems with MD RAID1 external journal devices
3707 Description: lconf's call to blkid is confused by RAID1 journal devices
3708 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3709 of returning all block devices that match the journal UUID.
3712 Frequency : always, for aggregate stripe size over 4GB
3714 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
3715 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
3716 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
3719 Frequency : always if "lfs find" used on a local file/directory
3721 Description: "lfs find" segfaults if used on a local file/directory
3722 Details : The case where a directory component was not specified wasn't
3723 handled correctly. Handle this properly.
3726 Frequency : always on ppc64
3728 Description: the write to an ext3 filesystem mounted with mballoc got stuck
3729 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3730 perform endianness conversion.
3733 Frequency : rarely (truncate to non-zero file size after write under load)
3734 Bugzilla : 10730, 10687
3735 Description: Files padded with zeros to next 4K multiple
3736 Details : With filesystems mounted using the "extents" option (2.6 kernels)
3737 it is possible that files that are truncated to a non-zero size
3738 immediately after being written are filled with zero bytes beyond
3739 the truncated size. No file data is lost.
3741 Severity : enhancement
3743 Description: Allow recovery/failover for liblustre clients.
3744 Details : liblustre clients were unaware of failover configurations until
3747 Severity : enhancement
3749 Description: user file locks should fail when not mounting with flock option
3750 Details : Set up an error-returning stub in ll_file_operations.lock field
3751 to prevent incorrect behaviour when client is mounted without
3752 flock option. Also, set up properly f_op->flock field for
3756 Frequency : always on ia64
3758 Description: "lfs df" loops on printing out MDS statfs information
3759 Details : The obd_ioctl_data was not initialized and in some systems
3760 this caused a failure during the ioctl that did not return
3761 an error. Initialize the struct and return an error on failure.
3764 Frequency : SLES 9 only
3766 Description: Error of copying files with lustre special EAs as root
3767 Details : Client side always return success for setxattr call for lustre
3768 special xattr (currently only "trusted.lov").
3771 Frequency : rarely on clusters with both ia64+i386 clients
3773 Description: ia64+i686 clients doing shared IO on the same file may LBUG
3774 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
3775 clients are doing concurrent writes to the same file it is
3776 possible that the ia64 clients may LASSERT because the OST
3777 extent locks are not PAGE_SIZE aligned. Ensure that grown
3778 locks are always aligned on the request boundary.
3781 Frequency : specific use, occasional
3783 Description: Overwriting in use executable truncates on-disk binary image
3784 Details : If one node attempts to overwrite an executable in use by
3785 another node, we now correctly return ETXTBSY instead of
3786 truncating the file.
3788 Severity : enhancement
3790 Description: Async OSC create to avoid the blocking unnecessarily.
3791 Details : If a OST has no remain object, system will block on the creating
3792 when need to create a new object on this OST. Now, ways use
3793 pre-created objects when available, instead of blocking on an
3794 empty osc while others are not empty. If we must block, we block
3795 for the shortest possible period of time.
3800 Description: chmod on Lustre root is propagated to other clients
3801 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
3802 invalid by the follow_mount time.
3805 Frequency : liblustre clients only
3807 Description: Race in 'instant cancel' lock handling could lead to such locks
3808 never to be granted in case of SMP MDS
3809 Details : Do not destroy not yet granted but cbpending locks in
3813 Frequency : replay/resend of open
3815 Description: non null lock assetion failure in mds_intent_policy
3816 Details : Trying to replay/resend lockless open requests resulted in
3817 mds_open() returning 0 with no lock. Now it sets a flag if
3818 a lock is going to be returned.
3820 Severity : enhancement
3822 Description: Checksum enhancements
3823 Details : New checksum enhancements allow for resending RPCs that failed
3826 Severity : enhancement
3828 Description: Tunables on number of dirty pages in cacche
3829 Details : Allow to set limit on number of dirty pages cached.
3834 Description: client crash on unmount - lock still has references
3835 Details : In some error handling cases it was possible to leak a lock
3836 reference on a client while accessing a file. This was not
3837 harmful to the client during operation, but would cause the
3838 client to crash when the filesystem is unmounted.
3841 Frequency : specific case, rare
3843 Description: ETXTBSY on mds though file not in use by client
3844 Details : ETXTBSY is no longer incorrectly returned when attempting to
3845 chmod or chown a directory that the user previously tried to
3846 execute or a currently-executing binary.
3849 Frequency : extremely rare except on liblustre-based clients
3851 Description: Lustre space not freed when files are deleted
3852 Details : Clean up open-unlinked files after client eviction. Previously
3853 the unlink was skipped and the files remained as orphans.
3858 Description: OST failure "would be an LBUG" in waiting_locks_callback()
3859 Details : In some cases it was possible to send a blocking callback to
3860 a client doing a glimpse, even though that client didn't get
3861 a lock granted. When the glimpse lock is cancelled on the OST
3862 the freed lock is left on the waiting list and corrupted the list.
3865 Frequency : all core dumps
3867 Description: Broke core dumps to lustre
3868 Details : Negative dentry may be unhashed if parent does not have UPDATE
3869 lock, but some callers, e.g. do_coredump, expect dentry to be
3870 hashed after successful create, hash it in ll_create_it.
3872 ------------------------------------------------------------------------------
3874 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
3876 * Support for kernels:
3877 2.6.9-42.0.2.EL (RHEL 4)
3878 2.6.5-7.276 (SLES 9)
3879 2.4.21-40.EL (RHEL 3)
3880 2.6.12.6 vanilla (kernel.org)
3884 Frequency : always on RHEL 3
3886 Description: Number of open files grows over time
3887 Details : The number of open files grows over time, whether or not
3888 Lustre is started. This was due to a filp leak introduced
3889 by one of our kernel patches.
3891 ------------------------------------------------------------------------------
3893 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
3895 * Support for kernels:
3896 2.6.9-42.EL (RHEL 4)
3897 2.6.5-7.276 (SLES 9)
3898 2.4.21-40.EL (RHEL 3)
3899 2.6.12.6 vanilla (kernel.org)
3904 Bugzilla : 5719, 9635, 9792, 9684
3905 Description: OST (or MDS) trips assertions in (re)connection under heavy load
3906 Details : If a server is under heavy load and cannot reply to new
3907 connection requests before the client resends the (re)connect,
3908 the connection handling code can behave badly if two service
3909 threads are concurrently handing separate (re)connections from
3910 the same client. Add better locking to the connection handling
3911 code, and ensure that only a single connection will be processed
3912 for a given client UUID, even if the lock is dropped.
3914 Severity : enhancement
3916 Description: add TCP zero-copy support to kernel
3917 Details : Add support to the kernel TCP stack to allow zero-copy bulk
3918 sends if the hardware supports scatter-gather and checksumming.
3919 This allows socklnd to do client-write and server-read more
3920 efficiently and reduce CPU utilization from skbuf copying.
3923 Frequency : only if NFS exporting from client
3925 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
3926 Details : If setting attributes on a file created under NFS that had
3927 never been opened it would be possible to oops the client
3928 if the file had no objects.
3931 Frequency : always for liblustre
3933 Description: liblustre client does MDS+OSTs setattr RPC for each write
3934 Details : When doing a write from a liblustre client, the client
3935 incorrectly issued an RPC to the MDS and each OST the file was
3936 striped over in order to update the timestamps. When writing
3937 with small chunks and many clients this could overwhelm the MDS
3938 with RPCs. In all cases it would slow down the write because
3939 these RPCs are unnecessary.
3941 Severity : enhancement
3943 Description: allow number of MDS service threads to be changed at module load
3944 Details : It is now possible to change the number of MDS service threads
3945 running. Adding "options mds mds_num_threads={N}" to the MDS's
3946 /etc/modprobe.conf will set the number of threads for the next
3947 time Lustre is restarted (assuming the "mds" module is also
3948 reloaded at that time). The default number of threads will
3949 stay the same, 32 for most systems.
3954 Description: OST crash if filesystem is unformatted or corrupt
3955 Details : If an OST is started on a device that has never been formatted
3956 or if the filesystem is corrupt and cannot even mount then the
3957 error handling cleanup routines would dereference a NULL pointer.
3962 Description: NULL pointer deref in llap_from_page.
3963 Details : get_cache_page_nowait can return a page with NULL (or otherwise
3964 incorrect) mapping if the page was truncated/reclaimed while it was
3965 searched for. Check for this condition and skip such pages when
3966 doing readahead. Introduce extra check to llap_from_page() to
3967 verify page->mapping->host is non-NULL (so page is not anonymous).
3970 Frequency : Sometimes when using sys_sendfile
3972 Description: "page not covered by a lock" warnings from ll_readpage
3973 Details : sendfile called ll_readpage without right page locks present.
3974 Now we introduced ll_file_sendfile that does necessary locking
3975 around call to generic_file_sendfile() much like we do in
3979 Frequency : with certain MDS communication failures at client mount time
3981 Description: NULL pointer deref after failed client mount
3982 Details : a client connection request may delayed by the network layer
3983 and not be sent until after the PTLRPC layer has timed out the
3984 request. If the client fails the mount immediately it will try
3985 to clean up before the network times out the request. Add a
3986 reference from the request import to the obd device and delay
3987 the cleanup until the network drops the request.
3990 Frequency : occasionally during client (re)connect
3992 Description: assertion failure during client (re)connect
3993 Details : processing a client connection request may be delayed by the
3994 client or server longer than the client connect timeout. This
3995 causes the client to resend the connection request. If the
3996 original connection request is replied in this interval, the
3997 client may trip an assertion failure in ptlrpc_connect_interpret()
3998 which thought it would be the only running connect process.
4001 Frequency : only with obd_echo servers and clients that are rebooted
4003 Description: kernel BUG accessing uninitialized data structure
4004 Details : When running an obd_echo server it did not start the ping_evictor
4005 thread, and when a client was evicted an uninitialized data
4006 structure was accessed. Start the ping_evictor in the RPC
4007 service startup instead of the OBD startup.
4009 Severity : enhancement
4010 Bugzilla : 10193 (patchless)
4011 Description: Remove dependency on various unexported kernel interfaces.
4012 Details : No longer need reparent_to_init, exit_mm, exit_files,
4013 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
4016 Frequency : rare (only users of deprecated and unsupported LDAP config)
4018 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
4019 Details : LDAP apparently contains 'lustreName' attributes instead of
4020 'name'. A simple remapping of the name is sufficient.
4023 Frequency : rare (only with non-default dump_on_timeout debug enabled)
4025 Description: waiting_locks_callback trips kernel BUG if client is evicted
4026 Details : Running with the dump_on_timeout debug flag turned on makes
4027 it possible that the waiting_locks_callback() can try to dump
4028 the Lustre kernel debug logs from an interrupt handler. Defer
4029 this log dumping to the expired_lock_main() thread.
4031 Severity : enhancement
4033 Description: Support NFS exporting on 2.6 kernels.
4034 Details : Implement non-rawops metadata methods for NFS server to use without
4035 changing NFS server code.
4038 Frequency : very rare (synthetic metadata workload only)
4040 Description: two racing renames might cause an MDS thread to deadlock
4041 Details : Running the "racer" program may cause one MDS thread to rename
4042 a file from being the source of a rename to being the target of
4043 a rename at exactly the same time that another thread is doing
4044 so, and the second thread has already enqueued these locks after
4045 doing a lookup of the target and is trying to relock them in
4046 order. Ensure that we don't try to re-lock the same resource.
4049 Frequency : only very large systems with liblustre clients
4051 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
4052 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
4053 all OSTs in parallel. This allows the eviction of stale liblustre
4054 clients to proceed much faster than if they were done in series,
4055 and also offers similar improvements for other set_info RPCs.
4060 Description: excessive CPU usage during initial read phase on client
4061 Details : During the initial read phase on a client, it would agressively
4062 retry readahead on the file, consuming too much CPU and impacting
4063 performance (since 1.4.5.8). Improve the readahead algorithm
4064 to avoid this, and also improve some other common cases (read
4065 of small files in particular, where "small" is files smaller than
4066 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
4071 Description: MDS crash when receiving packet with unknown intent.
4072 Details : Do not LBUG in unknown intent case, just return -EFAULT
4074 Severity : enhancement
4075 Bugzilla : 9293, 9385
4076 Description: MDS RPCs are serialised on client. This is unnecessary for some.
4077 Details : Do not serialize getattr (non-intent version) and statfs.
4080 Frequency : occasional, when OST network is overloaded/intermittent
4082 Description: client evicted by OST after bulk IO timeout
4083 Details : If a client sends a bulk IO request (read or write) the OST
4084 may evict the client if it is unresposive to its data GET/PUT
4085 request. This is incorrect if the network is overloaded (takes
4086 too long to transfer the RPC data) or dropped the OST GET/PUT
4087 request. There is no need to evict the client at all, since
4088 the pinger and/or lock callbacks will handle this, and the
4089 client can restart the bulk request.
