3 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
4 for currently supported client and server kernel versions.
5 * Server known to build on patched kernels:
6 2.6.32-431.29.2.el6 (RHEL6.5)
7 2.6.32-504.8.1.el6 (RHEL6.6)
8 3.0.101-0.46 (SLES11 SP3)
9 * Client known to build on unpatched kernels:
10 2.6.32-431.29.2.el6 (RHEL6.5)
11 2.6.32-504.8.1.el6 (RHEL6.6)
12 3.10.0-123.20.1.el7 (RHEL7)
13 3.0.101-0.46 (SLES11 SP3)
14 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
15 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
16 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
17 stack overflow. For more information, please refer to bugzilla 17630.
18 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
19 the re-exporting nodes to be mounted with "32bitapi" mount option
21 Severity : enhancement
23 Description: control OST-index in IDIF via ROCOMPAT flag.
24 Details : Introduce new flag OBD_ROCOMPAT_IDX_IN_IDIF that is stored in the
25 last_rcvd file. For new formatted OST device, it will be auto set;
26 for the case of upgrading from old OST device, you can enable it
27 via the lproc interface osd-ldiskfs.index_in_idif. With such flag
28 enabled, for new created OST-object, its IDIF-in-LMA will contain
29 the OST-index; for the existing OST-object, the OSD will convert
30 old format IDIF as new format IDIF with OST-index stored in the
31 LMA EA when accessing such OST-object or via OI scrub. Once such
32 flag is enabled, it cannot be reverted back, so the system cannot
33 be downgraded to the orignal incompatible version.
35 --------------------------------------------------------------------------------
37 07-30-2014 Intel Corporation
39 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
40 for currently supported client and server kernel versions.
41 * Server known to build on patched kernels:
42 2.6.32-431.20.3.el6 (RHEL6.5)
43 3.0.101-0.31 (SLES11 SP3)
44 * Client known to build on unpatched kernels:
45 2.6.32-431.20.3.el6 (RHEL6.5)
46 3.0.101-0.31 (SLES11 SP3)
49 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
50 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
51 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
52 stack overflow. For more information, please refer to bugzilla 17630.
53 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
54 the re-exporting nodes to be mounted with "32bitapi" mount option
56 --------------------------------------------------------------------------------
58 10-31-2013 Intel Corporation
60 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
61 * Server builds for kernels:
62 2.6.32-358.18.1.el6 (RHEL6.4)
63 * Client builds for unpatched kernels:
64 2.6.32-358.18.1.el6 (RHEL6.4)
65 3.0.92-0.8 (SLES11 SP3)
67 * Recommended e2fsprogs version: 1.42.7.wc2 or newer
68 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
69 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
70 stack overflow. For more information, please refer to bugzilla 17630.
71 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
72 the re-exporting nodes to be mounted with "32bitapi" mount option
74 --------------------------------------------------------------------------------
76 05-30-2013 Intel Corporation
78 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
79 * Server builds for kernels:
80 2.6.32-358.6.2.el6 (RHEL6.4)
81 * Client builds for unpatched kernels:
82 2.6.32-358.6.2.el6 (RHEL6.4)
83 3.0.74-0.6.6 (SLES11 SP2)
85 * Recommended e2fsprogs version: 1.42.7.wc1 or newer
86 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
87 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
88 stack overflow. For more information, please refer to bugzilla 17630.
89 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
90 the re-exporting nodes to be mounted with "32bitapi" mount option
92 --------------------------------------------------------------------------------
94 9-30-2012 Whamcloud, Inc.
96 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
97 * Server support for kernels:
98 2.6.32-279.5.1.el6 (RHEL6)
99 * Client support for unpatched kernels:
100 2.6.32-279.5.1.el6 (RHEL6)
102 2.6.18-238.19.1.el5 (RHEL5)
103 2.6.16.60-0.69.1 (SLES10)
104 * Recommended e2fsprogs version: 1.41.90.wc4
105 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
106 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
107 stack overflow. For more information, please refer to bugzilla 17630.
108 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
109 the re-exporting nodes to be mounted with "32bitapi" mount option
111 Severity : enhancement
113 Description: Update RHEL6.3 kernel to 2.6.32-279.5.1.el6.
115 Severity : enhancement
117 Description: Update RHEL6.3 kernel to 2.6.32-279.2.1.el6.
119 Severity : enhancement
121 Description: Update RHEL6.2 kernel to 2.6.32-220.17.1.el6
123 Severity : enhancement
125 Description: Update RHEL6.2 kernel to 2.6.32-220.13.1.el6
127 Severity : enhancement
129 Description: Update RHEL6.2 kernel to 2.6.32-220.7.1.el6
131 --------------------------------------------------------------------------------
133 04-30-2012 Whamcloud, Inc.
135 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
136 * Server support for kernels:
137 2.6.32-220.4.2.el6 (RHEL6)
138 * Client support for unpatched kernels:
139 2.6.32-220.4.2.el6 (RHEL6)
141 2.6.18-238.19.1.el5 (RHEL5)
142 2.6.16.60-0.69.1 (SLES10)
143 * Recommended e2fsprogs version: 1.41.90.wc4
144 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
145 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
146 stack overflow. For more information, please refer to bugzilla 17630.
147 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
148 the re-exporting nodes to be mounted with "32bitapi" mount option
150 Severity : enhancement
152 Description: Update RHEL6.2 kernel to 2.6.32-220.4.2.el6
154 Severity : enhancement
156 Description: Update RHEL6.2 kernel to 2.6.32-220.4.1.el6
158 Severity : enhancement
160 Description: Update RHEL6.2 kernel to 2.6.32-220.el6
162 Severity : enhancement
164 Description: Update RHEL6.1 kernel to 2.6.32-131.17.1.el6
166 --------------------------------------------------------------------------------
168 09-30-2011 Whamcloud, Inc.
170 * Server support for kernels:
171 2.6.18-238.19.1.el5 (RHEL5)
172 2.6.32-131.6.1.el6 (RHEL6)
173 * Client support for unpatched kernels:
174 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
175 2.6.18-238.19.1.el5 (RHEL5)
176 2.6.32-131.6.1.el6 (RHEL6)
177 2.6.29.4-167.fc11 (FC11)
178 2.6.16.60-0.69.1 (SLES10)
180 * Recommended e2fsprogs version: 1.41.90.wc3
181 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
182 Lustre filesystem with 4K stack may cause a stack overflow. For more
183 information, please refer to bugzilla 17630.
184 * NFSv4 reexport requires Lustre client on the re-exporting nodes to be
185 mounted with "32bitapi" mount option
187 Severity : enhancement
188 Description: Update RHEL6.1 kernel to 2.6.32-131.6.1.el6
190 Severity : enhancement
191 Description: Update RHEL5.6 kernel to 2.6.18-238.19.1.el5
193 Severity : enhancement
194 Description: Update RHEL5.6 kernel to 2.6.18-238.9.1.el5
197 Frequency : only with SLES10
199 Description: Use OFED "KMP" provided by Novell
200 Details : SLES10 SP3 ships with OFED in a separate "KMP" package.
201 Lustre is now built against this package. That means you need to
202 install the ofed-kmp package from Novell for the patchless client.
203 Note that the ofed-kmp that Novell ships may not exactly match the
204 kernel version but should still be compatible.
206 Severity : enhancement
208 Description: Update RHEL5.5 kernel to 2.6.18-194.17.1.el5
209 Update OEL5.5 kernel to 2.6.18-194.17.1.0.1.el5.
211 --------------------------------------------------------------------------------
213 2010-07-15 Oracle, Inc.
