1 09-03-2014 Intel Corporation
3 * See http://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
4 * Server support for kernels:
5 2.6.32-431.29.2.el6 (RHEL6)
6 3.0.101-0.35 (SLES11SP3)
7 * Client support for unpatched kernels:
8 2.6.32-431.29.2.el6 (RHEL6)
9 3.0.101-0.35 (SLES11SP3)
11 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
12 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
13 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
14 stack overflow. For more information, please refer to bugzilla 17630.
15 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
16 the re-exporting nodes to be mounted with "32bitapi" mount option
18 --------------------------------------------------------------------------------
20 06-23-2014 Intel Corporation
22 * See http://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
23 * Server support for kernels:
24 2.6.32-431.23.3.el6 (RHEL6)
25 3.0.101-0.35 (SLES11SP3)
26 * Client support for unpatched kernels:
27 2.6.32-431.23.3.el6 (RHEL6)
28 3.0.101-0.35 (SLES11SP3)
30 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
31 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
32 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
33 stack overflow. For more information, please refer to bugzilla 17630.
34 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
35 the re-exporting nodes to be mounted with "32bitapi" mount option
37 --------------------------------------------------------------------------------
39 03-17-2014 Intel Corporation
41 * See http://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
42 * Server support for kernels:
43 2.6.32-431.17.1.el6 (RHEL6)
44 3.0.101-0.29 (SLES11SP3)
45 * Client support for unpatched kernels:
46 2.6.32-431.17.1.el6 (RHEL6)
47 3.0.101-0.29 (SLES11SP3)
49 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
50 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
51 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
52 stack overflow. For more information, please refer to bugzilla 17630.
53 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
54 the re-exporting nodes to be mounted with "32bitapi" mount option
56 --------------------------------------------------------------------------------
58 10-31-2013 Intel Corporation
60 * See http://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
61 * Server builds for kernels:
62 2.6.32-358.18.1.el6 (RHEL6.4)
63 * Client builds for unpatched kernels:
64 2.6.32-279.19.1.el6 (RHEL6.3)
65 2.6.32-358.18.1.el6 (RHEL6.4)
66 3.0.93-0.5 (SLES11 SP2)
67 3.0.92-0.8 (SLES11 SP3)
69 * Recommended e2fsprogs version: 1.42.7.wc2 or newer
70 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
71 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
72 stack overflow. For more information, please refer to bugzilla 17630.
73 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
74 the re-exporting nodes to be mounted with "32bitapi" mount option
76 --------------------------------------------------------------------------------
78 05-30-2013 Intel Corporation
80 * See http://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
81 * Server support for kernels:
82 2.6.32-358.6.2.el6 (RHEL6)
83 * Client support for unpatched kernels:
84 2.6.32-358.6.2.el6 (RHEL6)
86 2.6.18-348.3.1.el5 (RHEL5)
88 * Recommended e2fsprogs version: 1.42.7.wc2 or newer
89 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
90 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
91 stack overflow. For more information, please refer to bugzilla 17630.
92 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
93 the re-exporting nodes to be mounted with "32bitapi" mount option
95 --------------------------------------------------------------------------------
97 9-30-2012 Whamcloud, Inc.
99 * See http://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
100 * Server support for kernels:
101 2.6.32-279.5.1.el6 (RHEL6)
102 * Client support for unpatched kernels:
103 2.6.32-279.5.1.el6 (RHEL6)
105 2.6.18-238.19.1.el5 (RHEL5)
106 2.6.16.60-0.69.1 (SLES10)
107 * Recommended e2fsprogs version: 1.41.90.wc4
108 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
109 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
110 stack overflow. For more information, please refer to bugzilla 17630.
111 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
112 the re-exporting nodes to be mounted with "32bitapi" mount option
114 Severity : enhancement
116 Description: Update RHEL6.3 kernel to 2.6.32-279.5.1.el6.
118 Severity : enhancement
120 Description: Update RHEL6.3 kernel to 2.6.32-279.2.1.el6.
122 Severity : enhancement
124 Description: Update RHEL6.2 kernel to 2.6.32-220.17.1.el6
126 Severity : enhancement
128 Description: Update RHEL6.2 kernel to 2.6.32-220.13.1.el6
130 Severity : enhancement
132 Description: Update RHEL6.2 kernel to 2.6.32-220.7.1.el6
134 --------------------------------------------------------------------------------
136 04-30-2012 Whamcloud, Inc.
138 * See http://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
139 * Server support for kernels:
140 2.6.32-220.4.2.el6 (RHEL6)
141 * Client support for unpatched kernels:
142 2.6.32-220.4.2.el6 (RHEL6)
144 2.6.18-238.19.1.el5 (RHEL5)
145 2.6.16.60-0.69.1 (SLES10)
146 * Recommended e2fsprogs version: 1.41.90.wc4
147 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
148 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
149 stack overflow. For more information, please refer to bugzilla 17630.
150 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
151 the re-exporting nodes to be mounted with "32bitapi" mount option
153 Severity : enhancement
155 Description: Update RHEL6.2 kernel to 2.6.32-220.4.2.el6
157 Severity : enhancement
159 Description: Update RHEL6.2 kernel to 2.6.32-220.4.1.el6
161 Severity : enhancement
163 Description: Update RHEL6.2 kernel to 2.6.32-220.el6
165 Severity : enhancement
167 Description: Update RHEL6.1 kernel to 2.6.32-131.17.1.el6
169 --------------------------------------------------------------------------------
171 09-30-2011 Whamcloud, Inc.
173 * Server support for kernels:
174 2.6.18-238.19.1.el5 (RHEL5)
175 2.6.32-131.6.1.el6 (RHEL6)
176 * Client support for unpatched kernels:
177 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
178 2.6.18-238.19.1.el5 (RHEL5)
179 2.6.32-131.6.1.el6 (RHEL6)
180 2.6.29.4-167.fc11 (FC11)
181 2.6.16.60-0.69.1 (SLES10)
183 * Recommended e2fsprogs version: 1.41.90.wc3
184 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
185 Lustre filesystem with 4K stack may cause a stack overflow. For more
186 information, please refer to bugzilla 17630.
187 * NFSv4 reexport requires Lustre client on the re-exporting nodes to be
188 mounted with "32bitapi" mount option
190 Severity : enhancement
191 Description: Update RHEL6.1 kernel to 2.6.32-131.6.1.el6
193 Severity : enhancement
194 Description: Update RHEL5.6 kernel to 2.6.18-238.19.1.el5
196 Severity : enhancement
197 Description: Update RHEL5.6 kernel to 2.6.18-238.9.1.el5
200 Frequency : only with SLES10
202 Description: Use OFED "KMP" provided by Novell
203 Details : SLES10 SP3 ships with OFED in a separate "KMP" package.
204 Lustre is now built against this package. That means you need to
205 install the ofed-kmp package from Novell for the patchless client.
206 Note that the ofed-kmp that Novell ships may not exactly match the
207 kernel version but should still be compatible.
209 Severity : enhancement
211 Description: Update RHEL5.5 kernel to 2.6.18-194.17.1.el5
212 Update OEL5.5 kernel to 2.6.18-194.17.1.0.1.el5.
214 --------------------------------------------------------------------------------
216 2010-07-15 Oracle, Inc.
218 * Server support for kernels:
219 2.6.18-164.11.1.el5 (RHEL 5)
220 2.6.18-164.11.1.0.1.el5 (OEL 5)
221 * Client support for unpatched kernels:
222 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
223 2.6.18-164.11.1.el5 (RHEL 5),
224 2.6.18-164.11.1.0.1.el5 (OEL 5)
225 2.6.16.60-0.42.8 (SLES 10),
227 2.6.29.4-167.fc11 (FC11)
228 * Recommended e2fsprogs version: 1.41.10-sun2
230 Severity : enhancement
232 Description: Distribute lfs_migrate script with lustre
233 Details : The lfs_migrate mentioned in the manual is now included in the
234 lustre tree in lustre/scripts and will distributed in lustre RPMs.
236 Severity : enhancement
238 Description: Failover nid list added to import proc file.
240 Severity : enhancement
242 Description: "som_preview" mount option required to use SOM
243 Details : The clients will need to use som_preview mount option
247 Description: Lustre does not do 1MB IOs to HW RAID
248 Details : Bump MAX_PHYS/HW_SEGMENTS and SG_ALL to 256 in the RHEL5 kernel.
250 Severity : enhancement
252 Description: lfs quota output is not very convenient for awk/sed-parsing
253 Details : Some positions in lfs quota output table could be empty or
254 non-empty which made it hard to parse it with scripts, now
255 a dash is put instead of space where there is not supposed
260 Description: Update to OFED 1.5.1
261 Details : OFED 1.5.1 will be used in environments that don't use
262 vendor supplied OFED.
266 Description: Remove lnet. prefix from lctl params display
267 Details : Also renamed lnet.memused to lnet_memused to
272 Description: Use conf_param -d to remove permanant settings.
276 Description: Add list_param -R
277 Details : lctl list_param now supports recursive mode.
280 Bugzilla : 21829, 22312
281 Description: Fix llobdstat.
282 Details : 1. Now updating the write_bytes counter
283 2. Port script fixes from 1.8.
287 Description: decrease the usage of memory on clients.
288 Details : 1. On clients, recycle dentries and inodes unused.
289 2. Delete the code related to ll_deathrow(att 6215 in bug 1443). It
294 Description: llobdstat fix and enhancement
295 Details : add a counter to set a limit to how many samples will be returned
296 fix a wildcard in the path to limit to obdfilter stats only
300 Description: lfs check only allowed for root
301 Details : Code cleanups and added a regresion test of non-root use of
306 Description: MDS Oopses in ptlrpc_dispatch_difficult_reply()/__list_add()
307 Details : Fix hr->hr_index overload.
311 Description: Fix FIEMAP functionality.
315 Description: FC11 patchless client support added
319 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and
320 OEL5.4 kernel to 2.6.18-164.11.1.0.1.el5.
324 Description: New acc-sm script and various test-framework enhancments.
328 Description: Previously landed patch was reverted.
332 Description: Update SLES10 kernel to 2.6.16.60-0.42.8 (patchless)
336 Description: Add ldiskfs maxdirsize mount option.
340 Description: Fix mkfs.lustre support for 16TB LUNs.
344 Description: Print mmp_check_interval
345 Details : print mmp_check_interval and make it possible to abort
346 mount operation in case it takes too long.
350 Description: Add cascading_rw to lustre mpi tests.
354 Description: Store SOM data in EA.
355 Details : Also several SOM related cleanups.
359 Description: Add refcount to md_open_data
363 Description: Use refcount on lov_request_set to prevent memory corruption.
367 Description: Statahead: use parent dir's i_mutex for synchronization.
371 Description: loadgen improvements.
375 Description: Enable peer health detection in ksocklnd and ko2iblnd.
379 Description: Control DCACHE_LUSTRE_INVALID flag with MDS_INODE_LOOKCUP lock
383 Description: Fix race on obd_devcie:md_stats
387 Description: Fix SLES 11 client builds.
388 Details : Resolved uninitalized pointer gcc warnings.
392 Description: Allow non-root access for "lfs check".
393 Details : Added a check in obd_class_ioctl() for OBD_IOC_PING_TARGET.
397 Description: Clear imp_force_reconnect correctly in ptlrpc_connect_interpret()
399 Severity : enhancement
401 Description: Update kernel to RHEL5.4 2.6.18-164.6.1.el5 and
402 OEL5 2.6.18-164.6.1.0.1.el5(Both in-kernel OFED enabled).
406 Description: Sanity test_27p fails.
407 Details : OBD_FAIL_OST_ENOSPC has been moved to filter_statfs().
408 Sanity tests 27 were cleaned up and improved.
412 Description: Unify req format on client/servers
413 Details : Use new req_capsule API [almost] everywhere instead of old PTLRPC
414 buffers and swabbers approach..
419 Description: drop join file stuff.
421 Severity : enhancement
423 Description: Add 2.6.27 and SLES11 (patchless client) support
427 Description: Fix in ptlrpc_expire_one_request() to print the signed time
430 Severity : enhancement
432 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
435 Severity : enhancement
437 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
439 Severity : enhancement
441 Description: correctly handle big reply message.
442 Details : send LNet event if reply is bigger then buffer and adjust this buffer
447 Description: Drop unnecessary __GFP_NOMEMALLOC flag from filter_get_page()
449 Severity : enhancement
451 Description: Add support for OEL5.
455 Description: Conf-sanity.sh 50g test - deactivated OST should not cause a panic.
457 Severity : enhancement
459 Description: "weak-modules" support
460 Details : Implement "weak-modules" support which enables kernel modules
461 to be used with any kernel that implements the same kABI. In
462 order to achieve this modules are now installed in
463 /lib/modules/$(uname -r)/updates/kernel on all distributions.
467 Description: client could not reconnect to OST because of an active request.
468 Details : The client now retries to reconnect to the same server, if a
469 connect request failed with EBUSY or -EAGAIN. Test case 35b
470 conf-sanity.sh has been created.
474 Description: Don't allow make backward step on assignin osc next id.
475 Details : race between allocation next id and ll_sync thread can be cause
476 of set wrong osc next id and can be kill valid ost objects.
480 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
484 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
488 Description: Error handling in osc_statfs_interpret() has been improved.
489 Details : Check in osc_statfs_interpret() for EBADR.
493 Description: Do not send statfs() requests to OSTs disabled by administrator.
494 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
496 Severity : enhancement
498 Description: Added sanity-quota test to for lfs(1) to work with numeric uid/gid.
500 Severity : enhancement
502 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
504 Severity : enhancement
506 Description: Enhancement for lfs(1) command to use numeric uid/gid.
510 Description: Improved cleanup procedure for conf-sanity test_46a.
514 Description: Do not start bulk transfer, if client tried to reconnect.
515 Details : Do not start bulk transfer, if client tried to reconnect,
516 but failed due to active rpcs. Check for exp_abort_active_req
517 has been implemented.
521 Description: Prevent inconsistences between linux and lustre mount structures.
522 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
523 Make the sleep interruptible.
525 Severity : enhancement
527 Description: provide server to client comms path
528 Details : server to client communications path via new LDLM_SET_INFO rpc
532 Description: truncate starts GFP_FS allocation under transaction causing deadlock
533 Details : ldiskfs_truncate calls grab_cache_page which may start page
534 allocation under an open transaction. This may lead to
535 calling prune_icache with consequent lustre reentrance.
539 Description: send correctly lsm on open replay
540 Details : MDS is trust to LSM size on replay open, but client can set wrong size
545 Description: enable client interface failover
546 Details : When a child reconnects from another NID, properly update export
547 nid hash position and ldlm reverse import.
549 Severity : enhancement
551 Description: Communicate OST degraded/readonly state via statfs to MDS
552 Details : Flags in the statfs returned from OSTs indicate whether the
553 OST is in a degraded RAID state, or if the filesystem has
554 turned read-only after a filesystem error is detected.
559 Description: don't panic if EPROTO was hit when reading symlink
560 Details : correctly handling request reference in error cases.
562 Severity : enhancement
564 Description: Add LustreNetLink, a kernel-userspace communcation path.
566 Severity : enhancement
568 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
573 Description: access to llog context before init.
574 Details : move handling CATALOGS file at osc layer and forbid access to llog
578 Frequency : rare, only when client is writing to OST on local node
580 Description: Avoid deadlock for local client writes
581 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
582 memory freeing context. This allows OST threads to set the
583 PF_MEMALLOC flag on task structures in order to allocate memory
584 from reserved pools and complete IO.
585 Use GFP_HIGHUSER for OST allocations for non-local client writes,
586 so that the OST threads generate memory pressure and allow
587 inactive pages to be reclaimed.
589 Severity : enhancement
591 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
596 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
597 Details : move ldlm namespace creation in setup phase to avoid grab
598 _lprocfs_lock with cli_sem held.
602 Description: Temporarily disable grant shrink.
603 Details : Disable the feature for debugging.
606 Frequency : only when formatting test filesystems
608 Description: Unable to run several mkfs.lustre on loop devices at the same time
609 Details : mkfs.lustre returns error 256 on the concurrent loop devices
610 formatting. The solution is to proper handle the error.
612 Severity : enhancement
614 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
616 Severity : enhancement
618 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
623 Description: resolve race between obd_disconnect and class_disconnect_exports
624 Details : if obd_disconnect will be called to already disconnected export he
625 forget release one reference and osc module can't unloaded.
627 Severity : enhancement
629 Description: Allow tuning service thread via /proc
630 Details : For each service a new
631 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
632 created that can be used to set min/max thread counts, and get the
633 current number of running threads.
638 Descriptoin: don't return error if have partially created objects for file.
639 Details : lov_update_create_set uses set->set_success as index for created
640 objects, so if some requests will be failed, they will have hole
641 at end of array and qos_shrink_lsm for allocate correct lsm.
643 Severity : enhancement
645 Description: Update OFED support to 1.4.1
650 Description: Reduce small size read RPC
651 Details : Set read-ahead limit for every file and only do read-ahead when
652 available read-ahead pages are bigger than 1M to avoid small size
655 Severity : enhancement
657 Description: move AT tunable parameters for more consistent usage
658 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
660 Severity : enhancement
662 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
663 Details : allow skip disconnected ost for send statfs request and hide error
669 Description: fix racy locking of mballoc block bitmaps causing BUG
670 Details : The locking of the mballoc buddy bitmap and the in-memory
671 block bitmap was using two different spin locks in some
672 cases. This made it possible to incorrectly access the
673 mballoc bitmap while another process was modifying it,
674 causing a sanity assertion to fail. While no on-disk corruption
675 was reported, there was some risk of this happening.
678 Frequency : rare, on llog test 6
680 Descriptoin: don't allow connect to already connected import
681 Details : allowing connect to already connected import is hide connecting problem.
686 Descriptoin: allow kill process which wait statahead result
687 Details : for some reasons 'ls' can stick in waiting result from statahead,
688 in this case need way for kill this process.
690 Severity : enhancement
692 Description: Add state history info file, enhance import info file
693 Details : Track import connection state changes in a new osc/mdc proc file;
694 add overview-type data to the osc/mdc import proc file.
696 Severity : enhancement
698 Description: MDS create should not wait for statfs RPC while holding DLM lock.
700 Severity : enhancement
702 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
707 Descriptoin: OST grant too much space to client even there are not enough space.
708 Details : Client will shrink its grant cache to OST if there are no write
709 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
710 this grant cache if there are already not enough avaible space
711 (left_space < total_clients * 32M).
714 Frequency : start MDS on uncleanly shutdowned MDS device
716 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
717 Details : stay in waiting mds<>ost recovery finished produce random bugs
718 due race between two ll_sync thread for one lov target. send
719 ACTIVATE event only if connect realy finished and import have
723 Frequency : rare, connect and disconnect target at same time
725 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
726 Details : don't call obd_disconnect under lov_lock. this long time
727 operation and can block ptlrpcd which answer to connect request.
732 Descriptoin: don't lose wakeup for imp_recovery_waitq
733 Details : recover_import_no_retry or invalidate_import and import_close can
734 both sleep on imp_recovery_waitq, but we was send only one wakeup
738 Frequency : always with long access acl
740 Descriptoin: mds can't pack reply with long acl.
741 Details : mds don't control size of acl but they limited by reint/getattr
744 Severity : enhancement
746 Description: Update to SLES10 kernel-2.6.16.60-0.33.
748 Severity : enhancement
750 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
753 Frequency : start MDS on uncleanly shutdowned MDS device
755 Descriptoin: aborting recovery hang on MDS
756 Details : don't throttle destroy RPCs for the MDT.
759 Frequency : on remount
761 Description: external journal device not working after the remount
762 Details : clear dev_rdonly flag for external journal devices in
768 Description: shutdown vs evict race
769 Details : client_disconnect_export vs connect request race.
770 if client will evicted at this time - we start invalidate
771 thread without referece to import and import can be freed
775 Frequency : rare, need acl's on inode.
777 Description: client can't handle ost additional correctly
778 Details : if ost was added after client connected to mds client can have
779 hit lnet_try_match_md ... to big messages to wide striped files.
780 in this case need teach client to handle config events about add
781 lov target and update client max ea size at that event.
783 Severity : enhancement
785 Description: Changelogs
786 Details : Changelogs are a lightweight mechanism to track filesystem
787 metadata and namespace changes. The changelog is recorded
788 permanently on the MDTs, and is periodically "consumed" / purged
789 when records are no longer needed.
791 Severity : enhancement
793 Description: compact fld format with extents
794 Details : Store range of seq rather than every seq in FLD. Seq
795 controller update FLD rather than clients. In Case of CMD, mdt0
796 has FLD, all other metadata server act as non persistent proxy
797 for FLD queries and cache fld entries in fld cache.
802 Description: don't skip ost target if they assigned to file
803 Details : Drop slow OSCs if we can, but not for requested start idx.
804 This means "if OSC is slow and it is not the requested
805 start OST, then it can be skipped, otherwise skip it only
806 if it is inactive/recovering/out-of-space.
810 Description: more cleanup in mds_lov
811 Details : not send LOV EA under replay, we can't know about they size at this
812 time. Don't allow client connect to mds before any ost connected,
813 for avoid problems with LOV EA size and returning EIO to client.
815 Severity : enhancement
817 Description: Interoperability at server side (Disk interoperability)
819 Severity : enhancement
821 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
823 Severity : enhancement
825 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
827 Severity : enhancement
829 Description: New client IO stack (CLIO).
831 Severity : enhancement
833 Description: Commit on sharing. Eliminate inter-client dependencies between
834 uncommitted transactions by doing transaction commits.
835 Thereby clients may recovery independently.
838 Frequency : Create a symlink file with a very long name
840 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
841 Details : If there is no extra space in the request for early cancels,
842 ldlm_req_handles_avail() returns 0 instead of a negative value.
844 Severity : enhancement
846 Description: Add /proc entry for import status
847 Details : The mdc, osc, and mgc import directories now have
848 an import directory that contains useful import data for debugging
851 Severity : enhancement
853 Description: Re-disable certain /proc logging
854 Details : Enable and disable client's offset_stats, extents_stats and
855 extents_stats_per_process stats logging on the fly.
858 Frequency : Only on FC kernels 2.6.22+
860 Description: oops in statahead
861 Details : Do not drop reference count for the dentry from VFS when lookup,
862 VFS will do that by itself.
864 Severity : enhancement
866 Description: Generic /proc file permissions
867 Details : Set /Proc file permissions in a more generic way to enable non-
868 root users operate on some /proc files.
872 Description: Hitting mdc_commit_close() ASSERTION
873 Details : Properly handle request reference release in
874 ll_release_openhandle().
878 Description: quota recovery deadlock during mds failover
879 Details : This patch includes att18982, att18236, att18237 in bz14840.
881 1. fix osts hang when mds does failover with quotaon
882 2. prevent watchdog storm when osts threads wait for the
887 Frequency : only patchless client
888 Description: add workaround for race between add/remove dentry from hash
890 Severity : enhancement
892 Description: Allow OST glimpses to return PW locks
896 Description: LBUG when llog conf file is full
897 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
901 Description: Prevent import from entering FULL state when server in recovery
905 Description: service mount cannot take device name with ":"
906 Details : Only when device name contains ":/" will mount treat it as
912 Description: replace ptlrpcd with the statahead thread to interpret the async
913 statahead RPC callback
917 Frequency : on recovery
918 Description: I/O failures after umount during fail back
919 Details : if client reconnected to restarted server we need join to recovery
920 instead of find server handler is changed and process self
921 eviction with cancel all locks.
923 Severity : enhancement
925 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
927 Severity : enhancement
929 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
931 Severity : enhancement
933 Description: Upcall on Lustre log has been dumped
934 Details : Allow for a user mode script to be called once a Lustre log has
935 been dumped. It passes the filename of the dumped log to the
936 script, the location of the script can be specified via
937 /proc/sys/lnet/debug_log_upcall.
942 Description: avoid idr_remove called for id which is not allocated.
943 Details : Move assigment s_dev for clustered nfs to end of initialization,
944 to avoid problem with error handling.
949 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
950 Details : When connection is reused this not moved from CONN_UNUSED_HASH
951 into CONN_USED_HASH and this prodice warning when put connection
952 again in unused hash.
955 Severity : enhancement
957 Description: File striping can now be set to use an arbitrary pool of OSTs.
959 Severity : enhancement
961 Description: Export bytes_read/bytes_write count on OSC/OST.
965 Description: Early reply size mismatch, MGC loses connection
966 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
967 the connect flags are properly negotiated.
