1 06-23-2014 Intel Corporation
3 * See http://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
4 * Server support for kernels:
5 2.6.32-431.23.3.el6 (RHEL6)
6 3.0.101-0.29 (SLES11SP3)
7 * Client support for unpatched kernels:
8 2.6.32-431.23.3.el6 (RHEL6)
9 3.0.101-0.29 (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 03-17-2014 Intel Corporation
22 * See http://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
23 * Server support for kernels:
24 2.6.32-431.17.1.el6 (RHEL6)
25 3.0.101-0.29 (SLES11SP3)
26 * Client support for unpatched kernels:
27 2.6.32-431.17.1.el6 (RHEL6)
28 3.0.101-0.29 (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 10-31-2013 Intel Corporation
41 * See http://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
42 * Server builds for kernels:
43 2.6.32-358.18.1.el6 (RHEL6.4)
44 * Client builds for unpatched kernels:
45 2.6.32-279.19.1.el6 (RHEL6.3)
46 2.6.32-358.18.1.el6 (RHEL6.4)
47 3.0.93-0.5 (SLES11 SP2)
48 3.0.92-0.8 (SLES11 SP3)
50 * Recommended e2fsprogs version: 1.42.7.wc2 or newer
51 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
52 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
53 stack overflow. For more information, please refer to bugzilla 17630.
54 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
55 the re-exporting nodes to be mounted with "32bitapi" mount option
57 --------------------------------------------------------------------------------
59 05-30-2013 Intel Corporation
61 * See http://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
62 * Server support for kernels:
63 2.6.32-358.6.2.el6 (RHEL6)
64 * Client support for unpatched kernels:
65 2.6.32-358.6.2.el6 (RHEL6)
67 2.6.18-348.3.1.el5 (RHEL5)
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 9-30-2012 Whamcloud, Inc.
80 * See http://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
81 * Server support for kernels:
82 2.6.32-279.5.1.el6 (RHEL6)
83 * Client support for unpatched kernels:
84 2.6.32-279.5.1.el6 (RHEL6)
86 2.6.18-238.19.1.el5 (RHEL5)
87 2.6.16.60-0.69.1 (SLES10)
88 * Recommended e2fsprogs version: 1.41.90.wc4
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 Severity : enhancement
97 Description: Update RHEL6.3 kernel to 2.6.32-279.5.1.el6.
99 Severity : enhancement
101 Description: Update RHEL6.3 kernel to 2.6.32-279.2.1.el6.
103 Severity : enhancement
105 Description: Update RHEL6.2 kernel to 2.6.32-220.17.1.el6
107 Severity : enhancement
109 Description: Update RHEL6.2 kernel to 2.6.32-220.13.1.el6
111 Severity : enhancement
113 Description: Update RHEL6.2 kernel to 2.6.32-220.7.1.el6
115 --------------------------------------------------------------------------------
117 04-30-2012 Whamcloud, Inc.
119 * See http://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
120 * Server support for kernels:
121 2.6.32-220.4.2.el6 (RHEL6)
122 * Client support for unpatched kernels:
123 2.6.32-220.4.2.el6 (RHEL6)
125 2.6.18-238.19.1.el5 (RHEL5)
126 2.6.16.60-0.69.1 (SLES10)
127 * Recommended e2fsprogs version: 1.41.90.wc4
128 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
129 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
130 stack overflow. For more information, please refer to bugzilla 17630.
131 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
132 the re-exporting nodes to be mounted with "32bitapi" mount option
134 Severity : enhancement
136 Description: Update RHEL6.2 kernel to 2.6.32-220.4.2.el6
138 Severity : enhancement
140 Description: Update RHEL6.2 kernel to 2.6.32-220.4.1.el6
142 Severity : enhancement
144 Description: Update RHEL6.2 kernel to 2.6.32-220.el6
146 Severity : enhancement
148 Description: Update RHEL6.1 kernel to 2.6.32-131.17.1.el6
150 --------------------------------------------------------------------------------
152 09-30-2011 Whamcloud, Inc.
154 * Server support for kernels:
155 2.6.18-238.19.1.el5 (RHEL5)
156 2.6.32-131.6.1.el6 (RHEL6)
157 * Client support for unpatched kernels:
158 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
159 2.6.18-238.19.1.el5 (RHEL5)
160 2.6.32-131.6.1.el6 (RHEL6)
161 2.6.29.4-167.fc11 (FC11)
162 2.6.16.60-0.69.1 (SLES10)
164 * Recommended e2fsprogs version: 1.41.90.wc3
165 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
166 Lustre filesystem with 4K stack may cause a stack overflow. For more
167 information, please refer to bugzilla 17630.
168 * NFSv4 reexport requires Lustre client on the re-exporting nodes to be
169 mounted with "32bitapi" mount option
171 Severity : enhancement
172 Description: Update RHEL6.1 kernel to 2.6.32-131.6.1.el6
174 Severity : enhancement
175 Description: Update RHEL5.6 kernel to 2.6.18-238.19.1.el5
177 Severity : enhancement
178 Description: Update RHEL5.6 kernel to 2.6.18-238.9.1.el5
181 Frequency : only with SLES10
183 Description: Use OFED "KMP" provided by Novell
184 Details : SLES10 SP3 ships with OFED in a separate "KMP" package.
185 Lustre is now built against this package. That means you need to
186 install the ofed-kmp package from Novell for the patchless client.
187 Note that the ofed-kmp that Novell ships may not exactly match the
188 kernel version but should still be compatible.
190 Severity : enhancement
192 Description: Update RHEL5.5 kernel to 2.6.18-194.17.1.el5
193 Update OEL5.5 kernel to 2.6.18-194.17.1.0.1.el5.
195 --------------------------------------------------------------------------------
197 2010-07-15 Oracle, Inc.
199 * Server support for kernels:
200 2.6.18-164.11.1.el5 (RHEL 5)
201 2.6.18-164.11.1.0.1.el5 (OEL 5)
202 * Client support for unpatched kernels:
203 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
204 2.6.18-164.11.1.el5 (RHEL 5),
205 2.6.18-164.11.1.0.1.el5 (OEL 5)
206 2.6.16.60-0.42.8 (SLES 10),
208 2.6.29.4-167.fc11 (FC11)
209 * Recommended e2fsprogs version: 1.41.10-sun2
211 Severity : enhancement
213 Description: Distribute lfs_migrate script with lustre
214 Details : The lfs_migrate mentioned in the manual is now included in the
215 lustre tree in lustre/scripts and will distributed in lustre RPMs.
217 Severity : enhancement
219 Description: Failover nid list added to import proc file.
221 Severity : enhancement
223 Description: "som_preview" mount option required to use SOM
224 Details : The clients will need to use som_preview mount option
228 Description: Lustre does not do 1MB IOs to HW RAID
229 Details : Bump MAX_PHYS/HW_SEGMENTS and SG_ALL to 256 in the RHEL5 kernel.
231 Severity : enhancement
233 Description: lfs quota output is not very convenient for awk/sed-parsing
234 Details : Some positions in lfs quota output table could be empty or
235 non-empty which made it hard to parse it with scripts, now
236 a dash is put instead of space where there is not supposed
241 Description: Update to OFED 1.5.1
242 Details : OFED 1.5.1 will be used in environments that don't use
243 vendor supplied OFED.
247 Description: Remove lnet. prefix from lctl params display
248 Details : Also renamed lnet.memused to lnet_memused to
253 Description: Use conf_param -d to remove permanant settings.
257 Description: Add list_param -R
258 Details : lctl list_param now supports recursive mode.
261 Bugzilla : 21829, 22312
262 Description: Fix llobdstat.
263 Details : 1. Now updating the write_bytes counter
264 2. Port script fixes from 1.8.
268 Description: decrease the usage of memory on clients.
269 Details : 1. On clients, recycle dentries and inodes unused.
270 2. Delete the code related to ll_deathrow(att 6215 in bug 1443). It
275 Description: llobdstat fix and enhancement
276 Details : add a counter to set a limit to how many samples will be returned
277 fix a wildcard in the path to limit to obdfilter stats only
281 Description: lfs check only allowed for root
282 Details : Code cleanups and added a regresion test of non-root use of
287 Description: MDS Oopses in ptlrpc_dispatch_difficult_reply()/__list_add()
288 Details : Fix hr->hr_index overload.
292 Description: Fix FIEMAP functionality.
296 Description: FC11 patchless client support added
300 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and
301 OEL5.4 kernel to 2.6.18-164.11.1.0.1.el5.
305 Description: New acc-sm script and various test-framework enhancments.
309 Description: Previously landed patch was reverted.
313 Description: Update SLES10 kernel to 2.6.16.60-0.42.8 (patchless)
317 Description: Add ldiskfs maxdirsize mount option.
321 Description: Fix mkfs.lustre support for 16TB LUNs.
325 Description: Print mmp_check_interval
326 Details : print mmp_check_interval and make it possible to abort
327 mount operation in case it takes too long.
331 Description: Add cascading_rw to lustre mpi tests.
335 Description: Store SOM data in EA.
336 Details : Also several SOM related cleanups.
340 Description: Add refcount to md_open_data
344 Description: Use refcount on lov_request_set to prevent memory corruption.
348 Description: Statahead: use parent dir's i_mutex for synchronization.
352 Description: loadgen improvements.
356 Description: Enable peer health detection in ksocklnd and ko2iblnd.
360 Description: Control DCACHE_LUSTRE_INVALID flag with MDS_INODE_LOOKCUP lock
364 Description: Fix race on obd_devcie:md_stats
368 Description: Fix SLES 11 client builds.
369 Details : Resolved uninitalized pointer gcc warnings.
373 Description: Allow non-root access for "lfs check".
374 Details : Added a check in obd_class_ioctl() for OBD_IOC_PING_TARGET.
378 Description: Clear imp_force_reconnect correctly in ptlrpc_connect_interpret()
380 Severity : enhancement
382 Description: Update kernel to RHEL5.4 2.6.18-164.6.1.el5 and
383 OEL5 2.6.18-164.6.1.0.1.el5(Both in-kernel OFED enabled).
387 Description: Sanity test_27p fails.
388 Details : OBD_FAIL_OST_ENOSPC has been moved to filter_statfs().
389 Sanity tests 27 were cleaned up and improved.
393 Description: Unify req format on client/servers
394 Details : Use new req_capsule API [almost] everywhere instead of old PTLRPC
395 buffers and swabbers approach..
400 Description: drop join file stuff.
402 Severity : enhancement
404 Description: Add 2.6.27 and SLES11 (patchless client) support
408 Description: Fix in ptlrpc_expire_one_request() to print the signed time
411 Severity : enhancement
413 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
416 Severity : enhancement
418 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
420 Severity : enhancement
422 Description: correctly handle big reply message.
423 Details : send LNet event if reply is bigger then buffer and adjust this buffer
428 Description: Drop unnecessary __GFP_NOMEMALLOC flag from filter_get_page()
430 Severity : enhancement
432 Description: Add support for OEL5.
436 Description: Conf-sanity.sh 50g test - deactivated OST should not cause a panic.
438 Severity : enhancement
440 Description: "weak-modules" support
441 Details : Implement "weak-modules" support which enables kernel modules
442 to be used with any kernel that implements the same kABI. In
443 order to achieve this modules are now installed in
444 /lib/modules/$(uname -r)/updates/kernel on all distributions.
448 Description: client could not reconnect to OST because of an active request.
449 Details : The client now retries to reconnect to the same server, if a
450 connect request failed with EBUSY or -EAGAIN. Test case 35b
451 conf-sanity.sh has been created.
455 Description: Don't allow make backward step on assignin osc next id.
456 Details : race between allocation next id and ll_sync thread can be cause
457 of set wrong osc next id and can be kill valid ost objects.
461 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
465 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
469 Description: Error handling in osc_statfs_interpret() has been improved.
470 Details : Check in osc_statfs_interpret() for EBADR.
474 Description: Do not send statfs() requests to OSTs disabled by administrator.
475 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
477 Severity : enhancement
479 Description: Added sanity-quota test to for lfs(1) to work with numeric uid/gid.
481 Severity : enhancement
483 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
485 Severity : enhancement
487 Description: Enhancement for lfs(1) command to use numeric uid/gid.
491 Description: Improved cleanup procedure for conf-sanity test_46a.
495 Description: Do not start bulk transfer, if client tried to reconnect.
496 Details : Do not start bulk transfer, if client tried to reconnect,
497 but failed due to active rpcs. Check for exp_abort_active_req
498 has been implemented.
502 Description: Prevent inconsistences between linux and lustre mount structures.
503 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
504 Make the sleep interruptible.
506 Severity : enhancement
508 Description: provide server to client comms path
509 Details : server to client communications path via new LDLM_SET_INFO rpc
513 Description: truncate starts GFP_FS allocation under transaction causing deadlock
514 Details : ldiskfs_truncate calls grab_cache_page which may start page
515 allocation under an open transaction. This may lead to
516 calling prune_icache with consequent lustre reentrance.
520 Description: send correctly lsm on open replay
521 Details : MDS is trust to LSM size on replay open, but client can set wrong size
526 Description: enable client interface failover
527 Details : When a child reconnects from another NID, properly update export
528 nid hash position and ldlm reverse import.
530 Severity : enhancement
532 Description: Communicate OST degraded/readonly state via statfs to MDS
533 Details : Flags in the statfs returned from OSTs indicate whether the
534 OST is in a degraded RAID state, or if the filesystem has
535 turned read-only after a filesystem error is detected.
540 Description: don't panic if EPROTO was hit when reading symlink
541 Details : correctly handling request reference in error cases.
543 Severity : enhancement
545 Description: Add LustreNetLink, a kernel-userspace communcation path.
547 Severity : enhancement
549 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
554 Description: access to llog context before init.
555 Details : move handling CATALOGS file at osc layer and forbid access to llog
559 Frequency : rare, only when client is writing to OST on local node
561 Description: Avoid deadlock for local client writes
562 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
563 memory freeing context. This allows OST threads to set the
564 PF_MEMALLOC flag on task structures in order to allocate memory
565 from reserved pools and complete IO.
566 Use GFP_HIGHUSER for OST allocations for non-local client writes,
567 so that the OST threads generate memory pressure and allow
568 inactive pages to be reclaimed.
570 Severity : enhancement
572 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
577 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
578 Details : move ldlm namespace creation in setup phase to avoid grab
579 _lprocfs_lock with cli_sem held.
583 Description: Temporarily disable grant shrink.
584 Details : Disable the feature for debugging.
587 Frequency : only when formatting test filesystems
589 Description: Unable to run several mkfs.lustre on loop devices at the same time
590 Details : mkfs.lustre returns error 256 on the concurrent loop devices
591 formatting. The solution is to proper handle the error.
593 Severity : enhancement
595 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
597 Severity : enhancement
599 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
604 Description: resolve race between obd_disconnect and class_disconnect_exports
605 Details : if obd_disconnect will be called to already disconnected export he
606 forget release one reference and osc module can't unloaded.
608 Severity : enhancement
610 Description: Allow tuning service thread via /proc
611 Details : For each service a new
612 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
613 created that can be used to set min/max thread counts, and get the
614 current number of running threads.
619 Descriptoin: don't return error if have partially created objects for file.
620 Details : lov_update_create_set uses set->set_success as index for created
621 objects, so if some requests will be failed, they will have hole
622 at end of array and qos_shrink_lsm for allocate correct lsm.
624 Severity : enhancement
626 Description: Update OFED support to 1.4.1
631 Description: Reduce small size read RPC
632 Details : Set read-ahead limit for every file and only do read-ahead when
633 available read-ahead pages are bigger than 1M to avoid small size
636 Severity : enhancement
638 Description: move AT tunable parameters for more consistent usage
639 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
641 Severity : enhancement
643 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
644 Details : allow skip disconnected ost for send statfs request and hide error
650 Description: fix racy locking of mballoc block bitmaps causing BUG
651 Details : The locking of the mballoc buddy bitmap and the in-memory
652 block bitmap was using two different spin locks in some
653 cases. This made it possible to incorrectly access the
654 mballoc bitmap while another process was modifying it,
655 causing a sanity assertion to fail. While no on-disk corruption
656 was reported, there was some risk of this happening.
659 Frequency : rare, on llog test 6
661 Descriptoin: don't allow connect to already connected import
662 Details : allowing connect to already connected import is hide connecting problem.
667 Descriptoin: allow kill process which wait statahead result
668 Details : for some reasons 'ls' can stick in waiting result from statahead,
669 in this case need way for kill this process.
671 Severity : enhancement
673 Description: Add state history info file, enhance import info file
674 Details : Track import connection state changes in a new osc/mdc proc file;
675 add overview-type data to the osc/mdc import proc file.
677 Severity : enhancement
679 Description: MDS create should not wait for statfs RPC while holding DLM lock.
681 Severity : enhancement
683 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
688 Descriptoin: OST grant too much space to client even there are not enough space.
689 Details : Client will shrink its grant cache to OST if there are no write
690 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
691 this grant cache if there are already not enough avaible space
692 (left_space < total_clients * 32M).
695 Frequency : start MDS on uncleanly shutdowned MDS device
697 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
698 Details : stay in waiting mds<>ost recovery finished produce random bugs
699 due race between two ll_sync thread for one lov target. send
700 ACTIVATE event only if connect realy finished and import have
704 Frequency : rare, connect and disconnect target at same time
706 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
707 Details : don't call obd_disconnect under lov_lock. this long time
708 operation and can block ptlrpcd which answer to connect request.
713 Descriptoin: don't lose wakeup for imp_recovery_waitq
714 Details : recover_import_no_retry or invalidate_import and import_close can
715 both sleep on imp_recovery_waitq, but we was send only one wakeup
719 Frequency : always with long access acl
721 Descriptoin: mds can't pack reply with long acl.
722 Details : mds don't control size of acl but they limited by reint/getattr
725 Severity : enhancement
727 Description: Update to SLES10 kernel-2.6.16.60-0.33.
729 Severity : enhancement
731 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
734 Frequency : start MDS on uncleanly shutdowned MDS device
736 Descriptoin: aborting recovery hang on MDS
737 Details : don't throttle destroy RPCs for the MDT.
740 Frequency : on remount
742 Description: external journal device not working after the remount
743 Details : clear dev_rdonly flag for external journal devices in
749 Description: shutdown vs evict race
750 Details : client_disconnect_export vs connect request race.
751 if client will evicted at this time - we start invalidate
752 thread without referece to import and import can be freed
756 Frequency : rare, need acl's on inode.
758 Description: client can't handle ost additional correctly
759 Details : if ost was added after client connected to mds client can have
760 hit lnet_try_match_md ... to big messages to wide striped files.
761 in this case need teach client to handle config events about add
762 lov target and update client max ea size at that event.
764 Severity : enhancement
766 Description: Changelogs
767 Details : Changelogs are a lightweight mechanism to track filesystem
768 metadata and namespace changes. The changelog is recorded
769 permanently on the MDTs, and is periodically "consumed" / purged
770 when records are no longer needed.
772 Severity : enhancement
774 Description: compact fld format with extents
775 Details : Store range of seq rather than every seq in FLD. Seq
776 controller update FLD rather than clients. In Case of CMD, mdt0
777 has FLD, all other metadata server act as non persistent proxy
778 for FLD queries and cache fld entries in fld cache.
783 Description: don't skip ost target if they assigned to file
784 Details : Drop slow OSCs if we can, but not for requested start idx.
785 This means "if OSC is slow and it is not the requested
786 start OST, then it can be skipped, otherwise skip it only
787 if it is inactive/recovering/out-of-space.
791 Description: more cleanup in mds_lov
792 Details : not send LOV EA under replay, we can't know about they size at this
793 time. Don't allow client connect to mds before any ost connected,
794 for avoid problems with LOV EA size and returning EIO to client.
796 Severity : enhancement
798 Description: Interoperability at server side (Disk interoperability)
800 Severity : enhancement
802 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
804 Severity : enhancement
806 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
808 Severity : enhancement
810 Description: New client IO stack (CLIO).
812 Severity : enhancement
814 Description: Commit on sharing. Eliminate inter-client dependencies between
815 uncommitted transactions by doing transaction commits.
816 Thereby clients may recovery independently.
819 Frequency : Create a symlink file with a very long name
821 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
822 Details : If there is no extra space in the request for early cancels,
823 ldlm_req_handles_avail() returns 0 instead of a negative value.
825 Severity : enhancement
827 Description: Add /proc entry for import status
828 Details : The mdc, osc, and mgc import directories now have
829 an import directory that contains useful import data for debugging
832 Severity : enhancement
834 Description: Re-disable certain /proc logging
835 Details : Enable and disable client's offset_stats, extents_stats and
836 extents_stats_per_process stats logging on the fly.
839 Frequency : Only on FC kernels 2.6.22+
841 Description: oops in statahead
842 Details : Do not drop reference count for the dentry from VFS when lookup,
843 VFS will do that by itself.
845 Severity : enhancement
847 Description: Generic /proc file permissions
848 Details : Set /Proc file permissions in a more generic way to enable non-
849 root users operate on some /proc files.
853 Description: Hitting mdc_commit_close() ASSERTION
854 Details : Properly handle request reference release in
855 ll_release_openhandle().
859 Description: quota recovery deadlock during mds failover
860 Details : This patch includes att18982, att18236, att18237 in bz14840.
862 1. fix osts hang when mds does failover with quotaon
863 2. prevent watchdog storm when osts threads wait for the
868 Frequency : only patchless client
869 Description: add workaround for race between add/remove dentry from hash
871 Severity : enhancement
873 Description: Allow OST glimpses to return PW locks
877 Description: LBUG when llog conf file is full
878 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
882 Description: Prevent import from entering FULL state when server in recovery
886 Description: service mount cannot take device name with ":"
887 Details : Only when device name contains ":/" will mount treat it as
893 Description: replace ptlrpcd with the statahead thread to interpret the async
894 statahead RPC callback
898 Frequency : on recovery
899 Description: I/O failures after umount during fail back
900 Details : if client reconnected to restarted server we need join to recovery
901 instead of find server handler is changed and process self
902 eviction with cancel all locks.
904 Severity : enhancement
906 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
908 Severity : enhancement
910 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
912 Severity : enhancement
914 Description: Upcall on Lustre log has been dumped
915 Details : Allow for a user mode script to be called once a Lustre log has
916 been dumped. It passes the filename of the dumped log to the
917 script, the location of the script can be specified via
918 /proc/sys/lnet/debug_log_upcall.
923 Description: avoid idr_remove called for id which is not allocated.
924 Details : Move assigment s_dev for clustered nfs to end of initialization,
925 to avoid problem with error handling.
930 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
931 Details : When connection is reused this not moved from CONN_UNUSED_HASH
932 into CONN_USED_HASH and this prodice warning when put connection
933 again in unused hash.
936 Severity : enhancement
938 Description: File striping can now be set to use an arbitrary pool of OSTs.
940 Severity : enhancement
942 Description: Export bytes_read/bytes_write count on OSC/OST.
946 Description: Early reply size mismatch, MGC loses connection
947 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
948 the connect flags are properly negotiated.
953 Description: quotas are not honored with O_DIRECT
954 Details : all writes with the flag O_DIRECT will use grants which leads to
955 this problem. Now using OBD_BRW_SYNC to guard this.
959 Description: add quota statistics
960 Details : 1. sort out quota proc entries and proc code.
961 2. add quota statistics
963 Severity : enhancement
965 Description: enable quota support for HEAD.
969 Description: Properly propagate oinfo flags from lov to osc for statfs
970 Details : restore missing copy oi_flags to lov requests.
972 Severity : enhancement
974 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
975 : l_getgroups(8), lst(8), routerstat(8)
976 Details : included man pages for llobdstat(8), llstat(8),
977 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
979 Severity : enhancement
981 Description: configure's --enable-quota should check the
982 : kernel .config for CONFIG_QUOTA
983 Details : configure is terminated if --enable-quota is passed but
984 : no quota support is in kernel
988 Description: Remove portals compatibility
989 Details : Remove portals compatibility, not interoperable with releases
994 Description: Resolve device initialization race
995 Details : Prevent proc handler from accessing devices added to the
996 obd_devs array but yet be intialized.