215 * Server support for kernels:
216 2.6.18-164.11.1.el5 (RHEL 5)
217 2.6.18-164.11.1.0.1.el5 (OEL 5)
218 * Client support for unpatched kernels:
219 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
220 2.6.18-164.11.1.el5 (RHEL 5),
221 2.6.18-164.11.1.0.1.el5 (OEL 5)
222 2.6.16.60-0.42.8 (SLES 10),
224 2.6.29.4-167.fc11 (FC11)
225 * Recommended e2fsprogs version: 1.41.10-sun2
227 Severity : enhancement
229 Description: Distribute lfs_migrate script with lustre
230 Details : The lfs_migrate mentioned in the manual is now included in the
231 lustre tree in lustre/scripts and will distributed in lustre RPMs.
233 Severity : enhancement
235 Description: Failover nid list added to import proc file.
237 Severity : enhancement
239 Description: "som_preview" mount option required to use SOM
240 Details : The clients will need to use som_preview mount option
244 Description: Lustre does not do 1MB IOs to HW RAID
245 Details : Bump MAX_PHYS/HW_SEGMENTS and SG_ALL to 256 in the RHEL5 kernel.
247 Severity : enhancement
249 Description: lfs quota output is not very convenient for awk/sed-parsing
250 Details : Some positions in lfs quota output table could be empty or
251 non-empty which made it hard to parse it with scripts, now
252 a dash is put instead of space where there is not supposed
257 Description: Update to OFED 1.5.1
258 Details : OFED 1.5.1 will be used in environments that don't use
259 vendor supplied OFED.
263 Description: Remove lnet. prefix from lctl params display
264 Details : Also renamed lnet.memused to lnet_memused to
269 Description: Use conf_param -d to remove permanant settings.
273 Description: Add list_param -R
274 Details : lctl list_param now supports recursive mode.
277 Bugzilla : 21829, 22312
278 Description: Fix llobdstat.
279 Details : 1. Now updating the write_bytes counter
280 2. Port script fixes from 1.8.
284 Description: decrease the usage of memory on clients.
285 Details : 1. On clients, recycle dentries and inodes unused.
286 2. Delete the code related to ll_deathrow(att 6215 in bug 1443). It
291 Description: llobdstat fix and enhancement
292 Details : add a counter to set a limit to how many samples will be returned
293 fix a wildcard in the path to limit to obdfilter stats only
297 Description: lfs check only allowed for root
298 Details : Code cleanups and added a regresion test of non-root use of
303 Description: MDS Oopses in ptlrpc_dispatch_difficult_reply()/__list_add()
304 Details : Fix hr->hr_index overload.
308 Description: Fix FIEMAP functionality.
312 Description: FC11 patchless client support added
316 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and
317 OEL5.4 kernel to 2.6.18-164.11.1.0.1.el5.
321 Description: New acc-sm script and various test-framework enhancments.
325 Description: Previously landed patch was reverted.
329 Description: Update SLES10 kernel to 2.6.16.60-0.42.8 (patchless)
333 Description: Add ldiskfs maxdirsize mount option.
337 Description: Fix mkfs.lustre support for 16TB LUNs.
341 Description: Print mmp_check_interval
342 Details : print mmp_check_interval and make it possible to abort
343 mount operation in case it takes too long.
347 Description: Add cascading_rw to lustre mpi tests.
351 Description: Store SOM data in EA.
352 Details : Also several SOM related cleanups.
356 Description: Add refcount to md_open_data
360 Description: Use refcount on lov_request_set to prevent memory corruption.
364 Description: Statahead: use parent dir's i_mutex for synchronization.
368 Description: loadgen improvements.
372 Description: Enable peer health detection in ksocklnd and ko2iblnd.
376 Description: Control DCACHE_LUSTRE_INVALID flag with MDS_INODE_LOOKCUP lock
380 Description: Fix race on obd_devcie:md_stats
384 Description: Fix SLES 11 client builds.
385 Details : Resolved uninitalized pointer gcc warnings.
389 Description: Allow non-root access for "lfs check".
390 Details : Added a check in obd_class_ioctl() for OBD_IOC_PING_TARGET.
394 Description: Clear imp_force_reconnect correctly in ptlrpc_connect_interpret()
396 Severity : enhancement
398 Description: Update kernel to RHEL5.4 2.6.18-164.6.1.el5 and
399 OEL5 2.6.18-164.6.1.0.1.el5(Both in-kernel OFED enabled).
403 Description: Sanity test_27p fails.
404 Details : OBD_FAIL_OST_ENOSPC has been moved to filter_statfs().
405 Sanity tests 27 were cleaned up and improved.
409 Description: Unify req format on client/servers
410 Details : Use new req_capsule API [almost] everywhere instead of old PTLRPC
411 buffers and swabbers approach..
416 Description: drop join file stuff.
418 Severity : enhancement
420 Description: Add 2.6.27 and SLES11 (patchless client) support
424 Description: Fix in ptlrpc_expire_one_request() to print the signed time
427 Severity : enhancement
429 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
432 Severity : enhancement
434 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
436 Severity : enhancement
438 Description: correctly handle big reply message.
439 Details : send LNet event if reply is bigger then buffer and adjust this buffer
444 Description: Drop unnecessary __GFP_NOMEMALLOC flag from filter_get_page()
446 Severity : enhancement
448 Description: Add support for OEL5.
452 Description: Conf-sanity.sh 50g test - deactivated OST should not cause a panic.
454 Severity : enhancement
456 Description: "weak-modules" support
457 Details : Implement "weak-modules" support which enables kernel modules
458 to be used with any kernel that implements the same kABI. In
459 order to achieve this modules are now installed in
460 /lib/modules/$(uname -r)/updates/kernel on all distributions.
464 Description: client could not reconnect to OST because of an active request.
465 Details : The client now retries to reconnect to the same server, if a
466 connect request failed with EBUSY or -EAGAIN. Test case 35b
467 conf-sanity.sh has been created.
471 Description: Don't allow make backward step on assignin osc next id.
472 Details : race between allocation next id and ll_sync thread can be cause
473 of set wrong osc next id and can be kill valid ost objects.
477 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
481 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
485 Description: Error handling in osc_statfs_interpret() has been improved.
486 Details : Check in osc_statfs_interpret() for EBADR.
490 Description: Do not send statfs() requests to OSTs disabled by administrator.
491 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
493 Severity : enhancement
495 Description: Added sanity-quota test to for lfs(1) to work with numeric uid/gid.
497 Severity : enhancement
499 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
501 Severity : enhancement
503 Description: Enhancement for lfs(1) command to use numeric uid/gid.
507 Description: Improved cleanup procedure for conf-sanity test_46a.
511 Description: Do not start bulk transfer, if client tried to reconnect.
512 Details : Do not start bulk transfer, if client tried to reconnect,
513 but failed due to active rpcs. Check for exp_abort_active_req
514 has been implemented.
518 Description: Prevent inconsistences between linux and lustre mount structures.
519 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
520 Make the sleep interruptible.
522 Severity : enhancement
524 Description: provide server to client comms path
525 Details : server to client communications path via new LDLM_SET_INFO rpc
529 Description: truncate starts GFP_FS allocation under transaction causing deadlock
530 Details : ldiskfs_truncate calls grab_cache_page which may start page
531 allocation under an open transaction. This may lead to
532 calling prune_icache with consequent lustre reentrance.
536 Description: send correctly lsm on open replay
537 Details : MDS is trust to LSM size on replay open, but client can set wrong size
542 Description: enable client interface failover
543 Details : When a child reconnects from another NID, properly update export
544 nid hash position and ldlm reverse import.
546 Severity : enhancement
548 Description: Communicate OST degraded/readonly state via statfs to MDS
549 Details : Flags in the statfs returned from OSTs indicate whether the
550 OST is in a degraded RAID state, or if the filesystem has
551 turned read-only after a filesystem error is detected.
556 Description: don't panic if EPROTO was hit when reading symlink
557 Details : correctly handling request reference in error cases.
559 Severity : enhancement
561 Description: Add LustreNetLink, a kernel-userspace communcation path.
563 Severity : enhancement
565 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
570 Description: access to llog context before init.
571 Details : move handling CATALOGS file at osc layer and forbid access to llog
575 Frequency : rare, only when client is writing to OST on local node
577 Description: Avoid deadlock for local client writes
578 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
579 memory freeing context. This allows OST threads to set the
580 PF_MEMALLOC flag on task structures in order to allocate memory
581 from reserved pools and complete IO.