4092 Frequency : Always when mmapping file with no objects
4094 Description: client crashes when mmapping file with no objects
4095 Details : Check that we actually have objects in a file before doing any
4096 operations on objects in ll_vm_open, ll_vm_close and
4102 Description: Request leak when working with deleted CWD
4103 Details : Introduce advanced request refcount tracking for requests
4104 referenced from lustre intent.
4106 Severity : Enhancement
4108 Description: Cache open file handles on client.
4109 Details : MDS now will return special lock along with openhandle, if
4110 requested and client is allowed to hold openhandle, even if unused,
4111 until such a lock is revoked. Helps NFS a lot, since NFS is opening
4112 closing files for every read/write openration.
4114 Severity : Enhancement
4116 Description: Cache open negative dentries on client when possible.
4117 Details : Guard negative dentries with UPDATE lock on parent dir, drop
4118 negative dentries on lock revocation.
4123 Description: Remounting a client read-only wasn't possible with a zconf mount
4124 Details : It wasn't possible to remount a client read-only with llmount.
4126 Severity : enhancement
4127 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
4128 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
4129 MPICH is included to add Lustre-specific ADIO interfaces.
4130 This is based closely on the UFS ADIO layer and only differs
4131 in file creation, in order to allow the OST striping to be set.
4132 This is user-contributed code and not supported by CFS.
4137 Description: extended inode attributes (immutable, append-only) work improperly
4138 when 2.4 and 2.6 kernels are used on client/server or vice versa
4139 Details : Introduce kernel-independent values for these flags.
4141 Severity : enhancement
4144 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
4145 Details : Many of the /proc/ tunables can only be tuned at a megabyte
4146 granularity. Now, Fractional MB granularity is be supported,
4147 this is very useful for low memory system.
4149 Severity : enhancement
4151 Description: Getattr by fid
4152 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
4153 locks, avoids extra getattr rpc requests to MDS, allows '/' to
4154 have locks and avoids getattr rpc requests for it on every stat.
4157 Frequency : Always, for filesystems larger than 2TB
4159 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4160 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
4161 filesystems larger than 2TB to be mounted with mballoc enabled.
4164 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
4166 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4167 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
4168 filesystems larger than 2TB is mounted then the kernel will
4169 silently corrupt the start of the filesystem. CONFIG_LBD is
4170 enabled for all CFS-supported kernels, but the possibility of
4171 this happening with a modified kernel config exists.
4173 Severity : enhancement
4175 Description: add client O_DIRECT support for 2.6 kernels
4176 Details : It is now possible to do O_DIRECT reads and writes to files
4177 in the Lustre client mountpoint on 2.6 kernel clients.
4179 Severity : enhancement
4181 Description: parallel glimpse, setattr, statfs, punch, destroy requests
4182 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
4183 parallel, not waiting for response from every OST before sending
4184 a rpc to the next OST.
4189 Description: setattr vs write race when updating file timestamps
4190 Details : Client processes that update a file timestamp into the past
4191 right after writing to the file (e.g. tar) it is possible that
4192 the updated file modification time can be reset to the current
4193 time due to a race between processing the setattr and write RPC.
4195 Severity : enhancement
4197 Description: Bring 'lfs find' closer in line with regular Linux find.
4198 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
4199 -print0 options and obtains all the needed info through the lustre
4202 Severity : enhancement
4204 Description: support up to 1024 configured devices on one node
4205 Details : change obd_dev array from statically allocated to dynamically
4206 allocated structs as they are first used to reduce memory usage
4211 Description: Flush dirty partially truncated pages during truncate
4212 Details : Immediatelly flush partially truncated pages in filter_setattr,
4213 this way we completely avoid having any pages in page cache on OST
4214 and can retire ugly workarounds during writes to flush such pages.
4219 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
4220 Details : protect inode from truncation within vfs_unlink() context
4221 just take a reference before calling vfs_unlink() and release it
4222 when parent's i_sem is free.
4227 Description: last_id value checked outside lock on OST caused LASSERT failure
4228 Details : If there were multiple MDS->OST object precreate requests in
4229 flight, it was possible that the OST's last object id was checked
4230 outside a lock and incorrectly tripped an assertion. Move checks
4231 inside locks, and discard old precreate requests.
4234 Frequency : always, if extents are used on OSTs
4236 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
4237 Details : OSTs using the extents format would not zero the high 16 bits of
4238 the index physical block number. This is not a problem for any
4239 OST filesystems smaller than 16TB, and no kernels support ext3
4240 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
4241 new/modified files) and can be fixed for existing filesystems
4242 with e2fsprogs-1.39-cfs1.
4247 Description: import connection selection may be incorrect if timer wraps
4248 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
4249 import connections to be ignored if the 32-bit jiffies counter
4250 wraps. Use a 64-bit jiffies counter.
4253 Frequency : very large clusters immediately after boot
4255 Description: LNET request buffers exhausted under heavy short-term load
4256 Details : If a large number of client requests are generated on a service
4257 that has previously never seen so many requests it is possible
4258 that the request buffer growth cannot keep up with the spike in
4259 demand. Instead of dropping incoming requests, they are held in
4260 the LND until the RPC service can accept more requests.
4263 Frequency : Sometimes during replay
4265 Description: Assertion failure in ll_local_open after replay.
4266 Details : If replay happened on an open request reply before we were able
4267 to set replay handler, reply will become not swabbed tripping the
4268 assertion in ll_local_open. Now we set the handler right after
4269 recognising of open request
4272 Frequency : very rare
4274 Description: kernel reports "badness in vsnprintf"
4275 Details : Reading from the "recovery_status" /proc file in small chunks
4276 may cause a negative length in lprocfs_obd_rd_recovery_status()
4277 call to vsnprintf() (which is otherwise harmless). Exit early
4278 if there is no more space in the output buffer.
4280 Severity : enhancement
4282 Description: clear OBD RPC statistics by writing to them
4283 Details : It is now possible to clear the OBD RPC statistics by writing
4284 to the "stats" file.
4289 Description: Client mtime is not the same on different clients after utimes
4290 Details : In some cases, the client was using the utimes() syscall on
4291 a file cached on another node. The clients now validate the
4292 ctime from the MDS + OSTs to determine which one is right.
4297 Description: Inability to activate failout mode
4298 Details : lconf script incorrectly assumed that in pythong string's numeric
4299 value is used in comparisons.
4302 Frequency : always with multiple stripes per file
4304 Description: Inefficient object allocation for mutli-stripe files
4305 Details : When selecting which OSTs to stripe files over, for files with
4306 a stripe count that divides evenly into the number of OSTs,
4307 the MDS is always picking the same starting OST for each file.
4308 Return the OST selection heuristic to the original design.
4313 Description: mount failures may take full timeout to return an error
4314 Details : Under some heavy load conditions it is possible that a
4315 failed mount can wait for the full obd_timeout interval,
4316 possibly several minutes, before reporting an error.
4317 Instead return an error as soon as the status is known.
4319 Frequency : quota enabled and large files being deleted
4321 Description: releasing more than 4GB of quota at once hangs OST
4322 Details : If a user deletes more than 4GB of files on a single OST it
4323 will cause the OST to spin in an infinite loop. Release
4324 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4329 Description: statfs data retrieved from /proc may be stale or zero
4330 Details : When reading per-device statfs data from /proc, in the
4331 {kbytes,files}_{total,free,avail} files, it may appear
4332 as zero or be out of date.
4335 Frequency : systems with MD RAID1 external journal devices
4337 Description: lconf's call to blkid is confused by RAID1 journal devices
4338 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4339 of returning all block devices that match the journal UUID.
4342 Frequency : always, for aggregate stripe size over 4GB
4344 Description: assertion fails when trying to use 4GB stripe size
4345 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
4346 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
4349 Frequency : always on ppc64
4351 Description: the first write on an ext3 filesystem with mballoc got stuck
4352 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4353 perform endianness conversion.
4355 ------------------------------------------------------------------------------
4357 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4359 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
4360 INTEROPERATE with older versions automatically. Please read the
4361 user documentation before upgrading any part of a live system.
4362 * WARNING: Lustre networking configuration changes are required with
4363 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
4366 * Support for kernels:
4367 2.6.9-22.0.2.EL (RHEL 4)
4368 2.6.5-7.244 (SLES 9)
4369 2.6.12.6 vanilla (kernel.org)
4372 Severity : enhancement
4373 Bugzilla : 7981/8208
4374 Description: Introduced Lustre Networking (LNET)
4375 Details : LNET is new networking infrastructure for Lustre, it includes
4376 a reorganized network configuration mode (see the user
4377 documentation for full details) as well as support for routing
4378 between different network fabrics. Lustre Networking Devices
4379 (LNDs) for the supported network fabrics have also been
4380 created for this new infrastructure.
4382 Severity : enhancement
4383 Description: Introduced Access control lists
4384 Details : clients can set ACLs on files and directories in order to have
4385 more fine-grained permissions than the standard Unix UGO+RWX.
4386 The MDS must be started with the "-o acl" mount option.
4388 Severity : enhancement
4389 Description: Introduced filesystem quotas
4390 Details : Administrators may now establish per-user quotas on the
4393 Severity : enhancement
4395 Description: Configuration change for the XT3
4396 The PTLLND is now used to run Lustre over Portals on the XT3
4397 The configure option(s) --with-cray-portals are no longer used.
4398 Rather --with-portals=<path-to-portals-includes> is used to
4399 enable building on the XT3. In addition to enable XT3 specific
4400 features the option --enable-cray-xt3 must be used.
4405 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
4406 Details : A race between allocating a new llcd and re-getting the llcd_lock
4407 allowed another thread to grab newly-allocated llcd.
4409 Severity : enhancement
4411 Description: 2.6 OST async journal commit and locking fix to improve performance
4412 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
4413 2.6 kernels are now async as they already were in 2.4 kernels so
4414 that they can commit concurrently with the network bulk transfer.
4415 For block-allocated files the filter allocation semaphore is held
4416 to avoid filesystem fragmentation during allocation. BKL lock
4417 removed for 2.6 xattr operations where it is no longer needed.
4422 Description: lconf incorrectly determined whether two IP networks could talk
4423 Details : In some more complicated routing and multiple-network
4424 configurations, lconf will avoid trying to make a network
4425 connection to a disjoint part of the IP space. It was doing the
4426 math incorrectly for one set of cases.
4431 Description: Fix for potential infinite loop processing records in an llog.
4432 Details : If an llog record is corrupted/zeroed, it is possible to loop
4433 forever in llog_process(). Validate the llog record length
4434 and skip the remainder of the block on error.
4437 Frequency : occasional (liblustre only)
4439 Description: liblustre could not open files whose last component is a symlink
4440 Details : sysio_path_walk() would incorrectly pass the open intent to
4441 intermediate path components.
4444 Frequency : rare (liblustre only with non-standard tuning)
4445 Bugzilla : 7201 (7350)
4446 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4447 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4448 after it has been released, only do this for referenced locks
4450 Severity : enhancement
4452 Description: specifying an (invalid) directory default stripe_size of -1
4453 would reset the directory default striping
4454 Details : stripe_size -1 was used internally to signal directory stripe
4455 removal, now use "all default" to signal dir stripe removal
4456 as a directory striping of "all default" is not useful
4459 Frequency : common for large clusters running liblustre clients
4461 Description: doing an ls when liblustre clients are running is slow
4462 Details : sending a glimpse AST to a liblustre client waits for every AST
4463 to time out, as liblustre clients will not respond. Since they
4464 cannot cache data we refresh the OST lock LVB from disk instead.
4466 Severity : enhancement
4468 Description: doing an ls at the same time as file IO can be slow
4469 Details : enqueue and other "small" requests can be blocked behind many
4470 large IO requests. Create a new OST IO portal for non-IO
4471 requests so they can be processed faster.
