3 * See http://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
4 * Server support for kernels:
5 2.6.32-431.11.2.el6 (RHEL6)
6 3.0.101-0.21 (SLES11SP3)
7 * Client support for unpatched kernels:
8 2.6.32-431.11.2.el6 (RHEL6)
9 3.0.101-0.21 (SLES11SP3)
10 2.6.18-348.3.1.el5 (RHEL5)
11 * Recommended e2fsprogs version: 1.42.9.wc1
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 10-31-2013 Intel Corporation
22 * See http://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
23 * Server support for kernels:
24 2.6.32-358.18.1.el6 (RHEL6)
25 * Client support for unpatched kernels:
26 2.6.32-358.18.1.el6 (RHEL6)
28 2.6.18-348.3.1.el5 (RHEL5)
29 2.6.16.60-0.69.1 (SLES10)
31 * Recommended e2fsprogs version: 1.42.7.wc1
32 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
33 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
34 stack overflow. For more information, please refer to bugzilla 17630.
35 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
36 the re-exporting nodes to be mounted with "32bitapi" mount option
38 --------------------------------------------------------------------------------
40 05-30-2013 Intel Corporation
42 * See http://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
43 * Server support for kernels:
44 2.6.32-358.6.1.el6 (RHEL6)
45 * Client support for unpatched kernels:
46 2.6.32-358.6.1.el6 (RHEL6)
48 2.6.18-348.3.1.el5 (RHEL5)
49 2.6.16.60-0.69.1 (SLES10)
51 * Recommended e2fsprogs version: 1.42.7.wc1
52 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
53 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
54 stack overflow. For more information, please refer to bugzilla 17630.
55 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
56 the re-exporting nodes to be mounted with "32bitapi" mount option
58 --------------------------------------------------------------------------------
60 9-30-2012 Whamcloud, Inc.
62 * See http://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
63 * Server support for kernels:
64 2.6.32-279.5.1.el6 (RHEL6)
65 * Client support for unpatched kernels:
66 2.6.32-279.5.1.el6 (RHEL6)
68 2.6.18-238.19.1.el5 (RHEL5)
69 2.6.16.60-0.69.1 (SLES10)
70 * Recommended e2fsprogs version: 1.41.90.wc4
71 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
72 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
73 stack overflow. For more information, please refer to bugzilla 17630.
74 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
75 the re-exporting nodes to be mounted with "32bitapi" mount option
77 Severity : enhancement
79 Description: Update RHEL6.3 kernel to 2.6.32-279.5.1.el6.
81 Severity : enhancement
83 Description: Update RHEL6.3 kernel to 2.6.32-279.2.1.el6.
85 Severity : enhancement
87 Description: Update RHEL6.2 kernel to 2.6.32-220.17.1.el6
89 Severity : enhancement
91 Description: Update RHEL6.2 kernel to 2.6.32-220.13.1.el6
93 Severity : enhancement
95 Description: Update RHEL6.2 kernel to 2.6.32-220.7.1.el6
97 --------------------------------------------------------------------------------
99 04-30-2012 Whamcloud, Inc.
101 * See http://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
102 * Server support for kernels:
103 2.6.32-220.4.2.el6 (RHEL6)
104 * Client support for unpatched kernels:
105 2.6.32-220.4.2.el6 (RHEL6)
107 2.6.18-238.19.1.el5 (RHEL5)
108 2.6.16.60-0.69.1 (SLES10)
109 * Recommended e2fsprogs version: 1.41.90.wc4
110 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
111 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
112 stack overflow. For more information, please refer to bugzilla 17630.
113 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
114 the re-exporting nodes to be mounted with "32bitapi" mount option
116 Severity : enhancement
118 Description: Update RHEL6.2 kernel to 2.6.32-220.4.2.el6
120 Severity : enhancement
122 Description: Update RHEL6.2 kernel to 2.6.32-220.4.1.el6
124 Severity : enhancement
126 Description: Update RHEL6.2 kernel to 2.6.32-220.el6
128 Severity : enhancement
130 Description: Update RHEL6.1 kernel to 2.6.32-131.17.1.el6
132 --------------------------------------------------------------------------------
134 09-30-2011 Whamcloud, Inc.
136 * Server support for kernels:
137 2.6.18-238.19.1.el5 (RHEL5)
138 2.6.32-131.6.1.el6 (RHEL6)
139 * Client support for unpatched kernels:
140 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
141 2.6.18-238.19.1.el5 (RHEL5)
142 2.6.32-131.6.1.el6 (RHEL6)
143 2.6.29.4-167.fc11 (FC11)
144 2.6.16.60-0.69.1 (SLES10)
146 * Recommended e2fsprogs version: 1.41.90.wc3
147 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
148 Lustre filesystem with 4K stack may cause a stack overflow. For more
149 information, please refer to bugzilla 17630.
150 * NFSv4 reexport requires Lustre client on the re-exporting nodes to be
151 mounted with "32bitapi" mount option
153 Severity : enhancement
154 Description: Update RHEL6.1 kernel to 2.6.32-131.6.1.el6
156 Severity : enhancement
157 Description: Update RHEL5.6 kernel to 2.6.18-238.19.1.el5
159 Severity : enhancement
160 Description: Update RHEL5.6 kernel to 2.6.18-238.9.1.el5
163 Frequency : only with SLES10
165 Description: Use OFED "KMP" provided by Novell
166 Details : SLES10 SP3 ships with OFED in a separate "KMP" package.
167 Lustre is now built against this package. That means you need to
168 install the ofed-kmp package from Novell for the patchless client.
169 Note that the ofed-kmp that Novell ships may not exactly match the
170 kernel version but should still be compatible.
172 Severity : enhancement
174 Description: Update RHEL5.5 kernel to 2.6.18-194.17.1.el5
175 Update OEL5.5 kernel to 2.6.18-194.17.1.0.1.el5.
177 --------------------------------------------------------------------------------
179 2010-07-15 Oracle, Inc.
181 * Server support for kernels:
182 2.6.18-164.11.1.el5 (RHEL 5)
183 2.6.18-164.11.1.0.1.el5 (OEL 5)
184 * Client support for unpatched kernels:
185 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
186 2.6.18-164.11.1.el5 (RHEL 5),
187 2.6.18-164.11.1.0.1.el5 (OEL 5)
188 2.6.16.60-0.42.8 (SLES 10),
190 2.6.29.4-167.fc11 (FC11)
191 * Recommended e2fsprogs version: 1.41.10-sun2
193 Severity : enhancement
195 Description: Distribute lfs_migrate script with lustre
196 Details : The lfs_migrate mentioned in the manual is now included in the
197 lustre tree in lustre/scripts and will distributed in lustre RPMs.
199 Severity : enhancement
201 Description: Failover nid list added to import proc file.
203 Severity : enhancement
205 Description: "som_preview" mount option required to use SOM
206 Details : The clients will need to use som_preview mount option
210 Description: Lustre does not do 1MB IOs to HW RAID
211 Details : Bump MAX_PHYS/HW_SEGMENTS and SG_ALL to 256 in the RHEL5 kernel.
213 Severity : enhancement
215 Description: lfs quota output is not very convenient for awk/sed-parsing
216 Details : Some positions in lfs quota output table could be empty or
217 non-empty which made it hard to parse it with scripts, now
218 a dash is put instead of space where there is not supposed
223 Description: Update to OFED 1.5.1
224 Details : OFED 1.5.1 will be used in environments that don't use
225 vendor supplied OFED.
229 Description: Remove lnet. prefix from lctl params display
230 Details : Also renamed lnet.memused to lnet_memused to
235 Description: Use conf_param -d to remove permanant settings.
239 Description: Add list_param -R
240 Details : lctl list_param now supports recursive mode.
243 Bugzilla : 21829, 22312
244 Description: Fix llobdstat.
245 Details : 1. Now updating the write_bytes counter
246 2. Port script fixes from 1.8.
250 Description: decrease the usage of memory on clients.
251 Details : 1. On clients, recycle dentries and inodes unused.
252 2. Delete the code related to ll_deathrow(att 6215 in bug 1443). It
257 Description: llobdstat fix and enhancement
258 Details : add a counter to set a limit to how many samples will be returned
259 fix a wildcard in the path to limit to obdfilter stats only
263 Description: lfs check only allowed for root
264 Details : Code cleanups and added a regresion test of non-root use of
269 Description: MDS Oopses in ptlrpc_dispatch_difficult_reply()/__list_add()
270 Details : Fix hr->hr_index overload.
274 Description: Fix FIEMAP functionality.
278 Description: FC11 patchless client support added
282 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and
283 OEL5.4 kernel to 2.6.18-164.11.1.0.1.el5.
287 Description: New acc-sm script and various test-framework enhancments.
291 Description: Previously landed patch was reverted.
295 Description: Update SLES10 kernel to 2.6.16.60-0.42.8 (patchless)
299 Description: Add ldiskfs maxdirsize mount option.
303 Description: Fix mkfs.lustre support for 16TB LUNs.
307 Description: Print mmp_check_interval
308 Details : print mmp_check_interval and make it possible to abort
309 mount operation in case it takes too long.
313 Description: Add cascading_rw to lustre mpi tests.
317 Description: Store SOM data in EA.
318 Details : Also several SOM related cleanups.
322 Description: Add refcount to md_open_data
326 Description: Use refcount on lov_request_set to prevent memory corruption.
330 Description: Statahead: use parent dir's i_mutex for synchronization.
334 Description: loadgen improvements.
338 Description: Enable peer health detection in ksocklnd and ko2iblnd.
342 Description: Control DCACHE_LUSTRE_INVALID flag with MDS_INODE_LOOKCUP lock
346 Description: Fix race on obd_devcie:md_stats
350 Description: Fix SLES 11 client builds.
351 Details : Resolved uninitalized pointer gcc warnings.
355 Description: Allow non-root access for "lfs check".
356 Details : Added a check in obd_class_ioctl() for OBD_IOC_PING_TARGET.
360 Description: Clear imp_force_reconnect correctly in ptlrpc_connect_interpret()
362 Severity : enhancement
364 Description: Update kernel to RHEL5.4 2.6.18-164.6.1.el5 and
365 OEL5 2.6.18-164.6.1.0.1.el5(Both in-kernel OFED enabled).
369 Description: Sanity test_27p fails.
370 Details : OBD_FAIL_OST_ENOSPC has been moved to filter_statfs().
371 Sanity tests 27 were cleaned up and improved.
375 Description: Unify req format on client/servers
376 Details : Use new req_capsule API [almost] everywhere instead of old PTLRPC
377 buffers and swabbers approach..
382 Description: drop join file stuff.
384 Severity : enhancement
386 Description: Add 2.6.27 and SLES11 (patchless client) support
390 Description: Fix in ptlrpc_expire_one_request() to print the signed time
393 Severity : enhancement
395 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
398 Severity : enhancement
400 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
402 Severity : enhancement
404 Description: correctly handle big reply message.
405 Details : send LNet event if reply is bigger then buffer and adjust this buffer
410 Description: Drop unnecessary __GFP_NOMEMALLOC flag from filter_get_page()
412 Severity : enhancement
414 Description: Add support for OEL5.
418 Description: Conf-sanity.sh 50g test - deactivated OST should not cause a panic.
420 Severity : enhancement
422 Description: "weak-modules" support
423 Details : Implement "weak-modules" support which enables kernel modules
424 to be used with any kernel that implements the same kABI. In
425 order to achieve this modules are now installed in
426 /lib/modules/$(uname -r)/updates/kernel on all distributions.
430 Description: client could not reconnect to OST because of an active request.
431 Details : The client now retries to reconnect to the same server, if a
432 connect request failed with EBUSY or -EAGAIN. Test case 35b
433 conf-sanity.sh has been created.
437 Description: Don't allow make backward step on assignin osc next id.
438 Details : race between allocation next id and ll_sync thread can be cause
439 of set wrong osc next id and can be kill valid ost objects.
443 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
447 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
451 Description: Error handling in osc_statfs_interpret() has been improved.
452 Details : Check in osc_statfs_interpret() for EBADR.
456 Description: Do not send statfs() requests to OSTs disabled by administrator.
457 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
459 Severity : enhancement
461 Description: Added sanity-quota test to for lfs(1) to work with numeric uid/gid.
463 Severity : enhancement
465 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
467 Severity : enhancement
469 Description: Enhancement for lfs(1) command to use numeric uid/gid.
473 Description: Improved cleanup procedure for conf-sanity test_46a.
477 Description: Do not start bulk transfer, if client tried to reconnect.
478 Details : Do not start bulk transfer, if client tried to reconnect,
479 but failed due to active rpcs. Check for exp_abort_active_req
480 has been implemented.
484 Description: Prevent inconsistences between linux and lustre mount structures.
485 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
486 Make the sleep interruptible.
488 Severity : enhancement
490 Description: provide server to client comms path
491 Details : server to client communications path via new LDLM_SET_INFO rpc
495 Description: truncate starts GFP_FS allocation under transaction causing deadlock
496 Details : ldiskfs_truncate calls grab_cache_page which may start page
497 allocation under an open transaction. This may lead to
498 calling prune_icache with consequent lustre reentrance.
502 Description: send correctly lsm on open replay
503 Details : MDS is trust to LSM size on replay open, but client can set wrong size
508 Description: enable client interface failover
509 Details : When a child reconnects from another NID, properly update export
510 nid hash position and ldlm reverse import.
512 Severity : enhancement
514 Description: Communicate OST degraded/readonly state via statfs to MDS
515 Details : Flags in the statfs returned from OSTs indicate whether the
516 OST is in a degraded RAID state, or if the filesystem has
517 turned read-only after a filesystem error is detected.
522 Description: don't panic if EPROTO was hit when reading symlink
523 Details : correctly handling request reference in error cases.
525 Severity : enhancement
527 Description: Add LustreNetLink, a kernel-userspace communcation path.
529 Severity : enhancement
531 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
536 Description: access to llog context before init.
537 Details : move handling CATALOGS file at osc layer and forbid access to llog
541 Frequency : rare, only when client is writing to OST on local node
543 Description: Avoid deadlock for local client writes
544 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
545 memory freeing context. This allows OST threads to set the
546 PF_MEMALLOC flag on task structures in order to allocate memory
547 from reserved pools and complete IO.
548 Use GFP_HIGHUSER for OST allocations for non-local client writes,
549 so that the OST threads generate memory pressure and allow
550 inactive pages to be reclaimed.
552 Severity : enhancement
554 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
559 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
560 Details : move ldlm namespace creation in setup phase to avoid grab
561 _lprocfs_lock with cli_sem held.
565 Description: Temporarily disable grant shrink.
566 Details : Disable the feature for debugging.
569 Frequency : only when formatting test filesystems
571 Description: Unable to run several mkfs.lustre on loop devices at the same time
572 Details : mkfs.lustre returns error 256 on the concurrent loop devices
573 formatting. The solution is to proper handle the error.
575 Severity : enhancement
577 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
579 Severity : enhancement
581 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
586 Description: resolve race between obd_disconnect and class_disconnect_exports
587 Details : if obd_disconnect will be called to already disconnected export he
588 forget release one reference and osc module can't unloaded.
590 Severity : enhancement
592 Description: Allow tuning service thread via /proc
593 Details : For each service a new
594 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
595 created that can be used to set min/max thread counts, and get the
596 current number of running threads.
601 Descriptoin: don't return error if have partially created objects for file.
602 Details : lov_update_create_set uses set->set_success as index for created
603 objects, so if some requests will be failed, they will have hole
604 at end of array and qos_shrink_lsm for allocate correct lsm.
606 Severity : enhancement
608 Description: Update OFED support to 1.4.1
613 Description: Reduce small size read RPC
614 Details : Set read-ahead limit for every file and only do read-ahead when
615 available read-ahead pages are bigger than 1M to avoid small size
618 Severity : enhancement
620 Description: move AT tunable parameters for more consistent usage
621 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
623 Severity : enhancement
625 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
626 Details : allow skip disconnected ost for send statfs request and hide error
632 Description: fix racy locking of mballoc block bitmaps causing BUG
633 Details : The locking of the mballoc buddy bitmap and the in-memory
634 block bitmap was using two different spin locks in some
635 cases. This made it possible to incorrectly access the
636 mballoc bitmap while another process was modifying it,
637 causing a sanity assertion to fail. While no on-disk corruption
638 was reported, there was some risk of this happening.
641 Frequency : rare, on llog test 6
643 Descriptoin: don't allow connect to already connected import
644 Details : allowing connect to already connected import is hide connecting problem.
649 Descriptoin: allow kill process which wait statahead result
650 Details : for some reasons 'ls' can stick in waiting result from statahead,
651 in this case need way for kill this process.
653 Severity : enhancement
655 Description: Add state history info file, enhance import info file
656 Details : Track import connection state changes in a new osc/mdc proc file;
657 add overview-type data to the osc/mdc import proc file.
659 Severity : enhancement
661 Description: MDS create should not wait for statfs RPC while holding DLM lock.
663 Severity : enhancement
665 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
670 Descriptoin: OST grant too much space to client even there are not enough space.
671 Details : Client will shrink its grant cache to OST if there are no write
672 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
673 this grant cache if there are already not enough avaible space
674 (left_space < total_clients * 32M).
677 Frequency : start MDS on uncleanly shutdowned MDS device
679 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
680 Details : stay in waiting mds<>ost recovery finished produce random bugs
681 due race between two ll_sync thread for one lov target. send
682 ACTIVATE event only if connect realy finished and import have
686 Frequency : rare, connect and disconnect target at same time
688 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
689 Details : don't call obd_disconnect under lov_lock. this long time
690 operation and can block ptlrpcd which answer to connect request.
695 Descriptoin: don't lose wakeup for imp_recovery_waitq
696 Details : recover_import_no_retry or invalidate_import and import_close can
697 both sleep on imp_recovery_waitq, but we was send only one wakeup
701 Frequency : always with long access acl
703 Descriptoin: mds can't pack reply with long acl.
704 Details : mds don't control size of acl but they limited by reint/getattr
707 Severity : enhancement
709 Description: Update to SLES10 kernel-2.6.16.60-0.33.
711 Severity : enhancement
713 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
716 Frequency : start MDS on uncleanly shutdowned MDS device
718 Descriptoin: aborting recovery hang on MDS
719 Details : don't throttle destroy RPCs for the MDT.
722 Frequency : on remount
724 Description: external journal device not working after the remount
725 Details : clear dev_rdonly flag for external journal devices in
731 Description: shutdown vs evict race
732 Details : client_disconnect_export vs connect request race.
733 if client will evicted at this time - we start invalidate
734 thread without referece to import and import can be freed
738 Frequency : rare, need acl's on inode.
740 Description: client can't handle ost additional correctly
741 Details : if ost was added after client connected to mds client can have
742 hit lnet_try_match_md ... to big messages to wide striped files.
743 in this case need teach client to handle config events about add
744 lov target and update client max ea size at that event.
746 Severity : enhancement
748 Description: Changelogs
749 Details : Changelogs are a lightweight mechanism to track filesystem
750 metadata and namespace changes. The changelog is recorded
751 permanently on the MDTs, and is periodically "consumed" / purged
752 when records are no longer needed.
754 Severity : enhancement
756 Description: compact fld format with extents
757 Details : Store range of seq rather than every seq in FLD. Seq
758 controller update FLD rather than clients. In Case of CMD, mdt0
759 has FLD, all other metadata server act as non persistent proxy
760 for FLD queries and cache fld entries in fld cache.
765 Description: don't skip ost target if they assigned to file
766 Details : Drop slow OSCs if we can, but not for requested start idx.
767 This means "if OSC is slow and it is not the requested
768 start OST, then it can be skipped, otherwise skip it only
769 if it is inactive/recovering/out-of-space.
773 Description: more cleanup in mds_lov
774 Details : not send LOV EA under replay, we can't know about they size at this
775 time. Don't allow client connect to mds before any ost connected,
776 for avoid problems with LOV EA size and returning EIO to client.
778 Severity : enhancement
780 Description: Interoperability at server side (Disk interoperability)
782 Severity : enhancement
784 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
786 Severity : enhancement
788 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
790 Severity : enhancement
792 Description: New client IO stack (CLIO).
794 Severity : enhancement
796 Description: Commit on sharing. Eliminate inter-client dependencies between
797 uncommitted transactions by doing transaction commits.
798 Thereby clients may recovery independently.
801 Frequency : Create a symlink file with a very long name
803 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
804 Details : If there is no extra space in the request for early cancels,
805 ldlm_req_handles_avail() returns 0 instead of a negative value.
807 Severity : enhancement
809 Description: Add /proc entry for import status
810 Details : The mdc, osc, and mgc import directories now have
811 an import directory that contains useful import data for debugging
814 Severity : enhancement
816 Description: Re-disable certain /proc logging
817 Details : Enable and disable client's offset_stats, extents_stats and
818 extents_stats_per_process stats logging on the fly.
821 Frequency : Only on FC kernels 2.6.22+
823 Description: oops in statahead
824 Details : Do not drop reference count for the dentry from VFS when lookup,
825 VFS will do that by itself.
827 Severity : enhancement
829 Description: Generic /proc file permissions
830 Details : Set /Proc file permissions in a more generic way to enable non-
831 root users operate on some /proc files.
835 Description: Hitting mdc_commit_close() ASSERTION
836 Details : Properly handle request reference release in
837 ll_release_openhandle().
841 Description: quota recovery deadlock during mds failover
842 Details : This patch includes att18982, att18236, att18237 in bz14840.
844 1. fix osts hang when mds does failover with quotaon
845 2. prevent watchdog storm when osts threads wait for the
850 Frequency : only patchless client
851 Description: add workaround for race between add/remove dentry from hash
853 Severity : enhancement
855 Description: Allow OST glimpses to return PW locks
859 Description: LBUG when llog conf file is full
860 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
864 Description: Prevent import from entering FULL state when server in recovery
868 Description: service mount cannot take device name with ":"
869 Details : Only when device name contains ":/" will mount treat it as
875 Description: replace ptlrpcd with the statahead thread to interpret the async
876 statahead RPC callback
880 Frequency : on recovery
881 Description: I/O failures after umount during fail back
882 Details : if client reconnected to restarted server we need join to recovery
883 instead of find server handler is changed and process self
884 eviction with cancel all locks.
886 Severity : enhancement
888 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
890 Severity : enhancement
892 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
894 Severity : enhancement
896 Description: Upcall on Lustre log has been dumped
897 Details : Allow for a user mode script to be called once a Lustre log has
898 been dumped. It passes the filename of the dumped log to the
899 script, the location of the script can be specified via
900 /proc/sys/lnet/debug_log_upcall.
905 Description: avoid idr_remove called for id which is not allocated.
906 Details : Move assigment s_dev for clustered nfs to end of initialization,
907 to avoid problem with error handling.
912 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
913 Details : When connection is reused this not moved from CONN_UNUSED_HASH
914 into CONN_USED_HASH and this prodice warning when put connection
915 again in unused hash.
918 Severity : enhancement
920 Description: File striping can now be set to use an arbitrary pool of OSTs.
922 Severity : enhancement
924 Description: Export bytes_read/bytes_write count on OSC/OST.
928 Description: Early reply size mismatch, MGC loses connection
929 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
930 the connect flags are properly negotiated.
935 Description: quotas are not honored with O_DIRECT
936 Details : all writes with the flag O_DIRECT will use grants which leads to
937 this problem. Now using OBD_BRW_SYNC to guard this.
941 Description: add quota statistics
942 Details : 1. sort out quota proc entries and proc code.
943 2. add quota statistics
945 Severity : enhancement
947 Description: enable quota support for HEAD.
951 Description: Properly propagate oinfo flags from lov to osc for statfs
952 Details : restore missing copy oi_flags to lov requests.
954 Severity : enhancement
956 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
957 : l_getgroups(8), lst(8), routerstat(8)
958 Details : included man pages for llobdstat(8), llstat(8),
959 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
961 Severity : enhancement
963 Description: configure's --enable-quota should check the
964 : kernel .config for CONFIG_QUOTA
965 Details : configure is terminated if --enable-quota is passed but
966 : no quota support is in kernel
970 Description: Remove portals compatibility
971 Details : Remove portals compatibility, not interoperable with releases
976 Description: Resolve device initialization race
977 Details : Prevent proc handler from accessing devices added to the
978 obd_devs array but yet be intialized.