582 Use GFP_HIGHUSER for OST allocations for non-local client writes,
583 so that the OST threads generate memory pressure and allow
584 inactive pages to be reclaimed.
586 Severity : enhancement
588 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
593 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
594 Details : move ldlm namespace creation in setup phase to avoid grab
595 _lprocfs_lock with cli_sem held.
599 Description: Temporarily disable grant shrink.
600 Details : Disable the feature for debugging.
603 Frequency : only when formatting test filesystems
605 Description: Unable to run several mkfs.lustre on loop devices at the same time
606 Details : mkfs.lustre returns error 256 on the concurrent loop devices
607 formatting. The solution is to proper handle the error.
609 Severity : enhancement
611 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
613 Severity : enhancement
615 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
620 Description: resolve race between obd_disconnect and class_disconnect_exports
621 Details : if obd_disconnect will be called to already disconnected export he
622 forget release one reference and osc module can't unloaded.
624 Severity : enhancement
626 Description: Allow tuning service thread via /proc
627 Details : For each service a new
628 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
629 created that can be used to set min/max thread counts, and get the
630 current number of running threads.
635 Descriptoin: don't return error if have partially created objects for file.
636 Details : lov_update_create_set uses set->set_success as index for created
637 objects, so if some requests will be failed, they will have hole
638 at end of array and qos_shrink_lsm for allocate correct lsm.
640 Severity : enhancement
642 Description: Update OFED support to 1.4.1
647 Description: Reduce small size read RPC
648 Details : Set read-ahead limit for every file and only do read-ahead when
649 available read-ahead pages are bigger than 1M to avoid small size
652 Severity : enhancement
654 Description: move AT tunable parameters for more consistent usage
655 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
657 Severity : enhancement
659 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
660 Details : allow skip disconnected ost for send statfs request and hide error
666 Description: fix racy locking of mballoc block bitmaps causing BUG
667 Details : The locking of the mballoc buddy bitmap and the in-memory
668 block bitmap was using two different spin locks in some
669 cases. This made it possible to incorrectly access the
670 mballoc bitmap while another process was modifying it,
671 causing a sanity assertion to fail. While no on-disk corruption
672 was reported, there was some risk of this happening.
675 Frequency : rare, on llog test 6
677 Descriptoin: don't allow connect to already connected import
678 Details : allowing connect to already connected import is hide connecting problem.
683 Descriptoin: allow kill process which wait statahead result
684 Details : for some reasons 'ls' can stick in waiting result from statahead,
685 in this case need way for kill this process.
687 Severity : enhancement
689 Description: Add state history info file, enhance import info file
690 Details : Track import connection state changes in a new osc/mdc proc file;
691 add overview-type data to the osc/mdc import proc file.
693 Severity : enhancement
695 Description: MDS create should not wait for statfs RPC while holding DLM lock.
697 Severity : enhancement
699 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
704 Descriptoin: OST grant too much space to client even there are not enough space.
705 Details : Client will shrink its grant cache to OST if there are no write
706 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
707 this grant cache if there are already not enough avaible space
708 (left_space < total_clients * 32M).
711 Frequency : start MDS on uncleanly shutdowned MDS device
713 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
714 Details : stay in waiting mds<>ost recovery finished produce random bugs
715 due race between two ll_sync thread for one lov target. send
716 ACTIVATE event only if connect realy finished and import have
720 Frequency : rare, connect and disconnect target at same time
722 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
723 Details : don't call obd_disconnect under lov_lock. this long time
724 operation and can block ptlrpcd which answer to connect request.
729 Descriptoin: don't lose wakeup for imp_recovery_waitq
730 Details : recover_import_no_retry or invalidate_import and import_close can
731 both sleep on imp_recovery_waitq, but we was send only one wakeup
735 Frequency : always with long access acl
737 Descriptoin: mds can't pack reply with long acl.
738 Details : mds don't control size of acl but they limited by reint/getattr
741 Severity : enhancement
743 Description: Update to SLES10 kernel-2.6.16.60-0.33.
745 Severity : enhancement
747 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
750 Frequency : start MDS on uncleanly shutdowned MDS device
752 Descriptoin: aborting recovery hang on MDS
753 Details : don't throttle destroy RPCs for the MDT.
756 Frequency : on remount
758 Description: external journal device not working after the remount
759 Details : clear dev_rdonly flag for external journal devices in
765 Description: shutdown vs evict race
766 Details : client_disconnect_export vs connect request race.
767 if client will evicted at this time - we start invalidate
768 thread without referece to import and import can be freed
772 Frequency : rare, need acl's on inode.
774 Description: client can't handle ost additional correctly
775 Details : if ost was added after client connected to mds client can have
776 hit lnet_try_match_md ... to big messages to wide striped files.
777 in this case need teach client to handle config events about add
778 lov target and update client max ea size at that event.
780 Severity : enhancement
782 Description: Changelogs
783 Details : Changelogs are a lightweight mechanism to track filesystem
784 metadata and namespace changes. The changelog is recorded
785 permanently on the MDTs, and is periodically "consumed" / purged
786 when records are no longer needed.
788 Severity : enhancement
790 Description: compact fld format with extents
791 Details : Store range of seq rather than every seq in FLD. Seq
792 controller update FLD rather than clients. In Case of CMD, mdt0
793 has FLD, all other metadata server act as non persistent proxy
794 for FLD queries and cache fld entries in fld cache.
799 Description: don't skip ost target if they assigned to file
800 Details : Drop slow OSCs if we can, but not for requested start idx.
801 This means "if OSC is slow and it is not the requested
802 start OST, then it can be skipped, otherwise skip it only
803 if it is inactive/recovering/out-of-space.
807 Description: more cleanup in mds_lov
808 Details : not send LOV EA under replay, we can't know about they size at this
809 time. Don't allow client connect to mds before any ost connected,
810 for avoid problems with LOV EA size and returning EIO to client.
812 Severity : enhancement
814 Description: Interoperability at server side (Disk interoperability)
816 Severity : enhancement
818 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
820 Severity : enhancement
822 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
824 Severity : enhancement
826 Description: New client IO stack (CLIO).
828 Severity : enhancement
830 Description: Commit on sharing. Eliminate inter-client dependencies between
831 uncommitted transactions by doing transaction commits.
832 Thereby clients may recovery independently.
835 Frequency : Create a symlink file with a very long name
837 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
838 Details : If there is no extra space in the request for early cancels,
839 ldlm_req_handles_avail() returns 0 instead of a negative value.
841 Severity : enhancement
843 Description: Add /proc entry for import status
844 Details : The mdc, osc, and mgc import directories now have
845 an import directory that contains useful import data for debugging
848 Severity : enhancement
850 Description: Re-disable certain /proc logging
851 Details : Enable and disable client's offset_stats, extents_stats and
852 extents_stats_per_process stats logging on the fly.
855 Frequency : Only on FC kernels 2.6.22+
857 Description: oops in statahead
858 Details : Do not drop reference count for the dentry from VFS when lookup,
859 VFS will do that by itself.
861 Severity : enhancement
863 Description: Generic /proc file permissions
864 Details : Set /Proc file permissions in a more generic way to enable non-
865 root users operate on some /proc files.
869 Description: Hitting mdc_commit_close() ASSERTION
870 Details : Properly handle request reference release in
871 ll_release_openhandle().