972 Description: quotas are not honored with O_DIRECT
973 Details : all writes with the flag O_DIRECT will use grants which leads to
974 this problem. Now using OBD_BRW_SYNC to guard this.
978 Description: add quota statistics
979 Details : 1. sort out quota proc entries and proc code.
980 2. add quota statistics
982 Severity : enhancement
984 Description: enable quota support for HEAD.
988 Description: Properly propagate oinfo flags from lov to osc for statfs
989 Details : restore missing copy oi_flags to lov requests.
991 Severity : enhancement
993 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
994 : l_getgroups(8), lst(8), routerstat(8)
995 Details : included man pages for llobdstat(8), llstat(8),
996 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
998 Severity : enhancement
1000 Description: configure's --enable-quota should check the
1001 : kernel .config for CONFIG_QUOTA
1002 Details : configure is terminated if --enable-quota is passed but
1003 : no quota support is in kernel
1007 Description: Remove portals compatibility
1008 Details : Remove portals compatibility, not interoperable with releases
1013 Description: Resolve device initialization race
1014 Details : Prevent proc handler from accessing devices added to the
1015 obd_devs array but yet be intialized.
1017 Severity : enhancement
1019 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
1021 Severity : enhancement
1023 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
1028 Description: Using wrong pointer in osc_brw_prep_request
1029 Details : Access to array[-1] can produce panic if kernel compiled with
1030 CONFIG_PAGE_ALLOC enabled
1034 Description: Client runs out of low memory
1035 Details : Consider only lowmem when counting initial number of llap pages
1039 Description: *optional* service tags registration
1040 Details : if the "service tags" package is installed on a Lustre node
1041 When the filesystem is mounted, a local-node service tag will
1042 be created. See http://inventory.sun.com/ for more information
1043 about the Service Tags asset management system.
1047 Description: Kernel BUG tries to release flock
1048 Details : Lustre does not destroy flock lock before last reference goes
1049 away. So always drop flock locks when client is evicted and
1050 perform unlock regardless of successfulness of speaking to MDS.
1054 Description: add refcount protection for osc callbacks, avoid panic on shutdown
1058 Description: sanity test 65a fails if stripecount of -1 is set
1059 Details : handle -1 striping on filesystem in ll_dirstripe_verify
1064 Description: ASSERTION(CheckWriteback(page,cmd)) failed
1065 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
1068 Severity : enhancement
1070 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
1074 Description: do not process already freed flock
1075 Details : flock can possibly be freed by another thread before it reaches
1076 to ldlm_flock_completion_ast.
1080 Description: LBUG during stress test
1081 Details : Need properly lock accesses the flock deadlock detection list.
1085 Description: oops in page fault handler
1086 Details : kernel page fault handler can return two special 'pages' in error
1087 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
1091 Description: timeout with invalidate import.
1092 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
1093 should be handled by ptlrpcd. This produce long age waiting and
1094 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
1096 Severity : enhancement
1098 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
1102 Description: enable MGS and MDT services start separately
1103 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
1104 a co-located MGS without starting the MGS, which is a complement
1105 to 'nosvc' mount option.
1109 Frequency : after recovery
1110 Description: precreate to many object's after del orphan.
1111 Details : del orphan st in oscc last_id == next_id and this triger growing
1112 count of precreated objects. Set flag LOW to skip increase count
1113 of precreated objects.
1117 Frequency : rare, on clear nid stats
1118 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
1119 Details : when clean nid stats sometimes try destroy live entry,
1120 and this produce panic in free.
1124 Description: Stack overflow during MDS log replay
1125 ease stack pressure by using a thread dealing llog_process.
1129 Description: wait until IO finished before start new when do lock cancel.
1130 Details : VM protocol want old IO finished before start new, in this case
1131 need wait until PG_writeback is cleared until check dirty flag and
1132 call writepages in lock cancel callback.
1134 Severity : enhancement
1136 Description: using special macro for print time and cleanup in includes.
1140 Description: mds_mfd_close() ASSERTION(rc == 0)
1141 Details : In mds_mfd_close(), we need protect inode's writecount change
1142 within its orphan write semaphore to prevent possible races.
1146 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
1150 Frequency : rare, on shutdown ost
1151 Description: don't hit live lock with umount ost.
1152 Details : shrink_dcache_parent can be in long loop with destroy dentries,
1153 use shrink_dcache_sb instead.
1157 Description: don't panic with use echo client
1158 Details : echo client pass NULL as client nid pointer and this produce null
1159 pointer dereference.
1163 Description: fix build on ppc32
1164 Details : compile code with -m64 flag produce wrong object file for ppc32.
1168 Description: add message levels for liblustreapi
1172 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
1173 Details : Move the 'good_osts' check before the 'total_bavail' check. This
1174 will result in an -EAGAIN and in the exit call path we call
1175 alloc_rr() which will with increasing aggressiveness attempt to
1176 aquire precreated objects on the minimum number of required OSCs.
1180 Description: Use old size assignment to avoid deadlock
1181 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
1182 the scheduling while holding a spinlock. We do not need locking
1183 for size in ll_update_inode() because size is only updated from
1184 the MDS for directories or files without objects, so there is no
1185 other place to do the update, and concurrent access to such inodes
1186 are protected by the inode lock.
1190 Description: resolve "_IOWR redefined" build error on SLES10
1194 Description: dump the memory debugging after all modules are unloaded to
1195 suppress false negative in conf_sanity test 39
1197 Severity : enhancement
1199 Description: build kernel-ib packages for OFED 1.3 in our release cycle
1204 Description: fix SLES kernel versioning
1205 Details : the kernel version for our SLES 10 kernel did not include a "-"
1206 before the "smp" at the end. while this was not a problem in
1207 general, it did mean that software trying to use the kernel
1208 version to try to detect a vendor specific kernel would fail.
1209 this was most evident by the OFED build scripts.
1213 Description: Don't update lov_desc members until making sure they are valid
1214 Details : When updating lov_desc members via proc fs, need fix their
1215 validities before doing the real update.
1219 Description: don't put request into delay list while invalidate in flight.
1220 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
1221 import in flight. this produce timeout for invalidate and sometimes
1222 can cause stale data.
1226 Frequency : on ppc only
1227 Description: not convert ost objects for directory because it's not exist.
1228 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
1232 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
1234 Severity : enhancement
1236 Description: Update to support 2.6.22.14 vanilla kernel.
1240 Frequency : rare, on recovery
1241 Description: read procfs can produce deadlock in some situation
1242 Details : Holding lprocfs lock which send rpc can produce block for destroy
1243 obd objects and this also block reconnect with -EALREADY. This
1244 isn't fix all lprocfs bugs - but make it rare.
1246 Severity : enhancement
1248 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
1251 Frequency : frequent on X2 node
1253 Description: mdc_set_open_replay_data LBUG
1254 Details : Set replay data for requests that are eligible for replay.
1258 Description: lustre_mgs: operation 101 on unconnected MGS
1259 Details : When MGC is disconnected from MGS long enough, MGS will evict the
1260 MGC, and late on MGC cannot successfully connect to MGS and a lot
1261 of the error messages complaining that MGS is not connected.
1265 Frequency : on network error
1266 Description: panic with double free request if network error
1267 Details : mdc_finish_enqueue is finish request if any network error ocuring,
1268 but it's true only for synchronus enqueue, for async enqueue
1269 (via ptlrpcd) this incorrect and ptlrpcd want finish request
1272 Severity : enhancement
1274 Description: client-side metadata stat-ahead during readdir(directory readahead)
1275 Details : perform client-side metadata stat-ahead when the client detects
1276 readdir and sequential stat of dir entries therein
1279 Frequency : on start mds
1281 Description: Implement get_info(last_id) in obdfilter.
1284 Frequency : occasional
1286 Description: Correctly check stale fid, not start epoch if ost not support SOM
1287 Details : open with flag O_CREATE need set old fid in op_fid3 because
1288 op_fid2 was overwritten with new generated fid, but mds can answer
1289 with one of these two fids and both is not stale. Setattr
1290 incorrectly started an epoch and assume will be called
1291 done_writing, but without SOM done_writing ever being called.
1294 Frequency : rare, depends on device drivers and load
1296 Description: MDS or OSS nodes crash due to stack overflow
1297 Details : Code changes in 1.8.0 increased the stack usage of some functions.
1298 In some cases, in conjunction with device drivers that use a lot
1299 of stack the MDS (or possibly OSS) service threads could overflow
1300 the stack. One change which was identified to consume additional
1301 stack has been reworked to avoid the extra stack usage.
1304 Frequency : occasional
1306 Description: Do not fail import if osc_interpret_create gets -EAGAIN
1307 Details : If osc_interpret_create got -EAGAIN it immediately exits and
1308 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
1309 oscc_has_objects and see OSC has no objests and call
1310 oscc_internal_create to resend create request.
1312 Severity : enhancement
1314 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
1316 Severity : enhancement
1318 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
1321 Frequency : very rare
1323 Description: Fix replay if there is an un-replied request and open
1324 Details : In some cases, older replay request will revert the
1325 mcd->mcd_last_xid on MDS which is used to record the client's
1326 latest sent request.
1328 Severity : enhancement
1330 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
1332 Severity : enhancement
1334 Description: Add rhel5 support to HEAD.
1336 Serverity : enhancement
1338 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
1343 Description: Don't allow skipping OSTs if index has been specified.
1344 Details : Don't allow skipping OSTs if index has been specified, make
1345 locking in internal create lots better.
1349 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
1350 Details : access to bitfield in structure is always rounded to long
1351 and this produce problem with not atomic change any bit.
1355 Description: Lustre make rpms failed.
1356 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
1357 builds Lustre rpms from tarball.
1360 Frequency : rare on shutdown ost
1362 Description: If llog cancel was not send before clean_exports phase, this can
1363 produce deadlock in llog code.
1364 Details : If llog thread has last reference to obd and call class_import_put
1365 this produce deadlock because llog_cleanup_commit_master wait when
1366 last llog_commit_thread exited, but this never success because was
1367 called from llog_commit_thread.
1371 Description: allow userland application know is lost one of stripes.
1372 Details : fill lvb_blocks with error code on ost and return it to
1373 application if error flag found.
1377 Description: NULL lov_tgts causing MDS oops
1378 Details : more safe checks for NULL lov_tgts for avoid oops.
1380 Severity : enhancement
1382 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
1386 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
1387 Details : If an OST is down the MDS will hang indefinitely in
1388 obd_statfs_rqset() waiting for the statfs data. While for
1389 MDS QOS usage of statfs, it should not stuck in waiting.
1391 Severity : enhancement
1393 Description: remote_acl support
1394 Details : Support ACL-based permission check for remote user.
1395 Support setfacl/getfacl for remote user with the utils
1396 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
1397 the system "{s,g}etfacl" utils.
1399 Severity : enhancement
1401 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
1403 Severity : enhancement
1405 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
1409 Description: make mgs_setparam() handle fsname containing dash
1410 Details : fsname containing a dash does not work with lctl conf_param
1412 Severity : enhancement
1414 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
1416 Severity : enhancement
1418 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
1422 Frequency : rare, at shutdown
1423 Description: access already free / zero obd_namespace.
1424 Details : if client_disconnect_export was called without force flag set,
1425 and exist connect request in flight, this can produce access to
1426 NULL pointer (or already free pointer) when connect_interpret
1427 store ocd flags in obd_namespace.
1431 Frequency : only at startup
1432 Description: not alloc memory with spinlock held.
1433 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
1434 if any spinlock held.
1436 Severity : enhancement
1438 Description: make lustre randomly fail allocating memory
1439 Details : Make lustre randomly failed allocating memory for testing purpose.
1441 Severity : enhancement
1443 Description: lost problems with lov objid file
1444 Details : Fixes some scability and access to not inited memory problems
1445 in work with lov objdid file.
1450 Description: lfs find does not continue on file error
1451 Details : Continue other files processing when a file/dir is absent.
1455 Description: Inconsistent usage of lustre_pack_reply()
1456 Details : Standardize the usage of lustre_pack_reply() such that it
1457 always generate a CERROR on failure.
1460 Frequency : occasional
1462 Description: MDS hang or stay in waiting lock
1463 Details : If client receive lock with CBPENDING flag ldlm need send lock
1464 cancel as separate rpc, to avoid situation when cancel request
1465 can't processed due all i/o threads stay in wait lock.
1469 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1473 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1477 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1481 Description: Update to RHEL4 latest kernel.
1483 Severity : enhancement
1485 Description: Build SLES10 patchless client fails
1486 Details : The configure was broken by run ./configure with
1487 --with-linux-obj=.... argument for patchless client. When the
1488 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1489 can't search header adequately. Use absolute path such as
1490 -I($LINUX)/include instead.
1494 Description: interrupt oig_wait produce painc on resend.
1495 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1496 private set, and this produce situation when rq_ptlrpcd_data not
1497 copyed to new allocated request and triggered LBUG on assert
1498 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1499 ptlrpcd set if request is changed and can be safety changed to use
1504 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1505 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1506 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1507 these macros and replace them with their difinitions instead.
1511 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1512 Details : Update which_patch & target file for SLES10 latest kernel.
1515 Description: Remove client patches from SLES 10 kernel.
1516 Details : This causes SLES 10 clients to behave as patchless clients
1517 even on a Lustre-patched (server) kernel.
1519 Severity : enhancement
1521 Description: self-adjustable client's lru lists
1522 Details : use adaptive algorithm for managing client cached locks lru
1523 lists according to current server load, other client's work
1524 pattern, memory activities, etc. Both, server and client
1525 side namespaces provide number of proc tunables for controlling
1528 Severity : enhancement
1530 Description: light-weight GSS support
1531 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
1532 and reduce performance overhead.
1534 Severity : enhancement
1536 Description: Linux keyring support
1537 Details : Support using service of Linux keyring for Lustre GSS internal
1538 context refresh/cache mechanism.
1542 Description: Fix errors in lfs documentation
1543 Details : Fixes man pages
1547 Description: don't use GFP_* in generic Lustre code.
1548 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
1552 Description: obdclass is limited by single OBD_ALLOC(idarray)
1553 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
1557 Description: warnings with build patchless client with vanila 2.6.19 and up
1558 Details : change old ctl_table style and replace ctl_table/ctl_table_header
1559 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
1563 Frequency : Only for SLES
1564 Description: sanity_quota fail test_1
1565 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
1566 file, which makes TSTID[2] inunique.
1570 Frequency : Only for SLES9
1571 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
1572 Details : Update target/ChangeLog/which_patch .
1576 Frequency : Only for test_10 in sanity-quota.sh
1577 Description: a bug in quota test code
1578 Details : checking a test flag in a wrong place
1582 Frequency : Only for test_13 in sanity-quota.sh
1583 Description: a bug in quota test script
1584 Details : shouldn't check overfull allocated quota
1588 Description: jbd statistics
1589 Details : Port older jbd statistics patch for sles10
1591 Severity : enhancement
1593 Description: Add printing inode info into message about error in writepage.
1598 Description: reply_lock_interpret crash due to race with it and lock cancel.
1599 Details : Do not replay locks that are being cancelled. Do not reference
1600 locks by their address during replay, just by their handle.
1604 Frequency : When flocks are used.
1605 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
1606 Details : Flock locks might destroy just granted lock if it could be merged
1607 with another existing flock, this is done in completion handler,
1608 so teach ldlm_cli_enquque_fini that this is a valid case for
1614 Description: Oops in read and write path when failing to allocate lock.
1615 Details : Check if lock allocation failed and return error back.
1619 Description: lstripe command fails for valid OST index
1620 Details : The stripe offset is compared to lov->desc.ld_tgt_count
1621 instead of lov->desc.ld_active_tgt_count.
1625 Description: sanity.sh failed test 103
1626 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
1631 Description: df doesn't work properly if diskfs blocksize != 4K
1632 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
1636 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
1637 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
1639 Severity : enhancement
1641 Description: Adaptive timeouts
1642 Details : RPC timeouts adapt to changing server load and network
1643 conditions to reduce resend attempts and improve recovery time.
1647 Description: llapi_file_create() does not allow some changes
1648 Details : add llapi_file_open() that allows specifying the mode and
1649 open flags, and also returns an open file handle.
1653 Description: merge and cleanup kernel patches.
1654 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
1658 Description: Add journal checksum support.(Kernel part)
1659 Details : The journal checksum feature adds two new flags i.e
1660 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
1661 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
1662 indicates that the commit block contains the checksum for
1663 the blocks described by the descriptor blocks. Now commit
1664 record can be sent to disk without waiting for descriptor
1665 blocks to be written to disk. This behavior is controlled
1666 using JBD2_FEATURE_ASYNC_COMMIT flag.
1670 Description: OSS needs mutliple precreate threads
1671 Details : Add ability to start more than one create thread per OSS.
1675 Description: Sanity.sh test_65k failed: setstripe should have succeeded
1676 Details : execute lfs setstripe on client
1680 Description: mds_obd_create error creating tmp object
1681 Details : When the user sets quota on root, llog will be affected and can't
1682 create files and write files.
1685 Frequency : Always on ia64 patchless client, and possibly others.
1687 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
1688 Details : This allows the patchless client to be loaded on architectures
1689 without this export.
1693 Description: RedHat Update kernel for RHEL5
1694 Details : Add the kernel config file for RHEL5.
1698 Description: RedHat Update kernel for RHEL5
1699 Details : Modify the kernel config file more closer RHEL5.
1703 Description: Build failure against Centos5 (RHEL5)
1704 Details : Define PAGE_SIZE when it isn't present.
1708 Description: test_7 on sanity-quota.sh can't be used on separate machines
1709 Details : add facet to handle it
1713 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
1714 Details : with NFS, the anon dentry's parent was set to itself in
1715 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
1716 corresponding dentry other than use rec->ur_name.
1718 Severity : enhancement
1720 Description: lfs setstripe enhancement
1721 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
1725 Description: enable data checksumming by default
1726 Details : enable checksum by default, allow --disable-checksum
1727 configure option and "-o nochecksum" mount option. Checksums
1728 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
1732 Description: lustre support for RHEL5
1733 Details : Add support for RHEL5.
1737 Description: Client eviction due to failover config
1738 Details : after a connection loss, the lustre client should attempt to
1739 reconnect to the last active server first before trying the
1740 other potential connections.
1744 Description: when mds and osts use different quota unit(32bit and 64bit),
1745 quota will be released repeatly.
1746 Details : void sending multiple quota reqs to mds, which will keep the
1747 status between the reqs.
1751 Description: osts not allocated evenly to files
1752 Details : change the condition to increase offset_idx
1755 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
1757 Description: Data corruption for OSTs that are formatted larger than 2TB
1759 Details : When generating the bio request for lustre file writes the
1760 sector number would overflow a temporary variable before being
1761 used for the IO. The data reads correctly from Lustre (which
1762 will overflow in a similar manner) but other file data or
1763 filesystem metadata may be corrupted in some cases.
1767 Description: Tune the kernel for good SCSI performance.
1768 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
1769 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
1774 Description: rewrite ext2-derived code in obdclass/uuid.c
1775 Details : rewrite inherited code (uuid parsing code from ext2 utils)
1776 from scratch preserving functionality.
1780 Description: "lfs find -obd UUID" prints directories
1781 Details : "lfs find -obd UUID" will return all directory names instead
1782 of just file names. It is incorrect because the directories
1783 do not reside on the OSTs.
1787 Description: use i_size_read and i_size_write in 2.6 port
1788 Details : replace inode->i_size access with i_size_read/write()
1791 Frequency : when removing large files
1793 Description: scheduling issue during removal of large Lustre files
1794 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
1795 It causes scheduling issues when removing large files (17TB in the
1799 Frequency : only with liblustre clients on XT3
1801 Description: evictions taking too long
1802 Details : allow llrd to evict clients directly on OSTs
1805 Frequency : only on ppc
1807 Description: /proc/fs/lustre/devices broken on ppc
1808 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
1809 We should make sure the type of 'index' is loff_t and then cast
1810 explicitly as needed below. Do not assign an explicitly cast
1814 Frequency : only for rhel5
1816 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
1817 Details : Modify the target file & which_kernel.
1819 Severity : enhancement
1821 Description: omit set fsid for export NFS
1822 Details : fix set/restore device id for avoid EMFILE error and mark lustre
1823 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
1827 Frequency : Always, for kernels after 2.6.16
1828 Description: Fix warning idr_remove called for id=.. which is not allocated.
1829 Details : Last kernels save old s_dev before kill super and not allow
1830 to restore from callback - restore it before call kill_anon_super.
1834 Description: buffer overruns could theoretically occur
1835 Details : llapi_semantic_traverse() modifies the "path" argument by
1836 appending values to the end of the origin string, and a
1837 overrun may occur. Adding buffer overrun check in liblustreapi.
1841 Description: Fix error on 'ls .' at the top of the Lustre mount.
1842 Details : Don't revalidate dentry if it is a root dentry.
1846 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
1847 Details : Modify vm-tunables-rhel4.patch.
1851 Description: Kernel config for 2.6.18-vanilla.
1852 Details : Modify targets/2.6-vanilla.target.in.
1853 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
1854 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
1855 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
1856 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
1860 Description: improve handling recoverable errors
1861 Details : if request processig with error which can be recoverable on server
1862 request should be resend, otherwise page released from cache and
1867 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
1868 Details : Modify target file & which_patch.
1869 A flaw was found in the IA32 system call emulation provided
1870 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
1871 value could be stored in the %RAX register, which could trigger an
1872 out-of-bounds system call table access. An untrusted local user
1873 could exploit this flaw to run code in the kernel
1874 (ie a root privilege escalation). (CVE-2007-4573).
1878 Description: O_DIRECT bypasses client statistics.
1879 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
1880 read_ahead_stats were updated. copied the stats section from
1881 osc_send_oap_rpc() into async_internal().
1885 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
1889 Description: change order of libsysio includes
1890 Details : '#include sysio.h' should always come before '#include xtio.h'
1894 Description: handle "serious error: objid * already exists" more gracefully
1895 Details : If LAST_ID value on disk is smaller than the objects existing in
1896 the O/0/d* directories, it indicates disk corruption and causes an
1897 LBUG(). If the object is 0-length, then we should use the existing
1898 object. This will help to avoid a full fsck in most cases.
1900 Severity : enhancement
1902 Description: adapt the lustre_config script to support the upgrade case
1903 Details : Add "-u" option for lustre_config script to support upgrading 1.4
1904 server targets to 1.6 in parallel.
1908 Description: To avoid grant space > avaible space when the disk is almost
1909 full. Without this patch you might see the error "grant XXXX >
1910 available" or some LBUG about grant, when the disk is almost
1912 Details : In filter_check_grant, for non_grant cache write, we should
1913 check the left space by if (*left > ungranted + bytes), instead
1914 of (*left > ungranted), because only we are sure the left space
1915 is enough for another "bytes", then the ungrant space should be
1916 increase. In client, we should update cl_avail_grant only there
1917 is OBD_MD_FLGRANT in the reply.
1921 Description: Update RHEL 4 kernel to fix local root privilege escalation.
1922 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
1923 described in CVE-2007-4573. This problem could allow untrusted
1924 local users to gain root access.
1927 Frequency : when using O_DIRECT and quotas
1929 Description: Incorrect file ownership on O_DIRECT output files
1930 Details : block usage reported by 'lfs quota' does not take into account
1931 files that have been written with O_DIRECT.
1936 Description: touch file failed when fs is not full
1937 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
1938 otherwise we can get ENOSP while fs is not full.
1942 Description: parallel lock callbacks
1943 Details : Instead of sending blocking and completion callbacks as separated
1944 requests, adding them to a set and sending in parallel.
1947 Frequency : only for Cray XT3
1948 Bugzilla : 12829/13455
1949 Description: Changing primary group doesn't change the group lustre assigns to
1951 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
1952 overridden with the primary group provided by the group upcall,
1953 whereas the supplied fsgid can be trusted if it is in the list of
1954 supplementary groups returned by the group upcall.
1956 Severity : enhancement
1958 Description: Allow masking D_WARNING, D_ERROR messages from console
1959 Details : Console messages can now be disabled via lnet.printk.
1963 Description: User code with malformed file open parameter crashes client node
1964 Details : Before packing join_file req, all the related reference should be
1965 checked carefully in case some malformed flags cause fake
1966 join_file req on client.
