3 * See http://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
4 for currently supported client and server kernel versions.
5 * Server known to build on patched kernels:
6 2.6.32-431.20.3.el6 (RHEL6.5)
7 3.0.101-0.31 (SLES11 SP3)
8 * Client known to build on unpatched kernels:
9 2.6.32-431.20.3.el6 (RHEL6.5)
10 3.0.101-0.31 (SLES11 SP3)
13 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
14 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
15 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
16 stack overflow. For more information, please refer to bugzilla 17630.
17 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
18 the re-exporting nodes to be mounted with "32bitapi" mount option
20 --------------------------------------------------------------------------------
22 10-31-2013 Intel Corporation
24 * See http://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
25 * Server builds for kernels:
26 2.6.32-358.18.1.el6 (RHEL6.4)
27 * Client builds for unpatched kernels:
28 2.6.32-358.18.1.el6 (RHEL6.4)
29 3.0.92-0.8 (SLES11 SP3)
31 * Recommended e2fsprogs version: 1.42.7.wc2 or newer
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 builds for kernels:
44 2.6.32-358.6.2.el6 (RHEL6.4)
45 * Client builds for unpatched kernels:
46 2.6.32-358.6.2.el6 (RHEL6.4)
47 3.0.74-0.6.6 (SLES11 SP2)
49 * Recommended e2fsprogs version: 1.42.7.wc1 or newer
50 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
51 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
52 stack overflow. For more information, please refer to bugzilla 17630.
53 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
54 the re-exporting nodes to be mounted with "32bitapi" mount option
56 --------------------------------------------------------------------------------
58 9-30-2012 Whamcloud, Inc.
60 * See http://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
61 * Server support for kernels:
62 2.6.32-279.5.1.el6 (RHEL6)
63 * Client support for unpatched kernels:
64 2.6.32-279.5.1.el6 (RHEL6)
66 2.6.18-238.19.1.el5 (RHEL5)
67 2.6.16.60-0.69.1 (SLES10)
68 * Recommended e2fsprogs version: 1.41.90.wc4
69 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
70 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
71 stack overflow. For more information, please refer to bugzilla 17630.
72 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
73 the re-exporting nodes to be mounted with "32bitapi" mount option
75 Severity : enhancement
77 Description: Update RHEL6.3 kernel to 2.6.32-279.5.1.el6.
79 Severity : enhancement
81 Description: Update RHEL6.3 kernel to 2.6.32-279.2.1.el6.
83 Severity : enhancement
85 Description: Update RHEL6.2 kernel to 2.6.32-220.17.1.el6
87 Severity : enhancement
89 Description: Update RHEL6.2 kernel to 2.6.32-220.13.1.el6
91 Severity : enhancement
93 Description: Update RHEL6.2 kernel to 2.6.32-220.7.1.el6
95 --------------------------------------------------------------------------------
97 04-30-2012 Whamcloud, Inc.
99 * See http://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
100 * Server support for kernels:
101 2.6.32-220.4.2.el6 (RHEL6)
102 * Client support for unpatched kernels:
103 2.6.32-220.4.2.el6 (RHEL6)
105 2.6.18-238.19.1.el5 (RHEL5)
106 2.6.16.60-0.69.1 (SLES10)
107 * Recommended e2fsprogs version: 1.41.90.wc4
108 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
109 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
110 stack overflow. For more information, please refer to bugzilla 17630.
111 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
112 the re-exporting nodes to be mounted with "32bitapi" mount option
114 Severity : enhancement
116 Description: Update RHEL6.2 kernel to 2.6.32-220.4.2.el6
118 Severity : enhancement
120 Description: Update RHEL6.2 kernel to 2.6.32-220.4.1.el6
122 Severity : enhancement
124 Description: Update RHEL6.2 kernel to 2.6.32-220.el6
126 Severity : enhancement
128 Description: Update RHEL6.1 kernel to 2.6.32-131.17.1.el6
130 --------------------------------------------------------------------------------
132 09-30-2011 Whamcloud, Inc.
134 * Server support for kernels:
135 2.6.18-238.19.1.el5 (RHEL5)
136 2.6.32-131.6.1.el6 (RHEL6)
137 * Client support for unpatched kernels:
138 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
139 2.6.18-238.19.1.el5 (RHEL5)
140 2.6.32-131.6.1.el6 (RHEL6)
141 2.6.29.4-167.fc11 (FC11)
142 2.6.16.60-0.69.1 (SLES10)
144 * Recommended e2fsprogs version: 1.41.90.wc3
145 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
146 Lustre filesystem with 4K stack may cause a stack overflow. For more
147 information, please refer to bugzilla 17630.
148 * NFSv4 reexport requires Lustre client on the re-exporting nodes to be
149 mounted with "32bitapi" mount option
151 Severity : enhancement
152 Description: Update RHEL6.1 kernel to 2.6.32-131.6.1.el6
154 Severity : enhancement
155 Description: Update RHEL5.6 kernel to 2.6.18-238.19.1.el5
157 Severity : enhancement
158 Description: Update RHEL5.6 kernel to 2.6.18-238.9.1.el5
161 Frequency : only with SLES10
163 Description: Use OFED "KMP" provided by Novell
164 Details : SLES10 SP3 ships with OFED in a separate "KMP" package.
165 Lustre is now built against this package. That means you need to
166 install the ofed-kmp package from Novell for the patchless client.
167 Note that the ofed-kmp that Novell ships may not exactly match the
168 kernel version but should still be compatible.
170 Severity : enhancement
172 Description: Update RHEL5.5 kernel to 2.6.18-194.17.1.el5
173 Update OEL5.5 kernel to 2.6.18-194.17.1.0.1.el5.
175 --------------------------------------------------------------------------------
177 2010-07-15 Oracle, Inc.
179 * Server support for kernels:
180 2.6.18-164.11.1.el5 (RHEL 5)
181 2.6.18-164.11.1.0.1.el5 (OEL 5)
182 * Client support for unpatched kernels:
183 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
184 2.6.18-164.11.1.el5 (RHEL 5),
185 2.6.18-164.11.1.0.1.el5 (OEL 5)
186 2.6.16.60-0.42.8 (SLES 10),
188 2.6.29.4-167.fc11 (FC11)
189 * Recommended e2fsprogs version: 1.41.10-sun2
191 Severity : enhancement
193 Description: Distribute lfs_migrate script with lustre
194 Details : The lfs_migrate mentioned in the manual is now included in the
195 lustre tree in lustre/scripts and will distributed in lustre RPMs.
197 Severity : enhancement
199 Description: Failover nid list added to import proc file.
201 Severity : enhancement
203 Description: "som_preview" mount option required to use SOM
204 Details : The clients will need to use som_preview mount option
208 Description: Lustre does not do 1MB IOs to HW RAID
209 Details : Bump MAX_PHYS/HW_SEGMENTS and SG_ALL to 256 in the RHEL5 kernel.
211 Severity : enhancement
213 Description: lfs quota output is not very convenient for awk/sed-parsing
214 Details : Some positions in lfs quota output table could be empty or
215 non-empty which made it hard to parse it with scripts, now
216 a dash is put instead of space where there is not supposed
221 Description: Update to OFED 1.5.1
222 Details : OFED 1.5.1 will be used in environments that don't use
223 vendor supplied OFED.
227 Description: Remove lnet. prefix from lctl params display
228 Details : Also renamed lnet.memused to lnet_memused to
233 Description: Use conf_param -d to remove permanant settings.
237 Description: Add list_param -R
238 Details : lctl list_param now supports recursive mode.
241 Bugzilla : 21829, 22312
242 Description: Fix llobdstat.
243 Details : 1. Now updating the write_bytes counter
244 2. Port script fixes from 1.8.
248 Description: decrease the usage of memory on clients.
249 Details : 1. On clients, recycle dentries and inodes unused.
250 2. Delete the code related to ll_deathrow(att 6215 in bug 1443). It
255 Description: llobdstat fix and enhancement
256 Details : add a counter to set a limit to how many samples will be returned
257 fix a wildcard in the path to limit to obdfilter stats only
261 Description: lfs check only allowed for root
262 Details : Code cleanups and added a regresion test of non-root use of
267 Description: MDS Oopses in ptlrpc_dispatch_difficult_reply()/__list_add()
268 Details : Fix hr->hr_index overload.
272 Description: Fix FIEMAP functionality.
276 Description: FC11 patchless client support added
280 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and
281 OEL5.4 kernel to 2.6.18-164.11.1.0.1.el5.
285 Description: New acc-sm script and various test-framework enhancments.
289 Description: Previously landed patch was reverted.
293 Description: Update SLES10 kernel to 2.6.16.60-0.42.8 (patchless)
297 Description: Add ldiskfs maxdirsize mount option.
301 Description: Fix mkfs.lustre support for 16TB LUNs.
305 Description: Print mmp_check_interval
306 Details : print mmp_check_interval and make it possible to abort
307 mount operation in case it takes too long.
311 Description: Add cascading_rw to lustre mpi tests.
315 Description: Store SOM data in EA.
316 Details : Also several SOM related cleanups.
320 Description: Add refcount to md_open_data
324 Description: Use refcount on lov_request_set to prevent memory corruption.
328 Description: Statahead: use parent dir's i_mutex for synchronization.
332 Description: loadgen improvements.
336 Description: Enable peer health detection in ksocklnd and ko2iblnd.
340 Description: Control DCACHE_LUSTRE_INVALID flag with MDS_INODE_LOOKCUP lock
344 Description: Fix race on obd_devcie:md_stats
348 Description: Fix SLES 11 client builds.
349 Details : Resolved uninitalized pointer gcc warnings.
353 Description: Allow non-root access for "lfs check".
354 Details : Added a check in obd_class_ioctl() for OBD_IOC_PING_TARGET.
358 Description: Clear imp_force_reconnect correctly in ptlrpc_connect_interpret()
360 Severity : enhancement
362 Description: Update kernel to RHEL5.4 2.6.18-164.6.1.el5 and
363 OEL5 2.6.18-164.6.1.0.1.el5(Both in-kernel OFED enabled).
367 Description: Sanity test_27p fails.
368 Details : OBD_FAIL_OST_ENOSPC has been moved to filter_statfs().
369 Sanity tests 27 were cleaned up and improved.
373 Description: Unify req format on client/servers
374 Details : Use new req_capsule API [almost] everywhere instead of old PTLRPC
375 buffers and swabbers approach..
380 Description: drop join file stuff.
382 Severity : enhancement
384 Description: Add 2.6.27 and SLES11 (patchless client) support
388 Description: Fix in ptlrpc_expire_one_request() to print the signed time
391 Severity : enhancement
393 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
396 Severity : enhancement
398 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
400 Severity : enhancement
402 Description: correctly handle big reply message.
403 Details : send LNet event if reply is bigger then buffer and adjust this buffer
408 Description: Drop unnecessary __GFP_NOMEMALLOC flag from filter_get_page()
410 Severity : enhancement
412 Description: Add support for OEL5.
416 Description: Conf-sanity.sh 50g test - deactivated OST should not cause a panic.
418 Severity : enhancement
420 Description: "weak-modules" support
421 Details : Implement "weak-modules" support which enables kernel modules
422 to be used with any kernel that implements the same kABI. In
423 order to achieve this modules are now installed in
424 /lib/modules/$(uname -r)/updates/kernel on all distributions.
428 Description: client could not reconnect to OST because of an active request.
429 Details : The client now retries to reconnect to the same server, if a
430 connect request failed with EBUSY or -EAGAIN. Test case 35b
431 conf-sanity.sh has been created.
435 Description: Don't allow make backward step on assignin osc next id.
436 Details : race between allocation next id and ll_sync thread can be cause
437 of set wrong osc next id and can be kill valid ost objects.
441 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
445 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
449 Description: Error handling in osc_statfs_interpret() has been improved.
450 Details : Check in osc_statfs_interpret() for EBADR.
454 Description: Do not send statfs() requests to OSTs disabled by administrator.
455 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
457 Severity : enhancement
459 Description: Added sanity-quota test to for lfs(1) to work with numeric uid/gid.
461 Severity : enhancement
463 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
465 Severity : enhancement
467 Description: Enhancement for lfs(1) command to use numeric uid/gid.
471 Description: Improved cleanup procedure for conf-sanity test_46a.
475 Description: Do not start bulk transfer, if client tried to reconnect.
476 Details : Do not start bulk transfer, if client tried to reconnect,
477 but failed due to active rpcs. Check for exp_abort_active_req
478 has been implemented.
482 Description: Prevent inconsistences between linux and lustre mount structures.
483 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
484 Make the sleep interruptible.
486 Severity : enhancement
488 Description: provide server to client comms path
489 Details : server to client communications path via new LDLM_SET_INFO rpc
493 Description: truncate starts GFP_FS allocation under transaction causing deadlock
494 Details : ldiskfs_truncate calls grab_cache_page which may start page
495 allocation under an open transaction. This may lead to
496 calling prune_icache with consequent lustre reentrance.
500 Description: send correctly lsm on open replay
501 Details : MDS is trust to LSM size on replay open, but client can set wrong size
506 Description: enable client interface failover
507 Details : When a child reconnects from another NID, properly update export
508 nid hash position and ldlm reverse import.
510 Severity : enhancement
512 Description: Communicate OST degraded/readonly state via statfs to MDS
513 Details : Flags in the statfs returned from OSTs indicate whether the
514 OST is in a degraded RAID state, or if the filesystem has
515 turned read-only after a filesystem error is detected.
520 Description: don't panic if EPROTO was hit when reading symlink
521 Details : correctly handling request reference in error cases.
523 Severity : enhancement
525 Description: Add LustreNetLink, a kernel-userspace communcation path.
527 Severity : enhancement
529 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
534 Description: access to llog context before init.
535 Details : move handling CATALOGS file at osc layer and forbid access to llog
539 Frequency : rare, only when client is writing to OST on local node
541 Description: Avoid deadlock for local client writes
542 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
543 memory freeing context. This allows OST threads to set the
544 PF_MEMALLOC flag on task structures in order to allocate memory
545 from reserved pools and complete IO.
546 Use GFP_HIGHUSER for OST allocations for non-local client writes,
547 so that the OST threads generate memory pressure and allow
548 inactive pages to be reclaimed.
550 Severity : enhancement
552 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
557 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
558 Details : move ldlm namespace creation in setup phase to avoid grab
559 _lprocfs_lock with cli_sem held.
563 Description: Temporarily disable grant shrink.
564 Details : Disable the feature for debugging.
567 Frequency : only when formatting test filesystems
569 Description: Unable to run several mkfs.lustre on loop devices at the same time
570 Details : mkfs.lustre returns error 256 on the concurrent loop devices
571 formatting. The solution is to proper handle the error.
573 Severity : enhancement
575 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
577 Severity : enhancement
579 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
584 Description: resolve race between obd_disconnect and class_disconnect_exports
585 Details : if obd_disconnect will be called to already disconnected export he
586 forget release one reference and osc module can't unloaded.
588 Severity : enhancement
590 Description: Allow tuning service thread via /proc
591 Details : For each service a new
592 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
593 created that can be used to set min/max thread counts, and get the
594 current number of running threads.
599 Descriptoin: don't return error if have partially created objects for file.
600 Details : lov_update_create_set uses set->set_success as index for created
601 objects, so if some requests will be failed, they will have hole
602 at end of array and qos_shrink_lsm for allocate correct lsm.
604 Severity : enhancement
606 Description: Update OFED support to 1.4.1
611 Description: Reduce small size read RPC
612 Details : Set read-ahead limit for every file and only do read-ahead when
613 available read-ahead pages are bigger than 1M to avoid small size
616 Severity : enhancement
618 Description: move AT tunable parameters for more consistent usage
619 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
621 Severity : enhancement
623 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
624 Details : allow skip disconnected ost for send statfs request and hide error
630 Description: fix racy locking of mballoc block bitmaps causing BUG
631 Details : The locking of the mballoc buddy bitmap and the in-memory
632 block bitmap was using two different spin locks in some
633 cases. This made it possible to incorrectly access the
634 mballoc bitmap while another process was modifying it,
635 causing a sanity assertion to fail. While no on-disk corruption
636 was reported, there was some risk of this happening.
639 Frequency : rare, on llog test 6
641 Descriptoin: don't allow connect to already connected import
642 Details : allowing connect to already connected import is hide connecting problem.
647 Descriptoin: allow kill process which wait statahead result
648 Details : for some reasons 'ls' can stick in waiting result from statahead,
649 in this case need way for kill this process.
651 Severity : enhancement
653 Description: Add state history info file, enhance import info file
654 Details : Track import connection state changes in a new osc/mdc proc file;
655 add overview-type data to the osc/mdc import proc file.
657 Severity : enhancement
659 Description: MDS create should not wait for statfs RPC while holding DLM lock.
661 Severity : enhancement
663 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
668 Descriptoin: OST grant too much space to client even there are not enough space.
669 Details : Client will shrink its grant cache to OST if there are no write
670 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
671 this grant cache if there are already not enough avaible space
672 (left_space < total_clients * 32M).
675 Frequency : start MDS on uncleanly shutdowned MDS device
677 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
678 Details : stay in waiting mds<>ost recovery finished produce random bugs
679 due race between two ll_sync thread for one lov target. send
680 ACTIVATE event only if connect realy finished and import have
684 Frequency : rare, connect and disconnect target at same time
686 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
687 Details : don't call obd_disconnect under lov_lock. this long time
688 operation and can block ptlrpcd which answer to connect request.
693 Descriptoin: don't lose wakeup for imp_recovery_waitq
694 Details : recover_import_no_retry or invalidate_import and import_close can
695 both sleep on imp_recovery_waitq, but we was send only one wakeup
699 Frequency : always with long access acl
701 Descriptoin: mds can't pack reply with long acl.
702 Details : mds don't control size of acl but they limited by reint/getattr
705 Severity : enhancement
707 Description: Update to SLES10 kernel-2.6.16.60-0.33.
709 Severity : enhancement
711 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
714 Frequency : start MDS on uncleanly shutdowned MDS device
716 Descriptoin: aborting recovery hang on MDS
717 Details : don't throttle destroy RPCs for the MDT.
720 Frequency : on remount
722 Description: external journal device not working after the remount
723 Details : clear dev_rdonly flag for external journal devices in
729 Description: shutdown vs evict race
730 Details : client_disconnect_export vs connect request race.
731 if client will evicted at this time - we start invalidate
732 thread without referece to import and import can be freed
736 Frequency : rare, need acl's on inode.
738 Description: client can't handle ost additional correctly
739 Details : if ost was added after client connected to mds client can have
740 hit lnet_try_match_md ... to big messages to wide striped files.
741 in this case need teach client to handle config events about add
742 lov target and update client max ea size at that event.
744 Severity : enhancement
746 Description: Changelogs
747 Details : Changelogs are a lightweight mechanism to track filesystem
748 metadata and namespace changes. The changelog is recorded
749 permanently on the MDTs, and is periodically "consumed" / purged
750 when records are no longer needed.
752 Severity : enhancement
754 Description: compact fld format with extents
755 Details : Store range of seq rather than every seq in FLD. Seq
756 controller update FLD rather than clients. In Case of CMD, mdt0
757 has FLD, all other metadata server act as non persistent proxy
758 for FLD queries and cache fld entries in fld cache.
763 Description: don't skip ost target if they assigned to file
764 Details : Drop slow OSCs if we can, but not for requested start idx.
765 This means "if OSC is slow and it is not the requested
766 start OST, then it can be skipped, otherwise skip it only
767 if it is inactive/recovering/out-of-space.
771 Description: more cleanup in mds_lov
772 Details : not send LOV EA under replay, we can't know about they size at this
773 time. Don't allow client connect to mds before any ost connected,
774 for avoid problems with LOV EA size and returning EIO to client.
776 Severity : enhancement
778 Description: Interoperability at server side (Disk interoperability)
780 Severity : enhancement
782 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
784 Severity : enhancement
786 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
788 Severity : enhancement
790 Description: New client IO stack (CLIO).
792 Severity : enhancement
794 Description: Commit on sharing. Eliminate inter-client dependencies between
795 uncommitted transactions by doing transaction commits.
796 Thereby clients may recovery independently.
799 Frequency : Create a symlink file with a very long name
801 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
802 Details : If there is no extra space in the request for early cancels,
803 ldlm_req_handles_avail() returns 0 instead of a negative value.
805 Severity : enhancement
807 Description: Add /proc entry for import status
808 Details : The mdc, osc, and mgc import directories now have
809 an import directory that contains useful import data for debugging
812 Severity : enhancement
814 Description: Re-disable certain /proc logging
815 Details : Enable and disable client's offset_stats, extents_stats and
816 extents_stats_per_process stats logging on the fly.
819 Frequency : Only on FC kernels 2.6.22+
821 Description: oops in statahead
822 Details : Do not drop reference count for the dentry from VFS when lookup,
823 VFS will do that by itself.
825 Severity : enhancement
827 Description: Generic /proc file permissions
828 Details : Set /Proc file permissions in a more generic way to enable non-
829 root users operate on some /proc files.
833 Description: Hitting mdc_commit_close() ASSERTION
834 Details : Properly handle request reference release in
835 ll_release_openhandle().
839 Description: quota recovery deadlock during mds failover
840 Details : This patch includes att18982, att18236, att18237 in bz14840.
842 1. fix osts hang when mds does failover with quotaon
843 2. prevent watchdog storm when osts threads wait for the
848 Frequency : only patchless client
849 Description: add workaround for race between add/remove dentry from hash
851 Severity : enhancement
853 Description: Allow OST glimpses to return PW locks
857 Description: LBUG when llog conf file is full
858 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
862 Description: Prevent import from entering FULL state when server in recovery
866 Description: service mount cannot take device name with ":"
867 Details : Only when device name contains ":/" will mount treat it as
873 Description: replace ptlrpcd with the statahead thread to interpret the async
874 statahead RPC callback
878 Frequency : on recovery
879 Description: I/O failures after umount during fail back
880 Details : if client reconnected to restarted server we need join to recovery
881 instead of find server handler is changed and process self
882 eviction with cancel all locks.
884 Severity : enhancement
886 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
888 Severity : enhancement
890 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
892 Severity : enhancement
894 Description: Upcall on Lustre log has been dumped
895 Details : Allow for a user mode script to be called once a Lustre log has
896 been dumped. It passes the filename of the dumped log to the
897 script, the location of the script can be specified via
898 /proc/sys/lnet/debug_log_upcall.
903 Description: avoid idr_remove called for id which is not allocated.
904 Details : Move assigment s_dev for clustered nfs to end of initialization,
905 to avoid problem with error handling.
910 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
911 Details : When connection is reused this not moved from CONN_UNUSED_HASH
912 into CONN_USED_HASH and this prodice warning when put connection
913 again in unused hash.
916 Severity : enhancement
918 Description: File striping can now be set to use an arbitrary pool of OSTs.
920 Severity : enhancement
922 Description: Export bytes_read/bytes_write count on OSC/OST.
926 Description: Early reply size mismatch, MGC loses connection
927 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
928 the connect flags are properly negotiated.
933 Description: quotas are not honored with O_DIRECT
934 Details : all writes with the flag O_DIRECT will use grants which leads to
935 this problem. Now using OBD_BRW_SYNC to guard this.
939 Description: add quota statistics
940 Details : 1. sort out quota proc entries and proc code.
941 2. add quota statistics
943 Severity : enhancement
945 Description: enable quota support for HEAD.
949 Description: Properly propagate oinfo flags from lov to osc for statfs
950 Details : restore missing copy oi_flags to lov requests.
952 Severity : enhancement
954 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
955 : l_getgroups(8), lst(8), routerstat(8)
956 Details : included man pages for llobdstat(8), llstat(8),
957 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
959 Severity : enhancement
961 Description: configure's --enable-quota should check the
962 : kernel .config for CONFIG_QUOTA
963 Details : configure is terminated if --enable-quota is passed but
964 : no quota support is in kernel
968 Description: Remove portals compatibility
969 Details : Remove portals compatibility, not interoperable with releases
974 Description: Resolve device initialization race
975 Details : Prevent proc handler from accessing devices added to the
976 obd_devs array but yet be intialized.