875 Description: quota recovery deadlock during mds failover
876 Details : This patch includes att18982, att18236, att18237 in bz14840.
878 1. fix osts hang when mds does failover with quotaon
879 2. prevent watchdog storm when osts threads wait for the
884 Frequency : only patchless client
885 Description: add workaround for race between add/remove dentry from hash
887 Severity : enhancement
889 Description: Allow OST glimpses to return PW locks
893 Description: LBUG when llog conf file is full
894 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
898 Description: Prevent import from entering FULL state when server in recovery
902 Description: service mount cannot take device name with ":"
903 Details : Only when device name contains ":/" will mount treat it as
909 Description: replace ptlrpcd with the statahead thread to interpret the async
910 statahead RPC callback
914 Frequency : on recovery
915 Description: I/O failures after umount during fail back
916 Details : if client reconnected to restarted server we need join to recovery
917 instead of find server handler is changed and process self
918 eviction with cancel all locks.
920 Severity : enhancement
922 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
924 Severity : enhancement
926 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
928 Severity : enhancement
930 Description: Upcall on Lustre log has been dumped
931 Details : Allow for a user mode script to be called once a Lustre log has
932 been dumped. It passes the filename of the dumped log to the
933 script, the location of the script can be specified via
934 /proc/sys/lnet/debug_log_upcall.
939 Description: avoid idr_remove called for id which is not allocated.
940 Details : Move assigment s_dev for clustered nfs to end of initialization,
941 to avoid problem with error handling.
946 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
947 Details : When connection is reused this not moved from CONN_UNUSED_HASH
948 into CONN_USED_HASH and this prodice warning when put connection
949 again in unused hash.
952 Severity : enhancement
954 Description: File striping can now be set to use an arbitrary pool of OSTs.
956 Severity : enhancement
958 Description: Export bytes_read/bytes_write count on OSC/OST.
962 Description: Early reply size mismatch, MGC loses connection
963 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
964 the connect flags are properly negotiated.
969 Description: quotas are not honored with O_DIRECT
970 Details : all writes with the flag O_DIRECT will use grants which leads to
971 this problem. Now using OBD_BRW_SYNC to guard this.
975 Description: add quota statistics
976 Details : 1. sort out quota proc entries and proc code.
977 2. add quota statistics
979 Severity : enhancement
981 Description: enable quota support for HEAD.
985 Description: Properly propagate oinfo flags from lov to osc for statfs
986 Details : restore missing copy oi_flags to lov requests.
988 Severity : enhancement
990 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
991 : l_getgroups(8), lst(8), routerstat(8)
992 Details : included man pages for llobdstat(8), llstat(8),
993 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
995 Severity : enhancement
997 Description: configure's --enable-quota should check the
998 : kernel .config for CONFIG_QUOTA
999 Details : configure is terminated if --enable-quota is passed but
1000 : no quota support is in kernel
1004 Description: Remove portals compatibility
1005 Details : Remove portals compatibility, not interoperable with releases
1010 Description: Resolve device initialization race
1011 Details : Prevent proc handler from accessing devices added to the
1012 obd_devs array but yet be intialized.
1014 Severity : enhancement
1016 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
1018 Severity : enhancement
1020 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
1025 Description: Using wrong pointer in osc_brw_prep_request
1026 Details : Access to array[-1] can produce panic if kernel compiled with
1027 CONFIG_PAGE_ALLOC enabled
1031 Description: Client runs out of low memory
1032 Details : Consider only lowmem when counting initial number of llap pages
1036 Description: *optional* service tags registration
1037 Details : if the "service tags" package is installed on a Lustre node
1038 When the filesystem is mounted, a local-node service tag will
1039 be created. See http://inventory.sun.com/ for more information
1040 about the Service Tags asset management system.
1044 Description: Kernel BUG tries to release flock
1045 Details : Lustre does not destroy flock lock before last reference goes
1046 away. So always drop flock locks when client is evicted and
1047 perform unlock regardless of successfulness of speaking to MDS.
1051 Description: add refcount protection for osc callbacks, avoid panic on shutdown
1055 Description: sanity test 65a fails if stripecount of -1 is set
1056 Details : handle -1 striping on filesystem in ll_dirstripe_verify
1061 Description: ASSERTION(CheckWriteback(page,cmd)) failed
1062 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
1065 Severity : enhancement
1067 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
1071 Description: do not process already freed flock
1072 Details : flock can possibly be freed by another thread before it reaches
1073 to ldlm_flock_completion_ast.
1077 Description: LBUG during stress test
1078 Details : Need properly lock accesses the flock deadlock detection list.
1082 Description: oops in page fault handler
1083 Details : kernel page fault handler can return two special 'pages' in error
1084 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
1088 Description: timeout with invalidate import.
1089 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
1090 should be handled by ptlrpcd. This produce long age waiting and
1091 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
1093 Severity : enhancement
1095 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
1099 Description: enable MGS and MDT services start separately
1100 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
1101 a co-located MGS without starting the MGS, which is a complement
1102 to 'nosvc' mount option.
1106 Frequency : after recovery
1107 Description: precreate to many object's after del orphan.
1108 Details : del orphan st in oscc last_id == next_id and this triger growing
1109 count of precreated objects. Set flag LOW to skip increase count
1110 of precreated objects.
1114 Frequency : rare, on clear nid stats
1115 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
1116 Details : when clean nid stats sometimes try destroy live entry,
1117 and this produce panic in free.
1121 Description: Stack overflow during MDS log replay
1122 ease stack pressure by using a thread dealing llog_process.
1126 Description: wait until IO finished before start new when do lock cancel.
1127 Details : VM protocol want old IO finished before start new, in this case
1128 need wait until PG_writeback is cleared until check dirty flag and
1129 call writepages in lock cancel callback.
1131 Severity : enhancement
1133 Description: using special macro for print time and cleanup in includes.
1137 Description: mds_mfd_close() ASSERTION(rc == 0)
1138 Details : In mds_mfd_close(), we need protect inode's writecount change
1139 within its orphan write semaphore to prevent possible races.
1143 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
1147 Frequency : rare, on shutdown ost
1148 Description: don't hit live lock with umount ost.
1149 Details : shrink_dcache_parent can be in long loop with destroy dentries,
1150 use shrink_dcache_sb instead.
1154 Description: don't panic with use echo client
1155 Details : echo client pass NULL as client nid pointer and this produce null
1156 pointer dereference.
1160 Description: fix build on ppc32
1161 Details : compile code with -m64 flag produce wrong object file for ppc32.
1165 Description: add message levels for liblustreapi
1169 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
1170 Details : Move the 'good_osts' check before the 'total_bavail' check. This
1171 will result in an -EAGAIN and in the exit call path we call
1172 alloc_rr() which will with increasing aggressiveness attempt to
1173 aquire precreated objects on the minimum number of required OSCs.
1177 Description: Use old size assignment to avoid deadlock
1178 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
1179 the scheduling while holding a spinlock. We do not need locking
1180 for size in ll_update_inode() because size is only updated from
1181 the MDS for directories or files without objects, so there is no
1182 other place to do the update, and concurrent access to such inodes
1183 are protected by the inode lock.
1187 Description: resolve "_IOWR redefined" build error on SLES10
1191 Description: dump the memory debugging after all modules are unloaded to
1192 suppress false negative in conf_sanity test 39
1194 Severity : enhancement
1196 Description: build kernel-ib packages for OFED 1.3 in our release cycle
1201 Description: fix SLES kernel versioning
1202 Details : the kernel version for our SLES 10 kernel did not include a "-"
1203 before the "smp" at the end. while this was not a problem in
1204 general, it did mean that software trying to use the kernel
1205 version to try to detect a vendor specific kernel would fail.
1206 this was most evident by the OFED build scripts.
1210 Description: Don't update lov_desc members until making sure they are valid
1211 Details : When updating lov_desc members via proc fs, need fix their
1212 validities before doing the real update.
1216 Description: don't put request into delay list while invalidate in flight.
1217 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
1218 import in flight. this produce timeout for invalidate and sometimes
1219 can cause stale data.