1970 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
1971 Details : ldlm_completion_ast() assumes that a lock is granted when the req
1972 mode is equal to the granted mode. However, it should also check
1973 that LDLM_FL_CP_REQD is not set.
1977 Description: Heavy nfs access might result in deadlocks
1978 Details : After ELC code landed, it is now improper to enqueue any mds
1979 locks under och_sem, because enqueue might want to decide to
1980 cancel open locks for same inode we are holding och_sem for.
1984 Description: Client eviction while running blogbench
1985 Details : A lot of unlink operations with concurrent I/O can lead to a
1986 deadlock causing evictions. To address the problem, the number of
1987 oustanding OST_DESTROY requests is now throttled to
1988 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
1989 callbacks are processed in priority.
1993 Description: enable ACLs on MDS by default
1994 Details : ACLs must be enabled on MDS by default.
1997 Frequency : PPC/PPC64 only
1999 Description: conflicts between asm-ppc64/types.h and lustre_types.h
2000 Details : fix duplicated definitions between asm-ppc64/types.h and
2001 lustre_types.h on PPC.
2004 Frequency : PPC/PPC64 only
2006 Description: asm-ppc/segment.h does not exist
2007 Details : fix compile issue on PPC.
2011 Description: better handle error messages in extents code
2014 Frequency : RHEL4 only
2016 Description: mkfs is very slow on IA64/RHEL4
2017 Details : A performance regression has been discovered in the MPT Fusion
2018 driver between versions 3.02.73rh and 3.02.99.00rh. As a
2019 consequence, we have downgraded the MPT Fusion driver in the RHEL4
2020 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
2022 Severity : enhancement
2024 Description: SNMP support enhancement
2025 Details : Adding total number of sampled request for an MDS node in snmp
2028 Severity : enhancement
2030 Description: Optimize ldlm waiting list processing for PR extent locks
2031 Details : When processing waiting list for read extent lock and meeting read
2032 lock that is same or wider to it that is not contended, skip
2033 processing rest of the list and immediatelly return current
2034 status of conflictness, since we are guaranteed there are no
2035 conflicting locks in the rest of the list.
2039 Description: Time out and refuse to reconnect
2040 Details : When the failover node is the primary node, it is possible
2041 to have two identical connections in imp_conn_list. We must
2042 compare not conn's pointers but NIDs, otherwise we can defeat
2043 connection throttling.
2047 Description: port llog fixes from b1_6 into HEAD
2048 Details : Port llog reference couting and some llog cleanups from b1_6
2049 (bug 10800) into HEAD, for protect from panic and access to already
2050 free llog structures.
2054 Description: Detect stride IO mode in read-ahead
2055 Details : When a client does stride read, read-ahead should detect that and
2056 read-ahead pages according to the detected stride pattern.
2060 Description: data checksumming impacts single node performance
2061 Details : add support for several checksum algorithm. Currently, only CRC32
2062 and Adler-32 are supported. The checksum type can be changed on
2063 the fly via /proc/fs/lustre/osc/*/checksum_type.
2067 Description: use adler32 for page checksums
2068 Details : when available, use the Adler-32 algorithm instead of CRC32 for
2073 Description: build for x2 fails
2074 Details : fix compile issue on Cray systems.
2078 Description: Properly match for duplicate locks
2079 Details : Due to different lock order from skiplists code, we need to
2080 traverse entire list for now
2083 Frequency : only on PPC/SLES10
2085 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
2086 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
2087 be defined. Add a hack in mkfs_lustre.c to work around this compile
2092 Description: LASSERT on MDS when client holding flock lock dies
2093 Details : ldlm pool logic depends on number of granted locks equal to
2094 number of released locks which is not true for flock locks, so
2095 just exclude such locks from consideration.
2099 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
2100 Details : Use fsfilt_sync() for both the whole filesystem sync and
2101 individual file sync to eliminate dangerous inode locking
2102 with I_LOCK that can lead to a deadlock.
2106 Description: performance in 1.6.3
2107 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
2108 sure that 1MB requests are not fragmented by the block layer.
2110 Severity : enhancement
2112 Description: organize the server-side client stats on per-nid basis
2113 Details : Change the structure of stats under obdfilter and mds to
2121 The "uuid"s file would list the uuids of _active_ exports.
2122 And the clear entry is to clear all stats and stale nids.
2124 Severity : enhancement
2126 Description: eliminate client locks in face of contention
2127 Details : file contention detection and lockless i/o implementation
2128 for contended files.
2132 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
2135 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
2137 Description: MDS may lose file striping (and hence file data) in some cases
2138 Details : If there are additional extended attributes stored on the MDS,
2139 in particular ACLs, SELinux, or user attributes (if user_xattr
2140 is specified for the client mount options) then there is a risk
2141 of attribute loss. Additionally, the Lustre file striping
2142 needs to be larger than default (e.g. striped over all OSTs),
2143 and an additional attribute must be stored initially in the
2144 inode and then increase in size enough to be moved to the
2145 external attribute block (e.g. ACL growing in size) for file
2150 Description: skiplist implementation simplification
2151 Details : skiplists are used to group compatible locks on granted list
2152 that was implemented as tracking first and last lock of each
2153 lock group the patch changes that to using doubly linked lists
2157 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
2158 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
2159 because LASSERT on both the data supplied by a client, and the
2160 data on disk is dangerous and incorrect.
2162 Severity : enhancement
2164 Description: Slow truncate/writes to huge files at high offsets.
2165 Details : Directly associate cached pages to lock that protect those pages,
2166 this allows us to quickly find what pages to write and remove
2167 once lock callback is received.
2171 Description: more ldlm soft lockups
2172 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
2173 starve other threads from the resource lock for a long time in
2174 case of long waiting queue, so change the debug level from
2175 D_OTHER to the less frequently used D_INFO.
2177 Severity : enhancement
2179 Description: add -gid, -group, -uid, -user options to lfs find
2183 Description: Hung threads in invalidate_inode_pages2_range
2184 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
2185 once one is completed. As a result, some RPCs are stuck in the
2186 queue and are never sent.
2190 Description: filter threads hungs on waiting journal commit
2191 Details : Cleanup filter group llog code, then only filter group llog will
2192 be only created in the MDS/OST syncing process.
2196 Description: Procfs and llog threads access destoryed import sometimes.
2197 Details : Sync the import destoryed process with procfs and llog threads by
2198 the import refcount and semaphore.
2200 Severity : enhancement
2202 Description: openlock cache of b1_6 port to HEAD
2207 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
2208 Details : If insertion of an extent fails, then discard the inode
2209 preallocation and free data blocks else it can lead to duplicate
2214 Description: don't always update ctime in ext3_xattr_set_handle()
2215 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
2216 In some cases the ctime should not be updated, for example for
2217 2.0->1.8 compatibility it is necessary to delete an xattr and it
2218 should not update the ctime.
2222 Bugzilla : 15713/16362
2223 Description: Assertion in iopen_connect_dentry in 1.6.3
2224 Details : looking up an inode via iopen with the wrong generation number can
2225 populate the dcache with a disconneced dentry while the inode
2226 number is in the process of being reallocated. This causes an
2227 assertion failure in iopen since the inode's dentry list contains
2228 both a connected and disconnected dentry.
2232 Description: assertion failure in ldlm_handle2lock()
2233 Details : fix a race between class_handle_unhash() and class_handle2object()
2234 introduced in lustre 1.6.5 by bug 13622.
2239 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
2240 Details : remove the limit of 256 scsi disks in the sd_iostat patch
2245 Description: oops in sd_iostats_seq_show()
2246 Details : unloading/reloading the scsi low level driver triggers a kernel
2247 bug when trying to access the sd iostat file.
2252 Description: Kernel panics during QLogic driver reload
2253 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
2254 patch, causing memory corruption.
2259 Description: journal_dev option does not work in b1_6
2260 Details : pass mount option during pre-mount.
2262 Severity : enhancement
2264 Description: Add a FIEMAP(FIle Extent MAP) ioctl
2265 Details : FIEMAP ioctl will allow an application to efficiently fetch the
2266 extent information of a file. It can be used to map logical blocks
2267 in a file to physical blocks in the block device.
2271 Description: LDLM soft lockups - improvement
2272 Details : It is be possible to send the lock handle along with each read
2273 or write request because the client is already doing a lock match
2274 itself so there isn't any reason the OST should have to re-do that
2278 Frequency : only on Cray X2
2280 Description: X2 build failures
2281 Details : fix build failures on Cray X2.
2285 Description: xid & resent requests
2286 Details : Initialize RPC XID from clock at startup (randomly if clock is
2289 Severity : enhancement
2291 Description: Add lustre_start utility to start or stop multiple Lustre servers
2296 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
2297 Details : In case of memory pressure, list_del() can be called twice on
2298 req->rq_history_list, causing a kernel oops.
2302 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
2303 Details : fix stack overflow in the distributed lock manager by defering
2304 export eviction after a failed AST to the elt thread instead of
2305 handling it in the dlm interpret routine.
2309 Description: Convert some comments to new format.
2310 Details : Update documenting comments to match doxygen conventions.
2314 Description: Grammar fixes.
2315 Details : A couple of trivial spelling fixes.
2319 Description: OSD_COUNTERS-mandatory
2320 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
2321 to simplify the code.
2325 Description: simplify cmm_device freeing logic.
2326 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
2330 Description: Add lockdep support to dt_object_operations locking interface.
2331 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
2332 parameter indicating lock ordering. Update mdd code to use new
2337 Description: Introduce failloc constants for lockless IO tests.
2338 Details : Add two new failloc constants to test lockless IO. Only one of
2339 them in implemented---another is checked in yet to be landed
2344 Description: Add lockdep support for inode mutex.
2345 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
2346 in the situations where Lustre has to lock more than one inode
2351 Description: Add optional invariants checking support.
2352 Details : Add new LINVRNT() macro, optional on new --enable-invariants
2353 configure switch. This macro is to be used for consistency and
2354 sanity checks that are too expensive to be left in `production'
2359 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
2360 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
2365 Description: Add lockdep support for ldlm_lock and ldlm_resource.
2366 Details : Use spin_lock_nested() in (the only) situation where more than
2367 one ldlm_lock is locked simultaneously. Also, fix possible
2368 dead-lock in ldlm_lock_change_resource() by enforcing particular
2373 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
2374 Details : Instead of specifying each ldlm_lock call-back through separate
2375 parameter, wrap them into struct ldlm_callback_suite.
2379 Description: Kill join_lru obd method and its callers.
2380 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
2381 in memory---a requirement implemented through ->o_join_lru() obd
2382 method in HEAD. Remove this method and its users.
2386 Description: Add asynchronous ldlm ENQUEUE completion handler.
2387 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
2388 case that stock ldlm_completion_ast() cannot handle as it waits
2389 until lock is granted. Introduce new ldlm_completion_ast_async()
2390 for this. Also comment ldlm_completion_ast().
2394 Description: ldlm_error <-> errno conversion.
2395 Details : Add functions to map (rather arbitrary) between LDLM error codes
2396 and standard errno values. CLIO needs this to prevent LDLM specific
2397 constants from escaping ldlm and osc.
2401 Description: Kill unused ldlm_handle2lock_ns() function.
2402 Details : Kill unused ldlm_handle2lock_ns() function.
2406 Description: Add lu_ref support to ldlm_lock
2407 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
2408 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
2409 and ldlm_resource. LDLM interface has to be changed, because code that
2410 releases a reference on a lock, has to "know" what reference this is.
2411 In the most frequent case
2413 lock = ldlm_handle2lock(handle);
2415 LDLM_LOCK_PUT(lock);
2417 no changes are required. When any other reference (received _not_ from
2418 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
2419 instead of LDLM_LOCK_PUT().
2421 Arguably, changes are pervasive, and interface requires some discipline
2422 for proper use. On the other hand, it was very instrumental in finding
2423 a few leaked lock references.
2427 Description: Add ldlm_lock_addref_try().
2428 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
2429 attempts to addref a lock that might be being canceled
2434 Description: Add ldlm_weigh_callback().
2435 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
2436 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
2437 hard-coded `number of pages' logic.
2441 Description: Add lockdep annotations to llog code.
2442 Details : Use appropriately tagged _nested() locking calls in the places
2443 where llog takes more than one ->lgh_lock lock.
2447 Description: Add loi_kms_set().
2448 Details : Wrap kms updates into a helper function.
2452 Description: Constify instances of struct lsm_operations.
2453 Details : Constify instances of struct lsm_operations.
2457 Description: lu_conf support.
2458 Details : On a server, a file system object is uniquely identified
2459 by a fid, which is sufficient to locate and load all object
2460 state (inode). On a client, on the other hand, more data are
2461 necessary instantiate an object. Change lu_object_find() and
2462 friends to take additional `lu_conf' argument describing object.
2463 Typically this includes layout information.
2467 Description: lu_context fixes.
2468 Details : Introduce new lu_context functions that are needed on the client
2469 side, where some system threads (ptlrpcd) are shared by multiple
2470 modules, and so cannot be stopped during module shutdown.
2474 Description: Add start and stop methods to lu_device_type_operations.
2475 Details : Introduce two new methods in lu_device_type_operations, that are
2476 invoked when first instance of a given type is created and last one
2477 is destroyed respectively. This is need by CLIO.
2481 Description: Add lu_ref support to struct lu_device.
2482 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
2483 track leaked references.
2487 Description: Introduce lu_kmem_descr.
2488 Details : lu_kmem_descr and its companion interface allow to create
2489 and destroy a number of kmem caches at once.
2493 Description: Fix lu_object finalization race.
2494 Details : Fix a race between lu_object_find() finding an object and its
2495 concurrent finalization. This race is (most likely) not possible
2496 on the server, but might happen on the client.
2500 Description: Introduce lu_ref interface.
2501 Details : lu_ref is a debugging module allowing to track references to
2502 a given object. It is quite cpu expensive, and has to be
2503 explicitly enabled with --enable-lu_ref. See usage description
2508 Description: Factor lu_site procfs stats into a separate function.
2509 Details : Separate lu_site stats printing code into a separate function
2510 that can be reused on a client.
2514 Description: Constify instances of struct {lu,dt,md}_device_operations.
2515 Details : Constify instances of struct {lu,dt,md}_device_operations.
2519 Description: Introduce struct md_site and move meta-data specific parts of
2520 struct lu_site here.
2521 Details : Move md-specific fields out of struct lu_site into special struct
2522 md_site, so that lu_site can be used on a client.
2526 Description: Kill mdd_lov_destroy().
2527 Details : Remove unused mdd code.
2531 Description: Add st_block checking to multistat.c.
2532 Details : Add st_block checking to multistat.c.
2536 Description: Add lu_ref support to struct obd_device.
2537 Details : Add lu_ref tracking to obd_device.
2541 Description: Kill obd_set_fail_loc().
2542 Details : Remove unused code.
2546 Description: Add special type for ptlrpc_request interpret functions.
2547 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
2548 there. Actual usage will be in CLIO.
2552 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
2553 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
2554 doesn't work with lockdep.
2558 Description: Add rwv.c test program.
2559 Details : New testing program exercising readv(2) and writev(2) (Qian).
2563 Description: Add sendfile.c test program.
2564 Details : New testing program exercising sendfile(2) (Jay).
2568 Description: Ratelimit a message that can be very frequent.
2569 Details : Ratelimit a memory allocation failure message that can
2574 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2575 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2579 Description: Liblustre build fixes.
2580 Details : Liblustre build fixes.
2584 Description: libcfs: add cfs_{need,cond}_resched() interface.
2585 Details : libcfs: add cfs_{need,cond}_resched() definition and
2586 implementations for Linux, NT, and liblustre.
2588 Severity : enhancement
2590 Description: More exported tunables for mballoc
2591 Details : Add support for tunable preallocation window and new tunables for
2592 large/small requests
2596 Description: Detect corruption of block bitmap and checking for preallocations
2597 Details : Checks validity of on-disk block bitmap. Also it does better
2598 checking of number of applied preallocations. When corruption is
2599 found, it turns filesystem readonly to prevent further corruptions.
2603 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
2604 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
2605 window to 0. If it does hit cache miss, and read-pattern is still
2606 stride-io mode, does not reset the stride window, but also does not
2607 increase the stride window length in this case.
2611 Frequency : only for big-endian servers
2612 Description: Check if system is big-endian while mounting fs with extents feature
2613 Details : Mounting a filesystem with extents feature will fail on big-endian
2614 systems since ext3-based ldiskfs is not supported on big-endian
2615 systems. This can be over-riden with "bigendian_extents" mount option.
2617 Severity : enhancement
2619 Description: The root squash functionality
2620 Details : A security feature, which is to prevent users from being able
2621 to mount lustre on their desktop, run as root, and delete
2622 all of the files in the filesystem. The goal is accomplished by
2623 remapping user id (UID) and group id (GID) of the root user to
2624 a UID and GID specified by the system administartor via Lustre
2625 configuration management server (MGS). The functionality also
2626 allows to specify sets of clients for which the remapping does
2631 Description: Excessive recovery window
2632 Details : With AT enabled, the recovery window can be excessively long (6000+
2633 seconds). To address this problem, we no longer use
2634 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
2635 timeout no longer depends on the service time, it is set to
2636 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
2637 time via pb_service_time.
2641 Description: Watchdog triggered on MDS failover
2642 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
2643 the OSTs know that the MDS "UUID" can be reused for the same export
2644 from a different NID, so we do not need to wait for the export to be
2648 Frequency : rare, only if using MMP with Linux RAID
2650 Description: MMP doesn't work with Linux RAID
2651 Details : While using HA for Lustre servers with Linux RAID, it is possible
2652 that MMP will not detect multiple mounts. To make this work we
2653 need to unplug the device queue in RAID when the MMP block is being
2654 written. Also while reading the MMP block, we should read it from
2655 disk and not the cached one.
2657 Severity : enhancement
2659 Description: open file using fid
2660 Details : A file can be opened using just its fid, like
2661 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
2664 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
2667 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
2668 Details : add proper locking for creation of jbd_4k slab cache
2672 Description: MMP check in ext3_remount() fails without displaying any error
2673 Details : When multiple mount protection fails during remount, proper error
2676 Severity : enhancement
2678 Description: Allow stripe size to be up to 4G-64k
2679 Details : Fix math logic to allow large stripe sizes.
2683 Description: add check for >8TB ldiskfs filesystems
2684 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
2685 Don't allow >8TB ldiskfs filesystems to be mounted without
2686 force_over_8tb mount option
2691 Descriptoin: OSS complains "too many blocks" and crashes
2692 Details : In filter_commitrw_write():
2693 1. lquota_chkquota();
2694 2. obdo_from_inode();
2695 3. lquota_pending_commit();
2696 "2." has changed uid and gid in this case, then caused
2697 lquota_chkquota() and lquota_pending_commit() to use
2698 different uid and gid. That is the root of the bug.
2702 Descriptoin: getstripe doesn't report stripe count = -1 sanely
2703 Details : Detect and print out the 'stripe count = -1' case properly.
2707 Description: Out or order replies might be lost on replay
2708 Details : In ptlrpc_retain_replayable_request if we cannot find retained
2709 request with tid smaller then one currently being added, add it
2710 to the start, not end of the list.
2712 --------------------------------------------------------------------------------
2714 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
2716 * Support for kernels:
2717 2.6.5-7.283 (SLES 9),
2718 2.6.9-55.EL (RHEL 4),
2719 2.6.16.46-0.14 (SLES 10),
2720 2.6.18.8 vanilla (kernel.org)
2721 * Client support for unpatched kernels:
2722 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2723 2.6.16 - 2.6.22 vanilla (kernel.org)
2724 * Due to recently discovered recovery problems, we do not recommend
2725 using patchless RHEL 4 clients with this or any earlier release.
2726 * Recommended e2fsprogs version: 1.39.cfs8
2727 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2728 * Starting with this release, the ldiskfs backing filesystem required
2729 by Lustre is now in its own package, lustre-ldiskfs. This package
2730 should be installed. It is versioned separately from Lustre and
2731 may be released separately in future.
2735 Description: block reactivating mgc import until all deactivates complete
2736 Details : Fix race when failing back MDT/MGS to itself (testing)
2739 Frequency : at statup only
2741 Description: mds_lov_synchronize race leads to various problems
2742 Details : simultaneous MDT->OST connections at startup can cause the
2743 sync to abort, leaving the OSC in a bad state.
2745 Severity : enhancement
2747 Description: add optional extra BUILD_VERSION info
2748 Details : add a new environment variable (namely LUSTRE_VERS) which allows
2749 to override the lustre version.
2752 Frequency : 2.6.18 servers only
2754 Description: ll_kern_mount() doesn't release the module reference
2755 Details : The ldiskfs module reference count never drops down to 0
2756 because ll_kern_mount() doesn't release the module reference.
2761 Description: server LBUG when using old ost_num_threads parameter
2762 Details : Accept the old ost_num_threads parameter but warn that it
2763 is deprecated, and fix an off-by-one error that caused an LBUG.
2768 Description: Transient SCSI error results in persistent IO issue
2769 Details : iobuf->dr_error is not reinitialized to 0 between two
2773 Frequency : sometimes when underlying device returns I/O errors
2775 Description: OSTs not going read-only during write failures
2776 Details : OSTs are not remounted read-only when the journal commit threads
2777 get I/O errors because fsfilt_ext3 calls journal_start/stop()
2778 instead of the ext3 wrappers.
2782 Description: poor connect scaling with increasing client count
2783 Details : Don't run filter_grant_sanity_check for more than 100 exports
2784 to improve scaling for large numbers of clients.
2787 Frequency : SLES10 only
2789 Description: sanity-quota.sh quotacheck failed: rc = -22
2790 Details : Quotas cannot be enabled on SLES10.
2793 Frequency : liblustre clients only
2795 Description: getdirentries does not give error when run on compute nodes
2796 Details : getdirentries does not fail when the size specified as an argument
2797 is too small to contain at least one entry
2799 Severity : enhancement
2801 Description: Add LNET router traceability for debug purposes
2802 Details : If a checksum failure occurs with a router as part of the
2803 IO path, the NID of the last router that forwarded the bulk data
2804 is printed so it can be identified.
2809 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
2810 Details : Due to a race condition, liblustre clients were occasionally
2811 evicted incorrectly.
2813 Severity : enhancement
2815 Description: lfs setstripe use optional parameters instead of postional
2818 Severity : enhancement
2820 Description: Nanosecond timestamp support for ldiskfs
2821 Details : The on-disk ldiskfs filesystem has added support for nanosecond
2822 resolution timestamps. There is not yet support for this at
2823 the Lustre filesystem level.
2826 Frequency : during server recovery
2828 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
2829 Details : request with rq_no_resend flag not awake l_wait_event if they get
2833 Frequency : nfs export on patchless client
2835 Description: connectathon hang when test nfs export over patchless client
2836 Details : Disconnected dentry cannot be found with lookup, so we do not need
2837 to unhash it or make it invalid
2840 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
2841 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
2842 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
2847 Description: open req refcounting wrong on reconnect
2848 Details : If reconnect happened between getting open reply from server and
2849 call to mdc_set_replay_data in ll_file_open, we will schedule
2850 replay for unreferenced request that we are about to free.
2851 Subsequent close will crash in variety of ways.
2852 Check that request is still eligible for replay in
2853 mdc_set_replay_data().
2858 Description: disable writes to filesystem when reading health_check file
2859 Details : the default for reading the health_check proc file has changed
2860 to NOT do a journal transaction and write to disk, because this
2861 can cause reads of the /proc file to hang and block HA state
2862 checking on a healthy but otherwise heavily loaded system. It
2863 is possible to return to the previous behaviour during configure
2864 with --enable-health-write.
2866 Severity : enhancement
2868 Description: 64-bit inode version
2869 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
2870 to the inode. This will be required for version-based recovery.
2875 Description: MDS fails to start if a duplicate client export is detected
2876 Details : in some rare cases it was possible for a client to connect to
2877 an MDS multiple times. Upon recovery the MDS would detect this
2878 and fail during startup. Handle this more gracefully.
2880 Severity : enhancement
2882 Description: Add -o localflock option to simulate old noflock
2884 Details : This will achieve local-only flock/fcntl locks
2890 Description: log_commit_thread vs filter_destroy race leads to crash
2891 Details : Take import reference before releasing llog record semaphore
2896 Description: Wrong request locking in request set processing
2897 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
2898 imp_delayed_list, in this place should be used imp_lock.