980 Severity : enhancement
982 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
984 Severity : enhancement
986 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
991 Description: Using wrong pointer in osc_brw_prep_request
992 Details : Access to array[-1] can produce panic if kernel compiled with
993 CONFIG_PAGE_ALLOC enabled
997 Description: Client runs out of low memory
998 Details : Consider only lowmem when counting initial number of llap pages
1002 Description: *optional* service tags registration
1003 Details : if the "service tags" package is installed on a Lustre node
1004 When the filesystem is mounted, a local-node service tag will
1005 be created. See http://inventory.sun.com/ for more information
1006 about the Service Tags asset management system.
1010 Description: Kernel BUG tries to release flock
1011 Details : Lustre does not destroy flock lock before last reference goes
1012 away. So always drop flock locks when client is evicted and
1013 perform unlock regardless of successfulness of speaking to MDS.
1017 Description: add refcount protection for osc callbacks, avoid panic on shutdown
1021 Description: sanity test 65a fails if stripecount of -1 is set
1022 Details : handle -1 striping on filesystem in ll_dirstripe_verify
1027 Description: ASSERTION(CheckWriteback(page,cmd)) failed
1028 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
1031 Severity : enhancement
1033 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
1037 Description: do not process already freed flock
1038 Details : flock can possibly be freed by another thread before it reaches
1039 to ldlm_flock_completion_ast.
1043 Description: LBUG during stress test
1044 Details : Need properly lock accesses the flock deadlock detection list.
1048 Description: oops in page fault handler
1049 Details : kernel page fault handler can return two special 'pages' in error
1050 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
1054 Description: timeout with invalidate import.
1055 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
1056 should be handled by ptlrpcd. This produce long age waiting and
1057 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
1059 Severity : enhancement
1061 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
1065 Description: enable MGS and MDT services start separately
1066 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
1067 a co-located MGS without starting the MGS, which is a complement
1068 to 'nosvc' mount option.
1072 Frequency : after recovery
1073 Description: precreate to many object's after del orphan.
1074 Details : del orphan st in oscc last_id == next_id and this triger growing
1075 count of precreated objects. Set flag LOW to skip increase count
1076 of precreated objects.
1080 Frequency : rare, on clear nid stats
1081 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
1082 Details : when clean nid stats sometimes try destroy live entry,
1083 and this produce panic in free.
1087 Description: Stack overflow during MDS log replay
1088 ease stack pressure by using a thread dealing llog_process.
1092 Description: wait until IO finished before start new when do lock cancel.
1093 Details : VM protocol want old IO finished before start new, in this case
1094 need wait until PG_writeback is cleared until check dirty flag and
1095 call writepages in lock cancel callback.
1097 Severity : enhancement
1099 Description: using special macro for print time and cleanup in includes.
1103 Description: mds_mfd_close() ASSERTION(rc == 0)
1104 Details : In mds_mfd_close(), we need protect inode's writecount change
1105 within its orphan write semaphore to prevent possible races.
1109 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
1113 Frequency : rare, on shutdown ost
1114 Description: don't hit live lock with umount ost.
1115 Details : shrink_dcache_parent can be in long loop with destroy dentries,
1116 use shrink_dcache_sb instead.
1120 Description: don't panic with use echo client
1121 Details : echo client pass NULL as client nid pointer and this produce null
1122 pointer dereference.
1126 Description: fix build on ppc32
1127 Details : compile code with -m64 flag produce wrong object file for ppc32.
1131 Description: add message levels for liblustreapi
1135 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
1136 Details : Move the 'good_osts' check before the 'total_bavail' check. This
1137 will result in an -EAGAIN and in the exit call path we call
1138 alloc_rr() which will with increasing aggressiveness attempt to
1139 aquire precreated objects on the minimum number of required OSCs.
1143 Description: Use old size assignment to avoid deadlock
1144 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
1145 the scheduling while holding a spinlock. We do not need locking
1146 for size in ll_update_inode() because size is only updated from
1147 the MDS for directories or files without objects, so there is no
1148 other place to do the update, and concurrent access to such inodes
1149 are protected by the inode lock.
1153 Description: resolve "_IOWR redefined" build error on SLES10
1157 Description: dump the memory debugging after all modules are unloaded to
1158 suppress false negative in conf_sanity test 39
1160 Severity : enhancement
1162 Description: build kernel-ib packages for OFED 1.3 in our release cycle
1167 Description: fix SLES kernel versioning
1168 Details : the kernel version for our SLES 10 kernel did not include a "-"
1169 before the "smp" at the end. while this was not a problem in
1170 general, it did mean that software trying to use the kernel
1171 version to try to detect a vendor specific kernel would fail.
1172 this was most evident by the OFED build scripts.
1176 Description: Don't update lov_desc members until making sure they are valid
1177 Details : When updating lov_desc members via proc fs, need fix their
1178 validities before doing the real update.
1182 Description: don't put request into delay list while invalidate in flight.
1183 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
1184 import in flight. this produce timeout for invalidate and sometimes
1185 can cause stale data.
1189 Frequency : on ppc only
1190 Description: not convert ost objects for directory because it's not exist.
1191 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
1195 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
1197 Severity : enhancement
1199 Description: Update to support 2.6.22.14 vanilla kernel.
1203 Frequency : rare, on recovery
1204 Description: read procfs can produce deadlock in some situation
1205 Details : Holding lprocfs lock which send rpc can produce block for destroy
1206 obd objects and this also block reconnect with -EALREADY. This
1207 isn't fix all lprocfs bugs - but make it rare.
1209 Severity : enhancement
1211 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
1214 Frequency : frequent on X2 node
1216 Description: mdc_set_open_replay_data LBUG
1217 Details : Set replay data for requests that are eligible for replay.
1221 Description: lustre_mgs: operation 101 on unconnected MGS
1222 Details : When MGC is disconnected from MGS long enough, MGS will evict the
1223 MGC, and late on MGC cannot successfully connect to MGS and a lot
1224 of the error messages complaining that MGS is not connected.
1228 Frequency : on network error
1229 Description: panic with double free request if network error
1230 Details : mdc_finish_enqueue is finish request if any network error ocuring,
1231 but it's true only for synchronus enqueue, for async enqueue
1232 (via ptlrpcd) this incorrect and ptlrpcd want finish request
1235 Severity : enhancement
1237 Description: client-side metadata stat-ahead during readdir(directory readahead)
1238 Details : perform client-side metadata stat-ahead when the client detects
1239 readdir and sequential stat of dir entries therein
1242 Frequency : on start mds
1244 Description: Implement get_info(last_id) in obdfilter.
1247 Frequency : occasional
1249 Description: Correctly check stale fid, not start epoch if ost not support SOM
1250 Details : open with flag O_CREATE need set old fid in op_fid3 because
1251 op_fid2 was overwritten with new generated fid, but mds can answer
1252 with one of these two fids and both is not stale. Setattr
1253 incorrectly started an epoch and assume will be called
1254 done_writing, but without SOM done_writing ever being called.
1257 Frequency : rare, depends on device drivers and load
1259 Description: MDS or OSS nodes crash due to stack overflow
1260 Details : Code changes in 1.8.0 increased the stack usage of some functions.
1261 In some cases, in conjunction with device drivers that use a lot
1262 of stack the MDS (or possibly OSS) service threads could overflow
1263 the stack. One change which was identified to consume additional
1264 stack has been reworked to avoid the extra stack usage.
1267 Frequency : occasional
1269 Description: Do not fail import if osc_interpret_create gets -EAGAIN
1270 Details : If osc_interpret_create got -EAGAIN it immediately exits and
1271 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
1272 oscc_has_objects and see OSC has no objests and call
1273 oscc_internal_create to resend create request.
1275 Severity : enhancement
1277 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
1279 Severity : enhancement
1281 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
1284 Frequency : very rare
1286 Description: Fix replay if there is an un-replied request and open
1287 Details : In some cases, older replay request will revert the
1288 mcd->mcd_last_xid on MDS which is used to record the client's
1289 latest sent request.
1291 Severity : enhancement
1293 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
1295 Severity : enhancement
1297 Description: Add rhel5 support to HEAD.
1299 Serverity : enhancement
1301 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
1306 Description: Don't allow skipping OSTs if index has been specified.
1307 Details : Don't allow skipping OSTs if index has been specified, make
1308 locking in internal create lots better.
1312 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
1313 Details : access to bitfield in structure is always rounded to long
1314 and this produce problem with not atomic change any bit.
1318 Description: Lustre make rpms failed.
1319 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
1320 builds Lustre rpms from tarball.
1323 Frequency : rare on shutdown ost
1325 Description: If llog cancel was not send before clean_exports phase, this can
1326 produce deadlock in llog code.
1327 Details : If llog thread has last reference to obd and call class_import_put
1328 this produce deadlock because llog_cleanup_commit_master wait when
1329 last llog_commit_thread exited, but this never success because was
1330 called from llog_commit_thread.
1334 Description: allow userland application know is lost one of stripes.
1335 Details : fill lvb_blocks with error code on ost and return it to
1336 application if error flag found.
1340 Description: NULL lov_tgts causing MDS oops
1341 Details : more safe checks for NULL lov_tgts for avoid oops.
1343 Severity : enhancement
1345 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
1349 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
1350 Details : If an OST is down the MDS will hang indefinitely in
1351 obd_statfs_rqset() waiting for the statfs data. While for
1352 MDS QOS usage of statfs, it should not stuck in waiting.
1354 Severity : enhancement
1356 Description: remote_acl support
1357 Details : Support ACL-based permission check for remote user.
1358 Support setfacl/getfacl for remote user with the utils
1359 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
1360 the system "{s,g}etfacl" utils.
1362 Severity : enhancement
1364 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
1366 Severity : enhancement
1368 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
1372 Description: make mgs_setparam() handle fsname containing dash
1373 Details : fsname containing a dash does not work with lctl conf_param
1375 Severity : enhancement
1377 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
1379 Severity : enhancement
1381 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
1385 Frequency : rare, at shutdown
1386 Description: access already free / zero obd_namespace.
1387 Details : if client_disconnect_export was called without force flag set,
1388 and exist connect request in flight, this can produce access to
1389 NULL pointer (or already free pointer) when connect_interpret
1390 store ocd flags in obd_namespace.
1394 Frequency : only at startup
1395 Description: not alloc memory with spinlock held.
1396 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
1397 if any spinlock held.
1399 Severity : enhancement
1401 Description: make lustre randomly fail allocating memory
1402 Details : Make lustre randomly failed allocating memory for testing purpose.
1404 Severity : enhancement
1406 Description: lost problems with lov objid file
1407 Details : Fixes some scability and access to not inited memory problems
1408 in work with lov objdid file.
1413 Description: lfs find does not continue on file error
1414 Details : Continue other files processing when a file/dir is absent.
1418 Description: Inconsistent usage of lustre_pack_reply()
1419 Details : Standardize the usage of lustre_pack_reply() such that it
1420 always generate a CERROR on failure.
1423 Frequency : occasional
1425 Description: MDS hang or stay in waiting lock
1426 Details : If client receive lock with CBPENDING flag ldlm need send lock
1427 cancel as separate rpc, to avoid situation when cancel request
1428 can't processed due all i/o threads stay in wait lock.
1432 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1436 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1440 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1444 Description: Update to RHEL4 latest kernel.
1446 Severity : enhancement
1448 Description: Build SLES10 patchless client fails
1449 Details : The configure was broken by run ./configure with
1450 --with-linux-obj=.... argument for patchless client. When the
1451 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1452 can't search header adequately. Use absolute path such as
1453 -I($LINUX)/include instead.
1457 Description: interrupt oig_wait produce painc on resend.
1458 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1459 private set, and this produce situation when rq_ptlrpcd_data not
1460 copyed to new allocated request and triggered LBUG on assert
1461 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1462 ptlrpcd set if request is changed and can be safety changed to use
1467 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1468 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1469 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1470 these macros and replace them with their difinitions instead.
1474 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1475 Details : Update which_patch & target file for SLES10 latest kernel.
1478 Description: Remove client patches from SLES 10 kernel.
1479 Details : This causes SLES 10 clients to behave as patchless clients
1480 even on a Lustre-patched (server) kernel.
1482 Severity : enhancement
1484 Description: self-adjustable client's lru lists
1485 Details : use adaptive algorithm for managing client cached locks lru
1486 lists according to current server load, other client's work
1487 pattern, memory activities, etc. Both, server and client
1488 side namespaces provide number of proc tunables for controlling
1491 Severity : enhancement
1493 Description: light-weight GSS support
1494 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
1495 and reduce performance overhead.
1497 Severity : enhancement
1499 Description: Linux keyring support
1500 Details : Support using service of Linux keyring for Lustre GSS internal
1501 context refresh/cache mechanism.
1505 Description: Fix errors in lfs documentation
1506 Details : Fixes man pages
1510 Description: don't use GFP_* in generic Lustre code.
1511 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
1515 Description: obdclass is limited by single OBD_ALLOC(idarray)
1516 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
1520 Description: warnings with build patchless client with vanila 2.6.19 and up
1521 Details : change old ctl_table style and replace ctl_table/ctl_table_header
1522 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
1526 Frequency : Only for SLES
1527 Description: sanity_quota fail test_1
1528 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
1529 file, which makes TSTID[2] inunique.
1533 Frequency : Only for SLES9
1534 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
1535 Details : Update target/ChangeLog/which_patch .
1539 Frequency : Only for test_10 in sanity-quota.sh
1540 Description: a bug in quota test code
1541 Details : checking a test flag in a wrong place
1545 Frequency : Only for test_13 in sanity-quota.sh
1546 Description: a bug in quota test script
1547 Details : shouldn't check overfull allocated quota
1551 Description: jbd statistics
1552 Details : Port older jbd statistics patch for sles10
1554 Severity : enhancement
1556 Description: Add printing inode info into message about error in writepage.
1561 Description: reply_lock_interpret crash due to race with it and lock cancel.
1562 Details : Do not replay locks that are being cancelled. Do not reference
1563 locks by their address during replay, just by their handle.
1567 Frequency : When flocks are used.
1568 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
1569 Details : Flock locks might destroy just granted lock if it could be merged
1570 with another existing flock, this is done in completion handler,
1571 so teach ldlm_cli_enquque_fini that this is a valid case for
1577 Description: Oops in read and write path when failing to allocate lock.
1578 Details : Check if lock allocation failed and return error back.
1582 Description: lstripe command fails for valid OST index
1583 Details : The stripe offset is compared to lov->desc.ld_tgt_count
1584 instead of lov->desc.ld_active_tgt_count.
1588 Description: sanity.sh failed test 103
1589 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
1594 Description: df doesn't work properly if diskfs blocksize != 4K
1595 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
1599 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
1600 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
1602 Severity : enhancement
1604 Description: Adaptive timeouts
1605 Details : RPC timeouts adapt to changing server load and network
1606 conditions to reduce resend attempts and improve recovery time.
1610 Description: llapi_file_create() does not allow some changes
1611 Details : add llapi_file_open() that allows specifying the mode and
1612 open flags, and also returns an open file handle.
1616 Description: merge and cleanup kernel patches.
1617 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
1621 Description: Add journal checksum support.(Kernel part)
1622 Details : The journal checksum feature adds two new flags i.e
1623 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
1624 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
1625 indicates that the commit block contains the checksum for
1626 the blocks described by the descriptor blocks. Now commit
1627 record can be sent to disk without waiting for descriptor
1628 blocks to be written to disk. This behavior is controlled
1629 using JBD2_FEATURE_ASYNC_COMMIT flag.
1633 Description: OSS needs mutliple precreate threads
1634 Details : Add ability to start more than one create thread per OSS.
1638 Description: Sanity.sh test_65k failed: setstripe should have succeeded
1639 Details : execute lfs setstripe on client
1643 Description: mds_obd_create error creating tmp object
1644 Details : When the user sets quota on root, llog will be affected and can't
1645 create files and write files.
1648 Frequency : Always on ia64 patchless client, and possibly others.
1650 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
1651 Details : This allows the patchless client to be loaded on architectures
1652 without this export.
1656 Description: RedHat Update kernel for RHEL5
1657 Details : Add the kernel config file for RHEL5.
1661 Description: RedHat Update kernel for RHEL5
1662 Details : Modify the kernel config file more closer RHEL5.
1666 Description: Build failure against Centos5 (RHEL5)
1667 Details : Define PAGE_SIZE when it isn't present.
1671 Description: test_7 on sanity-quota.sh can't be used on separate machines
1672 Details : add facet to handle it
1676 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
1677 Details : with NFS, the anon dentry's parent was set to itself in
1678 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
1679 corresponding dentry other than use rec->ur_name.
1681 Severity : enhancement
1683 Description: lfs setstripe enhancement
1684 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
1688 Description: enable data checksumming by default
1689 Details : enable checksum by default, allow --disable-checksum
1690 configure option and "-o nochecksum" mount option. Checksums
1691 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
1695 Description: lustre support for RHEL5
1696 Details : Add support for RHEL5.
1700 Description: Client eviction due to failover config
1701 Details : after a connection loss, the lustre client should attempt to
1702 reconnect to the last active server first before trying the
1703 other potential connections.
1707 Description: when mds and osts use different quota unit(32bit and 64bit),
1708 quota will be released repeatly.
1709 Details : void sending multiple quota reqs to mds, which will keep the
1710 status between the reqs.
1714 Description: osts not allocated evenly to files
1715 Details : change the condition to increase offset_idx
1718 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
1720 Description: Data corruption for OSTs that are formatted larger than 2TB
1722 Details : When generating the bio request for lustre file writes the
1723 sector number would overflow a temporary variable before being
1724 used for the IO. The data reads correctly from Lustre (which
1725 will overflow in a similar manner) but other file data or
1726 filesystem metadata may be corrupted in some cases.
1730 Description: Tune the kernel for good SCSI performance.
1731 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
1732 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
1737 Description: rewrite ext2-derived code in obdclass/uuid.c
1738 Details : rewrite inherited code (uuid parsing code from ext2 utils)
1739 from scratch preserving functionality.
1743 Description: "lfs find -obd UUID" prints directories
1744 Details : "lfs find -obd UUID" will return all directory names instead
1745 of just file names. It is incorrect because the directories
1746 do not reside on the OSTs.
1750 Description: use i_size_read and i_size_write in 2.6 port
1751 Details : replace inode->i_size access with i_size_read/write()
1754 Frequency : when removing large files
1756 Description: scheduling issue during removal of large Lustre files
1757 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
1758 It causes scheduling issues when removing large files (17TB in the
1762 Frequency : only with liblustre clients on XT3
1764 Description: evictions taking too long
1765 Details : allow llrd to evict clients directly on OSTs
1768 Frequency : only on ppc
1770 Description: /proc/fs/lustre/devices broken on ppc
1771 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
1772 We should make sure the type of 'index' is loff_t and then cast
1773 explicitly as needed below. Do not assign an explicitly cast
1777 Frequency : only for rhel5
1779 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
1780 Details : Modify the target file & which_kernel.
1782 Severity : enhancement
1784 Description: omit set fsid for export NFS
1785 Details : fix set/restore device id for avoid EMFILE error and mark lustre
1786 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
1790 Frequency : Always, for kernels after 2.6.16
1791 Description: Fix warning idr_remove called for id=.. which is not allocated.
1792 Details : Last kernels save old s_dev before kill super and not allow
1793 to restore from callback - restore it before call kill_anon_super.
1797 Description: buffer overruns could theoretically occur
1798 Details : llapi_semantic_traverse() modifies the "path" argument by
1799 appending values to the end of the origin string, and a
1800 overrun may occur. Adding buffer overrun check in liblustreapi.
1804 Description: Fix error on 'ls .' at the top of the Lustre mount.
1805 Details : Don't revalidate dentry if it is a root dentry.
1809 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
1810 Details : Modify vm-tunables-rhel4.patch.
1814 Description: Kernel config for 2.6.18-vanilla.
1815 Details : Modify targets/2.6-vanilla.target.in.
1816 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
1817 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
1818 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
1819 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
1823 Description: improve handling recoverable errors
1824 Details : if request processig with error which can be recoverable on server
1825 request should be resend, otherwise page released from cache and
1830 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
1831 Details : Modify target file & which_patch.
1832 A flaw was found in the IA32 system call emulation provided
1833 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
1834 value could be stored in the %RAX register, which could trigger an
1835 out-of-bounds system call table access. An untrusted local user
1836 could exploit this flaw to run code in the kernel
1837 (ie a root privilege escalation). (CVE-2007-4573).
1841 Description: O_DIRECT bypasses client statistics.
1842 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
1843 read_ahead_stats were updated. copied the stats section from
1844 osc_send_oap_rpc() into async_internal().
1848 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
1852 Description: change order of libsysio includes
1853 Details : '#include sysio.h' should always come before '#include xtio.h'
1857 Description: handle "serious error: objid * already exists" more gracefully
1858 Details : If LAST_ID value on disk is smaller than the objects existing in
1859 the O/0/d* directories, it indicates disk corruption and causes an
1860 LBUG(). If the object is 0-length, then we should use the existing
1861 object. This will help to avoid a full fsck in most cases.
1863 Severity : enhancement
1865 Description: adapt the lustre_config script to support the upgrade case
1866 Details : Add "-u" option for lustre_config script to support upgrading 1.4
1867 server targets to 1.6 in parallel.
1871 Description: To avoid grant space > avaible space when the disk is almost
1872 full. Without this patch you might see the error "grant XXXX >
1873 available" or some LBUG about grant, when the disk is almost
1875 Details : In filter_check_grant, for non_grant cache write, we should
1876 check the left space by if (*left > ungranted + bytes), instead
1877 of (*left > ungranted), because only we are sure the left space
1878 is enough for another "bytes", then the ungrant space should be
1879 increase. In client, we should update cl_avail_grant only there
1880 is OBD_MD_FLGRANT in the reply.
1884 Description: Update RHEL 4 kernel to fix local root privilege escalation.
1885 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
1886 described in CVE-2007-4573. This problem could allow untrusted
1887 local users to gain root access.
1890 Frequency : when using O_DIRECT and quotas
1892 Description: Incorrect file ownership on O_DIRECT output files
1893 Details : block usage reported by 'lfs quota' does not take into account
1894 files that have been written with O_DIRECT.
1899 Description: touch file failed when fs is not full
1900 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
1901 otherwise we can get ENOSP while fs is not full.
1905 Description: parallel lock callbacks
1906 Details : Instead of sending blocking and completion callbacks as separated
1907 requests, adding them to a set and sending in parallel.
1910 Frequency : only for Cray XT3
1911 Bugzilla : 12829/13455
1912 Description: Changing primary group doesn't change the group lustre assigns to
1914 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
1915 overridden with the primary group provided by the group upcall,
1916 whereas the supplied fsgid can be trusted if it is in the list of
1917 supplementary groups returned by the group upcall.
1919 Severity : enhancement
1921 Description: Allow masking D_WARNING, D_ERROR messages from console
1922 Details : Console messages can now be disabled via lnet.printk.
1926 Description: User code with malformed file open parameter crashes client node
1927 Details : Before packing join_file req, all the related reference should be
1928 checked carefully in case some malformed flags cause fake
1929 join_file req on client.
1933 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
1934 Details : ldlm_completion_ast() assumes that a lock is granted when the req
1935 mode is equal to the granted mode. However, it should also check
1936 that LDLM_FL_CP_REQD is not set.
1940 Description: Heavy nfs access might result in deadlocks
1941 Details : After ELC code landed, it is now improper to enqueue any mds
1942 locks under och_sem, because enqueue might want to decide to
1943 cancel open locks for same inode we are holding och_sem for.
1947 Description: Client eviction while running blogbench
1948 Details : A lot of unlink operations with concurrent I/O can lead to a
1949 deadlock causing evictions. To address the problem, the number of
1950 oustanding OST_DESTROY requests is now throttled to
1951 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
1952 callbacks are processed in priority.
1956 Description: enable ACLs on MDS by default
1957 Details : ACLs must be enabled on MDS by default.
1960 Frequency : PPC/PPC64 only
1962 Description: conflicts between asm-ppc64/types.h and lustre_types.h
1963 Details : fix duplicated definitions between asm-ppc64/types.h and
1964 lustre_types.h on PPC.