978 Severity : enhancement
980 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
982 Severity : enhancement
984 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
989 Description: Using wrong pointer in osc_brw_prep_request
990 Details : Access to array[-1] can produce panic if kernel compiled with
991 CONFIG_PAGE_ALLOC enabled
995 Description: Client runs out of low memory
996 Details : Consider only lowmem when counting initial number of llap pages
1000 Description: *optional* service tags registration
1001 Details : if the "service tags" package is installed on a Lustre node
1002 When the filesystem is mounted, a local-node service tag will
1003 be created. See http://inventory.sun.com/ for more information
1004 about the Service Tags asset management system.
1008 Description: Kernel BUG tries to release flock
1009 Details : Lustre does not destroy flock lock before last reference goes
1010 away. So always drop flock locks when client is evicted and
1011 perform unlock regardless of successfulness of speaking to MDS.
1015 Description: add refcount protection for osc callbacks, avoid panic on shutdown
1019 Description: sanity test 65a fails if stripecount of -1 is set
1020 Details : handle -1 striping on filesystem in ll_dirstripe_verify
1025 Description: ASSERTION(CheckWriteback(page,cmd)) failed
1026 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
1029 Severity : enhancement
1031 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
1035 Description: do not process already freed flock
1036 Details : flock can possibly be freed by another thread before it reaches
1037 to ldlm_flock_completion_ast.
1041 Description: LBUG during stress test
1042 Details : Need properly lock accesses the flock deadlock detection list.
1046 Description: oops in page fault handler
1047 Details : kernel page fault handler can return two special 'pages' in error
1048 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
1052 Description: timeout with invalidate import.
1053 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
1054 should be handled by ptlrpcd. This produce long age waiting and
1055 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
1057 Severity : enhancement
1059 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
1063 Description: enable MGS and MDT services start separately
1064 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
1065 a co-located MGS without starting the MGS, which is a complement
1066 to 'nosvc' mount option.
1070 Frequency : after recovery
1071 Description: precreate to many object's after del orphan.
1072 Details : del orphan st in oscc last_id == next_id and this triger growing
1073 count of precreated objects. Set flag LOW to skip increase count
1074 of precreated objects.
1078 Frequency : rare, on clear nid stats
1079 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
1080 Details : when clean nid stats sometimes try destroy live entry,
1081 and this produce panic in free.
1085 Description: Stack overflow during MDS log replay
1086 ease stack pressure by using a thread dealing llog_process.
1090 Description: wait until IO finished before start new when do lock cancel.
1091 Details : VM protocol want old IO finished before start new, in this case
1092 need wait until PG_writeback is cleared until check dirty flag and
1093 call writepages in lock cancel callback.
1095 Severity : enhancement
1097 Description: using special macro for print time and cleanup in includes.
1101 Description: mds_mfd_close() ASSERTION(rc == 0)
1102 Details : In mds_mfd_close(), we need protect inode's writecount change
1103 within its orphan write semaphore to prevent possible races.
1107 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
1111 Frequency : rare, on shutdown ost
1112 Description: don't hit live lock with umount ost.
1113 Details : shrink_dcache_parent can be in long loop with destroy dentries,
1114 use shrink_dcache_sb instead.
1118 Description: don't panic with use echo client
1119 Details : echo client pass NULL as client nid pointer and this produce null
1120 pointer dereference.
1124 Description: fix build on ppc32
1125 Details : compile code with -m64 flag produce wrong object file for ppc32.
1129 Description: add message levels for liblustreapi
1133 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
1134 Details : Move the 'good_osts' check before the 'total_bavail' check. This
1135 will result in an -EAGAIN and in the exit call path we call
1136 alloc_rr() which will with increasing aggressiveness attempt to
1137 aquire precreated objects on the minimum number of required OSCs.
1141 Description: Use old size assignment to avoid deadlock
1142 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
1143 the scheduling while holding a spinlock. We do not need locking
1144 for size in ll_update_inode() because size is only updated from
1145 the MDS for directories or files without objects, so there is no
1146 other place to do the update, and concurrent access to such inodes
1147 are protected by the inode lock.
1151 Description: resolve "_IOWR redefined" build error on SLES10
1155 Description: dump the memory debugging after all modules are unloaded to
1156 suppress false negative in conf_sanity test 39
1158 Severity : enhancement
1160 Description: build kernel-ib packages for OFED 1.3 in our release cycle
1165 Description: fix SLES kernel versioning
1166 Details : the kernel version for our SLES 10 kernel did not include a "-"
1167 before the "smp" at the end. while this was not a problem in
1168 general, it did mean that software trying to use the kernel
1169 version to try to detect a vendor specific kernel would fail.
1170 this was most evident by the OFED build scripts.
1174 Description: Don't update lov_desc members until making sure they are valid
1175 Details : When updating lov_desc members via proc fs, need fix their
1176 validities before doing the real update.
1180 Description: don't put request into delay list while invalidate in flight.
1181 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
1182 import in flight. this produce timeout for invalidate and sometimes
1183 can cause stale data.
1187 Frequency : on ppc only
1188 Description: not convert ost objects for directory because it's not exist.
1189 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
1193 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
1195 Severity : enhancement
1197 Description: Update to support 2.6.22.14 vanilla kernel.
1201 Frequency : rare, on recovery
1202 Description: read procfs can produce deadlock in some situation
1203 Details : Holding lprocfs lock which send rpc can produce block for destroy
1204 obd objects and this also block reconnect with -EALREADY. This
1205 isn't fix all lprocfs bugs - but make it rare.
1207 Severity : enhancement
1209 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
1212 Frequency : frequent on X2 node
1214 Description: mdc_set_open_replay_data LBUG
1215 Details : Set replay data for requests that are eligible for replay.
1219 Description: lustre_mgs: operation 101 on unconnected MGS
1220 Details : When MGC is disconnected from MGS long enough, MGS will evict the
1221 MGC, and late on MGC cannot successfully connect to MGS and a lot
1222 of the error messages complaining that MGS is not connected.
1226 Frequency : on network error
1227 Description: panic with double free request if network error
1228 Details : mdc_finish_enqueue is finish request if any network error ocuring,
1229 but it's true only for synchronus enqueue, for async enqueue
1230 (via ptlrpcd) this incorrect and ptlrpcd want finish request
1233 Severity : enhancement
1235 Description: client-side metadata stat-ahead during readdir(directory readahead)
1236 Details : perform client-side metadata stat-ahead when the client detects
1237 readdir and sequential stat of dir entries therein
1240 Frequency : on start mds
1242 Description: Implement get_info(last_id) in obdfilter.
1245 Frequency : occasional
1247 Description: Correctly check stale fid, not start epoch if ost not support SOM
1248 Details : open with flag O_CREATE need set old fid in op_fid3 because
1249 op_fid2 was overwritten with new generated fid, but mds can answer
1250 with one of these two fids and both is not stale. Setattr
1251 incorrectly started an epoch and assume will be called
1252 done_writing, but without SOM done_writing ever being called.
1255 Frequency : rare, depends on device drivers and load
1257 Description: MDS or OSS nodes crash due to stack overflow
1258 Details : Code changes in 1.8.0 increased the stack usage of some functions.
1259 In some cases, in conjunction with device drivers that use a lot
1260 of stack the MDS (or possibly OSS) service threads could overflow
1261 the stack. One change which was identified to consume additional
1262 stack has been reworked to avoid the extra stack usage.
1265 Frequency : occasional
1267 Description: Do not fail import if osc_interpret_create gets -EAGAIN
1268 Details : If osc_interpret_create got -EAGAIN it immediately exits and
1269 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
1270 oscc_has_objects and see OSC has no objests and call
1271 oscc_internal_create to resend create request.
1273 Severity : enhancement
1275 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
1277 Severity : enhancement
1279 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
1282 Frequency : very rare
1284 Description: Fix replay if there is an un-replied request and open
1285 Details : In some cases, older replay request will revert the
1286 mcd->mcd_last_xid on MDS which is used to record the client's
1287 latest sent request.
1289 Severity : enhancement
1291 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
1293 Severity : enhancement
1295 Description: Add rhel5 support to HEAD.
1297 Serverity : enhancement
1299 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
1304 Description: Don't allow skipping OSTs if index has been specified.
1305 Details : Don't allow skipping OSTs if index has been specified, make
1306 locking in internal create lots better.
1310 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
1311 Details : access to bitfield in structure is always rounded to long
1312 and this produce problem with not atomic change any bit.
1316 Description: Lustre make rpms failed.
1317 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
1318 builds Lustre rpms from tarball.
1321 Frequency : rare on shutdown ost
1323 Description: If llog cancel was not send before clean_exports phase, this can
1324 produce deadlock in llog code.
1325 Details : If llog thread has last reference to obd and call class_import_put
1326 this produce deadlock because llog_cleanup_commit_master wait when
1327 last llog_commit_thread exited, but this never success because was
1328 called from llog_commit_thread.
1332 Description: allow userland application know is lost one of stripes.
1333 Details : fill lvb_blocks with error code on ost and return it to
1334 application if error flag found.
1338 Description: NULL lov_tgts causing MDS oops
1339 Details : more safe checks for NULL lov_tgts for avoid oops.
1341 Severity : enhancement
1343 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
1347 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
1348 Details : If an OST is down the MDS will hang indefinitely in
1349 obd_statfs_rqset() waiting for the statfs data. While for
1350 MDS QOS usage of statfs, it should not stuck in waiting.
1352 Severity : enhancement
1354 Description: remote_acl support
1355 Details : Support ACL-based permission check for remote user.
1356 Support setfacl/getfacl for remote user with the utils
1357 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
1358 the system "{s,g}etfacl" utils.
1360 Severity : enhancement
1362 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
1364 Severity : enhancement
1366 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
1370 Description: make mgs_setparam() handle fsname containing dash
1371 Details : fsname containing a dash does not work with lctl conf_param
1373 Severity : enhancement
1375 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
1377 Severity : enhancement
1379 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
1383 Frequency : rare, at shutdown
1384 Description: access already free / zero obd_namespace.
1385 Details : if client_disconnect_export was called without force flag set,
1386 and exist connect request in flight, this can produce access to
1387 NULL pointer (or already free pointer) when connect_interpret
1388 store ocd flags in obd_namespace.
1392 Frequency : only at startup
1393 Description: not alloc memory with spinlock held.
1394 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
1395 if any spinlock held.
1397 Severity : enhancement
1399 Description: make lustre randomly fail allocating memory
1400 Details : Make lustre randomly failed allocating memory for testing purpose.
1402 Severity : enhancement
1404 Description: lost problems with lov objid file
1405 Details : Fixes some scability and access to not inited memory problems
1406 in work with lov objdid file.
1411 Description: lfs find does not continue on file error
1412 Details : Continue other files processing when a file/dir is absent.
1416 Description: Inconsistent usage of lustre_pack_reply()
1417 Details : Standardize the usage of lustre_pack_reply() such that it
1418 always generate a CERROR on failure.
1421 Frequency : occasional
1423 Description: MDS hang or stay in waiting lock
1424 Details : If client receive lock with CBPENDING flag ldlm need send lock
1425 cancel as separate rpc, to avoid situation when cancel request
1426 can't processed due all i/o threads stay in wait lock.
1430 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1434 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1438 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1442 Description: Update to RHEL4 latest kernel.
1444 Severity : enhancement
1446 Description: Build SLES10 patchless client fails
1447 Details : The configure was broken by run ./configure with
1448 --with-linux-obj=.... argument for patchless client. When the
1449 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1450 can't search header adequately. Use absolute path such as
1451 -I($LINUX)/include instead.
1455 Description: interrupt oig_wait produce painc on resend.
1456 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1457 private set, and this produce situation when rq_ptlrpcd_data not
1458 copyed to new allocated request and triggered LBUG on assert
1459 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1460 ptlrpcd set if request is changed and can be safety changed to use
1465 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1466 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1467 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1468 these macros and replace them with their difinitions instead.
1472 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1473 Details : Update which_patch & target file for SLES10 latest kernel.
1476 Description: Remove client patches from SLES 10 kernel.
1477 Details : This causes SLES 10 clients to behave as patchless clients
1478 even on a Lustre-patched (server) kernel.
1480 Severity : enhancement
1482 Description: self-adjustable client's lru lists
1483 Details : use adaptive algorithm for managing client cached locks lru
1484 lists according to current server load, other client's work
1485 pattern, memory activities, etc. Both, server and client
1486 side namespaces provide number of proc tunables for controlling
1489 Severity : enhancement
1491 Description: light-weight GSS support
1492 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
1493 and reduce performance overhead.
1495 Severity : enhancement
1497 Description: Linux keyring support
1498 Details : Support using service of Linux keyring for Lustre GSS internal
1499 context refresh/cache mechanism.
1503 Description: Fix errors in lfs documentation
1504 Details : Fixes man pages
1508 Description: don't use GFP_* in generic Lustre code.
1509 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
1513 Description: obdclass is limited by single OBD_ALLOC(idarray)
1514 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
1518 Description: warnings with build patchless client with vanila 2.6.19 and up
1519 Details : change old ctl_table style and replace ctl_table/ctl_table_header
1520 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
1524 Frequency : Only for SLES
1525 Description: sanity_quota fail test_1
1526 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
1527 file, which makes TSTID[2] inunique.
1531 Frequency : Only for SLES9
1532 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
1533 Details : Update target/ChangeLog/which_patch .
1537 Frequency : Only for test_10 in sanity-quota.sh
1538 Description: a bug in quota test code
1539 Details : checking a test flag in a wrong place
1543 Frequency : Only for test_13 in sanity-quota.sh
1544 Description: a bug in quota test script
1545 Details : shouldn't check overfull allocated quota
1549 Description: jbd statistics
1550 Details : Port older jbd statistics patch for sles10
1552 Severity : enhancement
1554 Description: Add printing inode info into message about error in writepage.
1559 Description: reply_lock_interpret crash due to race with it and lock cancel.
1560 Details : Do not replay locks that are being cancelled. Do not reference
1561 locks by their address during replay, just by their handle.
1565 Frequency : When flocks are used.
1566 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
1567 Details : Flock locks might destroy just granted lock if it could be merged
1568 with another existing flock, this is done in completion handler,
1569 so teach ldlm_cli_enquque_fini that this is a valid case for
1575 Description: Oops in read and write path when failing to allocate lock.
1576 Details : Check if lock allocation failed and return error back.
1580 Description: lstripe command fails for valid OST index
1581 Details : The stripe offset is compared to lov->desc.ld_tgt_count
1582 instead of lov->desc.ld_active_tgt_count.
1586 Description: sanity.sh failed test 103
1587 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
1592 Description: df doesn't work properly if diskfs blocksize != 4K
1593 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
1597 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
1598 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
1600 Severity : enhancement
1602 Description: Adaptive timeouts
1603 Details : RPC timeouts adapt to changing server load and network
1604 conditions to reduce resend attempts and improve recovery time.
1608 Description: llapi_file_create() does not allow some changes
1609 Details : add llapi_file_open() that allows specifying the mode and
1610 open flags, and also returns an open file handle.
1614 Description: merge and cleanup kernel patches.
1615 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
1619 Description: Add journal checksum support.(Kernel part)
1620 Details : The journal checksum feature adds two new flags i.e
1621 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
1622 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
1623 indicates that the commit block contains the checksum for
1624 the blocks described by the descriptor blocks. Now commit
1625 record can be sent to disk without waiting for descriptor
1626 blocks to be written to disk. This behavior is controlled
1627 using JBD2_FEATURE_ASYNC_COMMIT flag.
1631 Description: OSS needs mutliple precreate threads
1632 Details : Add ability to start more than one create thread per OSS.
1636 Description: Sanity.sh test_65k failed: setstripe should have succeeded
1637 Details : execute lfs setstripe on client
1641 Description: mds_obd_create error creating tmp object
1642 Details : When the user sets quota on root, llog will be affected and can't
1643 create files and write files.
1646 Frequency : Always on ia64 patchless client, and possibly others.
1648 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
1649 Details : This allows the patchless client to be loaded on architectures
1650 without this export.
1654 Description: RedHat Update kernel for RHEL5
1655 Details : Add the kernel config file for RHEL5.
1659 Description: RedHat Update kernel for RHEL5
1660 Details : Modify the kernel config file more closer RHEL5.
1664 Description: Build failure against Centos5 (RHEL5)
1665 Details : Define PAGE_SIZE when it isn't present.
1669 Description: test_7 on sanity-quota.sh can't be used on separate machines
1670 Details : add facet to handle it
1674 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
1675 Details : with NFS, the anon dentry's parent was set to itself in
1676 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
1677 corresponding dentry other than use rec->ur_name.
1679 Severity : enhancement
1681 Description: lfs setstripe enhancement
1682 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
1686 Description: enable data checksumming by default
1687 Details : enable checksum by default, allow --disable-checksum
1688 configure option and "-o nochecksum" mount option. Checksums
1689 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
1693 Description: lustre support for RHEL5
1694 Details : Add support for RHEL5.
1698 Description: Client eviction due to failover config
1699 Details : after a connection loss, the lustre client should attempt to
1700 reconnect to the last active server first before trying the
1701 other potential connections.
1705 Description: when mds and osts use different quota unit(32bit and 64bit),
1706 quota will be released repeatly.
1707 Details : void sending multiple quota reqs to mds, which will keep the
1708 status between the reqs.
1712 Description: osts not allocated evenly to files
1713 Details : change the condition to increase offset_idx
1716 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
1718 Description: Data corruption for OSTs that are formatted larger than 2TB
1720 Details : When generating the bio request for lustre file writes the
1721 sector number would overflow a temporary variable before being
1722 used for the IO. The data reads correctly from Lustre (which
1723 will overflow in a similar manner) but other file data or
1724 filesystem metadata may be corrupted in some cases.
1728 Description: Tune the kernel for good SCSI performance.
1729 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
1730 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
1735 Description: rewrite ext2-derived code in obdclass/uuid.c
1736 Details : rewrite inherited code (uuid parsing code from ext2 utils)
1737 from scratch preserving functionality.
1741 Description: "lfs find -obd UUID" prints directories
1742 Details : "lfs find -obd UUID" will return all directory names instead
1743 of just file names. It is incorrect because the directories
1744 do not reside on the OSTs.
1748 Description: use i_size_read and i_size_write in 2.6 port
1749 Details : replace inode->i_size access with i_size_read/write()
1752 Frequency : when removing large files
1754 Description: scheduling issue during removal of large Lustre files
1755 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
1756 It causes scheduling issues when removing large files (17TB in the
1760 Frequency : only with liblustre clients on XT3
1762 Description: evictions taking too long
1763 Details : allow llrd to evict clients directly on OSTs
1766 Frequency : only on ppc
1768 Description: /proc/fs/lustre/devices broken on ppc
1769 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
1770 We should make sure the type of 'index' is loff_t and then cast
1771 explicitly as needed below. Do not assign an explicitly cast
1775 Frequency : only for rhel5
1777 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
1778 Details : Modify the target file & which_kernel.
1780 Severity : enhancement
1782 Description: omit set fsid for export NFS
1783 Details : fix set/restore device id for avoid EMFILE error and mark lustre
1784 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
1788 Frequency : Always, for kernels after 2.6.16
1789 Description: Fix warning idr_remove called for id=.. which is not allocated.
1790 Details : Last kernels save old s_dev before kill super and not allow
1791 to restore from callback - restore it before call kill_anon_super.
1795 Description: buffer overruns could theoretically occur
1796 Details : llapi_semantic_traverse() modifies the "path" argument by
1797 appending values to the end of the origin string, and a
1798 overrun may occur. Adding buffer overrun check in liblustreapi.
1802 Description: Fix error on 'ls .' at the top of the Lustre mount.
1803 Details : Don't revalidate dentry if it is a root dentry.
1807 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
1808 Details : Modify vm-tunables-rhel4.patch.
1812 Description: Kernel config for 2.6.18-vanilla.
1813 Details : Modify targets/2.6-vanilla.target.in.
1814 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
1815 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
1816 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
1817 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
1821 Description: improve handling recoverable errors
1822 Details : if request processig with error which can be recoverable on server
1823 request should be resend, otherwise page released from cache and
1828 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
1829 Details : Modify target file & which_patch.
1830 A flaw was found in the IA32 system call emulation provided
1831 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
1832 value could be stored in the %RAX register, which could trigger an
1833 out-of-bounds system call table access. An untrusted local user
1834 could exploit this flaw to run code in the kernel
1835 (ie a root privilege escalation). (CVE-2007-4573).
1839 Description: O_DIRECT bypasses client statistics.
1840 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
1841 read_ahead_stats were updated. copied the stats section from
1842 osc_send_oap_rpc() into async_internal().
1846 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
1850 Description: change order of libsysio includes
1851 Details : '#include sysio.h' should always come before '#include xtio.h'
1855 Description: handle "serious error: objid * already exists" more gracefully
1856 Details : If LAST_ID value on disk is smaller than the objects existing in
1857 the O/0/d* directories, it indicates disk corruption and causes an
1858 LBUG(). If the object is 0-length, then we should use the existing
1859 object. This will help to avoid a full fsck in most cases.
1861 Severity : enhancement
1863 Description: adapt the lustre_config script to support the upgrade case
1864 Details : Add "-u" option for lustre_config script to support upgrading 1.4
1865 server targets to 1.6 in parallel.
1869 Description: To avoid grant space > avaible space when the disk is almost
1870 full. Without this patch you might see the error "grant XXXX >
1871 available" or some LBUG about grant, when the disk is almost
1873 Details : In filter_check_grant, for non_grant cache write, we should
1874 check the left space by if (*left > ungranted + bytes), instead
1875 of (*left > ungranted), because only we are sure the left space
1876 is enough for another "bytes", then the ungrant space should be
1877 increase. In client, we should update cl_avail_grant only there
1878 is OBD_MD_FLGRANT in the reply.
1882 Description: Update RHEL 4 kernel to fix local root privilege escalation.
1883 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
1884 described in CVE-2007-4573. This problem could allow untrusted
1885 local users to gain root access.
1888 Frequency : when using O_DIRECT and quotas
1890 Description: Incorrect file ownership on O_DIRECT output files
1891 Details : block usage reported by 'lfs quota' does not take into account
1892 files that have been written with O_DIRECT.
1897 Description: touch file failed when fs is not full
1898 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
1899 otherwise we can get ENOSP while fs is not full.
1903 Description: parallel lock callbacks
1904 Details : Instead of sending blocking and completion callbacks as separated
1905 requests, adding them to a set and sending in parallel.
1908 Frequency : only for Cray XT3
1909 Bugzilla : 12829/13455
1910 Description: Changing primary group doesn't change the group lustre assigns to
1912 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
1913 overridden with the primary group provided by the group upcall,
1914 whereas the supplied fsgid can be trusted if it is in the list of
1915 supplementary groups returned by the group upcall.
1917 Severity : enhancement
1919 Description: Allow masking D_WARNING, D_ERROR messages from console
1920 Details : Console messages can now be disabled via lnet.printk.
1924 Description: User code with malformed file open parameter crashes client node
1925 Details : Before packing join_file req, all the related reference should be
1926 checked carefully in case some malformed flags cause fake
1927 join_file req on client.
1931 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
1932 Details : ldlm_completion_ast() assumes that a lock is granted when the req
1933 mode is equal to the granted mode. However, it should also check
1934 that LDLM_FL_CP_REQD is not set.
1938 Description: Heavy nfs access might result in deadlocks
1939 Details : After ELC code landed, it is now improper to enqueue any mds
1940 locks under och_sem, because enqueue might want to decide to
1941 cancel open locks for same inode we are holding och_sem for.
1945 Description: Client eviction while running blogbench
1946 Details : A lot of unlink operations with concurrent I/O can lead to a
1947 deadlock causing evictions. To address the problem, the number of
1948 oustanding OST_DESTROY requests is now throttled to
1949 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
1950 callbacks are processed in priority.
1954 Description: enable ACLs on MDS by default
1955 Details : ACLs must be enabled on MDS by default.
1958 Frequency : PPC/PPC64 only
1960 Description: conflicts between asm-ppc64/types.h and lustre_types.h
1961 Details : fix duplicated definitions between asm-ppc64/types.h and
1962 lustre_types.h on PPC.