2901 Frequency : when reconnection
2903 Description: Grant Leak when osc reconnect to OST
2904 Details : When osc reconnect ost, OST(filter) should check whether it
2905 should grant more space to client by comparing fed_grant and
2906 cl_avail_grant, and return the granted space to client instead
2907 of "new granted" space, because client will call osc_init_grant
2908 to update the client grant space info.
2911 Frequency : when client reconnect to OST
2913 Description: Grant Leak when osc do resend and replay bulk write
2914 Details : When osc reconnect to OST, OST(filter)should clear grant info of
2915 bulk write request, because the grant info while be sync between
2916 OSC and OST when reconnect, and we should ignore the grant info
2917 these of resend/replay write req.
2922 Description: Grant space more than avaiable left space sometimes.
2923 Details : When then OST is about to be full, if two bulk writing from
2924 different clients came to OST. Accord the avaliable space of the
2925 OST, the first req should be permitted, and the second one
2926 should be denied by ENOSPC. But if the seconde arrived before
2927 the first one is commited. The OST might wrongly permit second
2928 writing, which will cause grant space > avaiable space.
2931 Frequency : when client is evicted
2933 Description: Grant might be wrongly erased when osc is evicted by OST
2934 Details : when the import is evicted by server, it will fork another
2935 thread ptlrpc_invalidate_import_thread to invalidate the
2936 import, where the grant will be set to 0. While the original
2937 thread will update the grant it got when connecting. So if
2938 the former happened latter, the grant will be wrongly errased
2939 because of this race.
2944 Description: Checking Stale with correct fid
2945 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
2946 to check whether it is stale, because sometimes, we want the
2947 enqueue happened anyway, and op_data.fid2 will not be initialized.
2949 Severity : enhancement
2951 Description: update patchless client
2952 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
2955 Frequency : only with 2.4 kernel
2957 Description: random memory corruption
2958 Details : size of struct ll_inode_info is to big for union inode.u and this
2959 can be cause of random memory corruption.
2964 Description: Memory leak in recovery
2965 Details : Lov_mds_md was not free in an error handler in mds_create_object.
2966 It should also check obd_fail before fsfilt_start, otherwise if
2967 fsfilt_start return -EROFS,(failover mds during mds recovery).
2968 then the req will return with repmsg->transno = 0 and rc = EROFS.
2969 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
2970 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
2971 be freed no matter whether fsfilt_commit success or not.
2974 Frequency : only with huge count clients
2976 Description: Prevents from taking the superblock lock in llap_from_page for
2978 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
2979 for llap_from_page call in ll_removepage prevents from taking the
2980 superblock lock for a soon died page.
2985 Description: Not check open intent error before release open handle
2986 Details : in some rare cases, the open intent error is not checked before
2987 release open handle, which may cause
2988 ASSERTION(open_req->rq_transno != 0), because it tries to release
2989 the failed open handle.
2994 Description: Set cat log bitmap only after create log success.
2995 Details : in some rare cases, the cat log bitmap is set too early. and it
2996 should be set only after create log success.
3000 Description: Accessing a block bevice can re-enable I/O when Lustre is
3001 tearing down a device.
3002 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
3007 Description: service threads may hog cpus when there are a lot of requests
3008 Details : Insert cond_resched to give other threads a chance to use some CPU
3013 Description: the cat log was not initialized in recovery
3014 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
3015 unlink log on mds will not be initialized until mds post
3016 recovery. And also in mds post recovery, the unlink log will
3017 initialization will be done asynchronausly, so there will be race
3018 between add unlink log and unlink log initialization.
3022 Description: brw_stats were being printed incorrectly
3023 Details : brw_stats were being printed as log2 but all of them were not
3024 recorded as log2. Also remove some code duplication arising from
3025 filter_tally_{read,write}.
3029 Frequency : rare, only in recovery.
3030 Description: ASSERTION(req->rq_type != LI_POISON) failed
3031 Details : imp_lock should be held while iterating over imp_sending_list for
3032 prevent destroy request after get timeout in ptlrpc_queue_wait.
3036 Description: replay-single.sh test 52 fails
3037 Details : A lock's skiplist need to be cleanup when it being unlinked
3038 from its resource list.
3042 Description: Short directio read returns full requested size rather than
3044 Details : Direct I/O operations should return actual amount of bytes
3045 transferred rather than requested size.
3047 Severity : enhancement
3049 Description: metadata RPC reduction (e.g. for rm performance)
3050 Details : decrease the amount of synchronous RPC between clients and servers
3051 by canceling conflicing lock before the operation on the client
3052 and packing thier handles into the main operation RPC to server.
3054 Severity : enhancement
3056 Description: Async OSC create to avoid the blocking unnecessarily.
3057 Details : If a OST has no remain object, system will block on the creating
3058 when need to create a new object on this OST. Now, ways use
3059 pre-created objects when available, instead of blocking on an
3060 empty osc while others are not empty. If we must block, we block
3061 for the shortest possible period of time.
3065 Description: improve handling recoverable errors
3066 Details : if request processig with error which can be recoverable on server
3067 request should be resend, otherwise page released from cache and
3070 Severity : enhancement
3072 Description: refine locking for avoid write wrong info into lov_objid file
3073 Details : fix possible races with add new target and write/update data in
3076 --------------------------------------------------------------------------------
3078 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
3083 Frequency : on some architectures
3085 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
3086 Details : Uninitialized flags sometimes cause configuration commands to
3090 Frequency : patchless clients only
3092 Description: missing __iget() symbol export
3093 Details : The __iget() symbol export is missing. To avoid the need for
3094 this on patchless clients the deathrow inode reaper is turned
3095 off, and we depend on the VM to clean up old inodes. This
3096 dependency was during via the fix for bug 12181.
3098 --------------------------------------------------------------------------------
3100 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
3102 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
3103 INTEROPERATE with older versions automatically. In many cases a
3104 special upgrade step is needed. Please read the
3105 user documentation before upgrading any part of a live system.
3106 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
3107 version will not interoperate with 1.6 betas before beta5 (1.5.95).
3108 * WARNING: Lustre configuration and startup changes are required with
3109 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
3113 Severity : enhancement
3115 Description: MountConf
3116 Details : Lustre configuration is now managed via mkfs and mount
3117 commands instead of lmc and lconf. New obd types (MGS, MGC)
3118 are added for dynamic configuration management. See
3119 https://mail.clusterfs.com/wikis/lustre/MountConf for
3122 Severity : enhancement
3124 Description: dynamic OST addition
3125 Details : OSTs can now be added to a live filesystem
3127 Severity : enhancement
3129 Description: startup order invariance
3130 Details : MDTs and OSTs can be started in any order. Clients only
3131 require the MDT to complete startup.
3133 Severity : enhancement
3135 Description: parallel, asynchronous orphan cleanup
3136 Details : orphan cleanup is now performed in separate threads for each
3137 OST, allowing parallel non-blocking operation.
3139 Severity : enhancement
3141 Description: optimized stripe assignment
3142 Details : stripe assignments are now made based on ost space available,
3143 ost previous usage, and OSS previous usage, in order to try
3144 to optimize storage space and networking resources.
3146 Severity : enhancement
3148 Description: Permanently set tunables
3149 Details : All writable /proc/fs/lustre tunables can now be permanently
3150 set on a per-server basis, at mkfs time or on a live system.
3152 Severity : enhancement
3154 Description: Lustre message v2
3155 Details : Add lustre message format v2.
3157 Severity : enhancement
3159 Description: client OST exclusion list
3160 Details : Clients can be started with a list of OSTs that should be
3161 declared "inactive" for known non-responsive OSTs.
3165 Description: SPEC SFS validation failure on NFS v2 over lustre.
3166 Details : Changes the blocksize for regular files to be 2x RPC size,
3167 and not depend on stripe size.
3169 Severity : enhancement
3171 Description: Multiple MD RPCs in flight.
3172 Details : Further unserialise some read-only MDS RPCs - learn about intents.
3173 To avoid overly-overloading MDS, introduce a limit on number of
3174 MDS RPCs in flight for a single client and add /proc controls
3175 to adjust this limit.
3177 Severity : enhancement
3179 Description: client read/write statistics
3180 Details : Add client read/write call usage stats for performance
3181 analysis of user processes.
3182 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
3183 file access. extents_stats shows chunk size distribution.
3184 extents_stats_per_process show chunk size distribution per
3187 Severity : enhancement
3189 Description: mds statistics
3190 Details : Add detailed mds operations statistics in
3191 /proc/fs/lustre/mds/*/stats.
3195 Description: Failure of copying files with lustre special EAs.
3196 Details : Client side always return success for setxattr call for lustre
3197 special xattr (currently only "trusted.lov").
3202 Description: Refcount LNET uuids
3203 Details : The global LNET uuid list grew linearly with every startup;
3204 refcount repeated list entries instead of always adding to
3207 Severity : enhancement
3209 Description: Dynamic service threads
3210 Details : Within a small range, start extra service threads
3211 automatically when the request queue builds up.
3214 Frequency : mixed-endian client/server environments
3216 Description: mixed-endian crashes
3217 Details : The new msg_v2 system had some failures in mixed-endian
3220 Severity : enhancement
3222 Description: Easy OST removal
3223 Details : OSTs can be permanently deactivated with e.g. 'lctl
3224 conf_param lustre-OST0001.osc.active=0'
3226 Severity : enhancement
3228 Description: MGS proc entries
3229 Details : Added basic proc entries for the MGS showing what filesystems
3232 Severity : enhancement
3234 Description: provide MGS failover
3235 Details : Added config lock reacquisition after MGS server failover.
3237 Severity : enhancement
3239 Description: add Linux 2.4 support
3240 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
3244 Description: a large application tries to do I/O to the same resource and dies
3245 in the middle of it.
3246 Details : Check the req->rq_arrival time after the call to
3247 ost_brw_lock_get(), but before we do anything about
3248 processing it & sending the BULK transfer request. This
3249 should help move old stale pending locks off the queue as
3250 quickly as obd_timeout.
3253 Frequency : when an incorrect nid is specified during startup
3255 Description: ptlrpc connect to non-existant node causes kernel crash
3256 Details : LNET can't be re-entered from an event callback, which
3257 happened when we expire a message after the export has been
3258 cleaned up. Instead, hand the zombie cleanup off to another
3261 Severity : enhancement
3263 Description: plain/inodebits lock performance improvement
3264 Details : Grouping plain/inodebits in granted list by their request modes
3265 and bits policy, thus improving the performance of search through
3269 Frequency : only if OST filesystem is corrupted
3271 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3272 Details : for a short time RPCs with bulk IO are in the replay list,
3273 but replay of bulk IOs is unimplemented. If the OST filesystem
3274 is corrupted due to disk cache incoherency and then replay is
3275 started it is possible to trip an assertion. Avoid putting
3276 committed RPCs into the replay list at all to avoid this issue.
3279 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3281 Description: System hang on startup
3282 Details : This bug allowed the liblustre (e.g. catamount) client to
3283 return to the app before handling all startup RPCs. This
3284 could leave the node unresponsive to lustre network traffic
3285 and manifested as a server ptllnd timeout.
3287 Severity : enhancement
3289 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3290 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
3291 boolean to control whether to print peer debug info when a
3292 client's RPC times out.
3295 Frequency : only for kernels with patches from Lustre below 1.4.3
3297 Description: Remove old rdonly API
3298 Details : Remove old rdonly API which unused from at least lustre 1.4.3
3301 Frequency : only for devices with external journals
3303 Description: Set external device read-only also
3304 Details : During a commanded failover stop, we set the disk device
3305 read-only while the server shuts down. We now also set any
3306 external journal device read-only at the same time.
3309 Frequency : when upgrading from 1.4 while trying to change parameters
3311 Description: The wrong (new) MDC name was used when setting parameters for
3312 upgraded MDT's. Also allows changing of OSC (and MDC)
3313 parameters if --writeconf is specified at tunefs upgrade time.
3316 Frequency : when setting specific ost indicies
3318 Description: QOS code breaks on skipped indicies
3319 Details : Add checks for missing OST indicies in the QOS code, so OSTs
3320 created with --index need not be sequential.
3322 Severity : enhancement
3324 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3325 Details : The uninit_groups feature works in conjunction with the kernel
3326 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3327 up the pass1 processing of e2fsck. This is a read-only feature
3328 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3329 mount filesystems that have had this feature enabled.
3331 Severity : enhancement
3333 Description: Improve multi-block allocation algorithm to avoid fragmentation
3334 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3335 allocation locality and avoid filesystem fragmentation.
3337 ------------------------------------------------------------------------------
3339 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
3341 * Support for kernels:
3342 2.4.21-47.0.1.EL (RHEL 3)
3343 2.6.5-7.283 (SLES 9)
3344 2.6.9-42.0.10.EL (RHEL 4)
3345 2.6.12.6 vanilla (kernel.org)
3346 2.6.16.27-0.9 (SLES 10)
3347 * Recommended e2fsprogs version: 1.39.cfs5
3349 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3353 Frequency : occasional, depends on client load and configuration
3354 Bugzilla : 12181, 12203
3355 Description: data loss for recently-modified files
3357 Details : In some cases it is possible that recently written or created
3358 files may not be written to disk in a timely manner (this should
3359 normally be within 30s unless client IO load is very high).
3360 The problem appears as zero-length files or files that are a
3361 multiple of 1MB in size after a client crash or client eviction
3362 that are missing data at the end of the file.
3364 This problem is more likely to be hit on clients where files are
3365 repeatedly created and unlinked in the same directory, clients
3366 have a large amount of RAM, have many CPUs, the filesystem has
3367 many OSTs, the clients are rebooted frequently, and/or the files
3368 are not accessed by other nodes after being written.
3370 The presence of the problem can be detected by looking at
3371 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
3372 smaller than the second (nr_unused) then dirty files will not
3373 be flushed automatically to disk. "sync; sleep 10" should be
3374 run several times on the node before unmounting it to update
3375 Lustre (this is also safe to run on nodes without this problem).
3377 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
3378 that can cause this same problem, so customers using that kernel
3379 also need to update the kernel in addition to Lustre. In order
3380 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
3382 It is normal that files written just before a client crash (less
3383 than 30s) may not yet have been flushed to disk, even for local
3387 Frequency : frequent on thin XT3 nodes
3389 Description: UUID collision on thin XT3 Linux nodes
3390 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
3391 randomly, since we relied on an insufficiently-seeded PRNG.
3396 Description: OSS hangs after "All ost request buffers busy"
3397 Details : A deadlock between quota and journal operations caused OSS
3398 hangs after printing "All ost request buffers busy."
3401 Frequency : always on liblustre builds
3403 Description: Cleanup compiler warnings on liblustre
3406 Frequency : always on liblustre builds on XT3
3408 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
3412 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
3413 > 32000 subdirectories
3414 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
3415 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
3416 subdirectory count crosses 32000. This will aid e2fsck to
3417 correctly handle more than 32000 subdirectories.
3420 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3422 Description: System hang on startup
3423 Details : This bug allowed the liblustre (e.g. catamount) client to
3424 return to the app before handling all startup RPCs. This
3425 could leave the node unresponsive to lustre network traffic
3426 and manifested as a server ptllnd timeout.
3428 Severity : enhancement
3430 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3431 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
3432 boolean to control whether to print peer debug info when a
3433 client's RPC times out.
3438 Description: make O_SYNC working on 2.6 kernels
3439 Details : 2.6 kernels use different method for mark pages for write,
3440 so need add a code to lustre for O_SYNC work.
3445 Description: Failure to close file and release space on NFS
3446 Details : Put inode details into lock acquired in ll_intent_file_open.
3447 Use mdc_intent_lock in ll_intent_open to properly
3448 detect all kind of errors unhandled by mdc_enqueue.
3453 Description: proc file read during shutdown sometimes raced obd removal,
3455 Details : Add lock to prevent obd access after proc file removal.
3458 Frequency : Only for files larger than 4GB on 32-bit clients.
3460 Description: improperly doing page alignment of locks
3461 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
3462 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3467 Description: RPCs being resent when they shouldn't be
3468 Details : Some RPCs that should not be resent are being resent. This
3469 can cause inconsistencies in the RPC state machine. Do not
3470 resend such requests.
3473 Frequency : rare, only with NFS export
3475 Description: Crash on NFS re-export node
3476 Details : under very unusual load conditions an assertion is hit in
3477 ll_intent_file_open()
3480 Frequency : only if OST filesystem is corrupted
3482 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3483 Details : for a short time RPCs with bulk IO are in the replay list,
3484 but replay of bulk IOs is unimplemented. If the OST filesystem
3485 is corrupted due to disk cache incoherency and then replay is
3486 started it is possible to trip an assertion. Avoid putting
3487 committed RPCs into the replay list at all to avoid this issue.
3492 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3493 Details : Large single O_DIRECT read and write calls can fail to allocate
3494 a sufficiently large buffer to process the request. In case of
3495 allocation failure the allocation is retried with a smaller
3496 buffer and broken into smaller requests.
3498 Severity : enhancement
3500 Description: Add -o localflock option to simulate old noflock behaviour.
3501 Details : This will achieve local-only flock/fcntl locks coherentness.
3506 Description: versioning check is incomplete
3507 Details : Checking the version difference of client vs. server, report
3508 error if the gap is too big.
3513 Description: add support PG_writeback bit
3514 Details : add support for PG_writeback bit for Lustre, for more carefull
3515 work with page cache in 2.6 kernel. This also fix some deadlocks
3516 and remove hack for work O_SYNC with 2.6 kernel.
3518 Severity : enhancement
3520 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3521 Details : The uninit_groups feature works in conjunction with the kernel
3522 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3523 up the pass1 processing of e2fsck. This is a read-only feature
3524 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3525 mount filesystems that have had this feature enabled.
3527 Severity : enhancement
3529 Description: Improve multi-block allocation algorithm to avoid fragmentation
3530 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3531 allocation locality and avoid filesystem fragmentation.
3533 ------------------------------------------------------------------------------
3535 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3537 * Support for kernels:
3538 2.6.9-42.0.3.EL (RHEL 4)
3539 2.6.5-7.276 (SLES 9)
3540 2.4.21-47.0.1.EL (RHEL 3)
3541 2.6.12.6 vanilla (kernel.org)
3542 2.6.16.21-0.8 (SLES10)
3543 * Recommended e2fsprogs version: 1.39.cfs2-0
3545 * The backwards-compatible /proc/sys/portals symlink has been removed
3546 in this release. Before upgrading, please ensure that you change
3547 any configuration scripts or /etc/sysctl.conf files that access
3548 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3549 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3550 in advance of the upgrade on any system running Lustre 1.4.6 or
3551 newer, since /proc/sys/lnet was added in that version.
3552 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3556 Frequency : only when quota is used
3558 Description: avoid scanning export list for quota master
3559 Details : Change the algorithms to avoid scanning export list in order
3560 to improve the efficiency.
3563 Frequency : MDS failover only, very rarely
3565 Description: "went back in time" messages on mds failover
3566 Details : The greatest transno may be lost when the current operation
3567 finishes with an error (transno==0) and the client's last_rcvd
3568 record is over-written. Save the greatest transno in the
3569 mds_last_transno for this case.
3572 Frequency : always for specific kernels and striping counts
3574 Description: client may get "Matching packet too big" without ACL support
3575 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
3576 when trying to access files in certain configurations. The
3577 clients should in fact be denied when mounting because they do
3578 not understand ACLs.
3581 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
3583 Description: many clients connecting with IO in progress causes connect timeouts
3584 Details : Avoid synchronous journal commits to avoid delays caused by many
3585 clients connecting/disconnecting when bulk IO is in progress.
3586 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
3587 OST_IO_PORTAL to avoid delays behind potentially many pending
3591 Frequency : occasionally with multiple writers to a single file
3593 Description: shared writes to file may result in wrong size reported by stat()
3594 Details : Allow growing of kms when extent lock is cancelled
3597 Frequency : always with random mmap IO to multi-striped file
3599 Description: mmap write might be lost if we are writing to a 'hole' in stripe
3600 Details : Only if the hole is at the end of OST object so that kms is too
3601 small. Fix is to increase kms accordingly in ll_nopage.
3604 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
3606 Description: writes to a missing object would leak memory on the OST
3607 Details : If there is an inconsistency between the MDS and OST filesystems,
3608 such that the MDS references an object that doesn't exist, writes
3609 to that object will leak memory due to incorrect cleanup in the
3610 error handling path, eventually running out of memory on the OST.
3615 Description: Creating too long symlink causes lustre errors
3616 Details : Check symlink and name lengths before sending requests to MDS.
3619 Frequency : only if flock is enabled (not on by default)
3621 Description: posix locks not released on fd closure on 2.6.9+
3622 Details : We failed to add posix locks to list of inode locks on 2.6.9+
3623 kernels, this caused such locks not to be released on fd close and
3624 then assertions on fs unmount about still used locks.
3627 Frequency : MDS failover only, very rarely
3629 Description: clients may get ASSERTION(granted_lock != NULL)
3630 Details : When request was taking a long time, and a client was resending
3631 a getattr by name lock request. The were multiple lock requests
3632 with the same client lock handle and
3633 mds_getattr_name->fixup_handle_for_resent_request found one of the
3634 lock handles but later failed with ASSERTION(granted_lock != NULL).
3639 Description: handle->h_buffer_credits > 0, assertion failure
3640 Details : h_buffer_credits is zero after truncate, causing assertion
3641 failure. This patch extends the transaction or creates a new
3645 Frequency : NFS re-export or patchless client
3646 Bugzilla : 11179, 10796
3647 Description: Crash on NFS re-export node (__d_move)
3648 Details : We do not want to hash the dentry if we don't have a lock.
3649 But if this dentry is later used in d_move, we'd hit uninitialised
3650 list head d_hash, so we just do this to init d_hash field but
3651 leave dentry unhashed.
3654 Frequency : NFS re-export or patchless client
3656 Description: NFS exports has problem with symbolic link
3657 Details : lustre client didn't properly install dentry when re-exported
3658 to NFS or running patchless client.
3661 Frequency : NFS re-export or patchless client
3663 Description: Various nfs/patchless fixes.
3664 Details : fixes reuse disconected alias for lookup process - this fixes
3665 warning "find_exported_dentry: npd != pd",
3666 fix permission error with open files at nfs.
3667 fix apply umask when do revalidate.
3670 Frequency : occasional
3672 Description: Crash on NFS re-export node
3673 Details : calling clear_page() on the wrong pointer triggered oops in
3674 generic_mapping_read().
3677 Frequency : rarely, using O_DIRECT IO
3679 Description: unaligned directio crashes client with LASSERT
3680 Details : check for unaligned buffers before trying any requests.
3683 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
3685 Description: stale data returned from RAID cache
3686 Details : If only a small amount of IO is done to the RAID device before
3687 reading it again it is possible to get stale data from the RAID
3688 cache instead of reading it from disk.
3691 Frequency : always for sles10 kernel
3693 Description: sles10 support
3694 Details : ll_follow_link: compile fixes and using of nd_set_link
3695 under newer kernels.
3698 Frequency : depends on arch, kernel and compiler version, always on sles10
3701 Description: recursive or deep enough symlinks cause stack overflow
3702 Details : getting rid of large stack-allocated variable in
3706 Frequency : depends on hardware
3708 Description: lustre write performance loss in the SLES10 kernel
3709 Details : the performance loss is caused by using of write barriers in the
3710 ext3 code. The SLES10 kernel turns barrier support on by
3711 default. The fix is to undo that change for ldiskfs.
3713 ------------------------------------------------------------------------------
3715 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
3717 * Support for kernels:
3718 2.6.9-42.0.3EL (RHEL 4)
3719 2.6.5-7.276 (SLES 9)
3720 2.4.21-47.0.1.EL (RHEL 3)
3721 2.6.12.6 vanilla (kernel.org)
3725 Frequency : quota enabled and large files being deleted
3727 Description: releasing more than 4GB of quota at once hangs OST
3728 Details : If a user deletes more than 4GB of files on a single OST it
3729 will cause the OST to spin in an infinite loop. Release
3730 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3735 Description: statfs data retrieved from /proc may be stale or zero
3736 Details : When reading per-device statfs data from /proc, in the
3737 {kbytes,files}_{total,free,avail} files, it may appear
3738 as zero or be out of date.
3741 Frequency : systems with MD RAID1 external journal devices
3743 Description: lconf's call to blkid is confused by RAID1 journal devices
3744 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3745 of returning all block devices that match the journal UUID.