998 Severity : enhancement
1000 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
1002 Severity : enhancement
1004 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
1009 Description: Using wrong pointer in osc_brw_prep_request
1010 Details : Access to array[-1] can produce panic if kernel compiled with
1011 CONFIG_PAGE_ALLOC enabled
1015 Description: Client runs out of low memory
1016 Details : Consider only lowmem when counting initial number of llap pages
1020 Description: *optional* service tags registration
1021 Details : if the "service tags" package is installed on a Lustre node
1022 When the filesystem is mounted, a local-node service tag will
1023 be created. See http://inventory.sun.com/ for more information
1024 about the Service Tags asset management system.
1028 Description: Kernel BUG tries to release flock
1029 Details : Lustre does not destroy flock lock before last reference goes
1030 away. So always drop flock locks when client is evicted and
1031 perform unlock regardless of successfulness of speaking to MDS.
1035 Description: add refcount protection for osc callbacks, avoid panic on shutdown
1039 Description: sanity test 65a fails if stripecount of -1 is set
1040 Details : handle -1 striping on filesystem in ll_dirstripe_verify
1045 Description: ASSERTION(CheckWriteback(page,cmd)) failed
1046 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
1049 Severity : enhancement
1051 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
1055 Description: do not process already freed flock
1056 Details : flock can possibly be freed by another thread before it reaches
1057 to ldlm_flock_completion_ast.
1061 Description: LBUG during stress test
1062 Details : Need properly lock accesses the flock deadlock detection list.
1066 Description: oops in page fault handler
1067 Details : kernel page fault handler can return two special 'pages' in error
1068 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
1072 Description: timeout with invalidate import.
1073 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
1074 should be handled by ptlrpcd. This produce long age waiting and
1075 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
1077 Severity : enhancement
1079 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
1083 Description: enable MGS and MDT services start separately
1084 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
1085 a co-located MGS without starting the MGS, which is a complement
1086 to 'nosvc' mount option.
1090 Frequency : after recovery
1091 Description: precreate to many object's after del orphan.
1092 Details : del orphan st in oscc last_id == next_id and this triger growing
1093 count of precreated objects. Set flag LOW to skip increase count
1094 of precreated objects.
1098 Frequency : rare, on clear nid stats
1099 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
1100 Details : when clean nid stats sometimes try destroy live entry,
1101 and this produce panic in free.
1105 Description: Stack overflow during MDS log replay
1106 ease stack pressure by using a thread dealing llog_process.
1110 Description: wait until IO finished before start new when do lock cancel.
1111 Details : VM protocol want old IO finished before start new, in this case
1112 need wait until PG_writeback is cleared until check dirty flag and
1113 call writepages in lock cancel callback.
1115 Severity : enhancement
1117 Description: using special macro for print time and cleanup in includes.
1121 Description: mds_mfd_close() ASSERTION(rc == 0)
1122 Details : In mds_mfd_close(), we need protect inode's writecount change
1123 within its orphan write semaphore to prevent possible races.
1127 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
1131 Frequency : rare, on shutdown ost
1132 Description: don't hit live lock with umount ost.
1133 Details : shrink_dcache_parent can be in long loop with destroy dentries,
1134 use shrink_dcache_sb instead.
1138 Description: don't panic with use echo client
1139 Details : echo client pass NULL as client nid pointer and this produce null
1140 pointer dereference.
1144 Description: fix build on ppc32
1145 Details : compile code with -m64 flag produce wrong object file for ppc32.
1149 Description: add message levels for liblustreapi
1153 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
1154 Details : Move the 'good_osts' check before the 'total_bavail' check. This
1155 will result in an -EAGAIN and in the exit call path we call
1156 alloc_rr() which will with increasing aggressiveness attempt to
1157 aquire precreated objects on the minimum number of required OSCs.
1161 Description: Use old size assignment to avoid deadlock
1162 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
1163 the scheduling while holding a spinlock. We do not need locking
1164 for size in ll_update_inode() because size is only updated from
1165 the MDS for directories or files without objects, so there is no
1166 other place to do the update, and concurrent access to such inodes
1167 are protected by the inode lock.
1171 Description: resolve "_IOWR redefined" build error on SLES10
1175 Description: dump the memory debugging after all modules are unloaded to
1176 suppress false negative in conf_sanity test 39
1178 Severity : enhancement
1180 Description: build kernel-ib packages for OFED 1.3 in our release cycle
1185 Description: fix SLES kernel versioning
1186 Details : the kernel version for our SLES 10 kernel did not include a "-"
1187 before the "smp" at the end. while this was not a problem in
1188 general, it did mean that software trying to use the kernel
1189 version to try to detect a vendor specific kernel would fail.
1190 this was most evident by the OFED build scripts.
1194 Description: Don't update lov_desc members until making sure they are valid
1195 Details : When updating lov_desc members via proc fs, need fix their
1196 validities before doing the real update.
1200 Description: don't put request into delay list while invalidate in flight.
1201 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
1202 import in flight. this produce timeout for invalidate and sometimes
1203 can cause stale data.
1207 Frequency : on ppc only
1208 Description: not convert ost objects for directory because it's not exist.
1209 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
1213 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
1215 Severity : enhancement
1217 Description: Update to support 2.6.22.14 vanilla kernel.
1221 Frequency : rare, on recovery
1222 Description: read procfs can produce deadlock in some situation
1223 Details : Holding lprocfs lock which send rpc can produce block for destroy
1224 obd objects and this also block reconnect with -EALREADY. This
1225 isn't fix all lprocfs bugs - but make it rare.
1227 Severity : enhancement
1229 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
1232 Frequency : frequent on X2 node
1234 Description: mdc_set_open_replay_data LBUG
1235 Details : Set replay data for requests that are eligible for replay.
1239 Description: lustre_mgs: operation 101 on unconnected MGS
1240 Details : When MGC is disconnected from MGS long enough, MGS will evict the
1241 MGC, and late on MGC cannot successfully connect to MGS and a lot
1242 of the error messages complaining that MGS is not connected.
1246 Frequency : on network error
1247 Description: panic with double free request if network error
1248 Details : mdc_finish_enqueue is finish request if any network error ocuring,
1249 but it's true only for synchronus enqueue, for async enqueue
1250 (via ptlrpcd) this incorrect and ptlrpcd want finish request
1253 Severity : enhancement
1255 Description: client-side metadata stat-ahead during readdir(directory readahead)
1256 Details : perform client-side metadata stat-ahead when the client detects
1257 readdir and sequential stat of dir entries therein
1260 Frequency : on start mds
1262 Description: Implement get_info(last_id) in obdfilter.
1265 Frequency : occasional
1267 Description: Correctly check stale fid, not start epoch if ost not support SOM
1268 Details : open with flag O_CREATE need set old fid in op_fid3 because
1269 op_fid2 was overwritten with new generated fid, but mds can answer
1270 with one of these two fids and both is not stale. Setattr
1271 incorrectly started an epoch and assume will be called
1272 done_writing, but without SOM done_writing ever being called.
1275 Frequency : rare, depends on device drivers and load
1277 Description: MDS or OSS nodes crash due to stack overflow
1278 Details : Code changes in 1.8.0 increased the stack usage of some functions.
1279 In some cases, in conjunction with device drivers that use a lot
1280 of stack the MDS (or possibly OSS) service threads could overflow
1281 the stack. One change which was identified to consume additional
1282 stack has been reworked to avoid the extra stack usage.
1285 Frequency : occasional
1287 Description: Do not fail import if osc_interpret_create gets -EAGAIN
1288 Details : If osc_interpret_create got -EAGAIN it immediately exits and
1289 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
1290 oscc_has_objects and see OSC has no objests and call
1291 oscc_internal_create to resend create request.
1293 Severity : enhancement
1295 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
1297 Severity : enhancement
1299 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
1302 Frequency : very rare
1304 Description: Fix replay if there is an un-replied request and open
1305 Details : In some cases, older replay request will revert the
1306 mcd->mcd_last_xid on MDS which is used to record the client's
1307 latest sent request.
1309 Severity : enhancement
1311 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
1313 Severity : enhancement
1315 Description: Add rhel5 support to HEAD.
1317 Serverity : enhancement
1319 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
1324 Description: Don't allow skipping OSTs if index has been specified.
1325 Details : Don't allow skipping OSTs if index has been specified, make
1326 locking in internal create lots better.
1330 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
1331 Details : access to bitfield in structure is always rounded to long
1332 and this produce problem with not atomic change any bit.
1336 Description: Lustre make rpms failed.
1337 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
1338 builds Lustre rpms from tarball.
1341 Frequency : rare on shutdown ost
1343 Description: If llog cancel was not send before clean_exports phase, this can
1344 produce deadlock in llog code.
1345 Details : If llog thread has last reference to obd and call class_import_put
1346 this produce deadlock because llog_cleanup_commit_master wait when
1347 last llog_commit_thread exited, but this never success because was
1348 called from llog_commit_thread.
1352 Description: allow userland application know is lost one of stripes.
1353 Details : fill lvb_blocks with error code on ost and return it to
1354 application if error flag found.
1358 Description: NULL lov_tgts causing MDS oops
1359 Details : more safe checks for NULL lov_tgts for avoid oops.
1361 Severity : enhancement
1363 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
1367 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
1368 Details : If an OST is down the MDS will hang indefinitely in
1369 obd_statfs_rqset() waiting for the statfs data. While for
1370 MDS QOS usage of statfs, it should not stuck in waiting.
1372 Severity : enhancement
1374 Description: remote_acl support
1375 Details : Support ACL-based permission check for remote user.
1376 Support setfacl/getfacl for remote user with the utils
1377 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
1378 the system "{s,g}etfacl" utils.
1380 Severity : enhancement
1382 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
1384 Severity : enhancement
1386 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
1390 Description: make mgs_setparam() handle fsname containing dash
1391 Details : fsname containing a dash does not work with lctl conf_param
1393 Severity : enhancement
1395 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
1397 Severity : enhancement
1399 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
1403 Frequency : rare, at shutdown
1404 Description: access already free / zero obd_namespace.
1405 Details : if client_disconnect_export was called without force flag set,
1406 and exist connect request in flight, this can produce access to
1407 NULL pointer (or already free pointer) when connect_interpret
1408 store ocd flags in obd_namespace.
1412 Frequency : only at startup
1413 Description: not alloc memory with spinlock held.
1414 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
1415 if any spinlock held.
1417 Severity : enhancement
1419 Description: make lustre randomly fail allocating memory
1420 Details : Make lustre randomly failed allocating memory for testing purpose.
1422 Severity : enhancement
1424 Description: lost problems with lov objid file
1425 Details : Fixes some scability and access to not inited memory problems
1426 in work with lov objdid file.
1431 Description: lfs find does not continue on file error
1432 Details : Continue other files processing when a file/dir is absent.
1436 Description: Inconsistent usage of lustre_pack_reply()
1437 Details : Standardize the usage of lustre_pack_reply() such that it
1438 always generate a CERROR on failure.
1441 Frequency : occasional
1443 Description: MDS hang or stay in waiting lock
1444 Details : If client receive lock with CBPENDING flag ldlm need send lock
1445 cancel as separate rpc, to avoid situation when cancel request
1446 can't processed due all i/o threads stay in wait lock.
1450 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1454 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1458 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1462 Description: Update to RHEL4 latest kernel.
1464 Severity : enhancement
1466 Description: Build SLES10 patchless client fails
1467 Details : The configure was broken by run ./configure with
1468 --with-linux-obj=.... argument for patchless client. When the
1469 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1470 can't search header adequately. Use absolute path such as
1471 -I($LINUX)/include instead.
1475 Description: interrupt oig_wait produce painc on resend.
1476 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1477 private set, and this produce situation when rq_ptlrpcd_data not
1478 copyed to new allocated request and triggered LBUG on assert
1479 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1480 ptlrpcd set if request is changed and can be safety changed to use
1485 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1486 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1487 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1488 these macros and replace them with their difinitions instead.
1492 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1493 Details : Update which_patch & target file for SLES10 latest kernel.
1496 Description: Remove client patches from SLES 10 kernel.
1497 Details : This causes SLES 10 clients to behave as patchless clients
1498 even on a Lustre-patched (server) kernel.
1500 Severity : enhancement
1502 Description: self-adjustable client's lru lists
1503 Details : use adaptive algorithm for managing client cached locks lru
1504 lists according to current server load, other client's work
1505 pattern, memory activities, etc. Both, server and client
1506 side namespaces provide number of proc tunables for controlling
1509 Severity : enhancement
1511 Description: light-weight GSS support
1512 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
1513 and reduce performance overhead.
1515 Severity : enhancement
1517 Description: Linux keyring support
1518 Details : Support using service of Linux keyring for Lustre GSS internal
1519 context refresh/cache mechanism.
1523 Description: Fix errors in lfs documentation
1524 Details : Fixes man pages
1528 Description: don't use GFP_* in generic Lustre code.
1529 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
1533 Description: obdclass is limited by single OBD_ALLOC(idarray)
1534 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
1538 Description: warnings with build patchless client with vanila 2.6.19 and up
1539 Details : change old ctl_table style and replace ctl_table/ctl_table_header
1540 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
1544 Frequency : Only for SLES
1545 Description: sanity_quota fail test_1
1546 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
1547 file, which makes TSTID[2] inunique.
1551 Frequency : Only for SLES9
1552 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
1553 Details : Update target/ChangeLog/which_patch .
1557 Frequency : Only for test_10 in sanity-quota.sh
1558 Description: a bug in quota test code
1559 Details : checking a test flag in a wrong place
1563 Frequency : Only for test_13 in sanity-quota.sh
1564 Description: a bug in quota test script
1565 Details : shouldn't check overfull allocated quota
1569 Description: jbd statistics
1570 Details : Port older jbd statistics patch for sles10
1572 Severity : enhancement
1574 Description: Add printing inode info into message about error in writepage.
1579 Description: reply_lock_interpret crash due to race with it and lock cancel.
1580 Details : Do not replay locks that are being cancelled. Do not reference
1581 locks by their address during replay, just by their handle.
1585 Frequency : When flocks are used.
1586 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
1587 Details : Flock locks might destroy just granted lock if it could be merged
1588 with another existing flock, this is done in completion handler,
1589 so teach ldlm_cli_enquque_fini that this is a valid case for
1595 Description: Oops in read and write path when failing to allocate lock.
1596 Details : Check if lock allocation failed and return error back.
1600 Description: lstripe command fails for valid OST index
1601 Details : The stripe offset is compared to lov->desc.ld_tgt_count
1602 instead of lov->desc.ld_active_tgt_count.
1606 Description: sanity.sh failed test 103
1607 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
1612 Description: df doesn't work properly if diskfs blocksize != 4K
1613 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
1617 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
1618 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
1620 Severity : enhancement
1622 Description: Adaptive timeouts
1623 Details : RPC timeouts adapt to changing server load and network
1624 conditions to reduce resend attempts and improve recovery time.
1628 Description: llapi_file_create() does not allow some changes
1629 Details : add llapi_file_open() that allows specifying the mode and
1630 open flags, and also returns an open file handle.
1634 Description: merge and cleanup kernel patches.
1635 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
1639 Description: Add journal checksum support.(Kernel part)
1640 Details : The journal checksum feature adds two new flags i.e
1641 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
1642 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
1643 indicates that the commit block contains the checksum for
1644 the blocks described by the descriptor blocks. Now commit
1645 record can be sent to disk without waiting for descriptor
1646 blocks to be written to disk. This behavior is controlled
1647 using JBD2_FEATURE_ASYNC_COMMIT flag.
1651 Description: OSS needs mutliple precreate threads
1652 Details : Add ability to start more than one create thread per OSS.
1656 Description: Sanity.sh test_65k failed: setstripe should have succeeded
1657 Details : execute lfs setstripe on client
1661 Description: mds_obd_create error creating tmp object
1662 Details : When the user sets quota on root, llog will be affected and can't
1663 create files and write files.
1666 Frequency : Always on ia64 patchless client, and possibly others.
1668 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
1669 Details : This allows the patchless client to be loaded on architectures
1670 without this export.
1674 Description: RedHat Update kernel for RHEL5
1675 Details : Add the kernel config file for RHEL5.
1679 Description: RedHat Update kernel for RHEL5
1680 Details : Modify the kernel config file more closer RHEL5.
1684 Description: Build failure against Centos5 (RHEL5)
1685 Details : Define PAGE_SIZE when it isn't present.
1689 Description: test_7 on sanity-quota.sh can't be used on separate machines
1690 Details : add facet to handle it
1694 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
1695 Details : with NFS, the anon dentry's parent was set to itself in
1696 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
1697 corresponding dentry other than use rec->ur_name.
1699 Severity : enhancement
1701 Description: lfs setstripe enhancement
1702 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
1706 Description: enable data checksumming by default
1707 Details : enable checksum by default, allow --disable-checksum
1708 configure option and "-o nochecksum" mount option. Checksums
1709 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
1713 Description: lustre support for RHEL5
1714 Details : Add support for RHEL5.
1718 Description: Client eviction due to failover config
1719 Details : after a connection loss, the lustre client should attempt to
1720 reconnect to the last active server first before trying the
1721 other potential connections.
1725 Description: when mds and osts use different quota unit(32bit and 64bit),
1726 quota will be released repeatly.
1727 Details : void sending multiple quota reqs to mds, which will keep the
1728 status between the reqs.
1732 Description: osts not allocated evenly to files
1733 Details : change the condition to increase offset_idx
1736 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
1738 Description: Data corruption for OSTs that are formatted larger than 2TB
1740 Details : When generating the bio request for lustre file writes the
1741 sector number would overflow a temporary variable before being
1742 used for the IO. The data reads correctly from Lustre (which
1743 will overflow in a similar manner) but other file data or
1744 filesystem metadata may be corrupted in some cases.
1748 Description: Tune the kernel for good SCSI performance.
1749 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
1750 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
1755 Description: rewrite ext2-derived code in obdclass/uuid.c
1756 Details : rewrite inherited code (uuid parsing code from ext2 utils)
1757 from scratch preserving functionality.
1761 Description: "lfs find -obd UUID" prints directories
1762 Details : "lfs find -obd UUID" will return all directory names instead
1763 of just file names. It is incorrect because the directories
1764 do not reside on the OSTs.
1768 Description: use i_size_read and i_size_write in 2.6 port
1769 Details : replace inode->i_size access with i_size_read/write()
1772 Frequency : when removing large files
1774 Description: scheduling issue during removal of large Lustre files
1775 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
1776 It causes scheduling issues when removing large files (17TB in the
1780 Frequency : only with liblustre clients on XT3
1782 Description: evictions taking too long
1783 Details : allow llrd to evict clients directly on OSTs
1786 Frequency : only on ppc
1788 Description: /proc/fs/lustre/devices broken on ppc
1789 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
1790 We should make sure the type of 'index' is loff_t and then cast
1791 explicitly as needed below. Do not assign an explicitly cast
1795 Frequency : only for rhel5
1797 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
1798 Details : Modify the target file & which_kernel.
1800 Severity : enhancement
1802 Description: omit set fsid for export NFS
1803 Details : fix set/restore device id for avoid EMFILE error and mark lustre
1804 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
1808 Frequency : Always, for kernels after 2.6.16
1809 Description: Fix warning idr_remove called for id=.. which is not allocated.
1810 Details : Last kernels save old s_dev before kill super and not allow
1811 to restore from callback - restore it before call kill_anon_super.
1815 Description: buffer overruns could theoretically occur
1816 Details : llapi_semantic_traverse() modifies the "path" argument by
1817 appending values to the end of the origin string, and a
1818 overrun may occur. Adding buffer overrun check in liblustreapi.
1822 Description: Fix error on 'ls .' at the top of the Lustre mount.
1823 Details : Don't revalidate dentry if it is a root dentry.
1827 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
1828 Details : Modify vm-tunables-rhel4.patch.
1832 Description: Kernel config for 2.6.18-vanilla.
1833 Details : Modify targets/2.6-vanilla.target.in.
1834 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
1835 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
1836 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
1837 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
1841 Description: improve handling recoverable errors
1842 Details : if request processig with error which can be recoverable on server
1843 request should be resend, otherwise page released from cache and
1848 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
1849 Details : Modify target file & which_patch.
1850 A flaw was found in the IA32 system call emulation provided
1851 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
1852 value could be stored in the %RAX register, which could trigger an
1853 out-of-bounds system call table access. An untrusted local user
1854 could exploit this flaw to run code in the kernel
1855 (ie a root privilege escalation). (CVE-2007-4573).
1859 Description: O_DIRECT bypasses client statistics.
1860 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
1861 read_ahead_stats were updated. copied the stats section from
1862 osc_send_oap_rpc() into async_internal().
1866 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
1870 Description: change order of libsysio includes
1871 Details : '#include sysio.h' should always come before '#include xtio.h'
1875 Description: handle "serious error: objid * already exists" more gracefully
1876 Details : If LAST_ID value on disk is smaller than the objects existing in
1877 the O/0/d* directories, it indicates disk corruption and causes an
1878 LBUG(). If the object is 0-length, then we should use the existing
1879 object. This will help to avoid a full fsck in most cases.
1881 Severity : enhancement
1883 Description: adapt the lustre_config script to support the upgrade case
1884 Details : Add "-u" option for lustre_config script to support upgrading 1.4
1885 server targets to 1.6 in parallel.
1889 Description: To avoid grant space > avaible space when the disk is almost
1890 full. Without this patch you might see the error "grant XXXX >
1891 available" or some LBUG about grant, when the disk is almost
1893 Details : In filter_check_grant, for non_grant cache write, we should
1894 check the left space by if (*left > ungranted + bytes), instead
1895 of (*left > ungranted), because only we are sure the left space
1896 is enough for another "bytes", then the ungrant space should be
1897 increase. In client, we should update cl_avail_grant only there
1898 is OBD_MD_FLGRANT in the reply.
1902 Description: Update RHEL 4 kernel to fix local root privilege escalation.
1903 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
1904 described in CVE-2007-4573. This problem could allow untrusted
1905 local users to gain root access.
1908 Frequency : when using O_DIRECT and quotas
1910 Description: Incorrect file ownership on O_DIRECT output files
1911 Details : block usage reported by 'lfs quota' does not take into account
1912 files that have been written with O_DIRECT.
1917 Description: touch file failed when fs is not full
1918 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
1919 otherwise we can get ENOSP while fs is not full.
1923 Description: parallel lock callbacks
1924 Details : Instead of sending blocking and completion callbacks as separated
1925 requests, adding them to a set and sending in parallel.
1928 Frequency : only for Cray XT3
1929 Bugzilla : 12829/13455
1930 Description: Changing primary group doesn't change the group lustre assigns to
1932 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
1933 overridden with the primary group provided by the group upcall,
1934 whereas the supplied fsgid can be trusted if it is in the list of
1935 supplementary groups returned by the group upcall.
1937 Severity : enhancement
1939 Description: Allow masking D_WARNING, D_ERROR messages from console
1940 Details : Console messages can now be disabled via lnet.printk.
1944 Description: User code with malformed file open parameter crashes client node
1945 Details : Before packing join_file req, all the related reference should be
1946 checked carefully in case some malformed flags cause fake
1947 join_file req on client.
1951 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
1952 Details : ldlm_completion_ast() assumes that a lock is granted when the req
1953 mode is equal to the granted mode. However, it should also check
1954 that LDLM_FL_CP_REQD is not set.
1958 Description: Heavy nfs access might result in deadlocks
1959 Details : After ELC code landed, it is now improper to enqueue any mds
1960 locks under och_sem, because enqueue might want to decide to
1961 cancel open locks for same inode we are holding och_sem for.
1965 Description: Client eviction while running blogbench
1966 Details : A lot of unlink operations with concurrent I/O can lead to a
1967 deadlock causing evictions. To address the problem, the number of
1968 oustanding OST_DESTROY requests is now throttled to
1969 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
1970 callbacks are processed in priority.
1974 Description: enable ACLs on MDS by default
1975 Details : ACLs must be enabled on MDS by default.
1978 Frequency : PPC/PPC64 only
1980 Description: conflicts between asm-ppc64/types.h and lustre_types.h
1981 Details : fix duplicated definitions between asm-ppc64/types.h and
1982 lustre_types.h on PPC.