4474 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4476 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4477 created file with mode 000
4478 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4479 allow an ftruncate() as a non-root user to a file with mode 000.
4480 Lustre now respects this flag to disable mode checks when
4481 truncating a file owned by the user
4484 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4487 Description: Revoking locks from clients that went dead or catatonic might take
4489 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4490 cancellation of such locks instant on servers without waiting for
4491 any reply from clients. Clients drops these locks when cancel
4492 notification from server is received without replying.
4495 Frequency : liblustre-only, when liblustre client dies or becomes busy
4497 Description: Doing ls on Linux clients can take a long time with active
4499 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4500 granting such locks to it in the first place if possible. Locks
4501 are taken by proxy on the OST during the read or write and
4502 dropped immediately afterward. Add connect flags handling, do
4503 not grant locks to liblustre clients for glimpse ASTs.
4505 Severity : enhancement
4507 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4508 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4509 sequential reads and behaves badly with applications doing random
4510 reads. Improve it by reading ahead at least read region, and
4511 avoiding excessive large RPC for small reads.
4513 Severity : enhancement
4515 Description: Creating more than 1000 files for a single job may cause a load
4516 imbalance on the OSTs if there are also a large number of OSTs.
4517 Details : qos_prep_create() uses an OST index reseed value that is an
4518 even multiple of the number of available OSTs so that if the
4519 reseed happens in the middle of the object allocation it will
4520 still utilize the OSTs as uniformly as possible.
4525 Description: OST or MDS may oops in ping_evictor_main()
4526 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
4527 but doesn't restart at beginning of obd_exports_timed list
4530 Severity : enhancement
4532 Description: improve by-nid export eviction on the MDS and OST
4533 Details : allow multiple exports with the same NID to be evicted at one
4534 time without re-searching the exports list.
4537 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
4539 Description: MDS may oops in groups_free()
4540 Details : in rare race conditions a newly allocated group_info struct is
4541 freed again, and this can be NULL. The 2.4 compatibility code
4542 for groups_free() checked for a NULL pointer, but 2.6 did not.
4545 Frequency : common for liblustre clients doing little filesystem IO
4546 Bugzilla : 9352, 7313
4547 Description: server may evict liblustre clients accessing contended locks
4548 Details : if a client is granted a lock or receives a completion AST
4549 with a blocking AST already set it would not reply to the AST
4550 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
4553 Frequency : lfs setstripe, only systems with more than 160 OSTs
4555 Description: unable to set striping with a starting offset beyond OST 160
4556 Details : llapi_create_file() incorrectly limited the starting stripe
4557 index to the maximum single-file stripe count.
4560 Frequency : LDAP users only
4562 Description: lconf did not handle in-kernel recovery with LDAP properly
4563 Details : lconf/LustreDB get_refs() is searching the wrong namespace
4565 Severity : enhancement
4567 Description: bind OST threads to NUMA nodes to improve performance
4568 Details : all OST threads are uniformly bound to CPUs on a single NUMA
4569 node and do their allocations there to localize memory access
4571 Severity : enhancement
4573 Description: llmount can determine client NID directly from Myrinet (GM)
4574 Details : the client NID code from gmnalnid was moved directly into
4575 llmount, removing the need to use this or specifying the
4576 client NID explicitly when mounting GM clients with zeroconf
4579 Frequency : if client is started with down MDS
4581 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
4582 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
4583 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
4584 ptlrpc_import_delayed_req() because we didn't check if the
4585 request was interrupted, and we also didn't break out of the
4586 event loop if there was no timeout
4591 Description: data loss during non-page-aligned writes to a single file from
4592 both multiple nodes and multiple threads on one node at same time
4593 Details : updates to KMS and lsm weren't protected by common lock. Resulting
4594 inconsistency led to false short-reads, that were cached and later
4595 used by ->prepare_write() to fill in partially written page,
4596 leading to data loss.
4599 Frequency : always, if lconf --abort_recovery used
4601 Description: lconf --abort_recovery fails with 'Operation not supported'
4602 Details : lconf was attempting to abort recovery on the MDT device and not
4605 Severity : enhancement
4607 Description: remove cleanup logs
4608 Details : replace lconf-generated cleanup logs with lustre internal
4609 cleanup routines. Eliminates the need for client-cleanup and
4612 Severity : enhancement
4614 Description: add support for EAs (user and system) on lustre filesystems
4615 Details : it is now possible to store extended attributes in the Lustre
4616 client filesystem, and with the user_xattr mount option it
4617 is possible to allow users to store EAs on their files also
4619 Severity : enhancement
4621 Description: Add possibility (config option) to show minimal available OST free
4623 Details : When compiled with --enable-mindf configure option, statfs(2)
4624 (and so, df) will return least minimal free space available from
4625 all OSTs as amount of free space on FS, instead of summary of
4626 free spaces of all OSTs.
4628 Severity : enhancement
4630 Description: do not expand extent locks acquired on OST-side
4631 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
4632 by server: they are not cached anyway.
4635 Frequency : when mmap is used/binaries executed from Lustre
4637 Description: Unmmap pages before throwing them away from read cache.
4638 Details : llap_shrink cache now attempts to unmap pages before discarding
4639 them (if unmapping failed - do not discard). SLES9 kernel has
4640 extra checks that trigger if this unmapping is not done first.
4645 Description: lconf didn't resolve symlinks before checking to see whether a
4646 given mountpoint was already in use
4649 Frequency : when migrating failover services
4650 Bugzilla : 6395, 9514
4651 Description: When migrating a subset of services from a node (e.g. failback
4652 from a failover service node) the remaining services would
4653 time out and evict clients.
4654 Details : lconf --force (implied by --failover) sets the global obd_timeout
4655 to 5 seconds in order to quickly disconnect, but this caused
4656 other RPCs to time out too quickly. Do not change the global
4657 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
4659 Severity : enhancement
4660 Frequency : if MDS is started with down OST
4661 Bugzilla : 9439,5706
4662 Description: Allow startup/shutdown of an MDS without depending on the
4663 availability of the OSTs.
4664 Details : Asynchronously call mds_lov_synchronize during MDS startup.
4665 Add appropriate locking and lov-osc refcounts for safe
4666 cleaning. Add osc abort_inflight calls in case the
4667 synchronize never started.
4670 Frequency : occasional (Cray XT3 only)
4672 Description: root not authorized to access files in CRAY_PORTALS environment
4673 Details : The client process capabilities were not honoured on the MDS in
4674 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
4675 been accessed by an authorized user then root was able to access
4676 the file on the local client also. The root user capabilities
4677 are now allowed on the MDS, as this environment has secure UID.
4680 Frequency : occasional
4682 Description: ldiskfs "too long searching" message happens too often
4683 Details : A debugging message (otherwise harmless) prints too often on
4684 the OST console. This has been reduced to only happen when
4685 there are fragmentation problems on the filesystem.
4690 Description: Division by zero in statfs when all OSCs are inactive
4691 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
4692 lov_statfs divides by value returned by lov_get_stripecnt().
4696 Bugzilla : 9489, 3273
4697 Description: First write from each client to each OST was only 4kB in size,
4698 to initialize client writeback cache, which caused sub-optimal
4699 RPCs and poor layout on disk for the first writen file.
4700 Details : Clients now request an initial cache grant at (re)connect time
4701 and so that they can start streaming writes to the cache right
4702 away and always do full-sized RPCs if there is enough data.
4703 If the OST is rebooted the client also re-establishes its grant
4704 so that client cached writes will be honoured under the grant.
4709 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
4710 Details : Now I/O RPCs go to different portal number and (presumably) fast
4711 lock requests (and glimses) and other RPCs get their own service
4712 threads pool that should be able to service those RPCs
4715 Severity : enhancement
4717 Description: Ability to exchange lustre version between client and servers and
4718 issue warnings at client side if client is too old. Also for
4719 liblustre clients there is ability to refuse connection of too old
4721 Details : New 'version' field is added to connect data structure that is
4722 filled with version info. That info is later checked by server and
4726 Frequency : rare, liblustre only.
4727 Bugzilla : 9296, 9581
4728 Description: Two simultaneous writes from liblustre at offset within same page
4729 might proceed at the same time overwriting eachother with stale
4731 Details : I/O lock withing llu_file_prwv was released too early, before data
4732 actually was hitting the wire. Extended lock-holding time until
4733 server acknowledges receiving data.
4736 Frequency : extremely rare. Never observed in practice.
4738 Description: avoid generating lustre_handle cookie of 0.
4739 Details : class_handle_hash() generates handle cookies by incrementing
4740 global counter, and can hit 0 occasionaly (this is unlikely, but
4741 not impossible, because initial value of cookie counter is
4742 selected randonly). Value of 0 is used as a sentinel meaning
4743 "unassigned handle" --- avoid it. Also coalesce two critical
4744 sections in this function into one.
4746 Severity : enhancement
4748 Description: allow liblustre clients to delegate truncate locking to OST
4749 Details : To avoid overhead of locking, liblustre client instructs OST to
4750 take extent lock in ost_punch() on client's behalf. New connection
4751 flag is added to handle backward compatibility.
4753 Severity : enhancement
4754 Bugzilla : 4928, 7341, 9758
4755 Description: allow number of OST service threads to be specified
4756 Details : a module parameter allows the number of OST service threads
4757 to be specified via "options ost ost_num_threads={N}" in the
4758 OSS's /etc/modules.conf or /etc/modprobe.conf.
4762 Bugzilla : 6146, 9635, 9895
4763 Description: servers crash with bad pointer in target_handle_connect()
4764 Details : In rare cases when a client is reconnecting it was possible that
4765 the connection request was the last reference for that export.
4766 We would temporarily drop the export reference and get a new
4767 one, but this may have been the last reference and the export
4768 was just destroyed. Get new reference before dropping old one.
4770 Severity : enhancement
4771 Frequency : if client is started with failover MDS
4773 Description: Allow multiple MDS hostnames in the mount command
4774 Details : Try to read the configuration from all specified MDS
4775 hostnames during a client mount in case the "primary"
4778 Severity : enhancement
4780 Description: Stop sending data to evicted clients as soon as possible.
4781 Details : Check if the client we are about to send or are sending data to
4782 was evicted already. (Check is done every second of waiting,
4783 for which l_wait_event interface was extended to allow checking
4784 of exit condition at specified intervals).
4787 Frequency : rare, normally only when NFS exporting is done from client
4789 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
4791 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
4794 Frequency : always, if srand() is called before liblustre initialization
4796 Description: Liblustre uses system PRNG disturbing its usage by user application
4797 Details : Introduce internal to lustre fast and high-quality PRNG for
4798 lustre usage and make liblustre and some other places in generic
4799 lustre code to use it.
4801 Severity : enhancement
4802 Bugzilla : 9477, 9557, 9870
4803 Description: Verify that the MDS configuration logs are updated when xml is
4804 Details : Check if the .xml configuration logs are newer than the config
4805 logs stored on the MDS and report an error if this is the case.
4806 Request --write-conf, or allow starting with --old_conf.
4808 Severity : enhancement
4810 Description: Handle symlinks in the path when checking if Lustre is mounted.
4811 Details : Resolve intermediate symlinks when checking if a client has
4812 mounted a filesystem to avoid duplicate client mounts.
4817 Description: lconf can hit an error exception but still return success.
4818 Details : The lconf command catches the Command error exception at the top
4819 level script context and will exit with the associated exit
4820 status, but doesn't ensure that this exit status is non-zero.
4825 Description: failure of ptlrpc thread startup can cause oops
4826 Details : Starting a ptlrpc service thread can fail if there are a large
4827 number of threads or the server memory is very fragmented.
4828 Handle this without oopsing.
4831 Frequency : always, only if liblustre and non-default acceptor port was used
4833 Description: liblustre cannot connect to servers with non-default acceptor port
4834 Details : tcpnal_set_default_params() was not called and was therefore
4835 ignoring the environment varaible TCPNAL_PORT, as well as other
4836 TCPNAL_ environment variables
4841 Description: two objects could be created on the same OST for a single file
4842 Details : If an OST is down, in some cases it was possible to create two
4843 objects on a single OST for a single file. No problems other
4844 than potential performance impact and spurious error messages.