1223 Frequency : on ppc only
1224 Description: not convert ost objects for directory because it's not exist.
1225 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
1229 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
1231 Severity : enhancement
1233 Description: Update to support 2.6.22.14 vanilla kernel.
1237 Frequency : rare, on recovery
1238 Description: read procfs can produce deadlock in some situation
1239 Details : Holding lprocfs lock which send rpc can produce block for destroy
1240 obd objects and this also block reconnect with -EALREADY. This
1241 isn't fix all lprocfs bugs - but make it rare.
1243 Severity : enhancement
1245 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
1248 Frequency : frequent on X2 node
1250 Description: mdc_set_open_replay_data LBUG
1251 Details : Set replay data for requests that are eligible for replay.
1255 Description: lustre_mgs: operation 101 on unconnected MGS
1256 Details : When MGC is disconnected from MGS long enough, MGS will evict the
1257 MGC, and late on MGC cannot successfully connect to MGS and a lot
1258 of the error messages complaining that MGS is not connected.
1262 Frequency : on network error
1263 Description: panic with double free request if network error
1264 Details : mdc_finish_enqueue is finish request if any network error ocuring,
1265 but it's true only for synchronus enqueue, for async enqueue
1266 (via ptlrpcd) this incorrect and ptlrpcd want finish request
1269 Severity : enhancement
1271 Description: client-side metadata stat-ahead during readdir(directory readahead)
1272 Details : perform client-side metadata stat-ahead when the client detects
1273 readdir and sequential stat of dir entries therein
1276 Frequency : on start mds
1278 Description: Implement get_info(last_id) in obdfilter.
1281 Frequency : occasional
1283 Description: Correctly check stale fid, not start epoch if ost not support SOM
1284 Details : open with flag O_CREATE need set old fid in op_fid3 because
1285 op_fid2 was overwritten with new generated fid, but mds can answer
1286 with one of these two fids and both is not stale. Setattr
1287 incorrectly started an epoch and assume will be called
1288 done_writing, but without SOM done_writing ever being called.
1291 Frequency : rare, depends on device drivers and load
1293 Description: MDS or OSS nodes crash due to stack overflow
1294 Details : Code changes in 1.8.0 increased the stack usage of some functions.
1295 In some cases, in conjunction with device drivers that use a lot
1296 of stack the MDS (or possibly OSS) service threads could overflow
1297 the stack. One change which was identified to consume additional
1298 stack has been reworked to avoid the extra stack usage.
1301 Frequency : occasional
1303 Description: Do not fail import if osc_interpret_create gets -EAGAIN
1304 Details : If osc_interpret_create got -EAGAIN it immediately exits and
1305 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
1306 oscc_has_objects and see OSC has no objests and call
1307 oscc_internal_create to resend create request.
1309 Severity : enhancement
1311 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
1313 Severity : enhancement
1315 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
1318 Frequency : very rare
1320 Description: Fix replay if there is an un-replied request and open
1321 Details : In some cases, older replay request will revert the
1322 mcd->mcd_last_xid on MDS which is used to record the client's
1323 latest sent request.
1325 Severity : enhancement
1327 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
1329 Severity : enhancement
1331 Description: Add rhel5 support to HEAD.
1333 Serverity : enhancement
1335 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
1340 Description: Don't allow skipping OSTs if index has been specified.
1341 Details : Don't allow skipping OSTs if index has been specified, make
1342 locking in internal create lots better.
1346 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
1347 Details : access to bitfield in structure is always rounded to long
1348 and this produce problem with not atomic change any bit.
1352 Description: Lustre make rpms failed.
1353 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
1354 builds Lustre rpms from tarball.
1357 Frequency : rare on shutdown ost
1359 Description: If llog cancel was not send before clean_exports phase, this can
1360 produce deadlock in llog code.
1361 Details : If llog thread has last reference to obd and call class_import_put
1362 this produce deadlock because llog_cleanup_commit_master wait when
1363 last llog_commit_thread exited, but this never success because was
1364 called from llog_commit_thread.
1368 Description: allow userland application know is lost one of stripes.
1369 Details : fill lvb_blocks with error code on ost and return it to
1370 application if error flag found.
1374 Description: NULL lov_tgts causing MDS oops
1375 Details : more safe checks for NULL lov_tgts for avoid oops.
1377 Severity : enhancement
1379 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
1383 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
1384 Details : If an OST is down the MDS will hang indefinitely in
1385 obd_statfs_rqset() waiting for the statfs data. While for
1386 MDS QOS usage of statfs, it should not stuck in waiting.
1388 Severity : enhancement
1390 Description: remote_acl support
1391 Details : Support ACL-based permission check for remote user.
1392 Support setfacl/getfacl for remote user with the utils
1393 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
1394 the system "{s,g}etfacl" utils.
1396 Severity : enhancement
1398 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
1400 Severity : enhancement
1402 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
1406 Description: make mgs_setparam() handle fsname containing dash
1407 Details : fsname containing a dash does not work with lctl conf_param
1409 Severity : enhancement
1411 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
1413 Severity : enhancement
1415 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
1419 Frequency : rare, at shutdown
1420 Description: access already free / zero obd_namespace.
1421 Details : if client_disconnect_export was called without force flag set,
1422 and exist connect request in flight, this can produce access to
1423 NULL pointer (or already free pointer) when connect_interpret
1424 store ocd flags in obd_namespace.
1428 Frequency : only at startup
1429 Description: not alloc memory with spinlock held.
1430 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
1431 if any spinlock held.
1433 Severity : enhancement
1435 Description: make lustre randomly fail allocating memory
1436 Details : Make lustre randomly failed allocating memory for testing purpose.
1438 Severity : enhancement
1440 Description: lost problems with lov objid file
1441 Details : Fixes some scability and access to not inited memory problems
1442 in work with lov objdid file.
1447 Description: lfs find does not continue on file error
1448 Details : Continue other files processing when a file/dir is absent.
1452 Description: Inconsistent usage of lustre_pack_reply()
1453 Details : Standardize the usage of lustre_pack_reply() such that it
1454 always generate a CERROR on failure.
1457 Frequency : occasional
1459 Description: MDS hang or stay in waiting lock
1460 Details : If client receive lock with CBPENDING flag ldlm need send lock
1461 cancel as separate rpc, to avoid situation when cancel request
1462 can't processed due all i/o threads stay in wait lock.
1466 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1470 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1474 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1478 Description: Update to RHEL4 latest kernel.
1480 Severity : enhancement
1482 Description: Build SLES10 patchless client fails
1483 Details : The configure was broken by run ./configure with
1484 --with-linux-obj=.... argument for patchless client. When the
1485 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1486 can't search header adequately. Use absolute path such as
1487 -I($LINUX)/include instead.
1491 Description: interrupt oig_wait produce painc on resend.
1492 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1493 private set, and this produce situation when rq_ptlrpcd_data not
1494 copyed to new allocated request and triggered LBUG on assert
1495 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1496 ptlrpcd set if request is changed and can be safety changed to use
1501 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1502 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1503 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1504 these macros and replace them with their difinitions instead.
1508 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1509 Details : Update which_patch & target file for SLES10 latest kernel.
1512 Description: Remove client patches from SLES 10 kernel.
1513 Details : This causes SLES 10 clients to behave as patchless clients
1514 even on a Lustre-patched (server) kernel.
1516 Severity : enhancement
1518 Description: self-adjustable client's lru lists
1519 Details : use adaptive algorithm for managing client cached locks lru
1520 lists according to current server load, other client's work
1521 pattern, memory activities, etc. Both, server and client
1522 side namespaces provide number of proc tunables for controlling
1525 Severity : enhancement
1527 Description: light-weight GSS support
1528 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
1529 and reduce performance overhead.
1531 Severity : enhancement
1533 Description: Linux keyring support
1534 Details : Support using service of Linux keyring for Lustre GSS internal
1535 context refresh/cache mechanism.