3748 Frequency : always, for aggregate stripe size over 4GB
3750 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
3751 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
3752 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
3755 Frequency : always if "lfs find" used on a local file/directory
3757 Description: "lfs find" segfaults if used on a local file/directory
3758 Details : The case where a directory component was not specified wasn't
3759 handled correctly. Handle this properly.
3762 Frequency : always on ppc64
3764 Description: the write to an ext3 filesystem mounted with mballoc got stuck
3765 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3766 perform endianness conversion.
3769 Frequency : rarely (truncate to non-zero file size after write under load)
3770 Bugzilla : 10730, 10687
3771 Description: Files padded with zeros to next 4K multiple
3772 Details : With filesystems mounted using the "extents" option (2.6 kernels)
3773 it is possible that files that are truncated to a non-zero size
3774 immediately after being written are filled with zero bytes beyond
3775 the truncated size. No file data is lost.
3777 Severity : enhancement
3779 Description: Allow recovery/failover for liblustre clients.
3780 Details : liblustre clients were unaware of failover configurations until
3783 Severity : enhancement
3785 Description: user file locks should fail when not mounting with flock option
3786 Details : Set up an error-returning stub in ll_file_operations.lock field
3787 to prevent incorrect behaviour when client is mounted without
3788 flock option. Also, set up properly f_op->flock field for
3792 Frequency : always on ia64
3794 Description: "lfs df" loops on printing out MDS statfs information
3795 Details : The obd_ioctl_data was not initialized and in some systems
3796 this caused a failure during the ioctl that did not return
3797 an error. Initialize the struct and return an error on failure.
3800 Frequency : SLES 9 only
3802 Description: Error of copying files with lustre special EAs as root
3803 Details : Client side always return success for setxattr call for lustre
3804 special xattr (currently only "trusted.lov").
3807 Frequency : rarely on clusters with both ia64+i386 clients
3809 Description: ia64+i686 clients doing shared IO on the same file may LBUG
3810 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
3811 clients are doing concurrent writes to the same file it is
3812 possible that the ia64 clients may LASSERT because the OST
3813 extent locks are not PAGE_SIZE aligned. Ensure that grown
3814 locks are always aligned on the request boundary.
3817 Frequency : specific use, occasional
3819 Description: Overwriting in use executable truncates on-disk binary image
3820 Details : If one node attempts to overwrite an executable in use by
3821 another node, we now correctly return ETXTBSY instead of
3822 truncating the file.
3824 Severity : enhancement
3826 Description: Async OSC create to avoid the blocking unnecessarily.
3827 Details : If a OST has no remain object, system will block on the creating
3828 when need to create a new object on this OST. Now, ways use
3829 pre-created objects when available, instead of blocking on an
3830 empty osc while others are not empty. If we must block, we block
3831 for the shortest possible period of time.
3836 Description: chmod on Lustre root is propagated to other clients
3837 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
3838 invalid by the follow_mount time.
3841 Frequency : liblustre clients only
3843 Description: Race in 'instant cancel' lock handling could lead to such locks
3844 never to be granted in case of SMP MDS
3845 Details : Do not destroy not yet granted but cbpending locks in
3849 Frequency : replay/resend of open
3851 Description: non null lock assetion failure in mds_intent_policy
3852 Details : Trying to replay/resend lockless open requests resulted in
3853 mds_open() returning 0 with no lock. Now it sets a flag if
3854 a lock is going to be returned.
3856 Severity : enhancement
3858 Description: Checksum enhancements
3859 Details : New checksum enhancements allow for resending RPCs that failed
3862 Severity : enhancement
3864 Description: Tunables on number of dirty pages in cacche
3865 Details : Allow to set limit on number of dirty pages cached.
3870 Description: client crash on unmount - lock still has references
3871 Details : In some error handling cases it was possible to leak a lock
3872 reference on a client while accessing a file. This was not
3873 harmful to the client during operation, but would cause the
3874 client to crash when the filesystem is unmounted.
3877 Frequency : specific case, rare
3879 Description: ETXTBSY on mds though file not in use by client
3880 Details : ETXTBSY is no longer incorrectly returned when attempting to
3881 chmod or chown a directory that the user previously tried to
3882 execute or a currently-executing binary.
3885 Frequency : extremely rare except on liblustre-based clients
3887 Description: Lustre space not freed when files are deleted
3888 Details : Clean up open-unlinked files after client eviction. Previously
3889 the unlink was skipped and the files remained as orphans.
3894 Description: OST failure "would be an LBUG" in waiting_locks_callback()
3895 Details : In some cases it was possible to send a blocking callback to
3896 a client doing a glimpse, even though that client didn't get
3897 a lock granted. When the glimpse lock is cancelled on the OST
3898 the freed lock is left on the waiting list and corrupted the list.
3901 Frequency : all core dumps
3903 Description: Broke core dumps to lustre
3904 Details : Negative dentry may be unhashed if parent does not have UPDATE
3905 lock, but some callers, e.g. do_coredump, expect dentry to be
3906 hashed after successful create, hash it in ll_create_it.
3908 ------------------------------------------------------------------------------
3910 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
3912 * Support for kernels:
3913 2.6.9-42.0.2.EL (RHEL 4)
3914 2.6.5-7.276 (SLES 9)
3915 2.4.21-40.EL (RHEL 3)
3916 2.6.12.6 vanilla (kernel.org)
3920 Frequency : always on RHEL 3
3922 Description: Number of open files grows over time
3923 Details : The number of open files grows over time, whether or not
3924 Lustre is started. This was due to a filp leak introduced
3925 by one of our kernel patches.
3927 ------------------------------------------------------------------------------
3929 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
3931 * Support for kernels:
3932 2.6.9-42.EL (RHEL 4)
3933 2.6.5-7.276 (SLES 9)
3934 2.4.21-40.EL (RHEL 3)
3935 2.6.12.6 vanilla (kernel.org)
3940 Bugzilla : 5719, 9635, 9792, 9684
3941 Description: OST (or MDS) trips assertions in (re)connection under heavy load
3942 Details : If a server is under heavy load and cannot reply to new
3943 connection requests before the client resends the (re)connect,
3944 the connection handling code can behave badly if two service
3945 threads are concurrently handing separate (re)connections from
3946 the same client. Add better locking to the connection handling
3947 code, and ensure that only a single connection will be processed
3948 for a given client UUID, even if the lock is dropped.
3950 Severity : enhancement
3952 Description: add TCP zero-copy support to kernel
3953 Details : Add support to the kernel TCP stack to allow zero-copy bulk
3954 sends if the hardware supports scatter-gather and checksumming.
3955 This allows socklnd to do client-write and server-read more
3956 efficiently and reduce CPU utilization from skbuf copying.
3959 Frequency : only if NFS exporting from client
3961 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
3962 Details : If setting attributes on a file created under NFS that had
3963 never been opened it would be possible to oops the client
3964 if the file had no objects.
3967 Frequency : always for liblustre
3969 Description: liblustre client does MDS+OSTs setattr RPC for each write
3970 Details : When doing a write from a liblustre client, the client
3971 incorrectly issued an RPC to the MDS and each OST the file was
3972 striped over in order to update the timestamps. When writing
3973 with small chunks and many clients this could overwhelm the MDS
3974 with RPCs. In all cases it would slow down the write because
3975 these RPCs are unnecessary.
3977 Severity : enhancement
3979 Description: allow number of MDS service threads to be changed at module load
3980 Details : It is now possible to change the number of MDS service threads
3981 running. Adding "options mds mds_num_threads={N}" to the MDS's
3982 /etc/modprobe.conf will set the number of threads for the next
3983 time Lustre is restarted (assuming the "mds" module is also
3984 reloaded at that time). The default number of threads will
3985 stay the same, 32 for most systems.
3990 Description: OST crash if filesystem is unformatted or corrupt
3991 Details : If an OST is started on a device that has never been formatted
3992 or if the filesystem is corrupt and cannot even mount then the
3993 error handling cleanup routines would dereference a NULL pointer.
3998 Description: NULL pointer deref in llap_from_page.
3999 Details : get_cache_page_nowait can return a page with NULL (or otherwise
4000 incorrect) mapping if the page was truncated/reclaimed while it was
4001 searched for. Check for this condition and skip such pages when
4002 doing readahead. Introduce extra check to llap_from_page() to
4003 verify page->mapping->host is non-NULL (so page is not anonymous).
4006 Frequency : Sometimes when using sys_sendfile
4008 Description: "page not covered by a lock" warnings from ll_readpage
4009 Details : sendfile called ll_readpage without right page locks present.
4010 Now we introduced ll_file_sendfile that does necessary locking
4011 around call to generic_file_sendfile() much like we do in
4015 Frequency : with certain MDS communication failures at client mount time
4017 Description: NULL pointer deref after failed client mount
4018 Details : a client connection request may delayed by the network layer
4019 and not be sent until after the PTLRPC layer has timed out the
4020 request. If the client fails the mount immediately it will try
4021 to clean up before the network times out the request. Add a
4022 reference from the request import to the obd device and delay
4023 the cleanup until the network drops the request.
4026 Frequency : occasionally during client (re)connect
4028 Description: assertion failure during client (re)connect
4029 Details : processing a client connection request may be delayed by the
4030 client or server longer than the client connect timeout. This
4031 causes the client to resend the connection request. If the
4032 original connection request is replied in this interval, the
4033 client may trip an assertion failure in ptlrpc_connect_interpret()
4034 which thought it would be the only running connect process.
4037 Frequency : only with obd_echo servers and clients that are rebooted
4039 Description: kernel BUG accessing uninitialized data structure
4040 Details : When running an obd_echo server it did not start the ping_evictor
4041 thread, and when a client was evicted an uninitialized data
4042 structure was accessed. Start the ping_evictor in the RPC
4043 service startup instead of the OBD startup.
4045 Severity : enhancement
4046 Bugzilla : 10193 (patchless)
4047 Description: Remove dependency on various unexported kernel interfaces.
4048 Details : No longer need reparent_to_init, exit_mm, exit_files,
4049 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
4052 Frequency : rare (only users of deprecated and unsupported LDAP config)
4054 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
4055 Details : LDAP apparently contains 'lustreName' attributes instead of
4056 'name'. A simple remapping of the name is sufficient.
4059 Frequency : rare (only with non-default dump_on_timeout debug enabled)
4061 Description: waiting_locks_callback trips kernel BUG if client is evicted
4062 Details : Running with the dump_on_timeout debug flag turned on makes
4063 it possible that the waiting_locks_callback() can try to dump
4064 the Lustre kernel debug logs from an interrupt handler. Defer
4065 this log dumping to the expired_lock_main() thread.
4067 Severity : enhancement
4069 Description: Support NFS exporting on 2.6 kernels.
4070 Details : Implement non-rawops metadata methods for NFS server to use without
4071 changing NFS server code.
4074 Frequency : very rare (synthetic metadata workload only)
4076 Description: two racing renames might cause an MDS thread to deadlock
4077 Details : Running the "racer" program may cause one MDS thread to rename
4078 a file from being the source of a rename to being the target of
4079 a rename at exactly the same time that another thread is doing
4080 so, and the second thread has already enqueued these locks after
4081 doing a lookup of the target and is trying to relock them in
4082 order. Ensure that we don't try to re-lock the same resource.
4085 Frequency : only very large systems with liblustre clients
4087 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
4088 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
4089 all OSTs in parallel. This allows the eviction of stale liblustre
4090 clients to proceed much faster than if they were done in series,
4091 and also offers similar improvements for other set_info RPCs.
4096 Description: excessive CPU usage during initial read phase on client
4097 Details : During the initial read phase on a client, it would agressively
4098 retry readahead on the file, consuming too much CPU and impacting
4099 performance (since 1.4.5.8). Improve the readahead algorithm
4100 to avoid this, and also improve some other common cases (read
4101 of small files in particular, where "small" is files smaller than
4102 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
4107 Description: MDS crash when receiving packet with unknown intent.
4108 Details : Do not LBUG in unknown intent case, just return -EFAULT
4110 Severity : enhancement
4111 Bugzilla : 9293, 9385
4112 Description: MDS RPCs are serialised on client. This is unnecessary for some.
4113 Details : Do not serialize getattr (non-intent version) and statfs.
4116 Frequency : occasional, when OST network is overloaded/intermittent
4118 Description: client evicted by OST after bulk IO timeout
4119 Details : If a client sends a bulk IO request (read or write) the OST
4120 may evict the client if it is unresposive to its data GET/PUT
4121 request. This is incorrect if the network is overloaded (takes
4122 too long to transfer the RPC data) or dropped the OST GET/PUT
4123 request. There is no need to evict the client at all, since
4124 the pinger and/or lock callbacks will handle this, and the
4125 client can restart the bulk request.
4128 Frequency : Always when mmapping file with no objects
4130 Description: client crashes when mmapping file with no objects
4131 Details : Check that we actually have objects in a file before doing any
4132 operations on objects in ll_vm_open, ll_vm_close and
4138 Description: Request leak when working with deleted CWD
4139 Details : Introduce advanced request refcount tracking for requests
4140 referenced from lustre intent.
4142 Severity : Enhancement
4144 Description: Cache open file handles on client.
4145 Details : MDS now will return special lock along with openhandle, if
4146 requested and client is allowed to hold openhandle, even if unused,
4147 until such a lock is revoked. Helps NFS a lot, since NFS is opening
4148 closing files for every read/write openration.
4150 Severity : Enhancement
4152 Description: Cache open negative dentries on client when possible.
4153 Details : Guard negative dentries with UPDATE lock on parent dir, drop
4154 negative dentries on lock revocation.
4159 Description: Remounting a client read-only wasn't possible with a zconf mount
4160 Details : It wasn't possible to remount a client read-only with llmount.
4162 Severity : enhancement
4163 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
4164 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
4165 MPICH is included to add Lustre-specific ADIO interfaces.
4166 This is based closely on the UFS ADIO layer and only differs
4167 in file creation, in order to allow the OST striping to be set.
4168 This is user-contributed code and not supported by CFS.
4173 Description: extended inode attributes (immutable, append-only) work improperly
4174 when 2.4 and 2.6 kernels are used on client/server or vice versa
4175 Details : Introduce kernel-independent values for these flags.
4177 Severity : enhancement
4180 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
4181 Details : Many of the /proc/ tunables can only be tuned at a megabyte
4182 granularity. Now, Fractional MB granularity is be supported,
4183 this is very useful for low memory system.
4185 Severity : enhancement
4187 Description: Getattr by fid
4188 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
4189 locks, avoids extra getattr rpc requests to MDS, allows '/' to
4190 have locks and avoids getattr rpc requests for it on every stat.
4193 Frequency : Always, for filesystems larger than 2TB
4195 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4196 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
4197 filesystems larger than 2TB to be mounted with mballoc enabled.
4200 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
4202 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4203 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
4204 filesystems larger than 2TB is mounted then the kernel will
4205 silently corrupt the start of the filesystem. CONFIG_LBD is
4206 enabled for all CFS-supported kernels, but the possibility of
4207 this happening with a modified kernel config exists.
4209 Severity : enhancement
4211 Description: add client O_DIRECT support for 2.6 kernels
4212 Details : It is now possible to do O_DIRECT reads and writes to files
4213 in the Lustre client mountpoint on 2.6 kernel clients.
4215 Severity : enhancement
4217 Description: parallel glimpse, setattr, statfs, punch, destroy requests
4218 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
4219 parallel, not waiting for response from every OST before sending
4220 a rpc to the next OST.
4225 Description: setattr vs write race when updating file timestamps
4226 Details : Client processes that update a file timestamp into the past
4227 right after writing to the file (e.g. tar) it is possible that
4228 the updated file modification time can be reset to the current
4229 time due to a race between processing the setattr and write RPC.
4231 Severity : enhancement
4233 Description: Bring 'lfs find' closer in line with regular Linux find.
4234 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
4235 -print0 options and obtains all the needed info through the lustre
4238 Severity : enhancement
4240 Description: support up to 1024 configured devices on one node
4241 Details : change obd_dev array from statically allocated to dynamically
4242 allocated structs as they are first used to reduce memory usage
4247 Description: Flush dirty partially truncated pages during truncate
4248 Details : Immediatelly flush partially truncated pages in filter_setattr,
4249 this way we completely avoid having any pages in page cache on OST
4250 and can retire ugly workarounds during writes to flush such pages.
4255 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
4256 Details : protect inode from truncation within vfs_unlink() context
4257 just take a reference before calling vfs_unlink() and release it
4258 when parent's i_sem is free.
4263 Description: last_id value checked outside lock on OST caused LASSERT failure
4264 Details : If there were multiple MDS->OST object precreate requests in
4265 flight, it was possible that the OST's last object id was checked
4266 outside a lock and incorrectly tripped an assertion. Move checks
4267 inside locks, and discard old precreate requests.
4270 Frequency : always, if extents are used on OSTs
4272 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
4273 Details : OSTs using the extents format would not zero the high 16 bits of
4274 the index physical block number. This is not a problem for any
4275 OST filesystems smaller than 16TB, and no kernels support ext3
4276 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
4277 new/modified files) and can be fixed for existing filesystems
4278 with e2fsprogs-1.39-cfs1.
4283 Description: import connection selection may be incorrect if timer wraps
4284 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
4285 import connections to be ignored if the 32-bit jiffies counter
4286 wraps. Use a 64-bit jiffies counter.
4289 Frequency : very large clusters immediately after boot
4291 Description: LNET request buffers exhausted under heavy short-term load
4292 Details : If a large number of client requests are generated on a service
4293 that has previously never seen so many requests it is possible
4294 that the request buffer growth cannot keep up with the spike in
4295 demand. Instead of dropping incoming requests, they are held in
4296 the LND until the RPC service can accept more requests.
4299 Frequency : Sometimes during replay
4301 Description: Assertion failure in ll_local_open after replay.
4302 Details : If replay happened on an open request reply before we were able
4303 to set replay handler, reply will become not swabbed tripping the
4304 assertion in ll_local_open. Now we set the handler right after
4305 recognising of open request
4308 Frequency : very rare
4310 Description: kernel reports "badness in vsnprintf"
4311 Details : Reading from the "recovery_status" /proc file in small chunks
4312 may cause a negative length in lprocfs_obd_rd_recovery_status()
4313 call to vsnprintf() (which is otherwise harmless). Exit early
4314 if there is no more space in the output buffer.
4316 Severity : enhancement
4318 Description: clear OBD RPC statistics by writing to them
4319 Details : It is now possible to clear the OBD RPC statistics by writing
4320 to the "stats" file.
4325 Description: Client mtime is not the same on different clients after utimes
4326 Details : In some cases, the client was using the utimes() syscall on
4327 a file cached on another node. The clients now validate the
4328 ctime from the MDS + OSTs to determine which one is right.
4333 Description: Inability to activate failout mode
4334 Details : lconf script incorrectly assumed that in pythong string's numeric
4335 value is used in comparisons.
4338 Frequency : always with multiple stripes per file
4340 Description: Inefficient object allocation for mutli-stripe files
4341 Details : When selecting which OSTs to stripe files over, for files with
4342 a stripe count that divides evenly into the number of OSTs,
4343 the MDS is always picking the same starting OST for each file.
4344 Return the OST selection heuristic to the original design.
4349 Description: mount failures may take full timeout to return an error
4350 Details : Under some heavy load conditions it is possible that a
4351 failed mount can wait for the full obd_timeout interval,
4352 possibly several minutes, before reporting an error.
4353 Instead return an error as soon as the status is known.
4355 Frequency : quota enabled and large files being deleted
4357 Description: releasing more than 4GB of quota at once hangs OST
4358 Details : If a user deletes more than 4GB of files on a single OST it
4359 will cause the OST to spin in an infinite loop. Release
4360 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4365 Description: statfs data retrieved from /proc may be stale or zero
4366 Details : When reading per-device statfs data from /proc, in the
4367 {kbytes,files}_{total,free,avail} files, it may appear
4368 as zero or be out of date.
4371 Frequency : systems with MD RAID1 external journal devices
4373 Description: lconf's call to blkid is confused by RAID1 journal devices
4374 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4375 of returning all block devices that match the journal UUID.
4378 Frequency : always, for aggregate stripe size over 4GB
4380 Description: assertion fails when trying to use 4GB stripe size
4381 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
4382 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
4385 Frequency : always on ppc64
4387 Description: the first write on an ext3 filesystem with mballoc got stuck
4388 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4389 perform endianness conversion.
4391 ------------------------------------------------------------------------------
4393 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4395 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
4396 INTEROPERATE with older versions automatically. Please read the
4397 user documentation before upgrading any part of a live system.
4398 * WARNING: Lustre networking configuration changes are required with
4399 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
4402 * Support for kernels:
4403 2.6.9-22.0.2.EL (RHEL 4)
4404 2.6.5-7.244 (SLES 9)
4405 2.6.12.6 vanilla (kernel.org)
4408 Severity : enhancement
4409 Bugzilla : 7981/8208
4410 Description: Introduced Lustre Networking (LNET)
4411 Details : LNET is new networking infrastructure for Lustre, it includes
4412 a reorganized network configuration mode (see the user
4413 documentation for full details) as well as support for routing
4414 between different network fabrics. Lustre Networking Devices
4415 (LNDs) for the supported network fabrics have also been
4416 created for this new infrastructure.
4418 Severity : enhancement
4419 Description: Introduced Access control lists
4420 Details : clients can set ACLs on files and directories in order to have
4421 more fine-grained permissions than the standard Unix UGO+RWX.
4422 The MDS must be started with the "-o acl" mount option.
4424 Severity : enhancement
4425 Description: Introduced filesystem quotas
4426 Details : Administrators may now establish per-user quotas on the
4429 Severity : enhancement
4431 Description: Configuration change for the XT3
4432 The PTLLND is now used to run Lustre over Portals on the XT3
4433 The configure option(s) --with-cray-portals are no longer used.
4434 Rather --with-portals=<path-to-portals-includes> is used to
4435 enable building on the XT3. In addition to enable XT3 specific
4436 features the option --enable-cray-xt3 must be used.
4441 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
4442 Details : A race between allocating a new llcd and re-getting the llcd_lock
4443 allowed another thread to grab newly-allocated llcd.
4445 Severity : enhancement
4447 Description: 2.6 OST async journal commit and locking fix to improve performance
4448 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
4449 2.6 kernels are now async as they already were in 2.4 kernels so
4450 that they can commit concurrently with the network bulk transfer.
4451 For block-allocated files the filter allocation semaphore is held
4452 to avoid filesystem fragmentation during allocation. BKL lock
4453 removed for 2.6 xattr operations where it is no longer needed.
4458 Description: lconf incorrectly determined whether two IP networks could talk
4459 Details : In some more complicated routing and multiple-network
4460 configurations, lconf will avoid trying to make a network
4461 connection to a disjoint part of the IP space. It was doing the
4462 math incorrectly for one set of cases.
4467 Description: Fix for potential infinite loop processing records in an llog.
4468 Details : If an llog record is corrupted/zeroed, it is possible to loop
4469 forever in llog_process(). Validate the llog record length
4470 and skip the remainder of the block on error.
4473 Frequency : occasional (liblustre only)
4475 Description: liblustre could not open files whose last component is a symlink
4476 Details : sysio_path_walk() would incorrectly pass the open intent to
4477 intermediate path components.
4480 Frequency : rare (liblustre only with non-standard tuning)
4481 Bugzilla : 7201 (7350)
4482 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4483 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4484 after it has been released, only do this for referenced locks
4486 Severity : enhancement
4488 Description: specifying an (invalid) directory default stripe_size of -1
4489 would reset the directory default striping
4490 Details : stripe_size -1 was used internally to signal directory stripe
4491 removal, now use "all default" to signal dir stripe removal
4492 as a directory striping of "all default" is not useful
4495 Frequency : common for large clusters running liblustre clients
4497 Description: doing an ls when liblustre clients are running is slow
4498 Details : sending a glimpse AST to a liblustre client waits for every AST
4499 to time out, as liblustre clients will not respond. Since they
4500 cannot cache data we refresh the OST lock LVB from disk instead.
4502 Severity : enhancement
4504 Description: doing an ls at the same time as file IO can be slow
4505 Details : enqueue and other "small" requests can be blocked behind many
4506 large IO requests. Create a new OST IO portal for non-IO
4507 requests so they can be processed faster.
4510 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4512 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4513 created file with mode 000
4514 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4515 allow an ftruncate() as a non-root user to a file with mode 000.