4848 Bugzilla : 5681, 9562
4849 Description: Client may oops in ll_unhash_aliases
4850 Details : Client dcache may become inconsistent in race condition.
4851 In some cases "getcwd" can fail if the current directory is
4857 Description: Inode refcounting problems in NFS export code
4858 Details : link_raw functions used to call d_instantiate without obtaining
4859 extra inode reference first.
4863 Bugzilla : 9942, 9903
4864 Description: Referencing freed requests leading to crash, memleaks with NFS.
4865 Details : We used to require that call to ll_revalidate_it was always
4866 followed by ll_lookup_it. Also with revalidate_special() it is
4867 possible to call ll_revalidate_it() twice for the same dentry
4868 even if first occurence returned success. This fix changes semantic
4869 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
4870 reference on a request referred from the intent.
4871 ll_intent_release() then releases such a request.
4874 Frequency : rare, normally benchmark loads only
4876 Description: unlinked inodes were kept in memory on the client
4877 Details : If a client is repeatedly creating and unlinking files it
4878 can accumulate a lot of stale inodes in the inode slab cache.
4879 If there is no other client load running this can cause the
4880 client node to run out of memory. Instead flush old inodes
4881 from client cache that have the same inode number as a new inode.
4884 Frequency : SLES9 2.6.5 kernel and long filenames only
4885 Bugzilla : 9969, 10379
4886 Description: utime reports stale NFS file handle
4887 Details : SLES9 uses out-of-dentry names in some cases, which confused
4888 the lustre dentry revalidation. Change it to always use the
4892 Frequency : rare, unless heavy write-truncate concurrency is continuous
4893 Bugzilla : 4180, 6984, 7171, 9963, 9331
4894 Description: OST becomes very slow and/or deadlocked during object unlink
4895 Details : filter_destroy() was holding onto the parent directory lock
4896 while truncating+unlinking objects. For very large objects this
4897 may block other threads for a long time and slow overall OST
4898 responsiveness. It may also be possible to get a lock ordering
4899 deadlock in this case, or run out of journal credits because of
4900 the combined truncate+unlink. Solution is to do object truncate
4901 first in one transaction without parent lock, and then do the
4902 final unlink in a new transaction with the parent lock. This
4903 reduces the lock hold time dramatically.
4906 Frequency : rare, 2.4 kernels only
4908 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
4909 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
4910 before unmounting their filesystem in order to avoid deadlock.
4911 The kernel_locked() function in 2.4 kernels only checks whether
4912 the kernel lock is held, not whether it is this process that is
4913 holding it as 2.6 kernels do.
4918 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
4919 Details : The client ptlrpc code may be trying to reconnect to a down
4920 server before a previous connection attempt has timed out.
4921 Increase the reconnect interval to be longer than the connection
4922 timeout interval to avoid sending duplicate connections to
4926 Frequency : echo_client brw_test command
4928 Description: fix echo_client to work with OST preallocated code
4929 Details : OST preallocation code (5137) didn't take echo_client IO path
4930 into account: echo_client calls filter methods outside of any
4931 OST thread and, hence, there is no per-thread preallocated
4932 pages and buffers to use. Solution: hijack pga pages for IO. As
4933 a byproduct, this avoids unnecessary data copying.
4937 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
4938 Description: Client can oops in mdc_commit_close() after open replay
4939 Details : It was possible for the MDS to return an open request with no
4940 transaction number in mds_finish_transno() if the client was
4941 evicted, but without actually returning an error. Clients
4942 would later try to replay that open and may trip an assertion
4943 Simplify the client close codepath, and always return an error
4944 from the MDS in case the open is not successful.
4947 Frequency : rare, 2.6 OSTs only
4949 Description: OST may deadlock under high load on fragmented files
4950 Details : If there was a heavy load and highly-fragmented OST filesystems
4951 it was possible to have all the OST threads deadlock waiting on
4952 allocation of biovecs, because the biovecs were not released
4953 until the entire RPC IO was completed. Instead, release biovecs
4954 as soon as they are complete to ensure forward IO progress.
4956 Severity : enhancement
4958 Description: Support for specifying external journal device at mount
4959 Details : If an OST or MDS device is formatted with an external journal
4960 device, this device major/minor is stored in the ext3 superblock
4961 and may not be valid for failover. Allow detecting and
4962 specifying the external journal at mount time.
4967 Description: Mounting an MDS with pending unlinked files may cause oops
4968 Details : target_finish_recovery() calls mds_postrecov() which returned
4969 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
4970 considers this an error and immediately begins cleaning up the
4971 lov, just after starting the mds_lov process
4973 Severity : enhancement
4975 Description: Implement 'lfs df' to report actual free space on per-OST basis
4976 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
4977 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
4978 usage of objects; '-h' to report in human readable format.
4980 ------------------------------------------------------------------------------
4982 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
4989 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
4990 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
4991 at startup during orphan unlinking. Instead, skip buddy update
4992 before setup, as it will be regenerated after recovery is complete.
4997 Description: If an OST is inactive, its locks might reference stale inodes.
4998 Details : lov_change_cbdata() must iterate over all namespaces, even if
4999 they are inactive to clear inode references from the lock.
5001 Severity : enhancement
5002 Frequency : occasional, if non-standard max_dirty_mb used
5004 Description: Client will block write RPCs if not enough grant
5005 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
5006 then the client will block writes while waiting for another RPC
5007 to complete instead of consuming its dirty limit. With change
5008 we get improved performance when max_dirty_mb is small.
5010 Severity : enhancement
5011 Bugzilla : 3389, 6253
5012 Description: Add support for supplementary groups on the MDS.
5013 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
5014 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
5015 lookups for user supplementary groups into a cache.
5019 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
5020 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
5021 for liblustre O_EXCL is enforced by MDS.
5025 Description: readdir never returns NULL in liblustre.
5026 Details : Corrected llu_iop_getdirentries logic, to return offset of next
5027 dentry in struct dirent.
5031 Frequency : liblustre only, depends on application IO pattern
5032 Description: liblustre clients evicted if not contacting servers
5033 Details : Don't put liblustre clients into the ping_evictor list, so
5034 they will not be evicted by the pinger ever.
5036 Severity : enhancement
5038 Description: Add ability to evict clients by NID from MDS.
5039 Details : By echoing "nid:$NID" string into
5040 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
5041 $NID would be instantly evicted from this MDS and from all active
5042 OSTs connected to it.
5046 Description: Do not query file size twice, somewhat slowing stat(2) calls.
5047 Details : lookup_it_finish() used to query file size from OSTs that was not
5052 Description: service threads change working directory to that of init
5053 Details : Starting lustre service threads may pin the working directory
5054 of the parent thread, making that filesystem busy. Threads
5055 now change to the working directory of init to avoid this.
5059 Frequency : during shutdown only
5060 Description: shutdown with a failed MDS or OST can cause unmount to hang
5061 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
5066 Frequency : 2.6 only
5067 Description: chmod/chown may include an extra supplementary group
5068 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
5069 supplementary group and if none is specified this is used.
5072 Bugzilla : 5479 (6816)
5074 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
5075 Details : If lookup is in progress on a file that is unlinked we might try
5076 to revalidate the inode and fail in revalidate after lookup is
5077 complete and ll_file_open() enqueues the open again but
5078 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
5081 Frequency : always, if lconf --abort_recovery used
5083 Description: lconf --abort_recovery fails with 'Operation not supported'
5084 Details : lconf was attempting to abort recovery on the MDT device and not
5087 ------------------------------------------------------------------------------
5089 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
5094 Frequency : rare (only unsupported configurations with a node running as an
5096 Bugzilla : 6514, 5137
5097 Description: Mounting a Lustre file system on a node running as an OST could
5099 Details : OSTs now preallocates memory needed to write out data at
5100 startup, instead of when needed, to avoid having to
5101 allocate memory in possibly low memory situations.
5102 Specifically, if the file system is mounted on on OST,
5103 memory pressure could force it to try to write out data,
5104 which it needed to allocate memory to do. Due to the low
5105 memory, it would be unable to do so and the node would
5106 become unresponsive.
5108 Severity : enhancement
5110 Description: Addition of lconf --service command line option
5111 Details : lconf now accepts a '--service <arg>' option, which is
5112 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
5114 Severity : enhancement
5116 Description: Failover mode is now the default for OSTs.
5117 Details : By default, OSTs will now run in failover mode. To return to
5118 the old behaviour, add '--failout' to the lmc line for OSTs.
5120 Severity : enhancement
5122 Description: Health checks are now provided for MDS and OSTs
5123 Details : Additional detailed health check information on MSD and OSTs
5124 is now provided through the procfs health_check value.
5127 Frequency : occasional, depends on IO load
5129 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
5130 after numerous create/delete cycles
5131 Details : The ext3 inode allocation policy would not allocate new inodes
5132 very well on the OSTs because there are no new directories
5133 being created. Instead we look for groups with free space if
5134 the parent directories are nearly full.
5139 Description: Network or server problems during mount may cause partially
5140 mounted clients instead of returning an error.
5141 Details : The config llog parsing code may overwrite the error return
5142 code during mount error handling, returning success instead
5148 Description: MDS can fail to allocate large reply buffers
5149 Details : After long uptimes the MDS can fail to allocate large reply
5150 buffers (e.g. zconf client mount config records) due to memory
5151 fragmentation or consumption by the buffer cache. Preallocate
5152 some large reply buffers so that these replies can be sent even
5153 under memory pressure.
5157 Frequency : rare (liblustre)
5158 Description: fsx running with liblustre complained that using truncate() to
5159 extend the file doesn't work. This patch corrects that issue.
5160 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
5161 ATTR_SIZE and lsm use in llu_setattr_raw.
5165 Frequency : rare, only 2.6 kernels
5166 Description: Unusual file access patterns on the MDS may result in inode
5167 data being lost in very rare circumstances.
5168 Details : Bad interaction between the ea-in-inode patch and the "no-read"
5169 code in the 2.6 kernel caused the inode and/or EA data not to
5170 be read from disk, causing single-file corruption.
5174 Frequency : rare, only 2.6 filesystems using extents
5175 Description: Heavy concurrent write and delete load may cause data corruption.
5176 Details : It was possible under high-load situations to have an extent
5177 metadata block in the block device cache from a just-unlinked
5178 file overwrite a newly-allocated data block. We now unmap any
5179 metadata buffers that alias just-allocated data blocks.
5183 Frequency : filesystems with default stripe_count larger than 77
5184 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
5185 Details : lconf specifies an inode size of 4096 bytes when the default
5186 stripe_count is larger than 77. This conflicts with the default
5187 inode density of 1 per 4096 bytes. Allocate smaller inodes in
5188 this case to avoid pinning too much memory for large EAs.
5190 ------------------------------------------------------------------------------
5192 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
5197 Frequency : rare (extremely heavy IO load with hundreds of clients)
5199 Description: Client is evicted, gets IO error writing to file
5200 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
5201 caused clients to be evicted for AST timeouts. The fixes in
5202 bug 5192 mean we no longer need to have such short AST timeouts
5203 so ldlm_timeout has been increased.
5206 Frequency : occasional during --force or --failover shutdown under load
5207 Bugzilla : 5949, 4834
5208 Description: Server oops/LBUG if stopped with --force or --failover under load
5209 Details : a collection of import/export refcount and cleanup ordering
5210 issues fixed for safer force cleanup
5213 Frequency : only filesystems larger than 120 OSTs
5214 Bugzilla : 5990, 6223
5215 Description: lfs getstripe would oops on a very large filesystem
5216 Details : lov_getconfig used kfree on vmalloc'd memory
5219 Frequency : only filesystems exporting via NFS to Solaris 10 clients
5220 Bugzilla : 6242, 6243
5221 Description: reading from files that had been truncated to a non-zero size
5222 but never opened returned no data
5223 Details : ll_file_read() reads zeros from no-object files to EOF
5228 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
5229 Details : There's little harm in aborting MDS/OSS recovery and letting it
5230 try again, so I removed the LASSERT and return an error instead.
5232 Severity : enhancement
5234 Description: New debugging infrastructure for tracking down data corruption
5235 Details : The I/O checksum code was replaced to: (a) control it at runtime,
5236 (b) cover more of the client-side code path, and (c) try to narrow
5237 down where problems occurred
5241 Bugzilla : 3819, 4364, 4397, 6313
5242 Description: Racing close and eviction MDS could cause assertion in mds_close
5243 Details : It was possible to get multiple mfd references during close and
5244 client eviction, leading to one thread referencing a freed mfd.