1965 Frequency : PPC/PPC64 only
1967 Description: asm-ppc/segment.h does not exist
1968 Details : fix compile issue on PPC.
1972 Description: better handle error messages in extents code
1975 Frequency : RHEL4 only
1977 Description: mkfs is very slow on IA64/RHEL4
1978 Details : A performance regression has been discovered in the MPT Fusion
1979 driver between versions 3.02.73rh and 3.02.99.00rh. As a
1980 consequence, we have downgraded the MPT Fusion driver in the RHEL4
1981 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
1983 Severity : enhancement
1985 Description: SNMP support enhancement
1986 Details : Adding total number of sampled request for an MDS node in snmp
1989 Severity : enhancement
1991 Description: Optimize ldlm waiting list processing for PR extent locks
1992 Details : When processing waiting list for read extent lock and meeting read
1993 lock that is same or wider to it that is not contended, skip
1994 processing rest of the list and immediatelly return current
1995 status of conflictness, since we are guaranteed there are no
1996 conflicting locks in the rest of the list.
2000 Description: Time out and refuse to reconnect
2001 Details : When the failover node is the primary node, it is possible
2002 to have two identical connections in imp_conn_list. We must
2003 compare not conn's pointers but NIDs, otherwise we can defeat
2004 connection throttling.
2008 Description: port llog fixes from b1_6 into HEAD
2009 Details : Port llog reference couting and some llog cleanups from b1_6
2010 (bug 10800) into HEAD, for protect from panic and access to already
2011 free llog structures.
2015 Description: Detect stride IO mode in read-ahead
2016 Details : When a client does stride read, read-ahead should detect that and
2017 read-ahead pages according to the detected stride pattern.
2021 Description: data checksumming impacts single node performance
2022 Details : add support for several checksum algorithm. Currently, only CRC32
2023 and Adler-32 are supported. The checksum type can be changed on
2024 the fly via /proc/fs/lustre/osc/*/checksum_type.
2028 Description: use adler32 for page checksums
2029 Details : when available, use the Adler-32 algorithm instead of CRC32 for
2034 Description: build for x2 fails
2035 Details : fix compile issue on Cray systems.
2039 Description: Properly match for duplicate locks
2040 Details : Due to different lock order from skiplists code, we need to
2041 traverse entire list for now
2044 Frequency : only on PPC/SLES10
2046 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
2047 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
2048 be defined. Add a hack in mkfs_lustre.c to work around this compile
2053 Description: LASSERT on MDS when client holding flock lock dies
2054 Details : ldlm pool logic depends on number of granted locks equal to
2055 number of released locks which is not true for flock locks, so
2056 just exclude such locks from consideration.
2060 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
2061 Details : Use fsfilt_sync() for both the whole filesystem sync and
2062 individual file sync to eliminate dangerous inode locking
2063 with I_LOCK that can lead to a deadlock.
2067 Description: performance in 1.6.3
2068 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
2069 sure that 1MB requests are not fragmented by the block layer.
2071 Severity : enhancement
2073 Description: organize the server-side client stats on per-nid basis
2074 Details : Change the structure of stats under obdfilter and mds to
2082 The "uuid"s file would list the uuids of _active_ exports.
2083 And the clear entry is to clear all stats and stale nids.
2085 Severity : enhancement
2087 Description: eliminate client locks in face of contention
2088 Details : file contention detection and lockless i/o implementation
2089 for contended files.
2093 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
2096 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
2098 Description: MDS may lose file striping (and hence file data) in some cases
2099 Details : If there are additional extended attributes stored on the MDS,
2100 in particular ACLs, SELinux, or user attributes (if user_xattr
2101 is specified for the client mount options) then there is a risk
2102 of attribute loss. Additionally, the Lustre file striping
2103 needs to be larger than default (e.g. striped over all OSTs),
2104 and an additional attribute must be stored initially in the
2105 inode and then increase in size enough to be moved to the
2106 external attribute block (e.g. ACL growing in size) for file
2111 Description: skiplist implementation simplification
2112 Details : skiplists are used to group compatible locks on granted list
2113 that was implemented as tracking first and last lock of each
2114 lock group the patch changes that to using doubly linked lists
2118 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
2119 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
2120 because LASSERT on both the data supplied by a client, and the
2121 data on disk is dangerous and incorrect.
2123 Severity : enhancement
2125 Description: Slow truncate/writes to huge files at high offsets.
2126 Details : Directly associate cached pages to lock that protect those pages,
2127 this allows us to quickly find what pages to write and remove
2128 once lock callback is received.
2132 Description: more ldlm soft lockups
2133 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
2134 starve other threads from the resource lock for a long time in
2135 case of long waiting queue, so change the debug level from
2136 D_OTHER to the less frequently used D_INFO.
2138 Severity : enhancement
2140 Description: add -gid, -group, -uid, -user options to lfs find
2144 Description: Hung threads in invalidate_inode_pages2_range
2145 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
2146 once one is completed. As a result, some RPCs are stuck in the
2147 queue and are never sent.
2151 Description: filter threads hungs on waiting journal commit
2152 Details : Cleanup filter group llog code, then only filter group llog will
2153 be only created in the MDS/OST syncing process.
2157 Description: Procfs and llog threads access destoryed import sometimes.
2158 Details : Sync the import destoryed process with procfs and llog threads by
2159 the import refcount and semaphore.
2161 Severity : enhancement
2163 Description: openlock cache of b1_6 port to HEAD
2168 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
2169 Details : If insertion of an extent fails, then discard the inode
2170 preallocation and free data blocks else it can lead to duplicate
2175 Description: don't always update ctime in ext3_xattr_set_handle()
2176 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
2177 In some cases the ctime should not be updated, for example for
2178 2.0->1.8 compatibility it is necessary to delete an xattr and it
2179 should not update the ctime.
2183 Bugzilla : 15713/16362
2184 Description: Assertion in iopen_connect_dentry in 1.6.3
2185 Details : looking up an inode via iopen with the wrong generation number can
2186 populate the dcache with a disconneced dentry while the inode
2187 number is in the process of being reallocated. This causes an
2188 assertion failure in iopen since the inode's dentry list contains
2189 both a connected and disconnected dentry.
2193 Description: assertion failure in ldlm_handle2lock()
2194 Details : fix a race between class_handle_unhash() and class_handle2object()
2195 introduced in lustre 1.6.5 by bug 13622.
2200 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
2201 Details : remove the limit of 256 scsi disks in the sd_iostat patch
2206 Description: oops in sd_iostats_seq_show()
2207 Details : unloading/reloading the scsi low level driver triggers a kernel
2208 bug when trying to access the sd iostat file.
2213 Description: Kernel panics during QLogic driver reload
2214 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
2215 patch, causing memory corruption.
2220 Description: journal_dev option does not work in b1_6
2221 Details : pass mount option during pre-mount.
2223 Severity : enhancement
2225 Description: Add a FIEMAP(FIle Extent MAP) ioctl
2226 Details : FIEMAP ioctl will allow an application to efficiently fetch the
2227 extent information of a file. It can be used to map logical blocks
2228 in a file to physical blocks in the block device.
2232 Description: LDLM soft lockups - improvement
2233 Details : It is be possible to send the lock handle along with each read
2234 or write request because the client is already doing a lock match
2235 itself so there isn't any reason the OST should have to re-do that
2239 Frequency : only on Cray X2
2241 Description: X2 build failures
2242 Details : fix build failures on Cray X2.
2246 Description: xid & resent requests
2247 Details : Initialize RPC XID from clock at startup (randomly if clock is
2250 Severity : enhancement
2252 Description: Add lustre_start utility to start or stop multiple Lustre servers
2257 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
2258 Details : In case of memory pressure, list_del() can be called twice on
2259 req->rq_history_list, causing a kernel oops.
2263 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
2264 Details : fix stack overflow in the distributed lock manager by defering
2265 export eviction after a failed AST to the elt thread instead of
2266 handling it in the dlm interpret routine.
2270 Description: Convert some comments to new format.
2271 Details : Update documenting comments to match doxygen conventions.
2275 Description: Grammar fixes.
2276 Details : A couple of trivial spelling fixes.
2280 Description: OSD_COUNTERS-mandatory
2281 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
2282 to simplify the code.
2286 Description: simplify cmm_device freeing logic.
2287 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
2291 Description: Add lockdep support to dt_object_operations locking interface.
2292 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
2293 parameter indicating lock ordering. Update mdd code to use new
2298 Description: Introduce failloc constants for lockless IO tests.
2299 Details : Add two new failloc constants to test lockless IO. Only one of
2300 them in implemented---another is checked in yet to be landed
2305 Description: Add lockdep support for inode mutex.
2306 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
2307 in the situations where Lustre has to lock more than one inode
2312 Description: Add optional invariants checking support.
2313 Details : Add new LINVRNT() macro, optional on new --enable-invariants
2314 configure switch. This macro is to be used for consistency and
2315 sanity checks that are too expensive to be left in `production'
2320 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
2321 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
2326 Description: Add lockdep support for ldlm_lock and ldlm_resource.
2327 Details : Use spin_lock_nested() in (the only) situation where more than
2328 one ldlm_lock is locked simultaneously. Also, fix possible
2329 dead-lock in ldlm_lock_change_resource() by enforcing particular
2334 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
2335 Details : Instead of specifying each ldlm_lock call-back through separate
2336 parameter, wrap them into struct ldlm_callback_suite.
2340 Description: Kill join_lru obd method and its callers.
2341 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
2342 in memory---a requirement implemented through ->o_join_lru() obd
2343 method in HEAD. Remove this method and its users.
2347 Description: Add asynchronous ldlm ENQUEUE completion handler.
2348 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
2349 case that stock ldlm_completion_ast() cannot handle as it waits
2350 until lock is granted. Introduce new ldlm_completion_ast_async()
2351 for this. Also comment ldlm_completion_ast().
2355 Description: ldlm_error <-> errno conversion.
2356 Details : Add functions to map (rather arbitrary) between LDLM error codes
2357 and standard errno values. CLIO needs this to prevent LDLM specific
2358 constants from escaping ldlm and osc.
2362 Description: Kill unused ldlm_handle2lock_ns() function.
2363 Details : Kill unused ldlm_handle2lock_ns() function.
2367 Description: Add lu_ref support to ldlm_lock
2368 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
2369 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
2370 and ldlm_resource. LDLM interface has to be changed, because code that
2371 releases a reference on a lock, has to "know" what reference this is.
2372 In the most frequent case
2374 lock = ldlm_handle2lock(handle);
2376 LDLM_LOCK_PUT(lock);
2378 no changes are required. When any other reference (received _not_ from
2379 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
2380 instead of LDLM_LOCK_PUT().
2382 Arguably, changes are pervasive, and interface requires some discipline
2383 for proper use. On the other hand, it was very instrumental in finding
2384 a few leaked lock references.
2388 Description: Add ldlm_lock_addref_try().
2389 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
2390 attempts to addref a lock that might be being canceled
2395 Description: Add ldlm_weigh_callback().
2396 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
2397 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
2398 hard-coded `number of pages' logic.
2402 Description: Add lockdep annotations to llog code.
2403 Details : Use appropriately tagged _nested() locking calls in the places
2404 where llog takes more than one ->lgh_lock lock.
2408 Description: Add loi_kms_set().
2409 Details : Wrap kms updates into a helper function.
2413 Description: Constify instances of struct lsm_operations.
2414 Details : Constify instances of struct lsm_operations.
2418 Description: lu_conf support.
2419 Details : On a server, a file system object is uniquely identified
2420 by a fid, which is sufficient to locate and load all object
2421 state (inode). On a client, on the other hand, more data are
2422 necessary instantiate an object. Change lu_object_find() and
2423 friends to take additional `lu_conf' argument describing object.
2424 Typically this includes layout information.
2428 Description: lu_context fixes.
2429 Details : Introduce new lu_context functions that are needed on the client
2430 side, where some system threads (ptlrpcd) are shared by multiple
2431 modules, and so cannot be stopped during module shutdown.
2435 Description: Add start and stop methods to lu_device_type_operations.
2436 Details : Introduce two new methods in lu_device_type_operations, that are
2437 invoked when first instance of a given type is created and last one
2438 is destroyed respectively. This is need by CLIO.
2442 Description: Add lu_ref support to struct lu_device.
2443 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
2444 track leaked references.
2448 Description: Introduce lu_kmem_descr.
2449 Details : lu_kmem_descr and its companion interface allow to create
2450 and destroy a number of kmem caches at once.
2454 Description: Fix lu_object finalization race.
2455 Details : Fix a race between lu_object_find() finding an object and its
2456 concurrent finalization. This race is (most likely) not possible
2457 on the server, but might happen on the client.
2461 Description: Introduce lu_ref interface.
2462 Details : lu_ref is a debugging module allowing to track references to
2463 a given object. It is quite cpu expensive, and has to be
2464 explicitly enabled with --enable-lu_ref. See usage description
2469 Description: Factor lu_site procfs stats into a separate function.
2470 Details : Separate lu_site stats printing code into a separate function
2471 that can be reused on a client.
2475 Description: Constify instances of struct {lu,dt,md}_device_operations.
2476 Details : Constify instances of struct {lu,dt,md}_device_operations.
2480 Description: Introduce struct md_site and move meta-data specific parts of
2481 struct lu_site here.
2482 Details : Move md-specific fields out of struct lu_site into special struct
2483 md_site, so that lu_site can be used on a client.
2487 Description: Kill mdd_lov_destroy().
2488 Details : Remove unused mdd code.
2492 Description: Add st_block checking to multistat.c.
2493 Details : Add st_block checking to multistat.c.
2497 Description: Add lu_ref support to struct obd_device.
2498 Details : Add lu_ref tracking to obd_device.
2502 Description: Kill obd_set_fail_loc().
2503 Details : Remove unused code.
2507 Description: Add special type for ptlrpc_request interpret functions.
2508 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
2509 there. Actual usage will be in CLIO.
2513 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
2514 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
2515 doesn't work with lockdep.
2519 Description: Add rwv.c test program.
2520 Details : New testing program exercising readv(2) and writev(2) (Qian).
2524 Description: Add sendfile.c test program.
2525 Details : New testing program exercising sendfile(2) (Jay).
2529 Description: Ratelimit a message that can be very frequent.
2530 Details : Ratelimit a memory allocation failure message that can
2535 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2536 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2540 Description: Liblustre build fixes.
2541 Details : Liblustre build fixes.
2545 Description: libcfs: add cfs_{need,cond}_resched() interface.
2546 Details : libcfs: add cfs_{need,cond}_resched() definition and
2547 implementations for Linux, NT, and liblustre.
2549 Severity : enhancement
2551 Description: More exported tunables for mballoc
2552 Details : Add support for tunable preallocation window and new tunables for
2553 large/small requests
2557 Description: Detect corruption of block bitmap and checking for preallocations
2558 Details : Checks validity of on-disk block bitmap. Also it does better
2559 checking of number of applied preallocations. When corruption is
2560 found, it turns filesystem readonly to prevent further corruptions.
2564 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
2565 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
2566 window to 0. If it does hit cache miss, and read-pattern is still
2567 stride-io mode, does not reset the stride window, but also does not
2568 increase the stride window length in this case.
2572 Frequency : only for big-endian servers
2573 Description: Check if system is big-endian while mounting fs with extents feature
2574 Details : Mounting a filesystem with extents feature will fail on big-endian
2575 systems since ext3-based ldiskfs is not supported on big-endian
2576 systems. This can be over-riden with "bigendian_extents" mount option.
2578 Severity : enhancement
2580 Description: The root squash functionality
2581 Details : A security feature, which is to prevent users from being able
2582 to mount lustre on their desktop, run as root, and delete
2583 all of the files in the filesystem. The goal is accomplished by
2584 remapping user id (UID) and group id (GID) of the root user to
2585 a UID and GID specified by the system administartor via Lustre
2586 configuration management server (MGS). The functionality also
2587 allows to specify sets of clients for which the remapping does
2592 Description: Excessive recovery window
2593 Details : With AT enabled, the recovery window can be excessively long (6000+
2594 seconds). To address this problem, we no longer use
2595 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
2596 timeout no longer depends on the service time, it is set to
2597 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
2598 time via pb_service_time.
2602 Description: Watchdog triggered on MDS failover
2603 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
2604 the OSTs know that the MDS "UUID" can be reused for the same export
2605 from a different NID, so we do not need to wait for the export to be
2609 Frequency : rare, only if using MMP with Linux RAID
2611 Description: MMP doesn't work with Linux RAID
2612 Details : While using HA for Lustre servers with Linux RAID, it is possible
2613 that MMP will not detect multiple mounts. To make this work we
2614 need to unplug the device queue in RAID when the MMP block is being
2615 written. Also while reading the MMP block, we should read it from
2616 disk and not the cached one.
2618 Severity : enhancement
2620 Description: open file using fid
2621 Details : A file can be opened using just its fid, like
2622 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
2625 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
2628 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
2629 Details : add proper locking for creation of jbd_4k slab cache
2633 Description: MMP check in ext3_remount() fails without displaying any error
2634 Details : When multiple mount protection fails during remount, proper error
2637 Severity : enhancement
2639 Description: Allow stripe size to be up to 4G-64k
2640 Details : Fix math logic to allow large stripe sizes.
2644 Description: add check for >8TB ldiskfs filesystems
2645 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
2646 Don't allow >8TB ldiskfs filesystems to be mounted without
2647 force_over_8tb mount option
2652 Descriptoin: OSS complains "too many blocks" and crashes
2653 Details : In filter_commitrw_write():
2654 1. lquota_chkquota();
2655 2. obdo_from_inode();
2656 3. lquota_pending_commit();
2657 "2." has changed uid and gid in this case, then caused
2658 lquota_chkquota() and lquota_pending_commit() to use
2659 different uid and gid. That is the root of the bug.
2663 Descriptoin: getstripe doesn't report stripe count = -1 sanely
2664 Details : Detect and print out the 'stripe count = -1' case properly.
2668 Description: Out or order replies might be lost on replay
2669 Details : In ptlrpc_retain_replayable_request if we cannot find retained
2670 request with tid smaller then one currently being added, add it
2671 to the start, not end of the list.
2673 --------------------------------------------------------------------------------
2675 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
2677 * Support for kernels:
2678 2.6.5-7.283 (SLES 9),
2679 2.6.9-55.EL (RHEL 4),
2680 2.6.16.46-0.14 (SLES 10),
2681 2.6.18.8 vanilla (kernel.org)
2682 * Client support for unpatched kernels:
2683 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2684 2.6.16 - 2.6.22 vanilla (kernel.org)
2685 * Due to recently discovered recovery problems, we do not recommend
2686 using patchless RHEL 4 clients with this or any earlier release.
2687 * Recommended e2fsprogs version: 1.39.cfs8
2688 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2689 * Starting with this release, the ldiskfs backing filesystem required
2690 by Lustre is now in its own package, lustre-ldiskfs. This package
2691 should be installed. It is versioned separately from Lustre and
2692 may be released separately in future.
2696 Description: block reactivating mgc import until all deactivates complete
2697 Details : Fix race when failing back MDT/MGS to itself (testing)
2700 Frequency : at statup only
2702 Description: mds_lov_synchronize race leads to various problems
2703 Details : simultaneous MDT->OST connections at startup can cause the
2704 sync to abort, leaving the OSC in a bad state.
2706 Severity : enhancement
2708 Description: add optional extra BUILD_VERSION info
2709 Details : add a new environment variable (namely LUSTRE_VERS) which allows
2710 to override the lustre version.
2713 Frequency : 2.6.18 servers only
2715 Description: ll_kern_mount() doesn't release the module reference
2716 Details : The ldiskfs module reference count never drops down to 0
2717 because ll_kern_mount() doesn't release the module reference.
2722 Description: server LBUG when using old ost_num_threads parameter
2723 Details : Accept the old ost_num_threads parameter but warn that it
2724 is deprecated, and fix an off-by-one error that caused an LBUG.
2729 Description: Transient SCSI error results in persistent IO issue
2730 Details : iobuf->dr_error is not reinitialized to 0 between two
2734 Frequency : sometimes when underlying device returns I/O errors
2736 Description: OSTs not going read-only during write failures
2737 Details : OSTs are not remounted read-only when the journal commit threads
2738 get I/O errors because fsfilt_ext3 calls journal_start/stop()
2739 instead of the ext3 wrappers.
2743 Description: poor connect scaling with increasing client count
2744 Details : Don't run filter_grant_sanity_check for more than 100 exports
2745 to improve scaling for large numbers of clients.
2748 Frequency : SLES10 only
2750 Description: sanity-quota.sh quotacheck failed: rc = -22
2751 Details : Quotas cannot be enabled on SLES10.
2754 Frequency : liblustre clients only
2756 Description: getdirentries does not give error when run on compute nodes
2757 Details : getdirentries does not fail when the size specified as an argument
2758 is too small to contain at least one entry
2760 Severity : enhancement
2762 Description: Add LNET router traceability for debug purposes
2763 Details : If a checksum failure occurs with a router as part of the
2764 IO path, the NID of the last router that forwarded the bulk data
2765 is printed so it can be identified.
2770 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
2771 Details : Due to a race condition, liblustre clients were occasionally
2772 evicted incorrectly.
2774 Severity : enhancement
2776 Description: lfs setstripe use optional parameters instead of postional
2779 Severity : enhancement
2781 Description: Nanosecond timestamp support for ldiskfs
2782 Details : The on-disk ldiskfs filesystem has added support for nanosecond
2783 resolution timestamps. There is not yet support for this at
2784 the Lustre filesystem level.
2787 Frequency : during server recovery
2789 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
2790 Details : request with rq_no_resend flag not awake l_wait_event if they get
2794 Frequency : nfs export on patchless client
2796 Description: connectathon hang when test nfs export over patchless client
2797 Details : Disconnected dentry cannot be found with lookup, so we do not need
2798 to unhash it or make it invalid
2801 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
2802 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
2803 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
2808 Description: open req refcounting wrong on reconnect
2809 Details : If reconnect happened between getting open reply from server and
2810 call to mdc_set_replay_data in ll_file_open, we will schedule
2811 replay for unreferenced request that we are about to free.
2812 Subsequent close will crash in variety of ways.
2813 Check that request is still eligible for replay in
2814 mdc_set_replay_data().
2819 Description: disable writes to filesystem when reading health_check file
2820 Details : the default for reading the health_check proc file has changed
2821 to NOT do a journal transaction and write to disk, because this
2822 can cause reads of the /proc file to hang and block HA state
2823 checking on a healthy but otherwise heavily loaded system. It
2824 is possible to return to the previous behaviour during configure
2825 with --enable-health-write.
2827 Severity : enhancement
2829 Description: 64-bit inode version
2830 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
2831 to the inode. This will be required for version-based recovery.
2836 Description: MDS fails to start if a duplicate client export is detected
2837 Details : in some rare cases it was possible for a client to connect to
2838 an MDS multiple times. Upon recovery the MDS would detect this
2839 and fail during startup. Handle this more gracefully.
2841 Severity : enhancement
2843 Description: Add -o localflock option to simulate old noflock
2845 Details : This will achieve local-only flock/fcntl locks
2851 Description: log_commit_thread vs filter_destroy race leads to crash
2852 Details : Take import reference before releasing llog record semaphore
2857 Description: Wrong request locking in request set processing
2858 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
2859 imp_delayed_list, in this place should be used imp_lock.
2862 Frequency : when reconnection
2864 Description: Grant Leak when osc reconnect to OST
2865 Details : When osc reconnect ost, OST(filter) should check whether it
2866 should grant more space to client by comparing fed_grant and
2867 cl_avail_grant, and return the granted space to client instead
2868 of "new granted" space, because client will call osc_init_grant
2869 to update the client grant space info.
2872 Frequency : when client reconnect to OST
2874 Description: Grant Leak when osc do resend and replay bulk write
2875 Details : When osc reconnect to OST, OST(filter)should clear grant info of
2876 bulk write request, because the grant info while be sync between
2877 OSC and OST when reconnect, and we should ignore the grant info
2878 these of resend/replay write req.
2883 Description: Grant space more than avaiable left space sometimes.
2884 Details : When then OST is about to be full, if two bulk writing from
2885 different clients came to OST. Accord the avaliable space of the
2886 OST, the first req should be permitted, and the second one
2887 should be denied by ENOSPC. But if the seconde arrived before
2888 the first one is commited. The OST might wrongly permit second
2889 writing, which will cause grant space > avaiable space.