1985 Frequency : PPC/PPC64 only
1987 Description: asm-ppc/segment.h does not exist
1988 Details : fix compile issue on PPC.
1992 Description: better handle error messages in extents code
1995 Frequency : RHEL4 only
1997 Description: mkfs is very slow on IA64/RHEL4
1998 Details : A performance regression has been discovered in the MPT Fusion
1999 driver between versions 3.02.73rh and 3.02.99.00rh. As a
2000 consequence, we have downgraded the MPT Fusion driver in the RHEL4
2001 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
2003 Severity : enhancement
2005 Description: SNMP support enhancement
2006 Details : Adding total number of sampled request for an MDS node in snmp
2009 Severity : enhancement
2011 Description: Optimize ldlm waiting list processing for PR extent locks
2012 Details : When processing waiting list for read extent lock and meeting read
2013 lock that is same or wider to it that is not contended, skip
2014 processing rest of the list and immediatelly return current
2015 status of conflictness, since we are guaranteed there are no
2016 conflicting locks in the rest of the list.
2020 Description: Time out and refuse to reconnect
2021 Details : When the failover node is the primary node, it is possible
2022 to have two identical connections in imp_conn_list. We must
2023 compare not conn's pointers but NIDs, otherwise we can defeat
2024 connection throttling.
2028 Description: port llog fixes from b1_6 into HEAD
2029 Details : Port llog reference couting and some llog cleanups from b1_6
2030 (bug 10800) into HEAD, for protect from panic and access to already
2031 free llog structures.
2035 Description: Detect stride IO mode in read-ahead
2036 Details : When a client does stride read, read-ahead should detect that and
2037 read-ahead pages according to the detected stride pattern.
2041 Description: data checksumming impacts single node performance
2042 Details : add support for several checksum algorithm. Currently, only CRC32
2043 and Adler-32 are supported. The checksum type can be changed on
2044 the fly via /proc/fs/lustre/osc/*/checksum_type.
2048 Description: use adler32 for page checksums
2049 Details : when available, use the Adler-32 algorithm instead of CRC32 for
2054 Description: build for x2 fails
2055 Details : fix compile issue on Cray systems.
2059 Description: Properly match for duplicate locks
2060 Details : Due to different lock order from skiplists code, we need to
2061 traverse entire list for now
2064 Frequency : only on PPC/SLES10
2066 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
2067 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
2068 be defined. Add a hack in mkfs_lustre.c to work around this compile
2073 Description: LASSERT on MDS when client holding flock lock dies
2074 Details : ldlm pool logic depends on number of granted locks equal to
2075 number of released locks which is not true for flock locks, so
2076 just exclude such locks from consideration.
2080 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
2081 Details : Use fsfilt_sync() for both the whole filesystem sync and
2082 individual file sync to eliminate dangerous inode locking
2083 with I_LOCK that can lead to a deadlock.
2087 Description: performance in 1.6.3
2088 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
2089 sure that 1MB requests are not fragmented by the block layer.
2091 Severity : enhancement
2093 Description: organize the server-side client stats on per-nid basis
2094 Details : Change the structure of stats under obdfilter and mds to
2102 The "uuid"s file would list the uuids of _active_ exports.
2103 And the clear entry is to clear all stats and stale nids.
2105 Severity : enhancement
2107 Description: eliminate client locks in face of contention
2108 Details : file contention detection and lockless i/o implementation
2109 for contended files.
2113 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
2116 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
2118 Description: MDS may lose file striping (and hence file data) in some cases
2119 Details : If there are additional extended attributes stored on the MDS,
2120 in particular ACLs, SELinux, or user attributes (if user_xattr
2121 is specified for the client mount options) then there is a risk
2122 of attribute loss. Additionally, the Lustre file striping
2123 needs to be larger than default (e.g. striped over all OSTs),
2124 and an additional attribute must be stored initially in the
2125 inode and then increase in size enough to be moved to the
2126 external attribute block (e.g. ACL growing in size) for file
2131 Description: skiplist implementation simplification
2132 Details : skiplists are used to group compatible locks on granted list
2133 that was implemented as tracking first and last lock of each
2134 lock group the patch changes that to using doubly linked lists
2138 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
2139 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
2140 because LASSERT on both the data supplied by a client, and the
2141 data on disk is dangerous and incorrect.
2143 Severity : enhancement
2145 Description: Slow truncate/writes to huge files at high offsets.
2146 Details : Directly associate cached pages to lock that protect those pages,
2147 this allows us to quickly find what pages to write and remove
2148 once lock callback is received.
2152 Description: more ldlm soft lockups
2153 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
2154 starve other threads from the resource lock for a long time in
2155 case of long waiting queue, so change the debug level from
2156 D_OTHER to the less frequently used D_INFO.
2158 Severity : enhancement
2160 Description: add -gid, -group, -uid, -user options to lfs find
2164 Description: Hung threads in invalidate_inode_pages2_range
2165 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
2166 once one is completed. As a result, some RPCs are stuck in the
2167 queue and are never sent.
2171 Description: filter threads hungs on waiting journal commit
2172 Details : Cleanup filter group llog code, then only filter group llog will
2173 be only created in the MDS/OST syncing process.
2177 Description: Procfs and llog threads access destoryed import sometimes.
2178 Details : Sync the import destoryed process with procfs and llog threads by
2179 the import refcount and semaphore.
2181 Severity : enhancement
2183 Description: openlock cache of b1_6 port to HEAD
2188 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
2189 Details : If insertion of an extent fails, then discard the inode
2190 preallocation and free data blocks else it can lead to duplicate
2195 Description: don't always update ctime in ext3_xattr_set_handle()
2196 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
2197 In some cases the ctime should not be updated, for example for
2198 2.0->1.8 compatibility it is necessary to delete an xattr and it
2199 should not update the ctime.
2203 Bugzilla : 15713/16362
2204 Description: Assertion in iopen_connect_dentry in 1.6.3
2205 Details : looking up an inode via iopen with the wrong generation number can
2206 populate the dcache with a disconneced dentry while the inode
2207 number is in the process of being reallocated. This causes an
2208 assertion failure in iopen since the inode's dentry list contains
2209 both a connected and disconnected dentry.
2213 Description: assertion failure in ldlm_handle2lock()
2214 Details : fix a race between class_handle_unhash() and class_handle2object()
2215 introduced in lustre 1.6.5 by bug 13622.
2220 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
2221 Details : remove the limit of 256 scsi disks in the sd_iostat patch
2226 Description: oops in sd_iostats_seq_show()
2227 Details : unloading/reloading the scsi low level driver triggers a kernel
2228 bug when trying to access the sd iostat file.
2233 Description: Kernel panics during QLogic driver reload
2234 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
2235 patch, causing memory corruption.
2240 Description: journal_dev option does not work in b1_6
2241 Details : pass mount option during pre-mount.
2243 Severity : enhancement
2245 Description: Add a FIEMAP(FIle Extent MAP) ioctl
2246 Details : FIEMAP ioctl will allow an application to efficiently fetch the
2247 extent information of a file. It can be used to map logical blocks
2248 in a file to physical blocks in the block device.
2252 Description: LDLM soft lockups - improvement
2253 Details : It is be possible to send the lock handle along with each read
2254 or write request because the client is already doing a lock match
2255 itself so there isn't any reason the OST should have to re-do that
2259 Frequency : only on Cray X2
2261 Description: X2 build failures
2262 Details : fix build failures on Cray X2.
2266 Description: xid & resent requests
2267 Details : Initialize RPC XID from clock at startup (randomly if clock is
2270 Severity : enhancement
2272 Description: Add lustre_start utility to start or stop multiple Lustre servers
2277 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
2278 Details : In case of memory pressure, list_del() can be called twice on
2279 req->rq_history_list, causing a kernel oops.
2283 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
2284 Details : fix stack overflow in the distributed lock manager by defering
2285 export eviction after a failed AST to the elt thread instead of
2286 handling it in the dlm interpret routine.
2290 Description: Convert some comments to new format.
2291 Details : Update documenting comments to match doxygen conventions.
2295 Description: Grammar fixes.
2296 Details : A couple of trivial spelling fixes.
2300 Description: OSD_COUNTERS-mandatory
2301 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
2302 to simplify the code.
2306 Description: simplify cmm_device freeing logic.
2307 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
2311 Description: Add lockdep support to dt_object_operations locking interface.
2312 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
2313 parameter indicating lock ordering. Update mdd code to use new
2318 Description: Introduce failloc constants for lockless IO tests.
2319 Details : Add two new failloc constants to test lockless IO. Only one of
2320 them in implemented---another is checked in yet to be landed
2325 Description: Add lockdep support for inode mutex.
2326 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
2327 in the situations where Lustre has to lock more than one inode
2332 Description: Add optional invariants checking support.
2333 Details : Add new LINVRNT() macro, optional on new --enable-invariants
2334 configure switch. This macro is to be used for consistency and
2335 sanity checks that are too expensive to be left in `production'
2340 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
2341 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
2346 Description: Add lockdep support for ldlm_lock and ldlm_resource.
2347 Details : Use spin_lock_nested() in (the only) situation where more than
2348 one ldlm_lock is locked simultaneously. Also, fix possible
2349 dead-lock in ldlm_lock_change_resource() by enforcing particular
2354 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
2355 Details : Instead of specifying each ldlm_lock call-back through separate
2356 parameter, wrap them into struct ldlm_callback_suite.
2360 Description: Kill join_lru obd method and its callers.
2361 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
2362 in memory---a requirement implemented through ->o_join_lru() obd
2363 method in HEAD. Remove this method and its users.
2367 Description: Add asynchronous ldlm ENQUEUE completion handler.
2368 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
2369 case that stock ldlm_completion_ast() cannot handle as it waits
2370 until lock is granted. Introduce new ldlm_completion_ast_async()
2371 for this. Also comment ldlm_completion_ast().
2375 Description: ldlm_error <-> errno conversion.
2376 Details : Add functions to map (rather arbitrary) between LDLM error codes
2377 and standard errno values. CLIO needs this to prevent LDLM specific
2378 constants from escaping ldlm and osc.
2382 Description: Kill unused ldlm_handle2lock_ns() function.
2383 Details : Kill unused ldlm_handle2lock_ns() function.
2387 Description: Add lu_ref support to ldlm_lock
2388 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
2389 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
2390 and ldlm_resource. LDLM interface has to be changed, because code that
2391 releases a reference on a lock, has to "know" what reference this is.
2392 In the most frequent case
2394 lock = ldlm_handle2lock(handle);
2396 LDLM_LOCK_PUT(lock);
2398 no changes are required. When any other reference (received _not_ from
2399 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
2400 instead of LDLM_LOCK_PUT().
2402 Arguably, changes are pervasive, and interface requires some discipline
2403 for proper use. On the other hand, it was very instrumental in finding
2404 a few leaked lock references.
2408 Description: Add ldlm_lock_addref_try().
2409 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
2410 attempts to addref a lock that might be being canceled
2415 Description: Add ldlm_weigh_callback().
2416 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
2417 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
2418 hard-coded `number of pages' logic.
2422 Description: Add lockdep annotations to llog code.
2423 Details : Use appropriately tagged _nested() locking calls in the places
2424 where llog takes more than one ->lgh_lock lock.
2428 Description: Add loi_kms_set().
2429 Details : Wrap kms updates into a helper function.
2433 Description: Constify instances of struct lsm_operations.
2434 Details : Constify instances of struct lsm_operations.
2438 Description: lu_conf support.
2439 Details : On a server, a file system object is uniquely identified
2440 by a fid, which is sufficient to locate and load all object
2441 state (inode). On a client, on the other hand, more data are
2442 necessary instantiate an object. Change lu_object_find() and
2443 friends to take additional `lu_conf' argument describing object.
2444 Typically this includes layout information.
2448 Description: lu_context fixes.
2449 Details : Introduce new lu_context functions that are needed on the client
2450 side, where some system threads (ptlrpcd) are shared by multiple
2451 modules, and so cannot be stopped during module shutdown.
2455 Description: Add start and stop methods to lu_device_type_operations.
2456 Details : Introduce two new methods in lu_device_type_operations, that are
2457 invoked when first instance of a given type is created and last one
2458 is destroyed respectively. This is need by CLIO.
2462 Description: Add lu_ref support to struct lu_device.
2463 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
2464 track leaked references.
2468 Description: Introduce lu_kmem_descr.
2469 Details : lu_kmem_descr and its companion interface allow to create
2470 and destroy a number of kmem caches at once.
2474 Description: Fix lu_object finalization race.
2475 Details : Fix a race between lu_object_find() finding an object and its
2476 concurrent finalization. This race is (most likely) not possible
2477 on the server, but might happen on the client.
2481 Description: Introduce lu_ref interface.
2482 Details : lu_ref is a debugging module allowing to track references to
2483 a given object. It is quite cpu expensive, and has to be
2484 explicitly enabled with --enable-lu_ref. See usage description
2489 Description: Factor lu_site procfs stats into a separate function.
2490 Details : Separate lu_site stats printing code into a separate function
2491 that can be reused on a client.
2495 Description: Constify instances of struct {lu,dt,md}_device_operations.
2496 Details : Constify instances of struct {lu,dt,md}_device_operations.
2500 Description: Introduce struct md_site and move meta-data specific parts of
2501 struct lu_site here.
2502 Details : Move md-specific fields out of struct lu_site into special struct
2503 md_site, so that lu_site can be used on a client.
2507 Description: Kill mdd_lov_destroy().
2508 Details : Remove unused mdd code.
2512 Description: Add st_block checking to multistat.c.
2513 Details : Add st_block checking to multistat.c.
2517 Description: Add lu_ref support to struct obd_device.
2518 Details : Add lu_ref tracking to obd_device.
2522 Description: Kill obd_set_fail_loc().
2523 Details : Remove unused code.
2527 Description: Add special type for ptlrpc_request interpret functions.
2528 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
2529 there. Actual usage will be in CLIO.
2533 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
2534 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
2535 doesn't work with lockdep.
2539 Description: Add rwv.c test program.
2540 Details : New testing program exercising readv(2) and writev(2) (Qian).
2544 Description: Add sendfile.c test program.
2545 Details : New testing program exercising sendfile(2) (Jay).
2549 Description: Ratelimit a message that can be very frequent.
2550 Details : Ratelimit a memory allocation failure message that can
2555 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2556 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2560 Description: Liblustre build fixes.
2561 Details : Liblustre build fixes.
2565 Description: libcfs: add cfs_{need,cond}_resched() interface.
2566 Details : libcfs: add cfs_{need,cond}_resched() definition and
2567 implementations for Linux, NT, and liblustre.
2569 Severity : enhancement
2571 Description: More exported tunables for mballoc
2572 Details : Add support for tunable preallocation window and new tunables for
2573 large/small requests
2577 Description: Detect corruption of block bitmap and checking for preallocations
2578 Details : Checks validity of on-disk block bitmap. Also it does better
2579 checking of number of applied preallocations. When corruption is
2580 found, it turns filesystem readonly to prevent further corruptions.
2584 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
2585 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
2586 window to 0. If it does hit cache miss, and read-pattern is still
2587 stride-io mode, does not reset the stride window, but also does not
2588 increase the stride window length in this case.
2592 Frequency : only for big-endian servers
2593 Description: Check if system is big-endian while mounting fs with extents feature
2594 Details : Mounting a filesystem with extents feature will fail on big-endian
2595 systems since ext3-based ldiskfs is not supported on big-endian
2596 systems. This can be over-riden with "bigendian_extents" mount option.
2598 Severity : enhancement
2600 Description: The root squash functionality
2601 Details : A security feature, which is to prevent users from being able
2602 to mount lustre on their desktop, run as root, and delete
2603 all of the files in the filesystem. The goal is accomplished by
2604 remapping user id (UID) and group id (GID) of the root user to
2605 a UID and GID specified by the system administartor via Lustre
2606 configuration management server (MGS). The functionality also
2607 allows to specify sets of clients for which the remapping does
2612 Description: Excessive recovery window
2613 Details : With AT enabled, the recovery window can be excessively long (6000+
2614 seconds). To address this problem, we no longer use
2615 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
2616 timeout no longer depends on the service time, it is set to
2617 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
2618 time via pb_service_time.
2622 Description: Watchdog triggered on MDS failover
2623 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
2624 the OSTs know that the MDS "UUID" can be reused for the same export
2625 from a different NID, so we do not need to wait for the export to be
2629 Frequency : rare, only if using MMP with Linux RAID
2631 Description: MMP doesn't work with Linux RAID
2632 Details : While using HA for Lustre servers with Linux RAID, it is possible
2633 that MMP will not detect multiple mounts. To make this work we
2634 need to unplug the device queue in RAID when the MMP block is being
2635 written. Also while reading the MMP block, we should read it from
2636 disk and not the cached one.
2638 Severity : enhancement
2640 Description: open file using fid
2641 Details : A file can be opened using just its fid, like
2642 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
2645 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
2648 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
2649 Details : add proper locking for creation of jbd_4k slab cache
2653 Description: MMP check in ext3_remount() fails without displaying any error
2654 Details : When multiple mount protection fails during remount, proper error
2657 Severity : enhancement
2659 Description: Allow stripe size to be up to 4G-64k
2660 Details : Fix math logic to allow large stripe sizes.
2664 Description: add check for >8TB ldiskfs filesystems
2665 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
2666 Don't allow >8TB ldiskfs filesystems to be mounted without
2667 force_over_8tb mount option
2672 Descriptoin: OSS complains "too many blocks" and crashes
2673 Details : In filter_commitrw_write():
2674 1. lquota_chkquota();
2675 2. obdo_from_inode();
2676 3. lquota_pending_commit();
2677 "2." has changed uid and gid in this case, then caused
2678 lquota_chkquota() and lquota_pending_commit() to use
2679 different uid and gid. That is the root of the bug.
2683 Descriptoin: getstripe doesn't report stripe count = -1 sanely
2684 Details : Detect and print out the 'stripe count = -1' case properly.
2688 Description: Out or order replies might be lost on replay
2689 Details : In ptlrpc_retain_replayable_request if we cannot find retained
2690 request with tid smaller then one currently being added, add it
2691 to the start, not end of the list.
2693 --------------------------------------------------------------------------------
2695 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
2697 * Support for kernels:
2698 2.6.5-7.283 (SLES 9),
2699 2.6.9-55.EL (RHEL 4),
2700 2.6.16.46-0.14 (SLES 10),
2701 2.6.18.8 vanilla (kernel.org)
2702 * Client support for unpatched kernels:
2703 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2704 2.6.16 - 2.6.22 vanilla (kernel.org)
2705 * Due to recently discovered recovery problems, we do not recommend
2706 using patchless RHEL 4 clients with this or any earlier release.
2707 * Recommended e2fsprogs version: 1.39.cfs8
2708 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2709 * Starting with this release, the ldiskfs backing filesystem required
2710 by Lustre is now in its own package, lustre-ldiskfs. This package
2711 should be installed. It is versioned separately from Lustre and
2712 may be released separately in future.
2716 Description: block reactivating mgc import until all deactivates complete
2717 Details : Fix race when failing back MDT/MGS to itself (testing)
2720 Frequency : at statup only
2722 Description: mds_lov_synchronize race leads to various problems
2723 Details : simultaneous MDT->OST connections at startup can cause the
2724 sync to abort, leaving the OSC in a bad state.
2726 Severity : enhancement
2728 Description: add optional extra BUILD_VERSION info
2729 Details : add a new environment variable (namely LUSTRE_VERS) which allows
2730 to override the lustre version.
2733 Frequency : 2.6.18 servers only
2735 Description: ll_kern_mount() doesn't release the module reference
2736 Details : The ldiskfs module reference count never drops down to 0
2737 because ll_kern_mount() doesn't release the module reference.
2742 Description: server LBUG when using old ost_num_threads parameter
2743 Details : Accept the old ost_num_threads parameter but warn that it
2744 is deprecated, and fix an off-by-one error that caused an LBUG.
2749 Description: Transient SCSI error results in persistent IO issue
2750 Details : iobuf->dr_error is not reinitialized to 0 between two
2754 Frequency : sometimes when underlying device returns I/O errors
2756 Description: OSTs not going read-only during write failures
2757 Details : OSTs are not remounted read-only when the journal commit threads
2758 get I/O errors because fsfilt_ext3 calls journal_start/stop()
2759 instead of the ext3 wrappers.
2763 Description: poor connect scaling with increasing client count
2764 Details : Don't run filter_grant_sanity_check for more than 100 exports
2765 to improve scaling for large numbers of clients.
2768 Frequency : SLES10 only
2770 Description: sanity-quota.sh quotacheck failed: rc = -22
2771 Details : Quotas cannot be enabled on SLES10.
2774 Frequency : liblustre clients only
2776 Description: getdirentries does not give error when run on compute nodes
2777 Details : getdirentries does not fail when the size specified as an argument
2778 is too small to contain at least one entry
2780 Severity : enhancement
2782 Description: Add LNET router traceability for debug purposes
2783 Details : If a checksum failure occurs with a router as part of the
2784 IO path, the NID of the last router that forwarded the bulk data
2785 is printed so it can be identified.
2790 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
2791 Details : Due to a race condition, liblustre clients were occasionally
2792 evicted incorrectly.
2794 Severity : enhancement
2796 Description: lfs setstripe use optional parameters instead of postional
2799 Severity : enhancement
2801 Description: Nanosecond timestamp support for ldiskfs
2802 Details : The on-disk ldiskfs filesystem has added support for nanosecond
2803 resolution timestamps. There is not yet support for this at
2804 the Lustre filesystem level.
2807 Frequency : during server recovery
2809 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
2810 Details : request with rq_no_resend flag not awake l_wait_event if they get
2814 Frequency : nfs export on patchless client
2816 Description: connectathon hang when test nfs export over patchless client
2817 Details : Disconnected dentry cannot be found with lookup, so we do not need
2818 to unhash it or make it invalid
2821 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
2822 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
2823 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
2828 Description: open req refcounting wrong on reconnect
2829 Details : If reconnect happened between getting open reply from server and
2830 call to mdc_set_replay_data in ll_file_open, we will schedule
2831 replay for unreferenced request that we are about to free.
2832 Subsequent close will crash in variety of ways.
2833 Check that request is still eligible for replay in
2834 mdc_set_replay_data().
2839 Description: disable writes to filesystem when reading health_check file
2840 Details : the default for reading the health_check proc file has changed
2841 to NOT do a journal transaction and write to disk, because this
2842 can cause reads of the /proc file to hang and block HA state
2843 checking on a healthy but otherwise heavily loaded system. It
2844 is possible to return to the previous behaviour during configure
2845 with --enable-health-write.
2847 Severity : enhancement
2849 Description: 64-bit inode version
2850 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
2851 to the inode. This will be required for version-based recovery.
2856 Description: MDS fails to start if a duplicate client export is detected
2857 Details : in some rare cases it was possible for a client to connect to
2858 an MDS multiple times. Upon recovery the MDS would detect this
2859 and fail during startup. Handle this more gracefully.
2861 Severity : enhancement
2863 Description: Add -o localflock option to simulate old noflock
2865 Details : This will achieve local-only flock/fcntl locks
2871 Description: log_commit_thread vs filter_destroy race leads to crash
2872 Details : Take import reference before releasing llog record semaphore
2877 Description: Wrong request locking in request set processing
2878 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
2879 imp_delayed_list, in this place should be used imp_lock.
2882 Frequency : when reconnection
2884 Description: Grant Leak when osc reconnect to OST
2885 Details : When osc reconnect ost, OST(filter) should check whether it
2886 should grant more space to client by comparing fed_grant and
2887 cl_avail_grant, and return the granted space to client instead
2888 of "new granted" space, because client will call osc_init_grant
2889 to update the client grant space info.
2892 Frequency : when client reconnect to OST
2894 Description: Grant Leak when osc do resend and replay bulk write
2895 Details : When osc reconnect to OST, OST(filter)should clear grant info of
2896 bulk write request, because the grant info while be sync between
2897 OSC and OST when reconnect, and we should ignore the grant info
2898 these of resend/replay write req.