1967 Frequency : PPC/PPC64 only
1969 Description: asm-ppc/segment.h does not exist
1970 Details : fix compile issue on PPC.
1974 Description: better handle error messages in extents code
1977 Frequency : RHEL4 only
1979 Description: mkfs is very slow on IA64/RHEL4
1980 Details : A performance regression has been discovered in the MPT Fusion
1981 driver between versions 3.02.73rh and 3.02.99.00rh. As a
1982 consequence, we have downgraded the MPT Fusion driver in the RHEL4
1983 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
1985 Severity : enhancement
1987 Description: SNMP support enhancement
1988 Details : Adding total number of sampled request for an MDS node in snmp
1991 Severity : enhancement
1993 Description: Optimize ldlm waiting list processing for PR extent locks
1994 Details : When processing waiting list for read extent lock and meeting read
1995 lock that is same or wider to it that is not contended, skip
1996 processing rest of the list and immediatelly return current
1997 status of conflictness, since we are guaranteed there are no
1998 conflicting locks in the rest of the list.
2002 Description: Time out and refuse to reconnect
2003 Details : When the failover node is the primary node, it is possible
2004 to have two identical connections in imp_conn_list. We must
2005 compare not conn's pointers but NIDs, otherwise we can defeat
2006 connection throttling.
2010 Description: port llog fixes from b1_6 into HEAD
2011 Details : Port llog reference couting and some llog cleanups from b1_6
2012 (bug 10800) into HEAD, for protect from panic and access to already
2013 free llog structures.
2017 Description: Detect stride IO mode in read-ahead
2018 Details : When a client does stride read, read-ahead should detect that and
2019 read-ahead pages according to the detected stride pattern.
2023 Description: data checksumming impacts single node performance
2024 Details : add support for several checksum algorithm. Currently, only CRC32
2025 and Adler-32 are supported. The checksum type can be changed on
2026 the fly via /proc/fs/lustre/osc/*/checksum_type.
2030 Description: use adler32 for page checksums
2031 Details : when available, use the Adler-32 algorithm instead of CRC32 for
2036 Description: build for x2 fails
2037 Details : fix compile issue on Cray systems.
2041 Description: Properly match for duplicate locks
2042 Details : Due to different lock order from skiplists code, we need to
2043 traverse entire list for now
2046 Frequency : only on PPC/SLES10
2048 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
2049 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
2050 be defined. Add a hack in mkfs_lustre.c to work around this compile
2055 Description: LASSERT on MDS when client holding flock lock dies
2056 Details : ldlm pool logic depends on number of granted locks equal to
2057 number of released locks which is not true for flock locks, so
2058 just exclude such locks from consideration.
2062 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
2063 Details : Use fsfilt_sync() for both the whole filesystem sync and
2064 individual file sync to eliminate dangerous inode locking
2065 with I_LOCK that can lead to a deadlock.
2069 Description: performance in 1.6.3
2070 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
2071 sure that 1MB requests are not fragmented by the block layer.
2073 Severity : enhancement
2075 Description: organize the server-side client stats on per-nid basis
2076 Details : Change the structure of stats under obdfilter and mds to
2084 The "uuid"s file would list the uuids of _active_ exports.
2085 And the clear entry is to clear all stats and stale nids.
2087 Severity : enhancement
2089 Description: eliminate client locks in face of contention
2090 Details : file contention detection and lockless i/o implementation
2091 for contended files.
2095 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
2098 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
2100 Description: MDS may lose file striping (and hence file data) in some cases
2101 Details : If there are additional extended attributes stored on the MDS,
2102 in particular ACLs, SELinux, or user attributes (if user_xattr
2103 is specified for the client mount options) then there is a risk
2104 of attribute loss. Additionally, the Lustre file striping
2105 needs to be larger than default (e.g. striped over all OSTs),
2106 and an additional attribute must be stored initially in the
2107 inode and then increase in size enough to be moved to the
2108 external attribute block (e.g. ACL growing in size) for file
2113 Description: skiplist implementation simplification
2114 Details : skiplists are used to group compatible locks on granted list
2115 that was implemented as tracking first and last lock of each
2116 lock group the patch changes that to using doubly linked lists
2120 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
2121 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
2122 because LASSERT on both the data supplied by a client, and the
2123 data on disk is dangerous and incorrect.
2125 Severity : enhancement
2127 Description: Slow truncate/writes to huge files at high offsets.
2128 Details : Directly associate cached pages to lock that protect those pages,
2129 this allows us to quickly find what pages to write and remove
2130 once lock callback is received.
2134 Description: more ldlm soft lockups
2135 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
2136 starve other threads from the resource lock for a long time in
2137 case of long waiting queue, so change the debug level from
2138 D_OTHER to the less frequently used D_INFO.
2140 Severity : enhancement
2142 Description: add -gid, -group, -uid, -user options to lfs find
2146 Description: Hung threads in invalidate_inode_pages2_range
2147 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
2148 once one is completed. As a result, some RPCs are stuck in the
2149 queue and are never sent.
2153 Description: filter threads hungs on waiting journal commit
2154 Details : Cleanup filter group llog code, then only filter group llog will
2155 be only created in the MDS/OST syncing process.
2159 Description: Procfs and llog threads access destoryed import sometimes.
2160 Details : Sync the import destoryed process with procfs and llog threads by
2161 the import refcount and semaphore.
2163 Severity : enhancement
2165 Description: openlock cache of b1_6 port to HEAD
2170 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
2171 Details : If insertion of an extent fails, then discard the inode
2172 preallocation and free data blocks else it can lead to duplicate
2177 Description: don't always update ctime in ext3_xattr_set_handle()
2178 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
2179 In some cases the ctime should not be updated, for example for
2180 2.0->1.8 compatibility it is necessary to delete an xattr and it
2181 should not update the ctime.
2185 Bugzilla : 15713/16362
2186 Description: Assertion in iopen_connect_dentry in 1.6.3
2187 Details : looking up an inode via iopen with the wrong generation number can
2188 populate the dcache with a disconneced dentry while the inode
2189 number is in the process of being reallocated. This causes an
2190 assertion failure in iopen since the inode's dentry list contains
2191 both a connected and disconnected dentry.
2195 Description: assertion failure in ldlm_handle2lock()
2196 Details : fix a race between class_handle_unhash() and class_handle2object()
2197 introduced in lustre 1.6.5 by bug 13622.
2202 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
2203 Details : remove the limit of 256 scsi disks in the sd_iostat patch
2208 Description: oops in sd_iostats_seq_show()
2209 Details : unloading/reloading the scsi low level driver triggers a kernel
2210 bug when trying to access the sd iostat file.
2215 Description: Kernel panics during QLogic driver reload
2216 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
2217 patch, causing memory corruption.
2222 Description: journal_dev option does not work in b1_6
2223 Details : pass mount option during pre-mount.
2225 Severity : enhancement
2227 Description: Add a FIEMAP(FIle Extent MAP) ioctl
2228 Details : FIEMAP ioctl will allow an application to efficiently fetch the
2229 extent information of a file. It can be used to map logical blocks
2230 in a file to physical blocks in the block device.
2234 Description: LDLM soft lockups - improvement
2235 Details : It is be possible to send the lock handle along with each read
2236 or write request because the client is already doing a lock match
2237 itself so there isn't any reason the OST should have to re-do that
2241 Frequency : only on Cray X2
2243 Description: X2 build failures
2244 Details : fix build failures on Cray X2.
2248 Description: xid & resent requests
2249 Details : Initialize RPC XID from clock at startup (randomly if clock is
2252 Severity : enhancement
2254 Description: Add lustre_start utility to start or stop multiple Lustre servers
2259 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
2260 Details : In case of memory pressure, list_del() can be called twice on
2261 req->rq_history_list, causing a kernel oops.
2265 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
2266 Details : fix stack overflow in the distributed lock manager by defering
2267 export eviction after a failed AST to the elt thread instead of
2268 handling it in the dlm interpret routine.
2272 Description: Convert some comments to new format.
2273 Details : Update documenting comments to match doxygen conventions.
2277 Description: Grammar fixes.
2278 Details : A couple of trivial spelling fixes.
2282 Description: OSD_COUNTERS-mandatory
2283 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
2284 to simplify the code.
2288 Description: simplify cmm_device freeing logic.
2289 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
2293 Description: Add lockdep support to dt_object_operations locking interface.
2294 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
2295 parameter indicating lock ordering. Update mdd code to use new
2300 Description: Introduce failloc constants for lockless IO tests.
2301 Details : Add two new failloc constants to test lockless IO. Only one of
2302 them in implemented---another is checked in yet to be landed
2307 Description: Add lockdep support for inode mutex.
2308 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
2309 in the situations where Lustre has to lock more than one inode
2314 Description: Add optional invariants checking support.
2315 Details : Add new LINVRNT() macro, optional on new --enable-invariants
2316 configure switch. This macro is to be used for consistency and
2317 sanity checks that are too expensive to be left in `production'
2322 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
2323 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
2328 Description: Add lockdep support for ldlm_lock and ldlm_resource.
2329 Details : Use spin_lock_nested() in (the only) situation where more than
2330 one ldlm_lock is locked simultaneously. Also, fix possible
2331 dead-lock in ldlm_lock_change_resource() by enforcing particular
2336 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
2337 Details : Instead of specifying each ldlm_lock call-back through separate
2338 parameter, wrap them into struct ldlm_callback_suite.
2342 Description: Kill join_lru obd method and its callers.
2343 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
2344 in memory---a requirement implemented through ->o_join_lru() obd
2345 method in HEAD. Remove this method and its users.
2349 Description: Add asynchronous ldlm ENQUEUE completion handler.
2350 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
2351 case that stock ldlm_completion_ast() cannot handle as it waits
2352 until lock is granted. Introduce new ldlm_completion_ast_async()
2353 for this. Also comment ldlm_completion_ast().
2357 Description: ldlm_error <-> errno conversion.
2358 Details : Add functions to map (rather arbitrary) between LDLM error codes
2359 and standard errno values. CLIO needs this to prevent LDLM specific
2360 constants from escaping ldlm and osc.
2364 Description: Kill unused ldlm_handle2lock_ns() function.
2365 Details : Kill unused ldlm_handle2lock_ns() function.
2369 Description: Add lu_ref support to ldlm_lock
2370 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
2371 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
2372 and ldlm_resource. LDLM interface has to be changed, because code that
2373 releases a reference on a lock, has to "know" what reference this is.
2374 In the most frequent case
2376 lock = ldlm_handle2lock(handle);
2378 LDLM_LOCK_PUT(lock);
2380 no changes are required. When any other reference (received _not_ from
2381 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
2382 instead of LDLM_LOCK_PUT().
2384 Arguably, changes are pervasive, and interface requires some discipline
2385 for proper use. On the other hand, it was very instrumental in finding
2386 a few leaked lock references.
2390 Description: Add ldlm_lock_addref_try().
2391 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
2392 attempts to addref a lock that might be being canceled
2397 Description: Add ldlm_weigh_callback().
2398 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
2399 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
2400 hard-coded `number of pages' logic.
2404 Description: Add lockdep annotations to llog code.
2405 Details : Use appropriately tagged _nested() locking calls in the places
2406 where llog takes more than one ->lgh_lock lock.
2410 Description: Add loi_kms_set().
2411 Details : Wrap kms updates into a helper function.
2415 Description: Constify instances of struct lsm_operations.
2416 Details : Constify instances of struct lsm_operations.
2420 Description: lu_conf support.
2421 Details : On a server, a file system object is uniquely identified
2422 by a fid, which is sufficient to locate and load all object
2423 state (inode). On a client, on the other hand, more data are
2424 necessary instantiate an object. Change lu_object_find() and
2425 friends to take additional `lu_conf' argument describing object.
2426 Typically this includes layout information.
2430 Description: lu_context fixes.
2431 Details : Introduce new lu_context functions that are needed on the client
2432 side, where some system threads (ptlrpcd) are shared by multiple
2433 modules, and so cannot be stopped during module shutdown.
2437 Description: Add start and stop methods to lu_device_type_operations.
2438 Details : Introduce two new methods in lu_device_type_operations, that are
2439 invoked when first instance of a given type is created and last one
2440 is destroyed respectively. This is need by CLIO.
2444 Description: Add lu_ref support to struct lu_device.
2445 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
2446 track leaked references.
2450 Description: Introduce lu_kmem_descr.
2451 Details : lu_kmem_descr and its companion interface allow to create
2452 and destroy a number of kmem caches at once.
2456 Description: Fix lu_object finalization race.
2457 Details : Fix a race between lu_object_find() finding an object and its
2458 concurrent finalization. This race is (most likely) not possible
2459 on the server, but might happen on the client.
2463 Description: Introduce lu_ref interface.
2464 Details : lu_ref is a debugging module allowing to track references to
2465 a given object. It is quite cpu expensive, and has to be
2466 explicitly enabled with --enable-lu_ref. See usage description
2471 Description: Factor lu_site procfs stats into a separate function.
2472 Details : Separate lu_site stats printing code into a separate function
2473 that can be reused on a client.
2477 Description: Constify instances of struct {lu,dt,md}_device_operations.
2478 Details : Constify instances of struct {lu,dt,md}_device_operations.
2482 Description: Introduce struct md_site and move meta-data specific parts of
2483 struct lu_site here.
2484 Details : Move md-specific fields out of struct lu_site into special struct
2485 md_site, so that lu_site can be used on a client.
2489 Description: Kill mdd_lov_destroy().
2490 Details : Remove unused mdd code.
2494 Description: Add st_block checking to multistat.c.
2495 Details : Add st_block checking to multistat.c.
2499 Description: Add lu_ref support to struct obd_device.
2500 Details : Add lu_ref tracking to obd_device.
2504 Description: Kill obd_set_fail_loc().
2505 Details : Remove unused code.
2509 Description: Add special type for ptlrpc_request interpret functions.
2510 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
2511 there. Actual usage will be in CLIO.
2515 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
2516 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
2517 doesn't work with lockdep.
2521 Description: Add rwv.c test program.
2522 Details : New testing program exercising readv(2) and writev(2) (Qian).
2526 Description: Add sendfile.c test program.
2527 Details : New testing program exercising sendfile(2) (Jay).
2531 Description: Ratelimit a message that can be very frequent.
2532 Details : Ratelimit a memory allocation failure message that can
2537 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2538 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2542 Description: Liblustre build fixes.
2543 Details : Liblustre build fixes.
2547 Description: libcfs: add cfs_{need,cond}_resched() interface.
2548 Details : libcfs: add cfs_{need,cond}_resched() definition and
2549 implementations for Linux, NT, and liblustre.
2551 Severity : enhancement
2553 Description: More exported tunables for mballoc
2554 Details : Add support for tunable preallocation window and new tunables for
2555 large/small requests
2559 Description: Detect corruption of block bitmap and checking for preallocations
2560 Details : Checks validity of on-disk block bitmap. Also it does better
2561 checking of number of applied preallocations. When corruption is
2562 found, it turns filesystem readonly to prevent further corruptions.
2566 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
2567 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
2568 window to 0. If it does hit cache miss, and read-pattern is still
2569 stride-io mode, does not reset the stride window, but also does not
2570 increase the stride window length in this case.
2574 Frequency : only for big-endian servers
2575 Description: Check if system is big-endian while mounting fs with extents feature
2576 Details : Mounting a filesystem with extents feature will fail on big-endian
2577 systems since ext3-based ldiskfs is not supported on big-endian
2578 systems. This can be over-riden with "bigendian_extents" mount option.
2580 Severity : enhancement
2582 Description: The root squash functionality
2583 Details : A security feature, which is to prevent users from being able
2584 to mount lustre on their desktop, run as root, and delete
2585 all of the files in the filesystem. The goal is accomplished by
2586 remapping user id (UID) and group id (GID) of the root user to
2587 a UID and GID specified by the system administartor via Lustre
2588 configuration management server (MGS). The functionality also
2589 allows to specify sets of clients for which the remapping does
2594 Description: Excessive recovery window
2595 Details : With AT enabled, the recovery window can be excessively long (6000+
2596 seconds). To address this problem, we no longer use
2597 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
2598 timeout no longer depends on the service time, it is set to
2599 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
2600 time via pb_service_time.
2604 Description: Watchdog triggered on MDS failover
2605 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
2606 the OSTs know that the MDS "UUID" can be reused for the same export
2607 from a different NID, so we do not need to wait for the export to be
2611 Frequency : rare, only if using MMP with Linux RAID
2613 Description: MMP doesn't work with Linux RAID
2614 Details : While using HA for Lustre servers with Linux RAID, it is possible
2615 that MMP will not detect multiple mounts. To make this work we
2616 need to unplug the device queue in RAID when the MMP block is being
2617 written. Also while reading the MMP block, we should read it from
2618 disk and not the cached one.
2620 Severity : enhancement
2622 Description: open file using fid
2623 Details : A file can be opened using just its fid, like
2624 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
2627 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
2630 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
2631 Details : add proper locking for creation of jbd_4k slab cache
2635 Description: MMP check in ext3_remount() fails without displaying any error
2636 Details : When multiple mount protection fails during remount, proper error
2639 Severity : enhancement
2641 Description: Allow stripe size to be up to 4G-64k
2642 Details : Fix math logic to allow large stripe sizes.
2646 Description: add check for >8TB ldiskfs filesystems
2647 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
2648 Don't allow >8TB ldiskfs filesystems to be mounted without
2649 force_over_8tb mount option
2654 Descriptoin: OSS complains "too many blocks" and crashes
2655 Details : In filter_commitrw_write():
2656 1. lquota_chkquota();
2657 2. obdo_from_inode();
2658 3. lquota_pending_commit();
2659 "2." has changed uid and gid in this case, then caused
2660 lquota_chkquota() and lquota_pending_commit() to use
2661 different uid and gid. That is the root of the bug.
2665 Descriptoin: getstripe doesn't report stripe count = -1 sanely
2666 Details : Detect and print out the 'stripe count = -1' case properly.
2670 Description: Out or order replies might be lost on replay
2671 Details : In ptlrpc_retain_replayable_request if we cannot find retained
2672 request with tid smaller then one currently being added, add it
2673 to the start, not end of the list.
2675 --------------------------------------------------------------------------------
2677 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
2679 * Support for kernels:
2680 2.6.5-7.283 (SLES 9),
2681 2.6.9-55.EL (RHEL 4),
2682 2.6.16.46-0.14 (SLES 10),
2683 2.6.18.8 vanilla (kernel.org)
2684 * Client support for unpatched kernels:
2685 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2686 2.6.16 - 2.6.22 vanilla (kernel.org)
2687 * Due to recently discovered recovery problems, we do not recommend
2688 using patchless RHEL 4 clients with this or any earlier release.
2689 * Recommended e2fsprogs version: 1.39.cfs8
2690 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2691 * Starting with this release, the ldiskfs backing filesystem required
2692 by Lustre is now in its own package, lustre-ldiskfs. This package
2693 should be installed. It is versioned separately from Lustre and
2694 may be released separately in future.
2698 Description: block reactivating mgc import until all deactivates complete
2699 Details : Fix race when failing back MDT/MGS to itself (testing)
2702 Frequency : at statup only
2704 Description: mds_lov_synchronize race leads to various problems
2705 Details : simultaneous MDT->OST connections at startup can cause the
2706 sync to abort, leaving the OSC in a bad state.
2708 Severity : enhancement
2710 Description: add optional extra BUILD_VERSION info
2711 Details : add a new environment variable (namely LUSTRE_VERS) which allows
2712 to override the lustre version.
2715 Frequency : 2.6.18 servers only
2717 Description: ll_kern_mount() doesn't release the module reference
2718 Details : The ldiskfs module reference count never drops down to 0
2719 because ll_kern_mount() doesn't release the module reference.
2724 Description: server LBUG when using old ost_num_threads parameter
2725 Details : Accept the old ost_num_threads parameter but warn that it
2726 is deprecated, and fix an off-by-one error that caused an LBUG.
2731 Description: Transient SCSI error results in persistent IO issue
2732 Details : iobuf->dr_error is not reinitialized to 0 between two
2736 Frequency : sometimes when underlying device returns I/O errors
2738 Description: OSTs not going read-only during write failures
2739 Details : OSTs are not remounted read-only when the journal commit threads
2740 get I/O errors because fsfilt_ext3 calls journal_start/stop()
2741 instead of the ext3 wrappers.
2745 Description: poor connect scaling with increasing client count
2746 Details : Don't run filter_grant_sanity_check for more than 100 exports
2747 to improve scaling for large numbers of clients.
2750 Frequency : SLES10 only
2752 Description: sanity-quota.sh quotacheck failed: rc = -22
2753 Details : Quotas cannot be enabled on SLES10.
2756 Frequency : liblustre clients only
2758 Description: getdirentries does not give error when run on compute nodes
2759 Details : getdirentries does not fail when the size specified as an argument
2760 is too small to contain at least one entry
2762 Severity : enhancement
2764 Description: Add LNET router traceability for debug purposes
2765 Details : If a checksum failure occurs with a router as part of the
2766 IO path, the NID of the last router that forwarded the bulk data
2767 is printed so it can be identified.
2772 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
2773 Details : Due to a race condition, liblustre clients were occasionally
2774 evicted incorrectly.
2776 Severity : enhancement
2778 Description: lfs setstripe use optional parameters instead of postional
2781 Severity : enhancement
2783 Description: Nanosecond timestamp support for ldiskfs
2784 Details : The on-disk ldiskfs filesystem has added support for nanosecond
2785 resolution timestamps. There is not yet support for this at
2786 the Lustre filesystem level.
2789 Frequency : during server recovery
2791 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
2792 Details : request with rq_no_resend flag not awake l_wait_event if they get
2796 Frequency : nfs export on patchless client
2798 Description: connectathon hang when test nfs export over patchless client
2799 Details : Disconnected dentry cannot be found with lookup, so we do not need
2800 to unhash it or make it invalid
2803 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
2804 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
2805 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
2810 Description: open req refcounting wrong on reconnect
2811 Details : If reconnect happened between getting open reply from server and
2812 call to mdc_set_replay_data in ll_file_open, we will schedule
2813 replay for unreferenced request that we are about to free.
2814 Subsequent close will crash in variety of ways.
2815 Check that request is still eligible for replay in
2816 mdc_set_replay_data().
2821 Description: disable writes to filesystem when reading health_check file
2822 Details : the default for reading the health_check proc file has changed
2823 to NOT do a journal transaction and write to disk, because this
2824 can cause reads of the /proc file to hang and block HA state
2825 checking on a healthy but otherwise heavily loaded system. It
2826 is possible to return to the previous behaviour during configure
2827 with --enable-health-write.
2829 Severity : enhancement
2831 Description: 64-bit inode version
2832 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
2833 to the inode. This will be required for version-based recovery.
2838 Description: MDS fails to start if a duplicate client export is detected
2839 Details : in some rare cases it was possible for a client to connect to
2840 an MDS multiple times. Upon recovery the MDS would detect this
2841 and fail during startup. Handle this more gracefully.
2843 Severity : enhancement
2845 Description: Add -o localflock option to simulate old noflock
2847 Details : This will achieve local-only flock/fcntl locks
2853 Description: log_commit_thread vs filter_destroy race leads to crash
2854 Details : Take import reference before releasing llog record semaphore
2859 Description: Wrong request locking in request set processing
2860 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
2861 imp_delayed_list, in this place should be used imp_lock.
2864 Frequency : when reconnection
2866 Description: Grant Leak when osc reconnect to OST
2867 Details : When osc reconnect ost, OST(filter) should check whether it
2868 should grant more space to client by comparing fed_grant and
2869 cl_avail_grant, and return the granted space to client instead
2870 of "new granted" space, because client will call osc_init_grant
2871 to update the client grant space info.
2874 Frequency : when client reconnect to OST
2876 Description: Grant Leak when osc do resend and replay bulk write
2877 Details : When osc reconnect to OST, OST(filter)should clear grant info of
2878 bulk write request, because the grant info while be sync between
2879 OSC and OST when reconnect, and we should ignore the grant info
2880 these of resend/replay write req.