2892 Frequency : when client is evicted
2894 Description: Grant might be wrongly erased when osc is evicted by OST
2895 Details : when the import is evicted by server, it will fork another
2896 thread ptlrpc_invalidate_import_thread to invalidate the
2897 import, where the grant will be set to 0. While the original
2898 thread will update the grant it got when connecting. So if
2899 the former happened latter, the grant will be wrongly errased
2900 because of this race.
2905 Description: Checking Stale with correct fid
2906 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
2907 to check whether it is stale, because sometimes, we want the
2908 enqueue happened anyway, and op_data.fid2 will not be initialized.
2910 Severity : enhancement
2912 Description: update patchless client
2913 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
2916 Frequency : only with 2.4 kernel
2918 Description: random memory corruption
2919 Details : size of struct ll_inode_info is to big for union inode.u and this
2920 can be cause of random memory corruption.
2925 Description: Memory leak in recovery
2926 Details : Lov_mds_md was not free in an error handler in mds_create_object.
2927 It should also check obd_fail before fsfilt_start, otherwise if
2928 fsfilt_start return -EROFS,(failover mds during mds recovery).
2929 then the req will return with repmsg->transno = 0 and rc = EROFS.
2930 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
2931 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
2932 be freed no matter whether fsfilt_commit success or not.
2935 Frequency : only with huge count clients
2937 Description: Prevents from taking the superblock lock in llap_from_page for
2939 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
2940 for llap_from_page call in ll_removepage prevents from taking the
2941 superblock lock for a soon died page.
2946 Description: Not check open intent error before release open handle
2947 Details : in some rare cases, the open intent error is not checked before
2948 release open handle, which may cause
2949 ASSERTION(open_req->rq_transno != 0), because it tries to release
2950 the failed open handle.
2955 Description: Set cat log bitmap only after create log success.
2956 Details : in some rare cases, the cat log bitmap is set too early. and it
2957 should be set only after create log success.
2961 Description: Accessing a block bevice can re-enable I/O when Lustre is
2962 tearing down a device.
2963 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
2968 Description: service threads may hog cpus when there are a lot of requests
2969 Details : Insert cond_resched to give other threads a chance to use some CPU
2974 Description: the cat log was not initialized in recovery
2975 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
2976 unlink log on mds will not be initialized until mds post
2977 recovery. And also in mds post recovery, the unlink log will
2978 initialization will be done asynchronausly, so there will be race
2979 between add unlink log and unlink log initialization.
2983 Description: brw_stats were being printed incorrectly
2984 Details : brw_stats were being printed as log2 but all of them were not
2985 recorded as log2. Also remove some code duplication arising from
2986 filter_tally_{read,write}.
2990 Frequency : rare, only in recovery.
2991 Description: ASSERTION(req->rq_type != LI_POISON) failed
2992 Details : imp_lock should be held while iterating over imp_sending_list for
2993 prevent destroy request after get timeout in ptlrpc_queue_wait.
2997 Description: replay-single.sh test 52 fails
2998 Details : A lock's skiplist need to be cleanup when it being unlinked
2999 from its resource list.
3003 Description: Short directio read returns full requested size rather than
3005 Details : Direct I/O operations should return actual amount of bytes
3006 transferred rather than requested size.
3008 Severity : enhancement
3010 Description: metadata RPC reduction (e.g. for rm performance)
3011 Details : decrease the amount of synchronous RPC between clients and servers
3012 by canceling conflicing lock before the operation on the client
3013 and packing thier handles into the main operation RPC to server.
3015 Severity : enhancement
3017 Description: Async OSC create to avoid the blocking unnecessarily.
3018 Details : If a OST has no remain object, system will block on the creating
3019 when need to create a new object on this OST. Now, ways use
3020 pre-created objects when available, instead of blocking on an
3021 empty osc while others are not empty. If we must block, we block
3022 for the shortest possible period of time.
3026 Description: improve handling recoverable errors
3027 Details : if request processig with error which can be recoverable on server
3028 request should be resend, otherwise page released from cache and
3031 Severity : enhancement
3033 Description: refine locking for avoid write wrong info into lov_objid file
3034 Details : fix possible races with add new target and write/update data in
3037 --------------------------------------------------------------------------------
3039 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
3044 Frequency : on some architectures
3046 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
3047 Details : Uninitialized flags sometimes cause configuration commands to
3051 Frequency : patchless clients only
3053 Description: missing __iget() symbol export
3054 Details : The __iget() symbol export is missing. To avoid the need for
3055 this on patchless clients the deathrow inode reaper is turned
3056 off, and we depend on the VM to clean up old inodes. This
3057 dependency was during via the fix for bug 12181.
3059 --------------------------------------------------------------------------------
3061 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
3063 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
3064 INTEROPERATE with older versions automatically. In many cases a
3065 special upgrade step is needed. Please read the
3066 user documentation before upgrading any part of a live system.
3067 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
3068 version will not interoperate with 1.6 betas before beta5 (1.5.95).
3069 * WARNING: Lustre configuration and startup changes are required with
3070 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
3074 Severity : enhancement
3076 Description: MountConf
3077 Details : Lustre configuration is now managed via mkfs and mount
3078 commands instead of lmc and lconf. New obd types (MGS, MGC)
3079 are added for dynamic configuration management. See
3080 https://mail.clusterfs.com/wikis/lustre/MountConf for
3083 Severity : enhancement
3085 Description: dynamic OST addition
3086 Details : OSTs can now be added to a live filesystem
3088 Severity : enhancement
3090 Description: startup order invariance
3091 Details : MDTs and OSTs can be started in any order. Clients only
3092 require the MDT to complete startup.
3094 Severity : enhancement
3096 Description: parallel, asynchronous orphan cleanup
3097 Details : orphan cleanup is now performed in separate threads for each
3098 OST, allowing parallel non-blocking operation.
3100 Severity : enhancement
3102 Description: optimized stripe assignment
3103 Details : stripe assignments are now made based on ost space available,
3104 ost previous usage, and OSS previous usage, in order to try
3105 to optimize storage space and networking resources.
3107 Severity : enhancement
3109 Description: Permanently set tunables
3110 Details : All writable /proc/fs/lustre tunables can now be permanently
3111 set on a per-server basis, at mkfs time or on a live system.
3113 Severity : enhancement
3115 Description: Lustre message v2
3116 Details : Add lustre message format v2.
3118 Severity : enhancement
3120 Description: client OST exclusion list
3121 Details : Clients can be started with a list of OSTs that should be
3122 declared "inactive" for known non-responsive OSTs.
3126 Description: SPEC SFS validation failure on NFS v2 over lustre.
3127 Details : Changes the blocksize for regular files to be 2x RPC size,
3128 and not depend on stripe size.
3130 Severity : enhancement
3132 Description: Multiple MD RPCs in flight.
3133 Details : Further unserialise some read-only MDS RPCs - learn about intents.
3134 To avoid overly-overloading MDS, introduce a limit on number of
3135 MDS RPCs in flight for a single client and add /proc controls
3136 to adjust this limit.
3138 Severity : enhancement
3140 Description: client read/write statistics
3141 Details : Add client read/write call usage stats for performance
3142 analysis of user processes.
3143 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
3144 file access. extents_stats shows chunk size distribution.
3145 extents_stats_per_process show chunk size distribution per
3148 Severity : enhancement
3150 Description: mds statistics
3151 Details : Add detailed mds operations statistics in
3152 /proc/fs/lustre/mds/*/stats.
3156 Description: Failure of copying files with lustre special EAs.
3157 Details : Client side always return success for setxattr call for lustre
3158 special xattr (currently only "trusted.lov").
3163 Description: Refcount LNET uuids
3164 Details : The global LNET uuid list grew linearly with every startup;
3165 refcount repeated list entries instead of always adding to
3168 Severity : enhancement
3170 Description: Dynamic service threads
3171 Details : Within a small range, start extra service threads
3172 automatically when the request queue builds up.
3175 Frequency : mixed-endian client/server environments
3177 Description: mixed-endian crashes
3178 Details : The new msg_v2 system had some failures in mixed-endian
3181 Severity : enhancement
3183 Description: Easy OST removal
3184 Details : OSTs can be permanently deactivated with e.g. 'lctl
3185 conf_param lustre-OST0001.osc.active=0'
3187 Severity : enhancement
3189 Description: MGS proc entries
3190 Details : Added basic proc entries for the MGS showing what filesystems
3193 Severity : enhancement
3195 Description: provide MGS failover
3196 Details : Added config lock reacquisition after MGS server failover.
3198 Severity : enhancement
3200 Description: add Linux 2.4 support
3201 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
3205 Description: a large application tries to do I/O to the same resource and dies
3206 in the middle of it.
3207 Details : Check the req->rq_arrival time after the call to
3208 ost_brw_lock_get(), but before we do anything about
3209 processing it & sending the BULK transfer request. This
3210 should help move old stale pending locks off the queue as
3211 quickly as obd_timeout.
3214 Frequency : when an incorrect nid is specified during startup
3216 Description: ptlrpc connect to non-existant node causes kernel crash
3217 Details : LNET can't be re-entered from an event callback, which
3218 happened when we expire a message after the export has been
3219 cleaned up. Instead, hand the zombie cleanup off to another
3222 Severity : enhancement
3224 Description: plain/inodebits lock performance improvement
3225 Details : Grouping plain/inodebits in granted list by their request modes
3226 and bits policy, thus improving the performance of search through
3230 Frequency : only if OST filesystem is corrupted
3232 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3233 Details : for a short time RPCs with bulk IO are in the replay list,
3234 but replay of bulk IOs is unimplemented. If the OST filesystem
3235 is corrupted due to disk cache incoherency and then replay is
3236 started it is possible to trip an assertion. Avoid putting
3237 committed RPCs into the replay list at all to avoid this issue.
3240 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3242 Description: System hang on startup
3243 Details : This bug allowed the liblustre (e.g. catamount) client to
3244 return to the app before handling all startup RPCs. This
3245 could leave the node unresponsive to lustre network traffic
3246 and manifested as a server ptllnd timeout.
3248 Severity : enhancement
3250 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3251 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
3252 boolean to control whether to print peer debug info when a
3253 client's RPC times out.
3256 Frequency : only for kernels with patches from Lustre below 1.4.3
3258 Description: Remove old rdonly API
3259 Details : Remove old rdonly API which unused from at least lustre 1.4.3
3262 Frequency : only for devices with external journals
3264 Description: Set external device read-only also
3265 Details : During a commanded failover stop, we set the disk device
3266 read-only while the server shuts down. We now also set any
3267 external journal device read-only at the same time.
3270 Frequency : when upgrading from 1.4 while trying to change parameters
3272 Description: The wrong (new) MDC name was used when setting parameters for
3273 upgraded MDT's. Also allows changing of OSC (and MDC)
3274 parameters if --writeconf is specified at tunefs upgrade time.
3277 Frequency : when setting specific ost indicies
3279 Description: QOS code breaks on skipped indicies
3280 Details : Add checks for missing OST indicies in the QOS code, so OSTs
3281 created with --index need not be sequential.
3283 Severity : enhancement
3285 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3286 Details : The uninit_groups feature works in conjunction with the kernel
3287 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3288 up the pass1 processing of e2fsck. This is a read-only feature
3289 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3290 mount filesystems that have had this feature enabled.
3292 Severity : enhancement
3294 Description: Improve multi-block allocation algorithm to avoid fragmentation
3295 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3296 allocation locality and avoid filesystem fragmentation.
3298 ------------------------------------------------------------------------------
3300 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
3302 * Support for kernels:
3303 2.4.21-47.0.1.EL (RHEL 3)
3304 2.6.5-7.283 (SLES 9)
3305 2.6.9-42.0.10.EL (RHEL 4)
3306 2.6.12.6 vanilla (kernel.org)
3307 2.6.16.27-0.9 (SLES 10)
3308 * Recommended e2fsprogs version: 1.39.cfs5
3310 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3314 Frequency : occasional, depends on client load and configuration
3315 Bugzilla : 12181, 12203
3316 Description: data loss for recently-modified files
3318 Details : In some cases it is possible that recently written or created
3319 files may not be written to disk in a timely manner (this should
3320 normally be within 30s unless client IO load is very high).
3321 The problem appears as zero-length files or files that are a
3322 multiple of 1MB in size after a client crash or client eviction
3323 that are missing data at the end of the file.
3325 This problem is more likely to be hit on clients where files are
3326 repeatedly created and unlinked in the same directory, clients
3327 have a large amount of RAM, have many CPUs, the filesystem has
3328 many OSTs, the clients are rebooted frequently, and/or the files
3329 are not accessed by other nodes after being written.
3331 The presence of the problem can be detected by looking at
3332 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
3333 smaller than the second (nr_unused) then dirty files will not
3334 be flushed automatically to disk. "sync; sleep 10" should be
3335 run several times on the node before unmounting it to update
3336 Lustre (this is also safe to run on nodes without this problem).
3338 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
3339 that can cause this same problem, so customers using that kernel
3340 also need to update the kernel in addition to Lustre. In order
3341 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
3343 It is normal that files written just before a client crash (less
3344 than 30s) may not yet have been flushed to disk, even for local
3348 Frequency : frequent on thin XT3 nodes
3350 Description: UUID collision on thin XT3 Linux nodes
3351 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
3352 randomly, since we relied on an insufficiently-seeded PRNG.
3357 Description: OSS hangs after "All ost request buffers busy"
3358 Details : A deadlock between quota and journal operations caused OSS
3359 hangs after printing "All ost request buffers busy."
3362 Frequency : always on liblustre builds
3364 Description: Cleanup compiler warnings on liblustre
3367 Frequency : always on liblustre builds on XT3
3369 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
3373 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
3374 > 32000 subdirectories
3375 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
3376 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
3377 subdirectory count crosses 32000. This will aid e2fsck to
3378 correctly handle more than 32000 subdirectories.
3381 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3383 Description: System hang on startup
3384 Details : This bug allowed the liblustre (e.g. catamount) client to
3385 return to the app before handling all startup RPCs. This
3386 could leave the node unresponsive to lustre network traffic
3387 and manifested as a server ptllnd timeout.
3389 Severity : enhancement
3391 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3392 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
3393 boolean to control whether to print peer debug info when a
3394 client's RPC times out.
3399 Description: make O_SYNC working on 2.6 kernels
3400 Details : 2.6 kernels use different method for mark pages for write,
3401 so need add a code to lustre for O_SYNC work.
3406 Description: Failure to close file and release space on NFS
3407 Details : Put inode details into lock acquired in ll_intent_file_open.
3408 Use mdc_intent_lock in ll_intent_open to properly
3409 detect all kind of errors unhandled by mdc_enqueue.
3414 Description: proc file read during shutdown sometimes raced obd removal,
3416 Details : Add lock to prevent obd access after proc file removal.
3419 Frequency : Only for files larger than 4GB on 32-bit clients.
3421 Description: improperly doing page alignment of locks
3422 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
3423 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3428 Description: RPCs being resent when they shouldn't be
3429 Details : Some RPCs that should not be resent are being resent. This
3430 can cause inconsistencies in the RPC state machine. Do not
3431 resend such requests.
3434 Frequency : rare, only with NFS export
3436 Description: Crash on NFS re-export node
3437 Details : under very unusual load conditions an assertion is hit in
3438 ll_intent_file_open()
3441 Frequency : only if OST filesystem is corrupted
3443 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3444 Details : for a short time RPCs with bulk IO are in the replay list,
3445 but replay of bulk IOs is unimplemented. If the OST filesystem
3446 is corrupted due to disk cache incoherency and then replay is
3447 started it is possible to trip an assertion. Avoid putting
3448 committed RPCs into the replay list at all to avoid this issue.
3453 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3454 Details : Large single O_DIRECT read and write calls can fail to allocate
3455 a sufficiently large buffer to process the request. In case of
3456 allocation failure the allocation is retried with a smaller
3457 buffer and broken into smaller requests.
3459 Severity : enhancement
3461 Description: Add -o localflock option to simulate old noflock behaviour.
3462 Details : This will achieve local-only flock/fcntl locks coherentness.
3467 Description: versioning check is incomplete
3468 Details : Checking the version difference of client vs. server, report
3469 error if the gap is too big.
3474 Description: add support PG_writeback bit
3475 Details : add support for PG_writeback bit for Lustre, for more carefull
3476 work with page cache in 2.6 kernel. This also fix some deadlocks
3477 and remove hack for work O_SYNC with 2.6 kernel.
3479 Severity : enhancement
3481 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3482 Details : The uninit_groups feature works in conjunction with the kernel
3483 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3484 up the pass1 processing of e2fsck. This is a read-only feature
3485 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3486 mount filesystems that have had this feature enabled.
3488 Severity : enhancement
3490 Description: Improve multi-block allocation algorithm to avoid fragmentation
3491 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3492 allocation locality and avoid filesystem fragmentation.
3494 ------------------------------------------------------------------------------
3496 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3498 * Support for kernels:
3499 2.6.9-42.0.3.EL (RHEL 4)
3500 2.6.5-7.276 (SLES 9)
3501 2.4.21-47.0.1.EL (RHEL 3)
3502 2.6.12.6 vanilla (kernel.org)
3503 2.6.16.21-0.8 (SLES10)
3504 * Recommended e2fsprogs version: 1.39.cfs2-0
3506 * The backwards-compatible /proc/sys/portals symlink has been removed
3507 in this release. Before upgrading, please ensure that you change
3508 any configuration scripts or /etc/sysctl.conf files that access
3509 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3510 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3511 in advance of the upgrade on any system running Lustre 1.4.6 or
3512 newer, since /proc/sys/lnet was added in that version.
3513 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3517 Frequency : only when quota is used
3519 Description: avoid scanning export list for quota master
3520 Details : Change the algorithms to avoid scanning export list in order
3521 to improve the efficiency.
3524 Frequency : MDS failover only, very rarely
3526 Description: "went back in time" messages on mds failover
3527 Details : The greatest transno may be lost when the current operation
3528 finishes with an error (transno==0) and the client's last_rcvd
3529 record is over-written. Save the greatest transno in the
3530 mds_last_transno for this case.
3533 Frequency : always for specific kernels and striping counts
3535 Description: client may get "Matching packet too big" without ACL support
3536 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
3537 when trying to access files in certain configurations. The
3538 clients should in fact be denied when mounting because they do
3539 not understand ACLs.
3542 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
3544 Description: many clients connecting with IO in progress causes connect timeouts
3545 Details : Avoid synchronous journal commits to avoid delays caused by many
3546 clients connecting/disconnecting when bulk IO is in progress.
3547 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
3548 OST_IO_PORTAL to avoid delays behind potentially many pending
3552 Frequency : occasionally with multiple writers to a single file
3554 Description: shared writes to file may result in wrong size reported by stat()
3555 Details : Allow growing of kms when extent lock is cancelled
3558 Frequency : always with random mmap IO to multi-striped file
3560 Description: mmap write might be lost if we are writing to a 'hole' in stripe
3561 Details : Only if the hole is at the end of OST object so that kms is too
3562 small. Fix is to increase kms accordingly in ll_nopage.
3565 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
3567 Description: writes to a missing object would leak memory on the OST
3568 Details : If there is an inconsistency between the MDS and OST filesystems,
3569 such that the MDS references an object that doesn't exist, writes
3570 to that object will leak memory due to incorrect cleanup in the
3571 error handling path, eventually running out of memory on the OST.
3576 Description: Creating too long symlink causes lustre errors
3577 Details : Check symlink and name lengths before sending requests to MDS.
3580 Frequency : only if flock is enabled (not on by default)
3582 Description: posix locks not released on fd closure on 2.6.9+
3583 Details : We failed to add posix locks to list of inode locks on 2.6.9+
3584 kernels, this caused such locks not to be released on fd close and
3585 then assertions on fs unmount about still used locks.
3588 Frequency : MDS failover only, very rarely
3590 Description: clients may get ASSERTION(granted_lock != NULL)
3591 Details : When request was taking a long time, and a client was resending
3592 a getattr by name lock request. The were multiple lock requests
3593 with the same client lock handle and
3594 mds_getattr_name->fixup_handle_for_resent_request found one of the
3595 lock handles but later failed with ASSERTION(granted_lock != NULL).
3600 Description: handle->h_buffer_credits > 0, assertion failure
3601 Details : h_buffer_credits is zero after truncate, causing assertion
3602 failure. This patch extends the transaction or creates a new
3606 Frequency : NFS re-export or patchless client
3607 Bugzilla : 11179, 10796
3608 Description: Crash on NFS re-export node (__d_move)
3609 Details : We do not want to hash the dentry if we don't have a lock.
3610 But if this dentry is later used in d_move, we'd hit uninitialised
3611 list head d_hash, so we just do this to init d_hash field but
3612 leave dentry unhashed.
3615 Frequency : NFS re-export or patchless client
3617 Description: NFS exports has problem with symbolic link
3618 Details : lustre client didn't properly install dentry when re-exported
3619 to NFS or running patchless client.
3622 Frequency : NFS re-export or patchless client
3624 Description: Various nfs/patchless fixes.
3625 Details : fixes reuse disconected alias for lookup process - this fixes
3626 warning "find_exported_dentry: npd != pd",
3627 fix permission error with open files at nfs.
3628 fix apply umask when do revalidate.
3631 Frequency : occasional
3633 Description: Crash on NFS re-export node
3634 Details : calling clear_page() on the wrong pointer triggered oops in
3635 generic_mapping_read().
3638 Frequency : rarely, using O_DIRECT IO
3640 Description: unaligned directio crashes client with LASSERT
3641 Details : check for unaligned buffers before trying any requests.
3644 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
3646 Description: stale data returned from RAID cache
3647 Details : If only a small amount of IO is done to the RAID device before
3648 reading it again it is possible to get stale data from the RAID
3649 cache instead of reading it from disk.
3652 Frequency : always for sles10 kernel
3654 Description: sles10 support
3655 Details : ll_follow_link: compile fixes and using of nd_set_link
3656 under newer kernels.
3659 Frequency : depends on arch, kernel and compiler version, always on sles10
3662 Description: recursive or deep enough symlinks cause stack overflow
3663 Details : getting rid of large stack-allocated variable in
3667 Frequency : depends on hardware
3669 Description: lustre write performance loss in the SLES10 kernel
3670 Details : the performance loss is caused by using of write barriers in the
3671 ext3 code. The SLES10 kernel turns barrier support on by
3672 default. The fix is to undo that change for ldiskfs.
3674 ------------------------------------------------------------------------------
3676 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
3678 * Support for kernels:
3679 2.6.9-42.0.3EL (RHEL 4)
3680 2.6.5-7.276 (SLES 9)
3681 2.4.21-47.0.1.EL (RHEL 3)
3682 2.6.12.6 vanilla (kernel.org)
3686 Frequency : quota enabled and large files being deleted
3688 Description: releasing more than 4GB of quota at once hangs OST
3689 Details : If a user deletes more than 4GB of files on a single OST it
3690 will cause the OST to spin in an infinite loop. Release
3691 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3696 Description: statfs data retrieved from /proc may be stale or zero
3697 Details : When reading per-device statfs data from /proc, in the
3698 {kbytes,files}_{total,free,avail} files, it may appear
3699 as zero or be out of date.
3702 Frequency : systems with MD RAID1 external journal devices
3704 Description: lconf's call to blkid is confused by RAID1 journal devices
3705 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3706 of returning all block devices that match the journal UUID.
3709 Frequency : always, for aggregate stripe size over 4GB
3711 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
3712 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
3713 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
3716 Frequency : always if "lfs find" used on a local file/directory
3718 Description: "lfs find" segfaults if used on a local file/directory
3719 Details : The case where a directory component was not specified wasn't
3720 handled correctly. Handle this properly.
3723 Frequency : always on ppc64
3725 Description: the write to an ext3 filesystem mounted with mballoc got stuck
3726 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3727 perform endianness conversion.
3730 Frequency : rarely (truncate to non-zero file size after write under load)
3731 Bugzilla : 10730, 10687
3732 Description: Files padded with zeros to next 4K multiple
3733 Details : With filesystems mounted using the "extents" option (2.6 kernels)
3734 it is possible that files that are truncated to a non-zero size
3735 immediately after being written are filled with zero bytes beyond
3736 the truncated size. No file data is lost.