2903 Description: Grant space more than avaiable left space sometimes.
2904 Details : When then OST is about to be full, if two bulk writing from
2905 different clients came to OST. Accord the avaliable space of the
2906 OST, the first req should be permitted, and the second one
2907 should be denied by ENOSPC. But if the seconde arrived before
2908 the first one is commited. The OST might wrongly permit second
2909 writing, which will cause grant space > avaiable space.
2912 Frequency : when client is evicted
2914 Description: Grant might be wrongly erased when osc is evicted by OST
2915 Details : when the import is evicted by server, it will fork another
2916 thread ptlrpc_invalidate_import_thread to invalidate the
2917 import, where the grant will be set to 0. While the original
2918 thread will update the grant it got when connecting. So if
2919 the former happened latter, the grant will be wrongly errased
2920 because of this race.
2925 Description: Checking Stale with correct fid
2926 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
2927 to check whether it is stale, because sometimes, we want the
2928 enqueue happened anyway, and op_data.fid2 will not be initialized.
2930 Severity : enhancement
2932 Description: update patchless client
2933 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
2936 Frequency : only with 2.4 kernel
2938 Description: random memory corruption
2939 Details : size of struct ll_inode_info is to big for union inode.u and this
2940 can be cause of random memory corruption.
2945 Description: Memory leak in recovery
2946 Details : Lov_mds_md was not free in an error handler in mds_create_object.
2947 It should also check obd_fail before fsfilt_start, otherwise if
2948 fsfilt_start return -EROFS,(failover mds during mds recovery).
2949 then the req will return with repmsg->transno = 0 and rc = EROFS.
2950 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
2951 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
2952 be freed no matter whether fsfilt_commit success or not.
2955 Frequency : only with huge count clients
2957 Description: Prevents from taking the superblock lock in llap_from_page for
2959 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
2960 for llap_from_page call in ll_removepage prevents from taking the
2961 superblock lock for a soon died page.
2966 Description: Not check open intent error before release open handle
2967 Details : in some rare cases, the open intent error is not checked before
2968 release open handle, which may cause
2969 ASSERTION(open_req->rq_transno != 0), because it tries to release
2970 the failed open handle.
2975 Description: Set cat log bitmap only after create log success.
2976 Details : in some rare cases, the cat log bitmap is set too early. and it
2977 should be set only after create log success.
2981 Description: Accessing a block bevice can re-enable I/O when Lustre is
2982 tearing down a device.
2983 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
2988 Description: service threads may hog cpus when there are a lot of requests
2989 Details : Insert cond_resched to give other threads a chance to use some CPU
2994 Description: the cat log was not initialized in recovery
2995 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
2996 unlink log on mds will not be initialized until mds post
2997 recovery. And also in mds post recovery, the unlink log will
2998 initialization will be done asynchronausly, so there will be race
2999 between add unlink log and unlink log initialization.
3003 Description: brw_stats were being printed incorrectly
3004 Details : brw_stats were being printed as log2 but all of them were not
3005 recorded as log2. Also remove some code duplication arising from
3006 filter_tally_{read,write}.
3010 Frequency : rare, only in recovery.
3011 Description: ASSERTION(req->rq_type != LI_POISON) failed
3012 Details : imp_lock should be held while iterating over imp_sending_list for
3013 prevent destroy request after get timeout in ptlrpc_queue_wait.
3017 Description: replay-single.sh test 52 fails
3018 Details : A lock's skiplist need to be cleanup when it being unlinked
3019 from its resource list.
3023 Description: Short directio read returns full requested size rather than
3025 Details : Direct I/O operations should return actual amount of bytes
3026 transferred rather than requested size.
3028 Severity : enhancement
3030 Description: metadata RPC reduction (e.g. for rm performance)
3031 Details : decrease the amount of synchronous RPC between clients and servers
3032 by canceling conflicing lock before the operation on the client
3033 and packing thier handles into the main operation RPC to server.
3035 Severity : enhancement
3037 Description: Async OSC create to avoid the blocking unnecessarily.
3038 Details : If a OST has no remain object, system will block on the creating
3039 when need to create a new object on this OST. Now, ways use
3040 pre-created objects when available, instead of blocking on an
3041 empty osc while others are not empty. If we must block, we block
3042 for the shortest possible period of time.
3046 Description: improve handling recoverable errors
3047 Details : if request processig with error which can be recoverable on server
3048 request should be resend, otherwise page released from cache and
3051 Severity : enhancement
3053 Description: refine locking for avoid write wrong info into lov_objid file
3054 Details : fix possible races with add new target and write/update data in
3057 --------------------------------------------------------------------------------
3059 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
3064 Frequency : on some architectures
3066 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
3067 Details : Uninitialized flags sometimes cause configuration commands to
3071 Frequency : patchless clients only
3073 Description: missing __iget() symbol export
3074 Details : The __iget() symbol export is missing. To avoid the need for
3075 this on patchless clients the deathrow inode reaper is turned
3076 off, and we depend on the VM to clean up old inodes. This
3077 dependency was during via the fix for bug 12181.
3079 --------------------------------------------------------------------------------
3081 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
3083 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
3084 INTEROPERATE with older versions automatically. In many cases a
3085 special upgrade step is needed. Please read the
3086 user documentation before upgrading any part of a live system.
3087 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
3088 version will not interoperate with 1.6 betas before beta5 (1.5.95).
3089 * WARNING: Lustre configuration and startup changes are required with
3090 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
3094 Severity : enhancement
3096 Description: MountConf
3097 Details : Lustre configuration is now managed via mkfs and mount
3098 commands instead of lmc and lconf. New obd types (MGS, MGC)
3099 are added for dynamic configuration management. See
3100 https://mail.clusterfs.com/wikis/lustre/MountConf for
3103 Severity : enhancement
3105 Description: dynamic OST addition
3106 Details : OSTs can now be added to a live filesystem
3108 Severity : enhancement
3110 Description: startup order invariance
3111 Details : MDTs and OSTs can be started in any order. Clients only
3112 require the MDT to complete startup.
3114 Severity : enhancement
3116 Description: parallel, asynchronous orphan cleanup
3117 Details : orphan cleanup is now performed in separate threads for each
3118 OST, allowing parallel non-blocking operation.
3120 Severity : enhancement
3122 Description: optimized stripe assignment
3123 Details : stripe assignments are now made based on ost space available,
3124 ost previous usage, and OSS previous usage, in order to try
3125 to optimize storage space and networking resources.
3127 Severity : enhancement
3129 Description: Permanently set tunables
3130 Details : All writable /proc/fs/lustre tunables can now be permanently
3131 set on a per-server basis, at mkfs time or on a live system.
3133 Severity : enhancement
3135 Description: Lustre message v2
3136 Details : Add lustre message format v2.
3138 Severity : enhancement
3140 Description: client OST exclusion list
3141 Details : Clients can be started with a list of OSTs that should be
3142 declared "inactive" for known non-responsive OSTs.
3146 Description: SPEC SFS validation failure on NFS v2 over lustre.
3147 Details : Changes the blocksize for regular files to be 2x RPC size,
3148 and not depend on stripe size.
3150 Severity : enhancement
3152 Description: Multiple MD RPCs in flight.
3153 Details : Further unserialise some read-only MDS RPCs - learn about intents.
3154 To avoid overly-overloading MDS, introduce a limit on number of
3155 MDS RPCs in flight for a single client and add /proc controls
3156 to adjust this limit.
3158 Severity : enhancement
3160 Description: client read/write statistics
3161 Details : Add client read/write call usage stats for performance
3162 analysis of user processes.
3163 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
3164 file access. extents_stats shows chunk size distribution.
3165 extents_stats_per_process show chunk size distribution per
3168 Severity : enhancement
3170 Description: mds statistics
3171 Details : Add detailed mds operations statistics in
3172 /proc/fs/lustre/mds/*/stats.
3176 Description: Failure of copying files with lustre special EAs.
3177 Details : Client side always return success for setxattr call for lustre
3178 special xattr (currently only "trusted.lov").
3183 Description: Refcount LNET uuids
3184 Details : The global LNET uuid list grew linearly with every startup;
3185 refcount repeated list entries instead of always adding to
3188 Severity : enhancement
3190 Description: Dynamic service threads
3191 Details : Within a small range, start extra service threads
3192 automatically when the request queue builds up.
3195 Frequency : mixed-endian client/server environments
3197 Description: mixed-endian crashes
3198 Details : The new msg_v2 system had some failures in mixed-endian
3201 Severity : enhancement
3203 Description: Easy OST removal
3204 Details : OSTs can be permanently deactivated with e.g. 'lctl
3205 conf_param lustre-OST0001.osc.active=0'
3207 Severity : enhancement
3209 Description: MGS proc entries
3210 Details : Added basic proc entries for the MGS showing what filesystems
3213 Severity : enhancement
3215 Description: provide MGS failover
3216 Details : Added config lock reacquisition after MGS server failover.
3218 Severity : enhancement
3220 Description: add Linux 2.4 support
3221 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
3225 Description: a large application tries to do I/O to the same resource and dies
3226 in the middle of it.
3227 Details : Check the req->rq_arrival time after the call to
3228 ost_brw_lock_get(), but before we do anything about
3229 processing it & sending the BULK transfer request. This
3230 should help move old stale pending locks off the queue as
3231 quickly as obd_timeout.
3234 Frequency : when an incorrect nid is specified during startup
3236 Description: ptlrpc connect to non-existant node causes kernel crash
3237 Details : LNET can't be re-entered from an event callback, which
3238 happened when we expire a message after the export has been
3239 cleaned up. Instead, hand the zombie cleanup off to another
3242 Severity : enhancement
3244 Description: plain/inodebits lock performance improvement
3245 Details : Grouping plain/inodebits in granted list by their request modes
3246 and bits policy, thus improving the performance of search through
3250 Frequency : only if OST filesystem is corrupted
3252 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3253 Details : for a short time RPCs with bulk IO are in the replay list,
3254 but replay of bulk IOs is unimplemented. If the OST filesystem
3255 is corrupted due to disk cache incoherency and then replay is
3256 started it is possible to trip an assertion. Avoid putting
3257 committed RPCs into the replay list at all to avoid this issue.
3260 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3262 Description: System hang on startup
3263 Details : This bug allowed the liblustre (e.g. catamount) client to
3264 return to the app before handling all startup RPCs. This
3265 could leave the node unresponsive to lustre network traffic
3266 and manifested as a server ptllnd timeout.
3268 Severity : enhancement
3270 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3271 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
3272 boolean to control whether to print peer debug info when a
3273 client's RPC times out.
3276 Frequency : only for kernels with patches from Lustre below 1.4.3
3278 Description: Remove old rdonly API
3279 Details : Remove old rdonly API which unused from at least lustre 1.4.3
3282 Frequency : only for devices with external journals
3284 Description: Set external device read-only also
3285 Details : During a commanded failover stop, we set the disk device
3286 read-only while the server shuts down. We now also set any
3287 external journal device read-only at the same time.
3290 Frequency : when upgrading from 1.4 while trying to change parameters
3292 Description: The wrong (new) MDC name was used when setting parameters for
3293 upgraded MDT's. Also allows changing of OSC (and MDC)
3294 parameters if --writeconf is specified at tunefs upgrade time.
3297 Frequency : when setting specific ost indicies
3299 Description: QOS code breaks on skipped indicies
3300 Details : Add checks for missing OST indicies in the QOS code, so OSTs
3301 created with --index need not be sequential.
3303 Severity : enhancement
3305 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3306 Details : The uninit_groups feature works in conjunction with the kernel
3307 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3308 up the pass1 processing of e2fsck. This is a read-only feature
3309 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3310 mount filesystems that have had this feature enabled.
3312 Severity : enhancement
3314 Description: Improve multi-block allocation algorithm to avoid fragmentation
3315 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3316 allocation locality and avoid filesystem fragmentation.
3318 ------------------------------------------------------------------------------
3320 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
3322 * Support for kernels:
3323 2.4.21-47.0.1.EL (RHEL 3)
3324 2.6.5-7.283 (SLES 9)
3325 2.6.9-42.0.10.EL (RHEL 4)
3326 2.6.12.6 vanilla (kernel.org)
3327 2.6.16.27-0.9 (SLES 10)
3328 * Recommended e2fsprogs version: 1.39.cfs5
3330 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3334 Frequency : occasional, depends on client load and configuration
3335 Bugzilla : 12181, 12203
3336 Description: data loss for recently-modified files
3338 Details : In some cases it is possible that recently written or created
3339 files may not be written to disk in a timely manner (this should
3340 normally be within 30s unless client IO load is very high).
3341 The problem appears as zero-length files or files that are a
3342 multiple of 1MB in size after a client crash or client eviction
3343 that are missing data at the end of the file.
3345 This problem is more likely to be hit on clients where files are
3346 repeatedly created and unlinked in the same directory, clients
3347 have a large amount of RAM, have many CPUs, the filesystem has
3348 many OSTs, the clients are rebooted frequently, and/or the files
3349 are not accessed by other nodes after being written.
3351 The presence of the problem can be detected by looking at
3352 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
3353 smaller than the second (nr_unused) then dirty files will not
3354 be flushed automatically to disk. "sync; sleep 10" should be
3355 run several times on the node before unmounting it to update
3356 Lustre (this is also safe to run on nodes without this problem).
3358 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
3359 that can cause this same problem, so customers using that kernel
3360 also need to update the kernel in addition to Lustre. In order
3361 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
3363 It is normal that files written just before a client crash (less
3364 than 30s) may not yet have been flushed to disk, even for local
3368 Frequency : frequent on thin XT3 nodes
3370 Description: UUID collision on thin XT3 Linux nodes
3371 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
3372 randomly, since we relied on an insufficiently-seeded PRNG.
3377 Description: OSS hangs after "All ost request buffers busy"
3378 Details : A deadlock between quota and journal operations caused OSS
3379 hangs after printing "All ost request buffers busy."
3382 Frequency : always on liblustre builds
3384 Description: Cleanup compiler warnings on liblustre
3387 Frequency : always on liblustre builds on XT3
3389 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
3393 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
3394 > 32000 subdirectories
3395 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
3396 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
3397 subdirectory count crosses 32000. This will aid e2fsck to
3398 correctly handle more than 32000 subdirectories.
3401 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3403 Description: System hang on startup
3404 Details : This bug allowed the liblustre (e.g. catamount) client to
3405 return to the app before handling all startup RPCs. This
3406 could leave the node unresponsive to lustre network traffic
3407 and manifested as a server ptllnd timeout.
3409 Severity : enhancement
3411 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3412 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
3413 boolean to control whether to print peer debug info when a
3414 client's RPC times out.
3419 Description: make O_SYNC working on 2.6 kernels
3420 Details : 2.6 kernels use different method for mark pages for write,
3421 so need add a code to lustre for O_SYNC work.
3426 Description: Failure to close file and release space on NFS
3427 Details : Put inode details into lock acquired in ll_intent_file_open.
3428 Use mdc_intent_lock in ll_intent_open to properly
3429 detect all kind of errors unhandled by mdc_enqueue.
3434 Description: proc file read during shutdown sometimes raced obd removal,
3436 Details : Add lock to prevent obd access after proc file removal.
3439 Frequency : Only for files larger than 4GB on 32-bit clients.
3441 Description: improperly doing page alignment of locks
3442 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
3443 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3448 Description: RPCs being resent when they shouldn't be
3449 Details : Some RPCs that should not be resent are being resent. This
3450 can cause inconsistencies in the RPC state machine. Do not
3451 resend such requests.
3454 Frequency : rare, only with NFS export
3456 Description: Crash on NFS re-export node
3457 Details : under very unusual load conditions an assertion is hit in
3458 ll_intent_file_open()
3461 Frequency : only if OST filesystem is corrupted
3463 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3464 Details : for a short time RPCs with bulk IO are in the replay list,
3465 but replay of bulk IOs is unimplemented. If the OST filesystem
3466 is corrupted due to disk cache incoherency and then replay is
3467 started it is possible to trip an assertion. Avoid putting
3468 committed RPCs into the replay list at all to avoid this issue.
3473 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3474 Details : Large single O_DIRECT read and write calls can fail to allocate
3475 a sufficiently large buffer to process the request. In case of
3476 allocation failure the allocation is retried with a smaller
3477 buffer and broken into smaller requests.
3479 Severity : enhancement
3481 Description: Add -o localflock option to simulate old noflock behaviour.
3482 Details : This will achieve local-only flock/fcntl locks coherentness.
3487 Description: versioning check is incomplete
3488 Details : Checking the version difference of client vs. server, report
3489 error if the gap is too big.
3494 Description: add support PG_writeback bit
3495 Details : add support for PG_writeback bit for Lustre, for more carefull
3496 work with page cache in 2.6 kernel. This also fix some deadlocks
3497 and remove hack for work O_SYNC with 2.6 kernel.
3499 Severity : enhancement
3501 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3502 Details : The uninit_groups feature works in conjunction with the kernel
3503 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3504 up the pass1 processing of e2fsck. This is a read-only feature
3505 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3506 mount filesystems that have had this feature enabled.
3508 Severity : enhancement
3510 Description: Improve multi-block allocation algorithm to avoid fragmentation
3511 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3512 allocation locality and avoid filesystem fragmentation.
3514 ------------------------------------------------------------------------------
3516 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3518 * Support for kernels:
3519 2.6.9-42.0.3.EL (RHEL 4)
3520 2.6.5-7.276 (SLES 9)
3521 2.4.21-47.0.1.EL (RHEL 3)
3522 2.6.12.6 vanilla (kernel.org)
3523 2.6.16.21-0.8 (SLES10)
3524 * Recommended e2fsprogs version: 1.39.cfs2-0
3526 * The backwards-compatible /proc/sys/portals symlink has been removed
3527 in this release. Before upgrading, please ensure that you change
3528 any configuration scripts or /etc/sysctl.conf files that access
3529 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3530 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3531 in advance of the upgrade on any system running Lustre 1.4.6 or
3532 newer, since /proc/sys/lnet was added in that version.
3533 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3537 Frequency : only when quota is used
3539 Description: avoid scanning export list for quota master
3540 Details : Change the algorithms to avoid scanning export list in order
3541 to improve the efficiency.
3544 Frequency : MDS failover only, very rarely
3546 Description: "went back in time" messages on mds failover
3547 Details : The greatest transno may be lost when the current operation
3548 finishes with an error (transno==0) and the client's last_rcvd
3549 record is over-written. Save the greatest transno in the
3550 mds_last_transno for this case.
3553 Frequency : always for specific kernels and striping counts
3555 Description: client may get "Matching packet too big" without ACL support
3556 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
3557 when trying to access files in certain configurations. The
3558 clients should in fact be denied when mounting because they do
3559 not understand ACLs.
3562 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
3564 Description: many clients connecting with IO in progress causes connect timeouts
3565 Details : Avoid synchronous journal commits to avoid delays caused by many
3566 clients connecting/disconnecting when bulk IO is in progress.
3567 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
3568 OST_IO_PORTAL to avoid delays behind potentially many pending
3572 Frequency : occasionally with multiple writers to a single file
3574 Description: shared writes to file may result in wrong size reported by stat()
3575 Details : Allow growing of kms when extent lock is cancelled
3578 Frequency : always with random mmap IO to multi-striped file
3580 Description: mmap write might be lost if we are writing to a 'hole' in stripe
3581 Details : Only if the hole is at the end of OST object so that kms is too
3582 small. Fix is to increase kms accordingly in ll_nopage.
3585 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
3587 Description: writes to a missing object would leak memory on the OST
3588 Details : If there is an inconsistency between the MDS and OST filesystems,
3589 such that the MDS references an object that doesn't exist, writes
3590 to that object will leak memory due to incorrect cleanup in the
3591 error handling path, eventually running out of memory on the OST.
3596 Description: Creating too long symlink causes lustre errors
3597 Details : Check symlink and name lengths before sending requests to MDS.
3600 Frequency : only if flock is enabled (not on by default)
3602 Description: posix locks not released on fd closure on 2.6.9+
3603 Details : We failed to add posix locks to list of inode locks on 2.6.9+
3604 kernels, this caused such locks not to be released on fd close and
3605 then assertions on fs unmount about still used locks.
3608 Frequency : MDS failover only, very rarely
3610 Description: clients may get ASSERTION(granted_lock != NULL)
3611 Details : When request was taking a long time, and a client was resending
3612 a getattr by name lock request. The were multiple lock requests
3613 with the same client lock handle and
3614 mds_getattr_name->fixup_handle_for_resent_request found one of the
3615 lock handles but later failed with ASSERTION(granted_lock != NULL).
3620 Description: handle->h_buffer_credits > 0, assertion failure
3621 Details : h_buffer_credits is zero after truncate, causing assertion
3622 failure. This patch extends the transaction or creates a new
3626 Frequency : NFS re-export or patchless client
3627 Bugzilla : 11179, 10796
3628 Description: Crash on NFS re-export node (__d_move)
3629 Details : We do not want to hash the dentry if we don't have a lock.
3630 But if this dentry is later used in d_move, we'd hit uninitialised
3631 list head d_hash, so we just do this to init d_hash field but
3632 leave dentry unhashed.
3635 Frequency : NFS re-export or patchless client
3637 Description: NFS exports has problem with symbolic link
3638 Details : lustre client didn't properly install dentry when re-exported
3639 to NFS or running patchless client.
3642 Frequency : NFS re-export or patchless client
3644 Description: Various nfs/patchless fixes.
3645 Details : fixes reuse disconected alias for lookup process - this fixes
3646 warning "find_exported_dentry: npd != pd",
3647 fix permission error with open files at nfs.
3648 fix apply umask when do revalidate.
3651 Frequency : occasional
3653 Description: Crash on NFS re-export node
3654 Details : calling clear_page() on the wrong pointer triggered oops in
3655 generic_mapping_read().
3658 Frequency : rarely, using O_DIRECT IO
3660 Description: unaligned directio crashes client with LASSERT
3661 Details : check for unaligned buffers before trying any requests.
3664 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
3666 Description: stale data returned from RAID cache
3667 Details : If only a small amount of IO is done to the RAID device before
3668 reading it again it is possible to get stale data from the RAID
3669 cache instead of reading it from disk.
3672 Frequency : always for sles10 kernel
3674 Description: sles10 support
3675 Details : ll_follow_link: compile fixes and using of nd_set_link
3676 under newer kernels.
3679 Frequency : depends on arch, kernel and compiler version, always on sles10
3682 Description: recursive or deep enough symlinks cause stack overflow
3683 Details : getting rid of large stack-allocated variable in
3687 Frequency : depends on hardware
3689 Description: lustre write performance loss in the SLES10 kernel
3690 Details : the performance loss is caused by using of write barriers in the
3691 ext3 code. The SLES10 kernel turns barrier support on by
3692 default. The fix is to undo that change for ldiskfs.
3694 ------------------------------------------------------------------------------
3696 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
3698 * Support for kernels:
3699 2.6.9-42.0.3EL (RHEL 4)
3700 2.6.5-7.276 (SLES 9)
3701 2.4.21-47.0.1.EL (RHEL 3)
3702 2.6.12.6 vanilla (kernel.org)
3706 Frequency : quota enabled and large files being deleted
3708 Description: releasing more than 4GB of quota at once hangs OST
3709 Details : If a user deletes more than 4GB of files on a single OST it
3710 will cause the OST to spin in an infinite loop. Release
3711 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3716 Description: statfs data retrieved from /proc may be stale or zero
3717 Details : When reading per-device statfs data from /proc, in the
3718 {kbytes,files}_{total,free,avail} files, it may appear
3719 as zero or be out of date.
3722 Frequency : systems with MD RAID1 external journal devices
3724 Description: lconf's call to blkid is confused by RAID1 journal devices
3725 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3726 of returning all block devices that match the journal UUID.
3729 Frequency : always, for aggregate stripe size over 4GB
3731 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
3732 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
3733 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
3736 Frequency : always if "lfs find" used on a local file/directory
3738 Description: "lfs find" segfaults if used on a local file/directory
3739 Details : The case where a directory component was not specified wasn't
3740 handled correctly. Handle this properly.
3743 Frequency : always on ppc64
3745 Description: the write to an ext3 filesystem mounted with mballoc got stuck
3746 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3747 perform endianness conversion.