4516 Lustre now respects this flag to disable mode checks when
4517 truncating a file owned by the user
4520 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4523 Description: Revoking locks from clients that went dead or catatonic might take
4525 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4526 cancellation of such locks instant on servers without waiting for
4527 any reply from clients. Clients drops these locks when cancel
4528 notification from server is received without replying.
4531 Frequency : liblustre-only, when liblustre client dies or becomes busy
4533 Description: Doing ls on Linux clients can take a long time with active
4535 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4536 granting such locks to it in the first place if possible. Locks
4537 are taken by proxy on the OST during the read or write and
4538 dropped immediately afterward. Add connect flags handling, do
4539 not grant locks to liblustre clients for glimpse ASTs.
4541 Severity : enhancement
4543 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4544 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4545 sequential reads and behaves badly with applications doing random
4546 reads. Improve it by reading ahead at least read region, and
4547 avoiding excessive large RPC for small reads.
4549 Severity : enhancement
4551 Description: Creating more than 1000 files for a single job may cause a load
4552 imbalance on the OSTs if there are also a large number of OSTs.
4553 Details : qos_prep_create() uses an OST index reseed value that is an
4554 even multiple of the number of available OSTs so that if the
4555 reseed happens in the middle of the object allocation it will
4556 still utilize the OSTs as uniformly as possible.
4561 Description: OST or MDS may oops in ping_evictor_main()
4562 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
4563 but doesn't restart at beginning of obd_exports_timed list
4566 Severity : enhancement
4568 Description: improve by-nid export eviction on the MDS and OST
4569 Details : allow multiple exports with the same NID to be evicted at one
4570 time without re-searching the exports list.
4573 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
4575 Description: MDS may oops in groups_free()
4576 Details : in rare race conditions a newly allocated group_info struct is
4577 freed again, and this can be NULL. The 2.4 compatibility code
4578 for groups_free() checked for a NULL pointer, but 2.6 did not.
4581 Frequency : common for liblustre clients doing little filesystem IO
4582 Bugzilla : 9352, 7313
4583 Description: server may evict liblustre clients accessing contended locks
4584 Details : if a client is granted a lock or receives a completion AST
4585 with a blocking AST already set it would not reply to the AST
4586 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
4589 Frequency : lfs setstripe, only systems with more than 160 OSTs
4591 Description: unable to set striping with a starting offset beyond OST 160
4592 Details : llapi_create_file() incorrectly limited the starting stripe
4593 index to the maximum single-file stripe count.
4596 Frequency : LDAP users only
4598 Description: lconf did not handle in-kernel recovery with LDAP properly
4599 Details : lconf/LustreDB get_refs() is searching the wrong namespace
4601 Severity : enhancement
4603 Description: bind OST threads to NUMA nodes to improve performance
4604 Details : all OST threads are uniformly bound to CPUs on a single NUMA
4605 node and do their allocations there to localize memory access
4607 Severity : enhancement
4609 Description: llmount can determine client NID directly from Myrinet (GM)
4610 Details : the client NID code from gmnalnid was moved directly into
4611 llmount, removing the need to use this or specifying the
4612 client NID explicitly when mounting GM clients with zeroconf
4615 Frequency : if client is started with down MDS
4617 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
4618 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
4619 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
4620 ptlrpc_import_delayed_req() because we didn't check if the
4621 request was interrupted, and we also didn't break out of the
4622 event loop if there was no timeout
4627 Description: data loss during non-page-aligned writes to a single file from
4628 both multiple nodes and multiple threads on one node at same time
4629 Details : updates to KMS and lsm weren't protected by common lock. Resulting
4630 inconsistency led to false short-reads, that were cached and later
4631 used by ->prepare_write() to fill in partially written page,
4632 leading to data loss.
4635 Frequency : always, if lconf --abort_recovery used
4637 Description: lconf --abort_recovery fails with 'Operation not supported'
4638 Details : lconf was attempting to abort recovery on the MDT device and not
4641 Severity : enhancement
4643 Description: remove cleanup logs
4644 Details : replace lconf-generated cleanup logs with lustre internal
4645 cleanup routines. Eliminates the need for client-cleanup and
4648 Severity : enhancement
4650 Description: add support for EAs (user and system) on lustre filesystems
4651 Details : it is now possible to store extended attributes in the Lustre
4652 client filesystem, and with the user_xattr mount option it
4653 is possible to allow users to store EAs on their files also
4655 Severity : enhancement
4657 Description: Add possibility (config option) to show minimal available OST free
4659 Details : When compiled with --enable-mindf configure option, statfs(2)
4660 (and so, df) will return least minimal free space available from
4661 all OSTs as amount of free space on FS, instead of summary of
4662 free spaces of all OSTs.
4664 Severity : enhancement
4666 Description: do not expand extent locks acquired on OST-side
4667 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
4668 by server: they are not cached anyway.
4671 Frequency : when mmap is used/binaries executed from Lustre
4673 Description: Unmmap pages before throwing them away from read cache.
4674 Details : llap_shrink cache now attempts to unmap pages before discarding
4675 them (if unmapping failed - do not discard). SLES9 kernel has
4676 extra checks that trigger if this unmapping is not done first.
4681 Description: lconf didn't resolve symlinks before checking to see whether a
4682 given mountpoint was already in use
4685 Frequency : when migrating failover services
4686 Bugzilla : 6395, 9514
4687 Description: When migrating a subset of services from a node (e.g. failback
4688 from a failover service node) the remaining services would
4689 time out and evict clients.
4690 Details : lconf --force (implied by --failover) sets the global obd_timeout
4691 to 5 seconds in order to quickly disconnect, but this caused
4692 other RPCs to time out too quickly. Do not change the global
4693 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
4695 Severity : enhancement
4696 Frequency : if MDS is started with down OST
4697 Bugzilla : 9439,5706
4698 Description: Allow startup/shutdown of an MDS without depending on the
4699 availability of the OSTs.
4700 Details : Asynchronously call mds_lov_synchronize during MDS startup.
4701 Add appropriate locking and lov-osc refcounts for safe
4702 cleaning. Add osc abort_inflight calls in case the
4703 synchronize never started.
4706 Frequency : occasional (Cray XT3 only)
4708 Description: root not authorized to access files in CRAY_PORTALS environment
4709 Details : The client process capabilities were not honoured on the MDS in
4710 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
4711 been accessed by an authorized user then root was able to access
4712 the file on the local client also. The root user capabilities
4713 are now allowed on the MDS, as this environment has secure UID.
4716 Frequency : occasional
4718 Description: ldiskfs "too long searching" message happens too often
4719 Details : A debugging message (otherwise harmless) prints too often on
4720 the OST console. This has been reduced to only happen when
4721 there are fragmentation problems on the filesystem.
4726 Description: Division by zero in statfs when all OSCs are inactive
4727 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
4728 lov_statfs divides by value returned by lov_get_stripecnt().
4732 Bugzilla : 9489, 3273
4733 Description: First write from each client to each OST was only 4kB in size,
4734 to initialize client writeback cache, which caused sub-optimal
4735 RPCs and poor layout on disk for the first writen file.
4736 Details : Clients now request an initial cache grant at (re)connect time
4737 and so that they can start streaming writes to the cache right
4738 away and always do full-sized RPCs if there is enough data.
4739 If the OST is rebooted the client also re-establishes its grant
4740 so that client cached writes will be honoured under the grant.
4745 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
4746 Details : Now I/O RPCs go to different portal number and (presumably) fast
4747 lock requests (and glimses) and other RPCs get their own service
4748 threads pool that should be able to service those RPCs
4751 Severity : enhancement
4753 Description: Ability to exchange lustre version between client and servers and
4754 issue warnings at client side if client is too old. Also for
4755 liblustre clients there is ability to refuse connection of too old
4757 Details : New 'version' field is added to connect data structure that is
4758 filled with version info. That info is later checked by server and
4762 Frequency : rare, liblustre only.
4763 Bugzilla : 9296, 9581
4764 Description: Two simultaneous writes from liblustre at offset within same page
4765 might proceed at the same time overwriting eachother with stale
4767 Details : I/O lock withing llu_file_prwv was released too early, before data
4768 actually was hitting the wire. Extended lock-holding time until
4769 server acknowledges receiving data.
4772 Frequency : extremely rare. Never observed in practice.
4774 Description: avoid generating lustre_handle cookie of 0.
4775 Details : class_handle_hash() generates handle cookies by incrementing
4776 global counter, and can hit 0 occasionaly (this is unlikely, but
4777 not impossible, because initial value of cookie counter is
4778 selected randonly). Value of 0 is used as a sentinel meaning
4779 "unassigned handle" --- avoid it. Also coalesce two critical
4780 sections in this function into one.
4782 Severity : enhancement
4784 Description: allow liblustre clients to delegate truncate locking to OST
4785 Details : To avoid overhead of locking, liblustre client instructs OST to
4786 take extent lock in ost_punch() on client's behalf. New connection
4787 flag is added to handle backward compatibility.
4789 Severity : enhancement
4790 Bugzilla : 4928, 7341, 9758
4791 Description: allow number of OST service threads to be specified
4792 Details : a module parameter allows the number of OST service threads
4793 to be specified via "options ost ost_num_threads={N}" in the
4794 OSS's /etc/modules.conf or /etc/modprobe.conf.
4798 Bugzilla : 6146, 9635, 9895
4799 Description: servers crash with bad pointer in target_handle_connect()
4800 Details : In rare cases when a client is reconnecting it was possible that
4801 the connection request was the last reference for that export.
4802 We would temporarily drop the export reference and get a new
4803 one, but this may have been the last reference and the export
4804 was just destroyed. Get new reference before dropping old one.
4806 Severity : enhancement
4807 Frequency : if client is started with failover MDS
4809 Description: Allow multiple MDS hostnames in the mount command
4810 Details : Try to read the configuration from all specified MDS
4811 hostnames during a client mount in case the "primary"
4814 Severity : enhancement
4816 Description: Stop sending data to evicted clients as soon as possible.
4817 Details : Check if the client we are about to send or are sending data to
4818 was evicted already. (Check is done every second of waiting,
4819 for which l_wait_event interface was extended to allow checking
4820 of exit condition at specified intervals).
4823 Frequency : rare, normally only when NFS exporting is done from client
4825 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
4827 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
4830 Frequency : always, if srand() is called before liblustre initialization
4832 Description: Liblustre uses system PRNG disturbing its usage by user application
4833 Details : Introduce internal to lustre fast and high-quality PRNG for
4834 lustre usage and make liblustre and some other places in generic
4835 lustre code to use it.
4837 Severity : enhancement
4838 Bugzilla : 9477, 9557, 9870
4839 Description: Verify that the MDS configuration logs are updated when xml is
4840 Details : Check if the .xml configuration logs are newer than the config
4841 logs stored on the MDS and report an error if this is the case.
4842 Request --write-conf, or allow starting with --old_conf.
4844 Severity : enhancement
4846 Description: Handle symlinks in the path when checking if Lustre is mounted.
4847 Details : Resolve intermediate symlinks when checking if a client has
4848 mounted a filesystem to avoid duplicate client mounts.
4853 Description: lconf can hit an error exception but still return success.
4854 Details : The lconf command catches the Command error exception at the top
4855 level script context and will exit with the associated exit
4856 status, but doesn't ensure that this exit status is non-zero.
4861 Description: failure of ptlrpc thread startup can cause oops
4862 Details : Starting a ptlrpc service thread can fail if there are a large
4863 number of threads or the server memory is very fragmented.
4864 Handle this without oopsing.
4867 Frequency : always, only if liblustre and non-default acceptor port was used
4869 Description: liblustre cannot connect to servers with non-default acceptor port
4870 Details : tcpnal_set_default_params() was not called and was therefore
4871 ignoring the environment varaible TCPNAL_PORT, as well as other
4872 TCPNAL_ environment variables
4877 Description: two objects could be created on the same OST for a single file
4878 Details : If an OST is down, in some cases it was possible to create two
4879 objects on a single OST for a single file. No problems other
4880 than potential performance impact and spurious error messages.
4884 Bugzilla : 5681, 9562
4885 Description: Client may oops in ll_unhash_aliases
4886 Details : Client dcache may become inconsistent in race condition.
4887 In some cases "getcwd" can fail if the current directory is
4893 Description: Inode refcounting problems in NFS export code
4894 Details : link_raw functions used to call d_instantiate without obtaining
4895 extra inode reference first.
4899 Bugzilla : 9942, 9903
4900 Description: Referencing freed requests leading to crash, memleaks with NFS.
4901 Details : We used to require that call to ll_revalidate_it was always
4902 followed by ll_lookup_it. Also with revalidate_special() it is
4903 possible to call ll_revalidate_it() twice for the same dentry
4904 even if first occurence returned success. This fix changes semantic
4905 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
4906 reference on a request referred from the intent.
4907 ll_intent_release() then releases such a request.
4910 Frequency : rare, normally benchmark loads only
4912 Description: unlinked inodes were kept in memory on the client
4913 Details : If a client is repeatedly creating and unlinking files it
4914 can accumulate a lot of stale inodes in the inode slab cache.
4915 If there is no other client load running this can cause the
4916 client node to run out of memory. Instead flush old inodes
4917 from client cache that have the same inode number as a new inode.
4920 Frequency : SLES9 2.6.5 kernel and long filenames only
4921 Bugzilla : 9969, 10379
4922 Description: utime reports stale NFS file handle
4923 Details : SLES9 uses out-of-dentry names in some cases, which confused
4924 the lustre dentry revalidation. Change it to always use the
4928 Frequency : rare, unless heavy write-truncate concurrency is continuous
4929 Bugzilla : 4180, 6984, 7171, 9963, 9331
4930 Description: OST becomes very slow and/or deadlocked during object unlink
4931 Details : filter_destroy() was holding onto the parent directory lock
4932 while truncating+unlinking objects. For very large objects this
4933 may block other threads for a long time and slow overall OST
4934 responsiveness. It may also be possible to get a lock ordering
4935 deadlock in this case, or run out of journal credits because of
4936 the combined truncate+unlink. Solution is to do object truncate
4937 first in one transaction without parent lock, and then do the
4938 final unlink in a new transaction with the parent lock. This
4939 reduces the lock hold time dramatically.
4942 Frequency : rare, 2.4 kernels only
4944 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
4945 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
4946 before unmounting their filesystem in order to avoid deadlock.
4947 The kernel_locked() function in 2.4 kernels only checks whether
4948 the kernel lock is held, not whether it is this process that is
4949 holding it as 2.6 kernels do.
4954 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
4955 Details : The client ptlrpc code may be trying to reconnect to a down
4956 server before a previous connection attempt has timed out.
4957 Increase the reconnect interval to be longer than the connection
4958 timeout interval to avoid sending duplicate connections to
4962 Frequency : echo_client brw_test command
4964 Description: fix echo_client to work with OST preallocated code
4965 Details : OST preallocation code (5137) didn't take echo_client IO path
4966 into account: echo_client calls filter methods outside of any
4967 OST thread and, hence, there is no per-thread preallocated
4968 pages and buffers to use. Solution: hijack pga pages for IO. As
4969 a byproduct, this avoids unnecessary data copying.
4973 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
4974 Description: Client can oops in mdc_commit_close() after open replay
4975 Details : It was possible for the MDS to return an open request with no
4976 transaction number in mds_finish_transno() if the client was
4977 evicted, but without actually returning an error. Clients
4978 would later try to replay that open and may trip an assertion
4979 Simplify the client close codepath, and always return an error
4980 from the MDS in case the open is not successful.
4983 Frequency : rare, 2.6 OSTs only
4985 Description: OST may deadlock under high load on fragmented files
4986 Details : If there was a heavy load and highly-fragmented OST filesystems
4987 it was possible to have all the OST threads deadlock waiting on
4988 allocation of biovecs, because the biovecs were not released
4989 until the entire RPC IO was completed. Instead, release biovecs
4990 as soon as they are complete to ensure forward IO progress.
4992 Severity : enhancement
4994 Description: Support for specifying external journal device at mount
4995 Details : If an OST or MDS device is formatted with an external journal
4996 device, this device major/minor is stored in the ext3 superblock
4997 and may not be valid for failover. Allow detecting and
4998 specifying the external journal at mount time.
5003 Description: Mounting an MDS with pending unlinked files may cause oops
5004 Details : target_finish_recovery() calls mds_postrecov() which returned
5005 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
5006 considers this an error and immediately begins cleaning up the
5007 lov, just after starting the mds_lov process
5009 Severity : enhancement
5011 Description: Implement 'lfs df' to report actual free space on per-OST basis
5012 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
5013 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
5014 usage of objects; '-h' to report in human readable format.
5016 ------------------------------------------------------------------------------
5018 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
5025 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
5026 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
5027 at startup during orphan unlinking. Instead, skip buddy update
5028 before setup, as it will be regenerated after recovery is complete.
5033 Description: If an OST is inactive, its locks might reference stale inodes.
5034 Details : lov_change_cbdata() must iterate over all namespaces, even if
5035 they are inactive to clear inode references from the lock.
5037 Severity : enhancement
5038 Frequency : occasional, if non-standard max_dirty_mb used
5040 Description: Client will block write RPCs if not enough grant
5041 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
5042 then the client will block writes while waiting for another RPC
5043 to complete instead of consuming its dirty limit. With change
5044 we get improved performance when max_dirty_mb is small.
5046 Severity : enhancement
5047 Bugzilla : 3389, 6253
5048 Description: Add support for supplementary groups on the MDS.
5049 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
5050 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
5051 lookups for user supplementary groups into a cache.
5055 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
5056 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
5057 for liblustre O_EXCL is enforced by MDS.
5061 Description: readdir never returns NULL in liblustre.
5062 Details : Corrected llu_iop_getdirentries logic, to return offset of next
5063 dentry in struct dirent.
5067 Frequency : liblustre only, depends on application IO pattern
5068 Description: liblustre clients evicted if not contacting servers
5069 Details : Don't put liblustre clients into the ping_evictor list, so
5070 they will not be evicted by the pinger ever.
5072 Severity : enhancement
5074 Description: Add ability to evict clients by NID from MDS.
5075 Details : By echoing "nid:$NID" string into
5076 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
5077 $NID would be instantly evicted from this MDS and from all active
5078 OSTs connected to it.
5082 Description: Do not query file size twice, somewhat slowing stat(2) calls.
5083 Details : lookup_it_finish() used to query file size from OSTs that was not
5088 Description: service threads change working directory to that of init
5089 Details : Starting lustre service threads may pin the working directory
5090 of the parent thread, making that filesystem busy. Threads
5091 now change to the working directory of init to avoid this.
5095 Frequency : during shutdown only
5096 Description: shutdown with a failed MDS or OST can cause unmount to hang
5097 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
5102 Frequency : 2.6 only
5103 Description: chmod/chown may include an extra supplementary group
5104 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
5105 supplementary group and if none is specified this is used.
5108 Bugzilla : 5479 (6816)
5110 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
5111 Details : If lookup is in progress on a file that is unlinked we might try
5112 to revalidate the inode and fail in revalidate after lookup is
5113 complete and ll_file_open() enqueues the open again but
5114 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
5117 Frequency : always, if lconf --abort_recovery used
5119 Description: lconf --abort_recovery fails with 'Operation not supported'
5120 Details : lconf was attempting to abort recovery on the MDT device and not
5123 ------------------------------------------------------------------------------
5125 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
5130 Frequency : rare (only unsupported configurations with a node running as an
5132 Bugzilla : 6514, 5137
5133 Description: Mounting a Lustre file system on a node running as an OST could
5135 Details : OSTs now preallocates memory needed to write out data at
5136 startup, instead of when needed, to avoid having to
5137 allocate memory in possibly low memory situations.
5138 Specifically, if the file system is mounted on on OST,
5139 memory pressure could force it to try to write out data,
5140 which it needed to allocate memory to do. Due to the low
5141 memory, it would be unable to do so and the node would
5142 become unresponsive.
5144 Severity : enhancement
5146 Description: Addition of lconf --service command line option
5147 Details : lconf now accepts a '--service <arg>' option, which is
5148 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
5150 Severity : enhancement
5152 Description: Failover mode is now the default for OSTs.
5153 Details : By default, OSTs will now run in failover mode. To return to
5154 the old behaviour, add '--failout' to the lmc line for OSTs.
5156 Severity : enhancement
5158 Description: Health checks are now provided for MDS and OSTs
5159 Details : Additional detailed health check information on MSD and OSTs
5160 is now provided through the procfs health_check value.
5163 Frequency : occasional, depends on IO load
5165 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
5166 after numerous create/delete cycles
5167 Details : The ext3 inode allocation policy would not allocate new inodes
5168 very well on the OSTs because there are no new directories
5169 being created. Instead we look for groups with free space if
5170 the parent directories are nearly full.
5175 Description: Network or server problems during mount may cause partially
5176 mounted clients instead of returning an error.
5177 Details : The config llog parsing code may overwrite the error return
5178 code during mount error handling, returning success instead
5184 Description: MDS can fail to allocate large reply buffers
5185 Details : After long uptimes the MDS can fail to allocate large reply
5186 buffers (e.g. zconf client mount config records) due to memory
5187 fragmentation or consumption by the buffer cache. Preallocate
5188 some large reply buffers so that these replies can be sent even
5189 under memory pressure.
5193 Frequency : rare (liblustre)
5194 Description: fsx running with liblustre complained that using truncate() to
5195 extend the file doesn't work. This patch corrects that issue.
5196 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
5197 ATTR_SIZE and lsm use in llu_setattr_raw.
5201 Frequency : rare, only 2.6 kernels
5202 Description: Unusual file access patterns on the MDS may result in inode
5203 data being lost in very rare circumstances.
5204 Details : Bad interaction between the ea-in-inode patch and the "no-read"
5205 code in the 2.6 kernel caused the inode and/or EA data not to
5206 be read from disk, causing single-file corruption.
5210 Frequency : rare, only 2.6 filesystems using extents
5211 Description: Heavy concurrent write and delete load may cause data corruption.
5212 Details : It was possible under high-load situations to have an extent
5213 metadata block in the block device cache from a just-unlinked
5214 file overwrite a newly-allocated data block. We now unmap any
5215 metadata buffers that alias just-allocated data blocks.
5219 Frequency : filesystems with default stripe_count larger than 77
5220 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
5221 Details : lconf specifies an inode size of 4096 bytes when the default
5222 stripe_count is larger than 77. This conflicts with the default
5223 inode density of 1 per 4096 bytes. Allocate smaller inodes in
5224 this case to avoid pinning too much memory for large EAs.
5226 ------------------------------------------------------------------------------
5228 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
5233 Frequency : rare (extremely heavy IO load with hundreds of clients)
5235 Description: Client is evicted, gets IO error writing to file
5236 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
5237 caused clients to be evicted for AST timeouts. The fixes in
5238 bug 5192 mean we no longer need to have such short AST timeouts
5239 so ldlm_timeout has been increased.
5242 Frequency : occasional during --force or --failover shutdown under load
5243 Bugzilla : 5949, 4834
5244 Description: Server oops/LBUG if stopped with --force or --failover under load
5245 Details : a collection of import/export refcount and cleanup ordering
5246 issues fixed for safer force cleanup
5249 Frequency : only filesystems larger than 120 OSTs
5250 Bugzilla : 5990, 6223
5251 Description: lfs getstripe would oops on a very large filesystem
5252 Details : lov_getconfig used kfree on vmalloc'd memory
5255 Frequency : only filesystems exporting via NFS to Solaris 10 clients
5256 Bugzilla : 6242, 6243
5257 Description: reading from files that had been truncated to a non-zero size
5258 but never opened returned no data
5259 Details : ll_file_read() reads zeros from no-object files to EOF
5264 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
5265 Details : There's little harm in aborting MDS/OSS recovery and letting it
5266 try again, so I removed the LASSERT and return an error instead.
5268 Severity : enhancement
5270 Description: New debugging infrastructure for tracking down data corruption
5271 Details : The I/O checksum code was replaced to: (a) control it at runtime,
5272 (b) cover more of the client-side code path, and (c) try to narrow
5273 down where problems occurred
5277 Bugzilla : 3819, 4364, 4397, 6313
5278 Description: Racing close and eviction MDS could cause assertion in mds_close
5279 Details : It was possible to get multiple mfd references during close and
5280 client eviction, leading to one thread referencing a freed mfd.