5246 Severity: : enhancement
5247 Bugzilla : 3262, 6359
5248 Description: Attempts to reconnect to servers are now more aggressive.
5249 Details : This builds on the enhanced upcall-less recovery that was added
5250 in 1.4.2. When trying to reconnect to servers, clients will
5251 now try each server in the failover group every 10 seconds. By
5252 default, clients would previously try one server every 25 seconds.
5257 Description: After recovery, certain operations trigger a failed
5258 assertion on a client.
5259 Details : Failing over an mds, using lconf -d --failover, while a
5260 client was doing a readdir() call would cause the client to
5261 LBUG after recovery completed and the readdir() was resent.
5263 Severity : enhancement
5265 Description: Default groups are now added by lconf
5266 Details : You can now run lconf --group <servicename> without having to
5267 manually add groups with lmc.
5270 Frequency : occasional
5272 Description: Nodes with an elan id of 0 trigger a failed assertion
5275 Frequency : always when accessing e.g. tty/console device nodes
5277 Description: tty and some other devices nodes cannot be used on lustre
5278 Details : file's private_data field is used by device data and lustre
5279 values in there got lost. New field was added to struct file to
5280 store fs-specific private data.
5283 Frequency : when exporting Lustre via NFS
5285 Description: NFSD failed occasionally when looking up a path component
5286 Details : NFSD is looking up ".." which was broken in ext3 directories
5287 that had grown large enough to become hashed.
5290 Frequency : Clusters with multiple interfaces not on the same subnet
5292 Description: Nodes will repeatedly try to reconnect to an interface which it
5293 cannot reach and report an error to the log.
5294 Details : Extra peer list entries will be created by lconf with some peers
5295 unreachable. lconf now validates the peer before adding it.
5298 Frequency : Only if a default stripe is set on the filesystem root.
5300 Description: Setting a default stripe on the filesystem root prevented the
5301 filesystem from being remounted.
5302 Details : The client was sending extra request flags in the root getattr
5303 request and did not allocate a reply buffer for the dir EA.
5306 Frequency : occasional, higher if lots of files are accessed by one client
5307 Bugzilla : 6159, 6097
5308 Description: Client trips assertion regarding lsm mismatch/magic
5309 Details : While revalidating inodes the VFS looks up inodes with ifind()
5310 and in rare cases can find an inode that is being freed.
5311 The ll_test_inode() code will free the lsm during ifind()
5312 when it finds an existing inode and then the VFS later attaches
5313 this free lsm to a new inode.
5317 Bugzilla : 6422, 7030
5318 Description: MDS deadlock between mkdir and client eviction
5319 Details : Creating a new file via mkdir or mknod (starting a transaction
5320 and getting the ns lock) can deadlock with client eviction
5321 (gets ns lock and trying to finish a synchronous transaction).
5324 Frequency : occasional
5325 Description: While starting a server, the fsfilt_ext3 module could not be
5327 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
5328 kernels. Previously, lconf would still use the ext3 name
5329 when trying to load modules. Now, it will correctly use
5330 ext3 on 2.4 and ldiskfs on 2.6.
5332 Severity : enhancement
5333 Description: The default stripe count has been changed to 1
5334 Details : The interpretation of the default stripe count (0, to lfs
5335 or lmc) has been changed to mean striping across a single
5336 OST, rather than all available. For general usage we have
5337 found a stripe count of 1 or 2 works best.
5339 Severity : enhancement
5340 Description: Add support for compiling against Cray portals.
5341 Details : Conditional compiling for some areas that are different
5345 Frequency : occasional
5346 Bugzilla : 6409, 6834
5347 Description: Creating files with an explicit stripe count may lead to
5348 a failed assertion on the MDS
5349 Details : If some OSTs are full or unavailable, creating files may
5350 trigger a failed assertion on the MDS. Now, Lustre will
5351 try to use other servers or return an error to the
5355 Frequency : occasional
5357 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
5358 caused lock timeout during readahead (since 1.4.2).
5359 Details : Processes doing readahead might match a lock that hasn't been
5360 granted yet if there are overlapping and conflicting lock
5361 requests. The readahead process waits on ungranted lock
5362 (original lock is CBPENDING), while OST waits for that process
5363 to cancel CBPENDING read lock and eventually evicts client.
5365 Severity : enhancement
5367 Description: Initial enabling of flock support for clients
5368 Details : Implements fcntl advisory locking and file status functions.
5369 This feature is provided as an optional mount flag (default
5370 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
5371 locking are implemented yet, and those that are are not guaranteed
5372 to be completely correct in production environments.
5373 mount -t lustre -o [flock|noflock] ...
5376 Frequency : occasional
5378 Description: OSTs running 2.4 kernels but with extents enabled might trip an
5379 assertion in the ext3 JBD (journaling) layer.
5380 Details : The b_committed_data struct is protected by the big kernel lock
5381 in 2.4 kernels, serializing journal_commit_transaction() and
5382 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
5383 access to this struct. In 2.6 kernels there is finer grained
5384 locking to improve SMP performance of the JBD layer.
5388 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
5393 Description: merge and cleanup kernel patches.
5396 -----------------------------------------------------------------------------
5398 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
5400 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
5401 versions of Lustre. Please update all servers and clients to
5402 version 1.4.2 or later at the same time. You must also run
5403 "lconf --write-conf {config}.xml" on the MDS while it is stopped
5404 to update the configuration logs.
5406 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
5407 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
5408 - swab llog records of type '0' so we get proper header size/idx (5861)
5409 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
5410 - fix rename of one directory over another leaking an inode (5953)
5411 - avoid SetPageDirty on 2.6 (5981)
5412 - don't re-add just-being-destroyed locks to the waiting list (5653)
5413 - when creating new directories, inherit the parent's custom
5414 striping settings if present parent (3048)
5415 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
5416 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
5417 - don't hold client locks on temporary worklist from l_lru (5666)
5418 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
5419 - automatically evict dead clients (5921)
5420 - Update file size properly in create+truncate+fstat case (6196)
5421 - Do not unhash mountpoint dentries, do not allow removal of
5423 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
5424 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
5425 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
5426 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
5427 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
5428 - Correctly handle reads of files with no objects (6243)
5429 - lctl recover will also mark a device active if deactivate used (5933)
5431 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
5432 - allow --write-conf on an MDS with different nettype than client (5619)
5433 - don't write config llogs to MDS for mounts not from that MDS (5617)
5434 - lconf should create multiple TCP connections from a client (5201)
5435 - init scripts are now turned off by default; run chkconfig --on
5436 lustre and chkconfig --on lustrefs to use them
5437 - upcalls are no longer needed for clients to recover to failover
5439 - add --abort-recovery option to lconf to abort recovery on device
5441 - add support for an arbitrary number of OSTs (3026)
5442 - Quota support protocol changes.
5443 - forward compatibility changes to wire structs (6007)
5444 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5445 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5446 - improved "lustre status" script
5447 - initialize blocksize for non-regular files (6062)
5448 - added --disable-server and --disable-client configure options (5782)
5449 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5450 - Vanilla 2.4.29 support
5451 - increase maximum number of obd devices to 520 (6242)
5452 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5453 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5454 SLES 9 2.6.5 kernels
5455 - limit stripes per file to 160 (the maximum EA size) (6093)
5457 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5460 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5461 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5462 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5463 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5464 - small ext3 extents cleanups and fixes (5733)
5465 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5466 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5467 - avoid needless client->OST connect, fix handle mismatch (5317)
5468 - fix DLM error path that led to out-of-sync client, long delays (5779)
5469 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5470 - fix several locking issues related to i_size (5492,5624,5654,5672)
5471 - don't move pending lock onto export if it is already evicted (5683)
5472 - fix kernel oops when creating .foo in unlinked directory (5548)
5473 - fix deadlock in obdfilter statistics vs. object create (5811)
5474 - use time_{before,after} to avoid timer jiffies wrap (5882)
5475 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5476 - Do not leak request if server was not able to process it (5154)
5477 - If mds_open unable to find parent dir, make that negative lookup(5154)
5478 - don't create new directories with extent-mapping (5909, 5936)
5480 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5481 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5482 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5483 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5484 - support for RHEL4 kernel on i686 (5773)
5485 - provide error messages when incompatible logs are encountered (5898)
5487 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5488 * version 1.4.0.10 (1.4.1 release candidate 1)
5490 - don't keep a lock reference when lock is not granted (4238)
5491 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5492 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5493 - drop import inflight refcount on signal_completed_replay error (5255)
5494 - unlock page after async write error during send (3677)
5495 - handle missing objects in filter_preprw_read properly (5265)
5496 - no transno return for symlink open, don't save no-trasno open (3440)
5497 - don't try to complete elan receive that already failed (4012)
5498 - free RPC server reply state on error (5406)
5499 - clean up thread from ptlrpc_start_thread() on error (5160)
5500 - readahead could read extra page into cache that wasn't ejected (5388)
5501 - prevent races in class_attach/setup/cleanup/detach (5260)
5502 - don't dereference de->d_inode after l_dput of de (5458)
5503 - use "int" for stripe value returned from lock_to_stripe (5544)
5504 - mballoc allocation and error-checking fixes in 2.6 (5504)
5505 - block device patches to fix I/O request sizes in 2.6 (5482)
5506 - look up hostnames for IB nals (5602)
5507 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5508 - don't start multiple acceptors for the same port (5277)
5509 - fix incorrect LASSERT in mds_getattr_name (5635)
5510 - export a proc file for general "ping" checking (5628)
5511 - fix "lfs check" to not block when the MDS is down (5628)
5513 - service request history (4965)
5514 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5515 - create an "evict_client" /proc entry on OSTs, like the MDS has
5516 - fix mount usage message, return errors per mount(8) (5168)
5517 - change grep [] to grep "[]" in tests so they work in more UMLs
5518 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5519 - remove ancient LOV_MAGIC_V0 EA support (5047)
5520 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5521 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5522 - allow readahead allocations to fail when low on memory (5383)
5523 - mmap locking landed again, after considerable improvement (2828)
5524 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
5526 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
5529 - send OST transaction number in read/write reply to free req (4966)
5530 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5531 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5532 - the watchdog thread now runs as interruptible (5246)
5533 - flock/lockf fixes (but it's still disabled, pending 5135)
5534 - don't use EXT3 constants in llite code (5094)
5535 - memory shortage at startup could cause assertion (5176)
5537 - reorganization of lov code
5538 - single portals codebase
5540 - add extents/mballoc support (5025)
5541 - direct I/O reads in the obdfilter (4048)
5542 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
5544 tbd Cluster File Systems, Inc. <info@clusterfs.com>
5547 - send OST transaction number in read/write reply to free req (4966)
5548 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5549 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5550 - the watchdog thread now runs as interruptible (5246)
5551 - handle missing objects in filter_preprw_read properly (5265)
5552 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5553 - drop import inflight refcount on signal_completed_replay error (5255)
5554 - unlock page after async write error during send (3677)
5555 - return original error code on reconstructed replies (3761)
5556 - no transno return for symlink open, don't save no-trasno open (3440)
5558 - add pid to ldlm debugging output (4922)
5559 - bump the watchdog timeouts -- we can't handle 30sec yet
5560 - extra debugging for orphan dentry/inode bug (5259)
5562 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
5565 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
5566 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
5567 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
5568 - fix NULL dereference in /proc/sys/portals/routes (4827)