2885 Description: Grant space more than avaiable left space sometimes.
2886 Details : When then OST is about to be full, if two bulk writing from
2887 different clients came to OST. Accord the avaliable space of the
2888 OST, the first req should be permitted, and the second one
2889 should be denied by ENOSPC. But if the seconde arrived before
2890 the first one is commited. The OST might wrongly permit second
2891 writing, which will cause grant space > avaiable space.
2894 Frequency : when client is evicted
2896 Description: Grant might be wrongly erased when osc is evicted by OST
2897 Details : when the import is evicted by server, it will fork another
2898 thread ptlrpc_invalidate_import_thread to invalidate the
2899 import, where the grant will be set to 0. While the original
2900 thread will update the grant it got when connecting. So if
2901 the former happened latter, the grant will be wrongly errased
2902 because of this race.
2907 Description: Checking Stale with correct fid
2908 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
2909 to check whether it is stale, because sometimes, we want the
2910 enqueue happened anyway, and op_data.fid2 will not be initialized.
2912 Severity : enhancement
2914 Description: update patchless client
2915 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
2918 Frequency : only with 2.4 kernel
2920 Description: random memory corruption
2921 Details : size of struct ll_inode_info is to big for union inode.u and this
2922 can be cause of random memory corruption.
2927 Description: Memory leak in recovery
2928 Details : Lov_mds_md was not free in an error handler in mds_create_object.
2929 It should also check obd_fail before fsfilt_start, otherwise if
2930 fsfilt_start return -EROFS,(failover mds during mds recovery).
2931 then the req will return with repmsg->transno = 0 and rc = EROFS.
2932 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
2933 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
2934 be freed no matter whether fsfilt_commit success or not.
2937 Frequency : only with huge count clients
2939 Description: Prevents from taking the superblock lock in llap_from_page for
2941 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
2942 for llap_from_page call in ll_removepage prevents from taking the
2943 superblock lock for a soon died page.
2948 Description: Not check open intent error before release open handle
2949 Details : in some rare cases, the open intent error is not checked before
2950 release open handle, which may cause
2951 ASSERTION(open_req->rq_transno != 0), because it tries to release
2952 the failed open handle.
2957 Description: Set cat log bitmap only after create log success.
2958 Details : in some rare cases, the cat log bitmap is set too early. and it
2959 should be set only after create log success.
2963 Description: Accessing a block bevice can re-enable I/O when Lustre is
2964 tearing down a device.
2965 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
2970 Description: service threads may hog cpus when there are a lot of requests
2971 Details : Insert cond_resched to give other threads a chance to use some CPU
2976 Description: the cat log was not initialized in recovery
2977 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
2978 unlink log on mds will not be initialized until mds post
2979 recovery. And also in mds post recovery, the unlink log will
2980 initialization will be done asynchronausly, so there will be race
2981 between add unlink log and unlink log initialization.
2985 Description: brw_stats were being printed incorrectly
2986 Details : brw_stats were being printed as log2 but all of them were not
2987 recorded as log2. Also remove some code duplication arising from
2988 filter_tally_{read,write}.
2992 Frequency : rare, only in recovery.
2993 Description: ASSERTION(req->rq_type != LI_POISON) failed
2994 Details : imp_lock should be held while iterating over imp_sending_list for
2995 prevent destroy request after get timeout in ptlrpc_queue_wait.
2999 Description: replay-single.sh test 52 fails
3000 Details : A lock's skiplist need to be cleanup when it being unlinked
3001 from its resource list.
3005 Description: Short directio read returns full requested size rather than
3007 Details : Direct I/O operations should return actual amount of bytes
3008 transferred rather than requested size.
3010 Severity : enhancement
3012 Description: metadata RPC reduction (e.g. for rm performance)
3013 Details : decrease the amount of synchronous RPC between clients and servers
3014 by canceling conflicing lock before the operation on the client
3015 and packing thier handles into the main operation RPC to server.
3017 Severity : enhancement
3019 Description: Async OSC create to avoid the blocking unnecessarily.
3020 Details : If a OST has no remain object, system will block on the creating
3021 when need to create a new object on this OST. Now, ways use
3022 pre-created objects when available, instead of blocking on an
3023 empty osc while others are not empty. If we must block, we block
3024 for the shortest possible period of time.
3028 Description: improve handling recoverable errors
3029 Details : if request processig with error which can be recoverable on server
3030 request should be resend, otherwise page released from cache and
3033 Severity : enhancement
3035 Description: refine locking for avoid write wrong info into lov_objid file
3036 Details : fix possible races with add new target and write/update data in
3039 --------------------------------------------------------------------------------
3041 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
3046 Frequency : on some architectures
3048 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
3049 Details : Uninitialized flags sometimes cause configuration commands to
3053 Frequency : patchless clients only
3055 Description: missing __iget() symbol export
3056 Details : The __iget() symbol export is missing. To avoid the need for
3057 this on patchless clients the deathrow inode reaper is turned
3058 off, and we depend on the VM to clean up old inodes. This
3059 dependency was during via the fix for bug 12181.
3061 --------------------------------------------------------------------------------
3063 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
3065 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
3066 INTEROPERATE with older versions automatically. In many cases a
3067 special upgrade step is needed. Please read the
3068 user documentation before upgrading any part of a live system.
3069 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
3070 version will not interoperate with 1.6 betas before beta5 (1.5.95).
3071 * WARNING: Lustre configuration and startup changes are required with
3072 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
3076 Severity : enhancement
3078 Description: MountConf
3079 Details : Lustre configuration is now managed via mkfs and mount
3080 commands instead of lmc and lconf. New obd types (MGS, MGC)
3081 are added for dynamic configuration management. See
3082 https://mail.clusterfs.com/wikis/lustre/MountConf for
3085 Severity : enhancement
3087 Description: dynamic OST addition
3088 Details : OSTs can now be added to a live filesystem
3090 Severity : enhancement
3092 Description: startup order invariance
3093 Details : MDTs and OSTs can be started in any order. Clients only
3094 require the MDT to complete startup.
3096 Severity : enhancement
3098 Description: parallel, asynchronous orphan cleanup
3099 Details : orphan cleanup is now performed in separate threads for each
3100 OST, allowing parallel non-blocking operation.
3102 Severity : enhancement
3104 Description: optimized stripe assignment
3105 Details : stripe assignments are now made based on ost space available,
3106 ost previous usage, and OSS previous usage, in order to try
3107 to optimize storage space and networking resources.
3109 Severity : enhancement
3111 Description: Permanently set tunables
3112 Details : All writable /proc/fs/lustre tunables can now be permanently
3113 set on a per-server basis, at mkfs time or on a live system.
3115 Severity : enhancement
3117 Description: Lustre message v2
3118 Details : Add lustre message format v2.
3120 Severity : enhancement
3122 Description: client OST exclusion list
3123 Details : Clients can be started with a list of OSTs that should be
3124 declared "inactive" for known non-responsive OSTs.
3128 Description: SPEC SFS validation failure on NFS v2 over lustre.
3129 Details : Changes the blocksize for regular files to be 2x RPC size,
3130 and not depend on stripe size.
3132 Severity : enhancement
3134 Description: Multiple MD RPCs in flight.
3135 Details : Further unserialise some read-only MDS RPCs - learn about intents.
3136 To avoid overly-overloading MDS, introduce a limit on number of
3137 MDS RPCs in flight for a single client and add /proc controls
3138 to adjust this limit.
3140 Severity : enhancement
3142 Description: client read/write statistics
3143 Details : Add client read/write call usage stats for performance
3144 analysis of user processes.
3145 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
3146 file access. extents_stats shows chunk size distribution.
3147 extents_stats_per_process show chunk size distribution per
3150 Severity : enhancement
3152 Description: mds statistics
3153 Details : Add detailed mds operations statistics in
3154 /proc/fs/lustre/mds/*/stats.
3158 Description: Failure of copying files with lustre special EAs.
3159 Details : Client side always return success for setxattr call for lustre
3160 special xattr (currently only "trusted.lov").
3165 Description: Refcount LNET uuids
3166 Details : The global LNET uuid list grew linearly with every startup;
3167 refcount repeated list entries instead of always adding to
3170 Severity : enhancement
3172 Description: Dynamic service threads
3173 Details : Within a small range, start extra service threads
3174 automatically when the request queue builds up.
3177 Frequency : mixed-endian client/server environments
3179 Description: mixed-endian crashes
3180 Details : The new msg_v2 system had some failures in mixed-endian
3183 Severity : enhancement
3185 Description: Easy OST removal
3186 Details : OSTs can be permanently deactivated with e.g. 'lctl
3187 conf_param lustre-OST0001.osc.active=0'
3189 Severity : enhancement
3191 Description: MGS proc entries
3192 Details : Added basic proc entries for the MGS showing what filesystems
3195 Severity : enhancement
3197 Description: provide MGS failover
3198 Details : Added config lock reacquisition after MGS server failover.
3200 Severity : enhancement
3202 Description: add Linux 2.4 support
3203 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
3207 Description: a large application tries to do I/O to the same resource and dies
3208 in the middle of it.
3209 Details : Check the req->rq_arrival time after the call to
3210 ost_brw_lock_get(), but before we do anything about
3211 processing it & sending the BULK transfer request. This
3212 should help move old stale pending locks off the queue as
3213 quickly as obd_timeout.
3216 Frequency : when an incorrect nid is specified during startup
3218 Description: ptlrpc connect to non-existant node causes kernel crash
3219 Details : LNET can't be re-entered from an event callback, which
3220 happened when we expire a message after the export has been
3221 cleaned up. Instead, hand the zombie cleanup off to another
3224 Severity : enhancement
3226 Description: plain/inodebits lock performance improvement
3227 Details : Grouping plain/inodebits in granted list by their request modes
3228 and bits policy, thus improving the performance of search through
3232 Frequency : only if OST filesystem is corrupted
3234 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3235 Details : for a short time RPCs with bulk IO are in the replay list,
3236 but replay of bulk IOs is unimplemented. If the OST filesystem
3237 is corrupted due to disk cache incoherency and then replay is
3238 started it is possible to trip an assertion. Avoid putting
3239 committed RPCs into the replay list at all to avoid this issue.
3242 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3244 Description: System hang on startup
3245 Details : This bug allowed the liblustre (e.g. catamount) client to
3246 return to the app before handling all startup RPCs. This
3247 could leave the node unresponsive to lustre network traffic
3248 and manifested as a server ptllnd timeout.
3250 Severity : enhancement
3252 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3253 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
3254 boolean to control whether to print peer debug info when a
3255 client's RPC times out.
3258 Frequency : only for kernels with patches from Lustre below 1.4.3
3260 Description: Remove old rdonly API
3261 Details : Remove old rdonly API which unused from at least lustre 1.4.3
3264 Frequency : only for devices with external journals
3266 Description: Set external device read-only also
3267 Details : During a commanded failover stop, we set the disk device
3268 read-only while the server shuts down. We now also set any
3269 external journal device read-only at the same time.
3272 Frequency : when upgrading from 1.4 while trying to change parameters
3274 Description: The wrong (new) MDC name was used when setting parameters for
3275 upgraded MDT's. Also allows changing of OSC (and MDC)
3276 parameters if --writeconf is specified at tunefs upgrade time.
3279 Frequency : when setting specific ost indicies
3281 Description: QOS code breaks on skipped indicies
3282 Details : Add checks for missing OST indicies in the QOS code, so OSTs
3283 created with --index need not be sequential.
3285 Severity : enhancement
3287 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3288 Details : The uninit_groups feature works in conjunction with the kernel
3289 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3290 up the pass1 processing of e2fsck. This is a read-only feature
3291 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3292 mount filesystems that have had this feature enabled.
3294 Severity : enhancement
3296 Description: Improve multi-block allocation algorithm to avoid fragmentation
3297 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3298 allocation locality and avoid filesystem fragmentation.
3300 ------------------------------------------------------------------------------
3302 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
3304 * Support for kernels:
3305 2.4.21-47.0.1.EL (RHEL 3)
3306 2.6.5-7.283 (SLES 9)
3307 2.6.9-42.0.10.EL (RHEL 4)
3308 2.6.12.6 vanilla (kernel.org)
3309 2.6.16.27-0.9 (SLES 10)
3310 * Recommended e2fsprogs version: 1.39.cfs5
3312 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3316 Frequency : occasional, depends on client load and configuration
3317 Bugzilla : 12181, 12203
3318 Description: data loss for recently-modified files
3320 Details : In some cases it is possible that recently written or created
3321 files may not be written to disk in a timely manner (this should
3322 normally be within 30s unless client IO load is very high).
3323 The problem appears as zero-length files or files that are a
3324 multiple of 1MB in size after a client crash or client eviction
3325 that are missing data at the end of the file.
3327 This problem is more likely to be hit on clients where files are
3328 repeatedly created and unlinked in the same directory, clients
3329 have a large amount of RAM, have many CPUs, the filesystem has
3330 many OSTs, the clients are rebooted frequently, and/or the files
3331 are not accessed by other nodes after being written.
3333 The presence of the problem can be detected by looking at
3334 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
3335 smaller than the second (nr_unused) then dirty files will not
3336 be flushed automatically to disk. "sync; sleep 10" should be
3337 run several times on the node before unmounting it to update
3338 Lustre (this is also safe to run on nodes without this problem).
3340 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
3341 that can cause this same problem, so customers using that kernel
3342 also need to update the kernel in addition to Lustre. In order
3343 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
3345 It is normal that files written just before a client crash (less
3346 than 30s) may not yet have been flushed to disk, even for local
3350 Frequency : frequent on thin XT3 nodes
3352 Description: UUID collision on thin XT3 Linux nodes
3353 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
3354 randomly, since we relied on an insufficiently-seeded PRNG.
3359 Description: OSS hangs after "All ost request buffers busy"
3360 Details : A deadlock between quota and journal operations caused OSS
3361 hangs after printing "All ost request buffers busy."
3364 Frequency : always on liblustre builds
3366 Description: Cleanup compiler warnings on liblustre
3369 Frequency : always on liblustre builds on XT3
3371 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
3375 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
3376 > 32000 subdirectories
3377 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
3378 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
3379 subdirectory count crosses 32000. This will aid e2fsck to
3380 correctly handle more than 32000 subdirectories.
3383 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3385 Description: System hang on startup
3386 Details : This bug allowed the liblustre (e.g. catamount) client to
3387 return to the app before handling all startup RPCs. This
3388 could leave the node unresponsive to lustre network traffic
3389 and manifested as a server ptllnd timeout.
3391 Severity : enhancement
3393 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3394 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
3395 boolean to control whether to print peer debug info when a
3396 client's RPC times out.
3401 Description: make O_SYNC working on 2.6 kernels
3402 Details : 2.6 kernels use different method for mark pages for write,
3403 so need add a code to lustre for O_SYNC work.
3408 Description: Failure to close file and release space on NFS
3409 Details : Put inode details into lock acquired in ll_intent_file_open.
3410 Use mdc_intent_lock in ll_intent_open to properly
3411 detect all kind of errors unhandled by mdc_enqueue.
3416 Description: proc file read during shutdown sometimes raced obd removal,
3418 Details : Add lock to prevent obd access after proc file removal.
3421 Frequency : Only for files larger than 4GB on 32-bit clients.
3423 Description: improperly doing page alignment of locks
3424 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
3425 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3430 Description: RPCs being resent when they shouldn't be
3431 Details : Some RPCs that should not be resent are being resent. This
3432 can cause inconsistencies in the RPC state machine. Do not
3433 resend such requests.
3436 Frequency : rare, only with NFS export
3438 Description: Crash on NFS re-export node
3439 Details : under very unusual load conditions an assertion is hit in
3440 ll_intent_file_open()
3443 Frequency : only if OST filesystem is corrupted
3445 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3446 Details : for a short time RPCs with bulk IO are in the replay list,
3447 but replay of bulk IOs is unimplemented. If the OST filesystem
3448 is corrupted due to disk cache incoherency and then replay is
3449 started it is possible to trip an assertion. Avoid putting
3450 committed RPCs into the replay list at all to avoid this issue.
3455 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3456 Details : Large single O_DIRECT read and write calls can fail to allocate
3457 a sufficiently large buffer to process the request. In case of
3458 allocation failure the allocation is retried with a smaller
3459 buffer and broken into smaller requests.
3461 Severity : enhancement
3463 Description: Add -o localflock option to simulate old noflock behaviour.
3464 Details : This will achieve local-only flock/fcntl locks coherentness.
3469 Description: versioning check is incomplete
3470 Details : Checking the version difference of client vs. server, report
3471 error if the gap is too big.
3476 Description: add support PG_writeback bit
3477 Details : add support for PG_writeback bit for Lustre, for more carefull
3478 work with page cache in 2.6 kernel. This also fix some deadlocks
3479 and remove hack for work O_SYNC with 2.6 kernel.
3481 Severity : enhancement
3483 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3484 Details : The uninit_groups feature works in conjunction with the kernel
3485 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3486 up the pass1 processing of e2fsck. This is a read-only feature
3487 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3488 mount filesystems that have had this feature enabled.
3490 Severity : enhancement
3492 Description: Improve multi-block allocation algorithm to avoid fragmentation
3493 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3494 allocation locality and avoid filesystem fragmentation.
3496 ------------------------------------------------------------------------------
3498 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3500 * Support for kernels:
3501 2.6.9-42.0.3.EL (RHEL 4)
3502 2.6.5-7.276 (SLES 9)
3503 2.4.21-47.0.1.EL (RHEL 3)
3504 2.6.12.6 vanilla (kernel.org)
3505 2.6.16.21-0.8 (SLES10)
3506 * Recommended e2fsprogs version: 1.39.cfs2-0
3508 * The backwards-compatible /proc/sys/portals symlink has been removed
3509 in this release. Before upgrading, please ensure that you change
3510 any configuration scripts or /etc/sysctl.conf files that access
3511 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3512 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3513 in advance of the upgrade on any system running Lustre 1.4.6 or
3514 newer, since /proc/sys/lnet was added in that version.
3515 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3519 Frequency : only when quota is used
3521 Description: avoid scanning export list for quota master
3522 Details : Change the algorithms to avoid scanning export list in order
3523 to improve the efficiency.
3526 Frequency : MDS failover only, very rarely
3528 Description: "went back in time" messages on mds failover
3529 Details : The greatest transno may be lost when the current operation
3530 finishes with an error (transno==0) and the client's last_rcvd
3531 record is over-written. Save the greatest transno in the
3532 mds_last_transno for this case.
3535 Frequency : always for specific kernels and striping counts
3537 Description: client may get "Matching packet too big" without ACL support
3538 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
3539 when trying to access files in certain configurations. The
3540 clients should in fact be denied when mounting because they do
3541 not understand ACLs.
3544 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
3546 Description: many clients connecting with IO in progress causes connect timeouts
3547 Details : Avoid synchronous journal commits to avoid delays caused by many
3548 clients connecting/disconnecting when bulk IO is in progress.
3549 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
3550 OST_IO_PORTAL to avoid delays behind potentially many pending
3554 Frequency : occasionally with multiple writers to a single file
3556 Description: shared writes to file may result in wrong size reported by stat()
3557 Details : Allow growing of kms when extent lock is cancelled
3560 Frequency : always with random mmap IO to multi-striped file
3562 Description: mmap write might be lost if we are writing to a 'hole' in stripe
3563 Details : Only if the hole is at the end of OST object so that kms is too
3564 small. Fix is to increase kms accordingly in ll_nopage.
3567 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
3569 Description: writes to a missing object would leak memory on the OST
3570 Details : If there is an inconsistency between the MDS and OST filesystems,
3571 such that the MDS references an object that doesn't exist, writes
3572 to that object will leak memory due to incorrect cleanup in the
3573 error handling path, eventually running out of memory on the OST.
3578 Description: Creating too long symlink causes lustre errors
3579 Details : Check symlink and name lengths before sending requests to MDS.
3582 Frequency : only if flock is enabled (not on by default)
3584 Description: posix locks not released on fd closure on 2.6.9+
3585 Details : We failed to add posix locks to list of inode locks on 2.6.9+
3586 kernels, this caused such locks not to be released on fd close and
3587 then assertions on fs unmount about still used locks.
3590 Frequency : MDS failover only, very rarely
3592 Description: clients may get ASSERTION(granted_lock != NULL)
3593 Details : When request was taking a long time, and a client was resending
3594 a getattr by name lock request. The were multiple lock requests
3595 with the same client lock handle and
3596 mds_getattr_name->fixup_handle_for_resent_request found one of the
3597 lock handles but later failed with ASSERTION(granted_lock != NULL).
3602 Description: handle->h_buffer_credits > 0, assertion failure
3603 Details : h_buffer_credits is zero after truncate, causing assertion
3604 failure. This patch extends the transaction or creates a new
3608 Frequency : NFS re-export or patchless client
3609 Bugzilla : 11179, 10796
3610 Description: Crash on NFS re-export node (__d_move)
3611 Details : We do not want to hash the dentry if we don't have a lock.
3612 But if this dentry is later used in d_move, we'd hit uninitialised
3613 list head d_hash, so we just do this to init d_hash field but
3614 leave dentry unhashed.
3617 Frequency : NFS re-export or patchless client
3619 Description: NFS exports has problem with symbolic link
3620 Details : lustre client didn't properly install dentry when re-exported
3621 to NFS or running patchless client.
3624 Frequency : NFS re-export or patchless client
3626 Description: Various nfs/patchless fixes.
3627 Details : fixes reuse disconected alias for lookup process - this fixes
3628 warning "find_exported_dentry: npd != pd",
3629 fix permission error with open files at nfs.
3630 fix apply umask when do revalidate.
3633 Frequency : occasional
3635 Description: Crash on NFS re-export node
3636 Details : calling clear_page() on the wrong pointer triggered oops in
3637 generic_mapping_read().
3640 Frequency : rarely, using O_DIRECT IO
3642 Description: unaligned directio crashes client with LASSERT
3643 Details : check for unaligned buffers before trying any requests.
3646 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
3648 Description: stale data returned from RAID cache
3649 Details : If only a small amount of IO is done to the RAID device before
3650 reading it again it is possible to get stale data from the RAID
3651 cache instead of reading it from disk.
3654 Frequency : always for sles10 kernel
3656 Description: sles10 support
3657 Details : ll_follow_link: compile fixes and using of nd_set_link
3658 under newer kernels.
3661 Frequency : depends on arch, kernel and compiler version, always on sles10
3664 Description: recursive or deep enough symlinks cause stack overflow
3665 Details : getting rid of large stack-allocated variable in
3669 Frequency : depends on hardware
3671 Description: lustre write performance loss in the SLES10 kernel
3672 Details : the performance loss is caused by using of write barriers in the
3673 ext3 code. The SLES10 kernel turns barrier support on by
3674 default. The fix is to undo that change for ldiskfs.
3676 ------------------------------------------------------------------------------
3678 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
3680 * Support for kernels:
3681 2.6.9-42.0.3EL (RHEL 4)
3682 2.6.5-7.276 (SLES 9)
3683 2.4.21-47.0.1.EL (RHEL 3)
3684 2.6.12.6 vanilla (kernel.org)
3688 Frequency : quota enabled and large files being deleted
3690 Description: releasing more than 4GB of quota at once hangs OST
3691 Details : If a user deletes more than 4GB of files on a single OST it
3692 will cause the OST to spin in an infinite loop. Release
3693 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3698 Description: statfs data retrieved from /proc may be stale or zero
3699 Details : When reading per-device statfs data from /proc, in the
3700 {kbytes,files}_{total,free,avail} files, it may appear
3701 as zero or be out of date.
3704 Frequency : systems with MD RAID1 external journal devices
3706 Description: lconf's call to blkid is confused by RAID1 journal devices
3707 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3708 of returning all block devices that match the journal UUID.
3711 Frequency : always, for aggregate stripe size over 4GB
3713 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
3714 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
3715 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
3718 Frequency : always if "lfs find" used on a local file/directory
3720 Description: "lfs find" segfaults if used on a local file/directory
3721 Details : The case where a directory component was not specified wasn't
3722 handled correctly. Handle this properly.
3725 Frequency : always on ppc64
3727 Description: the write to an ext3 filesystem mounted with mballoc got stuck
3728 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3729 perform endianness conversion.