3738 Severity : enhancement
3740 Description: Allow recovery/failover for liblustre clients.
3741 Details : liblustre clients were unaware of failover configurations until
3744 Severity : enhancement
3746 Description: user file locks should fail when not mounting with flock option
3747 Details : Set up an error-returning stub in ll_file_operations.lock field
3748 to prevent incorrect behaviour when client is mounted without
3749 flock option. Also, set up properly f_op->flock field for
3753 Frequency : always on ia64
3755 Description: "lfs df" loops on printing out MDS statfs information
3756 Details : The obd_ioctl_data was not initialized and in some systems
3757 this caused a failure during the ioctl that did not return
3758 an error. Initialize the struct and return an error on failure.
3761 Frequency : SLES 9 only
3763 Description: Error of copying files with lustre special EAs as root
3764 Details : Client side always return success for setxattr call for lustre
3765 special xattr (currently only "trusted.lov").
3768 Frequency : rarely on clusters with both ia64+i386 clients
3770 Description: ia64+i686 clients doing shared IO on the same file may LBUG
3771 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
3772 clients are doing concurrent writes to the same file it is
3773 possible that the ia64 clients may LASSERT because the OST
3774 extent locks are not PAGE_SIZE aligned. Ensure that grown
3775 locks are always aligned on the request boundary.
3778 Frequency : specific use, occasional
3780 Description: Overwriting in use executable truncates on-disk binary image
3781 Details : If one node attempts to overwrite an executable in use by
3782 another node, we now correctly return ETXTBSY instead of
3783 truncating the file.
3785 Severity : enhancement
3787 Description: Async OSC create to avoid the blocking unnecessarily.
3788 Details : If a OST has no remain object, system will block on the creating
3789 when need to create a new object on this OST. Now, ways use
3790 pre-created objects when available, instead of blocking on an
3791 empty osc while others are not empty. If we must block, we block
3792 for the shortest possible period of time.
3797 Description: chmod on Lustre root is propagated to other clients
3798 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
3799 invalid by the follow_mount time.
3802 Frequency : liblustre clients only
3804 Description: Race in 'instant cancel' lock handling could lead to such locks
3805 never to be granted in case of SMP MDS
3806 Details : Do not destroy not yet granted but cbpending locks in
3810 Frequency : replay/resend of open
3812 Description: non null lock assetion failure in mds_intent_policy
3813 Details : Trying to replay/resend lockless open requests resulted in
3814 mds_open() returning 0 with no lock. Now it sets a flag if
3815 a lock is going to be returned.
3817 Severity : enhancement
3819 Description: Checksum enhancements
3820 Details : New checksum enhancements allow for resending RPCs that failed
3823 Severity : enhancement
3825 Description: Tunables on number of dirty pages in cacche
3826 Details : Allow to set limit on number of dirty pages cached.
3831 Description: client crash on unmount - lock still has references
3832 Details : In some error handling cases it was possible to leak a lock
3833 reference on a client while accessing a file. This was not
3834 harmful to the client during operation, but would cause the
3835 client to crash when the filesystem is unmounted.
3838 Frequency : specific case, rare
3840 Description: ETXTBSY on mds though file not in use by client
3841 Details : ETXTBSY is no longer incorrectly returned when attempting to
3842 chmod or chown a directory that the user previously tried to
3843 execute or a currently-executing binary.
3846 Frequency : extremely rare except on liblustre-based clients
3848 Description: Lustre space not freed when files are deleted
3849 Details : Clean up open-unlinked files after client eviction. Previously
3850 the unlink was skipped and the files remained as orphans.
3855 Description: OST failure "would be an LBUG" in waiting_locks_callback()
3856 Details : In some cases it was possible to send a blocking callback to
3857 a client doing a glimpse, even though that client didn't get
3858 a lock granted. When the glimpse lock is cancelled on the OST
3859 the freed lock is left on the waiting list and corrupted the list.
3862 Frequency : all core dumps
3864 Description: Broke core dumps to lustre
3865 Details : Negative dentry may be unhashed if parent does not have UPDATE
3866 lock, but some callers, e.g. do_coredump, expect dentry to be
3867 hashed after successful create, hash it in ll_create_it.
3869 ------------------------------------------------------------------------------
3871 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
3873 * Support for kernels:
3874 2.6.9-42.0.2.EL (RHEL 4)
3875 2.6.5-7.276 (SLES 9)
3876 2.4.21-40.EL (RHEL 3)
3877 2.6.12.6 vanilla (kernel.org)
3881 Frequency : always on RHEL 3
3883 Description: Number of open files grows over time
3884 Details : The number of open files grows over time, whether or not
3885 Lustre is started. This was due to a filp leak introduced
3886 by one of our kernel patches.
3888 ------------------------------------------------------------------------------
3890 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
3892 * Support for kernels:
3893 2.6.9-42.EL (RHEL 4)
3894 2.6.5-7.276 (SLES 9)
3895 2.4.21-40.EL (RHEL 3)
3896 2.6.12.6 vanilla (kernel.org)
3901 Bugzilla : 5719, 9635, 9792, 9684
3902 Description: OST (or MDS) trips assertions in (re)connection under heavy load
3903 Details : If a server is under heavy load and cannot reply to new
3904 connection requests before the client resends the (re)connect,
3905 the connection handling code can behave badly if two service
3906 threads are concurrently handing separate (re)connections from
3907 the same client. Add better locking to the connection handling
3908 code, and ensure that only a single connection will be processed
3909 for a given client UUID, even if the lock is dropped.
3911 Severity : enhancement
3913 Description: add TCP zero-copy support to kernel
3914 Details : Add support to the kernel TCP stack to allow zero-copy bulk
3915 sends if the hardware supports scatter-gather and checksumming.
3916 This allows socklnd to do client-write and server-read more
3917 efficiently and reduce CPU utilization from skbuf copying.
3920 Frequency : only if NFS exporting from client
3922 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
3923 Details : If setting attributes on a file created under NFS that had
3924 never been opened it would be possible to oops the client
3925 if the file had no objects.
3928 Frequency : always for liblustre
3930 Description: liblustre client does MDS+OSTs setattr RPC for each write
3931 Details : When doing a write from a liblustre client, the client
3932 incorrectly issued an RPC to the MDS and each OST the file was
3933 striped over in order to update the timestamps. When writing
3934 with small chunks and many clients this could overwhelm the MDS
3935 with RPCs. In all cases it would slow down the write because
3936 these RPCs are unnecessary.
3938 Severity : enhancement
3940 Description: allow number of MDS service threads to be changed at module load
3941 Details : It is now possible to change the number of MDS service threads
3942 running. Adding "options mds mds_num_threads={N}" to the MDS's
3943 /etc/modprobe.conf will set the number of threads for the next
3944 time Lustre is restarted (assuming the "mds" module is also
3945 reloaded at that time). The default number of threads will
3946 stay the same, 32 for most systems.
3951 Description: OST crash if filesystem is unformatted or corrupt
3952 Details : If an OST is started on a device that has never been formatted
3953 or if the filesystem is corrupt and cannot even mount then the
3954 error handling cleanup routines would dereference a NULL pointer.
3959 Description: NULL pointer deref in llap_from_page.
3960 Details : get_cache_page_nowait can return a page with NULL (or otherwise
3961 incorrect) mapping if the page was truncated/reclaimed while it was
3962 searched for. Check for this condition and skip such pages when
3963 doing readahead. Introduce extra check to llap_from_page() to
3964 verify page->mapping->host is non-NULL (so page is not anonymous).
3967 Frequency : Sometimes when using sys_sendfile
3969 Description: "page not covered by a lock" warnings from ll_readpage
3970 Details : sendfile called ll_readpage without right page locks present.
3971 Now we introduced ll_file_sendfile that does necessary locking
3972 around call to generic_file_sendfile() much like we do in
3976 Frequency : with certain MDS communication failures at client mount time
3978 Description: NULL pointer deref after failed client mount
3979 Details : a client connection request may delayed by the network layer
3980 and not be sent until after the PTLRPC layer has timed out the
3981 request. If the client fails the mount immediately it will try
3982 to clean up before the network times out the request. Add a
3983 reference from the request import to the obd device and delay
3984 the cleanup until the network drops the request.
3987 Frequency : occasionally during client (re)connect
3989 Description: assertion failure during client (re)connect
3990 Details : processing a client connection request may be delayed by the
3991 client or server longer than the client connect timeout. This
3992 causes the client to resend the connection request. If the
3993 original connection request is replied in this interval, the
3994 client may trip an assertion failure in ptlrpc_connect_interpret()
3995 which thought it would be the only running connect process.
3998 Frequency : only with obd_echo servers and clients that are rebooted
4000 Description: kernel BUG accessing uninitialized data structure
4001 Details : When running an obd_echo server it did not start the ping_evictor
4002 thread, and when a client was evicted an uninitialized data
4003 structure was accessed. Start the ping_evictor in the RPC
4004 service startup instead of the OBD startup.
4006 Severity : enhancement
4007 Bugzilla : 10193 (patchless)
4008 Description: Remove dependency on various unexported kernel interfaces.
4009 Details : No longer need reparent_to_init, exit_mm, exit_files,
4010 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
4013 Frequency : rare (only users of deprecated and unsupported LDAP config)
4015 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
4016 Details : LDAP apparently contains 'lustreName' attributes instead of
4017 'name'. A simple remapping of the name is sufficient.
4020 Frequency : rare (only with non-default dump_on_timeout debug enabled)
4022 Description: waiting_locks_callback trips kernel BUG if client is evicted
4023 Details : Running with the dump_on_timeout debug flag turned on makes
4024 it possible that the waiting_locks_callback() can try to dump
4025 the Lustre kernel debug logs from an interrupt handler. Defer
4026 this log dumping to the expired_lock_main() thread.
4028 Severity : enhancement
4030 Description: Support NFS exporting on 2.6 kernels.
4031 Details : Implement non-rawops metadata methods for NFS server to use without
4032 changing NFS server code.
4035 Frequency : very rare (synthetic metadata workload only)
4037 Description: two racing renames might cause an MDS thread to deadlock
4038 Details : Running the "racer" program may cause one MDS thread to rename
4039 a file from being the source of a rename to being the target of
4040 a rename at exactly the same time that another thread is doing
4041 so, and the second thread has already enqueued these locks after
4042 doing a lookup of the target and is trying to relock them in
4043 order. Ensure that we don't try to re-lock the same resource.
4046 Frequency : only very large systems with liblustre clients
4048 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
4049 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
4050 all OSTs in parallel. This allows the eviction of stale liblustre
4051 clients to proceed much faster than if they were done in series,
4052 and also offers similar improvements for other set_info RPCs.
4057 Description: excessive CPU usage during initial read phase on client
4058 Details : During the initial read phase on a client, it would agressively
4059 retry readahead on the file, consuming too much CPU and impacting
4060 performance (since 1.4.5.8). Improve the readahead algorithm
4061 to avoid this, and also improve some other common cases (read
4062 of small files in particular, where "small" is files smaller than
4063 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
4068 Description: MDS crash when receiving packet with unknown intent.
4069 Details : Do not LBUG in unknown intent case, just return -EFAULT
4071 Severity : enhancement
4072 Bugzilla : 9293, 9385
4073 Description: MDS RPCs are serialised on client. This is unnecessary for some.
4074 Details : Do not serialize getattr (non-intent version) and statfs.
4077 Frequency : occasional, when OST network is overloaded/intermittent
4079 Description: client evicted by OST after bulk IO timeout
4080 Details : If a client sends a bulk IO request (read or write) the OST
4081 may evict the client if it is unresposive to its data GET/PUT
4082 request. This is incorrect if the network is overloaded (takes
4083 too long to transfer the RPC data) or dropped the OST GET/PUT
4084 request. There is no need to evict the client at all, since
4085 the pinger and/or lock callbacks will handle this, and the
4086 client can restart the bulk request.
4089 Frequency : Always when mmapping file with no objects
4091 Description: client crashes when mmapping file with no objects
4092 Details : Check that we actually have objects in a file before doing any
4093 operations on objects in ll_vm_open, ll_vm_close and
4099 Description: Request leak when working with deleted CWD
4100 Details : Introduce advanced request refcount tracking for requests
4101 referenced from lustre intent.
4103 Severity : Enhancement
4105 Description: Cache open file handles on client.
4106 Details : MDS now will return special lock along with openhandle, if
4107 requested and client is allowed to hold openhandle, even if unused,
4108 until such a lock is revoked. Helps NFS a lot, since NFS is opening
4109 closing files for every read/write openration.
4111 Severity : Enhancement
4113 Description: Cache open negative dentries on client when possible.
4114 Details : Guard negative dentries with UPDATE lock on parent dir, drop
4115 negative dentries on lock revocation.
4120 Description: Remounting a client read-only wasn't possible with a zconf mount
4121 Details : It wasn't possible to remount a client read-only with llmount.
4123 Severity : enhancement
4124 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
4125 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
4126 MPICH is included to add Lustre-specific ADIO interfaces.
4127 This is based closely on the UFS ADIO layer and only differs
4128 in file creation, in order to allow the OST striping to be set.
4129 This is user-contributed code and not supported by CFS.
4134 Description: extended inode attributes (immutable, append-only) work improperly
4135 when 2.4 and 2.6 kernels are used on client/server or vice versa
4136 Details : Introduce kernel-independent values for these flags.
4138 Severity : enhancement
4141 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
4142 Details : Many of the /proc/ tunables can only be tuned at a megabyte
4143 granularity. Now, Fractional MB granularity is be supported,
4144 this is very useful for low memory system.
4146 Severity : enhancement
4148 Description: Getattr by fid
4149 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
4150 locks, avoids extra getattr rpc requests to MDS, allows '/' to
4151 have locks and avoids getattr rpc requests for it on every stat.
4154 Frequency : Always, for filesystems larger than 2TB
4156 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4157 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
4158 filesystems larger than 2TB to be mounted with mballoc enabled.
4161 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
4163 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4164 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
4165 filesystems larger than 2TB is mounted then the kernel will
4166 silently corrupt the start of the filesystem. CONFIG_LBD is
4167 enabled for all CFS-supported kernels, but the possibility of
4168 this happening with a modified kernel config exists.
4170 Severity : enhancement
4172 Description: add client O_DIRECT support for 2.6 kernels
4173 Details : It is now possible to do O_DIRECT reads and writes to files
4174 in the Lustre client mountpoint on 2.6 kernel clients.
4176 Severity : enhancement
4178 Description: parallel glimpse, setattr, statfs, punch, destroy requests
4179 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
4180 parallel, not waiting for response from every OST before sending
4181 a rpc to the next OST.
4186 Description: setattr vs write race when updating file timestamps
4187 Details : Client processes that update a file timestamp into the past
4188 right after writing to the file (e.g. tar) it is possible that
4189 the updated file modification time can be reset to the current
4190 time due to a race between processing the setattr and write RPC.
4192 Severity : enhancement
4194 Description: Bring 'lfs find' closer in line with regular Linux find.
4195 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
4196 -print0 options and obtains all the needed info through the lustre
4199 Severity : enhancement
4201 Description: support up to 1024 configured devices on one node
4202 Details : change obd_dev array from statically allocated to dynamically
4203 allocated structs as they are first used to reduce memory usage
4208 Description: Flush dirty partially truncated pages during truncate
4209 Details : Immediatelly flush partially truncated pages in filter_setattr,
4210 this way we completely avoid having any pages in page cache on OST
4211 and can retire ugly workarounds during writes to flush such pages.
4216 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
4217 Details : protect inode from truncation within vfs_unlink() context
4218 just take a reference before calling vfs_unlink() and release it
4219 when parent's i_sem is free.
4224 Description: last_id value checked outside lock on OST caused LASSERT failure
4225 Details : If there were multiple MDS->OST object precreate requests in
4226 flight, it was possible that the OST's last object id was checked
4227 outside a lock and incorrectly tripped an assertion. Move checks
4228 inside locks, and discard old precreate requests.
4231 Frequency : always, if extents are used on OSTs
4233 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
4234 Details : OSTs using the extents format would not zero the high 16 bits of
4235 the index physical block number. This is not a problem for any
4236 OST filesystems smaller than 16TB, and no kernels support ext3
4237 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
4238 new/modified files) and can be fixed for existing filesystems
4239 with e2fsprogs-1.39-cfs1.
4244 Description: import connection selection may be incorrect if timer wraps
4245 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
4246 import connections to be ignored if the 32-bit jiffies counter
4247 wraps. Use a 64-bit jiffies counter.
4250 Frequency : very large clusters immediately after boot
4252 Description: LNET request buffers exhausted under heavy short-term load
4253 Details : If a large number of client requests are generated on a service
4254 that has previously never seen so many requests it is possible
4255 that the request buffer growth cannot keep up with the spike in
4256 demand. Instead of dropping incoming requests, they are held in
4257 the LND until the RPC service can accept more requests.
4260 Frequency : Sometimes during replay
4262 Description: Assertion failure in ll_local_open after replay.
4263 Details : If replay happened on an open request reply before we were able
4264 to set replay handler, reply will become not swabbed tripping the
4265 assertion in ll_local_open. Now we set the handler right after
4266 recognising of open request
4269 Frequency : very rare
4271 Description: kernel reports "badness in vsnprintf"
4272 Details : Reading from the "recovery_status" /proc file in small chunks
4273 may cause a negative length in lprocfs_obd_rd_recovery_status()
4274 call to vsnprintf() (which is otherwise harmless). Exit early
4275 if there is no more space in the output buffer.
4277 Severity : enhancement
4279 Description: clear OBD RPC statistics by writing to them
4280 Details : It is now possible to clear the OBD RPC statistics by writing
4281 to the "stats" file.
4286 Description: Client mtime is not the same on different clients after utimes
4287 Details : In some cases, the client was using the utimes() syscall on
4288 a file cached on another node. The clients now validate the
4289 ctime from the MDS + OSTs to determine which one is right.
4294 Description: Inability to activate failout mode
4295 Details : lconf script incorrectly assumed that in pythong string's numeric
4296 value is used in comparisons.
4299 Frequency : always with multiple stripes per file
4301 Description: Inefficient object allocation for mutli-stripe files
4302 Details : When selecting which OSTs to stripe files over, for files with
4303 a stripe count that divides evenly into the number of OSTs,
4304 the MDS is always picking the same starting OST for each file.
4305 Return the OST selection heuristic to the original design.
4310 Description: mount failures may take full timeout to return an error
4311 Details : Under some heavy load conditions it is possible that a
4312 failed mount can wait for the full obd_timeout interval,
4313 possibly several minutes, before reporting an error.
4314 Instead return an error as soon as the status is known.
4316 Frequency : quota enabled and large files being deleted
4318 Description: releasing more than 4GB of quota at once hangs OST
4319 Details : If a user deletes more than 4GB of files on a single OST it
4320 will cause the OST to spin in an infinite loop. Release
4321 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4326 Description: statfs data retrieved from /proc may be stale or zero
4327 Details : When reading per-device statfs data from /proc, in the
4328 {kbytes,files}_{total,free,avail} files, it may appear
4329 as zero or be out of date.
4332 Frequency : systems with MD RAID1 external journal devices
4334 Description: lconf's call to blkid is confused by RAID1 journal devices
4335 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4336 of returning all block devices that match the journal UUID.
4339 Frequency : always, for aggregate stripe size over 4GB
4341 Description: assertion fails when trying to use 4GB stripe size
4342 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
4343 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
4346 Frequency : always on ppc64
4348 Description: the first write on an ext3 filesystem with mballoc got stuck
4349 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4350 perform endianness conversion.
4352 ------------------------------------------------------------------------------
4354 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4356 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
4357 INTEROPERATE with older versions automatically. Please read the
4358 user documentation before upgrading any part of a live system.
4359 * WARNING: Lustre networking configuration changes are required with
4360 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
4363 * Support for kernels:
4364 2.6.9-22.0.2.EL (RHEL 4)
4365 2.6.5-7.244 (SLES 9)
4366 2.6.12.6 vanilla (kernel.org)
4369 Severity : enhancement
4370 Bugzilla : 7981/8208
4371 Description: Introduced Lustre Networking (LNET)
4372 Details : LNET is new networking infrastructure for Lustre, it includes
4373 a reorganized network configuration mode (see the user
4374 documentation for full details) as well as support for routing
4375 between different network fabrics. Lustre Networking Devices
4376 (LNDs) for the supported network fabrics have also been
4377 created for this new infrastructure.
4379 Severity : enhancement
4380 Description: Introduced Access control lists
4381 Details : clients can set ACLs on files and directories in order to have
4382 more fine-grained permissions than the standard Unix UGO+RWX.
4383 The MDS must be started with the "-o acl" mount option.
4385 Severity : enhancement
4386 Description: Introduced filesystem quotas
4387 Details : Administrators may now establish per-user quotas on the
4390 Severity : enhancement
4392 Description: Configuration change for the XT3
4393 The PTLLND is now used to run Lustre over Portals on the XT3
4394 The configure option(s) --with-cray-portals are no longer used.
4395 Rather --with-portals=<path-to-portals-includes> is used to
4396 enable building on the XT3. In addition to enable XT3 specific
4397 features the option --enable-cray-xt3 must be used.
4402 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
4403 Details : A race between allocating a new llcd and re-getting the llcd_lock
4404 allowed another thread to grab newly-allocated llcd.
4406 Severity : enhancement
4408 Description: 2.6 OST async journal commit and locking fix to improve performance
4409 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
4410 2.6 kernels are now async as they already were in 2.4 kernels so
4411 that they can commit concurrently with the network bulk transfer.
4412 For block-allocated files the filter allocation semaphore is held
4413 to avoid filesystem fragmentation during allocation. BKL lock
4414 removed for 2.6 xattr operations where it is no longer needed.
4419 Description: lconf incorrectly determined whether two IP networks could talk
4420 Details : In some more complicated routing and multiple-network
4421 configurations, lconf will avoid trying to make a network
4422 connection to a disjoint part of the IP space. It was doing the
4423 math incorrectly for one set of cases.
4428 Description: Fix for potential infinite loop processing records in an llog.
4429 Details : If an llog record is corrupted/zeroed, it is possible to loop
4430 forever in llog_process(). Validate the llog record length
4431 and skip the remainder of the block on error.
4434 Frequency : occasional (liblustre only)
4436 Description: liblustre could not open files whose last component is a symlink
4437 Details : sysio_path_walk() would incorrectly pass the open intent to
4438 intermediate path components.
4441 Frequency : rare (liblustre only with non-standard tuning)
4442 Bugzilla : 7201 (7350)
4443 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4444 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4445 after it has been released, only do this for referenced locks
4447 Severity : enhancement
4449 Description: specifying an (invalid) directory default stripe_size of -1
4450 would reset the directory default striping
4451 Details : stripe_size -1 was used internally to signal directory stripe
4452 removal, now use "all default" to signal dir stripe removal
4453 as a directory striping of "all default" is not useful
4456 Frequency : common for large clusters running liblustre clients
4458 Description: doing an ls when liblustre clients are running is slow
4459 Details : sending a glimpse AST to a liblustre client waits for every AST
4460 to time out, as liblustre clients will not respond. Since they
4461 cannot cache data we refresh the OST lock LVB from disk instead.
4463 Severity : enhancement
4465 Description: doing an ls at the same time as file IO can be slow
4466 Details : enqueue and other "small" requests can be blocked behind many
4467 large IO requests. Create a new OST IO portal for non-IO
4468 requests so they can be processed faster.
4471 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4473 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4474 created file with mode 000
4475 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4476 allow an ftruncate() as a non-root user to a file with mode 000.
4477 Lustre now respects this flag to disable mode checks when
4478 truncating a file owned by the user
4481 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4484 Description: Revoking locks from clients that went dead or catatonic might take
4486 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4487 cancellation of such locks instant on servers without waiting for
4488 any reply from clients. Clients drops these locks when cancel
4489 notification from server is received without replying.
4492 Frequency : liblustre-only, when liblustre client dies or becomes busy
4494 Description: Doing ls on Linux clients can take a long time with active
4496 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4497 granting such locks to it in the first place if possible. Locks
4498 are taken by proxy on the OST during the read or write and
4499 dropped immediately afterward. Add connect flags handling, do
4500 not grant locks to liblustre clients for glimpse ASTs.