3750 Frequency : rarely (truncate to non-zero file size after write under load)
3751 Bugzilla : 10730, 10687
3752 Description: Files padded with zeros to next 4K multiple
3753 Details : With filesystems mounted using the "extents" option (2.6 kernels)
3754 it is possible that files that are truncated to a non-zero size
3755 immediately after being written are filled with zero bytes beyond
3756 the truncated size. No file data is lost.
3758 Severity : enhancement
3760 Description: Allow recovery/failover for liblustre clients.
3761 Details : liblustre clients were unaware of failover configurations until
3764 Severity : enhancement
3766 Description: user file locks should fail when not mounting with flock option
3767 Details : Set up an error-returning stub in ll_file_operations.lock field
3768 to prevent incorrect behaviour when client is mounted without
3769 flock option. Also, set up properly f_op->flock field for
3773 Frequency : always on ia64
3775 Description: "lfs df" loops on printing out MDS statfs information
3776 Details : The obd_ioctl_data was not initialized and in some systems
3777 this caused a failure during the ioctl that did not return
3778 an error. Initialize the struct and return an error on failure.
3781 Frequency : SLES 9 only
3783 Description: Error of copying files with lustre special EAs as root
3784 Details : Client side always return success for setxattr call for lustre
3785 special xattr (currently only "trusted.lov").
3788 Frequency : rarely on clusters with both ia64+i386 clients
3790 Description: ia64+i686 clients doing shared IO on the same file may LBUG
3791 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
3792 clients are doing concurrent writes to the same file it is
3793 possible that the ia64 clients may LASSERT because the OST
3794 extent locks are not PAGE_SIZE aligned. Ensure that grown
3795 locks are always aligned on the request boundary.
3798 Frequency : specific use, occasional
3800 Description: Overwriting in use executable truncates on-disk binary image
3801 Details : If one node attempts to overwrite an executable in use by
3802 another node, we now correctly return ETXTBSY instead of
3803 truncating the file.
3805 Severity : enhancement
3807 Description: Async OSC create to avoid the blocking unnecessarily.
3808 Details : If a OST has no remain object, system will block on the creating
3809 when need to create a new object on this OST. Now, ways use
3810 pre-created objects when available, instead of blocking on an
3811 empty osc while others are not empty. If we must block, we block
3812 for the shortest possible period of time.
3817 Description: chmod on Lustre root is propagated to other clients
3818 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
3819 invalid by the follow_mount time.
3822 Frequency : liblustre clients only
3824 Description: Race in 'instant cancel' lock handling could lead to such locks
3825 never to be granted in case of SMP MDS
3826 Details : Do not destroy not yet granted but cbpending locks in
3830 Frequency : replay/resend of open
3832 Description: non null lock assetion failure in mds_intent_policy
3833 Details : Trying to replay/resend lockless open requests resulted in
3834 mds_open() returning 0 with no lock. Now it sets a flag if
3835 a lock is going to be returned.
3837 Severity : enhancement
3839 Description: Checksum enhancements
3840 Details : New checksum enhancements allow for resending RPCs that failed
3843 Severity : enhancement
3845 Description: Tunables on number of dirty pages in cacche
3846 Details : Allow to set limit on number of dirty pages cached.
3851 Description: client crash on unmount - lock still has references
3852 Details : In some error handling cases it was possible to leak a lock
3853 reference on a client while accessing a file. This was not
3854 harmful to the client during operation, but would cause the
3855 client to crash when the filesystem is unmounted.
3858 Frequency : specific case, rare
3860 Description: ETXTBSY on mds though file not in use by client
3861 Details : ETXTBSY is no longer incorrectly returned when attempting to
3862 chmod or chown a directory that the user previously tried to
3863 execute or a currently-executing binary.
3866 Frequency : extremely rare except on liblustre-based clients
3868 Description: Lustre space not freed when files are deleted
3869 Details : Clean up open-unlinked files after client eviction. Previously
3870 the unlink was skipped and the files remained as orphans.
3875 Description: OST failure "would be an LBUG" in waiting_locks_callback()
3876 Details : In some cases it was possible to send a blocking callback to
3877 a client doing a glimpse, even though that client didn't get
3878 a lock granted. When the glimpse lock is cancelled on the OST
3879 the freed lock is left on the waiting list and corrupted the list.
3882 Frequency : all core dumps
3884 Description: Broke core dumps to lustre
3885 Details : Negative dentry may be unhashed if parent does not have UPDATE
3886 lock, but some callers, e.g. do_coredump, expect dentry to be
3887 hashed after successful create, hash it in ll_create_it.
3889 ------------------------------------------------------------------------------
3891 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
3893 * Support for kernels:
3894 2.6.9-42.0.2.EL (RHEL 4)
3895 2.6.5-7.276 (SLES 9)
3896 2.4.21-40.EL (RHEL 3)
3897 2.6.12.6 vanilla (kernel.org)
3901 Frequency : always on RHEL 3
3903 Description: Number of open files grows over time
3904 Details : The number of open files grows over time, whether or not
3905 Lustre is started. This was due to a filp leak introduced
3906 by one of our kernel patches.
3908 ------------------------------------------------------------------------------
3910 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
3912 * Support for kernels:
3913 2.6.9-42.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)
3921 Bugzilla : 5719, 9635, 9792, 9684
3922 Description: OST (or MDS) trips assertions in (re)connection under heavy load
3923 Details : If a server is under heavy load and cannot reply to new
3924 connection requests before the client resends the (re)connect,
3925 the connection handling code can behave badly if two service
3926 threads are concurrently handing separate (re)connections from
3927 the same client. Add better locking to the connection handling
3928 code, and ensure that only a single connection will be processed
3929 for a given client UUID, even if the lock is dropped.
3931 Severity : enhancement
3933 Description: add TCP zero-copy support to kernel
3934 Details : Add support to the kernel TCP stack to allow zero-copy bulk
3935 sends if the hardware supports scatter-gather and checksumming.
3936 This allows socklnd to do client-write and server-read more
3937 efficiently and reduce CPU utilization from skbuf copying.
3940 Frequency : only if NFS exporting from client
3942 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
3943 Details : If setting attributes on a file created under NFS that had
3944 never been opened it would be possible to oops the client
3945 if the file had no objects.
3948 Frequency : always for liblustre
3950 Description: liblustre client does MDS+OSTs setattr RPC for each write
3951 Details : When doing a write from a liblustre client, the client
3952 incorrectly issued an RPC to the MDS and each OST the file was
3953 striped over in order to update the timestamps. When writing
3954 with small chunks and many clients this could overwhelm the MDS
3955 with RPCs. In all cases it would slow down the write because
3956 these RPCs are unnecessary.
3958 Severity : enhancement
3960 Description: allow number of MDS service threads to be changed at module load
3961 Details : It is now possible to change the number of MDS service threads
3962 running. Adding "options mds mds_num_threads={N}" to the MDS's
3963 /etc/modprobe.conf will set the number of threads for the next
3964 time Lustre is restarted (assuming the "mds" module is also
3965 reloaded at that time). The default number of threads will
3966 stay the same, 32 for most systems.
3971 Description: OST crash if filesystem is unformatted or corrupt
3972 Details : If an OST is started on a device that has never been formatted
3973 or if the filesystem is corrupt and cannot even mount then the
3974 error handling cleanup routines would dereference a NULL pointer.
3979 Description: NULL pointer deref in llap_from_page.
3980 Details : get_cache_page_nowait can return a page with NULL (or otherwise
3981 incorrect) mapping if the page was truncated/reclaimed while it was
3982 searched for. Check for this condition and skip such pages when
3983 doing readahead. Introduce extra check to llap_from_page() to
3984 verify page->mapping->host is non-NULL (so page is not anonymous).
3987 Frequency : Sometimes when using sys_sendfile
3989 Description: "page not covered by a lock" warnings from ll_readpage
3990 Details : sendfile called ll_readpage without right page locks present.
3991 Now we introduced ll_file_sendfile that does necessary locking
3992 around call to generic_file_sendfile() much like we do in
3996 Frequency : with certain MDS communication failures at client mount time
3998 Description: NULL pointer deref after failed client mount
3999 Details : a client connection request may delayed by the network layer
4000 and not be sent until after the PTLRPC layer has timed out the
4001 request. If the client fails the mount immediately it will try
4002 to clean up before the network times out the request. Add a
4003 reference from the request import to the obd device and delay
4004 the cleanup until the network drops the request.
4007 Frequency : occasionally during client (re)connect
4009 Description: assertion failure during client (re)connect
4010 Details : processing a client connection request may be delayed by the
4011 client or server longer than the client connect timeout. This
4012 causes the client to resend the connection request. If the
4013 original connection request is replied in this interval, the
4014 client may trip an assertion failure in ptlrpc_connect_interpret()
4015 which thought it would be the only running connect process.
4018 Frequency : only with obd_echo servers and clients that are rebooted
4020 Description: kernel BUG accessing uninitialized data structure
4021 Details : When running an obd_echo server it did not start the ping_evictor
4022 thread, and when a client was evicted an uninitialized data
4023 structure was accessed. Start the ping_evictor in the RPC
4024 service startup instead of the OBD startup.
4026 Severity : enhancement
4027 Bugzilla : 10193 (patchless)
4028 Description: Remove dependency on various unexported kernel interfaces.
4029 Details : No longer need reparent_to_init, exit_mm, exit_files,
4030 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
4033 Frequency : rare (only users of deprecated and unsupported LDAP config)
4035 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
4036 Details : LDAP apparently contains 'lustreName' attributes instead of
4037 'name'. A simple remapping of the name is sufficient.
4040 Frequency : rare (only with non-default dump_on_timeout debug enabled)
4042 Description: waiting_locks_callback trips kernel BUG if client is evicted
4043 Details : Running with the dump_on_timeout debug flag turned on makes
4044 it possible that the waiting_locks_callback() can try to dump
4045 the Lustre kernel debug logs from an interrupt handler. Defer
4046 this log dumping to the expired_lock_main() thread.
4048 Severity : enhancement
4050 Description: Support NFS exporting on 2.6 kernels.
4051 Details : Implement non-rawops metadata methods for NFS server to use without
4052 changing NFS server code.
4055 Frequency : very rare (synthetic metadata workload only)
4057 Description: two racing renames might cause an MDS thread to deadlock
4058 Details : Running the "racer" program may cause one MDS thread to rename
4059 a file from being the source of a rename to being the target of
4060 a rename at exactly the same time that another thread is doing
4061 so, and the second thread has already enqueued these locks after
4062 doing a lookup of the target and is trying to relock them in
4063 order. Ensure that we don't try to re-lock the same resource.
4066 Frequency : only very large systems with liblustre clients
4068 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
4069 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
4070 all OSTs in parallel. This allows the eviction of stale liblustre
4071 clients to proceed much faster than if they were done in series,
4072 and also offers similar improvements for other set_info RPCs.
4077 Description: excessive CPU usage during initial read phase on client
4078 Details : During the initial read phase on a client, it would agressively
4079 retry readahead on the file, consuming too much CPU and impacting
4080 performance (since 1.4.5.8). Improve the readahead algorithm
4081 to avoid this, and also improve some other common cases (read
4082 of small files in particular, where "small" is files smaller than
4083 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
4088 Description: MDS crash when receiving packet with unknown intent.
4089 Details : Do not LBUG in unknown intent case, just return -EFAULT
4091 Severity : enhancement
4092 Bugzilla : 9293, 9385
4093 Description: MDS RPCs are serialised on client. This is unnecessary for some.
4094 Details : Do not serialize getattr (non-intent version) and statfs.
4097 Frequency : occasional, when OST network is overloaded/intermittent
4099 Description: client evicted by OST after bulk IO timeout
4100 Details : If a client sends a bulk IO request (read or write) the OST
4101 may evict the client if it is unresposive to its data GET/PUT
4102 request. This is incorrect if the network is overloaded (takes
4103 too long to transfer the RPC data) or dropped the OST GET/PUT
4104 request. There is no need to evict the client at all, since
4105 the pinger and/or lock callbacks will handle this, and the
4106 client can restart the bulk request.
4109 Frequency : Always when mmapping file with no objects
4111 Description: client crashes when mmapping file with no objects
4112 Details : Check that we actually have objects in a file before doing any
4113 operations on objects in ll_vm_open, ll_vm_close and
4119 Description: Request leak when working with deleted CWD
4120 Details : Introduce advanced request refcount tracking for requests
4121 referenced from lustre intent.
4123 Severity : Enhancement
4125 Description: Cache open file handles on client.
4126 Details : MDS now will return special lock along with openhandle, if
4127 requested and client is allowed to hold openhandle, even if unused,
4128 until such a lock is revoked. Helps NFS a lot, since NFS is opening
4129 closing files for every read/write openration.
4131 Severity : Enhancement
4133 Description: Cache open negative dentries on client when possible.
4134 Details : Guard negative dentries with UPDATE lock on parent dir, drop
4135 negative dentries on lock revocation.
4140 Description: Remounting a client read-only wasn't possible with a zconf mount
4141 Details : It wasn't possible to remount a client read-only with llmount.
4143 Severity : enhancement
4144 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
4145 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
4146 MPICH is included to add Lustre-specific ADIO interfaces.
4147 This is based closely on the UFS ADIO layer and only differs
4148 in file creation, in order to allow the OST striping to be set.
4149 This is user-contributed code and not supported by CFS.
4154 Description: extended inode attributes (immutable, append-only) work improperly
4155 when 2.4 and 2.6 kernels are used on client/server or vice versa
4156 Details : Introduce kernel-independent values for these flags.
4158 Severity : enhancement
4161 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
4162 Details : Many of the /proc/ tunables can only be tuned at a megabyte
4163 granularity. Now, Fractional MB granularity is be supported,
4164 this is very useful for low memory system.
4166 Severity : enhancement
4168 Description: Getattr by fid
4169 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
4170 locks, avoids extra getattr rpc requests to MDS, allows '/' to
4171 have locks and avoids getattr rpc requests for it on every stat.
4174 Frequency : Always, for filesystems larger than 2TB
4176 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4177 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
4178 filesystems larger than 2TB to be mounted with mballoc enabled.
4181 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
4183 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4184 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
4185 filesystems larger than 2TB is mounted then the kernel will
4186 silently corrupt the start of the filesystem. CONFIG_LBD is
4187 enabled for all CFS-supported kernels, but the possibility of
4188 this happening with a modified kernel config exists.
4190 Severity : enhancement
4192 Description: add client O_DIRECT support for 2.6 kernels
4193 Details : It is now possible to do O_DIRECT reads and writes to files
4194 in the Lustre client mountpoint on 2.6 kernel clients.
4196 Severity : enhancement
4198 Description: parallel glimpse, setattr, statfs, punch, destroy requests
4199 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
4200 parallel, not waiting for response from every OST before sending
4201 a rpc to the next OST.
4206 Description: setattr vs write race when updating file timestamps
4207 Details : Client processes that update a file timestamp into the past
4208 right after writing to the file (e.g. tar) it is possible that
4209 the updated file modification time can be reset to the current
4210 time due to a race between processing the setattr and write RPC.
4212 Severity : enhancement
4214 Description: Bring 'lfs find' closer in line with regular Linux find.
4215 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
4216 -print0 options and obtains all the needed info through the lustre
4219 Severity : enhancement
4221 Description: support up to 1024 configured devices on one node
4222 Details : change obd_dev array from statically allocated to dynamically
4223 allocated structs as they are first used to reduce memory usage
4228 Description: Flush dirty partially truncated pages during truncate
4229 Details : Immediatelly flush partially truncated pages in filter_setattr,
4230 this way we completely avoid having any pages in page cache on OST
4231 and can retire ugly workarounds during writes to flush such pages.
4236 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
4237 Details : protect inode from truncation within vfs_unlink() context
4238 just take a reference before calling vfs_unlink() and release it
4239 when parent's i_sem is free.
4244 Description: last_id value checked outside lock on OST caused LASSERT failure
4245 Details : If there were multiple MDS->OST object precreate requests in
4246 flight, it was possible that the OST's last object id was checked
4247 outside a lock and incorrectly tripped an assertion. Move checks
4248 inside locks, and discard old precreate requests.
4251 Frequency : always, if extents are used on OSTs
4253 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
4254 Details : OSTs using the extents format would not zero the high 16 bits of
4255 the index physical block number. This is not a problem for any
4256 OST filesystems smaller than 16TB, and no kernels support ext3
4257 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
4258 new/modified files) and can be fixed for existing filesystems
4259 with e2fsprogs-1.39-cfs1.
4264 Description: import connection selection may be incorrect if timer wraps
4265 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
4266 import connections to be ignored if the 32-bit jiffies counter
4267 wraps. Use a 64-bit jiffies counter.
4270 Frequency : very large clusters immediately after boot
4272 Description: LNET request buffers exhausted under heavy short-term load
4273 Details : If a large number of client requests are generated on a service
4274 that has previously never seen so many requests it is possible
4275 that the request buffer growth cannot keep up with the spike in
4276 demand. Instead of dropping incoming requests, they are held in
4277 the LND until the RPC service can accept more requests.
4280 Frequency : Sometimes during replay
4282 Description: Assertion failure in ll_local_open after replay.
4283 Details : If replay happened on an open request reply before we were able
4284 to set replay handler, reply will become not swabbed tripping the
4285 assertion in ll_local_open. Now we set the handler right after
4286 recognising of open request
4289 Frequency : very rare
4291 Description: kernel reports "badness in vsnprintf"
4292 Details : Reading from the "recovery_status" /proc file in small chunks
4293 may cause a negative length in lprocfs_obd_rd_recovery_status()
4294 call to vsnprintf() (which is otherwise harmless). Exit early
4295 if there is no more space in the output buffer.
4297 Severity : enhancement
4299 Description: clear OBD RPC statistics by writing to them
4300 Details : It is now possible to clear the OBD RPC statistics by writing
4301 to the "stats" file.
4306 Description: Client mtime is not the same on different clients after utimes
4307 Details : In some cases, the client was using the utimes() syscall on
4308 a file cached on another node. The clients now validate the
4309 ctime from the MDS + OSTs to determine which one is right.
4314 Description: Inability to activate failout mode
4315 Details : lconf script incorrectly assumed that in pythong string's numeric
4316 value is used in comparisons.
4319 Frequency : always with multiple stripes per file
4321 Description: Inefficient object allocation for mutli-stripe files
4322 Details : When selecting which OSTs to stripe files over, for files with
4323 a stripe count that divides evenly into the number of OSTs,
4324 the MDS is always picking the same starting OST for each file.
4325 Return the OST selection heuristic to the original design.
4330 Description: mount failures may take full timeout to return an error
4331 Details : Under some heavy load conditions it is possible that a
4332 failed mount can wait for the full obd_timeout interval,
4333 possibly several minutes, before reporting an error.
4334 Instead return an error as soon as the status is known.
4336 Frequency : quota enabled and large files being deleted
4338 Description: releasing more than 4GB of quota at once hangs OST
4339 Details : If a user deletes more than 4GB of files on a single OST it
4340 will cause the OST to spin in an infinite loop. Release
4341 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4346 Description: statfs data retrieved from /proc may be stale or zero
4347 Details : When reading per-device statfs data from /proc, in the
4348 {kbytes,files}_{total,free,avail} files, it may appear
4349 as zero or be out of date.
4352 Frequency : systems with MD RAID1 external journal devices
4354 Description: lconf's call to blkid is confused by RAID1 journal devices
4355 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4356 of returning all block devices that match the journal UUID.
4359 Frequency : always, for aggregate stripe size over 4GB
4361 Description: assertion fails when trying to use 4GB stripe size
4362 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
4363 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
4366 Frequency : always on ppc64
4368 Description: the first write on an ext3 filesystem with mballoc got stuck
4369 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4370 perform endianness conversion.
4372 ------------------------------------------------------------------------------
4374 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4376 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
4377 INTEROPERATE with older versions automatically. Please read the
4378 user documentation before upgrading any part of a live system.
4379 * WARNING: Lustre networking configuration changes are required with
4380 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
4383 * Support for kernels:
4384 2.6.9-22.0.2.EL (RHEL 4)
4385 2.6.5-7.244 (SLES 9)
4386 2.6.12.6 vanilla (kernel.org)
4389 Severity : enhancement
4390 Bugzilla : 7981/8208
4391 Description: Introduced Lustre Networking (LNET)
4392 Details : LNET is new networking infrastructure for Lustre, it includes
4393 a reorganized network configuration mode (see the user
4394 documentation for full details) as well as support for routing
4395 between different network fabrics. Lustre Networking Devices
4396 (LNDs) for the supported network fabrics have also been
4397 created for this new infrastructure.
4399 Severity : enhancement
4400 Description: Introduced Access control lists
4401 Details : clients can set ACLs on files and directories in order to have
4402 more fine-grained permissions than the standard Unix UGO+RWX.
4403 The MDS must be started with the "-o acl" mount option.
4405 Severity : enhancement
4406 Description: Introduced filesystem quotas
4407 Details : Administrators may now establish per-user quotas on the
4410 Severity : enhancement
4412 Description: Configuration change for the XT3
4413 The PTLLND is now used to run Lustre over Portals on the XT3
4414 The configure option(s) --with-cray-portals are no longer used.
4415 Rather --with-portals=<path-to-portals-includes> is used to
4416 enable building on the XT3. In addition to enable XT3 specific
4417 features the option --enable-cray-xt3 must be used.
4422 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
4423 Details : A race between allocating a new llcd and re-getting the llcd_lock
4424 allowed another thread to grab newly-allocated llcd.
4426 Severity : enhancement
4428 Description: 2.6 OST async journal commit and locking fix to improve performance
4429 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
4430 2.6 kernels are now async as they already were in 2.4 kernels so
4431 that they can commit concurrently with the network bulk transfer.
4432 For block-allocated files the filter allocation semaphore is held
4433 to avoid filesystem fragmentation during allocation. BKL lock
4434 removed for 2.6 xattr operations where it is no longer needed.
4439 Description: lconf incorrectly determined whether two IP networks could talk
4440 Details : In some more complicated routing and multiple-network
4441 configurations, lconf will avoid trying to make a network
4442 connection to a disjoint part of the IP space. It was doing the
4443 math incorrectly for one set of cases.
4448 Description: Fix for potential infinite loop processing records in an llog.
4449 Details : If an llog record is corrupted/zeroed, it is possible to loop
4450 forever in llog_process(). Validate the llog record length
4451 and skip the remainder of the block on error.
4454 Frequency : occasional (liblustre only)
4456 Description: liblustre could not open files whose last component is a symlink
4457 Details : sysio_path_walk() would incorrectly pass the open intent to
4458 intermediate path components.
4461 Frequency : rare (liblustre only with non-standard tuning)
4462 Bugzilla : 7201 (7350)
4463 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4464 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4465 after it has been released, only do this for referenced locks
4467 Severity : enhancement
4469 Description: specifying an (invalid) directory default stripe_size of -1
4470 would reset the directory default striping
4471 Details : stripe_size -1 was used internally to signal directory stripe
4472 removal, now use "all default" to signal dir stripe removal
4473 as a directory striping of "all default" is not useful
4476 Frequency : common for large clusters running liblustre clients
4478 Description: doing an ls when liblustre clients are running is slow
4479 Details : sending a glimpse AST to a liblustre client waits for every AST
4480 to time out, as liblustre clients will not respond. Since they
4481 cannot cache data we refresh the OST lock LVB from disk instead.
4483 Severity : enhancement
4485 Description: doing an ls at the same time as file IO can be slow
4486 Details : enqueue and other "small" requests can be blocked behind many
4487 large IO requests. Create a new OST IO portal for non-IO
4488 requests so they can be processed faster.
4491 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4493 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4494 created file with mode 000
4495 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4496 allow an ftruncate() as a non-root user to a file with mode 000.
4497 Lustre now respects this flag to disable mode checks when
4498 truncating a file owned by the user
4501 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4504 Description: Revoking locks from clients that went dead or catatonic might take
4506 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4507 cancellation of such locks instant on servers without waiting for
4508 any reply from clients. Clients drops these locks when cancel
4509 notification from server is received without replying.
4512 Frequency : liblustre-only, when liblustre client dies or becomes busy
4514 Description: Doing ls on Linux clients can take a long time with active
4516 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4517 granting such locks to it in the first place if possible. Locks
4518 are taken by proxy on the OST during the read or write and
4519 dropped immediately afterward. Add connect flags handling, do
4520 not grant locks to liblustre clients for glimpse ASTs.