3732 Frequency : rarely (truncate to non-zero file size after write under load)
3733 Bugzilla : 10730, 10687
3734 Description: Files padded with zeros to next 4K multiple
3735 Details : With filesystems mounted using the "extents" option (2.6 kernels)
3736 it is possible that files that are truncated to a non-zero size
3737 immediately after being written are filled with zero bytes beyond
3738 the truncated size. No file data is lost.
3740 Severity : enhancement
3742 Description: Allow recovery/failover for liblustre clients.
3743 Details : liblustre clients were unaware of failover configurations until
3746 Severity : enhancement
3748 Description: user file locks should fail when not mounting with flock option
3749 Details : Set up an error-returning stub in ll_file_operations.lock field
3750 to prevent incorrect behaviour when client is mounted without
3751 flock option. Also, set up properly f_op->flock field for
3755 Frequency : always on ia64
3757 Description: "lfs df" loops on printing out MDS statfs information
3758 Details : The obd_ioctl_data was not initialized and in some systems
3759 this caused a failure during the ioctl that did not return
3760 an error. Initialize the struct and return an error on failure.
3763 Frequency : SLES 9 only
3765 Description: Error of copying files with lustre special EAs as root
3766 Details : Client side always return success for setxattr call for lustre
3767 special xattr (currently only "trusted.lov").
3770 Frequency : rarely on clusters with both ia64+i386 clients
3772 Description: ia64+i686 clients doing shared IO on the same file may LBUG
3773 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
3774 clients are doing concurrent writes to the same file it is
3775 possible that the ia64 clients may LASSERT because the OST
3776 extent locks are not PAGE_SIZE aligned. Ensure that grown
3777 locks are always aligned on the request boundary.
3780 Frequency : specific use, occasional
3782 Description: Overwriting in use executable truncates on-disk binary image
3783 Details : If one node attempts to overwrite an executable in use by
3784 another node, we now correctly return ETXTBSY instead of
3785 truncating the file.
3787 Severity : enhancement
3789 Description: Async OSC create to avoid the blocking unnecessarily.
3790 Details : If a OST has no remain object, system will block on the creating
3791 when need to create a new object on this OST. Now, ways use
3792 pre-created objects when available, instead of blocking on an
3793 empty osc while others are not empty. If we must block, we block
3794 for the shortest possible period of time.
3799 Description: chmod on Lustre root is propagated to other clients
3800 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
3801 invalid by the follow_mount time.
3804 Frequency : liblustre clients only
3806 Description: Race in 'instant cancel' lock handling could lead to such locks
3807 never to be granted in case of SMP MDS
3808 Details : Do not destroy not yet granted but cbpending locks in
3812 Frequency : replay/resend of open
3814 Description: non null lock assetion failure in mds_intent_policy
3815 Details : Trying to replay/resend lockless open requests resulted in
3816 mds_open() returning 0 with no lock. Now it sets a flag if
3817 a lock is going to be returned.
3819 Severity : enhancement
3821 Description: Checksum enhancements
3822 Details : New checksum enhancements allow for resending RPCs that failed
3825 Severity : enhancement
3827 Description: Tunables on number of dirty pages in cacche
3828 Details : Allow to set limit on number of dirty pages cached.
3833 Description: client crash on unmount - lock still has references
3834 Details : In some error handling cases it was possible to leak a lock
3835 reference on a client while accessing a file. This was not
3836 harmful to the client during operation, but would cause the
3837 client to crash when the filesystem is unmounted.
3840 Frequency : specific case, rare
3842 Description: ETXTBSY on mds though file not in use by client
3843 Details : ETXTBSY is no longer incorrectly returned when attempting to
3844 chmod or chown a directory that the user previously tried to
3845 execute or a currently-executing binary.
3848 Frequency : extremely rare except on liblustre-based clients
3850 Description: Lustre space not freed when files are deleted
3851 Details : Clean up open-unlinked files after client eviction. Previously
3852 the unlink was skipped and the files remained as orphans.
3857 Description: OST failure "would be an LBUG" in waiting_locks_callback()
3858 Details : In some cases it was possible to send a blocking callback to
3859 a client doing a glimpse, even though that client didn't get
3860 a lock granted. When the glimpse lock is cancelled on the OST
3861 the freed lock is left on the waiting list and corrupted the list.
3864 Frequency : all core dumps
3866 Description: Broke core dumps to lustre
3867 Details : Negative dentry may be unhashed if parent does not have UPDATE
3868 lock, but some callers, e.g. do_coredump, expect dentry to be
3869 hashed after successful create, hash it in ll_create_it.
3871 ------------------------------------------------------------------------------
3873 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
3875 * Support for kernels:
3876 2.6.9-42.0.2.EL (RHEL 4)
3877 2.6.5-7.276 (SLES 9)
3878 2.4.21-40.EL (RHEL 3)
3879 2.6.12.6 vanilla (kernel.org)
3883 Frequency : always on RHEL 3
3885 Description: Number of open files grows over time
3886 Details : The number of open files grows over time, whether or not
3887 Lustre is started. This was due to a filp leak introduced
3888 by one of our kernel patches.
3890 ------------------------------------------------------------------------------
3892 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
3894 * Support for kernels:
3895 2.6.9-42.EL (RHEL 4)
3896 2.6.5-7.276 (SLES 9)
3897 2.4.21-40.EL (RHEL 3)
3898 2.6.12.6 vanilla (kernel.org)
3903 Bugzilla : 5719, 9635, 9792, 9684
3904 Description: OST (or MDS) trips assertions in (re)connection under heavy load
3905 Details : If a server is under heavy load and cannot reply to new
3906 connection requests before the client resends the (re)connect,
3907 the connection handling code can behave badly if two service
3908 threads are concurrently handing separate (re)connections from
3909 the same client. Add better locking to the connection handling
3910 code, and ensure that only a single connection will be processed
3911 for a given client UUID, even if the lock is dropped.
3913 Severity : enhancement
3915 Description: add TCP zero-copy support to kernel
3916 Details : Add support to the kernel TCP stack to allow zero-copy bulk
3917 sends if the hardware supports scatter-gather and checksumming.
3918 This allows socklnd to do client-write and server-read more
3919 efficiently and reduce CPU utilization from skbuf copying.
3922 Frequency : only if NFS exporting from client
3924 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
3925 Details : If setting attributes on a file created under NFS that had
3926 never been opened it would be possible to oops the client
3927 if the file had no objects.
3930 Frequency : always for liblustre
3932 Description: liblustre client does MDS+OSTs setattr RPC for each write
3933 Details : When doing a write from a liblustre client, the client
3934 incorrectly issued an RPC to the MDS and each OST the file was
3935 striped over in order to update the timestamps. When writing
3936 with small chunks and many clients this could overwhelm the MDS
3937 with RPCs. In all cases it would slow down the write because
3938 these RPCs are unnecessary.
3940 Severity : enhancement
3942 Description: allow number of MDS service threads to be changed at module load
3943 Details : It is now possible to change the number of MDS service threads
3944 running. Adding "options mds mds_num_threads={N}" to the MDS's
3945 /etc/modprobe.conf will set the number of threads for the next
3946 time Lustre is restarted (assuming the "mds" module is also
3947 reloaded at that time). The default number of threads will
3948 stay the same, 32 for most systems.
3953 Description: OST crash if filesystem is unformatted or corrupt
3954 Details : If an OST is started on a device that has never been formatted
3955 or if the filesystem is corrupt and cannot even mount then the
3956 error handling cleanup routines would dereference a NULL pointer.
3961 Description: NULL pointer deref in llap_from_page.
3962 Details : get_cache_page_nowait can return a page with NULL (or otherwise
3963 incorrect) mapping if the page was truncated/reclaimed while it was
3964 searched for. Check for this condition and skip such pages when
3965 doing readahead. Introduce extra check to llap_from_page() to
3966 verify page->mapping->host is non-NULL (so page is not anonymous).
3969 Frequency : Sometimes when using sys_sendfile
3971 Description: "page not covered by a lock" warnings from ll_readpage
3972 Details : sendfile called ll_readpage without right page locks present.
3973 Now we introduced ll_file_sendfile that does necessary locking
3974 around call to generic_file_sendfile() much like we do in
3978 Frequency : with certain MDS communication failures at client mount time
3980 Description: NULL pointer deref after failed client mount
3981 Details : a client connection request may delayed by the network layer
3982 and not be sent until after the PTLRPC layer has timed out the
3983 request. If the client fails the mount immediately it will try
3984 to clean up before the network times out the request. Add a
3985 reference from the request import to the obd device and delay
3986 the cleanup until the network drops the request.
3989 Frequency : occasionally during client (re)connect
3991 Description: assertion failure during client (re)connect
3992 Details : processing a client connection request may be delayed by the
3993 client or server longer than the client connect timeout. This
3994 causes the client to resend the connection request. If the
3995 original connection request is replied in this interval, the
3996 client may trip an assertion failure in ptlrpc_connect_interpret()
3997 which thought it would be the only running connect process.
4000 Frequency : only with obd_echo servers and clients that are rebooted
4002 Description: kernel BUG accessing uninitialized data structure
4003 Details : When running an obd_echo server it did not start the ping_evictor
4004 thread, and when a client was evicted an uninitialized data
4005 structure was accessed. Start the ping_evictor in the RPC
4006 service startup instead of the OBD startup.
4008 Severity : enhancement
4009 Bugzilla : 10193 (patchless)
4010 Description: Remove dependency on various unexported kernel interfaces.
4011 Details : No longer need reparent_to_init, exit_mm, exit_files,
4012 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
4015 Frequency : rare (only users of deprecated and unsupported LDAP config)
4017 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
4018 Details : LDAP apparently contains 'lustreName' attributes instead of
4019 'name'. A simple remapping of the name is sufficient.
4022 Frequency : rare (only with non-default dump_on_timeout debug enabled)
4024 Description: waiting_locks_callback trips kernel BUG if client is evicted
4025 Details : Running with the dump_on_timeout debug flag turned on makes
4026 it possible that the waiting_locks_callback() can try to dump
4027 the Lustre kernel debug logs from an interrupt handler. Defer
4028 this log dumping to the expired_lock_main() thread.
4030 Severity : enhancement
4032 Description: Support NFS exporting on 2.6 kernels.
4033 Details : Implement non-rawops metadata methods for NFS server to use without
4034 changing NFS server code.
4037 Frequency : very rare (synthetic metadata workload only)
4039 Description: two racing renames might cause an MDS thread to deadlock
4040 Details : Running the "racer" program may cause one MDS thread to rename
4041 a file from being the source of a rename to being the target of
4042 a rename at exactly the same time that another thread is doing
4043 so, and the second thread has already enqueued these locks after
4044 doing a lookup of the target and is trying to relock them in
4045 order. Ensure that we don't try to re-lock the same resource.
4048 Frequency : only very large systems with liblustre clients
4050 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
4051 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
4052 all OSTs in parallel. This allows the eviction of stale liblustre
4053 clients to proceed much faster than if they were done in series,
4054 and also offers similar improvements for other set_info RPCs.
4059 Description: excessive CPU usage during initial read phase on client
4060 Details : During the initial read phase on a client, it would agressively
4061 retry readahead on the file, consuming too much CPU and impacting
4062 performance (since 1.4.5.8). Improve the readahead algorithm
4063 to avoid this, and also improve some other common cases (read
4064 of small files in particular, where "small" is files smaller than
4065 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
4070 Description: MDS crash when receiving packet with unknown intent.
4071 Details : Do not LBUG in unknown intent case, just return -EFAULT
4073 Severity : enhancement
4074 Bugzilla : 9293, 9385
4075 Description: MDS RPCs are serialised on client. This is unnecessary for some.
4076 Details : Do not serialize getattr (non-intent version) and statfs.
4079 Frequency : occasional, when OST network is overloaded/intermittent
4081 Description: client evicted by OST after bulk IO timeout
4082 Details : If a client sends a bulk IO request (read or write) the OST
4083 may evict the client if it is unresposive to its data GET/PUT
4084 request. This is incorrect if the network is overloaded (takes
4085 too long to transfer the RPC data) or dropped the OST GET/PUT
4086 request. There is no need to evict the client at all, since
4087 the pinger and/or lock callbacks will handle this, and the
4088 client can restart the bulk request.
4091 Frequency : Always when mmapping file with no objects
4093 Description: client crashes when mmapping file with no objects
4094 Details : Check that we actually have objects in a file before doing any
4095 operations on objects in ll_vm_open, ll_vm_close and
4101 Description: Request leak when working with deleted CWD
4102 Details : Introduce advanced request refcount tracking for requests
4103 referenced from lustre intent.
4105 Severity : Enhancement
4107 Description: Cache open file handles on client.
4108 Details : MDS now will return special lock along with openhandle, if
4109 requested and client is allowed to hold openhandle, even if unused,
4110 until such a lock is revoked. Helps NFS a lot, since NFS is opening
4111 closing files for every read/write openration.
4113 Severity : Enhancement
4115 Description: Cache open negative dentries on client when possible.
4116 Details : Guard negative dentries with UPDATE lock on parent dir, drop
4117 negative dentries on lock revocation.
4122 Description: Remounting a client read-only wasn't possible with a zconf mount
4123 Details : It wasn't possible to remount a client read-only with llmount.
4125 Severity : enhancement
4126 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
4127 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
4128 MPICH is included to add Lustre-specific ADIO interfaces.
4129 This is based closely on the UFS ADIO layer and only differs
4130 in file creation, in order to allow the OST striping to be set.
4131 This is user-contributed code and not supported by CFS.
4136 Description: extended inode attributes (immutable, append-only) work improperly
4137 when 2.4 and 2.6 kernels are used on client/server or vice versa
4138 Details : Introduce kernel-independent values for these flags.
4140 Severity : enhancement
4143 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
4144 Details : Many of the /proc/ tunables can only be tuned at a megabyte
4145 granularity. Now, Fractional MB granularity is be supported,
4146 this is very useful for low memory system.
4148 Severity : enhancement
4150 Description: Getattr by fid
4151 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
4152 locks, avoids extra getattr rpc requests to MDS, allows '/' to
4153 have locks and avoids getattr rpc requests for it on every stat.
4156 Frequency : Always, for filesystems larger than 2TB
4158 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4159 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
4160 filesystems larger than 2TB to be mounted with mballoc enabled.
4163 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
4165 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4166 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
4167 filesystems larger than 2TB is mounted then the kernel will
4168 silently corrupt the start of the filesystem. CONFIG_LBD is
4169 enabled for all CFS-supported kernels, but the possibility of
4170 this happening with a modified kernel config exists.
4172 Severity : enhancement
4174 Description: add client O_DIRECT support for 2.6 kernels
4175 Details : It is now possible to do O_DIRECT reads and writes to files
4176 in the Lustre client mountpoint on 2.6 kernel clients.
4178 Severity : enhancement
4180 Description: parallel glimpse, setattr, statfs, punch, destroy requests
4181 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
4182 parallel, not waiting for response from every OST before sending
4183 a rpc to the next OST.
4188 Description: setattr vs write race when updating file timestamps
4189 Details : Client processes that update a file timestamp into the past
4190 right after writing to the file (e.g. tar) it is possible that
4191 the updated file modification time can be reset to the current
4192 time due to a race between processing the setattr and write RPC.
4194 Severity : enhancement
4196 Description: Bring 'lfs find' closer in line with regular Linux find.
4197 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
4198 -print0 options and obtains all the needed info through the lustre
4201 Severity : enhancement
4203 Description: support up to 1024 configured devices on one node
4204 Details : change obd_dev array from statically allocated to dynamically
4205 allocated structs as they are first used to reduce memory usage
4210 Description: Flush dirty partially truncated pages during truncate
4211 Details : Immediatelly flush partially truncated pages in filter_setattr,
4212 this way we completely avoid having any pages in page cache on OST
4213 and can retire ugly workarounds during writes to flush such pages.
4218 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
4219 Details : protect inode from truncation within vfs_unlink() context
4220 just take a reference before calling vfs_unlink() and release it
4221 when parent's i_sem is free.
4226 Description: last_id value checked outside lock on OST caused LASSERT failure
4227 Details : If there were multiple MDS->OST object precreate requests in
4228 flight, it was possible that the OST's last object id was checked
4229 outside a lock and incorrectly tripped an assertion. Move checks
4230 inside locks, and discard old precreate requests.
4233 Frequency : always, if extents are used on OSTs
4235 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
4236 Details : OSTs using the extents format would not zero the high 16 bits of
4237 the index physical block number. This is not a problem for any
4238 OST filesystems smaller than 16TB, and no kernels support ext3
4239 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
4240 new/modified files) and can be fixed for existing filesystems
4241 with e2fsprogs-1.39-cfs1.
4246 Description: import connection selection may be incorrect if timer wraps
4247 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
4248 import connections to be ignored if the 32-bit jiffies counter
4249 wraps. Use a 64-bit jiffies counter.
4252 Frequency : very large clusters immediately after boot
4254 Description: LNET request buffers exhausted under heavy short-term load
4255 Details : If a large number of client requests are generated on a service
4256 that has previously never seen so many requests it is possible
4257 that the request buffer growth cannot keep up with the spike in
4258 demand. Instead of dropping incoming requests, they are held in
4259 the LND until the RPC service can accept more requests.
4262 Frequency : Sometimes during replay
4264 Description: Assertion failure in ll_local_open after replay.
4265 Details : If replay happened on an open request reply before we were able
4266 to set replay handler, reply will become not swabbed tripping the
4267 assertion in ll_local_open. Now we set the handler right after
4268 recognising of open request
4271 Frequency : very rare
4273 Description: kernel reports "badness in vsnprintf"
4274 Details : Reading from the "recovery_status" /proc file in small chunks
4275 may cause a negative length in lprocfs_obd_rd_recovery_status()
4276 call to vsnprintf() (which is otherwise harmless). Exit early
4277 if there is no more space in the output buffer.
4279 Severity : enhancement
4281 Description: clear OBD RPC statistics by writing to them
4282 Details : It is now possible to clear the OBD RPC statistics by writing
4283 to the "stats" file.
4288 Description: Client mtime is not the same on different clients after utimes
4289 Details : In some cases, the client was using the utimes() syscall on
4290 a file cached on another node. The clients now validate the
4291 ctime from the MDS + OSTs to determine which one is right.
4296 Description: Inability to activate failout mode
4297 Details : lconf script incorrectly assumed that in pythong string's numeric
4298 value is used in comparisons.
4301 Frequency : always with multiple stripes per file
4303 Description: Inefficient object allocation for mutli-stripe files
4304 Details : When selecting which OSTs to stripe files over, for files with
4305 a stripe count that divides evenly into the number of OSTs,
4306 the MDS is always picking the same starting OST for each file.
4307 Return the OST selection heuristic to the original design.
4312 Description: mount failures may take full timeout to return an error
4313 Details : Under some heavy load conditions it is possible that a
4314 failed mount can wait for the full obd_timeout interval,
4315 possibly several minutes, before reporting an error.
4316 Instead return an error as soon as the status is known.
4318 Frequency : quota enabled and large files being deleted
4320 Description: releasing more than 4GB of quota at once hangs OST
4321 Details : If a user deletes more than 4GB of files on a single OST it
4322 will cause the OST to spin in an infinite loop. Release
4323 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4328 Description: statfs data retrieved from /proc may be stale or zero
4329 Details : When reading per-device statfs data from /proc, in the
4330 {kbytes,files}_{total,free,avail} files, it may appear
4331 as zero or be out of date.
4334 Frequency : systems with MD RAID1 external journal devices
4336 Description: lconf's call to blkid is confused by RAID1 journal devices
4337 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4338 of returning all block devices that match the journal UUID.
4341 Frequency : always, for aggregate stripe size over 4GB
4343 Description: assertion fails when trying to use 4GB stripe size
4344 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
4345 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
4348 Frequency : always on ppc64
4350 Description: the first write on an ext3 filesystem with mballoc got stuck
4351 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4352 perform endianness conversion.
4354 ------------------------------------------------------------------------------
4356 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4358 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
4359 INTEROPERATE with older versions automatically. Please read the
4360 user documentation before upgrading any part of a live system.
4361 * WARNING: Lustre networking configuration changes are required with
4362 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
4365 * Support for kernels:
4366 2.6.9-22.0.2.EL (RHEL 4)
4367 2.6.5-7.244 (SLES 9)
4368 2.6.12.6 vanilla (kernel.org)
4371 Severity : enhancement
4372 Bugzilla : 7981/8208
4373 Description: Introduced Lustre Networking (LNET)
4374 Details : LNET is new networking infrastructure for Lustre, it includes
4375 a reorganized network configuration mode (see the user
4376 documentation for full details) as well as support for routing
4377 between different network fabrics. Lustre Networking Devices
4378 (LNDs) for the supported network fabrics have also been
4379 created for this new infrastructure.
4381 Severity : enhancement
4382 Description: Introduced Access control lists
4383 Details : clients can set ACLs on files and directories in order to have
4384 more fine-grained permissions than the standard Unix UGO+RWX.
4385 The MDS must be started with the "-o acl" mount option.
4387 Severity : enhancement
4388 Description: Introduced filesystem quotas
4389 Details : Administrators may now establish per-user quotas on the
4392 Severity : enhancement
4394 Description: Configuration change for the XT3
4395 The PTLLND is now used to run Lustre over Portals on the XT3
4396 The configure option(s) --with-cray-portals are no longer used.
4397 Rather --with-portals=<path-to-portals-includes> is used to
4398 enable building on the XT3. In addition to enable XT3 specific
4399 features the option --enable-cray-xt3 must be used.
4404 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
4405 Details : A race between allocating a new llcd and re-getting the llcd_lock
4406 allowed another thread to grab newly-allocated llcd.
4408 Severity : enhancement
4410 Description: 2.6 OST async journal commit and locking fix to improve performance
4411 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
4412 2.6 kernels are now async as they already were in 2.4 kernels so
4413 that they can commit concurrently with the network bulk transfer.
4414 For block-allocated files the filter allocation semaphore is held
4415 to avoid filesystem fragmentation during allocation. BKL lock
4416 removed for 2.6 xattr operations where it is no longer needed.
4421 Description: lconf incorrectly determined whether two IP networks could talk
4422 Details : In some more complicated routing and multiple-network
4423 configurations, lconf will avoid trying to make a network
4424 connection to a disjoint part of the IP space. It was doing the
4425 math incorrectly for one set of cases.
4430 Description: Fix for potential infinite loop processing records in an llog.
4431 Details : If an llog record is corrupted/zeroed, it is possible to loop
4432 forever in llog_process(). Validate the llog record length
4433 and skip the remainder of the block on error.
4436 Frequency : occasional (liblustre only)
4438 Description: liblustre could not open files whose last component is a symlink
4439 Details : sysio_path_walk() would incorrectly pass the open intent to
4440 intermediate path components.
4443 Frequency : rare (liblustre only with non-standard tuning)
4444 Bugzilla : 7201 (7350)
4445 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4446 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4447 after it has been released, only do this for referenced locks
4449 Severity : enhancement
4451 Description: specifying an (invalid) directory default stripe_size of -1
4452 would reset the directory default striping
4453 Details : stripe_size -1 was used internally to signal directory stripe
4454 removal, now use "all default" to signal dir stripe removal
4455 as a directory striping of "all default" is not useful
4458 Frequency : common for large clusters running liblustre clients
4460 Description: doing an ls when liblustre clients are running is slow
4461 Details : sending a glimpse AST to a liblustre client waits for every AST
4462 to time out, as liblustre clients will not respond. Since they
4463 cannot cache data we refresh the OST lock LVB from disk instead.
4465 Severity : enhancement
4467 Description: doing an ls at the same time as file IO can be slow
4468 Details : enqueue and other "small" requests can be blocked behind many
4469 large IO requests. Create a new OST IO portal for non-IO
4470 requests so they can be processed faster.
4473 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4475 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4476 created file with mode 000
4477 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4478 allow an ftruncate() as a non-root user to a file with mode 000.
4479 Lustre now respects this flag to disable mode checks when
4480 truncating a file owned by the user
4483 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4486 Description: Revoking locks from clients that went dead or catatonic might take
4488 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4489 cancellation of such locks instant on servers without waiting for
4490 any reply from clients. Clients drops these locks when cancel
4491 notification from server is received without replying.