4502 Severity : enhancement
4504 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4505 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4506 sequential reads and behaves badly with applications doing random
4507 reads. Improve it by reading ahead at least read region, and
4508 avoiding excessive large RPC for small reads.
4510 Severity : enhancement
4512 Description: Creating more than 1000 files for a single job may cause a load
4513 imbalance on the OSTs if there are also a large number of OSTs.
4514 Details : qos_prep_create() uses an OST index reseed value that is an
4515 even multiple of the number of available OSTs so that if the
4516 reseed happens in the middle of the object allocation it will
4517 still utilize the OSTs as uniformly as possible.
4522 Description: OST or MDS may oops in ping_evictor_main()
4523 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
4524 but doesn't restart at beginning of obd_exports_timed list
4527 Severity : enhancement
4529 Description: improve by-nid export eviction on the MDS and OST
4530 Details : allow multiple exports with the same NID to be evicted at one
4531 time without re-searching the exports list.
4534 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
4536 Description: MDS may oops in groups_free()
4537 Details : in rare race conditions a newly allocated group_info struct is
4538 freed again, and this can be NULL. The 2.4 compatibility code
4539 for groups_free() checked for a NULL pointer, but 2.6 did not.
4542 Frequency : common for liblustre clients doing little filesystem IO
4543 Bugzilla : 9352, 7313
4544 Description: server may evict liblustre clients accessing contended locks
4545 Details : if a client is granted a lock or receives a completion AST
4546 with a blocking AST already set it would not reply to the AST
4547 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
4550 Frequency : lfs setstripe, only systems with more than 160 OSTs
4552 Description: unable to set striping with a starting offset beyond OST 160
4553 Details : llapi_create_file() incorrectly limited the starting stripe
4554 index to the maximum single-file stripe count.
4557 Frequency : LDAP users only
4559 Description: lconf did not handle in-kernel recovery with LDAP properly
4560 Details : lconf/LustreDB get_refs() is searching the wrong namespace
4562 Severity : enhancement
4564 Description: bind OST threads to NUMA nodes to improve performance
4565 Details : all OST threads are uniformly bound to CPUs on a single NUMA
4566 node and do their allocations there to localize memory access
4568 Severity : enhancement
4570 Description: llmount can determine client NID directly from Myrinet (GM)
4571 Details : the client NID code from gmnalnid was moved directly into
4572 llmount, removing the need to use this or specifying the
4573 client NID explicitly when mounting GM clients with zeroconf
4576 Frequency : if client is started with down MDS
4578 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
4579 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
4580 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
4581 ptlrpc_import_delayed_req() because we didn't check if the
4582 request was interrupted, and we also didn't break out of the
4583 event loop if there was no timeout
4588 Description: data loss during non-page-aligned writes to a single file from
4589 both multiple nodes and multiple threads on one node at same time
4590 Details : updates to KMS and lsm weren't protected by common lock. Resulting
4591 inconsistency led to false short-reads, that were cached and later
4592 used by ->prepare_write() to fill in partially written page,
4593 leading to data loss.
4596 Frequency : always, if lconf --abort_recovery used
4598 Description: lconf --abort_recovery fails with 'Operation not supported'
4599 Details : lconf was attempting to abort recovery on the MDT device and not
4602 Severity : enhancement
4604 Description: remove cleanup logs
4605 Details : replace lconf-generated cleanup logs with lustre internal
4606 cleanup routines. Eliminates the need for client-cleanup and
4609 Severity : enhancement
4611 Description: add support for EAs (user and system) on lustre filesystems
4612 Details : it is now possible to store extended attributes in the Lustre
4613 client filesystem, and with the user_xattr mount option it
4614 is possible to allow users to store EAs on their files also
4616 Severity : enhancement
4618 Description: Add possibility (config option) to show minimal available OST free
4620 Details : When compiled with --enable-mindf configure option, statfs(2)
4621 (and so, df) will return least minimal free space available from
4622 all OSTs as amount of free space on FS, instead of summary of
4623 free spaces of all OSTs.
4625 Severity : enhancement
4627 Description: do not expand extent locks acquired on OST-side
4628 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
4629 by server: they are not cached anyway.
4632 Frequency : when mmap is used/binaries executed from Lustre
4634 Description: Unmmap pages before throwing them away from read cache.
4635 Details : llap_shrink cache now attempts to unmap pages before discarding
4636 them (if unmapping failed - do not discard). SLES9 kernel has
4637 extra checks that trigger if this unmapping is not done first.
4642 Description: lconf didn't resolve symlinks before checking to see whether a
4643 given mountpoint was already in use
4646 Frequency : when migrating failover services
4647 Bugzilla : 6395, 9514
4648 Description: When migrating a subset of services from a node (e.g. failback
4649 from a failover service node) the remaining services would
4650 time out and evict clients.
4651 Details : lconf --force (implied by --failover) sets the global obd_timeout
4652 to 5 seconds in order to quickly disconnect, but this caused
4653 other RPCs to time out too quickly. Do not change the global
4654 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
4656 Severity : enhancement
4657 Frequency : if MDS is started with down OST
4658 Bugzilla : 9439,5706
4659 Description: Allow startup/shutdown of an MDS without depending on the
4660 availability of the OSTs.
4661 Details : Asynchronously call mds_lov_synchronize during MDS startup.
4662 Add appropriate locking and lov-osc refcounts for safe
4663 cleaning. Add osc abort_inflight calls in case the
4664 synchronize never started.
4667 Frequency : occasional (Cray XT3 only)
4669 Description: root not authorized to access files in CRAY_PORTALS environment
4670 Details : The client process capabilities were not honoured on the MDS in
4671 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
4672 been accessed by an authorized user then root was able to access
4673 the file on the local client also. The root user capabilities
4674 are now allowed on the MDS, as this environment has secure UID.
4677 Frequency : occasional
4679 Description: ldiskfs "too long searching" message happens too often
4680 Details : A debugging message (otherwise harmless) prints too often on
4681 the OST console. This has been reduced to only happen when
4682 there are fragmentation problems on the filesystem.
4687 Description: Division by zero in statfs when all OSCs are inactive
4688 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
4689 lov_statfs divides by value returned by lov_get_stripecnt().
4693 Bugzilla : 9489, 3273
4694 Description: First write from each client to each OST was only 4kB in size,
4695 to initialize client writeback cache, which caused sub-optimal
4696 RPCs and poor layout on disk for the first writen file.
4697 Details : Clients now request an initial cache grant at (re)connect time
4698 and so that they can start streaming writes to the cache right
4699 away and always do full-sized RPCs if there is enough data.
4700 If the OST is rebooted the client also re-establishes its grant
4701 so that client cached writes will be honoured under the grant.
4706 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
4707 Details : Now I/O RPCs go to different portal number and (presumably) fast
4708 lock requests (and glimses) and other RPCs get their own service
4709 threads pool that should be able to service those RPCs
4712 Severity : enhancement
4714 Description: Ability to exchange lustre version between client and servers and
4715 issue warnings at client side if client is too old. Also for
4716 liblustre clients there is ability to refuse connection of too old
4718 Details : New 'version' field is added to connect data structure that is
4719 filled with version info. That info is later checked by server and
4723 Frequency : rare, liblustre only.
4724 Bugzilla : 9296, 9581
4725 Description: Two simultaneous writes from liblustre at offset within same page
4726 might proceed at the same time overwriting eachother with stale
4728 Details : I/O lock withing llu_file_prwv was released too early, before data
4729 actually was hitting the wire. Extended lock-holding time until
4730 server acknowledges receiving data.
4733 Frequency : extremely rare. Never observed in practice.
4735 Description: avoid generating lustre_handle cookie of 0.
4736 Details : class_handle_hash() generates handle cookies by incrementing
4737 global counter, and can hit 0 occasionaly (this is unlikely, but
4738 not impossible, because initial value of cookie counter is
4739 selected randonly). Value of 0 is used as a sentinel meaning
4740 "unassigned handle" --- avoid it. Also coalesce two critical
4741 sections in this function into one.
4743 Severity : enhancement
4745 Description: allow liblustre clients to delegate truncate locking to OST
4746 Details : To avoid overhead of locking, liblustre client instructs OST to
4747 take extent lock in ost_punch() on client's behalf. New connection
4748 flag is added to handle backward compatibility.
4750 Severity : enhancement
4751 Bugzilla : 4928, 7341, 9758
4752 Description: allow number of OST service threads to be specified
4753 Details : a module parameter allows the number of OST service threads
4754 to be specified via "options ost ost_num_threads={N}" in the
4755 OSS's /etc/modules.conf or /etc/modprobe.conf.
4759 Bugzilla : 6146, 9635, 9895
4760 Description: servers crash with bad pointer in target_handle_connect()
4761 Details : In rare cases when a client is reconnecting it was possible that
4762 the connection request was the last reference for that export.
4763 We would temporarily drop the export reference and get a new
4764 one, but this may have been the last reference and the export
4765 was just destroyed. Get new reference before dropping old one.
4767 Severity : enhancement
4768 Frequency : if client is started with failover MDS
4770 Description: Allow multiple MDS hostnames in the mount command
4771 Details : Try to read the configuration from all specified MDS
4772 hostnames during a client mount in case the "primary"
4775 Severity : enhancement
4777 Description: Stop sending data to evicted clients as soon as possible.
4778 Details : Check if the client we are about to send or are sending data to
4779 was evicted already. (Check is done every second of waiting,
4780 for which l_wait_event interface was extended to allow checking
4781 of exit condition at specified intervals).
4784 Frequency : rare, normally only when NFS exporting is done from client
4786 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
4788 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
4791 Frequency : always, if srand() is called before liblustre initialization
4793 Description: Liblustre uses system PRNG disturbing its usage by user application
4794 Details : Introduce internal to lustre fast and high-quality PRNG for
4795 lustre usage and make liblustre and some other places in generic
4796 lustre code to use it.
4798 Severity : enhancement
4799 Bugzilla : 9477, 9557, 9870
4800 Description: Verify that the MDS configuration logs are updated when xml is
4801 Details : Check if the .xml configuration logs are newer than the config
4802 logs stored on the MDS and report an error if this is the case.
4803 Request --write-conf, or allow starting with --old_conf.
4805 Severity : enhancement
4807 Description: Handle symlinks in the path when checking if Lustre is mounted.
4808 Details : Resolve intermediate symlinks when checking if a client has
4809 mounted a filesystem to avoid duplicate client mounts.
4814 Description: lconf can hit an error exception but still return success.
4815 Details : The lconf command catches the Command error exception at the top
4816 level script context and will exit with the associated exit
4817 status, but doesn't ensure that this exit status is non-zero.
4822 Description: failure of ptlrpc thread startup can cause oops
4823 Details : Starting a ptlrpc service thread can fail if there are a large
4824 number of threads or the server memory is very fragmented.
4825 Handle this without oopsing.
4828 Frequency : always, only if liblustre and non-default acceptor port was used
4830 Description: liblustre cannot connect to servers with non-default acceptor port
4831 Details : tcpnal_set_default_params() was not called and was therefore
4832 ignoring the environment varaible TCPNAL_PORT, as well as other
4833 TCPNAL_ environment variables
4838 Description: two objects could be created on the same OST for a single file
4839 Details : If an OST is down, in some cases it was possible to create two
4840 objects on a single OST for a single file. No problems other
4841 than potential performance impact and spurious error messages.
4845 Bugzilla : 5681, 9562
4846 Description: Client may oops in ll_unhash_aliases
4847 Details : Client dcache may become inconsistent in race condition.
4848 In some cases "getcwd" can fail if the current directory is
4854 Description: Inode refcounting problems in NFS export code
4855 Details : link_raw functions used to call d_instantiate without obtaining
4856 extra inode reference first.
4860 Bugzilla : 9942, 9903
4861 Description: Referencing freed requests leading to crash, memleaks with NFS.
4862 Details : We used to require that call to ll_revalidate_it was always
4863 followed by ll_lookup_it. Also with revalidate_special() it is
4864 possible to call ll_revalidate_it() twice for the same dentry
4865 even if first occurence returned success. This fix changes semantic
4866 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
4867 reference on a request referred from the intent.
4868 ll_intent_release() then releases such a request.
4871 Frequency : rare, normally benchmark loads only
4873 Description: unlinked inodes were kept in memory on the client
4874 Details : If a client is repeatedly creating and unlinking files it
4875 can accumulate a lot of stale inodes in the inode slab cache.
4876 If there is no other client load running this can cause the
4877 client node to run out of memory. Instead flush old inodes
4878 from client cache that have the same inode number as a new inode.
4881 Frequency : SLES9 2.6.5 kernel and long filenames only
4882 Bugzilla : 9969, 10379
4883 Description: utime reports stale NFS file handle
4884 Details : SLES9 uses out-of-dentry names in some cases, which confused
4885 the lustre dentry revalidation. Change it to always use the
4889 Frequency : rare, unless heavy write-truncate concurrency is continuous
4890 Bugzilla : 4180, 6984, 7171, 9963, 9331
4891 Description: OST becomes very slow and/or deadlocked during object unlink
4892 Details : filter_destroy() was holding onto the parent directory lock
4893 while truncating+unlinking objects. For very large objects this
4894 may block other threads for a long time and slow overall OST
4895 responsiveness. It may also be possible to get a lock ordering
4896 deadlock in this case, or run out of journal credits because of
4897 the combined truncate+unlink. Solution is to do object truncate
4898 first in one transaction without parent lock, and then do the
4899 final unlink in a new transaction with the parent lock. This
4900 reduces the lock hold time dramatically.
4903 Frequency : rare, 2.4 kernels only
4905 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
4906 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
4907 before unmounting their filesystem in order to avoid deadlock.
4908 The kernel_locked() function in 2.4 kernels only checks whether
4909 the kernel lock is held, not whether it is this process that is
4910 holding it as 2.6 kernels do.
4915 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
4916 Details : The client ptlrpc code may be trying to reconnect to a down
4917 server before a previous connection attempt has timed out.
4918 Increase the reconnect interval to be longer than the connection
4919 timeout interval to avoid sending duplicate connections to
4923 Frequency : echo_client brw_test command
4925 Description: fix echo_client to work with OST preallocated code
4926 Details : OST preallocation code (5137) didn't take echo_client IO path
4927 into account: echo_client calls filter methods outside of any
4928 OST thread and, hence, there is no per-thread preallocated
4929 pages and buffers to use. Solution: hijack pga pages for IO. As
4930 a byproduct, this avoids unnecessary data copying.
4934 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
4935 Description: Client can oops in mdc_commit_close() after open replay
4936 Details : It was possible for the MDS to return an open request with no
4937 transaction number in mds_finish_transno() if the client was
4938 evicted, but without actually returning an error. Clients
4939 would later try to replay that open and may trip an assertion
4940 Simplify the client close codepath, and always return an error
4941 from the MDS in case the open is not successful.
4944 Frequency : rare, 2.6 OSTs only
4946 Description: OST may deadlock under high load on fragmented files
4947 Details : If there was a heavy load and highly-fragmented OST filesystems
4948 it was possible to have all the OST threads deadlock waiting on
4949 allocation of biovecs, because the biovecs were not released
4950 until the entire RPC IO was completed. Instead, release biovecs
4951 as soon as they are complete to ensure forward IO progress.
4953 Severity : enhancement
4955 Description: Support for specifying external journal device at mount
4956 Details : If an OST or MDS device is formatted with an external journal
4957 device, this device major/minor is stored in the ext3 superblock
4958 and may not be valid for failover. Allow detecting and
4959 specifying the external journal at mount time.
4964 Description: Mounting an MDS with pending unlinked files may cause oops
4965 Details : target_finish_recovery() calls mds_postrecov() which returned
4966 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
4967 considers this an error and immediately begins cleaning up the
4968 lov, just after starting the mds_lov process
4970 Severity : enhancement
4972 Description: Implement 'lfs df' to report actual free space on per-OST basis
4973 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
4974 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
4975 usage of objects; '-h' to report in human readable format.
4977 ------------------------------------------------------------------------------
4979 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
4986 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
4987 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
4988 at startup during orphan unlinking. Instead, skip buddy update
4989 before setup, as it will be regenerated after recovery is complete.
4994 Description: If an OST is inactive, its locks might reference stale inodes.
4995 Details : lov_change_cbdata() must iterate over all namespaces, even if
4996 they are inactive to clear inode references from the lock.
4998 Severity : enhancement
4999 Frequency : occasional, if non-standard max_dirty_mb used
5001 Description: Client will block write RPCs if not enough grant
5002 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
5003 then the client will block writes while waiting for another RPC
5004 to complete instead of consuming its dirty limit. With change
5005 we get improved performance when max_dirty_mb is small.
5007 Severity : enhancement
5008 Bugzilla : 3389, 6253
5009 Description: Add support for supplementary groups on the MDS.
5010 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
5011 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
5012 lookups for user supplementary groups into a cache.
5016 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
5017 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
5018 for liblustre O_EXCL is enforced by MDS.
5022 Description: readdir never returns NULL in liblustre.
5023 Details : Corrected llu_iop_getdirentries logic, to return offset of next
5024 dentry in struct dirent.
5028 Frequency : liblustre only, depends on application IO pattern
5029 Description: liblustre clients evicted if not contacting servers
5030 Details : Don't put liblustre clients into the ping_evictor list, so
5031 they will not be evicted by the pinger ever.
5033 Severity : enhancement
5035 Description: Add ability to evict clients by NID from MDS.
5036 Details : By echoing "nid:$NID" string into
5037 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
5038 $NID would be instantly evicted from this MDS and from all active
5039 OSTs connected to it.
5043 Description: Do not query file size twice, somewhat slowing stat(2) calls.
5044 Details : lookup_it_finish() used to query file size from OSTs that was not
5049 Description: service threads change working directory to that of init
5050 Details : Starting lustre service threads may pin the working directory
5051 of the parent thread, making that filesystem busy. Threads
5052 now change to the working directory of init to avoid this.
5056 Frequency : during shutdown only
5057 Description: shutdown with a failed MDS or OST can cause unmount to hang
5058 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
5063 Frequency : 2.6 only
5064 Description: chmod/chown may include an extra supplementary group
5065 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
5066 supplementary group and if none is specified this is used.
5069 Bugzilla : 5479 (6816)
5071 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
5072 Details : If lookup is in progress on a file that is unlinked we might try
5073 to revalidate the inode and fail in revalidate after lookup is
5074 complete and ll_file_open() enqueues the open again but
5075 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
5078 Frequency : always, if lconf --abort_recovery used
5080 Description: lconf --abort_recovery fails with 'Operation not supported'
5081 Details : lconf was attempting to abort recovery on the MDT device and not
5084 ------------------------------------------------------------------------------
5086 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
5091 Frequency : rare (only unsupported configurations with a node running as an
5093 Bugzilla : 6514, 5137
5094 Description: Mounting a Lustre file system on a node running as an OST could
5096 Details : OSTs now preallocates memory needed to write out data at
5097 startup, instead of when needed, to avoid having to
5098 allocate memory in possibly low memory situations.
5099 Specifically, if the file system is mounted on on OST,
5100 memory pressure could force it to try to write out data,
5101 which it needed to allocate memory to do. Due to the low
5102 memory, it would be unable to do so and the node would
5103 become unresponsive.
5105 Severity : enhancement
5107 Description: Addition of lconf --service command line option
5108 Details : lconf now accepts a '--service <arg>' option, which is
5109 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
5111 Severity : enhancement
5113 Description: Failover mode is now the default for OSTs.
5114 Details : By default, OSTs will now run in failover mode. To return to
5115 the old behaviour, add '--failout' to the lmc line for OSTs.
5117 Severity : enhancement
5119 Description: Health checks are now provided for MDS and OSTs
5120 Details : Additional detailed health check information on MSD and OSTs
5121 is now provided through the procfs health_check value.
5124 Frequency : occasional, depends on IO load
5126 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
5127 after numerous create/delete cycles
5128 Details : The ext3 inode allocation policy would not allocate new inodes
5129 very well on the OSTs because there are no new directories
5130 being created. Instead we look for groups with free space if
5131 the parent directories are nearly full.
5136 Description: Network or server problems during mount may cause partially
5137 mounted clients instead of returning an error.
5138 Details : The config llog parsing code may overwrite the error return
5139 code during mount error handling, returning success instead
5145 Description: MDS can fail to allocate large reply buffers
5146 Details : After long uptimes the MDS can fail to allocate large reply
5147 buffers (e.g. zconf client mount config records) due to memory
5148 fragmentation or consumption by the buffer cache. Preallocate
5149 some large reply buffers so that these replies can be sent even
5150 under memory pressure.
5154 Frequency : rare (liblustre)
5155 Description: fsx running with liblustre complained that using truncate() to
5156 extend the file doesn't work. This patch corrects that issue.
5157 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
5158 ATTR_SIZE and lsm use in llu_setattr_raw.
5162 Frequency : rare, only 2.6 kernels
5163 Description: Unusual file access patterns on the MDS may result in inode
5164 data being lost in very rare circumstances.
5165 Details : Bad interaction between the ea-in-inode patch and the "no-read"
5166 code in the 2.6 kernel caused the inode and/or EA data not to
5167 be read from disk, causing single-file corruption.
5171 Frequency : rare, only 2.6 filesystems using extents
5172 Description: Heavy concurrent write and delete load may cause data corruption.
5173 Details : It was possible under high-load situations to have an extent
5174 metadata block in the block device cache from a just-unlinked
5175 file overwrite a newly-allocated data block. We now unmap any
5176 metadata buffers that alias just-allocated data blocks.
5180 Frequency : filesystems with default stripe_count larger than 77
5181 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
5182 Details : lconf specifies an inode size of 4096 bytes when the default
5183 stripe_count is larger than 77. This conflicts with the default
5184 inode density of 1 per 4096 bytes. Allocate smaller inodes in
5185 this case to avoid pinning too much memory for large EAs.
5187 ------------------------------------------------------------------------------
5189 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
5194 Frequency : rare (extremely heavy IO load with hundreds of clients)
5196 Description: Client is evicted, gets IO error writing to file
5197 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
5198 caused clients to be evicted for AST timeouts. The fixes in
5199 bug 5192 mean we no longer need to have such short AST timeouts
5200 so ldlm_timeout has been increased.
5203 Frequency : occasional during --force or --failover shutdown under load
5204 Bugzilla : 5949, 4834
5205 Description: Server oops/LBUG if stopped with --force or --failover under load
5206 Details : a collection of import/export refcount and cleanup ordering
5207 issues fixed for safer force cleanup
5210 Frequency : only filesystems larger than 120 OSTs
5211 Bugzilla : 5990, 6223
5212 Description: lfs getstripe would oops on a very large filesystem
5213 Details : lov_getconfig used kfree on vmalloc'd memory
5216 Frequency : only filesystems exporting via NFS to Solaris 10 clients
5217 Bugzilla : 6242, 6243
5218 Description: reading from files that had been truncated to a non-zero size
5219 but never opened returned no data
5220 Details : ll_file_read() reads zeros from no-object files to EOF
5225 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
5226 Details : There's little harm in aborting MDS/OSS recovery and letting it
5227 try again, so I removed the LASSERT and return an error instead.
5229 Severity : enhancement
5231 Description: New debugging infrastructure for tracking down data corruption
5232 Details : The I/O checksum code was replaced to: (a) control it at runtime,
5233 (b) cover more of the client-side code path, and (c) try to narrow
5234 down where problems occurred
5238 Bugzilla : 3819, 4364, 4397, 6313
5239 Description: Racing close and eviction MDS could cause assertion in mds_close
5240 Details : It was possible to get multiple mfd references during close and
5241 client eviction, leading to one thread referencing a freed mfd.
5243 Severity: : enhancement
5244 Bugzilla : 3262, 6359
5245 Description: Attempts to reconnect to servers are now more aggressive.
5246 Details : This builds on the enhanced upcall-less recovery that was added
5247 in 1.4.2. When trying to reconnect to servers, clients will
5248 now try each server in the failover group every 10 seconds. By
5249 default, clients would previously try one server every 25 seconds.
5254 Description: After recovery, certain operations trigger a failed
5255 assertion on a client.