4522 Severity : enhancement
4524 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4525 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4526 sequential reads and behaves badly with applications doing random
4527 reads. Improve it by reading ahead at least read region, and
4528 avoiding excessive large RPC for small reads.
4530 Severity : enhancement
4532 Description: Creating more than 1000 files for a single job may cause a load
4533 imbalance on the OSTs if there are also a large number of OSTs.
4534 Details : qos_prep_create() uses an OST index reseed value that is an
4535 even multiple of the number of available OSTs so that if the
4536 reseed happens in the middle of the object allocation it will
4537 still utilize the OSTs as uniformly as possible.
4542 Description: OST or MDS may oops in ping_evictor_main()
4543 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
4544 but doesn't restart at beginning of obd_exports_timed list
4547 Severity : enhancement
4549 Description: improve by-nid export eviction on the MDS and OST
4550 Details : allow multiple exports with the same NID to be evicted at one
4551 time without re-searching the exports list.
4554 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
4556 Description: MDS may oops in groups_free()
4557 Details : in rare race conditions a newly allocated group_info struct is
4558 freed again, and this can be NULL. The 2.4 compatibility code
4559 for groups_free() checked for a NULL pointer, but 2.6 did not.
4562 Frequency : common for liblustre clients doing little filesystem IO
4563 Bugzilla : 9352, 7313
4564 Description: server may evict liblustre clients accessing contended locks
4565 Details : if a client is granted a lock or receives a completion AST
4566 with a blocking AST already set it would not reply to the AST
4567 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
4570 Frequency : lfs setstripe, only systems with more than 160 OSTs
4572 Description: unable to set striping with a starting offset beyond OST 160
4573 Details : llapi_create_file() incorrectly limited the starting stripe
4574 index to the maximum single-file stripe count.
4577 Frequency : LDAP users only
4579 Description: lconf did not handle in-kernel recovery with LDAP properly
4580 Details : lconf/LustreDB get_refs() is searching the wrong namespace
4582 Severity : enhancement
4584 Description: bind OST threads to NUMA nodes to improve performance
4585 Details : all OST threads are uniformly bound to CPUs on a single NUMA
4586 node and do their allocations there to localize memory access
4588 Severity : enhancement
4590 Description: llmount can determine client NID directly from Myrinet (GM)
4591 Details : the client NID code from gmnalnid was moved directly into
4592 llmount, removing the need to use this or specifying the
4593 client NID explicitly when mounting GM clients with zeroconf
4596 Frequency : if client is started with down MDS
4598 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
4599 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
4600 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
4601 ptlrpc_import_delayed_req() because we didn't check if the
4602 request was interrupted, and we also didn't break out of the
4603 event loop if there was no timeout
4608 Description: data loss during non-page-aligned writes to a single file from
4609 both multiple nodes and multiple threads on one node at same time
4610 Details : updates to KMS and lsm weren't protected by common lock. Resulting
4611 inconsistency led to false short-reads, that were cached and later
4612 used by ->prepare_write() to fill in partially written page,
4613 leading to data loss.
4616 Frequency : always, if lconf --abort_recovery used
4618 Description: lconf --abort_recovery fails with 'Operation not supported'
4619 Details : lconf was attempting to abort recovery on the MDT device and not
4622 Severity : enhancement
4624 Description: remove cleanup logs
4625 Details : replace lconf-generated cleanup logs with lustre internal
4626 cleanup routines. Eliminates the need for client-cleanup and
4629 Severity : enhancement
4631 Description: add support for EAs (user and system) on lustre filesystems
4632 Details : it is now possible to store extended attributes in the Lustre
4633 client filesystem, and with the user_xattr mount option it
4634 is possible to allow users to store EAs on their files also
4636 Severity : enhancement
4638 Description: Add possibility (config option) to show minimal available OST free
4640 Details : When compiled with --enable-mindf configure option, statfs(2)
4641 (and so, df) will return least minimal free space available from
4642 all OSTs as amount of free space on FS, instead of summary of
4643 free spaces of all OSTs.
4645 Severity : enhancement
4647 Description: do not expand extent locks acquired on OST-side
4648 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
4649 by server: they are not cached anyway.
4652 Frequency : when mmap is used/binaries executed from Lustre
4654 Description: Unmmap pages before throwing them away from read cache.
4655 Details : llap_shrink cache now attempts to unmap pages before discarding
4656 them (if unmapping failed - do not discard). SLES9 kernel has
4657 extra checks that trigger if this unmapping is not done first.
4662 Description: lconf didn't resolve symlinks before checking to see whether a
4663 given mountpoint was already in use
4666 Frequency : when migrating failover services
4667 Bugzilla : 6395, 9514
4668 Description: When migrating a subset of services from a node (e.g. failback
4669 from a failover service node) the remaining services would
4670 time out and evict clients.
4671 Details : lconf --force (implied by --failover) sets the global obd_timeout
4672 to 5 seconds in order to quickly disconnect, but this caused
4673 other RPCs to time out too quickly. Do not change the global
4674 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
4676 Severity : enhancement
4677 Frequency : if MDS is started with down OST
4678 Bugzilla : 9439,5706
4679 Description: Allow startup/shutdown of an MDS without depending on the
4680 availability of the OSTs.
4681 Details : Asynchronously call mds_lov_synchronize during MDS startup.
4682 Add appropriate locking and lov-osc refcounts for safe
4683 cleaning. Add osc abort_inflight calls in case the
4684 synchronize never started.
4687 Frequency : occasional (Cray XT3 only)
4689 Description: root not authorized to access files in CRAY_PORTALS environment
4690 Details : The client process capabilities were not honoured on the MDS in
4691 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
4692 been accessed by an authorized user then root was able to access
4693 the file on the local client also. The root user capabilities
4694 are now allowed on the MDS, as this environment has secure UID.
4697 Frequency : occasional
4699 Description: ldiskfs "too long searching" message happens too often
4700 Details : A debugging message (otherwise harmless) prints too often on
4701 the OST console. This has been reduced to only happen when
4702 there are fragmentation problems on the filesystem.
4707 Description: Division by zero in statfs when all OSCs are inactive
4708 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
4709 lov_statfs divides by value returned by lov_get_stripecnt().
4713 Bugzilla : 9489, 3273
4714 Description: First write from each client to each OST was only 4kB in size,
4715 to initialize client writeback cache, which caused sub-optimal
4716 RPCs and poor layout on disk for the first writen file.
4717 Details : Clients now request an initial cache grant at (re)connect time
4718 and so that they can start streaming writes to the cache right
4719 away and always do full-sized RPCs if there is enough data.
4720 If the OST is rebooted the client also re-establishes its grant
4721 so that client cached writes will be honoured under the grant.
4726 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
4727 Details : Now I/O RPCs go to different portal number and (presumably) fast
4728 lock requests (and glimses) and other RPCs get their own service
4729 threads pool that should be able to service those RPCs
4732 Severity : enhancement
4734 Description: Ability to exchange lustre version between client and servers and
4735 issue warnings at client side if client is too old. Also for
4736 liblustre clients there is ability to refuse connection of too old
4738 Details : New 'version' field is added to connect data structure that is
4739 filled with version info. That info is later checked by server and
4743 Frequency : rare, liblustre only.
4744 Bugzilla : 9296, 9581
4745 Description: Two simultaneous writes from liblustre at offset within same page
4746 might proceed at the same time overwriting eachother with stale
4748 Details : I/O lock withing llu_file_prwv was released too early, before data
4749 actually was hitting the wire. Extended lock-holding time until
4750 server acknowledges receiving data.
4753 Frequency : extremely rare. Never observed in practice.
4755 Description: avoid generating lustre_handle cookie of 0.
4756 Details : class_handle_hash() generates handle cookies by incrementing
4757 global counter, and can hit 0 occasionaly (this is unlikely, but
4758 not impossible, because initial value of cookie counter is
4759 selected randonly). Value of 0 is used as a sentinel meaning
4760 "unassigned handle" --- avoid it. Also coalesce two critical
4761 sections in this function into one.
4763 Severity : enhancement
4765 Description: allow liblustre clients to delegate truncate locking to OST
4766 Details : To avoid overhead of locking, liblustre client instructs OST to
4767 take extent lock in ost_punch() on client's behalf. New connection
4768 flag is added to handle backward compatibility.
4770 Severity : enhancement
4771 Bugzilla : 4928, 7341, 9758
4772 Description: allow number of OST service threads to be specified
4773 Details : a module parameter allows the number of OST service threads
4774 to be specified via "options ost ost_num_threads={N}" in the
4775 OSS's /etc/modules.conf or /etc/modprobe.conf.
4779 Bugzilla : 6146, 9635, 9895
4780 Description: servers crash with bad pointer in target_handle_connect()
4781 Details : In rare cases when a client is reconnecting it was possible that
4782 the connection request was the last reference for that export.
4783 We would temporarily drop the export reference and get a new
4784 one, but this may have been the last reference and the export
4785 was just destroyed. Get new reference before dropping old one.
4787 Severity : enhancement
4788 Frequency : if client is started with failover MDS
4790 Description: Allow multiple MDS hostnames in the mount command
4791 Details : Try to read the configuration from all specified MDS
4792 hostnames during a client mount in case the "primary"
4795 Severity : enhancement
4797 Description: Stop sending data to evicted clients as soon as possible.
4798 Details : Check if the client we are about to send or are sending data to
4799 was evicted already. (Check is done every second of waiting,
4800 for which l_wait_event interface was extended to allow checking
4801 of exit condition at specified intervals).
4804 Frequency : rare, normally only when NFS exporting is done from client
4806 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
4808 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
4811 Frequency : always, if srand() is called before liblustre initialization
4813 Description: Liblustre uses system PRNG disturbing its usage by user application
4814 Details : Introduce internal to lustre fast and high-quality PRNG for
4815 lustre usage and make liblustre and some other places in generic
4816 lustre code to use it.
4818 Severity : enhancement
4819 Bugzilla : 9477, 9557, 9870
4820 Description: Verify that the MDS configuration logs are updated when xml is
4821 Details : Check if the .xml configuration logs are newer than the config
4822 logs stored on the MDS and report an error if this is the case.
4823 Request --write-conf, or allow starting with --old_conf.
4825 Severity : enhancement
4827 Description: Handle symlinks in the path when checking if Lustre is mounted.
4828 Details : Resolve intermediate symlinks when checking if a client has
4829 mounted a filesystem to avoid duplicate client mounts.
4834 Description: lconf can hit an error exception but still return success.
4835 Details : The lconf command catches the Command error exception at the top
4836 level script context and will exit with the associated exit
4837 status, but doesn't ensure that this exit status is non-zero.
4842 Description: failure of ptlrpc thread startup can cause oops
4843 Details : Starting a ptlrpc service thread can fail if there are a large
4844 number of threads or the server memory is very fragmented.
4845 Handle this without oopsing.
4848 Frequency : always, only if liblustre and non-default acceptor port was used
4850 Description: liblustre cannot connect to servers with non-default acceptor port
4851 Details : tcpnal_set_default_params() was not called and was therefore
4852 ignoring the environment varaible TCPNAL_PORT, as well as other
4853 TCPNAL_ environment variables
4858 Description: two objects could be created on the same OST for a single file
4859 Details : If an OST is down, in some cases it was possible to create two
4860 objects on a single OST for a single file. No problems other
4861 than potential performance impact and spurious error messages.
4865 Bugzilla : 5681, 9562
4866 Description: Client may oops in ll_unhash_aliases
4867 Details : Client dcache may become inconsistent in race condition.
4868 In some cases "getcwd" can fail if the current directory is
4874 Description: Inode refcounting problems in NFS export code
4875 Details : link_raw functions used to call d_instantiate without obtaining
4876 extra inode reference first.
4880 Bugzilla : 9942, 9903
4881 Description: Referencing freed requests leading to crash, memleaks with NFS.
4882 Details : We used to require that call to ll_revalidate_it was always
4883 followed by ll_lookup_it. Also with revalidate_special() it is
4884 possible to call ll_revalidate_it() twice for the same dentry
4885 even if first occurence returned success. This fix changes semantic
4886 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
4887 reference on a request referred from the intent.
4888 ll_intent_release() then releases such a request.
4891 Frequency : rare, normally benchmark loads only
4893 Description: unlinked inodes were kept in memory on the client
4894 Details : If a client is repeatedly creating and unlinking files it
4895 can accumulate a lot of stale inodes in the inode slab cache.
4896 If there is no other client load running this can cause the
4897 client node to run out of memory. Instead flush old inodes
4898 from client cache that have the same inode number as a new inode.
4901 Frequency : SLES9 2.6.5 kernel and long filenames only
4902 Bugzilla : 9969, 10379
4903 Description: utime reports stale NFS file handle
4904 Details : SLES9 uses out-of-dentry names in some cases, which confused
4905 the lustre dentry revalidation. Change it to always use the
4909 Frequency : rare, unless heavy write-truncate concurrency is continuous
4910 Bugzilla : 4180, 6984, 7171, 9963, 9331
4911 Description: OST becomes very slow and/or deadlocked during object unlink
4912 Details : filter_destroy() was holding onto the parent directory lock
4913 while truncating+unlinking objects. For very large objects this
4914 may block other threads for a long time and slow overall OST
4915 responsiveness. It may also be possible to get a lock ordering
4916 deadlock in this case, or run out of journal credits because of
4917 the combined truncate+unlink. Solution is to do object truncate
4918 first in one transaction without parent lock, and then do the
4919 final unlink in a new transaction with the parent lock. This
4920 reduces the lock hold time dramatically.
4923 Frequency : rare, 2.4 kernels only
4925 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
4926 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
4927 before unmounting their filesystem in order to avoid deadlock.
4928 The kernel_locked() function in 2.4 kernels only checks whether
4929 the kernel lock is held, not whether it is this process that is
4930 holding it as 2.6 kernels do.
4935 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
4936 Details : The client ptlrpc code may be trying to reconnect to a down
4937 server before a previous connection attempt has timed out.
4938 Increase the reconnect interval to be longer than the connection
4939 timeout interval to avoid sending duplicate connections to
4943 Frequency : echo_client brw_test command
4945 Description: fix echo_client to work with OST preallocated code
4946 Details : OST preallocation code (5137) didn't take echo_client IO path
4947 into account: echo_client calls filter methods outside of any
4948 OST thread and, hence, there is no per-thread preallocated
4949 pages and buffers to use. Solution: hijack pga pages for IO. As
4950 a byproduct, this avoids unnecessary data copying.
4954 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
4955 Description: Client can oops in mdc_commit_close() after open replay
4956 Details : It was possible for the MDS to return an open request with no
4957 transaction number in mds_finish_transno() if the client was
4958 evicted, but without actually returning an error. Clients
4959 would later try to replay that open and may trip an assertion
4960 Simplify the client close codepath, and always return an error
4961 from the MDS in case the open is not successful.
4964 Frequency : rare, 2.6 OSTs only
4966 Description: OST may deadlock under high load on fragmented files
4967 Details : If there was a heavy load and highly-fragmented OST filesystems
4968 it was possible to have all the OST threads deadlock waiting on
4969 allocation of biovecs, because the biovecs were not released
4970 until the entire RPC IO was completed. Instead, release biovecs
4971 as soon as they are complete to ensure forward IO progress.
4973 Severity : enhancement
4975 Description: Support for specifying external journal device at mount
4976 Details : If an OST or MDS device is formatted with an external journal
4977 device, this device major/minor is stored in the ext3 superblock
4978 and may not be valid for failover. Allow detecting and
4979 specifying the external journal at mount time.
4984 Description: Mounting an MDS with pending unlinked files may cause oops
4985 Details : target_finish_recovery() calls mds_postrecov() which returned
4986 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
4987 considers this an error and immediately begins cleaning up the
4988 lov, just after starting the mds_lov process
4990 Severity : enhancement
4992 Description: Implement 'lfs df' to report actual free space on per-OST basis
4993 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
4994 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
4995 usage of objects; '-h' to report in human readable format.
4997 ------------------------------------------------------------------------------
4999 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
5006 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
5007 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
5008 at startup during orphan unlinking. Instead, skip buddy update
5009 before setup, as it will be regenerated after recovery is complete.
5014 Description: If an OST is inactive, its locks might reference stale inodes.
5015 Details : lov_change_cbdata() must iterate over all namespaces, even if
5016 they are inactive to clear inode references from the lock.
5018 Severity : enhancement
5019 Frequency : occasional, if non-standard max_dirty_mb used
5021 Description: Client will block write RPCs if not enough grant
5022 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
5023 then the client will block writes while waiting for another RPC
5024 to complete instead of consuming its dirty limit. With change
5025 we get improved performance when max_dirty_mb is small.
5027 Severity : enhancement
5028 Bugzilla : 3389, 6253
5029 Description: Add support for supplementary groups on the MDS.
5030 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
5031 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
5032 lookups for user supplementary groups into a cache.
5036 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
5037 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
5038 for liblustre O_EXCL is enforced by MDS.
5042 Description: readdir never returns NULL in liblustre.
5043 Details : Corrected llu_iop_getdirentries logic, to return offset of next
5044 dentry in struct dirent.
5048 Frequency : liblustre only, depends on application IO pattern
5049 Description: liblustre clients evicted if not contacting servers
5050 Details : Don't put liblustre clients into the ping_evictor list, so
5051 they will not be evicted by the pinger ever.
5053 Severity : enhancement
5055 Description: Add ability to evict clients by NID from MDS.
5056 Details : By echoing "nid:$NID" string into
5057 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
5058 $NID would be instantly evicted from this MDS and from all active
5059 OSTs connected to it.
5063 Description: Do not query file size twice, somewhat slowing stat(2) calls.
5064 Details : lookup_it_finish() used to query file size from OSTs that was not
5069 Description: service threads change working directory to that of init
5070 Details : Starting lustre service threads may pin the working directory
5071 of the parent thread, making that filesystem busy. Threads
5072 now change to the working directory of init to avoid this.
5076 Frequency : during shutdown only
5077 Description: shutdown with a failed MDS or OST can cause unmount to hang
5078 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
5083 Frequency : 2.6 only
5084 Description: chmod/chown may include an extra supplementary group
5085 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
5086 supplementary group and if none is specified this is used.
5089 Bugzilla : 5479 (6816)
5091 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
5092 Details : If lookup is in progress on a file that is unlinked we might try
5093 to revalidate the inode and fail in revalidate after lookup is
5094 complete and ll_file_open() enqueues the open again but
5095 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
5098 Frequency : always, if lconf --abort_recovery used
5100 Description: lconf --abort_recovery fails with 'Operation not supported'
5101 Details : lconf was attempting to abort recovery on the MDT device and not
5104 ------------------------------------------------------------------------------
5106 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
5111 Frequency : rare (only unsupported configurations with a node running as an
5113 Bugzilla : 6514, 5137
5114 Description: Mounting a Lustre file system on a node running as an OST could
5116 Details : OSTs now preallocates memory needed to write out data at
5117 startup, instead of when needed, to avoid having to
5118 allocate memory in possibly low memory situations.
5119 Specifically, if the file system is mounted on on OST,
5120 memory pressure could force it to try to write out data,
5121 which it needed to allocate memory to do. Due to the low
5122 memory, it would be unable to do so and the node would
5123 become unresponsive.
5125 Severity : enhancement
5127 Description: Addition of lconf --service command line option
5128 Details : lconf now accepts a '--service <arg>' option, which is
5129 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
5131 Severity : enhancement
5133 Description: Failover mode is now the default for OSTs.
5134 Details : By default, OSTs will now run in failover mode. To return to
5135 the old behaviour, add '--failout' to the lmc line for OSTs.
5137 Severity : enhancement
5139 Description: Health checks are now provided for MDS and OSTs
5140 Details : Additional detailed health check information on MSD and OSTs
5141 is now provided through the procfs health_check value.
5144 Frequency : occasional, depends on IO load
5146 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
5147 after numerous create/delete cycles
5148 Details : The ext3 inode allocation policy would not allocate new inodes
5149 very well on the OSTs because there are no new directories
5150 being created. Instead we look for groups with free space if
5151 the parent directories are nearly full.
5156 Description: Network or server problems during mount may cause partially
5157 mounted clients instead of returning an error.
5158 Details : The config llog parsing code may overwrite the error return
5159 code during mount error handling, returning success instead
5165 Description: MDS can fail to allocate large reply buffers
5166 Details : After long uptimes the MDS can fail to allocate large reply
5167 buffers (e.g. zconf client mount config records) due to memory
5168 fragmentation or consumption by the buffer cache. Preallocate
5169 some large reply buffers so that these replies can be sent even
5170 under memory pressure.
5174 Frequency : rare (liblustre)
5175 Description: fsx running with liblustre complained that using truncate() to
5176 extend the file doesn't work. This patch corrects that issue.
5177 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
5178 ATTR_SIZE and lsm use in llu_setattr_raw.
5182 Frequency : rare, only 2.6 kernels
5183 Description: Unusual file access patterns on the MDS may result in inode
5184 data being lost in very rare circumstances.
5185 Details : Bad interaction between the ea-in-inode patch and the "no-read"
5186 code in the 2.6 kernel caused the inode and/or EA data not to
5187 be read from disk, causing single-file corruption.
5191 Frequency : rare, only 2.6 filesystems using extents
5192 Description: Heavy concurrent write and delete load may cause data corruption.
5193 Details : It was possible under high-load situations to have an extent
5194 metadata block in the block device cache from a just-unlinked
5195 file overwrite a newly-allocated data block. We now unmap any
5196 metadata buffers that alias just-allocated data blocks.
5200 Frequency : filesystems with default stripe_count larger than 77
5201 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
5202 Details : lconf specifies an inode size of 4096 bytes when the default
5203 stripe_count is larger than 77. This conflicts with the default
5204 inode density of 1 per 4096 bytes. Allocate smaller inodes in
5205 this case to avoid pinning too much memory for large EAs.
5207 ------------------------------------------------------------------------------
5209 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
5214 Frequency : rare (extremely heavy IO load with hundreds of clients)
5216 Description: Client is evicted, gets IO error writing to file
5217 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
5218 caused clients to be evicted for AST timeouts. The fixes in
5219 bug 5192 mean we no longer need to have such short AST timeouts
5220 so ldlm_timeout has been increased.
5223 Frequency : occasional during --force or --failover shutdown under load
5224 Bugzilla : 5949, 4834
5225 Description: Server oops/LBUG if stopped with --force or --failover under load
5226 Details : a collection of import/export refcount and cleanup ordering
5227 issues fixed for safer force cleanup
5230 Frequency : only filesystems larger than 120 OSTs
5231 Bugzilla : 5990, 6223
5232 Description: lfs getstripe would oops on a very large filesystem
5233 Details : lov_getconfig used kfree on vmalloc'd memory
5236 Frequency : only filesystems exporting via NFS to Solaris 10 clients
5237 Bugzilla : 6242, 6243
5238 Description: reading from files that had been truncated to a non-zero size
5239 but never opened returned no data
5240 Details : ll_file_read() reads zeros from no-object files to EOF
5245 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
5246 Details : There's little harm in aborting MDS/OSS recovery and letting it
5247 try again, so I removed the LASSERT and return an error instead.
5249 Severity : enhancement
5251 Description: New debugging infrastructure for tracking down data corruption
5252 Details : The I/O checksum code was replaced to: (a) control it at runtime,
5253 (b) cover more of the client-side code path, and (c) try to narrow
5254 down where problems occurred
5258 Bugzilla : 3819, 4364, 4397, 6313
5259 Description: Racing close and eviction MDS could cause assertion in mds_close
5260 Details : It was possible to get multiple mfd references during close and
5261 client eviction, leading to one thread referencing a freed mfd.
5263 Severity: : enhancement
5264 Bugzilla : 3262, 6359
5265 Description: Attempts to reconnect to servers are now more aggressive.
5266 Details : This builds on the enhanced upcall-less recovery that was added
5267 in 1.4.2. When trying to reconnect to servers, clients will
5268 now try each server in the failover group every 10 seconds. By
5269 default, clients would previously try one server every 25 seconds.
5274 Description: After recovery, certain operations trigger a failed
5275 assertion on a client.