4494 Frequency : liblustre-only, when liblustre client dies or becomes busy
4496 Description: Doing ls on Linux clients can take a long time with active
4498 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4499 granting such locks to it in the first place if possible. Locks
4500 are taken by proxy on the OST during the read or write and
4501 dropped immediately afterward. Add connect flags handling, do
4502 not grant locks to liblustre clients for glimpse ASTs.
4504 Severity : enhancement
4506 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4507 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4508 sequential reads and behaves badly with applications doing random
4509 reads. Improve it by reading ahead at least read region, and
4510 avoiding excessive large RPC for small reads.
4512 Severity : enhancement
4514 Description: Creating more than 1000 files for a single job may cause a load
4515 imbalance on the OSTs if there are also a large number of OSTs.
4516 Details : qos_prep_create() uses an OST index reseed value that is an
4517 even multiple of the number of available OSTs so that if the
4518 reseed happens in the middle of the object allocation it will
4519 still utilize the OSTs as uniformly as possible.
4524 Description: OST or MDS may oops in ping_evictor_main()
4525 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
4526 but doesn't restart at beginning of obd_exports_timed list
4529 Severity : enhancement
4531 Description: improve by-nid export eviction on the MDS and OST
4532 Details : allow multiple exports with the same NID to be evicted at one
4533 time without re-searching the exports list.
4536 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
4538 Description: MDS may oops in groups_free()
4539 Details : in rare race conditions a newly allocated group_info struct is
4540 freed again, and this can be NULL. The 2.4 compatibility code
4541 for groups_free() checked for a NULL pointer, but 2.6 did not.
4544 Frequency : common for liblustre clients doing little filesystem IO
4545 Bugzilla : 9352, 7313
4546 Description: server may evict liblustre clients accessing contended locks
4547 Details : if a client is granted a lock or receives a completion AST
4548 with a blocking AST already set it would not reply to the AST
4549 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
4552 Frequency : lfs setstripe, only systems with more than 160 OSTs
4554 Description: unable to set striping with a starting offset beyond OST 160
4555 Details : llapi_create_file() incorrectly limited the starting stripe
4556 index to the maximum single-file stripe count.
4559 Frequency : LDAP users only
4561 Description: lconf did not handle in-kernel recovery with LDAP properly
4562 Details : lconf/LustreDB get_refs() is searching the wrong namespace
4564 Severity : enhancement
4566 Description: bind OST threads to NUMA nodes to improve performance
4567 Details : all OST threads are uniformly bound to CPUs on a single NUMA
4568 node and do their allocations there to localize memory access
4570 Severity : enhancement
4572 Description: llmount can determine client NID directly from Myrinet (GM)
4573 Details : the client NID code from gmnalnid was moved directly into
4574 llmount, removing the need to use this or specifying the
4575 client NID explicitly when mounting GM clients with zeroconf
4578 Frequency : if client is started with down MDS
4580 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
4581 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
4582 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
4583 ptlrpc_import_delayed_req() because we didn't check if the
4584 request was interrupted, and we also didn't break out of the
4585 event loop if there was no timeout
4590 Description: data loss during non-page-aligned writes to a single file from
4591 both multiple nodes and multiple threads on one node at same time
4592 Details : updates to KMS and lsm weren't protected by common lock. Resulting
4593 inconsistency led to false short-reads, that were cached and later
4594 used by ->prepare_write() to fill in partially written page,
4595 leading to data loss.
4598 Frequency : always, if lconf --abort_recovery used
4600 Description: lconf --abort_recovery fails with 'Operation not supported'
4601 Details : lconf was attempting to abort recovery on the MDT device and not
4604 Severity : enhancement
4606 Description: remove cleanup logs
4607 Details : replace lconf-generated cleanup logs with lustre internal
4608 cleanup routines. Eliminates the need for client-cleanup and
4611 Severity : enhancement
4613 Description: add support for EAs (user and system) on lustre filesystems
4614 Details : it is now possible to store extended attributes in the Lustre
4615 client filesystem, and with the user_xattr mount option it
4616 is possible to allow users to store EAs on their files also
4618 Severity : enhancement
4620 Description: Add possibility (config option) to show minimal available OST free
4622 Details : When compiled with --enable-mindf configure option, statfs(2)
4623 (and so, df) will return least minimal free space available from
4624 all OSTs as amount of free space on FS, instead of summary of
4625 free spaces of all OSTs.
4627 Severity : enhancement
4629 Description: do not expand extent locks acquired on OST-side
4630 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
4631 by server: they are not cached anyway.
4634 Frequency : when mmap is used/binaries executed from Lustre
4636 Description: Unmmap pages before throwing them away from read cache.
4637 Details : llap_shrink cache now attempts to unmap pages before discarding
4638 them (if unmapping failed - do not discard). SLES9 kernel has
4639 extra checks that trigger if this unmapping is not done first.
4644 Description: lconf didn't resolve symlinks before checking to see whether a
4645 given mountpoint was already in use
4648 Frequency : when migrating failover services
4649 Bugzilla : 6395, 9514
4650 Description: When migrating a subset of services from a node (e.g. failback
4651 from a failover service node) the remaining services would
4652 time out and evict clients.
4653 Details : lconf --force (implied by --failover) sets the global obd_timeout
4654 to 5 seconds in order to quickly disconnect, but this caused
4655 other RPCs to time out too quickly. Do not change the global
4656 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
4658 Severity : enhancement
4659 Frequency : if MDS is started with down OST
4660 Bugzilla : 9439,5706
4661 Description: Allow startup/shutdown of an MDS without depending on the
4662 availability of the OSTs.
4663 Details : Asynchronously call mds_lov_synchronize during MDS startup.
4664 Add appropriate locking and lov-osc refcounts for safe
4665 cleaning. Add osc abort_inflight calls in case the
4666 synchronize never started.
4669 Frequency : occasional (Cray XT3 only)
4671 Description: root not authorized to access files in CRAY_PORTALS environment
4672 Details : The client process capabilities were not honoured on the MDS in
4673 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
4674 been accessed by an authorized user then root was able to access
4675 the file on the local client also. The root user capabilities
4676 are now allowed on the MDS, as this environment has secure UID.
4679 Frequency : occasional
4681 Description: ldiskfs "too long searching" message happens too often
4682 Details : A debugging message (otherwise harmless) prints too often on
4683 the OST console. This has been reduced to only happen when
4684 there are fragmentation problems on the filesystem.
4689 Description: Division by zero in statfs when all OSCs are inactive
4690 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
4691 lov_statfs divides by value returned by lov_get_stripecnt().
4695 Bugzilla : 9489, 3273
4696 Description: First write from each client to each OST was only 4kB in size,
4697 to initialize client writeback cache, which caused sub-optimal
4698 RPCs and poor layout on disk for the first writen file.
4699 Details : Clients now request an initial cache grant at (re)connect time
4700 and so that they can start streaming writes to the cache right
4701 away and always do full-sized RPCs if there is enough data.
4702 If the OST is rebooted the client also re-establishes its grant
4703 so that client cached writes will be honoured under the grant.
4708 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
4709 Details : Now I/O RPCs go to different portal number and (presumably) fast
4710 lock requests (and glimses) and other RPCs get their own service
4711 threads pool that should be able to service those RPCs
4714 Severity : enhancement
4716 Description: Ability to exchange lustre version between client and servers and
4717 issue warnings at client side if client is too old. Also for
4718 liblustre clients there is ability to refuse connection of too old
4720 Details : New 'version' field is added to connect data structure that is
4721 filled with version info. That info is later checked by server and
4725 Frequency : rare, liblustre only.
4726 Bugzilla : 9296, 9581
4727 Description: Two simultaneous writes from liblustre at offset within same page
4728 might proceed at the same time overwriting eachother with stale
4730 Details : I/O lock withing llu_file_prwv was released too early, before data
4731 actually was hitting the wire. Extended lock-holding time until
4732 server acknowledges receiving data.
4735 Frequency : extremely rare. Never observed in practice.
4737 Description: avoid generating lustre_handle cookie of 0.
4738 Details : class_handle_hash() generates handle cookies by incrementing
4739 global counter, and can hit 0 occasionaly (this is unlikely, but
4740 not impossible, because initial value of cookie counter is
4741 selected randonly). Value of 0 is used as a sentinel meaning
4742 "unassigned handle" --- avoid it. Also coalesce two critical
4743 sections in this function into one.
4745 Severity : enhancement
4747 Description: allow liblustre clients to delegate truncate locking to OST
4748 Details : To avoid overhead of locking, liblustre client instructs OST to
4749 take extent lock in ost_punch() on client's behalf. New connection
4750 flag is added to handle backward compatibility.
4752 Severity : enhancement
4753 Bugzilla : 4928, 7341, 9758
4754 Description: allow number of OST service threads to be specified
4755 Details : a module parameter allows the number of OST service threads
4756 to be specified via "options ost ost_num_threads={N}" in the
4757 OSS's /etc/modules.conf or /etc/modprobe.conf.
4761 Bugzilla : 6146, 9635, 9895
4762 Description: servers crash with bad pointer in target_handle_connect()
4763 Details : In rare cases when a client is reconnecting it was possible that
4764 the connection request was the last reference for that export.
4765 We would temporarily drop the export reference and get a new
4766 one, but this may have been the last reference and the export
4767 was just destroyed. Get new reference before dropping old one.
4769 Severity : enhancement
4770 Frequency : if client is started with failover MDS
4772 Description: Allow multiple MDS hostnames in the mount command
4773 Details : Try to read the configuration from all specified MDS
4774 hostnames during a client mount in case the "primary"
4777 Severity : enhancement
4779 Description: Stop sending data to evicted clients as soon as possible.
4780 Details : Check if the client we are about to send or are sending data to
4781 was evicted already. (Check is done every second of waiting,
4782 for which l_wait_event interface was extended to allow checking
4783 of exit condition at specified intervals).
4786 Frequency : rare, normally only when NFS exporting is done from client
4788 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
4790 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
4793 Frequency : always, if srand() is called before liblustre initialization
4795 Description: Liblustre uses system PRNG disturbing its usage by user application
4796 Details : Introduce internal to lustre fast and high-quality PRNG for
4797 lustre usage and make liblustre and some other places in generic
4798 lustre code to use it.
4800 Severity : enhancement
4801 Bugzilla : 9477, 9557, 9870
4802 Description: Verify that the MDS configuration logs are updated when xml is
4803 Details : Check if the .xml configuration logs are newer than the config
4804 logs stored on the MDS and report an error if this is the case.
4805 Request --write-conf, or allow starting with --old_conf.
4807 Severity : enhancement
4809 Description: Handle symlinks in the path when checking if Lustre is mounted.
4810 Details : Resolve intermediate symlinks when checking if a client has
4811 mounted a filesystem to avoid duplicate client mounts.
4816 Description: lconf can hit an error exception but still return success.
4817 Details : The lconf command catches the Command error exception at the top
4818 level script context and will exit with the associated exit
4819 status, but doesn't ensure that this exit status is non-zero.
4824 Description: failure of ptlrpc thread startup can cause oops
4825 Details : Starting a ptlrpc service thread can fail if there are a large
4826 number of threads or the server memory is very fragmented.
4827 Handle this without oopsing.
4830 Frequency : always, only if liblustre and non-default acceptor port was used
4832 Description: liblustre cannot connect to servers with non-default acceptor port
4833 Details : tcpnal_set_default_params() was not called and was therefore
4834 ignoring the environment varaible TCPNAL_PORT, as well as other
4835 TCPNAL_ environment variables
4840 Description: two objects could be created on the same OST for a single file
4841 Details : If an OST is down, in some cases it was possible to create two
4842 objects on a single OST for a single file. No problems other
4843 than potential performance impact and spurious error messages.
4847 Bugzilla : 5681, 9562
4848 Description: Client may oops in ll_unhash_aliases
4849 Details : Client dcache may become inconsistent in race condition.
4850 In some cases "getcwd" can fail if the current directory is
4856 Description: Inode refcounting problems in NFS export code
4857 Details : link_raw functions used to call d_instantiate without obtaining
4858 extra inode reference first.
4862 Bugzilla : 9942, 9903
4863 Description: Referencing freed requests leading to crash, memleaks with NFS.
4864 Details : We used to require that call to ll_revalidate_it was always
4865 followed by ll_lookup_it. Also with revalidate_special() it is
4866 possible to call ll_revalidate_it() twice for the same dentry
4867 even if first occurence returned success. This fix changes semantic
4868 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
4869 reference on a request referred from the intent.
4870 ll_intent_release() then releases such a request.
4873 Frequency : rare, normally benchmark loads only
4875 Description: unlinked inodes were kept in memory on the client
4876 Details : If a client is repeatedly creating and unlinking files it
4877 can accumulate a lot of stale inodes in the inode slab cache.
4878 If there is no other client load running this can cause the
4879 client node to run out of memory. Instead flush old inodes
4880 from client cache that have the same inode number as a new inode.
4883 Frequency : SLES9 2.6.5 kernel and long filenames only
4884 Bugzilla : 9969, 10379
4885 Description: utime reports stale NFS file handle
4886 Details : SLES9 uses out-of-dentry names in some cases, which confused
4887 the lustre dentry revalidation. Change it to always use the
4891 Frequency : rare, unless heavy write-truncate concurrency is continuous
4892 Bugzilla : 4180, 6984, 7171, 9963, 9331
4893 Description: OST becomes very slow and/or deadlocked during object unlink
4894 Details : filter_destroy() was holding onto the parent directory lock
4895 while truncating+unlinking objects. For very large objects this
4896 may block other threads for a long time and slow overall OST
4897 responsiveness. It may also be possible to get a lock ordering
4898 deadlock in this case, or run out of journal credits because of
4899 the combined truncate+unlink. Solution is to do object truncate
4900 first in one transaction without parent lock, and then do the
4901 final unlink in a new transaction with the parent lock. This
4902 reduces the lock hold time dramatically.
4905 Frequency : rare, 2.4 kernels only
4907 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
4908 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
4909 before unmounting their filesystem in order to avoid deadlock.
4910 The kernel_locked() function in 2.4 kernels only checks whether
4911 the kernel lock is held, not whether it is this process that is
4912 holding it as 2.6 kernels do.
4917 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
4918 Details : The client ptlrpc code may be trying to reconnect to a down
4919 server before a previous connection attempt has timed out.
4920 Increase the reconnect interval to be longer than the connection
4921 timeout interval to avoid sending duplicate connections to
4925 Frequency : echo_client brw_test command
4927 Description: fix echo_client to work with OST preallocated code
4928 Details : OST preallocation code (5137) didn't take echo_client IO path
4929 into account: echo_client calls filter methods outside of any
4930 OST thread and, hence, there is no per-thread preallocated
4931 pages and buffers to use. Solution: hijack pga pages for IO. As
4932 a byproduct, this avoids unnecessary data copying.
4936 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
4937 Description: Client can oops in mdc_commit_close() after open replay
4938 Details : It was possible for the MDS to return an open request with no
4939 transaction number in mds_finish_transno() if the client was
4940 evicted, but without actually returning an error. Clients
4941 would later try to replay that open and may trip an assertion
4942 Simplify the client close codepath, and always return an error
4943 from the MDS in case the open is not successful.
4946 Frequency : rare, 2.6 OSTs only
4948 Description: OST may deadlock under high load on fragmented files
4949 Details : If there was a heavy load and highly-fragmented OST filesystems
4950 it was possible to have all the OST threads deadlock waiting on
4951 allocation of biovecs, because the biovecs were not released
4952 until the entire RPC IO was completed. Instead, release biovecs
4953 as soon as they are complete to ensure forward IO progress.
4955 Severity : enhancement
4957 Description: Support for specifying external journal device at mount
4958 Details : If an OST or MDS device is formatted with an external journal
4959 device, this device major/minor is stored in the ext3 superblock
4960 and may not be valid for failover. Allow detecting and
4961 specifying the external journal at mount time.
4966 Description: Mounting an MDS with pending unlinked files may cause oops
4967 Details : target_finish_recovery() calls mds_postrecov() which returned
4968 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
4969 considers this an error and immediately begins cleaning up the
4970 lov, just after starting the mds_lov process
4972 Severity : enhancement
4974 Description: Implement 'lfs df' to report actual free space on per-OST basis
4975 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
4976 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
4977 usage of objects; '-h' to report in human readable format.
4979 ------------------------------------------------------------------------------
4981 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
4988 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
4989 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
4990 at startup during orphan unlinking. Instead, skip buddy update
4991 before setup, as it will be regenerated after recovery is complete.
4996 Description: If an OST is inactive, its locks might reference stale inodes.
4997 Details : lov_change_cbdata() must iterate over all namespaces, even if
4998 they are inactive to clear inode references from the lock.
5000 Severity : enhancement
5001 Frequency : occasional, if non-standard max_dirty_mb used
5003 Description: Client will block write RPCs if not enough grant
5004 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
5005 then the client will block writes while waiting for another RPC
5006 to complete instead of consuming its dirty limit. With change
5007 we get improved performance when max_dirty_mb is small.
5009 Severity : enhancement
5010 Bugzilla : 3389, 6253
5011 Description: Add support for supplementary groups on the MDS.
5012 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
5013 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
5014 lookups for user supplementary groups into a cache.
5018 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
5019 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
5020 for liblustre O_EXCL is enforced by MDS.
5024 Description: readdir never returns NULL in liblustre.
5025 Details : Corrected llu_iop_getdirentries logic, to return offset of next
5026 dentry in struct dirent.
5030 Frequency : liblustre only, depends on application IO pattern
5031 Description: liblustre clients evicted if not contacting servers
5032 Details : Don't put liblustre clients into the ping_evictor list, so
5033 they will not be evicted by the pinger ever.
5035 Severity : enhancement
5037 Description: Add ability to evict clients by NID from MDS.
5038 Details : By echoing "nid:$NID" string into
5039 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
5040 $NID would be instantly evicted from this MDS and from all active
5041 OSTs connected to it.
5045 Description: Do not query file size twice, somewhat slowing stat(2) calls.
5046 Details : lookup_it_finish() used to query file size from OSTs that was not
5051 Description: service threads change working directory to that of init
5052 Details : Starting lustre service threads may pin the working directory
5053 of the parent thread, making that filesystem busy. Threads
5054 now change to the working directory of init to avoid this.
5058 Frequency : during shutdown only
5059 Description: shutdown with a failed MDS or OST can cause unmount to hang
5060 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
5065 Frequency : 2.6 only
5066 Description: chmod/chown may include an extra supplementary group
5067 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
5068 supplementary group and if none is specified this is used.
5071 Bugzilla : 5479 (6816)
5073 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
5074 Details : If lookup is in progress on a file that is unlinked we might try
5075 to revalidate the inode and fail in revalidate after lookup is
5076 complete and ll_file_open() enqueues the open again but
5077 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
5080 Frequency : always, if lconf --abort_recovery used
5082 Description: lconf --abort_recovery fails with 'Operation not supported'
5083 Details : lconf was attempting to abort recovery on the MDT device and not
5086 ------------------------------------------------------------------------------
5088 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
5093 Frequency : rare (only unsupported configurations with a node running as an
5095 Bugzilla : 6514, 5137
5096 Description: Mounting a Lustre file system on a node running as an OST could
5098 Details : OSTs now preallocates memory needed to write out data at
5099 startup, instead of when needed, to avoid having to
5100 allocate memory in possibly low memory situations.
5101 Specifically, if the file system is mounted on on OST,
5102 memory pressure could force it to try to write out data,
5103 which it needed to allocate memory to do. Due to the low
5104 memory, it would be unable to do so and the node would
5105 become unresponsive.
5107 Severity : enhancement
5109 Description: Addition of lconf --service command line option
5110 Details : lconf now accepts a '--service <arg>' option, which is
5111 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
5113 Severity : enhancement
5115 Description: Failover mode is now the default for OSTs.
5116 Details : By default, OSTs will now run in failover mode. To return to
5117 the old behaviour, add '--failout' to the lmc line for OSTs.
5119 Severity : enhancement
5121 Description: Health checks are now provided for MDS and OSTs
5122 Details : Additional detailed health check information on MSD and OSTs
5123 is now provided through the procfs health_check value.
5126 Frequency : occasional, depends on IO load
5128 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
5129 after numerous create/delete cycles
5130 Details : The ext3 inode allocation policy would not allocate new inodes
5131 very well on the OSTs because there are no new directories
5132 being created. Instead we look for groups with free space if
5133 the parent directories are nearly full.
5138 Description: Network or server problems during mount may cause partially
5139 mounted clients instead of returning an error.
5140 Details : The config llog parsing code may overwrite the error return
5141 code during mount error handling, returning success instead
5147 Description: MDS can fail to allocate large reply buffers
5148 Details : After long uptimes the MDS can fail to allocate large reply
5149 buffers (e.g. zconf client mount config records) due to memory
5150 fragmentation or consumption by the buffer cache. Preallocate
5151 some large reply buffers so that these replies can be sent even
5152 under memory pressure.
5156 Frequency : rare (liblustre)
5157 Description: fsx running with liblustre complained that using truncate() to
5158 extend the file doesn't work. This patch corrects that issue.
5159 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
5160 ATTR_SIZE and lsm use in llu_setattr_raw.
5164 Frequency : rare, only 2.6 kernels
5165 Description: Unusual file access patterns on the MDS may result in inode
5166 data being lost in very rare circumstances.
5167 Details : Bad interaction between the ea-in-inode patch and the "no-read"
5168 code in the 2.6 kernel caused the inode and/or EA data not to
5169 be read from disk, causing single-file corruption.
5173 Frequency : rare, only 2.6 filesystems using extents
5174 Description: Heavy concurrent write and delete load may cause data corruption.
5175 Details : It was possible under high-load situations to have an extent
5176 metadata block in the block device cache from a just-unlinked
5177 file overwrite a newly-allocated data block. We now unmap any
5178 metadata buffers that alias just-allocated data blocks.
5182 Frequency : filesystems with default stripe_count larger than 77
5183 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
5184 Details : lconf specifies an inode size of 4096 bytes when the default
5185 stripe_count is larger than 77. This conflicts with the default
5186 inode density of 1 per 4096 bytes. Allocate smaller inodes in
5187 this case to avoid pinning too much memory for large EAs.
5189 ------------------------------------------------------------------------------
5191 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
5196 Frequency : rare (extremely heavy IO load with hundreds of clients)
5198 Description: Client is evicted, gets IO error writing to file
5199 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
5200 caused clients to be evicted for AST timeouts. The fixes in
5201 bug 5192 mean we no longer need to have such short AST timeouts
5202 so ldlm_timeout has been increased.
5205 Frequency : occasional during --force or --failover shutdown under load
5206 Bugzilla : 5949, 4834
5207 Description: Server oops/LBUG if stopped with --force or --failover under load
5208 Details : a collection of import/export refcount and cleanup ordering
5209 issues fixed for safer force cleanup
5212 Frequency : only filesystems larger than 120 OSTs
5213 Bugzilla : 5990, 6223
5214 Description: lfs getstripe would oops on a very large filesystem
5215 Details : lov_getconfig used kfree on vmalloc'd memory
5218 Frequency : only filesystems exporting via NFS to Solaris 10 clients
5219 Bugzilla : 6242, 6243
5220 Description: reading from files that had been truncated to a non-zero size
5221 but never opened returned no data
5222 Details : ll_file_read() reads zeros from no-object files to EOF
5227 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
5228 Details : There's little harm in aborting MDS/OSS recovery and letting it
5229 try again, so I removed the LASSERT and return an error instead.
5231 Severity : enhancement
5233 Description: New debugging infrastructure for tracking down data corruption
5234 Details : The I/O checksum code was replaced to: (a) control it at runtime,
5235 (b) cover more of the client-side code path, and (c) try to narrow
5236 down where problems occurred
5240 Bugzilla : 3819, 4364, 4397, 6313
5241 Description: Racing close and eviction MDS could cause assertion in mds_close
5242 Details : It was possible to get multiple mfd references during close and
5243 client eviction, leading to one thread referencing a freed mfd.
5245 Severity: : enhancement
5246 Bugzilla : 3262, 6359
5247 Description: Attempts to reconnect to servers are now more aggressive.
5248 Details : This builds on the enhanced upcall-less recovery that was added
5249 in 1.4.2. When trying to reconnect to servers, clients will
5250 now try each server in the failover group every 10 seconds. By
5251 default, clients would previously try one server every 25 seconds.