5256 Details : Failing over an mds, using lconf -d --failover, while a
5257 client was doing a readdir() call would cause the client to
5258 LBUG after recovery completed and the readdir() was resent.
5260 Severity : enhancement
5262 Description: Default groups are now added by lconf
5263 Details : You can now run lconf --group <servicename> without having to
5264 manually add groups with lmc.
5267 Frequency : occasional
5269 Description: Nodes with an elan id of 0 trigger a failed assertion
5272 Frequency : always when accessing e.g. tty/console device nodes
5274 Description: tty and some other devices nodes cannot be used on lustre
5275 Details : file's private_data field is used by device data and lustre
5276 values in there got lost. New field was added to struct file to
5277 store fs-specific private data.
5280 Frequency : when exporting Lustre via NFS
5282 Description: NFSD failed occasionally when looking up a path component
5283 Details : NFSD is looking up ".." which was broken in ext3 directories
5284 that had grown large enough to become hashed.
5287 Frequency : Clusters with multiple interfaces not on the same subnet
5289 Description: Nodes will repeatedly try to reconnect to an interface which it
5290 cannot reach and report an error to the log.
5291 Details : Extra peer list entries will be created by lconf with some peers
5292 unreachable. lconf now validates the peer before adding it.
5295 Frequency : Only if a default stripe is set on the filesystem root.
5297 Description: Setting a default stripe on the filesystem root prevented the
5298 filesystem from being remounted.
5299 Details : The client was sending extra request flags in the root getattr
5300 request and did not allocate a reply buffer for the dir EA.
5303 Frequency : occasional, higher if lots of files are accessed by one client
5304 Bugzilla : 6159, 6097
5305 Description: Client trips assertion regarding lsm mismatch/magic
5306 Details : While revalidating inodes the VFS looks up inodes with ifind()
5307 and in rare cases can find an inode that is being freed.
5308 The ll_test_inode() code will free the lsm during ifind()
5309 when it finds an existing inode and then the VFS later attaches
5310 this free lsm to a new inode.
5314 Bugzilla : 6422, 7030
5315 Description: MDS deadlock between mkdir and client eviction
5316 Details : Creating a new file via mkdir or mknod (starting a transaction
5317 and getting the ns lock) can deadlock with client eviction
5318 (gets ns lock and trying to finish a synchronous transaction).
5321 Frequency : occasional
5322 Description: While starting a server, the fsfilt_ext3 module could not be
5324 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
5325 kernels. Previously, lconf would still use the ext3 name
5326 when trying to load modules. Now, it will correctly use
5327 ext3 on 2.4 and ldiskfs on 2.6.
5329 Severity : enhancement
5330 Description: The default stripe count has been changed to 1
5331 Details : The interpretation of the default stripe count (0, to lfs
5332 or lmc) has been changed to mean striping across a single
5333 OST, rather than all available. For general usage we have
5334 found a stripe count of 1 or 2 works best.
5336 Severity : enhancement
5337 Description: Add support for compiling against Cray portals.
5338 Details : Conditional compiling for some areas that are different
5342 Frequency : occasional
5343 Bugzilla : 6409, 6834
5344 Description: Creating files with an explicit stripe count may lead to
5345 a failed assertion on the MDS
5346 Details : If some OSTs are full or unavailable, creating files may
5347 trigger a failed assertion on the MDS. Now, Lustre will
5348 try to use other servers or return an error to the
5352 Frequency : occasional
5354 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
5355 caused lock timeout during readahead (since 1.4.2).
5356 Details : Processes doing readahead might match a lock that hasn't been
5357 granted yet if there are overlapping and conflicting lock
5358 requests. The readahead process waits on ungranted lock
5359 (original lock is CBPENDING), while OST waits for that process
5360 to cancel CBPENDING read lock and eventually evicts client.
5362 Severity : enhancement
5364 Description: Initial enabling of flock support for clients
5365 Details : Implements fcntl advisory locking and file status functions.
5366 This feature is provided as an optional mount flag (default
5367 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
5368 locking are implemented yet, and those that are are not guaranteed
5369 to be completely correct in production environments.
5370 mount -t lustre -o [flock|noflock] ...
5373 Frequency : occasional
5375 Description: OSTs running 2.4 kernels but with extents enabled might trip an
5376 assertion in the ext3 JBD (journaling) layer.
5377 Details : The b_committed_data struct is protected by the big kernel lock
5378 in 2.4 kernels, serializing journal_commit_transaction() and
5379 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
5380 access to this struct. In 2.6 kernels there is finer grained
5381 locking to improve SMP performance of the JBD layer.
5385 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
5390 Description: merge and cleanup kernel patches.
5393 -----------------------------------------------------------------------------
5395 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
5397 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
5398 versions of Lustre. Please update all servers and clients to
5399 version 1.4.2 or later at the same time. You must also run
5400 "lconf --write-conf {config}.xml" on the MDS while it is stopped
5401 to update the configuration logs.
5403 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
5404 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
5405 - swab llog records of type '0' so we get proper header size/idx (5861)
5406 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
5407 - fix rename of one directory over another leaking an inode (5953)
5408 - avoid SetPageDirty on 2.6 (5981)
5409 - don't re-add just-being-destroyed locks to the waiting list (5653)
5410 - when creating new directories, inherit the parent's custom
5411 striping settings if present parent (3048)
5412 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
5413 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
5414 - don't hold client locks on temporary worklist from l_lru (5666)
5415 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
5416 - automatically evict dead clients (5921)
5417 - Update file size properly in create+truncate+fstat case (6196)
5418 - Do not unhash mountpoint dentries, do not allow removal of
5420 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
5421 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
5422 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
5423 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
5424 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
5425 - Correctly handle reads of files with no objects (6243)
5426 - lctl recover will also mark a device active if deactivate used (5933)
5428 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
5429 - allow --write-conf on an MDS with different nettype than client (5619)
5430 - don't write config llogs to MDS for mounts not from that MDS (5617)
5431 - lconf should create multiple TCP connections from a client (5201)
5432 - init scripts are now turned off by default; run chkconfig --on
5433 lustre and chkconfig --on lustrefs to use them
5434 - upcalls are no longer needed for clients to recover to failover
5436 - add --abort-recovery option to lconf to abort recovery on device
5438 - add support for an arbitrary number of OSTs (3026)
5439 - Quota support protocol changes.
5440 - forward compatibility changes to wire structs (6007)
5441 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5442 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5443 - improved "lustre status" script
5444 - initialize blocksize for non-regular files (6062)
5445 - added --disable-server and --disable-client configure options (5782)
5446 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5447 - Vanilla 2.4.29 support
5448 - increase maximum number of obd devices to 520 (6242)
5449 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5450 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5451 SLES 9 2.6.5 kernels
5452 - limit stripes per file to 160 (the maximum EA size) (6093)
5454 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5457 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5458 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5459 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5460 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5461 - small ext3 extents cleanups and fixes (5733)
5462 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5463 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5464 - avoid needless client->OST connect, fix handle mismatch (5317)
5465 - fix DLM error path that led to out-of-sync client, long delays (5779)
5466 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5467 - fix several locking issues related to i_size (5492,5624,5654,5672)
5468 - don't move pending lock onto export if it is already evicted (5683)
5469 - fix kernel oops when creating .foo in unlinked directory (5548)
5470 - fix deadlock in obdfilter statistics vs. object create (5811)
5471 - use time_{before,after} to avoid timer jiffies wrap (5882)
5472 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5473 - Do not leak request if server was not able to process it (5154)
5474 - If mds_open unable to find parent dir, make that negative lookup(5154)
5475 - don't create new directories with extent-mapping (5909, 5936)
5477 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5478 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5479 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5480 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5481 - support for RHEL4 kernel on i686 (5773)
5482 - provide error messages when incompatible logs are encountered (5898)
5484 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5485 * version 1.4.0.10 (1.4.1 release candidate 1)
5487 - don't keep a lock reference when lock is not granted (4238)
5488 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5489 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5490 - drop import inflight refcount on signal_completed_replay error (5255)
5491 - unlock page after async write error during send (3677)
5492 - handle missing objects in filter_preprw_read properly (5265)
5493 - no transno return for symlink open, don't save no-trasno open (3440)
5494 - don't try to complete elan receive that already failed (4012)
5495 - free RPC server reply state on error (5406)
5496 - clean up thread from ptlrpc_start_thread() on error (5160)
5497 - readahead could read extra page into cache that wasn't ejected (5388)
5498 - prevent races in class_attach/setup/cleanup/detach (5260)
5499 - don't dereference de->d_inode after l_dput of de (5458)
5500 - use "int" for stripe value returned from lock_to_stripe (5544)
5501 - mballoc allocation and error-checking fixes in 2.6 (5504)
5502 - block device patches to fix I/O request sizes in 2.6 (5482)
5503 - look up hostnames for IB nals (5602)
5504 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5505 - don't start multiple acceptors for the same port (5277)
5506 - fix incorrect LASSERT in mds_getattr_name (5635)
5507 - export a proc file for general "ping" checking (5628)
5508 - fix "lfs check" to not block when the MDS is down (5628)
5510 - service request history (4965)
5511 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5512 - create an "evict_client" /proc entry on OSTs, like the MDS has
5513 - fix mount usage message, return errors per mount(8) (5168)
5514 - change grep [] to grep "[]" in tests so they work in more UMLs
5515 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5516 - remove ancient LOV_MAGIC_V0 EA support (5047)
5517 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5518 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5519 - allow readahead allocations to fail when low on memory (5383)
5520 - mmap locking landed again, after considerable improvement (2828)
5521 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
5523 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
5526 - send OST transaction number in read/write reply to free req (4966)
5527 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5528 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5529 - the watchdog thread now runs as interruptible (5246)
5530 - flock/lockf fixes (but it's still disabled, pending 5135)
5531 - don't use EXT3 constants in llite code (5094)
5532 - memory shortage at startup could cause assertion (5176)
5534 - reorganization of lov code
5535 - single portals codebase
5537 - add extents/mballoc support (5025)
5538 - direct I/O reads in the obdfilter (4048)
5539 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
5541 tbd Cluster File Systems, Inc. <info@clusterfs.com>
5544 - send OST transaction number in read/write reply to free req (4966)
5545 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5546 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5547 - the watchdog thread now runs as interruptible (5246)
5548 - handle missing objects in filter_preprw_read properly (5265)
5549 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5550 - drop import inflight refcount on signal_completed_replay error (5255)
5551 - unlock page after async write error during send (3677)
5552 - return original error code on reconstructed replies (3761)
5553 - no transno return for symlink open, don't save no-trasno open (3440)
5555 - add pid to ldlm debugging output (4922)
5556 - bump the watchdog timeouts -- we can't handle 30sec yet
5557 - extra debugging for orphan dentry/inode bug (5259)
5559 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
5562 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
5563 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
5564 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
5565 - fix NULL dereference in /proc/sys/portals/routes (4827)
5566 - allow failed mdc_close() operations to be interrupted (4561)
5567 - stop precreate on OST before MDS would time out on it (4778)
5568 - don't send partial-page writes before EOF from client (4410)
5569 - discard client grant for sub-page writes on large-page clients (4520)
5570 - don't free dentries not owned by NFS code, check generation (4806)
5571 - fix lsm leak if mds_create_objects() fails (4801)
5572 - limit debug_daemon file size, always print CERROR messages (4789)
5573 - use transno after validating reply (3892)
5574 - process timed out requests if import state changes (3754)
5575 - update mtime on OST during writes, return in glimpse (4829)
5576 - add mkfsoptions to LDAP (4679)
5577 - use ->max_readahead method instead of zapping global ra (5039)
5578 - don't interrupt __l_wait_event() during strace
5580 - add software watchdogs to catch hung threads quickly (4941)
5581 - make lustrefs init script start after nfs is mounted
5582 - fix CWARN/ERROR duplication (4930)
5583 - return async write errors to application if possible (2248)
5584 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
5585 - print NAL number in %x format (4645)
5586 - update barely-supported suse-2.4.21-171 series (4842)
5587 - support for sles 9 %post scripts
5588 - support for building 2.6 kernel-source packages
5589 - support for sles km_* packages
5591 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
5594 - ignore -ENOENT errors in osc_destroy (3639)
5595 - notify osc create thread that OSC is being cleaned up (4600)
5596 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
5597 - reconstruct ost_handle() like mds_handle() (4657)
5598 - create a new thread to do import eviction to avoid deadlock (3969)
5599 - let lconf resolve symlinked-to devices (4629)
5600 - don't unlink "objects" from directory with default EA (4554)
5601 - hold socknal file ref over connect in case target is down (4394)
5602 - allow more than 32000 subdirectories in a single directory (3244)
5603 - fix blocks count for O_DIRECT writes (3751)
5604 - OST returns ENOSPC from object create when no space left (4539)
5605 - don't send truncate RPC if file size isn't changing (4410)
5606 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
5607 - bind to privileged port in socknal and tcpnal (3689)
5609 - rate limit CERROR/CWARN console message to avoid overload (4519)
5610 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
5611 - basic mmap support (3918)
5612 - kernel patch series update from b1_4 (4711)
5614 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
5617 - avoid crash during MDS cleanup with OST shut down (2775)
5618 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
5619 - don't use bad inodes on the MDS (3744)
5620 - dynamic object preallocation to improve recovery speed (4236)
5621 - don't hold spinlock over lock dumping or change debug flags (4401)
5622 - don't zero obd_dev when it is force cleaned (3651)
5623 - print grants to console if they go negative (4431)
5624 - "lctl deactivate" will stop automatic recovery attempts (3406)
5625 - look for existing locks in ldlm_handle_enqueue() (3764)
5626 - don't resolve lock handle twice in recovery avoiding race (4401)
5627 - revalidate should check working dir is a directory (4134)
5629 - don't always mark "slow" obdfilter messages as errors (4418)
5631 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
5634 - don't close LustreDB during write_conf until it is done (3860)
5635 - fix typo in lconf for_each_profile (3821)
5636 - allow dumping logs from multiple threads at one time (3820)
5637 - don't allow multiple threads in OSC recovery (3812)
5638 - fix debug_size parameters (3864)
5639 - fix mds_postrecov to initialize import for llog ctxt (3121)
5640 - replace config semaphore with spinlock (3306)
5641 - be sure to send a reply for a CANCEL rpc with bad export (3863)
5642 - don't allow enqueue to complete on a destroyed export (3822)
5643 - down write_lock before checking llog header bitmap (3825)
5644 - recover from lock replay timeout (3764)
5645 - up llog sem before sending rpc (3652)
5646 - reduce ns lock hold times when setting kms (3267)
5647 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
5648 - fix NULL deref and obd_dev leak on setup error (3312)
5649 - replace some LBUG about llog ops with error handling (3841)
5650 - don't match INVALID dentries from d_lookup and spin (3784)
5651 - hold dcache_lock while marking dentries INVALID and hashing (4255)
5652 - fix invalid assertion in ptlrpc_set_wait (3880)
5654 - add libwrap support for the TCP acceptor (3996)
5655 - add /proc/sys/portals/routes for non-root route listing (3994)
5656 - allow setting MDS UUID in .xml (2580)
5657 - print the stack of a process that LBUGs (4228)
5659 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
5662 - don't cleanup request in ll_file_open() on failed MDS open (3430)
5663 - make sure to unset replay flag from failed open requests (3440)
5664 - if default stripe count is 0, use OST count for inode size (3636)
5665 - update parent mtime/ctime on client for create/unlink (2611)
5666 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
5667 - free recovery state on server during a forced cleanup (3571)
5668 - unregister_reply for resent reqs (3063)
5669 - loop back devices mounting and status check on 2.6 (3563)
5670 - fix resource-creation race that can provoke i_size == 0 (3513)
5671 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
5672 - more debugging for page-accounting assertion (3746)
5673 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
5674 - avoid deadlock after precreation failure (3758)
5675 - fix race and lock order deadlock in orphan handling (3450, 3750)
5676 - add validity checks when grabbing inodes from l_ast_data (3599)
5678 - add /proc/.../recovery_status to obdfilter (3428)
5679 - lightweight CDEBUG infrastructure, debug daemon (3668)
5680 - change default OSC RPC parameters to be better on small clusters
5681 - turn off OST read cache for files smaller than 32MB
5682 - install man pages and include them in rpms (3100)
5683 - add new init script for (un)mounting lustre filesystems (2593)
5684 - run chkconfig in %post for init scripts (3701)
5685 - drop scimac NAL (unmaintained)
5687 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
5690 - clean kiobufs before and after use (3485)
5691 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
5692 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
5693 - change init script to fail more clearly if not run as root (1528)
5694 - allow clients to reconnect during replay (1742)
5695 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
5696 - don't do DNS lookups on NIDs too small for IP addresses (3442)
5697 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
5698 - fix cond_resched (3554)
5699 - only evict unfinished clients after recovery (3515)
5700 - allow bulk resend, prevent data loss (3570)
5701 - dynamic ptlrpc request buffer allocation (2102)
5702 - don't allow unlinking open directory if it isn't empty (2904)
5703 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
5704 - remove extraneous obd dereference causing LASSERT failure (3334)
5705 - don't use get_cycles() when creating temp. files on the mds (3156)
5706 - hold i_sem when setting i_size in ll_extent_lock() (3564)
5707 - handle EEXIST for set-stripe, set proper directory name (3336)
5709 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
5710 - fix ksocknal_fmb_callback() error messages (2918)
5712 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
5715 - don't copy lvb into (possibly NULL) reply on error (2983)
5716 - don't deref dentry after dput, don't free lvb on error (2922)
5717 - use the kms to determine writeback rpc length (2947)
5718 - increment oti_logcookies when osc is inactive (2948)
5719 - update client's i_blocks count via lvb messages (2543)
5720 - handle intent open/close of special files properly (1557)
5721 - mount MDS with errors=remount-ro, like obdfilter (2009)
5722 - initialize lock handle to avoid ASSERT on error cleanup (3057)
5723 - don't use cancelling-locks' kms values (2947)
5724 - use highest lock extent for kms, not last one (2925)
5725 - don't dereference ERR_PTR() dentry in error handling path (3107)
5726 - fix thread race in portals_debug_dumplog() (3122)
5727 - create lprocfs device entries at setup instead of at attach (1519)
5728 - common AST error handler, don't evict client on completion race (3145)
5729 - zero nameidata in detach_mnt in 2.6 (3118)
5730 - verify d_inode after revalidate_special is valid in 2.6 (3116)
5731 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
5732 - initialize RPC timeout timer earlier for 2.6 (3219)
5733 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
5734 - print nal/nid for unknown nid (3258)
5735 - additional checks for oscc recovery before doing precreate (3284)
5736 - fix ll_extent_lock() error return code for 64-bit systems (3043)
5737 - don't crash in mdc_close for bad permissions on open (3285)
5738 - zero i_rdev for non-device files (3147)
5739 - clear page->private before handing to FS, better assertion (3119)
5740 - tune the read pipeline (3236)
5741 - fix incorrect decref of invalidated dentry (2350)
5742 - provide read-ahead stats and refine rpc in flight stats (3328)
5743 - don't hold journal transaction open across create RPC (3313)
5744 - update atime on MDS at close time (3265)
5745 - close LDAP connection when recovering to avoid server load (3315)
5746 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
5747 - don't leak open file on MDS after open resend (3325)
5748 - serialize filter_precreate and filter_destroy_precreated (3329)
5749 - loop device shouldn't call sync_dev() for nul device (3092)
5750 - clear page cache after eviction (2766)
5751 - resynchronize MDS->OST in background (2824)
5752 - refuse to mount the same filesystem twice on same mountpoint (3394)
5753 - allow llmount to create routes for mounting behind routers (3320)
5754 - push lock cancellation to blocking thread for glimpse ASTs (3409)
5755 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
5756 - fix rare problem with rename on htree directories (3417)
5758 - allow default OST striping configuration per directory (1414)
5759 - fix compilation for qswnal for 2.6 kernels (3125)
5760 - increase maximum number of MDS request buffers for large systems
5761 - change liblustreapi to be useful for external progs like lfsck (3098)
5762 - increase local configuration timeout for slow disks (3353)
5763 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
5765 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5768 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
5769 - don't overwrite extent policy data in reply if lock was blocked (2901)
5770 - drop filter export grants atomically with removal from device (2663)
5771 - del obd_self_export from work_list in class_disconnect_exports (2908)
5772 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
5773 - swab reply message in mdc_close, other PPC fixes (2464)
5774 - fix destroying of named logs (2325)
5775 - overwrite old logs when running lconf --write_conf (2264)
5776 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
5777 - fix race in target_handle_connect (2898)
5778 - mds_reint_create() should take same inode create lock (2926)
5779 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
5780 - don't close files for self_export to avoid uninitialized obd (2936)
5781 - allow MDS with the same name as client node (2939)
5782 - hold dentry reference for closed log files for unlink (2325)
5783 - reserve space for all logs during transactions (2059)
5784 - don't evict page beyond end of stripe extent (2925)
5785 - don't oops on a deleted current working directory (2399)
5786 - handle hard links to targets without a parent properly (2517)
5787 - don't dereference NULL lock when racing during eviction (2867)
5788 - don't grow lock extents when lots of conflicting locks (2919)
5790 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
5793 - account for cache space usage on clients to avoid data loss (974)
5794 - lfsck support in lustre kernel code (2349)
5795 - reduce journal credits needed for BRW writes (2370)
5796 - orphan handling to avoid losing space on client/server crashes
5797 - ptlrpcd can be blocked, stopping ALL progress (2477)
5798 - use lock value blocks to assist in proper KMS, faster stat (1021)
5799 - takes i_sem instead of DLM locks internally on obdfilter (2720)
5800 - recovery for initial connections (2355)
5801 - fixes for mds_cleanup_orphans (1934)
5802 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
5803 - block all file creations until orphan recovery completes (1901)
5804 - client remove rq_connection from request struct (2423)
5805 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
5806 - recovery timer race (2670)
5807 - mdc_close recovey bug (2532)
5808 - ptlrpc cleanup bug (2710)
5809 - mds timeout on local locks (2588)
5810 - namespace lock held during RPCs (2431)
5811 - handle interrupted sync write properly (2503)
5812 - don't try to handle a message that hasn't been replied to (2699)
5813 - client assert failure during cleanup after abort recovery (2701)
5814 - leak mdc device after failed mount (2712)
5815 - ptlrpc_check_set allows timedout requests to complete (2714)
5816 - wait for inflight reqs when ptlrpcd finishes (2710)
5817 - make sure unregistered services are removed from the srv_list
5818 - reset bulk XID's when resending them (caught by 1138 test)
5819 - unregister_bulk after timeout
5820 - fix lconf error (2694)
5821 - handle write after unfinished setstripe, stripe-only getstripe (2388)
5822 - readahead locks pages, leaves pending causing memory pressure (2673)
5823 - increase OST request buffers to 4096 on large machines (2729)
5824 - fix up permission of existing directories in simple_mkdir (2661)
5825 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
5826 - don't assign transno to errored transactions (2742)
5827 - don't delete objects on OST if given a bogus objid from MDS (2751)
5828 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
5829 - if rq_no_resend, then timeout request after recovery (2432)
5830 - fix MDS llog_logid record size, 64-bit array alignment (2733)
5831 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
5832 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