5276 Details : Failing over an mds, using lconf -d --failover, while a
5277 client was doing a readdir() call would cause the client to
5278 LBUG after recovery completed and the readdir() was resent.
5280 Severity : enhancement
5282 Description: Default groups are now added by lconf
5283 Details : You can now run lconf --group <servicename> without having to
5284 manually add groups with lmc.
5287 Frequency : occasional
5289 Description: Nodes with an elan id of 0 trigger a failed assertion
5292 Frequency : always when accessing e.g. tty/console device nodes
5294 Description: tty and some other devices nodes cannot be used on lustre
5295 Details : file's private_data field is used by device data and lustre
5296 values in there got lost. New field was added to struct file to
5297 store fs-specific private data.
5300 Frequency : when exporting Lustre via NFS
5302 Description: NFSD failed occasionally when looking up a path component
5303 Details : NFSD is looking up ".." which was broken in ext3 directories
5304 that had grown large enough to become hashed.
5307 Frequency : Clusters with multiple interfaces not on the same subnet
5309 Description: Nodes will repeatedly try to reconnect to an interface which it
5310 cannot reach and report an error to the log.
5311 Details : Extra peer list entries will be created by lconf with some peers
5312 unreachable. lconf now validates the peer before adding it.
5315 Frequency : Only if a default stripe is set on the filesystem root.
5317 Description: Setting a default stripe on the filesystem root prevented the
5318 filesystem from being remounted.
5319 Details : The client was sending extra request flags in the root getattr
5320 request and did not allocate a reply buffer for the dir EA.
5323 Frequency : occasional, higher if lots of files are accessed by one client
5324 Bugzilla : 6159, 6097
5325 Description: Client trips assertion regarding lsm mismatch/magic
5326 Details : While revalidating inodes the VFS looks up inodes with ifind()
5327 and in rare cases can find an inode that is being freed.
5328 The ll_test_inode() code will free the lsm during ifind()
5329 when it finds an existing inode and then the VFS later attaches
5330 this free lsm to a new inode.
5334 Bugzilla : 6422, 7030
5335 Description: MDS deadlock between mkdir and client eviction
5336 Details : Creating a new file via mkdir or mknod (starting a transaction
5337 and getting the ns lock) can deadlock with client eviction
5338 (gets ns lock and trying to finish a synchronous transaction).
5341 Frequency : occasional
5342 Description: While starting a server, the fsfilt_ext3 module could not be
5344 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
5345 kernels. Previously, lconf would still use the ext3 name
5346 when trying to load modules. Now, it will correctly use
5347 ext3 on 2.4 and ldiskfs on 2.6.
5349 Severity : enhancement
5350 Description: The default stripe count has been changed to 1
5351 Details : The interpretation of the default stripe count (0, to lfs
5352 or lmc) has been changed to mean striping across a single
5353 OST, rather than all available. For general usage we have
5354 found a stripe count of 1 or 2 works best.
5356 Severity : enhancement
5357 Description: Add support for compiling against Cray portals.
5358 Details : Conditional compiling for some areas that are different
5362 Frequency : occasional
5363 Bugzilla : 6409, 6834
5364 Description: Creating files with an explicit stripe count may lead to
5365 a failed assertion on the MDS
5366 Details : If some OSTs are full or unavailable, creating files may
5367 trigger a failed assertion on the MDS. Now, Lustre will
5368 try to use other servers or return an error to the
5372 Frequency : occasional
5374 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
5375 caused lock timeout during readahead (since 1.4.2).
5376 Details : Processes doing readahead might match a lock that hasn't been
5377 granted yet if there are overlapping and conflicting lock
5378 requests. The readahead process waits on ungranted lock
5379 (original lock is CBPENDING), while OST waits for that process
5380 to cancel CBPENDING read lock and eventually evicts client.
5382 Severity : enhancement
5384 Description: Initial enabling of flock support for clients
5385 Details : Implements fcntl advisory locking and file status functions.
5386 This feature is provided as an optional mount flag (default
5387 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
5388 locking are implemented yet, and those that are are not guaranteed
5389 to be completely correct in production environments.
5390 mount -t lustre -o [flock|noflock] ...
5393 Frequency : occasional
5395 Description: OSTs running 2.4 kernels but with extents enabled might trip an
5396 assertion in the ext3 JBD (journaling) layer.
5397 Details : The b_committed_data struct is protected by the big kernel lock
5398 in 2.4 kernels, serializing journal_commit_transaction() and
5399 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
5400 access to this struct. In 2.6 kernels there is finer grained
5401 locking to improve SMP performance of the JBD layer.
5405 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
5410 Description: merge and cleanup kernel patches.
5413 -----------------------------------------------------------------------------
5415 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
5417 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
5418 versions of Lustre. Please update all servers and clients to
5419 version 1.4.2 or later at the same time. You must also run
5420 "lconf --write-conf {config}.xml" on the MDS while it is stopped
5421 to update the configuration logs.
5423 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
5424 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
5425 - swab llog records of type '0' so we get proper header size/idx (5861)
5426 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
5427 - fix rename of one directory over another leaking an inode (5953)
5428 - avoid SetPageDirty on 2.6 (5981)
5429 - don't re-add just-being-destroyed locks to the waiting list (5653)
5430 - when creating new directories, inherit the parent's custom
5431 striping settings if present parent (3048)
5432 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
5433 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
5434 - don't hold client locks on temporary worklist from l_lru (5666)
5435 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
5436 - automatically evict dead clients (5921)
5437 - Update file size properly in create+truncate+fstat case (6196)
5438 - Do not unhash mountpoint dentries, do not allow removal of
5440 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
5441 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
5442 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
5443 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
5444 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
5445 - Correctly handle reads of files with no objects (6243)
5446 - lctl recover will also mark a device active if deactivate used (5933)
5448 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
5449 - allow --write-conf on an MDS with different nettype than client (5619)
5450 - don't write config llogs to MDS for mounts not from that MDS (5617)
5451 - lconf should create multiple TCP connections from a client (5201)
5452 - init scripts are now turned off by default; run chkconfig --on
5453 lustre and chkconfig --on lustrefs to use them
5454 - upcalls are no longer needed for clients to recover to failover
5456 - add --abort-recovery option to lconf to abort recovery on device
5458 - add support for an arbitrary number of OSTs (3026)
5459 - Quota support protocol changes.
5460 - forward compatibility changes to wire structs (6007)
5461 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5462 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5463 - improved "lustre status" script
5464 - initialize blocksize for non-regular files (6062)
5465 - added --disable-server and --disable-client configure options (5782)
5466 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5467 - Vanilla 2.4.29 support
5468 - increase maximum number of obd devices to 520 (6242)
5469 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5470 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5471 SLES 9 2.6.5 kernels
5472 - limit stripes per file to 160 (the maximum EA size) (6093)
5474 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5477 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5478 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5479 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5480 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5481 - small ext3 extents cleanups and fixes (5733)
5482 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5483 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5484 - avoid needless client->OST connect, fix handle mismatch (5317)
5485 - fix DLM error path that led to out-of-sync client, long delays (5779)
5486 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5487 - fix several locking issues related to i_size (5492,5624,5654,5672)
5488 - don't move pending lock onto export if it is already evicted (5683)
5489 - fix kernel oops when creating .foo in unlinked directory (5548)
5490 - fix deadlock in obdfilter statistics vs. object create (5811)
5491 - use time_{before,after} to avoid timer jiffies wrap (5882)
5492 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5493 - Do not leak request if server was not able to process it (5154)
5494 - If mds_open unable to find parent dir, make that negative lookup(5154)
5495 - don't create new directories with extent-mapping (5909, 5936)
5497 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5498 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5499 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5500 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5501 - support for RHEL4 kernel on i686 (5773)
5502 - provide error messages when incompatible logs are encountered (5898)
5504 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5505 * version 1.4.0.10 (1.4.1 release candidate 1)
5507 - don't keep a lock reference when lock is not granted (4238)
5508 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5509 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5510 - drop import inflight refcount on signal_completed_replay error (5255)
5511 - unlock page after async write error during send (3677)
5512 - handle missing objects in filter_preprw_read properly (5265)
5513 - no transno return for symlink open, don't save no-trasno open (3440)
5514 - don't try to complete elan receive that already failed (4012)
5515 - free RPC server reply state on error (5406)
5516 - clean up thread from ptlrpc_start_thread() on error (5160)
5517 - readahead could read extra page into cache that wasn't ejected (5388)
5518 - prevent races in class_attach/setup/cleanup/detach (5260)
5519 - don't dereference de->d_inode after l_dput of de (5458)
5520 - use "int" for stripe value returned from lock_to_stripe (5544)
5521 - mballoc allocation and error-checking fixes in 2.6 (5504)
5522 - block device patches to fix I/O request sizes in 2.6 (5482)
5523 - look up hostnames for IB nals (5602)
5524 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5525 - don't start multiple acceptors for the same port (5277)
5526 - fix incorrect LASSERT in mds_getattr_name (5635)
5527 - export a proc file for general "ping" checking (5628)
5528 - fix "lfs check" to not block when the MDS is down (5628)
5530 - service request history (4965)
5531 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5532 - create an "evict_client" /proc entry on OSTs, like the MDS has
5533 - fix mount usage message, return errors per mount(8) (5168)
5534 - change grep [] to grep "[]" in tests so they work in more UMLs
5535 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5536 - remove ancient LOV_MAGIC_V0 EA support (5047)
5537 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5538 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5539 - allow readahead allocations to fail when low on memory (5383)
5540 - mmap locking landed again, after considerable improvement (2828)
5541 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
5543 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
5546 - send OST transaction number in read/write reply to free req (4966)
5547 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5548 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5549 - the watchdog thread now runs as interruptible (5246)
5550 - flock/lockf fixes (but it's still disabled, pending 5135)
5551 - don't use EXT3 constants in llite code (5094)
5552 - memory shortage at startup could cause assertion (5176)
5554 - reorganization of lov code
5555 - single portals codebase
5557 - add extents/mballoc support (5025)
5558 - direct I/O reads in the obdfilter (4048)
5559 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
5561 tbd Cluster File Systems, Inc. <info@clusterfs.com>
5564 - send OST transaction number in read/write reply to free req (4966)
5565 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5566 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5567 - the watchdog thread now runs as interruptible (5246)
5568 - handle missing objects in filter_preprw_read properly (5265)
5569 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5570 - drop import inflight refcount on signal_completed_replay error (5255)
5571 - unlock page after async write error during send (3677)
5572 - return original error code on reconstructed replies (3761)
5573 - no transno return for symlink open, don't save no-trasno open (3440)
5575 - add pid to ldlm debugging output (4922)
5576 - bump the watchdog timeouts -- we can't handle 30sec yet
5577 - extra debugging for orphan dentry/inode bug (5259)
5579 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
5582 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
5583 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
5584 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
5585 - fix NULL dereference in /proc/sys/portals/routes (4827)
5586 - allow failed mdc_close() operations to be interrupted (4561)
5587 - stop precreate on OST before MDS would time out on it (4778)
5588 - don't send partial-page writes before EOF from client (4410)
5589 - discard client grant for sub-page writes on large-page clients (4520)
5590 - don't free dentries not owned by NFS code, check generation (4806)
5591 - fix lsm leak if mds_create_objects() fails (4801)
5592 - limit debug_daemon file size, always print CERROR messages (4789)
5593 - use transno after validating reply (3892)
5594 - process timed out requests if import state changes (3754)
5595 - update mtime on OST during writes, return in glimpse (4829)
5596 - add mkfsoptions to LDAP (4679)
5597 - use ->max_readahead method instead of zapping global ra (5039)
5598 - don't interrupt __l_wait_event() during strace
5600 - add software watchdogs to catch hung threads quickly (4941)
5601 - make lustrefs init script start after nfs is mounted
5602 - fix CWARN/ERROR duplication (4930)
5603 - return async write errors to application if possible (2248)
5604 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
5605 - print NAL number in %x format (4645)
5606 - update barely-supported suse-2.4.21-171 series (4842)
5607 - support for sles 9 %post scripts
5608 - support for building 2.6 kernel-source packages
5609 - support for sles km_* packages
5611 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
5614 - ignore -ENOENT errors in osc_destroy (3639)
5615 - notify osc create thread that OSC is being cleaned up (4600)
5616 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
5617 - reconstruct ost_handle() like mds_handle() (4657)
5618 - create a new thread to do import eviction to avoid deadlock (3969)
5619 - let lconf resolve symlinked-to devices (4629)
5620 - don't unlink "objects" from directory with default EA (4554)
5621 - hold socknal file ref over connect in case target is down (4394)
5622 - allow more than 32000 subdirectories in a single directory (3244)
5623 - fix blocks count for O_DIRECT writes (3751)
5624 - OST returns ENOSPC from object create when no space left (4539)
5625 - don't send truncate RPC if file size isn't changing (4410)
5626 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
5627 - bind to privileged port in socknal and tcpnal (3689)
5629 - rate limit CERROR/CWARN console message to avoid overload (4519)
5630 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
5631 - basic mmap support (3918)
5632 - kernel patch series update from b1_4 (4711)
5634 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
5637 - avoid crash during MDS cleanup with OST shut down (2775)
5638 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
5639 - don't use bad inodes on the MDS (3744)
5640 - dynamic object preallocation to improve recovery speed (4236)
5641 - don't hold spinlock over lock dumping or change debug flags (4401)
5642 - don't zero obd_dev when it is force cleaned (3651)
5643 - print grants to console if they go negative (4431)
5644 - "lctl deactivate" will stop automatic recovery attempts (3406)
5645 - look for existing locks in ldlm_handle_enqueue() (3764)
5646 - don't resolve lock handle twice in recovery avoiding race (4401)
5647 - revalidate should check working dir is a directory (4134)
5649 - don't always mark "slow" obdfilter messages as errors (4418)
5651 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
5654 - don't close LustreDB during write_conf until it is done (3860)
5655 - fix typo in lconf for_each_profile (3821)
5656 - allow dumping logs from multiple threads at one time (3820)
5657 - don't allow multiple threads in OSC recovery (3812)
5658 - fix debug_size parameters (3864)
5659 - fix mds_postrecov to initialize import for llog ctxt (3121)
5660 - replace config semaphore with spinlock (3306)
5661 - be sure to send a reply for a CANCEL rpc with bad export (3863)
5662 - don't allow enqueue to complete on a destroyed export (3822)
5663 - down write_lock before checking llog header bitmap (3825)
5664 - recover from lock replay timeout (3764)
5665 - up llog sem before sending rpc (3652)
5666 - reduce ns lock hold times when setting kms (3267)
5667 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
5668 - fix NULL deref and obd_dev leak on setup error (3312)
5669 - replace some LBUG about llog ops with error handling (3841)
5670 - don't match INVALID dentries from d_lookup and spin (3784)
5671 - hold dcache_lock while marking dentries INVALID and hashing (4255)
5672 - fix invalid assertion in ptlrpc_set_wait (3880)
5674 - add libwrap support for the TCP acceptor (3996)
5675 - add /proc/sys/portals/routes for non-root route listing (3994)
5676 - allow setting MDS UUID in .xml (2580)
5677 - print the stack of a process that LBUGs (4228)
5679 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
5682 - don't cleanup request in ll_file_open() on failed MDS open (3430)
5683 - make sure to unset replay flag from failed open requests (3440)
5684 - if default stripe count is 0, use OST count for inode size (3636)
5685 - update parent mtime/ctime on client for create/unlink (2611)
5686 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
5687 - free recovery state on server during a forced cleanup (3571)
5688 - unregister_reply for resent reqs (3063)
5689 - loop back devices mounting and status check on 2.6 (3563)
5690 - fix resource-creation race that can provoke i_size == 0 (3513)
5691 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
5692 - more debugging for page-accounting assertion (3746)
5693 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
5694 - avoid deadlock after precreation failure (3758)
5695 - fix race and lock order deadlock in orphan handling (3450, 3750)
5696 - add validity checks when grabbing inodes from l_ast_data (3599)
5698 - add /proc/.../recovery_status to obdfilter (3428)
5699 - lightweight CDEBUG infrastructure, debug daemon (3668)
5700 - change default OSC RPC parameters to be better on small clusters
5701 - turn off OST read cache for files smaller than 32MB
5702 - install man pages and include them in rpms (3100)
5703 - add new init script for (un)mounting lustre filesystems (2593)
5704 - run chkconfig in %post for init scripts (3701)
5705 - drop scimac NAL (unmaintained)
5707 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
5710 - clean kiobufs before and after use (3485)
5711 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
5712 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
5713 - change init script to fail more clearly if not run as root (1528)
5714 - allow clients to reconnect during replay (1742)
5715 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
5716 - don't do DNS lookups on NIDs too small for IP addresses (3442)
5717 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
5718 - fix cond_resched (3554)
5719 - only evict unfinished clients after recovery (3515)
5720 - allow bulk resend, prevent data loss (3570)
5721 - dynamic ptlrpc request buffer allocation (2102)
5722 - don't allow unlinking open directory if it isn't empty (2904)
5723 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
5724 - remove extraneous obd dereference causing LASSERT failure (3334)
5725 - don't use get_cycles() when creating temp. files on the mds (3156)
5726 - hold i_sem when setting i_size in ll_extent_lock() (3564)
5727 - handle EEXIST for set-stripe, set proper directory name (3336)
5729 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
5730 - fix ksocknal_fmb_callback() error messages (2918)
5732 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
5735 - don't copy lvb into (possibly NULL) reply on error (2983)
5736 - don't deref dentry after dput, don't free lvb on error (2922)
5737 - use the kms to determine writeback rpc length (2947)
5738 - increment oti_logcookies when osc is inactive (2948)
5739 - update client's i_blocks count via lvb messages (2543)
5740 - handle intent open/close of special files properly (1557)
5741 - mount MDS with errors=remount-ro, like obdfilter (2009)
5742 - initialize lock handle to avoid ASSERT on error cleanup (3057)
5743 - don't use cancelling-locks' kms values (2947)
5744 - use highest lock extent for kms, not last one (2925)
5745 - don't dereference ERR_PTR() dentry in error handling path (3107)
5746 - fix thread race in portals_debug_dumplog() (3122)
5747 - create lprocfs device entries at setup instead of at attach (1519)
5748 - common AST error handler, don't evict client on completion race (3145)
5749 - zero nameidata in detach_mnt in 2.6 (3118)
5750 - verify d_inode after revalidate_special is valid in 2.6 (3116)
5751 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
5752 - initialize RPC timeout timer earlier for 2.6 (3219)
5753 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
5754 - print nal/nid for unknown nid (3258)
5755 - additional checks for oscc recovery before doing precreate (3284)
5756 - fix ll_extent_lock() error return code for 64-bit systems (3043)
5757 - don't crash in mdc_close for bad permissions on open (3285)
5758 - zero i_rdev for non-device files (3147)
5759 - clear page->private before handing to FS, better assertion (3119)
5760 - tune the read pipeline (3236)
5761 - fix incorrect decref of invalidated dentry (2350)
5762 - provide read-ahead stats and refine rpc in flight stats (3328)
5763 - don't hold journal transaction open across create RPC (3313)
5764 - update atime on MDS at close time (3265)
5765 - close LDAP connection when recovering to avoid server load (3315)
5766 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
5767 - don't leak open file on MDS after open resend (3325)
5768 - serialize filter_precreate and filter_destroy_precreated (3329)
5769 - loop device shouldn't call sync_dev() for nul device (3092)
5770 - clear page cache after eviction (2766)
5771 - resynchronize MDS->OST in background (2824)
5772 - refuse to mount the same filesystem twice on same mountpoint (3394)
5773 - allow llmount to create routes for mounting behind routers (3320)
5774 - push lock cancellation to blocking thread for glimpse ASTs (3409)
5775 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
5776 - fix rare problem with rename on htree directories (3417)
5778 - allow default OST striping configuration per directory (1414)
5779 - fix compilation for qswnal for 2.6 kernels (3125)
5780 - increase maximum number of MDS request buffers for large systems
5781 - change liblustreapi to be useful for external progs like lfsck (3098)
5782 - increase local configuration timeout for slow disks (3353)
5783 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
5785 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5788 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
5789 - don't overwrite extent policy data in reply if lock was blocked (2901)
5790 - drop filter export grants atomically with removal from device (2663)
5791 - del obd_self_export from work_list in class_disconnect_exports (2908)
5792 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
5793 - swab reply message in mdc_close, other PPC fixes (2464)
5794 - fix destroying of named logs (2325)
5795 - overwrite old logs when running lconf --write_conf (2264)
5796 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
5797 - fix race in target_handle_connect (2898)
5798 - mds_reint_create() should take same inode create lock (2926)
5799 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
5800 - don't close files for self_export to avoid uninitialized obd (2936)
5801 - allow MDS with the same name as client node (2939)
5802 - hold dentry reference for closed log files for unlink (2325)
5803 - reserve space for all logs during transactions (2059)
5804 - don't evict page beyond end of stripe extent (2925)
5805 - don't oops on a deleted current working directory (2399)
5806 - handle hard links to targets without a parent properly (2517)
5807 - don't dereference NULL lock when racing during eviction (2867)
5808 - don't grow lock extents when lots of conflicting locks (2919)
5810 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
5813 - account for cache space usage on clients to avoid data loss (974)
5814 - lfsck support in lustre kernel code (2349)
5815 - reduce journal credits needed for BRW writes (2370)
5816 - orphan handling to avoid losing space on client/server crashes
5817 - ptlrpcd can be blocked, stopping ALL progress (2477)
5818 - use lock value blocks to assist in proper KMS, faster stat (1021)
5819 - takes i_sem instead of DLM locks internally on obdfilter (2720)
5820 - recovery for initial connections (2355)
5821 - fixes for mds_cleanup_orphans (1934)
5822 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
5823 - block all file creations until orphan recovery completes (1901)
5824 - client remove rq_connection from request struct (2423)
5825 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
5826 - recovery timer race (2670)
5827 - mdc_close recovey bug (2532)
5828 - ptlrpc cleanup bug (2710)
5829 - mds timeout on local locks (2588)
5830 - namespace lock held during RPCs (2431)
5831 - handle interrupted sync write properly (2503)
5832 - don't try to handle a message that hasn't been replied to (2699)
5833 - client assert failure during cleanup after abort recovery (2701)
5834 - leak mdc device after failed mount (2712)
5835 - ptlrpc_check_set allows timedout requests to complete (2714)
5836 - wait for inflight reqs when ptlrpcd finishes (2710)
5837 - make sure unregistered services are removed from the srv_list
5838 - reset bulk XID's when resending them (caught by 1138 test)
5839 - unregister_bulk after timeout
5840 - fix lconf error (2694)
5841 - handle write after unfinished setstripe, stripe-only getstripe (2388)
5842 - readahead locks pages, leaves pending causing memory pressure (2673)
5843 - increase OST request buffers to 4096 on large machines (2729)
5844 - fix up permission of existing directories in simple_mkdir (2661)
5845 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
5846 - don't assign transno to errored transactions (2742)
5847 - don't delete objects on OST if given a bogus objid from MDS (2751)
5848 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
5849 - if rq_no_resend, then timeout request after recovery (2432)
5850 - fix MDS llog_logid record size, 64-bit array alignment (2733)
5851 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
5852 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