5256 Description: After recovery, certain operations trigger a failed
5257 assertion on a client.
5258 Details : Failing over an mds, using lconf -d --failover, while a
5259 client was doing a readdir() call would cause the client to
5260 LBUG after recovery completed and the readdir() was resent.
5262 Severity : enhancement
5264 Description: Default groups are now added by lconf
5265 Details : You can now run lconf --group <servicename> without having to
5266 manually add groups with lmc.
5269 Frequency : occasional
5271 Description: Nodes with an elan id of 0 trigger a failed assertion
5274 Frequency : always when accessing e.g. tty/console device nodes
5276 Description: tty and some other devices nodes cannot be used on lustre
5277 Details : file's private_data field is used by device data and lustre
5278 values in there got lost. New field was added to struct file to
5279 store fs-specific private data.
5282 Frequency : when exporting Lustre via NFS
5284 Description: NFSD failed occasionally when looking up a path component
5285 Details : NFSD is looking up ".." which was broken in ext3 directories
5286 that had grown large enough to become hashed.
5289 Frequency : Clusters with multiple interfaces not on the same subnet
5291 Description: Nodes will repeatedly try to reconnect to an interface which it
5292 cannot reach and report an error to the log.
5293 Details : Extra peer list entries will be created by lconf with some peers
5294 unreachable. lconf now validates the peer before adding it.
5297 Frequency : Only if a default stripe is set on the filesystem root.
5299 Description: Setting a default stripe on the filesystem root prevented the
5300 filesystem from being remounted.
5301 Details : The client was sending extra request flags in the root getattr
5302 request and did not allocate a reply buffer for the dir EA.
5305 Frequency : occasional, higher if lots of files are accessed by one client
5306 Bugzilla : 6159, 6097
5307 Description: Client trips assertion regarding lsm mismatch/magic
5308 Details : While revalidating inodes the VFS looks up inodes with ifind()
5309 and in rare cases can find an inode that is being freed.
5310 The ll_test_inode() code will free the lsm during ifind()
5311 when it finds an existing inode and then the VFS later attaches
5312 this free lsm to a new inode.
5316 Bugzilla : 6422, 7030
5317 Description: MDS deadlock between mkdir and client eviction
5318 Details : Creating a new file via mkdir or mknod (starting a transaction
5319 and getting the ns lock) can deadlock with client eviction
5320 (gets ns lock and trying to finish a synchronous transaction).
5323 Frequency : occasional
5324 Description: While starting a server, the fsfilt_ext3 module could not be
5326 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
5327 kernels. Previously, lconf would still use the ext3 name
5328 when trying to load modules. Now, it will correctly use
5329 ext3 on 2.4 and ldiskfs on 2.6.
5331 Severity : enhancement
5332 Description: The default stripe count has been changed to 1
5333 Details : The interpretation of the default stripe count (0, to lfs
5334 or lmc) has been changed to mean striping across a single
5335 OST, rather than all available. For general usage we have
5336 found a stripe count of 1 or 2 works best.
5338 Severity : enhancement
5339 Description: Add support for compiling against Cray portals.
5340 Details : Conditional compiling for some areas that are different
5344 Frequency : occasional
5345 Bugzilla : 6409, 6834
5346 Description: Creating files with an explicit stripe count may lead to
5347 a failed assertion on the MDS
5348 Details : If some OSTs are full or unavailable, creating files may
5349 trigger a failed assertion on the MDS. Now, Lustre will
5350 try to use other servers or return an error to the
5354 Frequency : occasional
5356 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
5357 caused lock timeout during readahead (since 1.4.2).
5358 Details : Processes doing readahead might match a lock that hasn't been
5359 granted yet if there are overlapping and conflicting lock
5360 requests. The readahead process waits on ungranted lock
5361 (original lock is CBPENDING), while OST waits for that process
5362 to cancel CBPENDING read lock and eventually evicts client.
5364 Severity : enhancement
5366 Description: Initial enabling of flock support for clients
5367 Details : Implements fcntl advisory locking and file status functions.
5368 This feature is provided as an optional mount flag (default
5369 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
5370 locking are implemented yet, and those that are are not guaranteed
5371 to be completely correct in production environments.
5372 mount -t lustre -o [flock|noflock] ...
5375 Frequency : occasional
5377 Description: OSTs running 2.4 kernels but with extents enabled might trip an
5378 assertion in the ext3 JBD (journaling) layer.
5379 Details : The b_committed_data struct is protected by the big kernel lock
5380 in 2.4 kernels, serializing journal_commit_transaction() and
5381 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
5382 access to this struct. In 2.6 kernels there is finer grained
5383 locking to improve SMP performance of the JBD layer.
5387 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
5392 Description: merge and cleanup kernel patches.
5395 -----------------------------------------------------------------------------
5397 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
5399 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
5400 versions of Lustre. Please update all servers and clients to
5401 version 1.4.2 or later at the same time. You must also run
5402 "lconf --write-conf {config}.xml" on the MDS while it is stopped
5403 to update the configuration logs.
5405 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
5406 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
5407 - swab llog records of type '0' so we get proper header size/idx (5861)
5408 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
5409 - fix rename of one directory over another leaking an inode (5953)
5410 - avoid SetPageDirty on 2.6 (5981)
5411 - don't re-add just-being-destroyed locks to the waiting list (5653)
5412 - when creating new directories, inherit the parent's custom
5413 striping settings if present parent (3048)
5414 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
5415 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
5416 - don't hold client locks on temporary worklist from l_lru (5666)
5417 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
5418 - automatically evict dead clients (5921)
5419 - Update file size properly in create+truncate+fstat case (6196)
5420 - Do not unhash mountpoint dentries, do not allow removal of
5422 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
5423 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
5424 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
5425 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
5426 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
5427 - Correctly handle reads of files with no objects (6243)
5428 - lctl recover will also mark a device active if deactivate used (5933)
5430 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
5431 - allow --write-conf on an MDS with different nettype than client (5619)
5432 - don't write config llogs to MDS for mounts not from that MDS (5617)
5433 - lconf should create multiple TCP connections from a client (5201)
5434 - init scripts are now turned off by default; run chkconfig --on
5435 lustre and chkconfig --on lustrefs to use them
5436 - upcalls are no longer needed for clients to recover to failover
5438 - add --abort-recovery option to lconf to abort recovery on device
5440 - add support for an arbitrary number of OSTs (3026)
5441 - Quota support protocol changes.
5442 - forward compatibility changes to wire structs (6007)
5443 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5444 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5445 - improved "lustre status" script
5446 - initialize blocksize for non-regular files (6062)
5447 - added --disable-server and --disable-client configure options (5782)
5448 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5449 - Vanilla 2.4.29 support
5450 - increase maximum number of obd devices to 520 (6242)
5451 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5452 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5453 SLES 9 2.6.5 kernels
5454 - limit stripes per file to 160 (the maximum EA size) (6093)
5456 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5459 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5460 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5461 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5462 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5463 - small ext3 extents cleanups and fixes (5733)
5464 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5465 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5466 - avoid needless client->OST connect, fix handle mismatch (5317)
5467 - fix DLM error path that led to out-of-sync client, long delays (5779)
5468 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5469 - fix several locking issues related to i_size (5492,5624,5654,5672)
5470 - don't move pending lock onto export if it is already evicted (5683)
5471 - fix kernel oops when creating .foo in unlinked directory (5548)
5472 - fix deadlock in obdfilter statistics vs. object create (5811)
5473 - use time_{before,after} to avoid timer jiffies wrap (5882)
5474 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5475 - Do not leak request if server was not able to process it (5154)
5476 - If mds_open unable to find parent dir, make that negative lookup(5154)
5477 - don't create new directories with extent-mapping (5909, 5936)
5479 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5480 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5481 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5482 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5483 - support for RHEL4 kernel on i686 (5773)
5484 - provide error messages when incompatible logs are encountered (5898)
5486 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5487 * version 1.4.0.10 (1.4.1 release candidate 1)
5489 - don't keep a lock reference when lock is not granted (4238)
5490 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5491 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5492 - drop import inflight refcount on signal_completed_replay error (5255)
5493 - unlock page after async write error during send (3677)
5494 - handle missing objects in filter_preprw_read properly (5265)
5495 - no transno return for symlink open, don't save no-trasno open (3440)
5496 - don't try to complete elan receive that already failed (4012)
5497 - free RPC server reply state on error (5406)
5498 - clean up thread from ptlrpc_start_thread() on error (5160)
5499 - readahead could read extra page into cache that wasn't ejected (5388)
5500 - prevent races in class_attach/setup/cleanup/detach (5260)
5501 - don't dereference de->d_inode after l_dput of de (5458)
5502 - use "int" for stripe value returned from lock_to_stripe (5544)
5503 - mballoc allocation and error-checking fixes in 2.6 (5504)
5504 - block device patches to fix I/O request sizes in 2.6 (5482)
5505 - look up hostnames for IB nals (5602)
5506 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5507 - don't start multiple acceptors for the same port (5277)
5508 - fix incorrect LASSERT in mds_getattr_name (5635)
5509 - export a proc file for general "ping" checking (5628)
5510 - fix "lfs check" to not block when the MDS is down (5628)
5512 - service request history (4965)
5513 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5514 - create an "evict_client" /proc entry on OSTs, like the MDS has
5515 - fix mount usage message, return errors per mount(8) (5168)
5516 - change grep [] to grep "[]" in tests so they work in more UMLs
5517 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5518 - remove ancient LOV_MAGIC_V0 EA support (5047)
5519 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5520 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5521 - allow readahead allocations to fail when low on memory (5383)
5522 - mmap locking landed again, after considerable improvement (2828)
5523 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
5525 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
5528 - send OST transaction number in read/write reply to free req (4966)
5529 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5530 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5531 - the watchdog thread now runs as interruptible (5246)
5532 - flock/lockf fixes (but it's still disabled, pending 5135)
5533 - don't use EXT3 constants in llite code (5094)
5534 - memory shortage at startup could cause assertion (5176)
5536 - reorganization of lov code
5537 - single portals codebase
5539 - add extents/mballoc support (5025)
5540 - direct I/O reads in the obdfilter (4048)
5541 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
5543 tbd 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 - handle missing objects in filter_preprw_read properly (5265)
5551 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5552 - drop import inflight refcount on signal_completed_replay error (5255)
5553 - unlock page after async write error during send (3677)
5554 - return original error code on reconstructed replies (3761)
5555 - no transno return for symlink open, don't save no-trasno open (3440)
5557 - add pid to ldlm debugging output (4922)
5558 - bump the watchdog timeouts -- we can't handle 30sec yet
5559 - extra debugging for orphan dentry/inode bug (5259)
5561 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
5564 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
5565 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
5566 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
5567 - fix NULL dereference in /proc/sys/portals/routes (4827)
5568 - allow failed mdc_close() operations to be interrupted (4561)
5569 - stop precreate on OST before MDS would time out on it (4778)
5570 - don't send partial-page writes before EOF from client (4410)
5571 - discard client grant for sub-page writes on large-page clients (4520)
5572 - don't free dentries not owned by NFS code, check generation (4806)
5573 - fix lsm leak if mds_create_objects() fails (4801)
5574 - limit debug_daemon file size, always print CERROR messages (4789)
5575 - use transno after validating reply (3892)
5576 - process timed out requests if import state changes (3754)
5577 - update mtime on OST during writes, return in glimpse (4829)
5578 - add mkfsoptions to LDAP (4679)
5579 - use ->max_readahead method instead of zapping global ra (5039)
5580 - don't interrupt __l_wait_event() during strace
5582 - add software watchdogs to catch hung threads quickly (4941)
5583 - make lustrefs init script start after nfs is mounted
5584 - fix CWARN/ERROR duplication (4930)
5585 - return async write errors to application if possible (2248)
5586 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
5587 - print NAL number in %x format (4645)
5588 - update barely-supported suse-2.4.21-171 series (4842)
5589 - support for sles 9 %post scripts
5590 - support for building 2.6 kernel-source packages
5591 - support for sles km_* packages
5593 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
5596 - ignore -ENOENT errors in osc_destroy (3639)
5597 - notify osc create thread that OSC is being cleaned up (4600)
5598 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
5599 - reconstruct ost_handle() like mds_handle() (4657)
5600 - create a new thread to do import eviction to avoid deadlock (3969)
5601 - let lconf resolve symlinked-to devices (4629)
5602 - don't unlink "objects" from directory with default EA (4554)
5603 - hold socknal file ref over connect in case target is down (4394)
5604 - allow more than 32000 subdirectories in a single directory (3244)
5605 - fix blocks count for O_DIRECT writes (3751)
5606 - OST returns ENOSPC from object create when no space left (4539)
5607 - don't send truncate RPC if file size isn't changing (4410)
5608 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
5609 - bind to privileged port in socknal and tcpnal (3689)
5611 - rate limit CERROR/CWARN console message to avoid overload (4519)
5612 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
5613 - basic mmap support (3918)
5614 - kernel patch series update from b1_4 (4711)
5616 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
5619 - avoid crash during MDS cleanup with OST shut down (2775)
5620 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
5621 - don't use bad inodes on the MDS (3744)
5622 - dynamic object preallocation to improve recovery speed (4236)
5623 - don't hold spinlock over lock dumping or change debug flags (4401)
5624 - don't zero obd_dev when it is force cleaned (3651)
5625 - print grants to console if they go negative (4431)
5626 - "lctl deactivate" will stop automatic recovery attempts (3406)
5627 - look for existing locks in ldlm_handle_enqueue() (3764)
5628 - don't resolve lock handle twice in recovery avoiding race (4401)
5629 - revalidate should check working dir is a directory (4134)
5631 - don't always mark "slow" obdfilter messages as errors (4418)
5633 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
5636 - don't close LustreDB during write_conf until it is done (3860)
5637 - fix typo in lconf for_each_profile (3821)
5638 - allow dumping logs from multiple threads at one time (3820)
5639 - don't allow multiple threads in OSC recovery (3812)
5640 - fix debug_size parameters (3864)
5641 - fix mds_postrecov to initialize import for llog ctxt (3121)
5642 - replace config semaphore with spinlock (3306)
5643 - be sure to send a reply for a CANCEL rpc with bad export (3863)
5644 - don't allow enqueue to complete on a destroyed export (3822)
5645 - down write_lock before checking llog header bitmap (3825)
5646 - recover from lock replay timeout (3764)
5647 - up llog sem before sending rpc (3652)
5648 - reduce ns lock hold times when setting kms (3267)
5649 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
5650 - fix NULL deref and obd_dev leak on setup error (3312)
5651 - replace some LBUG about llog ops with error handling (3841)
5652 - don't match INVALID dentries from d_lookup and spin (3784)
5653 - hold dcache_lock while marking dentries INVALID and hashing (4255)
5654 - fix invalid assertion in ptlrpc_set_wait (3880)
5656 - add libwrap support for the TCP acceptor (3996)
5657 - add /proc/sys/portals/routes for non-root route listing (3994)
5658 - allow setting MDS UUID in .xml (2580)
5659 - print the stack of a process that LBUGs (4228)
5661 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
5664 - don't cleanup request in ll_file_open() on failed MDS open (3430)
5665 - make sure to unset replay flag from failed open requests (3440)
5666 - if default stripe count is 0, use OST count for inode size (3636)
5667 - update parent mtime/ctime on client for create/unlink (2611)
5668 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
5669 - free recovery state on server during a forced cleanup (3571)
5670 - unregister_reply for resent reqs (3063)
5671 - loop back devices mounting and status check on 2.6 (3563)
5672 - fix resource-creation race that can provoke i_size == 0 (3513)
5673 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
5674 - more debugging for page-accounting assertion (3746)
5675 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
5676 - avoid deadlock after precreation failure (3758)
5677 - fix race and lock order deadlock in orphan handling (3450, 3750)
5678 - add validity checks when grabbing inodes from l_ast_data (3599)
5680 - add /proc/.../recovery_status to obdfilter (3428)
5681 - lightweight CDEBUG infrastructure, debug daemon (3668)
5682 - change default OSC RPC parameters to be better on small clusters
5683 - turn off OST read cache for files smaller than 32MB
5684 - install man pages and include them in rpms (3100)
5685 - add new init script for (un)mounting lustre filesystems (2593)
5686 - run chkconfig in %post for init scripts (3701)
5687 - drop scimac NAL (unmaintained)
5689 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
5692 - clean kiobufs before and after use (3485)
5693 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
5694 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
5695 - change init script to fail more clearly if not run as root (1528)
5696 - allow clients to reconnect during replay (1742)
5697 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
5698 - don't do DNS lookups on NIDs too small for IP addresses (3442)
5699 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
5700 - fix cond_resched (3554)
5701 - only evict unfinished clients after recovery (3515)
5702 - allow bulk resend, prevent data loss (3570)
5703 - dynamic ptlrpc request buffer allocation (2102)
5704 - don't allow unlinking open directory if it isn't empty (2904)
5705 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
5706 - remove extraneous obd dereference causing LASSERT failure (3334)
5707 - don't use get_cycles() when creating temp. files on the mds (3156)
5708 - hold i_sem when setting i_size in ll_extent_lock() (3564)
5709 - handle EEXIST for set-stripe, set proper directory name (3336)
5711 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
5712 - fix ksocknal_fmb_callback() error messages (2918)
5714 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
5717 - don't copy lvb into (possibly NULL) reply on error (2983)
5718 - don't deref dentry after dput, don't free lvb on error (2922)
5719 - use the kms to determine writeback rpc length (2947)
5720 - increment oti_logcookies when osc is inactive (2948)
5721 - update client's i_blocks count via lvb messages (2543)
5722 - handle intent open/close of special files properly (1557)
5723 - mount MDS with errors=remount-ro, like obdfilter (2009)
5724 - initialize lock handle to avoid ASSERT on error cleanup (3057)
5725 - don't use cancelling-locks' kms values (2947)
5726 - use highest lock extent for kms, not last one (2925)
5727 - don't dereference ERR_PTR() dentry in error handling path (3107)
5728 - fix thread race in portals_debug_dumplog() (3122)
5729 - create lprocfs device entries at setup instead of at attach (1519)
5730 - common AST error handler, don't evict client on completion race (3145)
5731 - zero nameidata in detach_mnt in 2.6 (3118)
5732 - verify d_inode after revalidate_special is valid in 2.6 (3116)
5733 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
5734 - initialize RPC timeout timer earlier for 2.6 (3219)
5735 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
5736 - print nal/nid for unknown nid (3258)
5737 - additional checks for oscc recovery before doing precreate (3284)
5738 - fix ll_extent_lock() error return code for 64-bit systems (3043)
5739 - don't crash in mdc_close for bad permissions on open (3285)
5740 - zero i_rdev for non-device files (3147)
5741 - clear page->private before handing to FS, better assertion (3119)
5742 - tune the read pipeline (3236)
5743 - fix incorrect decref of invalidated dentry (2350)
5744 - provide read-ahead stats and refine rpc in flight stats (3328)
5745 - don't hold journal transaction open across create RPC (3313)
5746 - update atime on MDS at close time (3265)
5747 - close LDAP connection when recovering to avoid server load (3315)
5748 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
5749 - don't leak open file on MDS after open resend (3325)
5750 - serialize filter_precreate and filter_destroy_precreated (3329)
5751 - loop device shouldn't call sync_dev() for nul device (3092)
5752 - clear page cache after eviction (2766)
5753 - resynchronize MDS->OST in background (2824)
5754 - refuse to mount the same filesystem twice on same mountpoint (3394)
5755 - allow llmount to create routes for mounting behind routers (3320)
5756 - push lock cancellation to blocking thread for glimpse ASTs (3409)
5757 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
5758 - fix rare problem with rename on htree directories (3417)
5760 - allow default OST striping configuration per directory (1414)
5761 - fix compilation for qswnal for 2.6 kernels (3125)
5762 - increase maximum number of MDS request buffers for large systems
5763 - change liblustreapi to be useful for external progs like lfsck (3098)
5764 - increase local configuration timeout for slow disks (3353)
5765 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
5767 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5770 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
5771 - don't overwrite extent policy data in reply if lock was blocked (2901)
5772 - drop filter export grants atomically with removal from device (2663)
5773 - del obd_self_export from work_list in class_disconnect_exports (2908)
5774 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
5775 - swab reply message in mdc_close, other PPC fixes (2464)
5776 - fix destroying of named logs (2325)
5777 - overwrite old logs when running lconf --write_conf (2264)
5778 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
5779 - fix race in target_handle_connect (2898)
5780 - mds_reint_create() should take same inode create lock (2926)
5781 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
5782 - don't close files for self_export to avoid uninitialized obd (2936)
5783 - allow MDS with the same name as client node (2939)
5784 - hold dentry reference for closed log files for unlink (2325)
5785 - reserve space for all logs during transactions (2059)
5786 - don't evict page beyond end of stripe extent (2925)
5787 - don't oops on a deleted current working directory (2399)
5788 - handle hard links to targets without a parent properly (2517)
5789 - don't dereference NULL lock when racing during eviction (2867)
5790 - don't grow lock extents when lots of conflicting locks (2919)
5792 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
5795 - account for cache space usage on clients to avoid data loss (974)
5796 - lfsck support in lustre kernel code (2349)
5797 - reduce journal credits needed for BRW writes (2370)
5798 - orphan handling to avoid losing space on client/server crashes
5799 - ptlrpcd can be blocked, stopping ALL progress (2477)
5800 - use lock value blocks to assist in proper KMS, faster stat (1021)
5801 - takes i_sem instead of DLM locks internally on obdfilter (2720)
5802 - recovery for initial connections (2355)
5803 - fixes for mds_cleanup_orphans (1934)
5804 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
5805 - block all file creations until orphan recovery completes (1901)
5806 - client remove rq_connection from request struct (2423)
5807 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
5808 - recovery timer race (2670)
5809 - mdc_close recovey bug (2532)
5810 - ptlrpc cleanup bug (2710)
5811 - mds timeout on local locks (2588)
5812 - namespace lock held during RPCs (2431)
5813 - handle interrupted sync write properly (2503)
5814 - don't try to handle a message that hasn't been replied to (2699)
5815 - client assert failure during cleanup after abort recovery (2701)
5816 - leak mdc device after failed mount (2712)
5817 - ptlrpc_check_set allows timedout requests to complete (2714)
5818 - wait for inflight reqs when ptlrpcd finishes (2710)
5819 - make sure unregistered services are removed from the srv_list
5820 - reset bulk XID's when resending them (caught by 1138 test)
5821 - unregister_bulk after timeout
5822 - fix lconf error (2694)
5823 - handle write after unfinished setstripe, stripe-only getstripe (2388)
5824 - readahead locks pages, leaves pending causing memory pressure (2673)
5825 - increase OST request buffers to 4096 on large machines (2729)
5826 - fix up permission of existing directories in simple_mkdir (2661)
5827 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
5828 - don't assign transno to errored transactions (2742)
5829 - don't delete objects on OST if given a bogus objid from MDS (2751)
5830 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
5831 - if rq_no_resend, then timeout request after recovery (2432)
5832 - fix MDS llog_logid record size, 64-bit array alignment (2733)
5833 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
5834 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
5835 - MDS recovery shouldn't delete objects that it has given out (2730)