1539 Description: Fix errors in lfs documentation
1540 Details : Fixes man pages
1544 Description: don't use GFP_* in generic Lustre code.
1545 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
1549 Description: obdclass is limited by single OBD_ALLOC(idarray)
1550 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
1554 Description: warnings with build patchless client with vanila 2.6.19 and up
1555 Details : change old ctl_table style and replace ctl_table/ctl_table_header
1556 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
1560 Frequency : Only for SLES
1561 Description: sanity_quota fail test_1
1562 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
1563 file, which makes TSTID[2] inunique.
1567 Frequency : Only for SLES9
1568 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
1569 Details : Update target/ChangeLog/which_patch .
1573 Frequency : Only for test_10 in sanity-quota.sh
1574 Description: a bug in quota test code
1575 Details : checking a test flag in a wrong place
1579 Frequency : Only for test_13 in sanity-quota.sh
1580 Description: a bug in quota test script
1581 Details : shouldn't check overfull allocated quota
1585 Description: jbd statistics
1586 Details : Port older jbd statistics patch for sles10
1588 Severity : enhancement
1590 Description: Add printing inode info into message about error in writepage.
1595 Description: reply_lock_interpret crash due to race with it and lock cancel.
1596 Details : Do not replay locks that are being cancelled. Do not reference
1597 locks by their address during replay, just by their handle.
1601 Frequency : When flocks are used.
1602 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
1603 Details : Flock locks might destroy just granted lock if it could be merged
1604 with another existing flock, this is done in completion handler,
1605 so teach ldlm_cli_enquque_fini that this is a valid case for
1611 Description: Oops in read and write path when failing to allocate lock.
1612 Details : Check if lock allocation failed and return error back.
1616 Description: lstripe command fails for valid OST index
1617 Details : The stripe offset is compared to lov->desc.ld_tgt_count
1618 instead of lov->desc.ld_active_tgt_count.
1622 Description: sanity.sh failed test 103
1623 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
1628 Description: df doesn't work properly if diskfs blocksize != 4K
1629 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
1633 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
1634 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
1636 Severity : enhancement
1638 Description: Adaptive timeouts
1639 Details : RPC timeouts adapt to changing server load and network
1640 conditions to reduce resend attempts and improve recovery time.
1644 Description: llapi_file_create() does not allow some changes
1645 Details : add llapi_file_open() that allows specifying the mode and
1646 open flags, and also returns an open file handle.
1650 Description: merge and cleanup kernel patches.
1651 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
1655 Description: Add journal checksum support.(Kernel part)
1656 Details : The journal checksum feature adds two new flags i.e
1657 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
1658 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
1659 indicates that the commit block contains the checksum for
1660 the blocks described by the descriptor blocks. Now commit
1661 record can be sent to disk without waiting for descriptor
1662 blocks to be written to disk. This behavior is controlled
1663 using JBD2_FEATURE_ASYNC_COMMIT flag.
1667 Description: OSS needs mutliple precreate threads
1668 Details : Add ability to start more than one create thread per OSS.
1672 Description: Sanity.sh test_65k failed: setstripe should have succeeded
1673 Details : execute lfs setstripe on client
1677 Description: mds_obd_create error creating tmp object
1678 Details : When the user sets quota on root, llog will be affected and can't
1679 create files and write files.
1682 Frequency : Always on ia64 patchless client, and possibly others.
1684 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
1685 Details : This allows the patchless client to be loaded on architectures
1686 without this export.
1690 Description: RedHat Update kernel for RHEL5
1691 Details : Add the kernel config file for RHEL5.
1695 Description: RedHat Update kernel for RHEL5
1696 Details : Modify the kernel config file more closer RHEL5.
1700 Description: Build failure against Centos5 (RHEL5)
1701 Details : Define PAGE_SIZE when it isn't present.
1705 Description: test_7 on sanity-quota.sh can't be used on separate machines
1706 Details : add facet to handle it
1710 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
1711 Details : with NFS, the anon dentry's parent was set to itself in
1712 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
1713 corresponding dentry other than use rec->ur_name.
1715 Severity : enhancement
1717 Description: lfs setstripe enhancement
1718 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
1722 Description: enable data checksumming by default
1723 Details : enable checksum by default, allow --disable-checksum
1724 configure option and "-o nochecksum" mount option. Checksums
1725 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
1729 Description: lustre support for RHEL5
1730 Details : Add support for RHEL5.
1734 Description: Client eviction due to failover config
1735 Details : after a connection loss, the lustre client should attempt to
1736 reconnect to the last active server first before trying the
1737 other potential connections.
1741 Description: when mds and osts use different quota unit(32bit and 64bit),
1742 quota will be released repeatly.
1743 Details : void sending multiple quota reqs to mds, which will keep the
1744 status between the reqs.
1748 Description: osts not allocated evenly to files
1749 Details : change the condition to increase offset_idx
1752 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
1754 Description: Data corruption for OSTs that are formatted larger than 2TB
1756 Details : When generating the bio request for lustre file writes the
1757 sector number would overflow a temporary variable before being
1758 used for the IO. The data reads correctly from Lustre (which
1759 will overflow in a similar manner) but other file data or
1760 filesystem metadata may be corrupted in some cases.
1764 Description: Tune the kernel for good SCSI performance.
1765 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
1766 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
1771 Description: rewrite ext2-derived code in obdclass/uuid.c
1772 Details : rewrite inherited code (uuid parsing code from ext2 utils)
1773 from scratch preserving functionality.
1777 Description: "lfs find -obd UUID" prints directories
1778 Details : "lfs find -obd UUID" will return all directory names instead
1779 of just file names. It is incorrect because the directories
1780 do not reside on the OSTs.
1784 Description: use i_size_read and i_size_write in 2.6 port
1785 Details : replace inode->i_size access with i_size_read/write()
1788 Frequency : when removing large files
1790 Description: scheduling issue during removal of large Lustre files
1791 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
1792 It causes scheduling issues when removing large files (17TB in the
1796 Frequency : only with liblustre clients on XT3
1798 Description: evictions taking too long
1799 Details : allow llrd to evict clients directly on OSTs
1802 Frequency : only on ppc
1804 Description: /proc/fs/lustre/devices broken on ppc
1805 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
1806 We should make sure the type of 'index' is loff_t and then cast
1807 explicitly as needed below. Do not assign an explicitly cast
1811 Frequency : only for rhel5
1813 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
1814 Details : Modify the target file & which_kernel.
1816 Severity : enhancement
1818 Description: omit set fsid for export NFS
1819 Details : fix set/restore device id for avoid EMFILE error and mark lustre
1820 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
1824 Frequency : Always, for kernels after 2.6.16
1825 Description: Fix warning idr_remove called for id=.. which is not allocated.
1826 Details : Last kernels save old s_dev before kill super and not allow
1827 to restore from callback - restore it before call kill_anon_super.
1831 Description: buffer overruns could theoretically occur
1832 Details : llapi_semantic_traverse() modifies the "path" argument by
1833 appending values to the end of the origin string, and a
1834 overrun may occur. Adding buffer overrun check in liblustreapi.
1838 Description: Fix error on 'ls .' at the top of the Lustre mount.
1839 Details : Don't revalidate dentry if it is a root dentry.
1843 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
1844 Details : Modify vm-tunables-rhel4.patch.
1848 Description: Kernel config for 2.6.18-vanilla.
1849 Details : Modify targets/2.6-vanilla.target.in.
1850 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
1851 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
1852 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
1853 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
1857 Description: improve handling recoverable errors
1858 Details : if request processig with error which can be recoverable on server
1859 request should be resend, otherwise page released from cache and
1864 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
1865 Details : Modify target file & which_patch.
1866 A flaw was found in the IA32 system call emulation provided
1867 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
1868 value could be stored in the %RAX register, which could trigger an
1869 out-of-bounds system call table access. An untrusted local user
1870 could exploit this flaw to run code in the kernel
1871 (ie a root privilege escalation). (CVE-2007-4573).
1875 Description: O_DIRECT bypasses client statistics.
1876 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
1877 read_ahead_stats were updated. copied the stats section from
1878 osc_send_oap_rpc() into async_internal().