5853 - MDS recovery shouldn't delete objects that it has given out (2730)
5854 - if enqueue arrives after completion, don't clobber LVB (2819)
5855 - don't unlock pages twice when trigger_group_io returns error (2814)
5856 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
5857 - don't write pages to disk if there was an error (1450)
5858 - don't ping imports that have recovery disabled (2676)
5859 - take buffered bytes into account when balancing socknal conn (2817)
5860 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
5861 - reconnect unlink llog connection after MDS reconnects to OST (2816)
5862 - remove little-endian swabbing of llog records (1987)
5863 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
5864 - retry reposting request buffers if they fail (1191)
5865 - grow extent at grant time to avoid granting a revoked lock (2809)
5866 - lock revoke doesn't evict page if covered by a second lock (2765)
5867 - disable VM readahead to avoid reading outside lock extents (2805)
5869 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
5870 - updated kernel patches for hp-2.4.20 kernel (2681)
5872 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
5875 - fix truncated write corruption (2366)
5876 - fix for failed assertion in iopen_connect_dentry (1792,2517)
5878 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
5879 - ensure len(uuid) < 37 in lmc (1171)
5880 - fix ia64 OOPS in llog_test (2255)
5881 - zero end of page at obdfilter for partial page writes (2648)
5882 - don't leave stale dentries around after renames (bug 2428)
5883 - fix timeouts when evicting a client with a single lock held (2642)
5884 - set deadline for the initial HELLO message to drain (2634)
5885 - print out dotted-quad IP addresses in the socknal (2302)
5887 - additional debugging for MDS client eviction problem (2443)
5888 - fix mkfsoptions support for osts (2603, 2604)
5890 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
5893 - add series for the vanilla 2.6.0 kernel
5894 - add series for the vanilla 2.4.24 kernel
5895 - add series for a cray x86/64 UL kernel drop
5896 - fix xattr patches for the vanilla 2.4.19 series
5898 - generate true UUIDs in lmc (1171)
5899 - have portals stack dumping break in UML (2466)
5900 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
5901 - allocate with _NOFS in ldlm to avoid deadlock (1933)
5902 - wake callback waiting threads on client eviction (2460)
5903 - Add --ptldebug and --subsystem to lmc (1719)
5904 - update assertion to allow safe interrupt allocation
5905 - set rq_no_resend for cancel requests (2432)
5906 - recalculate ptlrpcd timeout after resend (2494)
5907 - call vfs_rmdir when removing pending directories (2368)
5908 - fix renaming a file to itself (2429)
5909 - lmc creates a default one-stripe lov (2454)
5910 - expand procfs space to handle large clusters (2326)
5911 - increase UML stack to avoid overflow
5912 - update lconf's list of debug and subsystem masks
5913 - fix lfs find --obd (2510)
5914 - /proc tunable for disabling filter read caching (2591)
5915 - stop rpm packages from altering slapd.conf (2301)
5916 - disable nagle in the socknal under 0conf (2578)
5917 - choose mds inode size based on stripe count (2572)
5918 - fix kernel-source rpm problems (2516)
5920 - add --disable-doc to avoid pdf generation (2421)
5921 - update documentation, tests, type-os, comments
5922 - avoid format warnings on ia64
5923 - remove the TOE NAL
5924 - tiny code cleanups by removing unused fields
5926 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
5929 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
5930 - remove the most blatant lies from BUILDING file (2371)
5931 - change default debug level to reasonable production setting
5932 - reduce client side cache size to reduce cache flush time
5933 - reduce max RPCs in flight to avoid unnecessary file fragmentation
5934 - make TCP zerocopy and pinger support enabled by default (2476)
5935 - sync writes completed after process exits caused crashes (2319)
5936 - maintain correct mount count on the MDS (2356)
5937 - backout 1557, because 2316 wasn't really fixed
5938 - better file I/O statistics gathering in /proc
5939 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
5940 - another kernel patch to fix zero-copy TCP function export
5941 - don't take duplicate lock when processing re-sent getattr (2420)
5942 - lctl uses obd_self_export instead of creating new conn (2353)
5943 - MDS/OST recovery case which requires object creation asserted (2425)
5944 - move lfs from /usr/sbin to /usr/bin in packages
5945 - fix race between mds_client_add and mds_client_free (2417)
5946 - use kmalloc instead of slabs in portals (2430)
5947 - don't create duplicate records when a failover MDS is present (2442)
5948 - remove unnecessary mount age check (2332)
5949 - don't remove directory inodes from locks prematurely (2451)
5950 - don't break if MDS service name is the same as hostname (2103)
5951 - fix races in client write RPC generation when cache full (2482)
5953 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
5956 - remove now-unused request->rq_obd (278)
5957 - if an allocation fails, print out how much memory we've used (1933)
5958 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
5959 - add the "configurable stack size" patch to most series files (1256)
5960 - ability to write large log records, for 100+ OST configs (2306)
5961 - fix NULL deref when filter_prep fails (2314)
5962 - fix operator precedence error in filter_sync
5963 - dynamic allocation of socknal TX descriptors (2315)
5964 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
5965 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
5966 - ensure that CWARN messages in Portals always get to the syslog
5967 - __init/__exit are not for prototype decls (ldlm_init/exit)
5968 - x86-64 compile warning fixes
5969 - fix gateway LMC keyword conflict (2318)
5970 - fix MDS lock inversions in getattr/reint paths (1844)
5971 - fix a rare lock re-ordering bug, which caused deadlock (2322)
5972 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
5973 - DLM race condition prevented some lock evictions (2328)
5974 - ENOMEM detection and retry on socknal sends (2230)
5975 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
5976 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
5977 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
5978 - better error messages when a client is rejected during recovery (1505)
5979 - avoid cancelling locks which were never granted, after failure (2330)
5980 - fix i_sem/journal inversion in mds_client_add (2333)
5981 - fix truncate/getattr lock cycle deadlock (2334)
5982 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
5983 - fix two rare exit paths which could leak an l_lock() ref (2321)
5984 - fencepost error in MDS/OST orphan recovery (2226)
5985 - make log record alignment 8 bytes (1988)
5986 - lstripe now fails when requested offset > ost_count (2237)
5987 - ensure that all kernel series have a complete list.h (1607)
5988 - fix crashes in special-file operations (2316)
5989 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
5991 - allow configurable automake binary, for testing new versions
5992 - small update to the lfs documentation
5994 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
5996 * fix negative export reference count in fsfilt_sync (2312)
5998 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
5999 * release candidate 0.9.1
6001 - orphans are moved into the PENDING directory for possible recovery
6002 - replayed opens now open by fid for orphan/rename safety (1042)
6003 - last close of an orphan inode generates a transno (683)
6004 - chdir() and mount() now pin the directory entry (1020)
6005 - avoid CERROR in normal ll_setattr_raw() error case (1500)
6006 - discard very old requests without processing them (1502)
6007 - remove some common, well-understood CERRORs (1505)
6008 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
6009 - clear "grant" flags in OST replies until OST grant code lands (1644)
6010 - fix read performance by not clobbering i_blksize on client (1598)
6011 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
6012 - make LRU size a /proc tunable, clears locks when reduced (707)
6013 - fix some lprocfs rot that prevented ptlbd from loading (1732)
6014 - server locks take references on exports now (1558)
6015 - build fixes for 2.4.20-rh trees (1663)
6016 - return an error from lov_create if all OSCs are inactive (1751)
6017 - fix import levels when a reconnect happens without a timeout (1597)
6018 - exit early from mds_open if we get a lookup error (1749)
6019 - partial page read at EOF wouldn't wait for disk before sending (1642)
6020 - avoid NULL deref in obdfilter when reading page past EOF (1592)
6021 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
6022 - fix LBUG in ll_it_open_error with rc = -2 (1861)
6023 - write/truncate lock inversion (1639)
6024 - Don't auto-load obdclass, portals modules during cleanup (1495)
6025 - fix timestamps from jumping to "now" (1763)
6026 - extra journal assertions (1648)
6027 - add an extra multiunlink test (1771)
6028 - fix read_record/write_record API (1776)
6029 - fix leak of offset_extent, possible incorrect i_size later (1772)
6030 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
6031 - Debugging for the kqswnal_get_idle_tx problems (1820)
6032 - Allow recovery to be attempted multiple times (1536)
6033 - Write out MDS last_rcvd file after it is first created (1600)
6034 - Fix tx_descriptor leak in failed transmit situations (1827)
6035 - ext3 journaling fixes for assertion failure after IO error (1871)
6036 - class_export_put() on freed export after completion AST error (1896)
6037 - Fix revalidate looping in VFS (1322)
6038 - Don't access a freed export during MDS_REINT timeout (1521)
6039 - Add open-unlink recovery support on the MDS (1673,1764)
6040 - Return an error if no MDS data was read from last_rcvd (1946)
6041 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
6042 - Don't setup a disk device that doesn't match exported UUID (317)
6043 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
6044 - avoid committing NULL handle in force close
6045 - local.sh is now a one-stripe LOV configuration
6046 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
6047 - fix ext3 htree duplicate directory entry corruption (1516)
6048 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
6049 - update to new LOV EA format (2097)
6050 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
6052 2003-06-15 Phil Schwan <phil@clusterfs.com>
6055 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
6056 - per-import/export recovery handling (958, 931, 959)
6057 - multiple last-rcvd slots, for serving multiple FSes (949)
6058 - connections are again shared between multiple imp/exports (963, 964)
6059 - "umount -f" would hang if any requests needed to be sent (393, 978)
6060 - avoid pinning large req buffer by copying for queued messages (989)
6061 - add "uuid" to "lctl device" command to help upcalls (991)
6062 - "open" RPCs with transnos would confuse recovery counters (1037)
6063 - do proper endian conversion of all wire messages (288, 340, 891)
6064 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
6065 - call waiting locks callback from LDLM recovery thread (1127, 1151)
6066 - fix ptlrpc_connection leak in target_handle_connect (1174)
6067 - fix import refcounting bug in OST and MDS cleanup (1134)
6068 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
6069 - fix very unlikely obd_types race condition (501)
6070 - remove osc_open hack for echo_client (1187)
6071 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
6072 - a failure in read_inode2 leads to deadlock (1139)
6073 - cancel ack-locks as soon as transaction is committed (1072)
6074 - fix major leaks and crashes in the bulk I/O path (937, 1057)
6075 - make sure to commitrw after any preprw to avoid deadlock (1162)
6076 - failing to execute a file in a lustre FS would lock inode (1203)
6077 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
6078 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
6079 - obd_uuid2tgt would walk off the end of the list (1255)
6080 - on IA64 the osc would give portals incorrect bulk size (1258)
6081 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
6082 - fix lock inversion caused by new llite matching code (1282)
6083 - limit the number of dirty pages on a client to 10MB (1286)
6084 - timed out locks were not being corrected cancelled (1289)
6085 - fix O_DIRECT above 4GB on IA-32 (1292)
6086 * major user-visible changes
6087 - fail out/fail over policy now controlled by the upcall (993)
6089 - add OBD_PING to check server availability and failure (954)
6090 - lustre messages are now sent in sending host order (288, 340, 891)
6091 - add eadatalen to MDS getattr reply (340)
6092 - OST read replies may contain second buffer, with per-page status (593)
6094 2003-03-11 Phil Schwan <phil@clusterfs.com>
6097 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
6098 - failed open()s could cause deadlock; fixed (867, 869)
6099 - stop cancelling OST locks when files are closed (481)
6100 - overlapping XID spaces caused network corruption (851, 853)
6101 - fix unsafe fsfilt counter arithmetic; change to atomic_t
6102 - setattr_raw added, to do single-RPC, server-side setattrs
6103 - lmc/lconf syntax change for OST UUIDs
6104 - fix crashy race condition between ptlrpc_free_req and osc_close
6105 - don't use request in mdc_enqueue if we hit a timeout (889)
6106 - don't set the inode i_size for regular files from the MDS (896)
6107 - handle out of order completion AST (842)
6108 - don't LBUG if a lock request times out after receiving AST (913)
6109 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
6110 - if a bad lock AST arrives, send an error instead of dropping entirely
6111 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
6112 - fix leak in bulk IO when only partially completed (899, 900, 926)
6113 - fix O_DIRECT for ia64 (55)
6114 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
6115 - C-z after timeout could hang a process forever; fixed (977)
6117 - client-side I/O cache (678, 924, 929, 941, 970)
6119 - READPAGE and SETATTRs which don't take server-side locks get
6122 2003-02-11 Phil Schwan <phil@clusterfs.com>
6125 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
6126 - Fix fsfilt_extN_readpage() to read a full page of directory
6127 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
6128 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
6129 - Fix imbalanced LOV object allocation and out-of-bound access (469)
6130 - Most intent operations were removed, in favour of a new RPC mode
6131 that does a single RPC to the server and bypasses most of the VFS
6132 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
6133 - Aggressively cancel local locks on DLM servers
6134 - mds_reint_unlink sends EA to the client if it's the last nlink.
6135 client uses that EA to unlink OST objects.
6136 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
6137 - recursive symlinks were fixed (439)
6138 - fixed NULL deref in DEBUG_REQ
6139 - filter_update_lastobjid no longer calls sync, which annoyed extN
6140 - fixed multi-client small-writes to a single file problem (445)
6141 - fixed mtime updates during file writes (607)
6142 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
6143 - fixed bug in obd_brw_read/write() (under guise of testing 367)
6144 - fixed Linux OST size reporting problem (444, 656)
6145 - OST now updates object mtime with writes or setattr (607, 619)
6146 - client verifies file size before zeroing page past EOF (445)
6147 - OST now writes last allocated objid to disk with allocation (108)
6148 - LOV on echo now works (409)
6150 - mds_reint_unlink sends a new buffer, with the EA included. this
6151 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
6152 set if a regular file was being unlinked, and it was the last link
6153 - use PtlGet from the target for bulk writes (315)
6154 - OST now updates object mtime with writes or setattr (607, 619)
6155 - LDLM now has a grant-time callback to revalidate locked items, if
6157 - Many MDS operations were reorganized to combat race conditions
6159 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
6160 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
6162 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
6165 - Fully reactivate OST imports after reconnection (512, others)
6166 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
6167 - More graceful error handling for truncating on dead OST (515)
6168 - Don't error out unless we're actually accessing dead stripes (474)
6169 - Fix garbage sizes when stripes are missing (410)
6170 - LRU counters were broken, causing constant lock purge (433, 432)
6171 - garbage on read from stripes with failed OSTs (441)
6172 - mark OSCs as active before reconnecting during recovery (438)
6173 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
6174 - lfind did not preserve OST order in output (443)
6175 - symlinks cause hung clients, incorrect data (439)
6176 - stop dereferencing request after dropping refcount (457)
6177 - don't LASSERT(spin_is_locked) on non-SMP (455)
6178 - fixes for many rename() bugs
6179 - fstat didn't correctly synchronize attributes (399)
6180 - server must handle lock cancellation during blocking AST prep (487)
6181 - bulk descriptors were free()d too soon (511)
6182 - fix paths in lconf, which would load incorrect modules (451, 507)
6183 - fix confusing lconf 'host not found' error message (386)
6184 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
6185 - fix race condition in mdc_blocking_ast() for inode access (526)
6186 - fix lov_unpackmd() unpacking wrong number of stripes (537)
6187 - fix lov_set_osc_active() marking wrong OSC inactive (440)
6188 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
6189 - fix multiple writes of stripe MD to MDS (358, maybe 519)
6190 - fix lstripe in several ways (kernel side) (527)
6191 - fix request leak in ldlm_cli_enqueue (262)
6192 - incorrect OSC was marked inactive after OST failure
6193 - call mds_fs_cleanup before unmounting filesystem (524)
6194 - fix races between taking ns_lock and ldlm_lock_change_resource
6195 - fix races updating LOV export open file list
6196 - fix lov_enqueue error path, avoid decref-ing bad lock handle
6197 - fix recovery NULL deref in ldlm_cli_cancel_unused
6198 - fix some DLM races by using new hash table for lock handles (419)
6199 - permit the client to specify desired inodes, at replay
6200 - duplicate requests when we queue them for replay reintegration
6201 - fix last_rcvd offset calculation
6202 - sync after each recovered transaction, so we always make progress
6203 - never, not always, ERESTART requests without transnos
6204 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
6206 - skip replay if the MDS says that the client is already connected
6207 - don't check for a recovery-enabled export to match lctl's UUID
6208 - don't INC_USE_COUNT for phantom exports
6209 - don't crash when cleaning up phantom exports (567)
6210 - don't double-finish or set replay data for errored mdc_open requests
6211 - abort requests when they time out, so we don't get old replies
6212 - send/receive replies for AST messages again
6213 - if the client says that it doesn't have the lock, cancel it on the
6215 - if we timeout during I/O, don't try to cancel an in-use lock; instead
6216 mark it as destroyed, it will all work out when decref is called
6217 - fix module use counts (22, 581)
6219 - ASTs now expect a reply (server cancels lock on error reply)
6221 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
6224 - fix many simultaneous client startup (392)
6225 - fix dentry->d_it clobbering
6226 - credentials weren't being shipped for readdir/getattr operations
6227 - remove invalid assertions triggered during some concurrent MD
6229 - proper Lustre versions added (336, 389)
6230 - fix memory leak for create error case (398)
6231 - fix LOV locking bug that would get cli/srv out of sync
6232 - fix echo client over LOV (409)
6233 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
6234 - fix double-O_EXCL intent crash (424)
6235 - avoid sending multiple lock CANCELs (352)
6237 - MDS can do multi-client recovery (modulo bugs in new code)
6239 - many updates, edits, cleanups
6241 2002-11-18 Phil Schwan <phil@clusterfs.com>
6244 - fix null d_it dereference (346)
6245 - fix full OST/dbench hang (333)
6246 - fix permission problem with file removal (286)
6247 - fix removal of OSCs from LOV when they fail
6248 - fix NULL deref during bulk timeout (214)
6249 - fix problems related to multiple filesystems on one MDS (241)
6250 - fixed serious subtle metadata locking bugs
6251 - free locks on clients when inodes are removed due to memory
6253 - fix inode pointer in lock data (285)
6254 - partial support for multiple MDS on a single host (241)
6255 - data locks weren't cancelled at clear_inode time (290, 311)
6256 - intent locks could lead to unbounded lock growth (205)
6257 - added a maximum lock count, an LRU list, and a flusher
6258 - fix multiple rename (365)
6259 - properly abstracted the echo client
6260 - OSC locked 1 byte too many; fixed
6261 - rewrote brw callback code:
6262 - fixed recovery bugs related to LOVs (306)
6263 - fixed too-many-pages-in-one-write crash (191)
6264 - fixed (again) crash in sync_io_timeout (214)
6265 - probably fixed callback-related race (385)
6267 - Add capability to MDS protocol
6268 - LDLM cancellations and callbacks on different portals
6270 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
6273 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
6274 - defer open object destruction to close time (601981, 138)
6275 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
6276 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
6277 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
6278 - don't call obd_setattr in ll_file_release for destroyed objects
6279 * protocol change to lustre_msg: move |version| and add |flags|
6280 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
6281 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
6282 * added replay of create, unlink, link and rename operations during
6283 MDS failover; recovery should be much more robust now
6284 * remove failed OSCs from LOVs (only lov_create uses this so far)
6285 * the lustre-HOWTO was brought (more) up to date (582544)
6287 2002-10-23 Phil Schwan <phil@clusterfs.com>
6290 - in-use dentries weren't being reused properly (617851)
6291 - prevent multiple LDLM setup (599178)
6292 - fix LOV size calculations for truncate (617853)
6293 - fix client handling of MDS intent errors (POSIX)
6294 - fix permission bug in lovstripe.c test (624321)
6295 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
6296 - truncate past end of file could corrupt data
6297 - proper cleanup after timeouts, crashes, etc (592524, 550815)
6298 - a race in recovery could return ETIMEDOUT to apps (623947)
6299 - building outside the source directory was fixed
6300 * the lustre-HOWTO was brought (more) up to date (582544)
6301 * major progress was made on recovery functionality
6303 2002-10-10 Phil Schwan <phil@clusterfs.com>
6306 - recovery deadlock fix
6307 - rm -rf causes LBUG fix (617817)
6308 - file open by multiple tasks fix (618962)
6309 - directory permissions bugs (602707 and 620007)
6310 - journal_stop fixed with locking (611313)
6311 - O_APPEND failures resolved (618273, perhaps 614459)
6312 - lconf PATH fix (619770)
6313 - IA64 build fix (621450)
6314 - RPC buffer sizes scale with amount of memory
6316 2002-10-01 Phil Schwan <phil@clusterfs.com>
6319 - locks would be cancelled without throwing away data pages,
6320 resulting in inconsistent data (605627)
6321 - inode attributes were not always being refreshed (605627, 612449)
6322 - lconf now continues to cleanup after lctl reports an error
6323 - MDS now enforces user permissions (602707)
6324 - lprocfs cleanup fixed, but not yet enabled (614157)
6325 - fixed infinite server hang, should a client not respond to an AST
6326 - avoid going into recovery if user calls readlink() with a buffer
6327 that's too small (613941)
6328 - AST RPCs no longer require replies (614867) -- this may be changed
6329 - don't crash server if client sends an IOV that's too big (611336)
6330 - fixed lock conversion deadlock (611892)
6331 - fixed the following of symlinks (614622)
6332 * recovery: the server can remove locks from a client that dies, other
6333 clients can make progress
6334 * more extN patch fixes
6335 * compile-time configurable ptlrpc buffer allocations
6337 - collaborative read cache document
6338 - Lustre Lite Performance CDR document-in-progress
6340 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6343 - fix typo in patch-2.4.18
6345 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6348 - clear ptlrpc request each time in handle_incoming_request()
6349 - unlink of files now destroys the object on the OST
6351 2002-09-19 Peter Braam <braam@clusterfs.com>
6353 * add hard link support
6354 * change obdfile creation method
6355 * kernel patch changed
6357 2002-09-19 Peter Braam <braam@clusterfs.com>
6360 - stack overflow bug in extN fixed
6362 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
6364 * documentation updates
6365 - add man pages for config tools
6366 - update tests/README to describe testing with new config tools
6367 - finish metadata API descriptions
6368 * bug fixes and cleanups
6369 - statfs workaround for 16TB limit
6370 - LOV stripe allocation improved, can stripe on subset of OSTs
6371 - LOV file size/IO offset was wrong for files > 4GB in size
6372 - object EA data was being dropped, caused files to be unreadable
6373 - memory overflow with non-LOV OST caused memory corruption
6374 - fixed regression tests to work with new config tools, obdfilter
6375 - fixed bug when directory size became larger than 1 block
6376 - fixed bug (for single client case) when PWD was deleted
6377 - invalidate local directory pages when doing intent-based ops
6378 - avoid LDLM oops when lock callback contained bad data
6380 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
6382 * documentation updates
6383 * bug fixes and cleanups
6384 - configuration tools
6387 - 64-bit compile warnings
6388 - 64-bit internal statfs data
6390 * test_brw on persistent OST devices
6392 * lprocfs (disabled)
6394 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
6396 * documentation updates
6397 * bug fixes and cleanups
6398 * configuration tools
6400 2002-08-30 Peter J. Braam <braam@clusterfs.com>
6403 * many small fixes to 0_5_4
6404 * io/network handling
6405 * thinkos in MDS operations
6407 2002-08-24 Peter J. Braam <braam@clusterfs.com>
6410 * crucial basic fixes to 0.5.3
6411 * IOR, Iozone work over Elan
6414 2002-08-07 Phil Schwan <phil@clusterfs.com>
6415 * version 0_5_3, our first alpha
6416 * we use the new Portals iovs
6417 * documentation updates
6418 * bug fixes and cleanups
6419 * small changes in the DLM wire protocol
6421 2002-07-25 Peter J. Braam <braam@clusterfs.com>
6422 * version 0_5_1 with some initial stability,
6423 * locking on MD and file I/O.
6424 * documentation updates
6425 * several bug fixes since 0.5.0
6426 * small changes in wire protocol
6428 2002-07-18 Phil Schwan <phil@clusterfs.com>
6430 * delivered as Lustre Light Alpha
6431 * fixed a crash after handling invalid MDS requests
6432 * fixed directory pages for architectures with non-4k pages sizes
6434 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
6435 * release version v0_4_4
6436 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
6438 * Fixes a number of interruption problems with OST operations.
6439 * Update documentation for portals header changes
6440 * Move all wire protocol structs/defines to lustre_idl.h
6441 * Fixes symlink length bug.
6442 * Add tcpdump to repository.
6444 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
6445 * release version v0_4_3
6446 * Fixes statfs for inodes on extN.
6447 * Fixes bug in runtests which would delete /etc/hosts.
6448 * Use 64-bit object IDs wherever possible (not into VFS though)
6449 Remove ost_get_info, which is unused by lustre, and out of date.
6451 2002-07-03 Peter Braam <braam@clusterfs.com>
6452 * release version v0_4_2 Fixes a lookup error (type not passed)
6453 * move forward to head of Portals
6454 * move forward to latest Lustre kernel
6456 2002-06-25 Peter Braam <braam@clusterfs.com>
6457 * release version v0_4_1. Hopefully stable on single node use.