5569 - allow failed mdc_close() operations to be interrupted (4561)
5570 - stop precreate on OST before MDS would time out on it (4778)
5571 - don't send partial-page writes before EOF from client (4410)
5572 - discard client grant for sub-page writes on large-page clients (4520)
5573 - don't free dentries not owned by NFS code, check generation (4806)
5574 - fix lsm leak if mds_create_objects() fails (4801)
5575 - limit debug_daemon file size, always print CERROR messages (4789)
5576 - use transno after validating reply (3892)
5577 - process timed out requests if import state changes (3754)
5578 - update mtime on OST during writes, return in glimpse (4829)
5579 - add mkfsoptions to LDAP (4679)
5580 - use ->max_readahead method instead of zapping global ra (5039)
5581 - don't interrupt __l_wait_event() during strace
5583 - add software watchdogs to catch hung threads quickly (4941)
5584 - make lustrefs init script start after nfs is mounted
5585 - fix CWARN/ERROR duplication (4930)
5586 - return async write errors to application if possible (2248)
5587 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
5588 - print NAL number in %x format (4645)
5589 - update barely-supported suse-2.4.21-171 series (4842)
5590 - support for sles 9 %post scripts
5591 - support for building 2.6 kernel-source packages
5592 - support for sles km_* packages
5594 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
5597 - ignore -ENOENT errors in osc_destroy (3639)
5598 - notify osc create thread that OSC is being cleaned up (4600)
5599 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
5600 - reconstruct ost_handle() like mds_handle() (4657)
5601 - create a new thread to do import eviction to avoid deadlock (3969)
5602 - let lconf resolve symlinked-to devices (4629)
5603 - don't unlink "objects" from directory with default EA (4554)
5604 - hold socknal file ref over connect in case target is down (4394)
5605 - allow more than 32000 subdirectories in a single directory (3244)
5606 - fix blocks count for O_DIRECT writes (3751)
5607 - OST returns ENOSPC from object create when no space left (4539)
5608 - don't send truncate RPC if file size isn't changing (4410)
5609 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
5610 - bind to privileged port in socknal and tcpnal (3689)
5612 - rate limit CERROR/CWARN console message to avoid overload (4519)
5613 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
5614 - basic mmap support (3918)
5615 - kernel patch series update from b1_4 (4711)
5617 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
5620 - avoid crash during MDS cleanup with OST shut down (2775)
5621 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
5622 - don't use bad inodes on the MDS (3744)
5623 - dynamic object preallocation to improve recovery speed (4236)
5624 - don't hold spinlock over lock dumping or change debug flags (4401)
5625 - don't zero obd_dev when it is force cleaned (3651)
5626 - print grants to console if they go negative (4431)
5627 - "lctl deactivate" will stop automatic recovery attempts (3406)
5628 - look for existing locks in ldlm_handle_enqueue() (3764)
5629 - don't resolve lock handle twice in recovery avoiding race (4401)
5630 - revalidate should check working dir is a directory (4134)
5632 - don't always mark "slow" obdfilter messages as errors (4418)
5634 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
5637 - don't close LustreDB during write_conf until it is done (3860)
5638 - fix typo in lconf for_each_profile (3821)
5639 - allow dumping logs from multiple threads at one time (3820)
5640 - don't allow multiple threads in OSC recovery (3812)
5641 - fix debug_size parameters (3864)
5642 - fix mds_postrecov to initialize import for llog ctxt (3121)
5643 - replace config semaphore with spinlock (3306)
5644 - be sure to send a reply for a CANCEL rpc with bad export (3863)
5645 - don't allow enqueue to complete on a destroyed export (3822)
5646 - down write_lock before checking llog header bitmap (3825)
5647 - recover from lock replay timeout (3764)
5648 - up llog sem before sending rpc (3652)
5649 - reduce ns lock hold times when setting kms (3267)
5650 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
5651 - fix NULL deref and obd_dev leak on setup error (3312)
5652 - replace some LBUG about llog ops with error handling (3841)
5653 - don't match INVALID dentries from d_lookup and spin (3784)
5654 - hold dcache_lock while marking dentries INVALID and hashing (4255)
5655 - fix invalid assertion in ptlrpc_set_wait (3880)
5657 - add libwrap support for the TCP acceptor (3996)
5658 - add /proc/sys/portals/routes for non-root route listing (3994)
5659 - allow setting MDS UUID in .xml (2580)
5660 - print the stack of a process that LBUGs (4228)
5662 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
5665 - don't cleanup request in ll_file_open() on failed MDS open (3430)
5666 - make sure to unset replay flag from failed open requests (3440)
5667 - if default stripe count is 0, use OST count for inode size (3636)
5668 - update parent mtime/ctime on client for create/unlink (2611)
5669 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
5670 - free recovery state on server during a forced cleanup (3571)
5671 - unregister_reply for resent reqs (3063)
5672 - loop back devices mounting and status check on 2.6 (3563)
5673 - fix resource-creation race that can provoke i_size == 0 (3513)
5674 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
5675 - more debugging for page-accounting assertion (3746)
5676 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
5677 - avoid deadlock after precreation failure (3758)
5678 - fix race and lock order deadlock in orphan handling (3450, 3750)
5679 - add validity checks when grabbing inodes from l_ast_data (3599)
5681 - add /proc/.../recovery_status to obdfilter (3428)
5682 - lightweight CDEBUG infrastructure, debug daemon (3668)
5683 - change default OSC RPC parameters to be better on small clusters
5684 - turn off OST read cache for files smaller than 32MB
5685 - install man pages and include them in rpms (3100)
5686 - add new init script for (un)mounting lustre filesystems (2593)
5687 - run chkconfig in %post for init scripts (3701)
5688 - drop scimac NAL (unmaintained)
5690 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
5693 - clean kiobufs before and after use (3485)
5694 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
5695 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
5696 - change init script to fail more clearly if not run as root (1528)
5697 - allow clients to reconnect during replay (1742)
5698 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
5699 - don't do DNS lookups on NIDs too small for IP addresses (3442)
5700 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
5701 - fix cond_resched (3554)
5702 - only evict unfinished clients after recovery (3515)
5703 - allow bulk resend, prevent data loss (3570)
5704 - dynamic ptlrpc request buffer allocation (2102)
5705 - don't allow unlinking open directory if it isn't empty (2904)
5706 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
5707 - remove extraneous obd dereference causing LASSERT failure (3334)
5708 - don't use get_cycles() when creating temp. files on the mds (3156)
5709 - hold i_sem when setting i_size in ll_extent_lock() (3564)
5710 - handle EEXIST for set-stripe, set proper directory name (3336)
5712 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
5713 - fix ksocknal_fmb_callback() error messages (2918)
5715 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
5718 - don't copy lvb into (possibly NULL) reply on error (2983)
5719 - don't deref dentry after dput, don't free lvb on error (2922)
5720 - use the kms to determine writeback rpc length (2947)
5721 - increment oti_logcookies when osc is inactive (2948)
5722 - update client's i_blocks count via lvb messages (2543)
5723 - handle intent open/close of special files properly (1557)
5724 - mount MDS with errors=remount-ro, like obdfilter (2009)
5725 - initialize lock handle to avoid ASSERT on error cleanup (3057)
5726 - don't use cancelling-locks' kms values (2947)
5727 - use highest lock extent for kms, not last one (2925)
5728 - don't dereference ERR_PTR() dentry in error handling path (3107)
5729 - fix thread race in portals_debug_dumplog() (3122)
5730 - create lprocfs device entries at setup instead of at attach (1519)
5731 - common AST error handler, don't evict client on completion race (3145)
5732 - zero nameidata in detach_mnt in 2.6 (3118)
5733 - verify d_inode after revalidate_special is valid in 2.6 (3116)
5734 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
5735 - initialize RPC timeout timer earlier for 2.6 (3219)
5736 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
5737 - print nal/nid for unknown nid (3258)
5738 - additional checks for oscc recovery before doing precreate (3284)
5739 - fix ll_extent_lock() error return code for 64-bit systems (3043)
5740 - don't crash in mdc_close for bad permissions on open (3285)
5741 - zero i_rdev for non-device files (3147)
5742 - clear page->private before handing to FS, better assertion (3119)
5743 - tune the read pipeline (3236)
5744 - fix incorrect decref of invalidated dentry (2350)
5745 - provide read-ahead stats and refine rpc in flight stats (3328)
5746 - don't hold journal transaction open across create RPC (3313)
5747 - update atime on MDS at close time (3265)
5748 - close LDAP connection when recovering to avoid server load (3315)
5749 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
5750 - don't leak open file on MDS after open resend (3325)
5751 - serialize filter_precreate and filter_destroy_precreated (3329)
5752 - loop device shouldn't call sync_dev() for nul device (3092)
5753 - clear page cache after eviction (2766)
5754 - resynchronize MDS->OST in background (2824)
5755 - refuse to mount the same filesystem twice on same mountpoint (3394)
5756 - allow llmount to create routes for mounting behind routers (3320)
5757 - push lock cancellation to blocking thread for glimpse ASTs (3409)
5758 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
5759 - fix rare problem with rename on htree directories (3417)
5761 - allow default OST striping configuration per directory (1414)
5762 - fix compilation for qswnal for 2.6 kernels (3125)
5763 - increase maximum number of MDS request buffers for large systems
5764 - change liblustreapi to be useful for external progs like lfsck (3098)
5765 - increase local configuration timeout for slow disks (3353)
5766 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
5768 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5771 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
5772 - don't overwrite extent policy data in reply if lock was blocked (2901)
5773 - drop filter export grants atomically with removal from device (2663)
5774 - del obd_self_export from work_list in class_disconnect_exports (2908)
5775 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
5776 - swab reply message in mdc_close, other PPC fixes (2464)
5777 - fix destroying of named logs (2325)
5778 - overwrite old logs when running lconf --write_conf (2264)
5779 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
5780 - fix race in target_handle_connect (2898)
5781 - mds_reint_create() should take same inode create lock (2926)
5782 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
5783 - don't close files for self_export to avoid uninitialized obd (2936)
5784 - allow MDS with the same name as client node (2939)
5785 - hold dentry reference for closed log files for unlink (2325)
5786 - reserve space for all logs during transactions (2059)
5787 - don't evict page beyond end of stripe extent (2925)
5788 - don't oops on a deleted current working directory (2399)
5789 - handle hard links to targets without a parent properly (2517)
5790 - don't dereference NULL lock when racing during eviction (2867)
5791 - don't grow lock extents when lots of conflicting locks (2919)
5793 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
5796 - account for cache space usage on clients to avoid data loss (974)
5797 - lfsck support in lustre kernel code (2349)
5798 - reduce journal credits needed for BRW writes (2370)
5799 - orphan handling to avoid losing space on client/server crashes
5800 - ptlrpcd can be blocked, stopping ALL progress (2477)
5801 - use lock value blocks to assist in proper KMS, faster stat (1021)
5802 - takes i_sem instead of DLM locks internally on obdfilter (2720)
5803 - recovery for initial connections (2355)
5804 - fixes for mds_cleanup_orphans (1934)
5805 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
5806 - block all file creations until orphan recovery completes (1901)
5807 - client remove rq_connection from request struct (2423)
5808 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
5809 - recovery timer race (2670)
5810 - mdc_close recovey bug (2532)
5811 - ptlrpc cleanup bug (2710)
5812 - mds timeout on local locks (2588)
5813 - namespace lock held during RPCs (2431)
5814 - handle interrupted sync write properly (2503)
5815 - don't try to handle a message that hasn't been replied to (2699)
5816 - client assert failure during cleanup after abort recovery (2701)
5817 - leak mdc device after failed mount (2712)
5818 - ptlrpc_check_set allows timedout requests to complete (2714)
5819 - wait for inflight reqs when ptlrpcd finishes (2710)
5820 - make sure unregistered services are removed from the srv_list
5821 - reset bulk XID's when resending them (caught by 1138 test)
5822 - unregister_bulk after timeout
5823 - fix lconf error (2694)
5824 - handle write after unfinished setstripe, stripe-only getstripe (2388)
5825 - readahead locks pages, leaves pending causing memory pressure (2673)
5826 - increase OST request buffers to 4096 on large machines (2729)
5827 - fix up permission of existing directories in simple_mkdir (2661)
5828 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
5829 - don't assign transno to errored transactions (2742)
5830 - don't delete objects on OST if given a bogus objid from MDS (2751)
5831 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
5832 - if rq_no_resend, then timeout request after recovery (2432)
5833 - fix MDS llog_logid record size, 64-bit array alignment (2733)
5834 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
5835 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
5836 - MDS recovery shouldn't delete objects that it has given out (2730)
5837 - if enqueue arrives after completion, don't clobber LVB (2819)
5838 - don't unlock pages twice when trigger_group_io returns error (2814)
5839 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
5840 - don't write pages to disk if there was an error (1450)