5833 - MDS recovery shouldn't delete objects that it has given out (2730)
5834 - if enqueue arrives after completion, don't clobber LVB (2819)
5835 - don't unlock pages twice when trigger_group_io returns error (2814)
5836 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
5837 - don't write pages to disk if there was an error (1450)
5838 - don't ping imports that have recovery disabled (2676)
5839 - take buffered bytes into account when balancing socknal conn (2817)
5840 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
5841 - reconnect unlink llog connection after MDS reconnects to OST (2816)
5842 - remove little-endian swabbing of llog records (1987)
5843 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
5844 - retry reposting request buffers if they fail (1191)
5845 - grow extent at grant time to avoid granting a revoked lock (2809)
5846 - lock revoke doesn't evict page if covered by a second lock (2765)
5847 - disable VM readahead to avoid reading outside lock extents (2805)
5849 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
5850 - updated kernel patches for hp-2.4.20 kernel (2681)
5852 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
5855 - fix truncated write corruption (2366)
5856 - fix for failed assertion in iopen_connect_dentry (1792,2517)
5858 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
5859 - ensure len(uuid) < 37 in lmc (1171)
5860 - fix ia64 OOPS in llog_test (2255)
5861 - zero end of page at obdfilter for partial page writes (2648)
5862 - don't leave stale dentries around after renames (bug 2428)
5863 - fix timeouts when evicting a client with a single lock held (2642)
5864 - set deadline for the initial HELLO message to drain (2634)
5865 - print out dotted-quad IP addresses in the socknal (2302)
5867 - additional debugging for MDS client eviction problem (2443)
5868 - fix mkfsoptions support for osts (2603, 2604)
5870 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
5873 - add series for the vanilla 2.6.0 kernel
5874 - add series for the vanilla 2.4.24 kernel
5875 - add series for a cray x86/64 UL kernel drop
5876 - fix xattr patches for the vanilla 2.4.19 series
5878 - generate true UUIDs in lmc (1171)
5879 - have portals stack dumping break in UML (2466)
5880 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
5881 - allocate with _NOFS in ldlm to avoid deadlock (1933)
5882 - wake callback waiting threads on client eviction (2460)
5883 - Add --ptldebug and --subsystem to lmc (1719)
5884 - update assertion to allow safe interrupt allocation
5885 - set rq_no_resend for cancel requests (2432)
5886 - recalculate ptlrpcd timeout after resend (2494)
5887 - call vfs_rmdir when removing pending directories (2368)
5888 - fix renaming a file to itself (2429)
5889 - lmc creates a default one-stripe lov (2454)
5890 - expand procfs space to handle large clusters (2326)
5891 - increase UML stack to avoid overflow
5892 - update lconf's list of debug and subsystem masks
5893 - fix lfs find --obd (2510)
5894 - /proc tunable for disabling filter read caching (2591)
5895 - stop rpm packages from altering slapd.conf (2301)
5896 - disable nagle in the socknal under 0conf (2578)
5897 - choose mds inode size based on stripe count (2572)
5898 - fix kernel-source rpm problems (2516)
5900 - add --disable-doc to avoid pdf generation (2421)
5901 - update documentation, tests, type-os, comments
5902 - avoid format warnings on ia64
5903 - remove the TOE NAL
5904 - tiny code cleanups by removing unused fields
5906 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
5909 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
5910 - remove the most blatant lies from BUILDING file (2371)
5911 - change default debug level to reasonable production setting
5912 - reduce client side cache size to reduce cache flush time
5913 - reduce max RPCs in flight to avoid unnecessary file fragmentation
5914 - make TCP zerocopy and pinger support enabled by default (2476)
5915 - sync writes completed after process exits caused crashes (2319)
5916 - maintain correct mount count on the MDS (2356)
5917 - backout 1557, because 2316 wasn't really fixed
5918 - better file I/O statistics gathering in /proc
5919 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
5920 - another kernel patch to fix zero-copy TCP function export
5921 - don't take duplicate lock when processing re-sent getattr (2420)
5922 - lctl uses obd_self_export instead of creating new conn (2353)
5923 - MDS/OST recovery case which requires object creation asserted (2425)
5924 - move lfs from /usr/sbin to /usr/bin in packages
5925 - fix race between mds_client_add and mds_client_free (2417)
5926 - use kmalloc instead of slabs in portals (2430)
5927 - don't create duplicate records when a failover MDS is present (2442)
5928 - remove unnecessary mount age check (2332)
5929 - don't remove directory inodes from locks prematurely (2451)
5930 - don't break if MDS service name is the same as hostname (2103)
5931 - fix races in client write RPC generation when cache full (2482)
5933 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
5936 - remove now-unused request->rq_obd (278)
5937 - if an allocation fails, print out how much memory we've used (1933)
5938 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
5939 - add the "configurable stack size" patch to most series files (1256)
5940 - ability to write large log records, for 100+ OST configs (2306)
5941 - fix NULL deref when filter_prep fails (2314)
5942 - fix operator precedence error in filter_sync
5943 - dynamic allocation of socknal TX descriptors (2315)
5944 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
5945 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
5946 - ensure that CWARN messages in Portals always get to the syslog
5947 - __init/__exit are not for prototype decls (ldlm_init/exit)
5948 - x86-64 compile warning fixes
5949 - fix gateway LMC keyword conflict (2318)
5950 - fix MDS lock inversions in getattr/reint paths (1844)
5951 - fix a rare lock re-ordering bug, which caused deadlock (2322)
5952 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
5953 - DLM race condition prevented some lock evictions (2328)
5954 - ENOMEM detection and retry on socknal sends (2230)
5955 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
5956 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
5957 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
5958 - better error messages when a client is rejected during recovery (1505)
5959 - avoid cancelling locks which were never granted, after failure (2330)
5960 - fix i_sem/journal inversion in mds_client_add (2333)
5961 - fix truncate/getattr lock cycle deadlock (2334)
5962 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
5963 - fix two rare exit paths which could leak an l_lock() ref (2321)
5964 - fencepost error in MDS/OST orphan recovery (2226)
5965 - make log record alignment 8 bytes (1988)
5966 - lstripe now fails when requested offset > ost_count (2237)
5967 - ensure that all kernel series have a complete list.h (1607)
5968 - fix crashes in special-file operations (2316)
5969 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
5971 - allow configurable automake binary, for testing new versions
5972 - small update to the lfs documentation
5974 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
5976 * fix negative export reference count in fsfilt_sync (2312)
5978 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
5979 * release candidate 0.9.1
5981 - orphans are moved into the PENDING directory for possible recovery
5982 - replayed opens now open by fid for orphan/rename safety (1042)
5983 - last close of an orphan inode generates a transno (683)
5984 - chdir() and mount() now pin the directory entry (1020)
5985 - avoid CERROR in normal ll_setattr_raw() error case (1500)
5986 - discard very old requests without processing them (1502)
5987 - remove some common, well-understood CERRORs (1505)
5988 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
5989 - clear "grant" flags in OST replies until OST grant code lands (1644)
5990 - fix read performance by not clobbering i_blksize on client (1598)
5991 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
5992 - make LRU size a /proc tunable, clears locks when reduced (707)
5993 - fix some lprocfs rot that prevented ptlbd from loading (1732)
5994 - server locks take references on exports now (1558)
5995 - build fixes for 2.4.20-rh trees (1663)
5996 - return an error from lov_create if all OSCs are inactive (1751)
5997 - fix import levels when a reconnect happens without a timeout (1597)
5998 - exit early from mds_open if we get a lookup error (1749)
5999 - partial page read at EOF wouldn't wait for disk before sending (1642)
6000 - avoid NULL deref in obdfilter when reading page past EOF (1592)
6001 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
6002 - fix LBUG in ll_it_open_error with rc = -2 (1861)
6003 - write/truncate lock inversion (1639)
6004 - Don't auto-load obdclass, portals modules during cleanup (1495)
6005 - fix timestamps from jumping to "now" (1763)
6006 - extra journal assertions (1648)
6007 - add an extra multiunlink test (1771)
6008 - fix read_record/write_record API (1776)
6009 - fix leak of offset_extent, possible incorrect i_size later (1772)
6010 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
6011 - Debugging for the kqswnal_get_idle_tx problems (1820)
6012 - Allow recovery to be attempted multiple times (1536)
6013 - Write out MDS last_rcvd file after it is first created (1600)
6014 - Fix tx_descriptor leak in failed transmit situations (1827)
6015 - ext3 journaling fixes for assertion failure after IO error (1871)
6016 - class_export_put() on freed export after completion AST error (1896)
6017 - Fix revalidate looping in VFS (1322)
6018 - Don't access a freed export during MDS_REINT timeout (1521)
6019 - Add open-unlink recovery support on the MDS (1673,1764)
6020 - Return an error if no MDS data was read from last_rcvd (1946)
6021 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
6022 - Don't setup a disk device that doesn't match exported UUID (317)
6023 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
6024 - avoid committing NULL handle in force close
6025 - local.sh is now a one-stripe LOV configuration
6026 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
6027 - fix ext3 htree duplicate directory entry corruption (1516)
6028 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
6029 - update to new LOV EA format (2097)
6030 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
6032 2003-06-15 Phil Schwan <phil@clusterfs.com>
6035 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
6036 - per-import/export recovery handling (958, 931, 959)
6037 - multiple last-rcvd slots, for serving multiple FSes (949)
6038 - connections are again shared between multiple imp/exports (963, 964)
6039 - "umount -f" would hang if any requests needed to be sent (393, 978)
6040 - avoid pinning large req buffer by copying for queued messages (989)
6041 - add "uuid" to "lctl device" command to help upcalls (991)
6042 - "open" RPCs with transnos would confuse recovery counters (1037)
6043 - do proper endian conversion of all wire messages (288, 340, 891)
6044 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
6045 - call waiting locks callback from LDLM recovery thread (1127, 1151)
6046 - fix ptlrpc_connection leak in target_handle_connect (1174)
6047 - fix import refcounting bug in OST and MDS cleanup (1134)
6048 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
6049 - fix very unlikely obd_types race condition (501)
6050 - remove osc_open hack for echo_client (1187)
6051 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
6052 - a failure in read_inode2 leads to deadlock (1139)
6053 - cancel ack-locks as soon as transaction is committed (1072)
6054 - fix major leaks and crashes in the bulk I/O path (937, 1057)
6055 - make sure to commitrw after any preprw to avoid deadlock (1162)
6056 - failing to execute a file in a lustre FS would lock inode (1203)
6057 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
6058 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
6059 - obd_uuid2tgt would walk off the end of the list (1255)
6060 - on IA64 the osc would give portals incorrect bulk size (1258)
6061 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
6062 - fix lock inversion caused by new llite matching code (1282)
6063 - limit the number of dirty pages on a client to 10MB (1286)
6064 - timed out locks were not being corrected cancelled (1289)
6065 - fix O_DIRECT above 4GB on IA-32 (1292)
6066 * major user-visible changes
6067 - fail out/fail over policy now controlled by the upcall (993)
6069 - add OBD_PING to check server availability and failure (954)
6070 - lustre messages are now sent in sending host order (288, 340, 891)
6071 - add eadatalen to MDS getattr reply (340)
6072 - OST read replies may contain second buffer, with per-page status (593)
6074 2003-03-11 Phil Schwan <phil@clusterfs.com>
6077 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
6078 - failed open()s could cause deadlock; fixed (867, 869)
6079 - stop cancelling OST locks when files are closed (481)
6080 - overlapping XID spaces caused network corruption (851, 853)
6081 - fix unsafe fsfilt counter arithmetic; change to atomic_t
6082 - setattr_raw added, to do single-RPC, server-side setattrs
6083 - lmc/lconf syntax change for OST UUIDs
6084 - fix crashy race condition between ptlrpc_free_req and osc_close
6085 - don't use request in mdc_enqueue if we hit a timeout (889)
6086 - don't set the inode i_size for regular files from the MDS (896)
6087 - handle out of order completion AST (842)
6088 - don't LBUG if a lock request times out after receiving AST (913)
6089 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
6090 - if a bad lock AST arrives, send an error instead of dropping entirely
6091 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
6092 - fix leak in bulk IO when only partially completed (899, 900, 926)
6093 - fix O_DIRECT for ia64 (55)
6094 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
6095 - C-z after timeout could hang a process forever; fixed (977)
6097 - client-side I/O cache (678, 924, 929, 941, 970)
6099 - READPAGE and SETATTRs which don't take server-side locks get
6102 2003-02-11 Phil Schwan <phil@clusterfs.com>
6105 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
6106 - Fix fsfilt_extN_readpage() to read a full page of directory
6107 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
6108 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
6109 - Fix imbalanced LOV object allocation and out-of-bound access (469)
6110 - Most intent operations were removed, in favour of a new RPC mode
6111 that does a single RPC to the server and bypasses most of the VFS
6112 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
6113 - Aggressively cancel local locks on DLM servers
6114 - mds_reint_unlink sends EA to the client if it's the last nlink.
6115 client uses that EA to unlink OST objects.
6116 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
6117 - recursive symlinks were fixed (439)
6118 - fixed NULL deref in DEBUG_REQ
6119 - filter_update_lastobjid no longer calls sync, which annoyed extN
6120 - fixed multi-client small-writes to a single file problem (445)
6121 - fixed mtime updates during file writes (607)
6122 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
6123 - fixed bug in obd_brw_read/write() (under guise of testing 367)
6124 - fixed Linux OST size reporting problem (444, 656)
6125 - OST now updates object mtime with writes or setattr (607, 619)
6126 - client verifies file size before zeroing page past EOF (445)
6127 - OST now writes last allocated objid to disk with allocation (108)
6128 - LOV on echo now works (409)
6130 - mds_reint_unlink sends a new buffer, with the EA included. this
6131 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
6132 set if a regular file was being unlinked, and it was the last link
6133 - use PtlGet from the target for bulk writes (315)
6134 - OST now updates object mtime with writes or setattr (607, 619)
6135 - LDLM now has a grant-time callback to revalidate locked items, if
6137 - Many MDS operations were reorganized to combat race conditions
6139 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
6140 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
6142 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
6145 - Fully reactivate OST imports after reconnection (512, others)
6146 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
6147 - More graceful error handling for truncating on dead OST (515)
6148 - Don't error out unless we're actually accessing dead stripes (474)
6149 - Fix garbage sizes when stripes are missing (410)
6150 - LRU counters were broken, causing constant lock purge (433, 432)
6151 - garbage on read from stripes with failed OSTs (441)
6152 - mark OSCs as active before reconnecting during recovery (438)
6153 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
6154 - lfind did not preserve OST order in output (443)
6155 - symlinks cause hung clients, incorrect data (439)
6156 - stop dereferencing request after dropping refcount (457)
6157 - don't LASSERT(spin_is_locked) on non-SMP (455)
6158 - fixes for many rename() bugs
6159 - fstat didn't correctly synchronize attributes (399)
6160 - server must handle lock cancellation during blocking AST prep (487)
6161 - bulk descriptors were free()d too soon (511)
6162 - fix paths in lconf, which would load incorrect modules (451, 507)
6163 - fix confusing lconf 'host not found' error message (386)
6164 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
6165 - fix race condition in mdc_blocking_ast() for inode access (526)
6166 - fix lov_unpackmd() unpacking wrong number of stripes (537)
6167 - fix lov_set_osc_active() marking wrong OSC inactive (440)
6168 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
6169 - fix multiple writes of stripe MD to MDS (358, maybe 519)
6170 - fix lstripe in several ways (kernel side) (527)
6171 - fix request leak in ldlm_cli_enqueue (262)
6172 - incorrect OSC was marked inactive after OST failure
6173 - call mds_fs_cleanup before unmounting filesystem (524)
6174 - fix races between taking ns_lock and ldlm_lock_change_resource
6175 - fix races updating LOV export open file list
6176 - fix lov_enqueue error path, avoid decref-ing bad lock handle
6177 - fix recovery NULL deref in ldlm_cli_cancel_unused
6178 - fix some DLM races by using new hash table for lock handles (419)
6179 - permit the client to specify desired inodes, at replay
6180 - duplicate requests when we queue them for replay reintegration
6181 - fix last_rcvd offset calculation
6182 - sync after each recovered transaction, so we always make progress
6183 - never, not always, ERESTART requests without transnos
6184 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
6186 - skip replay if the MDS says that the client is already connected
6187 - don't check for a recovery-enabled export to match lctl's UUID
6188 - don't INC_USE_COUNT for phantom exports
6189 - don't crash when cleaning up phantom exports (567)
6190 - don't double-finish or set replay data for errored mdc_open requests
6191 - abort requests when they time out, so we don't get old replies
6192 - send/receive replies for AST messages again
6193 - if the client says that it doesn't have the lock, cancel it on the
6195 - if we timeout during I/O, don't try to cancel an in-use lock; instead
6196 mark it as destroyed, it will all work out when decref is called
6197 - fix module use counts (22, 581)
6199 - ASTs now expect a reply (server cancels lock on error reply)
6201 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
6204 - fix many simultaneous client startup (392)
6205 - fix dentry->d_it clobbering
6206 - credentials weren't being shipped for readdir/getattr operations
6207 - remove invalid assertions triggered during some concurrent MD
6209 - proper Lustre versions added (336, 389)
6210 - fix memory leak for create error case (398)
6211 - fix LOV locking bug that would get cli/srv out of sync
6212 - fix echo client over LOV (409)
6213 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
6214 - fix double-O_EXCL intent crash (424)
6215 - avoid sending multiple lock CANCELs (352)
6217 - MDS can do multi-client recovery (modulo bugs in new code)
6219 - many updates, edits, cleanups
6221 2002-11-18 Phil Schwan <phil@clusterfs.com>
6224 - fix null d_it dereference (346)
6225 - fix full OST/dbench hang (333)
6226 - fix permission problem with file removal (286)
6227 - fix removal of OSCs from LOV when they fail
6228 - fix NULL deref during bulk timeout (214)
6229 - fix problems related to multiple filesystems on one MDS (241)
6230 - fixed serious subtle metadata locking bugs
6231 - free locks on clients when inodes are removed due to memory
6233 - fix inode pointer in lock data (285)
6234 - partial support for multiple MDS on a single host (241)
6235 - data locks weren't cancelled at clear_inode time (290, 311)
6236 - intent locks could lead to unbounded lock growth (205)
6237 - added a maximum lock count, an LRU list, and a flusher
6238 - fix multiple rename (365)
6239 - properly abstracted the echo client
6240 - OSC locked 1 byte too many; fixed
6241 - rewrote brw callback code:
6242 - fixed recovery bugs related to LOVs (306)
6243 - fixed too-many-pages-in-one-write crash (191)
6244 - fixed (again) crash in sync_io_timeout (214)
6245 - probably fixed callback-related race (385)
6247 - Add capability to MDS protocol
6248 - LDLM cancellations and callbacks on different portals
6250 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
6253 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
6254 - defer open object destruction to close time (601981, 138)
6255 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
6256 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
6257 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
6258 - don't call obd_setattr in ll_file_release for destroyed objects
6259 * protocol change to lustre_msg: move |version| and add |flags|
6260 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
6261 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
6262 * added replay of create, unlink, link and rename operations during
6263 MDS failover; recovery should be much more robust now
6264 * remove failed OSCs from LOVs (only lov_create uses this so far)
6265 * the lustre-HOWTO was brought (more) up to date (582544)
6267 2002-10-23 Phil Schwan <phil@clusterfs.com>
6270 - in-use dentries weren't being reused properly (617851)
6271 - prevent multiple LDLM setup (599178)
6272 - fix LOV size calculations for truncate (617853)
6273 - fix client handling of MDS intent errors (POSIX)
6274 - fix permission bug in lovstripe.c test (624321)
6275 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
6276 - truncate past end of file could corrupt data
6277 - proper cleanup after timeouts, crashes, etc (592524, 550815)
6278 - a race in recovery could return ETIMEDOUT to apps (623947)
6279 - building outside the source directory was fixed
6280 * the lustre-HOWTO was brought (more) up to date (582544)
6281 * major progress was made on recovery functionality
6283 2002-10-10 Phil Schwan <phil@clusterfs.com>
6286 - recovery deadlock fix
6287 - rm -rf causes LBUG fix (617817)
6288 - file open by multiple tasks fix (618962)
6289 - directory permissions bugs (602707 and 620007)
6290 - journal_stop fixed with locking (611313)
6291 - O_APPEND failures resolved (618273, perhaps 614459)
6292 - lconf PATH fix (619770)
6293 - IA64 build fix (621450)
6294 - RPC buffer sizes scale with amount of memory
6296 2002-10-01 Phil Schwan <phil@clusterfs.com>
6299 - locks would be cancelled without throwing away data pages,
6300 resulting in inconsistent data (605627)
6301 - inode attributes were not always being refreshed (605627, 612449)
6302 - lconf now continues to cleanup after lctl reports an error
6303 - MDS now enforces user permissions (602707)
6304 - lprocfs cleanup fixed, but not yet enabled (614157)
6305 - fixed infinite server hang, should a client not respond to an AST
6306 - avoid going into recovery if user calls readlink() with a buffer
6307 that's too small (613941)
6308 - AST RPCs no longer require replies (614867) -- this may be changed
6309 - don't crash server if client sends an IOV that's too big (611336)
6310 - fixed lock conversion deadlock (611892)
6311 - fixed the following of symlinks (614622)
6312 * recovery: the server can remove locks from a client that dies, other
6313 clients can make progress
6314 * more extN patch fixes
6315 * compile-time configurable ptlrpc buffer allocations
6317 - collaborative read cache document
6318 - Lustre Lite Performance CDR document-in-progress
6320 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6323 - fix typo in patch-2.4.18
6325 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6328 - clear ptlrpc request each time in handle_incoming_request()
6329 - unlink of files now destroys the object on the OST
6331 2002-09-19 Peter Braam <braam@clusterfs.com>
6333 * add hard link support
6334 * change obdfile creation method
6335 * kernel patch changed
6337 2002-09-19 Peter Braam <braam@clusterfs.com>
6340 - stack overflow bug in extN fixed
6342 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
6344 * documentation updates
6345 - add man pages for config tools
6346 - update tests/README to describe testing with new config tools
6347 - finish metadata API descriptions
6348 * bug fixes and cleanups
6349 - statfs workaround for 16TB limit
6350 - LOV stripe allocation improved, can stripe on subset of OSTs
6351 - LOV file size/IO offset was wrong for files > 4GB in size
6352 - object EA data was being dropped, caused files to be unreadable
6353 - memory overflow with non-LOV OST caused memory corruption
6354 - fixed regression tests to work with new config tools, obdfilter
6355 - fixed bug when directory size became larger than 1 block
6356 - fixed bug (for single client case) when PWD was deleted
6357 - invalidate local directory pages when doing intent-based ops
6358 - avoid LDLM oops when lock callback contained bad data
6360 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
6362 * documentation updates
6363 * bug fixes and cleanups
6364 - configuration tools
6367 - 64-bit compile warnings
6368 - 64-bit internal statfs data
6370 * test_brw on persistent OST devices
6372 * lprocfs (disabled)
6374 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
6376 * documentation updates
6377 * bug fixes and cleanups
6378 * configuration tools
6380 2002-08-30 Peter J. Braam <braam@clusterfs.com>
6383 * many small fixes to 0_5_4
6384 * io/network handling
6385 * thinkos in MDS operations
6387 2002-08-24 Peter J. Braam <braam@clusterfs.com>
6390 * crucial basic fixes to 0.5.3
6391 * IOR, Iozone work over Elan
6394 2002-08-07 Phil Schwan <phil@clusterfs.com>
6395 * version 0_5_3, our first alpha
6396 * we use the new Portals iovs
6397 * documentation updates
6398 * bug fixes and cleanups
6399 * small changes in the DLM wire protocol
6401 2002-07-25 Peter J. Braam <braam@clusterfs.com>
6402 * version 0_5_1 with some initial stability,
6403 * locking on MD and file I/O.
6404 * documentation updates
6405 * several bug fixes since 0.5.0
6406 * small changes in wire protocol
6408 2002-07-18 Phil Schwan <phil@clusterfs.com>
6410 * delivered as Lustre Light Alpha
6411 * fixed a crash after handling invalid MDS requests
6412 * fixed directory pages for architectures with non-4k pages sizes
6414 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
6415 * release version v0_4_4
6416 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
6418 * Fixes a number of interruption problems with OST operations.
6419 * Update documentation for portals header changes
6420 * Move all wire protocol structs/defines to lustre_idl.h
6421 * Fixes symlink length bug.
6422 * Add tcpdump to repository.
6424 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
6425 * release version v0_4_3
6426 * Fixes statfs for inodes on extN.
6427 * Fixes bug in runtests which would delete /etc/hosts.
6428 * Use 64-bit object IDs wherever possible (not into VFS though)
6429 Remove ost_get_info, which is unused by lustre, and out of date.
6431 2002-07-03 Peter Braam <braam@clusterfs.com>
6432 * release version v0_4_2 Fixes a lookup error (type not passed)
6433 * move forward to head of Portals
6434 * move forward to latest Lustre kernel
6436 2002-06-25 Peter Braam <braam@clusterfs.com>
6437 * release version v0_4_1. Hopefully stable on single node use.