5282 Severity: : enhancement
5283 Bugzilla : 3262, 6359
5284 Description: Attempts to reconnect to servers are now more aggressive.
5285 Details : This builds on the enhanced upcall-less recovery that was added
5286 in 1.4.2. When trying to reconnect to servers, clients will
5287 now try each server in the failover group every 10 seconds. By
5288 default, clients would previously try one server every 25 seconds.
5293 Description: After recovery, certain operations trigger a failed
5294 assertion on a client.
5295 Details : Failing over an mds, using lconf -d --failover, while a
5296 client was doing a readdir() call would cause the client to
5297 LBUG after recovery completed and the readdir() was resent.
5299 Severity : enhancement
5301 Description: Default groups are now added by lconf
5302 Details : You can now run lconf --group <servicename> without having to
5303 manually add groups with lmc.
5306 Frequency : occasional
5308 Description: Nodes with an elan id of 0 trigger a failed assertion
5311 Frequency : always when accessing e.g. tty/console device nodes
5313 Description: tty and some other devices nodes cannot be used on lustre
5314 Details : file's private_data field is used by device data and lustre
5315 values in there got lost. New field was added to struct file to
5316 store fs-specific private data.
5319 Frequency : when exporting Lustre via NFS
5321 Description: NFSD failed occasionally when looking up a path component
5322 Details : NFSD is looking up ".." which was broken in ext3 directories
5323 that had grown large enough to become hashed.
5326 Frequency : Clusters with multiple interfaces not on the same subnet
5328 Description: Nodes will repeatedly try to reconnect to an interface which it
5329 cannot reach and report an error to the log.
5330 Details : Extra peer list entries will be created by lconf with some peers
5331 unreachable. lconf now validates the peer before adding it.
5334 Frequency : Only if a default stripe is set on the filesystem root.
5336 Description: Setting a default stripe on the filesystem root prevented the
5337 filesystem from being remounted.
5338 Details : The client was sending extra request flags in the root getattr
5339 request and did not allocate a reply buffer for the dir EA.
5342 Frequency : occasional, higher if lots of files are accessed by one client
5343 Bugzilla : 6159, 6097
5344 Description: Client trips assertion regarding lsm mismatch/magic
5345 Details : While revalidating inodes the VFS looks up inodes with ifind()
5346 and in rare cases can find an inode that is being freed.
5347 The ll_test_inode() code will free the lsm during ifind()
5348 when it finds an existing inode and then the VFS later attaches
5349 this free lsm to a new inode.
5353 Bugzilla : 6422, 7030
5354 Description: MDS deadlock between mkdir and client eviction
5355 Details : Creating a new file via mkdir or mknod (starting a transaction
5356 and getting the ns lock) can deadlock with client eviction
5357 (gets ns lock and trying to finish a synchronous transaction).
5360 Frequency : occasional
5361 Description: While starting a server, the fsfilt_ext3 module could not be
5363 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
5364 kernels. Previously, lconf would still use the ext3 name
5365 when trying to load modules. Now, it will correctly use
5366 ext3 on 2.4 and ldiskfs on 2.6.
5368 Severity : enhancement
5369 Description: The default stripe count has been changed to 1
5370 Details : The interpretation of the default stripe count (0, to lfs
5371 or lmc) has been changed to mean striping across a single
5372 OST, rather than all available. For general usage we have
5373 found a stripe count of 1 or 2 works best.
5375 Severity : enhancement
5376 Description: Add support for compiling against Cray portals.
5377 Details : Conditional compiling for some areas that are different
5381 Frequency : occasional
5382 Bugzilla : 6409, 6834
5383 Description: Creating files with an explicit stripe count may lead to
5384 a failed assertion on the MDS
5385 Details : If some OSTs are full or unavailable, creating files may
5386 trigger a failed assertion on the MDS. Now, Lustre will
5387 try to use other servers or return an error to the
5391 Frequency : occasional
5393 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
5394 caused lock timeout during readahead (since 1.4.2).
5395 Details : Processes doing readahead might match a lock that hasn't been
5396 granted yet if there are overlapping and conflicting lock
5397 requests. The readahead process waits on ungranted lock
5398 (original lock is CBPENDING), while OST waits for that process
5399 to cancel CBPENDING read lock and eventually evicts client.
5401 Severity : enhancement
5403 Description: Initial enabling of flock support for clients
5404 Details : Implements fcntl advisory locking and file status functions.
5405 This feature is provided as an optional mount flag (default
5406 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
5407 locking are implemented yet, and those that are are not guaranteed
5408 to be completely correct in production environments.
5409 mount -t lustre -o [flock|noflock] ...
5412 Frequency : occasional
5414 Description: OSTs running 2.4 kernels but with extents enabled might trip an
5415 assertion in the ext3 JBD (journaling) layer.
5416 Details : The b_committed_data struct is protected by the big kernel lock
5417 in 2.4 kernels, serializing journal_commit_transaction() and
5418 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
5419 access to this struct. In 2.6 kernels there is finer grained
5420 locking to improve SMP performance of the JBD layer.
5424 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
5429 Description: merge and cleanup kernel patches.
5432 -----------------------------------------------------------------------------
5434 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
5436 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
5437 versions of Lustre. Please update all servers and clients to
5438 version 1.4.2 or later at the same time. You must also run
5439 "lconf --write-conf {config}.xml" on the MDS while it is stopped
5440 to update the configuration logs.
5442 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
5443 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
5444 - swab llog records of type '0' so we get proper header size/idx (5861)
5445 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
5446 - fix rename of one directory over another leaking an inode (5953)
5447 - avoid SetPageDirty on 2.6 (5981)
5448 - don't re-add just-being-destroyed locks to the waiting list (5653)
5449 - when creating new directories, inherit the parent's custom
5450 striping settings if present parent (3048)
5451 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
5452 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
5453 - don't hold client locks on temporary worklist from l_lru (5666)
5454 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
5455 - automatically evict dead clients (5921)
5456 - Update file size properly in create+truncate+fstat case (6196)
5457 - Do not unhash mountpoint dentries, do not allow removal of
5459 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
5460 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
5461 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
5462 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
5463 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
5464 - Correctly handle reads of files with no objects (6243)
5465 - lctl recover will also mark a device active if deactivate used (5933)
5467 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
5468 - allow --write-conf on an MDS with different nettype than client (5619)
5469 - don't write config llogs to MDS for mounts not from that MDS (5617)
5470 - lconf should create multiple TCP connections from a client (5201)
5471 - init scripts are now turned off by default; run chkconfig --on
5472 lustre and chkconfig --on lustrefs to use them
5473 - upcalls are no longer needed for clients to recover to failover
5475 - add --abort-recovery option to lconf to abort recovery on device
5477 - add support for an arbitrary number of OSTs (3026)
5478 - Quota support protocol changes.
5479 - forward compatibility changes to wire structs (6007)
5480 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5481 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5482 - improved "lustre status" script
5483 - initialize blocksize for non-regular files (6062)
5484 - added --disable-server and --disable-client configure options (5782)
5485 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5486 - Vanilla 2.4.29 support
5487 - increase maximum number of obd devices to 520 (6242)
5488 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5489 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5490 SLES 9 2.6.5 kernels
5491 - limit stripes per file to 160 (the maximum EA size) (6093)
5493 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5496 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5497 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5498 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5499 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5500 - small ext3 extents cleanups and fixes (5733)
5501 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5502 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5503 - avoid needless client->OST connect, fix handle mismatch (5317)
5504 - fix DLM error path that led to out-of-sync client, long delays (5779)
5505 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5506 - fix several locking issues related to i_size (5492,5624,5654,5672)
5507 - don't move pending lock onto export if it is already evicted (5683)
5508 - fix kernel oops when creating .foo in unlinked directory (5548)
5509 - fix deadlock in obdfilter statistics vs. object create (5811)
5510 - use time_{before,after} to avoid timer jiffies wrap (5882)
5511 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5512 - Do not leak request if server was not able to process it (5154)
5513 - If mds_open unable to find parent dir, make that negative lookup(5154)
5514 - don't create new directories with extent-mapping (5909, 5936)
5516 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5517 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5518 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5519 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5520 - support for RHEL4 kernel on i686 (5773)
5521 - provide error messages when incompatible logs are encountered (5898)
5523 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5524 * version 1.4.0.10 (1.4.1 release candidate 1)
5526 - don't keep a lock reference when lock is not granted (4238)
5527 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5528 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5529 - drop import inflight refcount on signal_completed_replay error (5255)
5530 - unlock page after async write error during send (3677)
5531 - handle missing objects in filter_preprw_read properly (5265)
5532 - no transno return for symlink open, don't save no-trasno open (3440)
5533 - don't try to complete elan receive that already failed (4012)
5534 - free RPC server reply state on error (5406)
5535 - clean up thread from ptlrpc_start_thread() on error (5160)
5536 - readahead could read extra page into cache that wasn't ejected (5388)
5537 - prevent races in class_attach/setup/cleanup/detach (5260)
5538 - don't dereference de->d_inode after l_dput of de (5458)
5539 - use "int" for stripe value returned from lock_to_stripe (5544)
5540 - mballoc allocation and error-checking fixes in 2.6 (5504)
5541 - block device patches to fix I/O request sizes in 2.6 (5482)
5542 - look up hostnames for IB nals (5602)
5543 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5544 - don't start multiple acceptors for the same port (5277)
5545 - fix incorrect LASSERT in mds_getattr_name (5635)
5546 - export a proc file for general "ping" checking (5628)
5547 - fix "lfs check" to not block when the MDS is down (5628)
5549 - service request history (4965)
5550 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5551 - create an "evict_client" /proc entry on OSTs, like the MDS has
5552 - fix mount usage message, return errors per mount(8) (5168)
5553 - change grep [] to grep "[]" in tests so they work in more UMLs
5554 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5555 - remove ancient LOV_MAGIC_V0 EA support (5047)
5556 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5557 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5558 - allow readahead allocations to fail when low on memory (5383)
5559 - mmap locking landed again, after considerable improvement (2828)
5560 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
5562 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
5565 - send OST transaction number in read/write reply to free req (4966)
5566 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5567 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5568 - the watchdog thread now runs as interruptible (5246)
5569 - flock/lockf fixes (but it's still disabled, pending 5135)
5570 - don't use EXT3 constants in llite code (5094)
5571 - memory shortage at startup could cause assertion (5176)
5573 - reorganization of lov code
5574 - single portals codebase
5576 - add extents/mballoc support (5025)
5577 - direct I/O reads in the obdfilter (4048)
5578 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
5580 tbd Cluster File Systems, Inc. <info@clusterfs.com>
5583 - send OST transaction number in read/write reply to free req (4966)
5584 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5585 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5586 - the watchdog thread now runs as interruptible (5246)
5587 - handle missing objects in filter_preprw_read properly (5265)
5588 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5589 - drop import inflight refcount on signal_completed_replay error (5255)
5590 - unlock page after async write error during send (3677)
5591 - return original error code on reconstructed replies (3761)
5592 - no transno return for symlink open, don't save no-trasno open (3440)
5594 - add pid to ldlm debugging output (4922)
5595 - bump the watchdog timeouts -- we can't handle 30sec yet
5596 - extra debugging for orphan dentry/inode bug (5259)
5598 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
5601 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
5602 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
5603 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
5604 - fix NULL dereference in /proc/sys/portals/routes (4827)
5605 - allow failed mdc_close() operations to be interrupted (4561)
5606 - stop precreate on OST before MDS would time out on it (4778)
5607 - don't send partial-page writes before EOF from client (4410)
5608 - discard client grant for sub-page writes on large-page clients (4520)
5609 - don't free dentries not owned by NFS code, check generation (4806)
5610 - fix lsm leak if mds_create_objects() fails (4801)
5611 - limit debug_daemon file size, always print CERROR messages (4789)
5612 - use transno after validating reply (3892)
5613 - process timed out requests if import state changes (3754)
5614 - update mtime on OST during writes, return in glimpse (4829)
5615 - add mkfsoptions to LDAP (4679)
5616 - use ->max_readahead method instead of zapping global ra (5039)
5617 - don't interrupt __l_wait_event() during strace
5619 - add software watchdogs to catch hung threads quickly (4941)
5620 - make lustrefs init script start after nfs is mounted
5621 - fix CWARN/ERROR duplication (4930)
5622 - return async write errors to application if possible (2248)
5623 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
5624 - print NAL number in %x format (4645)
5625 - update barely-supported suse-2.4.21-171 series (4842)
5626 - support for sles 9 %post scripts
5627 - support for building 2.6 kernel-source packages
5628 - support for sles km_* packages
5630 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
5633 - ignore -ENOENT errors in osc_destroy (3639)
5634 - notify osc create thread that OSC is being cleaned up (4600)
5635 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
5636 - reconstruct ost_handle() like mds_handle() (4657)
5637 - create a new thread to do import eviction to avoid deadlock (3969)
5638 - let lconf resolve symlinked-to devices (4629)
5639 - don't unlink "objects" from directory with default EA (4554)
5640 - hold socknal file ref over connect in case target is down (4394)
5641 - allow more than 32000 subdirectories in a single directory (3244)
5642 - fix blocks count for O_DIRECT writes (3751)
5643 - OST returns ENOSPC from object create when no space left (4539)
5644 - don't send truncate RPC if file size isn't changing (4410)
5645 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
5646 - bind to privileged port in socknal and tcpnal (3689)
5648 - rate limit CERROR/CWARN console message to avoid overload (4519)
5649 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
5650 - basic mmap support (3918)
5651 - kernel patch series update from b1_4 (4711)
5653 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
5656 - avoid crash during MDS cleanup with OST shut down (2775)
5657 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
5658 - don't use bad inodes on the MDS (3744)
5659 - dynamic object preallocation to improve recovery speed (4236)
5660 - don't hold spinlock over lock dumping or change debug flags (4401)
5661 - don't zero obd_dev when it is force cleaned (3651)
5662 - print grants to console if they go negative (4431)
5663 - "lctl deactivate" will stop automatic recovery attempts (3406)
5664 - look for existing locks in ldlm_handle_enqueue() (3764)
5665 - don't resolve lock handle twice in recovery avoiding race (4401)
5666 - revalidate should check working dir is a directory (4134)
5668 - don't always mark "slow" obdfilter messages as errors (4418)
5670 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
5673 - don't close LustreDB during write_conf until it is done (3860)
5674 - fix typo in lconf for_each_profile (3821)
5675 - allow dumping logs from multiple threads at one time (3820)
5676 - don't allow multiple threads in OSC recovery (3812)
5677 - fix debug_size parameters (3864)
5678 - fix mds_postrecov to initialize import for llog ctxt (3121)
5679 - replace config semaphore with spinlock (3306)
5680 - be sure to send a reply for a CANCEL rpc with bad export (3863)
5681 - don't allow enqueue to complete on a destroyed export (3822)
5682 - down write_lock before checking llog header bitmap (3825)
5683 - recover from lock replay timeout (3764)
5684 - up llog sem before sending rpc (3652)
5685 - reduce ns lock hold times when setting kms (3267)
5686 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
5687 - fix NULL deref and obd_dev leak on setup error (3312)
5688 - replace some LBUG about llog ops with error handling (3841)
5689 - don't match INVALID dentries from d_lookup and spin (3784)
5690 - hold dcache_lock while marking dentries INVALID and hashing (4255)
5691 - fix invalid assertion in ptlrpc_set_wait (3880)
5693 - add libwrap support for the TCP acceptor (3996)
5694 - add /proc/sys/portals/routes for non-root route listing (3994)
5695 - allow setting MDS UUID in .xml (2580)
5696 - print the stack of a process that LBUGs (4228)
5698 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
5701 - don't cleanup request in ll_file_open() on failed MDS open (3430)
5702 - make sure to unset replay flag from failed open requests (3440)
5703 - if default stripe count is 0, use OST count for inode size (3636)
5704 - update parent mtime/ctime on client for create/unlink (2611)
5705 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
5706 - free recovery state on server during a forced cleanup (3571)
5707 - unregister_reply for resent reqs (3063)
5708 - loop back devices mounting and status check on 2.6 (3563)
5709 - fix resource-creation race that can provoke i_size == 0 (3513)
5710 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
5711 - more debugging for page-accounting assertion (3746)
5712 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
5713 - avoid deadlock after precreation failure (3758)
5714 - fix race and lock order deadlock in orphan handling (3450, 3750)
5715 - add validity checks when grabbing inodes from l_ast_data (3599)
5717 - add /proc/.../recovery_status to obdfilter (3428)
5718 - lightweight CDEBUG infrastructure, debug daemon (3668)
5719 - change default OSC RPC parameters to be better on small clusters
5720 - turn off OST read cache for files smaller than 32MB
5721 - install man pages and include them in rpms (3100)
5722 - add new init script for (un)mounting lustre filesystems (2593)
5723 - run chkconfig in %post for init scripts (3701)
5724 - drop scimac NAL (unmaintained)
5726 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
5729 - clean kiobufs before and after use (3485)
5730 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
5731 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
5732 - change init script to fail more clearly if not run as root (1528)
5733 - allow clients to reconnect during replay (1742)
5734 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
5735 - don't do DNS lookups on NIDs too small for IP addresses (3442)
5736 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
5737 - fix cond_resched (3554)
5738 - only evict unfinished clients after recovery (3515)
5739 - allow bulk resend, prevent data loss (3570)
5740 - dynamic ptlrpc request buffer allocation (2102)
5741 - don't allow unlinking open directory if it isn't empty (2904)
5742 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
5743 - remove extraneous obd dereference causing LASSERT failure (3334)
5744 - don't use get_cycles() when creating temp. files on the mds (3156)
5745 - hold i_sem when setting i_size in ll_extent_lock() (3564)
5746 - handle EEXIST for set-stripe, set proper directory name (3336)
5748 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
5749 - fix ksocknal_fmb_callback() error messages (2918)
5751 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
5754 - don't copy lvb into (possibly NULL) reply on error (2983)
5755 - don't deref dentry after dput, don't free lvb on error (2922)
5756 - use the kms to determine writeback rpc length (2947)
5757 - increment oti_logcookies when osc is inactive (2948)
5758 - update client's i_blocks count via lvb messages (2543)
5759 - handle intent open/close of special files properly (1557)
5760 - mount MDS with errors=remount-ro, like obdfilter (2009)
5761 - initialize lock handle to avoid ASSERT on error cleanup (3057)
5762 - don't use cancelling-locks' kms values (2947)
5763 - use highest lock extent for kms, not last one (2925)
5764 - don't dereference ERR_PTR() dentry in error handling path (3107)
5765 - fix thread race in portals_debug_dumplog() (3122)
5766 - create lprocfs device entries at setup instead of at attach (1519)
5767 - common AST error handler, don't evict client on completion race (3145)
5768 - zero nameidata in detach_mnt in 2.6 (3118)
5769 - verify d_inode after revalidate_special is valid in 2.6 (3116)
5770 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
5771 - initialize RPC timeout timer earlier for 2.6 (3219)
5772 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
5773 - print nal/nid for unknown nid (3258)
5774 - additional checks for oscc recovery before doing precreate (3284)
5775 - fix ll_extent_lock() error return code for 64-bit systems (3043)
5776 - don't crash in mdc_close for bad permissions on open (3285)
5777 - zero i_rdev for non-device files (3147)
5778 - clear page->private before handing to FS, better assertion (3119)
5779 - tune the read pipeline (3236)
5780 - fix incorrect decref of invalidated dentry (2350)
5781 - provide read-ahead stats and refine rpc in flight stats (3328)
5782 - don't hold journal transaction open across create RPC (3313)
5783 - update atime on MDS at close time (3265)
5784 - close LDAP connection when recovering to avoid server load (3315)
5785 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
5786 - don't leak open file on MDS after open resend (3325)
5787 - serialize filter_precreate and filter_destroy_precreated (3329)
5788 - loop device shouldn't call sync_dev() for nul device (3092)
5789 - clear page cache after eviction (2766)
5790 - resynchronize MDS->OST in background (2824)
5791 - refuse to mount the same filesystem twice on same mountpoint (3394)
5792 - allow llmount to create routes for mounting behind routers (3320)
5793 - push lock cancellation to blocking thread for glimpse ASTs (3409)
5794 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
5795 - fix rare problem with rename on htree directories (3417)
5797 - allow default OST striping configuration per directory (1414)
5798 - fix compilation for qswnal for 2.6 kernels (3125)
5799 - increase maximum number of MDS request buffers for large systems
5800 - change liblustreapi to be useful for external progs like lfsck (3098)
5801 - increase local configuration timeout for slow disks (3353)
5802 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
5804 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5807 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
5808 - don't overwrite extent policy data in reply if lock was blocked (2901)
5809 - drop filter export grants atomically with removal from device (2663)
5810 - del obd_self_export from work_list in class_disconnect_exports (2908)
5811 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
5812 - swab reply message in mdc_close, other PPC fixes (2464)
5813 - fix destroying of named logs (2325)
5814 - overwrite old logs when running lconf --write_conf (2264)
5815 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
5816 - fix race in target_handle_connect (2898)
5817 - mds_reint_create() should take same inode create lock (2926)
5818 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
5819 - don't close files for self_export to avoid uninitialized obd (2936)
5820 - allow MDS with the same name as client node (2939)
5821 - hold dentry reference for closed log files for unlink (2325)
5822 - reserve space for all logs during transactions (2059)
5823 - don't evict page beyond end of stripe extent (2925)
5824 - don't oops on a deleted current working directory (2399)
5825 - handle hard links to targets without a parent properly (2517)
5826 - don't dereference NULL lock when racing during eviction (2867)
5827 - don't grow lock extents when lots of conflicting locks (2919)
5829 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
5832 - account for cache space usage on clients to avoid data loss (974)
5833 - lfsck support in lustre kernel code (2349)
5834 - reduce journal credits needed for BRW writes (2370)
5835 - orphan handling to avoid losing space on client/server crashes
5836 - ptlrpcd can be blocked, stopping ALL progress (2477)
5837 - use lock value blocks to assist in proper KMS, faster stat (1021)
5838 - takes i_sem instead of DLM locks internally on obdfilter (2720)
5839 - recovery for initial connections (2355)
5840 - fixes for mds_cleanup_orphans (1934)
5841 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
5842 - block all file creations until orphan recovery completes (1901)
5843 - client remove rq_connection from request struct (2423)
5844 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
5845 - recovery timer race (2670)
5846 - mdc_close recovey bug (2532)
5847 - ptlrpc cleanup bug (2710)
5848 - mds timeout on local locks (2588)
5849 - namespace lock held during RPCs (2431)
5850 - handle interrupted sync write properly (2503)
5851 - don't try to handle a message that hasn't been replied to (2699)
5852 - client assert failure during cleanup after abort recovery (2701)
5853 - leak mdc device after failed mount (2712)
5854 - ptlrpc_check_set allows timedout requests to complete (2714)
5855 - wait for inflight reqs when ptlrpcd finishes (2710)
5856 - make sure unregistered services are removed from the srv_list
5857 - reset bulk XID's when resending them (caught by 1138 test)
5858 - unregister_bulk after timeout
5859 - fix lconf error (2694)
5860 - handle write after unfinished setstripe, stripe-only getstripe (2388)
5861 - readahead locks pages, leaves pending causing memory pressure (2673)
5862 - increase OST request buffers to 4096 on large machines (2729)
5863 - fix up permission of existing directories in simple_mkdir (2661)
5864 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
5865 - don't assign transno to errored transactions (2742)
5866 - don't delete objects on OST if given a bogus objid from MDS (2751)
5867 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