5841 - don't ping imports that have recovery disabled (2676)
5842 - take buffered bytes into account when balancing socknal conn (2817)
5843 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
5844 - reconnect unlink llog connection after MDS reconnects to OST (2816)
5845 - remove little-endian swabbing of llog records (1987)
5846 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
5847 - retry reposting request buffers if they fail (1191)
5848 - grow extent at grant time to avoid granting a revoked lock (2809)
5849 - lock revoke doesn't evict page if covered by a second lock (2765)
5850 - disable VM readahead to avoid reading outside lock extents (2805)
5852 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
5853 - updated kernel patches for hp-2.4.20 kernel (2681)
5855 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
5858 - fix truncated write corruption (2366)
5859 - fix for failed assertion in iopen_connect_dentry (1792,2517)
5861 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
5862 - ensure len(uuid) < 37 in lmc (1171)
5863 - fix ia64 OOPS in llog_test (2255)
5864 - zero end of page at obdfilter for partial page writes (2648)
5865 - don't leave stale dentries around after renames (bug 2428)
5866 - fix timeouts when evicting a client with a single lock held (2642)
5867 - set deadline for the initial HELLO message to drain (2634)
5868 - print out dotted-quad IP addresses in the socknal (2302)
5870 - additional debugging for MDS client eviction problem (2443)
5871 - fix mkfsoptions support for osts (2603, 2604)
5873 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
5876 - add series for the vanilla 2.6.0 kernel
5877 - add series for the vanilla 2.4.24 kernel
5878 - add series for a cray x86/64 UL kernel drop
5879 - fix xattr patches for the vanilla 2.4.19 series
5881 - generate true UUIDs in lmc (1171)
5882 - have portals stack dumping break in UML (2466)
5883 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
5884 - allocate with _NOFS in ldlm to avoid deadlock (1933)
5885 - wake callback waiting threads on client eviction (2460)
5886 - Add --ptldebug and --subsystem to lmc (1719)
5887 - update assertion to allow safe interrupt allocation
5888 - set rq_no_resend for cancel requests (2432)
5889 - recalculate ptlrpcd timeout after resend (2494)
5890 - call vfs_rmdir when removing pending directories (2368)
5891 - fix renaming a file to itself (2429)
5892 - lmc creates a default one-stripe lov (2454)
5893 - expand procfs space to handle large clusters (2326)
5894 - increase UML stack to avoid overflow
5895 - update lconf's list of debug and subsystem masks
5896 - fix lfs find --obd (2510)
5897 - /proc tunable for disabling filter read caching (2591)
5898 - stop rpm packages from altering slapd.conf (2301)
5899 - disable nagle in the socknal under 0conf (2578)
5900 - choose mds inode size based on stripe count (2572)
5901 - fix kernel-source rpm problems (2516)
5903 - add --disable-doc to avoid pdf generation (2421)
5904 - update documentation, tests, type-os, comments
5905 - avoid format warnings on ia64
5906 - remove the TOE NAL
5907 - tiny code cleanups by removing unused fields
5909 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
5912 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
5913 - remove the most blatant lies from BUILDING file (2371)
5914 - change default debug level to reasonable production setting
5915 - reduce client side cache size to reduce cache flush time
5916 - reduce max RPCs in flight to avoid unnecessary file fragmentation
5917 - make TCP zerocopy and pinger support enabled by default (2476)
5918 - sync writes completed after process exits caused crashes (2319)
5919 - maintain correct mount count on the MDS (2356)
5920 - backout 1557, because 2316 wasn't really fixed
5921 - better file I/O statistics gathering in /proc
5922 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
5923 - another kernel patch to fix zero-copy TCP function export
5924 - don't take duplicate lock when processing re-sent getattr (2420)
5925 - lctl uses obd_self_export instead of creating new conn (2353)
5926 - MDS/OST recovery case which requires object creation asserted (2425)
5927 - move lfs from /usr/sbin to /usr/bin in packages
5928 - fix race between mds_client_add and mds_client_free (2417)
5929 - use kmalloc instead of slabs in portals (2430)
5930 - don't create duplicate records when a failover MDS is present (2442)
5931 - remove unnecessary mount age check (2332)
5932 - don't remove directory inodes from locks prematurely (2451)
5933 - don't break if MDS service name is the same as hostname (2103)
5934 - fix races in client write RPC generation when cache full (2482)
5936 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
5939 - remove now-unused request->rq_obd (278)
5940 - if an allocation fails, print out how much memory we've used (1933)
5941 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
5942 - add the "configurable stack size" patch to most series files (1256)
5943 - ability to write large log records, for 100+ OST configs (2306)
5944 - fix NULL deref when filter_prep fails (2314)
5945 - fix operator precedence error in filter_sync
5946 - dynamic allocation of socknal TX descriptors (2315)
5947 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
5948 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
5949 - ensure that CWARN messages in Portals always get to the syslog
5950 - __init/__exit are not for prototype decls (ldlm_init/exit)
5951 - x86-64 compile warning fixes
5952 - fix gateway LMC keyword conflict (2318)
5953 - fix MDS lock inversions in getattr/reint paths (1844)
5954 - fix a rare lock re-ordering bug, which caused deadlock (2322)
5955 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
5956 - DLM race condition prevented some lock evictions (2328)
5957 - ENOMEM detection and retry on socknal sends (2230)
5958 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
5959 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
5960 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
5961 - better error messages when a client is rejected during recovery (1505)
5962 - avoid cancelling locks which were never granted, after failure (2330)
5963 - fix i_sem/journal inversion in mds_client_add (2333)
5964 - fix truncate/getattr lock cycle deadlock (2334)
5965 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
5966 - fix two rare exit paths which could leak an l_lock() ref (2321)
5967 - fencepost error in MDS/OST orphan recovery (2226)
5968 - make log record alignment 8 bytes (1988)
5969 - lstripe now fails when requested offset > ost_count (2237)
5970 - ensure that all kernel series have a complete list.h (1607)
5971 - fix crashes in special-file operations (2316)
5972 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
5974 - allow configurable automake binary, for testing new versions
5975 - small update to the lfs documentation
5977 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
5979 * fix negative export reference count in fsfilt_sync (2312)
5981 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
5982 * release candidate 0.9.1
5984 - orphans are moved into the PENDING directory for possible recovery
5985 - replayed opens now open by fid for orphan/rename safety (1042)
5986 - last close of an orphan inode generates a transno (683)
5987 - chdir() and mount() now pin the directory entry (1020)
5988 - avoid CERROR in normal ll_setattr_raw() error case (1500)
5989 - discard very old requests without processing them (1502)
5990 - remove some common, well-understood CERRORs (1505)
5991 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
5992 - clear "grant" flags in OST replies until OST grant code lands (1644)
5993 - fix read performance by not clobbering i_blksize on client (1598)
5994 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
5995 - make LRU size a /proc tunable, clears locks when reduced (707)
5996 - fix some lprocfs rot that prevented ptlbd from loading (1732)
5997 - server locks take references on exports now (1558)
5998 - build fixes for 2.4.20-rh trees (1663)
5999 - return an error from lov_create if all OSCs are inactive (1751)
6000 - fix import levels when a reconnect happens without a timeout (1597)
6001 - exit early from mds_open if we get a lookup error (1749)
6002 - partial page read at EOF wouldn't wait for disk before sending (1642)
6003 - avoid NULL deref in obdfilter when reading page past EOF (1592)
6004 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
6005 - fix LBUG in ll_it_open_error with rc = -2 (1861)
6006 - write/truncate lock inversion (1639)
6007 - Don't auto-load obdclass, portals modules during cleanup (1495)
6008 - fix timestamps from jumping to "now" (1763)
6009 - extra journal assertions (1648)
6010 - add an extra multiunlink test (1771)
6011 - fix read_record/write_record API (1776)
6012 - fix leak of offset_extent, possible incorrect i_size later (1772)
6013 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
6014 - Debugging for the kqswnal_get_idle_tx problems (1820)
6015 - Allow recovery to be attempted multiple times (1536)
6016 - Write out MDS last_rcvd file after it is first created (1600)
6017 - Fix tx_descriptor leak in failed transmit situations (1827)
6018 - ext3 journaling fixes for assertion failure after IO error (1871)
6019 - class_export_put() on freed export after completion AST error (1896)
6020 - Fix revalidate looping in VFS (1322)
6021 - Don't access a freed export during MDS_REINT timeout (1521)
6022 - Add open-unlink recovery support on the MDS (1673,1764)
6023 - Return an error if no MDS data was read from last_rcvd (1946)
6024 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
6025 - Don't setup a disk device that doesn't match exported UUID (317)
6026 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
6027 - avoid committing NULL handle in force close
6028 - local.sh is now a one-stripe LOV configuration
6029 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
6030 - fix ext3 htree duplicate directory entry corruption (1516)
6031 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
6032 - update to new LOV EA format (2097)
6033 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
6035 2003-06-15 Phil Schwan <phil@clusterfs.com>
6038 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
6039 - per-import/export recovery handling (958, 931, 959)
6040 - multiple last-rcvd slots, for serving multiple FSes (949)
6041 - connections are again shared between multiple imp/exports (963, 964)
6042 - "umount -f" would hang if any requests needed to be sent (393, 978)
6043 - avoid pinning large req buffer by copying for queued messages (989)
6044 - add "uuid" to "lctl device" command to help upcalls (991)
6045 - "open" RPCs with transnos would confuse recovery counters (1037)
6046 - do proper endian conversion of all wire messages (288, 340, 891)
6047 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
6048 - call waiting locks callback from LDLM recovery thread (1127, 1151)
6049 - fix ptlrpc_connection leak in target_handle_connect (1174)
6050 - fix import refcounting bug in OST and MDS cleanup (1134)
6051 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
6052 - fix very unlikely obd_types race condition (501)
6053 - remove osc_open hack for echo_client (1187)
6054 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
6055 - a failure in read_inode2 leads to deadlock (1139)
6056 - cancel ack-locks as soon as transaction is committed (1072)
6057 - fix major leaks and crashes in the bulk I/O path (937, 1057)
6058 - make sure to commitrw after any preprw to avoid deadlock (1162)
6059 - failing to execute a file in a lustre FS would lock inode (1203)
6060 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
6061 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
6062 - obd_uuid2tgt would walk off the end of the list (1255)
6063 - on IA64 the osc would give portals incorrect bulk size (1258)
6064 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
6065 - fix lock inversion caused by new llite matching code (1282)
6066 - limit the number of dirty pages on a client to 10MB (1286)
6067 - timed out locks were not being corrected cancelled (1289)
6068 - fix O_DIRECT above 4GB on IA-32 (1292)
6069 * major user-visible changes
6070 - fail out/fail over policy now controlled by the upcall (993)
6072 - add OBD_PING to check server availability and failure (954)
6073 - lustre messages are now sent in sending host order (288, 340, 891)
6074 - add eadatalen to MDS getattr reply (340)
6075 - OST read replies may contain second buffer, with per-page status (593)
6077 2003-03-11 Phil Schwan <phil@clusterfs.com>
6080 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
6081 - failed open()s could cause deadlock; fixed (867, 869)
6082 - stop cancelling OST locks when files are closed (481)
6083 - overlapping XID spaces caused network corruption (851, 853)
6084 - fix unsafe fsfilt counter arithmetic; change to atomic_t
6085 - setattr_raw added, to do single-RPC, server-side setattrs
6086 - lmc/lconf syntax change for OST UUIDs
6087 - fix crashy race condition between ptlrpc_free_req and osc_close
6088 - don't use request in mdc_enqueue if we hit a timeout (889)
6089 - don't set the inode i_size for regular files from the MDS (896)
6090 - handle out of order completion AST (842)
6091 - don't LBUG if a lock request times out after receiving AST (913)
6092 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
6093 - if a bad lock AST arrives, send an error instead of dropping entirely
6094 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
6095 - fix leak in bulk IO when only partially completed (899, 900, 926)
6096 - fix O_DIRECT for ia64 (55)
6097 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
6098 - C-z after timeout could hang a process forever; fixed (977)
6100 - client-side I/O cache (678, 924, 929, 941, 970)
6102 - READPAGE and SETATTRs which don't take server-side locks get
6105 2003-02-11 Phil Schwan <phil@clusterfs.com>
6108 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
6109 - Fix fsfilt_extN_readpage() to read a full page of directory
6110 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
6111 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
6112 - Fix imbalanced LOV object allocation and out-of-bound access (469)
6113 - Most intent operations were removed, in favour of a new RPC mode
6114 that does a single RPC to the server and bypasses most of the VFS
6115 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
6116 - Aggressively cancel local locks on DLM servers
6117 - mds_reint_unlink sends EA to the client if it's the last nlink.
6118 client uses that EA to unlink OST objects.