1882 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
1886 Description: change order of libsysio includes
1887 Details : '#include sysio.h' should always come before '#include xtio.h'
1891 Description: handle "serious error: objid * already exists" more gracefully
1892 Details : If LAST_ID value on disk is smaller than the objects existing in
1893 the O/0/d* directories, it indicates disk corruption and causes an
1894 LBUG(). If the object is 0-length, then we should use the existing
1895 object. This will help to avoid a full fsck in most cases.
1897 Severity : enhancement
1899 Description: adapt the lustre_config script to support the upgrade case
1900 Details : Add "-u" option for lustre_config script to support upgrading 1.4
1901 server targets to 1.6 in parallel.
1905 Description: To avoid grant space > avaible space when the disk is almost
1906 full. Without this patch you might see the error "grant XXXX >
1907 available" or some LBUG about grant, when the disk is almost
1909 Details : In filter_check_grant, for non_grant cache write, we should
1910 check the left space by if (*left > ungranted + bytes), instead
1911 of (*left > ungranted), because only we are sure the left space
1912 is enough for another "bytes", then the ungrant space should be
1913 increase. In client, we should update cl_avail_grant only there
1914 is OBD_MD_FLGRANT in the reply.
1918 Description: Update RHEL 4 kernel to fix local root privilege escalation.
1919 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
1920 described in CVE-2007-4573. This problem could allow untrusted
1921 local users to gain root access.
1924 Frequency : when using O_DIRECT and quotas
1926 Description: Incorrect file ownership on O_DIRECT output files
1927 Details : block usage reported by 'lfs quota' does not take into account
1928 files that have been written with O_DIRECT.
1933 Description: touch file failed when fs is not full
1934 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
1935 otherwise we can get ENOSP while fs is not full.
1939 Description: parallel lock callbacks
1940 Details : Instead of sending blocking and completion callbacks as separated
1941 requests, adding them to a set and sending in parallel.
1944 Frequency : only for Cray XT3
1945 Bugzilla : 12829/13455
1946 Description: Changing primary group doesn't change the group lustre assigns to
1948 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
1949 overridden with the primary group provided by the group upcall,
1950 whereas the supplied fsgid can be trusted if it is in the list of
1951 supplementary groups returned by the group upcall.
1953 Severity : enhancement
1955 Description: Allow masking D_WARNING, D_ERROR messages from console
1956 Details : Console messages can now be disabled via lnet.printk.
1960 Description: User code with malformed file open parameter crashes client node
1961 Details : Before packing join_file req, all the related reference should be
1962 checked carefully in case some malformed flags cause fake
1963 join_file req on client.
1967 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
1968 Details : ldlm_completion_ast() assumes that a lock is granted when the req
1969 mode is equal to the granted mode. However, it should also check
1970 that LDLM_FL_CP_REQD is not set.
1974 Description: Heavy nfs access might result in deadlocks
1975 Details : After ELC code landed, it is now improper to enqueue any mds
1976 locks under och_sem, because enqueue might want to decide to
1977 cancel open locks for same inode we are holding och_sem for.
1981 Description: Client eviction while running blogbench
1982 Details : A lot of unlink operations with concurrent I/O can lead to a
1983 deadlock causing evictions. To address the problem, the number of
1984 oustanding OST_DESTROY requests is now throttled to
1985 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
1986 callbacks are processed in priority.
1990 Description: enable ACLs on MDS by default
1991 Details : ACLs must be enabled on MDS by default.
1994 Frequency : PPC/PPC64 only
1996 Description: conflicts between asm-ppc64/types.h and lustre_types.h
1997 Details : fix duplicated definitions between asm-ppc64/types.h and
1998 lustre_types.h on PPC.
2001 Frequency : PPC/PPC64 only
2003 Description: asm-ppc/segment.h does not exist
2004 Details : fix compile issue on PPC.
2008 Description: better handle error messages in extents code
2011 Frequency : RHEL4 only
2013 Description: mkfs is very slow on IA64/RHEL4
2014 Details : A performance regression has been discovered in the MPT Fusion
2015 driver between versions 3.02.73rh and 3.02.99.00rh. As a
2016 consequence, we have downgraded the MPT Fusion driver in the RHEL4
2017 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
2019 Severity : enhancement
2021 Description: SNMP support enhancement
2022 Details : Adding total number of sampled request for an MDS node in snmp
2025 Severity : enhancement
2027 Description: Optimize ldlm waiting list processing for PR extent locks
2028 Details : When processing waiting list for read extent lock and meeting read
2029 lock that is same or wider to it that is not contended, skip
2030 processing rest of the list and immediatelly return current
2031 status of conflictness, since we are guaranteed there are no
2032 conflicting locks in the rest of the list.
2036 Description: Time out and refuse to reconnect
2037 Details : When the failover node is the primary node, it is possible
2038 to have two identical connections in imp_conn_list. We must
2039 compare not conn's pointers but NIDs, otherwise we can defeat
2040 connection throttling.
2044 Description: port llog fixes from b1_6 into HEAD
2045 Details : Port llog reference couting and some llog cleanups from b1_6
2046 (bug 10800) into HEAD, for protect from panic and access to already
2047 free llog structures.
2051 Description: Detect stride IO mode in read-ahead
2052 Details : When a client does stride read, read-ahead should detect that and
2053 read-ahead pages according to the detected stride pattern.
2057 Description: data checksumming impacts single node performance
2058 Details : add support for several checksum algorithm. Currently, only CRC32
2059 and Adler-32 are supported. The checksum type can be changed on
2060 the fly via /proc/fs/lustre/osc/*/checksum_type.
2064 Description: use adler32 for page checksums
2065 Details : when available, use the Adler-32 algorithm instead of CRC32 for
2070 Description: build for x2 fails
2071 Details : fix compile issue on Cray systems.
2075 Description: Properly match for duplicate locks
2076 Details : Due to different lock order from skiplists code, we need to
2077 traverse entire list for now
2080 Frequency : only on PPC/SLES10
2082 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
2083 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
2084 be defined. Add a hack in mkfs_lustre.c to work around this compile
2089 Description: LASSERT on MDS when client holding flock lock dies
2090 Details : ldlm pool logic depends on number of granted locks equal to
2091 number of released locks which is not true for flock locks, so
2092 just exclude such locks from consideration.
2096 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
2097 Details : Use fsfilt_sync() for both the whole filesystem sync and
2098 individual file sync to eliminate dangerous inode locking
2099 with I_LOCK that can lead to a deadlock.
2103 Description: performance in 1.6.3
2104 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
2105 sure that 1MB requests are not fragmented by the block layer.
2107 Severity : enhancement
2109 Description: organize the server-side client stats on per-nid basis
2110 Details : Change the structure of stats under obdfilter and mds to
2118 The "uuid"s file would list the uuids of _active_ exports.
2119 And the clear entry is to clear all stats and stale nids.
2121 Severity : enhancement
2123 Description: eliminate client locks in face of contention
2124 Details : file contention detection and lockless i/o implementation
2125 for contended files.
2129 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
2132 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
2134 Description: MDS may lose file striping (and hence file data) in some cases
2135 Details : If there are additional extended attributes stored on the MDS,
2136 in particular ACLs, SELinux, or user attributes (if user_xattr
2137 is specified for the client mount options) then there is a risk
2138 of attribute loss. Additionally, the Lustre file striping
2139 needs to be larger than default (e.g. striped over all OSTs),
2140 and an additional attribute must be stored initially in the
2141 inode and then increase in size enough to be moved to the
2142 external attribute block (e.g. ACL growing in size) for file
2147 Description: skiplist implementation simplification
2148 Details : skiplists are used to group compatible locks on granted list
2149 that was implemented as tracking first and last lock of each
2150 lock group the patch changes that to using doubly linked lists
2154 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
2155 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
2156 because LASSERT on both the data supplied by a client, and the
2157 data on disk is dangerous and incorrect.