5836 - if enqueue arrives after completion, don't clobber LVB (2819)
5837 - don't unlock pages twice when trigger_group_io returns error (2814)
5838 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
5839 - don't write pages to disk if there was an error (1450)
5840 - don't ping imports that have recovery disabled (2676)
5841 - take buffered bytes into account when balancing socknal conn (2817)
5842 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
5843 - reconnect unlink llog connection after MDS reconnects to OST (2816)
5844 - remove little-endian swabbing of llog records (1987)
5845 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
5846 - retry reposting request buffers if they fail (1191)
5847 - grow extent at grant time to avoid granting a revoked lock (2809)
5848 - lock revoke doesn't evict page if covered by a second lock (2765)
5849 - disable VM readahead to avoid reading outside lock extents (2805)
5851 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
5852 - updated kernel patches for hp-2.4.20 kernel (2681)
5854 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
5857 - fix truncated write corruption (2366)
5858 - fix for failed assertion in iopen_connect_dentry (1792,2517)
5860 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
5861 - ensure len(uuid) < 37 in lmc (1171)
5862 - fix ia64 OOPS in llog_test (2255)
5863 - zero end of page at obdfilter for partial page writes (2648)
5864 - don't leave stale dentries around after renames (bug 2428)
5865 - fix timeouts when evicting a client with a single lock held (2642)
5866 - set deadline for the initial HELLO message to drain (2634)
5867 - print out dotted-quad IP addresses in the socknal (2302)
5869 - additional debugging for MDS client eviction problem (2443)
5870 - fix mkfsoptions support for osts (2603, 2604)
5872 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
5875 - add series for the vanilla 2.6.0 kernel
5876 - add series for the vanilla 2.4.24 kernel
5877 - add series for a cray x86/64 UL kernel drop
5878 - fix xattr patches for the vanilla 2.4.19 series
5880 - generate true UUIDs in lmc (1171)
5881 - have portals stack dumping break in UML (2466)
5882 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
5883 - allocate with _NOFS in ldlm to avoid deadlock (1933)
5884 - wake callback waiting threads on client eviction (2460)
5885 - Add --ptldebug and --subsystem to lmc (1719)
5886 - update assertion to allow safe interrupt allocation
5887 - set rq_no_resend for cancel requests (2432)
5888 - recalculate ptlrpcd timeout after resend (2494)
5889 - call vfs_rmdir when removing pending directories (2368)
5890 - fix renaming a file to itself (2429)
5891 - lmc creates a default one-stripe lov (2454)
5892 - expand procfs space to handle large clusters (2326)
5893 - increase UML stack to avoid overflow
5894 - update lconf's list of debug and subsystem masks
5895 - fix lfs find --obd (2510)
5896 - /proc tunable for disabling filter read caching (2591)
5897 - stop rpm packages from altering slapd.conf (2301)
5898 - disable nagle in the socknal under 0conf (2578)
5899 - choose mds inode size based on stripe count (2572)
5900 - fix kernel-source rpm problems (2516)
5902 - add --disable-doc to avoid pdf generation (2421)
5903 - update documentation, tests, type-os, comments
5904 - avoid format warnings on ia64
5905 - remove the TOE NAL
5906 - tiny code cleanups by removing unused fields
5908 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
5911 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
5912 - remove the most blatant lies from BUILDING file (2371)
5913 - change default debug level to reasonable production setting
5914 - reduce client side cache size to reduce cache flush time
5915 - reduce max RPCs in flight to avoid unnecessary file fragmentation
5916 - make TCP zerocopy and pinger support enabled by default (2476)
5917 - sync writes completed after process exits caused crashes (2319)
5918 - maintain correct mount count on the MDS (2356)
5919 - backout 1557, because 2316 wasn't really fixed
5920 - better file I/O statistics gathering in /proc
5921 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
5922 - another kernel patch to fix zero-copy TCP function export
5923 - don't take duplicate lock when processing re-sent getattr (2420)
5924 - lctl uses obd_self_export instead of creating new conn (2353)
5925 - MDS/OST recovery case which requires object creation asserted (2425)
5926 - move lfs from /usr/sbin to /usr/bin in packages
5927 - fix race between mds_client_add and mds_client_free (2417)
5928 - use kmalloc instead of slabs in portals (2430)
5929 - don't create duplicate records when a failover MDS is present (2442)
5930 - remove unnecessary mount age check (2332)
5931 - don't remove directory inodes from locks prematurely (2451)
5932 - don't break if MDS service name is the same as hostname (2103)
5933 - fix races in client write RPC generation when cache full (2482)
5935 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
5938 - remove now-unused request->rq_obd (278)
5939 - if an allocation fails, print out how much memory we've used (1933)
5940 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
5941 - add the "configurable stack size" patch to most series files (1256)
5942 - ability to write large log records, for 100+ OST configs (2306)
5943 - fix NULL deref when filter_prep fails (2314)
5944 - fix operator precedence error in filter_sync
5945 - dynamic allocation of socknal TX descriptors (2315)
5946 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
5947 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
5948 - ensure that CWARN messages in Portals always get to the syslog
5949 - __init/__exit are not for prototype decls (ldlm_init/exit)
5950 - x86-64 compile warning fixes
5951 - fix gateway LMC keyword conflict (2318)
5952 - fix MDS lock inversions in getattr/reint paths (1844)
5953 - fix a rare lock re-ordering bug, which caused deadlock (2322)
5954 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
5955 - DLM race condition prevented some lock evictions (2328)
5956 - ENOMEM detection and retry on socknal sends (2230)
5957 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
5958 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
5959 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
5960 - better error messages when a client is rejected during recovery (1505)
5961 - avoid cancelling locks which were never granted, after failure (2330)
5962 - fix i_sem/journal inversion in mds_client_add (2333)
5963 - fix truncate/getattr lock cycle deadlock (2334)
5964 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
5965 - fix two rare exit paths which could leak an l_lock() ref (2321)
5966 - fencepost error in MDS/OST orphan recovery (2226)
5967 - make log record alignment 8 bytes (1988)
5968 - lstripe now fails when requested offset > ost_count (2237)
5969 - ensure that all kernel series have a complete list.h (1607)
5970 - fix crashes in special-file operations (2316)
5971 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
5973 - allow configurable automake binary, for testing new versions
5974 - small update to the lfs documentation
5976 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
5978 * fix negative export reference count in fsfilt_sync (2312)
5980 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
5981 * release candidate 0.9.1
5983 - orphans are moved into the PENDING directory for possible recovery
5984 - replayed opens now open by fid for orphan/rename safety (1042)
5985 - last close of an orphan inode generates a transno (683)
5986 - chdir() and mount() now pin the directory entry (1020)
5987 - avoid CERROR in normal ll_setattr_raw() error case (1500)
5988 - discard very old requests without processing them (1502)
5989 - remove some common, well-understood CERRORs (1505)
5990 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
5991 - clear "grant" flags in OST replies until OST grant code lands (1644)
5992 - fix read performance by not clobbering i_blksize on client (1598)
5993 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
5994 - make LRU size a /proc tunable, clears locks when reduced (707)
5995 - fix some lprocfs rot that prevented ptlbd from loading (1732)
5996 - server locks take references on exports now (1558)
5997 - build fixes for 2.4.20-rh trees (1663)
5998 - return an error from lov_create if all OSCs are inactive (1751)
5999 - fix import levels when a reconnect happens without a timeout (1597)
6000 - exit early from mds_open if we get a lookup error (1749)
6001 - partial page read at EOF wouldn't wait for disk before sending (1642)
6002 - avoid NULL deref in obdfilter when reading page past EOF (1592)
6003 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
6004 - fix LBUG in ll_it_open_error with rc = -2 (1861)
6005 - write/truncate lock inversion (1639)
6006 - Don't auto-load obdclass, portals modules during cleanup (1495)
6007 - fix timestamps from jumping to "now" (1763)
6008 - extra journal assertions (1648)
6009 - add an extra multiunlink test (1771)
6010 - fix read_record/write_record API (1776)
6011 - fix leak of offset_extent, possible incorrect i_size later (1772)
6012 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
6013 - Debugging for the kqswnal_get_idle_tx problems (1820)
6014 - Allow recovery to be attempted multiple times (1536)
6015 - Write out MDS last_rcvd file after it is first created (1600)
6016 - Fix tx_descriptor leak in failed transmit situations (1827)
6017 - ext3 journaling fixes for assertion failure after IO error (1871)
6018 - class_export_put() on freed export after completion AST error (1896)
6019 - Fix revalidate looping in VFS (1322)
6020 - Don't access a freed export during MDS_REINT timeout (1521)
6021 - Add open-unlink recovery support on the MDS (1673,1764)
6022 - Return an error if no MDS data was read from last_rcvd (1946)
6023 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
6024 - Don't setup a disk device that doesn't match exported UUID (317)
6025 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
6026 - avoid committing NULL handle in force close
6027 - local.sh is now a one-stripe LOV configuration
6028 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
6029 - fix ext3 htree duplicate directory entry corruption (1516)
6030 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
6031 - update to new LOV EA format (2097)
6032 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
6034 2003-06-15 Phil Schwan <phil@clusterfs.com>
6037 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
6038 - per-import/export recovery handling (958, 931, 959)
6039 - multiple last-rcvd slots, for serving multiple FSes (949)
6040 - connections are again shared between multiple imp/exports (963, 964)
6041 - "umount -f" would hang if any requests needed to be sent (393, 978)
6042 - avoid pinning large req buffer by copying for queued messages (989)
6043 - add "uuid" to "lctl device" command to help upcalls (991)
6044 - "open" RPCs with transnos would confuse recovery counters (1037)
6045 - do proper endian conversion of all wire messages (288, 340, 891)
6046 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
6047 - call waiting locks callback from LDLM recovery thread (1127, 1151)
6048 - fix ptlrpc_connection leak in target_handle_connect (1174)
6049 - fix import refcounting bug in OST and MDS cleanup (1134)
6050 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
6051 - fix very unlikely obd_types race condition (501)
6052 - remove osc_open hack for echo_client (1187)
6053 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
6054 - a failure in read_inode2 leads to deadlock (1139)
6055 - cancel ack-locks as soon as transaction is committed (1072)
6056 - fix major leaks and crashes in the bulk I/O path (937, 1057)
6057 - make sure to commitrw after any preprw to avoid deadlock (1162)
6058 - failing to execute a file in a lustre FS would lock inode (1203)
6059 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
6060 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
6061 - obd_uuid2tgt would walk off the end of the list (1255)
6062 - on IA64 the osc would give portals incorrect bulk size (1258)
6063 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
6064 - fix lock inversion caused by new llite matching code (1282)
6065 - limit the number of dirty pages on a client to 10MB (1286)
6066 - timed out locks were not being corrected cancelled (1289)
6067 - fix O_DIRECT above 4GB on IA-32 (1292)
6068 * major user-visible changes
6069 - fail out/fail over policy now controlled by the upcall (993)
6071 - add OBD_PING to check server availability and failure (954)
6072 - lustre messages are now sent in sending host order (288, 340, 891)
6073 - add eadatalen to MDS getattr reply (340)
6074 - OST read replies may contain second buffer, with per-page status (593)
6076 2003-03-11 Phil Schwan <phil@clusterfs.com>
6079 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
6080 - failed open()s could cause deadlock; fixed (867, 869)
6081 - stop cancelling OST locks when files are closed (481)
6082 - overlapping XID spaces caused network corruption (851, 853)
6083 - fix unsafe fsfilt counter arithmetic; change to atomic_t
6084 - setattr_raw added, to do single-RPC, server-side setattrs
6085 - lmc/lconf syntax change for OST UUIDs
6086 - fix crashy race condition between ptlrpc_free_req and osc_close
6087 - don't use request in mdc_enqueue if we hit a timeout (889)
6088 - don't set the inode i_size for regular files from the MDS (896)
6089 - handle out of order completion AST (842)
6090 - don't LBUG if a lock request times out after receiving AST (913)
6091 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
6092 - if a bad lock AST arrives, send an error instead of dropping entirely
6093 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
6094 - fix leak in bulk IO when only partially completed (899, 900, 926)
6095 - fix O_DIRECT for ia64 (55)
6096 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
6097 - C-z after timeout could hang a process forever; fixed (977)
6099 - client-side I/O cache (678, 924, 929, 941, 970)
6101 - READPAGE and SETATTRs which don't take server-side locks get
6104 2003-02-11 Phil Schwan <phil@clusterfs.com>
6107 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
6108 - Fix fsfilt_extN_readpage() to read a full page of directory
6109 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
6110 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
6111 - Fix imbalanced LOV object allocation and out-of-bound access (469)
6112 - Most intent operations were removed, in favour of a new RPC mode
6113 that does a single RPC to the server and bypasses most of the VFS
6114 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
6115 - Aggressively cancel local locks on DLM servers
6116 - mds_reint_unlink sends EA to the client if it's the last nlink.
6117 client uses that EA to unlink OST objects.
6118 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
6119 - recursive symlinks were fixed (439)
6120 - fixed NULL deref in DEBUG_REQ
6121 - filter_update_lastobjid no longer calls sync, which annoyed extN
6122 - fixed multi-client small-writes to a single file problem (445)
6123 - fixed mtime updates during file writes (607)
6124 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
6125 - fixed bug in obd_brw_read/write() (under guise of testing 367)
6126 - fixed Linux OST size reporting problem (444, 656)
6127 - OST now updates object mtime with writes or setattr (607, 619)
6128 - client verifies file size before zeroing page past EOF (445)
6129 - OST now writes last allocated objid to disk with allocation (108)
6130 - LOV on echo now works (409)
6132 - mds_reint_unlink sends a new buffer, with the EA included. this
6133 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
6134 set if a regular file was being unlinked, and it was the last link
6135 - use PtlGet from the target for bulk writes (315)
6136 - OST now updates object mtime with writes or setattr (607, 619)
6137 - LDLM now has a grant-time callback to revalidate locked items, if
6139 - Many MDS operations were reorganized to combat race conditions
6141 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
6142 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
6144 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
6147 - Fully reactivate OST imports after reconnection (512, others)
6148 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
6149 - More graceful error handling for truncating on dead OST (515)
6150 - Don't error out unless we're actually accessing dead stripes (474)
6151 - Fix garbage sizes when stripes are missing (410)
6152 - LRU counters were broken, causing constant lock purge (433, 432)
6153 - garbage on read from stripes with failed OSTs (441)
6154 - mark OSCs as active before reconnecting during recovery (438)
6155 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
6156 - lfind did not preserve OST order in output (443)
6157 - symlinks cause hung clients, incorrect data (439)
6158 - stop dereferencing request after dropping refcount (457)
6159 - don't LASSERT(spin_is_locked) on non-SMP (455)
6160 - fixes for many rename() bugs
6161 - fstat didn't correctly synchronize attributes (399)
6162 - server must handle lock cancellation during blocking AST prep (487)
6163 - bulk descriptors were free()d too soon (511)
6164 - fix paths in lconf, which would load incorrect modules (451, 507)
6165 - fix confusing lconf 'host not found' error message (386)
6166 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
6167 - fix race condition in mdc_blocking_ast() for inode access (526)
6168 - fix lov_unpackmd() unpacking wrong number of stripes (537)
6169 - fix lov_set_osc_active() marking wrong OSC inactive (440)
6170 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
6171 - fix multiple writes of stripe MD to MDS (358, maybe 519)
6172 - fix lstripe in several ways (kernel side) (527)
6173 - fix request leak in ldlm_cli_enqueue (262)
6174 - incorrect OSC was marked inactive after OST failure
6175 - call mds_fs_cleanup before unmounting filesystem (524)
6176 - fix races between taking ns_lock and ldlm_lock_change_resource
6177 - fix races updating LOV export open file list
6178 - fix lov_enqueue error path, avoid decref-ing bad lock handle
6179 - fix recovery NULL deref in ldlm_cli_cancel_unused
6180 - fix some DLM races by using new hash table for lock handles (419)
6181 - permit the client to specify desired inodes, at replay
6182 - duplicate requests when we queue them for replay reintegration
6183 - fix last_rcvd offset calculation
6184 - sync after each recovered transaction, so we always make progress
6185 - never, not always, ERESTART requests without transnos
6186 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
6188 - skip replay if the MDS says that the client is already connected
6189 - don't check for a recovery-enabled export to match lctl's UUID
6190 - don't INC_USE_COUNT for phantom exports
6191 - don't crash when cleaning up phantom exports (567)
6192 - don't double-finish or set replay data for errored mdc_open requests
6193 - abort requests when they time out, so we don't get old replies
6194 - send/receive replies for AST messages again
6195 - if the client says that it doesn't have the lock, cancel it on the
6197 - if we timeout during I/O, don't try to cancel an in-use lock; instead
6198 mark it as destroyed, it will all work out when decref is called
6199 - fix module use counts (22, 581)
6201 - ASTs now expect a reply (server cancels lock on error reply)
6203 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
6206 - fix many simultaneous client startup (392)
6207 - fix dentry->d_it clobbering
6208 - credentials weren't being shipped for readdir/getattr operations
6209 - remove invalid assertions triggered during some concurrent MD
6211 - proper Lustre versions added (336, 389)
6212 - fix memory leak for create error case (398)
6213 - fix LOV locking bug that would get cli/srv out of sync
6214 - fix echo client over LOV (409)
6215 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
6216 - fix double-O_EXCL intent crash (424)
6217 - avoid sending multiple lock CANCELs (352)
6219 - MDS can do multi-client recovery (modulo bugs in new code)
6221 - many updates, edits, cleanups
6223 2002-11-18 Phil Schwan <phil@clusterfs.com>
6226 - fix null d_it dereference (346)
6227 - fix full OST/dbench hang (333)
6228 - fix permission problem with file removal (286)
6229 - fix removal of OSCs from LOV when they fail
6230 - fix NULL deref during bulk timeout (214)
6231 - fix problems related to multiple filesystems on one MDS (241)
6232 - fixed serious subtle metadata locking bugs
6233 - free locks on clients when inodes are removed due to memory
6235 - fix inode pointer in lock data (285)
6236 - partial support for multiple MDS on a single host (241)
6237 - data locks weren't cancelled at clear_inode time (290, 311)
6238 - intent locks could lead to unbounded lock growth (205)
6239 - added a maximum lock count, an LRU list, and a flusher
6240 - fix multiple rename (365)
6241 - properly abstracted the echo client
6242 - OSC locked 1 byte too many; fixed
6243 - rewrote brw callback code:
6244 - fixed recovery bugs related to LOVs (306)
6245 - fixed too-many-pages-in-one-write crash (191)
6246 - fixed (again) crash in sync_io_timeout (214)
6247 - probably fixed callback-related race (385)
6249 - Add capability to MDS protocol
6250 - LDLM cancellations and callbacks on different portals
6252 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
6255 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
6256 - defer open object destruction to close time (601981, 138)
6257 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
6258 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
6259 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
6260 - don't call obd_setattr in ll_file_release for destroyed objects
6261 * protocol change to lustre_msg: move |version| and add |flags|
6262 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
6263 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
6264 * added replay of create, unlink, link and rename operations during
6265 MDS failover; recovery should be much more robust now
6266 * remove failed OSCs from LOVs (only lov_create uses this so far)
6267 * the lustre-HOWTO was brought (more) up to date (582544)
6269 2002-10-23 Phil Schwan <phil@clusterfs.com>
6272 - in-use dentries weren't being reused properly (617851)
6273 - prevent multiple LDLM setup (599178)
6274 - fix LOV size calculations for truncate (617853)
6275 - fix client handling of MDS intent errors (POSIX)
6276 - fix permission bug in lovstripe.c test (624321)
6277 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
6278 - truncate past end of file could corrupt data
6279 - proper cleanup after timeouts, crashes, etc (592524, 550815)
6280 - a race in recovery could return ETIMEDOUT to apps (623947)
6281 - building outside the source directory was fixed
6282 * the lustre-HOWTO was brought (more) up to date (582544)
6283 * major progress was made on recovery functionality
6285 2002-10-10 Phil Schwan <phil@clusterfs.com>
6288 - recovery deadlock fix
6289 - rm -rf causes LBUG fix (617817)
6290 - file open by multiple tasks fix (618962)
6291 - directory permissions bugs (602707 and 620007)
6292 - journal_stop fixed with locking (611313)
6293 - O_APPEND failures resolved (618273, perhaps 614459)
6294 - lconf PATH fix (619770)
6295 - IA64 build fix (621450)
6296 - RPC buffer sizes scale with amount of memory
6298 2002-10-01 Phil Schwan <phil@clusterfs.com>
6301 - locks would be cancelled without throwing away data pages,
6302 resulting in inconsistent data (605627)
6303 - inode attributes were not always being refreshed (605627, 612449)
6304 - lconf now continues to cleanup after lctl reports an error
6305 - MDS now enforces user permissions (602707)
6306 - lprocfs cleanup fixed, but not yet enabled (614157)
6307 - fixed infinite server hang, should a client not respond to an AST
6308 - avoid going into recovery if user calls readlink() with a buffer
6309 that's too small (613941)
6310 - AST RPCs no longer require replies (614867) -- this may be changed
6311 - don't crash server if client sends an IOV that's too big (611336)
6312 - fixed lock conversion deadlock (611892)
6313 - fixed the following of symlinks (614622)
6314 * recovery: the server can remove locks from a client that dies, other
6315 clients can make progress
6316 * more extN patch fixes
6317 * compile-time configurable ptlrpc buffer allocations
6319 - collaborative read cache document
6320 - Lustre Lite Performance CDR document-in-progress
6322 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6325 - fix typo in patch-2.4.18
6327 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6330 - clear ptlrpc request each time in handle_incoming_request()
6331 - unlink of files now destroys the object on the OST
6333 2002-09-19 Peter Braam <braam@clusterfs.com>
6335 * add hard link support
6336 * change obdfile creation method
6337 * kernel patch changed
6339 2002-09-19 Peter Braam <braam@clusterfs.com>
6342 - stack overflow bug in extN fixed
6344 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
6346 * documentation updates
6347 - add man pages for config tools
6348 - update tests/README to describe testing with new config tools
6349 - finish metadata API descriptions
6350 * bug fixes and cleanups
6351 - statfs workaround for 16TB limit
6352 - LOV stripe allocation improved, can stripe on subset of OSTs
6353 - LOV file size/IO offset was wrong for files > 4GB in size
6354 - object EA data was being dropped, caused files to be unreadable
6355 - memory overflow with non-LOV OST caused memory corruption
6356 - fixed regression tests to work with new config tools, obdfilter
6357 - fixed bug when directory size became larger than 1 block
6358 - fixed bug (for single client case) when PWD was deleted
6359 - invalidate local directory pages when doing intent-based ops
6360 - avoid LDLM oops when lock callback contained bad data
6362 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
6364 * documentation updates
6365 * bug fixes and cleanups
6366 - configuration tools
6369 - 64-bit compile warnings
6370 - 64-bit internal statfs data
6372 * test_brw on persistent OST devices
6374 * lprocfs (disabled)
6376 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
6378 * documentation updates
6379 * bug fixes and cleanups
6380 * configuration tools
6382 2002-08-30 Peter J. Braam <braam@clusterfs.com>
6385 * many small fixes to 0_5_4
6386 * io/network handling
6387 * thinkos in MDS operations
6389 2002-08-24 Peter J. Braam <braam@clusterfs.com>
6392 * crucial basic fixes to 0.5.3
6393 * IOR, Iozone work over Elan
6396 2002-08-07 Phil Schwan <phil@clusterfs.com>
6397 * version 0_5_3, our first alpha
6398 * we use the new Portals iovs
6399 * documentation updates
6400 * bug fixes and cleanups
6401 * small changes in the DLM wire protocol
6403 2002-07-25 Peter J. Braam <braam@clusterfs.com>
6404 * version 0_5_1 with some initial stability,
6405 * locking on MD and file I/O.
6406 * documentation updates
6407 * several bug fixes since 0.5.0
6408 * small changes in wire protocol
6410 2002-07-18 Phil Schwan <phil@clusterfs.com>
6412 * delivered as Lustre Light Alpha
6413 * fixed a crash after handling invalid MDS requests
6414 * fixed directory pages for architectures with non-4k pages sizes
6416 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
6417 * release version v0_4_4
6418 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
6420 * Fixes a number of interruption problems with OST operations.
6421 * Update documentation for portals header changes
6422 * Move all wire protocol structs/defines to lustre_idl.h
6423 * Fixes symlink length bug.
6424 * Add tcpdump to repository.
6426 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
6427 * release version v0_4_3
6428 * Fixes statfs for inodes on extN.
6429 * Fixes bug in runtests which would delete /etc/hosts.
6430 * Use 64-bit object IDs wherever possible (not into VFS though)
6431 Remove ost_get_info, which is unused by lustre, and out of date.
6433 2002-07-03 Peter Braam <braam@clusterfs.com>
6434 * release version v0_4_2 Fixes a lookup error (type not passed)
6435 * move forward to head of Portals
6436 * move forward to latest Lustre kernel
6438 2002-06-25 Peter Braam <braam@clusterfs.com>
6439 * release version v0_4_1. Hopefully stable on single node use.