6119 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
6120 - recursive symlinks were fixed (439)
6121 - fixed NULL deref in DEBUG_REQ
6122 - filter_update_lastobjid no longer calls sync, which annoyed extN
6123 - fixed multi-client small-writes to a single file problem (445)
6124 - fixed mtime updates during file writes (607)
6125 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
6126 - fixed bug in obd_brw_read/write() (under guise of testing 367)
6127 - fixed Linux OST size reporting problem (444, 656)
6128 - OST now updates object mtime with writes or setattr (607, 619)
6129 - client verifies file size before zeroing page past EOF (445)
6130 - OST now writes last allocated objid to disk with allocation (108)
6131 - LOV on echo now works (409)
6133 - mds_reint_unlink sends a new buffer, with the EA included. this
6134 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
6135 set if a regular file was being unlinked, and it was the last link
6136 - use PtlGet from the target for bulk writes (315)
6137 - OST now updates object mtime with writes or setattr (607, 619)
6138 - LDLM now has a grant-time callback to revalidate locked items, if
6140 - Many MDS operations were reorganized to combat race conditions
6142 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
6143 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
6145 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
6148 - Fully reactivate OST imports after reconnection (512, others)
6149 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
6150 - More graceful error handling for truncating on dead OST (515)
6151 - Don't error out unless we're actually accessing dead stripes (474)
6152 - Fix garbage sizes when stripes are missing (410)
6153 - LRU counters were broken, causing constant lock purge (433, 432)
6154 - garbage on read from stripes with failed OSTs (441)
6155 - mark OSCs as active before reconnecting during recovery (438)
6156 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
6157 - lfind did not preserve OST order in output (443)
6158 - symlinks cause hung clients, incorrect data (439)
6159 - stop dereferencing request after dropping refcount (457)
6160 - don't LASSERT(spin_is_locked) on non-SMP (455)
6161 - fixes for many rename() bugs
6162 - fstat didn't correctly synchronize attributes (399)
6163 - server must handle lock cancellation during blocking AST prep (487)
6164 - bulk descriptors were free()d too soon (511)
6165 - fix paths in lconf, which would load incorrect modules (451, 507)
6166 - fix confusing lconf 'host not found' error message (386)
6167 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
6168 - fix race condition in mdc_blocking_ast() for inode access (526)
6169 - fix lov_unpackmd() unpacking wrong number of stripes (537)
6170 - fix lov_set_osc_active() marking wrong OSC inactive (440)
6171 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
6172 - fix multiple writes of stripe MD to MDS (358, maybe 519)
6173 - fix lstripe in several ways (kernel side) (527)
6174 - fix request leak in ldlm_cli_enqueue (262)
6175 - incorrect OSC was marked inactive after OST failure
6176 - call mds_fs_cleanup before unmounting filesystem (524)
6177 - fix races between taking ns_lock and ldlm_lock_change_resource
6178 - fix races updating LOV export open file list
6179 - fix lov_enqueue error path, avoid decref-ing bad lock handle
6180 - fix recovery NULL deref in ldlm_cli_cancel_unused
6181 - fix some DLM races by using new hash table for lock handles (419)
6182 - permit the client to specify desired inodes, at replay
6183 - duplicate requests when we queue them for replay reintegration
6184 - fix last_rcvd offset calculation
6185 - sync after each recovered transaction, so we always make progress
6186 - never, not always, ERESTART requests without transnos
6187 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
6189 - skip replay if the MDS says that the client is already connected
6190 - don't check for a recovery-enabled export to match lctl's UUID
6191 - don't INC_USE_COUNT for phantom exports
6192 - don't crash when cleaning up phantom exports (567)
6193 - don't double-finish or set replay data for errored mdc_open requests
6194 - abort requests when they time out, so we don't get old replies
6195 - send/receive replies for AST messages again
6196 - if the client says that it doesn't have the lock, cancel it on the
6198 - if we timeout during I/O, don't try to cancel an in-use lock; instead
6199 mark it as destroyed, it will all work out when decref is called
6200 - fix module use counts (22, 581)
6202 - ASTs now expect a reply (server cancels lock on error reply)
6204 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
6207 - fix many simultaneous client startup (392)
6208 - fix dentry->d_it clobbering
6209 - credentials weren't being shipped for readdir/getattr operations
6210 - remove invalid assertions triggered during some concurrent MD
6212 - proper Lustre versions added (336, 389)
6213 - fix memory leak for create error case (398)
6214 - fix LOV locking bug that would get cli/srv out of sync
6215 - fix echo client over LOV (409)
6216 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
6217 - fix double-O_EXCL intent crash (424)
6218 - avoid sending multiple lock CANCELs (352)
6220 - MDS can do multi-client recovery (modulo bugs in new code)
6222 - many updates, edits, cleanups
6224 2002-11-18 Phil Schwan <phil@clusterfs.com>
6227 - fix null d_it dereference (346)
6228 - fix full OST/dbench hang (333)
6229 - fix permission problem with file removal (286)
6230 - fix removal of OSCs from LOV when they fail
6231 - fix NULL deref during bulk timeout (214)
6232 - fix problems related to multiple filesystems on one MDS (241)
6233 - fixed serious subtle metadata locking bugs
6234 - free locks on clients when inodes are removed due to memory
6236 - fix inode pointer in lock data (285)
6237 - partial support for multiple MDS on a single host (241)
6238 - data locks weren't cancelled at clear_inode time (290, 311)
6239 - intent locks could lead to unbounded lock growth (205)
6240 - added a maximum lock count, an LRU list, and a flusher
6241 - fix multiple rename (365)
6242 - properly abstracted the echo client
6243 - OSC locked 1 byte too many; fixed
6244 - rewrote brw callback code:
6245 - fixed recovery bugs related to LOVs (306)
6246 - fixed too-many-pages-in-one-write crash (191)
6247 - fixed (again) crash in sync_io_timeout (214)
6248 - probably fixed callback-related race (385)
6250 - Add capability to MDS protocol
6251 - LDLM cancellations and callbacks on different portals
6253 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
6256 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
6257 - defer open object destruction to close time (601981, 138)
6258 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
6259 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
6260 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
6261 - don't call obd_setattr in ll_file_release for destroyed objects
6262 * protocol change to lustre_msg: move |version| and add |flags|
6263 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
6264 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
6265 * added replay of create, unlink, link and rename operations during
6266 MDS failover; recovery should be much more robust now
6267 * remove failed OSCs from LOVs (only lov_create uses this so far)
6268 * the lustre-HOWTO was brought (more) up to date (582544)
6270 2002-10-23 Phil Schwan <phil@clusterfs.com>
6273 - in-use dentries weren't being reused properly (617851)
6274 - prevent multiple LDLM setup (599178)
6275 - fix LOV size calculations for truncate (617853)
6276 - fix client handling of MDS intent errors (POSIX)
6277 - fix permission bug in lovstripe.c test (624321)
6278 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
6279 - truncate past end of file could corrupt data
6280 - proper cleanup after timeouts, crashes, etc (592524, 550815)
6281 - a race in recovery could return ETIMEDOUT to apps (623947)
6282 - building outside the source directory was fixed
6283 * the lustre-HOWTO was brought (more) up to date (582544)
6284 * major progress was made on recovery functionality
6286 2002-10-10 Phil Schwan <phil@clusterfs.com>
6289 - recovery deadlock fix
6290 - rm -rf causes LBUG fix (617817)
6291 - file open by multiple tasks fix (618962)
6292 - directory permissions bugs (602707 and 620007)
6293 - journal_stop fixed with locking (611313)
6294 - O_APPEND failures resolved (618273, perhaps 614459)
6295 - lconf PATH fix (619770)
6296 - IA64 build fix (621450)
6297 - RPC buffer sizes scale with amount of memory
6299 2002-10-01 Phil Schwan <phil@clusterfs.com>
6302 - locks would be cancelled without throwing away data pages,
6303 resulting in inconsistent data (605627)
6304 - inode attributes were not always being refreshed (605627, 612449)
6305 - lconf now continues to cleanup after lctl reports an error
6306 - MDS now enforces user permissions (602707)
6307 - lprocfs cleanup fixed, but not yet enabled (614157)
6308 - fixed infinite server hang, should a client not respond to an AST
6309 - avoid going into recovery if user calls readlink() with a buffer
6310 that's too small (613941)
6311 - AST RPCs no longer require replies (614867) -- this may be changed
6312 - don't crash server if client sends an IOV that's too big (611336)
6313 - fixed lock conversion deadlock (611892)
6314 - fixed the following of symlinks (614622)
6315 * recovery: the server can remove locks from a client that dies, other
6316 clients can make progress
6317 * more extN patch fixes
6318 * compile-time configurable ptlrpc buffer allocations
6320 - collaborative read cache document
6321 - Lustre Lite Performance CDR document-in-progress
6323 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6326 - fix typo in patch-2.4.18
6328 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6331 - clear ptlrpc request each time in handle_incoming_request()
6332 - unlink of files now destroys the object on the OST
6334 2002-09-19 Peter Braam <braam@clusterfs.com>
6336 * add hard link support
6337 * change obdfile creation method
6338 * kernel patch changed
6340 2002-09-19 Peter Braam <braam@clusterfs.com>
6343 - stack overflow bug in extN fixed
6345 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
6347 * documentation updates
6348 - add man pages for config tools
6349 - update tests/README to describe testing with new config tools
6350 - finish metadata API descriptions
6351 * bug fixes and cleanups
6352 - statfs workaround for 16TB limit
6353 - LOV stripe allocation improved, can stripe on subset of OSTs
6354 - LOV file size/IO offset was wrong for files > 4GB in size
6355 - object EA data was being dropped, caused files to be unreadable
6356 - memory overflow with non-LOV OST caused memory corruption
6357 - fixed regression tests to work with new config tools, obdfilter
6358 - fixed bug when directory size became larger than 1 block
6359 - fixed bug (for single client case) when PWD was deleted
6360 - invalidate local directory pages when doing intent-based ops
6361 - avoid LDLM oops when lock callback contained bad data
6363 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
6365 * documentation updates
6366 * bug fixes and cleanups
6367 - configuration tools
6370 - 64-bit compile warnings
6371 - 64-bit internal statfs data
6373 * test_brw on persistent OST devices
6375 * lprocfs (disabled)
6377 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
6379 * documentation updates
6380 * bug fixes and cleanups
6381 * configuration tools
6383 2002-08-30 Peter J. Braam <braam@clusterfs.com>
6386 * many small fixes to 0_5_4
6387 * io/network handling
6388 * thinkos in MDS operations
6390 2002-08-24 Peter J. Braam <braam@clusterfs.com>
6393 * crucial basic fixes to 0.5.3
6394 * IOR, Iozone work over Elan
6397 2002-08-07 Phil Schwan <phil@clusterfs.com>
6398 * version 0_5_3, our first alpha
6399 * we use the new Portals iovs
6400 * documentation updates
6401 * bug fixes and cleanups
6402 * small changes in the DLM wire protocol
6404 2002-07-25 Peter J. Braam <braam@clusterfs.com>
6405 * version 0_5_1 with some initial stability,
6406 * locking on MD and file I/O.
6407 * documentation updates
6408 * several bug fixes since 0.5.0
6409 * small changes in wire protocol
6411 2002-07-18 Phil Schwan <phil@clusterfs.com>
6413 * delivered as Lustre Light Alpha
6414 * fixed a crash after handling invalid MDS requests
6415 * fixed directory pages for architectures with non-4k pages sizes
6417 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
6418 * release version v0_4_4
6419 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
6421 * Fixes a number of interruption problems with OST operations.
6422 * Update documentation for portals header changes
6423 * Move all wire protocol structs/defines to lustre_idl.h
6424 * Fixes symlink length bug.
6425 * Add tcpdump to repository.
6427 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
6428 * release version v0_4_3
6429 * Fixes statfs for inodes on extN.
6430 * Fixes bug in runtests which would delete /etc/hosts.
6431 * Use 64-bit object IDs wherever possible (not into VFS though)
6432 Remove ost_get_info, which is unused by lustre, and out of date.
6434 2002-07-03 Peter Braam <braam@clusterfs.com>
6435 * release version v0_4_2 Fixes a lookup error (type not passed)
6436 * move forward to head of Portals
6437 * move forward to latest Lustre kernel
6439 2002-06-25 Peter Braam <braam@clusterfs.com>
6440 * release version v0_4_1. Hopefully stable on single node use.