5868 - if rq_no_resend, then timeout request after recovery (2432)
5869 - fix MDS llog_logid record size, 64-bit array alignment (2733)
5870 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
5871 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
5872 - MDS recovery shouldn't delete objects that it has given out (2730)
5873 - if enqueue arrives after completion, don't clobber LVB (2819)
5874 - don't unlock pages twice when trigger_group_io returns error (2814)
5875 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
5876 - don't write pages to disk if there was an error (1450)
5877 - don't ping imports that have recovery disabled (2676)
5878 - take buffered bytes into account when balancing socknal conn (2817)
5879 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
5880 - reconnect unlink llog connection after MDS reconnects to OST (2816)
5881 - remove little-endian swabbing of llog records (1987)
5882 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
5883 - retry reposting request buffers if they fail (1191)
5884 - grow extent at grant time to avoid granting a revoked lock (2809)
5885 - lock revoke doesn't evict page if covered by a second lock (2765)
5886 - disable VM readahead to avoid reading outside lock extents (2805)
5888 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
5889 - updated kernel patches for hp-2.4.20 kernel (2681)
5891 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
5894 - fix truncated write corruption (2366)
5895 - fix for failed assertion in iopen_connect_dentry (1792,2517)
5897 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
5898 - ensure len(uuid) < 37 in lmc (1171)
5899 - fix ia64 OOPS in llog_test (2255)
5900 - zero end of page at obdfilter for partial page writes (2648)
5901 - don't leave stale dentries around after renames (bug 2428)
5902 - fix timeouts when evicting a client with a single lock held (2642)
5903 - set deadline for the initial HELLO message to drain (2634)
5904 - print out dotted-quad IP addresses in the socknal (2302)
5906 - additional debugging for MDS client eviction problem (2443)
5907 - fix mkfsoptions support for osts (2603, 2604)
5909 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
5912 - add series for the vanilla 2.6.0 kernel
5913 - add series for the vanilla 2.4.24 kernel
5914 - add series for a cray x86/64 UL kernel drop
5915 - fix xattr patches for the vanilla 2.4.19 series
5917 - generate true UUIDs in lmc (1171)
5918 - have portals stack dumping break in UML (2466)
5919 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
5920 - allocate with _NOFS in ldlm to avoid deadlock (1933)
5921 - wake callback waiting threads on client eviction (2460)
5922 - Add --ptldebug and --subsystem to lmc (1719)
5923 - update assertion to allow safe interrupt allocation
5924 - set rq_no_resend for cancel requests (2432)
5925 - recalculate ptlrpcd timeout after resend (2494)
5926 - call vfs_rmdir when removing pending directories (2368)
5927 - fix renaming a file to itself (2429)
5928 - lmc creates a default one-stripe lov (2454)
5929 - expand procfs space to handle large clusters (2326)
5930 - increase UML stack to avoid overflow
5931 - update lconf's list of debug and subsystem masks
5932 - fix lfs find --obd (2510)
5933 - /proc tunable for disabling filter read caching (2591)
5934 - stop rpm packages from altering slapd.conf (2301)
5935 - disable nagle in the socknal under 0conf (2578)
5936 - choose mds inode size based on stripe count (2572)
5937 - fix kernel-source rpm problems (2516)
5939 - add --disable-doc to avoid pdf generation (2421)
5940 - update documentation, tests, type-os, comments
5941 - avoid format warnings on ia64
5942 - remove the TOE NAL
5943 - tiny code cleanups by removing unused fields
5945 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
5948 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
5949 - remove the most blatant lies from BUILDING file (2371)
5950 - change default debug level to reasonable production setting
5951 - reduce client side cache size to reduce cache flush time
5952 - reduce max RPCs in flight to avoid unnecessary file fragmentation
5953 - make TCP zerocopy and pinger support enabled by default (2476)
5954 - sync writes completed after process exits caused crashes (2319)
5955 - maintain correct mount count on the MDS (2356)
5956 - backout 1557, because 2316 wasn't really fixed
5957 - better file I/O statistics gathering in /proc
5958 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
5959 - another kernel patch to fix zero-copy TCP function export
5960 - don't take duplicate lock when processing re-sent getattr (2420)
5961 - lctl uses obd_self_export instead of creating new conn (2353)
5962 - MDS/OST recovery case which requires object creation asserted (2425)
5963 - move lfs from /usr/sbin to /usr/bin in packages
5964 - fix race between mds_client_add and mds_client_free (2417)
5965 - use kmalloc instead of slabs in portals (2430)
5966 - don't create duplicate records when a failover MDS is present (2442)
5967 - remove unnecessary mount age check (2332)
5968 - don't remove directory inodes from locks prematurely (2451)
5969 - don't break if MDS service name is the same as hostname (2103)
5970 - fix races in client write RPC generation when cache full (2482)
5972 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
5975 - remove now-unused request->rq_obd (278)
5976 - if an allocation fails, print out how much memory we've used (1933)
5977 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
5978 - add the "configurable stack size" patch to most series files (1256)
5979 - ability to write large log records, for 100+ OST configs (2306)
5980 - fix NULL deref when filter_prep fails (2314)
5981 - fix operator precedence error in filter_sync
5982 - dynamic allocation of socknal TX descriptors (2315)
5983 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
5984 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
5985 - ensure that CWARN messages in Portals always get to the syslog
5986 - __init/__exit are not for prototype decls (ldlm_init/exit)
5987 - x86-64 compile warning fixes
5988 - fix gateway LMC keyword conflict (2318)
5989 - fix MDS lock inversions in getattr/reint paths (1844)
5990 - fix a rare lock re-ordering bug, which caused deadlock (2322)
5991 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
5992 - DLM race condition prevented some lock evictions (2328)
5993 - ENOMEM detection and retry on socknal sends (2230)
5994 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
5995 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
5996 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
5997 - better error messages when a client is rejected during recovery (1505)
5998 - avoid cancelling locks which were never granted, after failure (2330)
5999 - fix i_sem/journal inversion in mds_client_add (2333)
6000 - fix truncate/getattr lock cycle deadlock (2334)
6001 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
6002 - fix two rare exit paths which could leak an l_lock() ref (2321)
6003 - fencepost error in MDS/OST orphan recovery (2226)
6004 - make log record alignment 8 bytes (1988)
6005 - lstripe now fails when requested offset > ost_count (2237)
6006 - ensure that all kernel series have a complete list.h (1607)
6007 - fix crashes in special-file operations (2316)
6008 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
6010 - allow configurable automake binary, for testing new versions
6011 - small update to the lfs documentation
6013 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
6015 * fix negative export reference count in fsfilt_sync (2312)
6017 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
6018 * release candidate 0.9.1
6020 - orphans are moved into the PENDING directory for possible recovery
6021 - replayed opens now open by fid for orphan/rename safety (1042)
6022 - last close of an orphan inode generates a transno (683)
6023 - chdir() and mount() now pin the directory entry (1020)
6024 - avoid CERROR in normal ll_setattr_raw() error case (1500)
6025 - discard very old requests without processing them (1502)
6026 - remove some common, well-understood CERRORs (1505)
6027 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
6028 - clear "grant" flags in OST replies until OST grant code lands (1644)
6029 - fix read performance by not clobbering i_blksize on client (1598)
6030 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
6031 - make LRU size a /proc tunable, clears locks when reduced (707)
6032 - fix some lprocfs rot that prevented ptlbd from loading (1732)
6033 - server locks take references on exports now (1558)
6034 - build fixes for 2.4.20-rh trees (1663)
6035 - return an error from lov_create if all OSCs are inactive (1751)
6036 - fix import levels when a reconnect happens without a timeout (1597)
6037 - exit early from mds_open if we get a lookup error (1749)
6038 - partial page read at EOF wouldn't wait for disk before sending (1642)
6039 - avoid NULL deref in obdfilter when reading page past EOF (1592)
6040 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
6041 - fix LBUG in ll_it_open_error with rc = -2 (1861)
6042 - write/truncate lock inversion (1639)
6043 - Don't auto-load obdclass, portals modules during cleanup (1495)
6044 - fix timestamps from jumping to "now" (1763)
6045 - extra journal assertions (1648)
6046 - add an extra multiunlink test (1771)
6047 - fix read_record/write_record API (1776)
6048 - fix leak of offset_extent, possible incorrect i_size later (1772)
6049 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
6050 - Debugging for the kqswnal_get_idle_tx problems (1820)
6051 - Allow recovery to be attempted multiple times (1536)
6052 - Write out MDS last_rcvd file after it is first created (1600)
6053 - Fix tx_descriptor leak in failed transmit situations (1827)
6054 - ext3 journaling fixes for assertion failure after IO error (1871)
6055 - class_export_put() on freed export after completion AST error (1896)
6056 - Fix revalidate looping in VFS (1322)
6057 - Don't access a freed export during MDS_REINT timeout (1521)
6058 - Add open-unlink recovery support on the MDS (1673,1764)
6059 - Return an error if no MDS data was read from last_rcvd (1946)
6060 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
6061 - Don't setup a disk device that doesn't match exported UUID (317)
6062 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
6063 - avoid committing NULL handle in force close
6064 - local.sh is now a one-stripe LOV configuration
6065 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
6066 - fix ext3 htree duplicate directory entry corruption (1516)
6067 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
6068 - update to new LOV EA format (2097)
6069 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
6071 2003-06-15 Phil Schwan <phil@clusterfs.com>
6074 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
6075 - per-import/export recovery handling (958, 931, 959)
6076 - multiple last-rcvd slots, for serving multiple FSes (949)
6077 - connections are again shared between multiple imp/exports (963, 964)
6078 - "umount -f" would hang if any requests needed to be sent (393, 978)
6079 - avoid pinning large req buffer by copying for queued messages (989)
6080 - add "uuid" to "lctl device" command to help upcalls (991)
6081 - "open" RPCs with transnos would confuse recovery counters (1037)
6082 - do proper endian conversion of all wire messages (288, 340, 891)
6083 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
6084 - call waiting locks callback from LDLM recovery thread (1127, 1151)
6085 - fix ptlrpc_connection leak in target_handle_connect (1174)
6086 - fix import refcounting bug in OST and MDS cleanup (1134)
6087 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
6088 - fix very unlikely obd_types race condition (501)
6089 - remove osc_open hack for echo_client (1187)
6090 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
6091 - a failure in read_inode2 leads to deadlock (1139)
6092 - cancel ack-locks as soon as transaction is committed (1072)
6093 - fix major leaks and crashes in the bulk I/O path (937, 1057)
6094 - make sure to commitrw after any preprw to avoid deadlock (1162)
6095 - failing to execute a file in a lustre FS would lock inode (1203)
6096 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
6097 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
6098 - obd_uuid2tgt would walk off the end of the list (1255)
6099 - on IA64 the osc would give portals incorrect bulk size (1258)
6100 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
6101 - fix lock inversion caused by new llite matching code (1282)
6102 - limit the number of dirty pages on a client to 10MB (1286)
6103 - timed out locks were not being corrected cancelled (1289)
6104 - fix O_DIRECT above 4GB on IA-32 (1292)
6105 * major user-visible changes
6106 - fail out/fail over policy now controlled by the upcall (993)
6108 - add OBD_PING to check server availability and failure (954)
6109 - lustre messages are now sent in sending host order (288, 340, 891)
6110 - add eadatalen to MDS getattr reply (340)
6111 - OST read replies may contain second buffer, with per-page status (593)
6113 2003-03-11 Phil Schwan <phil@clusterfs.com>
6116 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
6117 - failed open()s could cause deadlock; fixed (867, 869)
6118 - stop cancelling OST locks when files are closed (481)
6119 - overlapping XID spaces caused network corruption (851, 853)
6120 - fix unsafe fsfilt counter arithmetic; change to atomic_t
6121 - setattr_raw added, to do single-RPC, server-side setattrs
6122 - lmc/lconf syntax change for OST UUIDs
6123 - fix crashy race condition between ptlrpc_free_req and osc_close
6124 - don't use request in mdc_enqueue if we hit a timeout (889)
6125 - don't set the inode i_size for regular files from the MDS (896)
6126 - handle out of order completion AST (842)
6127 - don't LBUG if a lock request times out after receiving AST (913)
6128 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
6129 - if a bad lock AST arrives, send an error instead of dropping entirely
6130 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
6131 - fix leak in bulk IO when only partially completed (899, 900, 926)
6132 - fix O_DIRECT for ia64 (55)
6133 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
6134 - C-z after timeout could hang a process forever; fixed (977)
6136 - client-side I/O cache (678, 924, 929, 941, 970)
6138 - READPAGE and SETATTRs which don't take server-side locks get
6141 2003-02-11 Phil Schwan <phil@clusterfs.com>
6144 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
6145 - Fix fsfilt_extN_readpage() to read a full page of directory
6146 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
6147 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
6148 - Fix imbalanced LOV object allocation and out-of-bound access (469)
6149 - Most intent operations were removed, in favour of a new RPC mode
6150 that does a single RPC to the server and bypasses most of the VFS
6151 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
6152 - Aggressively cancel local locks on DLM servers
6153 - mds_reint_unlink sends EA to the client if it's the last nlink.
6154 client uses that EA to unlink OST objects.
6155 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
6156 - recursive symlinks were fixed (439)
6157 - fixed NULL deref in DEBUG_REQ
6158 - filter_update_lastobjid no longer calls sync, which annoyed extN
6159 - fixed multi-client small-writes to a single file problem (445)
6160 - fixed mtime updates during file writes (607)
6161 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
6162 - fixed bug in obd_brw_read/write() (under guise of testing 367)
6163 - fixed Linux OST size reporting problem (444, 656)
6164 - OST now updates object mtime with writes or setattr (607, 619)
6165 - client verifies file size before zeroing page past EOF (445)
6166 - OST now writes last allocated objid to disk with allocation (108)
6167 - LOV on echo now works (409)
6169 - mds_reint_unlink sends a new buffer, with the EA included. this
6170 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
6171 set if a regular file was being unlinked, and it was the last link
6172 - use PtlGet from the target for bulk writes (315)
6173 - OST now updates object mtime with writes or setattr (607, 619)
6174 - LDLM now has a grant-time callback to revalidate locked items, if
6176 - Many MDS operations were reorganized to combat race conditions
6178 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
6179 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
6181 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
6184 - Fully reactivate OST imports after reconnection (512, others)
6185 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
6186 - More graceful error handling for truncating on dead OST (515)
6187 - Don't error out unless we're actually accessing dead stripes (474)
6188 - Fix garbage sizes when stripes are missing (410)
6189 - LRU counters were broken, causing constant lock purge (433, 432)
6190 - garbage on read from stripes with failed OSTs (441)
6191 - mark OSCs as active before reconnecting during recovery (438)
6192 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
6193 - lfind did not preserve OST order in output (443)
6194 - symlinks cause hung clients, incorrect data (439)
6195 - stop dereferencing request after dropping refcount (457)
6196 - don't LASSERT(spin_is_locked) on non-SMP (455)
6197 - fixes for many rename() bugs
6198 - fstat didn't correctly synchronize attributes (399)
6199 - server must handle lock cancellation during blocking AST prep (487)
6200 - bulk descriptors were free()d too soon (511)
6201 - fix paths in lconf, which would load incorrect modules (451, 507)
6202 - fix confusing lconf 'host not found' error message (386)
6203 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
6204 - fix race condition in mdc_blocking_ast() for inode access (526)
6205 - fix lov_unpackmd() unpacking wrong number of stripes (537)
6206 - fix lov_set_osc_active() marking wrong OSC inactive (440)
6207 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
6208 - fix multiple writes of stripe MD to MDS (358, maybe 519)
6209 - fix lstripe in several ways (kernel side) (527)
6210 - fix request leak in ldlm_cli_enqueue (262)
6211 - incorrect OSC was marked inactive after OST failure
6212 - call mds_fs_cleanup before unmounting filesystem (524)
6213 - fix races between taking ns_lock and ldlm_lock_change_resource
6214 - fix races updating LOV export open file list
6215 - fix lov_enqueue error path, avoid decref-ing bad lock handle
6216 - fix recovery NULL deref in ldlm_cli_cancel_unused
6217 - fix some DLM races by using new hash table for lock handles (419)
6218 - permit the client to specify desired inodes, at replay
6219 - duplicate requests when we queue them for replay reintegration
6220 - fix last_rcvd offset calculation
6221 - sync after each recovered transaction, so we always make progress
6222 - never, not always, ERESTART requests without transnos
6223 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
6225 - skip replay if the MDS says that the client is already connected
6226 - don't check for a recovery-enabled export to match lctl's UUID
6227 - don't INC_USE_COUNT for phantom exports
6228 - don't crash when cleaning up phantom exports (567)
6229 - don't double-finish or set replay data for errored mdc_open requests
6230 - abort requests when they time out, so we don't get old replies
6231 - send/receive replies for AST messages again
6232 - if the client says that it doesn't have the lock, cancel it on the
6234 - if we timeout during I/O, don't try to cancel an in-use lock; instead
6235 mark it as destroyed, it will all work out when decref is called
6236 - fix module use counts (22, 581)
6238 - ASTs now expect a reply (server cancels lock on error reply)
6240 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
6243 - fix many simultaneous client startup (392)
6244 - fix dentry->d_it clobbering
6245 - credentials weren't being shipped for readdir/getattr operations
6246 - remove invalid assertions triggered during some concurrent MD
6248 - proper Lustre versions added (336, 389)
6249 - fix memory leak for create error case (398)
6250 - fix LOV locking bug that would get cli/srv out of sync
6251 - fix echo client over LOV (409)
6252 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
6253 - fix double-O_EXCL intent crash (424)
6254 - avoid sending multiple lock CANCELs (352)
6256 - MDS can do multi-client recovery (modulo bugs in new code)
6258 - many updates, edits, cleanups
6260 2002-11-18 Phil Schwan <phil@clusterfs.com>
6263 - fix null d_it dereference (346)
6264 - fix full OST/dbench hang (333)
6265 - fix permission problem with file removal (286)
6266 - fix removal of OSCs from LOV when they fail
6267 - fix NULL deref during bulk timeout (214)
6268 - fix problems related to multiple filesystems on one MDS (241)
6269 - fixed serious subtle metadata locking bugs
6270 - free locks on clients when inodes are removed due to memory
6272 - fix inode pointer in lock data (285)
6273 - partial support for multiple MDS on a single host (241)
6274 - data locks weren't cancelled at clear_inode time (290, 311)
6275 - intent locks could lead to unbounded lock growth (205)
6276 - added a maximum lock count, an LRU list, and a flusher
6277 - fix multiple rename (365)
6278 - properly abstracted the echo client
6279 - OSC locked 1 byte too many; fixed
6280 - rewrote brw callback code:
6281 - fixed recovery bugs related to LOVs (306)
6282 - fixed too-many-pages-in-one-write crash (191)
6283 - fixed (again) crash in sync_io_timeout (214)
6284 - probably fixed callback-related race (385)
6286 - Add capability to MDS protocol
6287 - LDLM cancellations and callbacks on different portals
6289 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
6292 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
6293 - defer open object destruction to close time (601981, 138)
6294 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
6295 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
6296 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
6297 - don't call obd_setattr in ll_file_release for destroyed objects
6298 * protocol change to lustre_msg: move |version| and add |flags|
6299 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
6300 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
6301 * added replay of create, unlink, link and rename operations during
6302 MDS failover; recovery should be much more robust now
6303 * remove failed OSCs from LOVs (only lov_create uses this so far)
6304 * the lustre-HOWTO was brought (more) up to date (582544)
6306 2002-10-23 Phil Schwan <phil@clusterfs.com>
6309 - in-use dentries weren't being reused properly (617851)
6310 - prevent multiple LDLM setup (599178)
6311 - fix LOV size calculations for truncate (617853)
6312 - fix client handling of MDS intent errors (POSIX)
6313 - fix permission bug in lovstripe.c test (624321)
6314 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
6315 - truncate past end of file could corrupt data
6316 - proper cleanup after timeouts, crashes, etc (592524, 550815)
6317 - a race in recovery could return ETIMEDOUT to apps (623947)
6318 - building outside the source directory was fixed
6319 * the lustre-HOWTO was brought (more) up to date (582544)
6320 * major progress was made on recovery functionality
6322 2002-10-10 Phil Schwan <phil@clusterfs.com>
6325 - recovery deadlock fix
6326 - rm -rf causes LBUG fix (617817)
6327 - file open by multiple tasks fix (618962)
6328 - directory permissions bugs (602707 and 620007)
6329 - journal_stop fixed with locking (611313)
6330 - O_APPEND failures resolved (618273, perhaps 614459)
6331 - lconf PATH fix (619770)
6332 - IA64 build fix (621450)
6333 - RPC buffer sizes scale with amount of memory
6335 2002-10-01 Phil Schwan <phil@clusterfs.com>
6338 - locks would be cancelled without throwing away data pages,
6339 resulting in inconsistent data (605627)
6340 - inode attributes were not always being refreshed (605627, 612449)
6341 - lconf now continues to cleanup after lctl reports an error
6342 - MDS now enforces user permissions (602707)
6343 - lprocfs cleanup fixed, but not yet enabled (614157)
6344 - fixed infinite server hang, should a client not respond to an AST
6345 - avoid going into recovery if user calls readlink() with a buffer
6346 that's too small (613941)
6347 - AST RPCs no longer require replies (614867) -- this may be changed
6348 - don't crash server if client sends an IOV that's too big (611336)
6349 - fixed lock conversion deadlock (611892)
6350 - fixed the following of symlinks (614622)
6351 * recovery: the server can remove locks from a client that dies, other
6352 clients can make progress
6353 * more extN patch fixes
6354 * compile-time configurable ptlrpc buffer allocations
6356 - collaborative read cache document
6357 - Lustre Lite Performance CDR document-in-progress
6359 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6362 - fix typo in patch-2.4.18
6364 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6367 - clear ptlrpc request each time in handle_incoming_request()
6368 - unlink of files now destroys the object on the OST
6370 2002-09-19 Peter Braam <braam@clusterfs.com>
6372 * add hard link support
6373 * change obdfile creation method
6374 * kernel patch changed
6376 2002-09-19 Peter Braam <braam@clusterfs.com>
6379 - stack overflow bug in extN fixed
6381 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
6383 * documentation updates
6384 - add man pages for config tools
6385 - update tests/README to describe testing with new config tools
6386 - finish metadata API descriptions
6387 * bug fixes and cleanups
6388 - statfs workaround for 16TB limit
6389 - LOV stripe allocation improved, can stripe on subset of OSTs
6390 - LOV file size/IO offset was wrong for files > 4GB in size
6391 - object EA data was being dropped, caused files to be unreadable
6392 - memory overflow with non-LOV OST caused memory corruption
6393 - fixed regression tests to work with new config tools, obdfilter
6394 - fixed bug when directory size became larger than 1 block
6395 - fixed bug (for single client case) when PWD was deleted
6396 - invalidate local directory pages when doing intent-based ops
6397 - avoid LDLM oops when lock callback contained bad data
6399 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
6401 * documentation updates
6402 * bug fixes and cleanups
6403 - configuration tools
6406 - 64-bit compile warnings
6407 - 64-bit internal statfs data
6409 * test_brw on persistent OST devices
6411 * lprocfs (disabled)
6413 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
6415 * documentation updates
6416 * bug fixes and cleanups
6417 * configuration tools
6419 2002-08-30 Peter J. Braam <braam@clusterfs.com>
6422 * many small fixes to 0_5_4
6423 * io/network handling
6424 * thinkos in MDS operations
6426 2002-08-24 Peter J. Braam <braam@clusterfs.com>
6429 * crucial basic fixes to 0.5.3
6430 * IOR, Iozone work over Elan
6433 2002-08-07 Phil Schwan <phil@clusterfs.com>
6434 * version 0_5_3, our first alpha
6435 * we use the new Portals iovs
6436 * documentation updates
6437 * bug fixes and cleanups
6438 * small changes in the DLM wire protocol
6440 2002-07-25 Peter J. Braam <braam@clusterfs.com>
6441 * version 0_5_1 with some initial stability,
6442 * locking on MD and file I/O.
6443 * documentation updates
6444 * several bug fixes since 0.5.0
6445 * small changes in wire protocol
6447 2002-07-18 Phil Schwan <phil@clusterfs.com>
6449 * delivered as Lustre Light Alpha
6450 * fixed a crash after handling invalid MDS requests
6451 * fixed directory pages for architectures with non-4k pages sizes
6453 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
6454 * release version v0_4_4
6455 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
6457 * Fixes a number of interruption problems with OST operations.
6458 * Update documentation for portals header changes
6459 * Move all wire protocol structs/defines to lustre_idl.h
6460 * Fixes symlink length bug.
6461 * Add tcpdump to repository.
6463 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
6464 * release version v0_4_3
6465 * Fixes statfs for inodes on extN.
6466 * Fixes bug in runtests which would delete /etc/hosts.
6467 * Use 64-bit object IDs wherever possible (not into VFS though)
6468 Remove ost_get_info, which is unused by lustre, and out of date.
6470 2002-07-03 Peter Braam <braam@clusterfs.com>
6471 * release version v0_4_2 Fixes a lookup error (type not passed)
6472 * move forward to head of Portals
6473 * move forward to latest Lustre kernel
6475 2002-06-25 Peter Braam <braam@clusterfs.com>
6476 * release version v0_4_1. Hopefully stable on single node use.