2159 Severity : enhancement
2161 Description: Slow truncate/writes to huge files at high offsets.
2162 Details : Directly associate cached pages to lock that protect those pages,
2163 this allows us to quickly find what pages to write and remove
2164 once lock callback is received.
2168 Description: more ldlm soft lockups
2169 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
2170 starve other threads from the resource lock for a long time in
2171 case of long waiting queue, so change the debug level from
2172 D_OTHER to the less frequently used D_INFO.
2174 Severity : enhancement
2176 Description: add -gid, -group, -uid, -user options to lfs find
2180 Description: Hung threads in invalidate_inode_pages2_range
2181 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
2182 once one is completed. As a result, some RPCs are stuck in the
2183 queue and are never sent.
2187 Description: filter threads hungs on waiting journal commit
2188 Details : Cleanup filter group llog code, then only filter group llog will
2189 be only created in the MDS/OST syncing process.
2193 Description: Procfs and llog threads access destoryed import sometimes.
2194 Details : Sync the import destoryed process with procfs and llog threads by
2195 the import refcount and semaphore.
2197 Severity : enhancement
2199 Description: openlock cache of b1_6 port to HEAD
2204 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
2205 Details : If insertion of an extent fails, then discard the inode
2206 preallocation and free data blocks else it can lead to duplicate
2211 Description: don't always update ctime in ext3_xattr_set_handle()
2212 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
2213 In some cases the ctime should not be updated, for example for
2214 2.0->1.8 compatibility it is necessary to delete an xattr and it
2215 should not update the ctime.
2219 Bugzilla : 15713/16362
2220 Description: Assertion in iopen_connect_dentry in 1.6.3
2221 Details : looking up an inode via iopen with the wrong generation number can
2222 populate the dcache with a disconneced dentry while the inode
2223 number is in the process of being reallocated. This causes an
2224 assertion failure in iopen since the inode's dentry list contains
2225 both a connected and disconnected dentry.
2229 Description: assertion failure in ldlm_handle2lock()
2230 Details : fix a race between class_handle_unhash() and class_handle2object()
2231 introduced in lustre 1.6.5 by bug 13622.
2236 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
2237 Details : remove the limit of 256 scsi disks in the sd_iostat patch
2242 Description: oops in sd_iostats_seq_show()
2243 Details : unloading/reloading the scsi low level driver triggers a kernel
2244 bug when trying to access the sd iostat file.
2249 Description: Kernel panics during QLogic driver reload
2250 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
2251 patch, causing memory corruption.
2256 Description: journal_dev option does not work in b1_6
2257 Details : pass mount option during pre-mount.
2259 Severity : enhancement
2261 Description: Add a FIEMAP(FIle Extent MAP) ioctl
2262 Details : FIEMAP ioctl will allow an application to efficiently fetch the
2263 extent information of a file. It can be used to map logical blocks
2264 in a file to physical blocks in the block device.
2268 Description: LDLM soft lockups - improvement
2269 Details : It is be possible to send the lock handle along with each read
2270 or write request because the client is already doing a lock match
2271 itself so there isn't any reason the OST should have to re-do that
2275 Frequency : only on Cray X2
2277 Description: X2 build failures
2278 Details : fix build failures on Cray X2.
2282 Description: xid & resent requests
2283 Details : Initialize RPC XID from clock at startup (randomly if clock is
2286 Severity : enhancement
2288 Description: Add lustre_start utility to start or stop multiple Lustre servers
2293 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
2294 Details : In case of memory pressure, list_del() can be called twice on
2295 req->rq_history_list, causing a kernel oops.
2299 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
2300 Details : fix stack overflow in the distributed lock manager by defering
2301 export eviction after a failed AST to the elt thread instead of
2302 handling it in the dlm interpret routine.
2306 Description: Convert some comments to new format.
2307 Details : Update documenting comments to match doxygen conventions.
2311 Description: Grammar fixes.
2312 Details : A couple of trivial spelling fixes.
2316 Description: OSD_COUNTERS-mandatory
2317 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
2318 to simplify the code.
2322 Description: simplify cmm_device freeing logic.
2323 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
2327 Description: Add lockdep support to dt_object_operations locking interface.
2328 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
2329 parameter indicating lock ordering. Update mdd code to use new
2334 Description: Introduce failloc constants for lockless IO tests.
2335 Details : Add two new failloc constants to test lockless IO. Only one of
2336 them in implemented---another is checked in yet to be landed
2341 Description: Add lockdep support for inode mutex.
2342 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
2343 in the situations where Lustre has to lock more than one inode
2348 Description: Add optional invariants checking support.
2349 Details : Add new LINVRNT() macro, optional on new --enable-invariants
2350 configure switch. This macro is to be used for consistency and
2351 sanity checks that are too expensive to be left in `production'
2356 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
2357 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
2362 Description: Add lockdep support for ldlm_lock and ldlm_resource.
2363 Details : Use spin_lock_nested() in (the only) situation where more than
2364 one ldlm_lock is locked simultaneously. Also, fix possible
2365 dead-lock in ldlm_lock_change_resource() by enforcing particular
2370 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
2371 Details : Instead of specifying each ldlm_lock call-back through separate
2372 parameter, wrap them into struct ldlm_callback_suite.
2376 Description: Kill join_lru obd method and its callers.
2377 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
2378 in memory---a requirement implemented through ->o_join_lru() obd
2379 method in HEAD. Remove this method and its users.
2383 Description: Add asynchronous ldlm ENQUEUE completion handler.
2384 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
2385 case that stock ldlm_completion_ast() cannot handle as it waits
2386 until lock is granted. Introduce new ldlm_completion_ast_async()
2387 for this. Also comment ldlm_completion_ast().
2391 Description: ldlm_error <-> errno conversion.
2392 Details : Add functions to map (rather arbitrary) between LDLM error codes
2393 and standard errno values. CLIO needs this to prevent LDLM specific
2394 constants from escaping ldlm and osc.
2398 Description: Kill unused ldlm_handle2lock_ns() function.
2399 Details : Kill unused ldlm_handle2lock_ns() function.
2403 Description: Add lu_ref support to ldlm_lock
2404 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
2405 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
2406 and ldlm_resource. LDLM interface has to be changed, because code that
2407 releases a reference on a lock, has to "know" what reference this is.
2408 In the most frequent case
2410 lock = ldlm_handle2lock(handle);
2412 LDLM_LOCK_PUT(lock);
2414 no changes are required. When any other reference (received _not_ from
2415 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
2416 instead of LDLM_LOCK_PUT().
2418 Arguably, changes are pervasive, and interface requires some discipline
2419 for proper use. On the other hand, it was very instrumental in finding
2420 a few leaked lock references.
2424 Description: Add ldlm_lock_addref_try().
2425 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
2426 attempts to addref a lock that might be being canceled
2431 Description: Add ldlm_weigh_callback().
2432 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
2433 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
2434 hard-coded `number of pages' logic.
2438 Description: Add lockdep annotations to llog code.
2439 Details : Use appropriately tagged _nested() locking calls in the places
2440 where llog takes more than one ->lgh_lock lock.
2444 Description: Add loi_kms_set().
2445 Details : Wrap kms updates into a helper function.
2449 Description: Constify instances of struct lsm_operations.
2450 Details : Constify instances of struct lsm_operations.
2454 Description: lu_conf support.
2455 Details : On a server, a file system object is uniquely identified
2456 by a fid, which is sufficient to locate and load all object
2457 state (inode). On a client, on the other hand, more data are
2458 necessary instantiate an object. Change lu_object_find() and
2459 friends to take additional `lu_conf' argument describing object.
2460 Typically this includes layout information.
2464 Description: lu_context fixes.
2465 Details : Introduce new lu_context functions that are needed on the client
2466 side, where some system threads (ptlrpcd) are shared by multiple
2467 modules, and so cannot be stopped during module shutdown.
2471 Description: Add start and stop methods to lu_device_type_operations.
2472 Details : Introduce two new methods in lu_device_type_operations, that are
2473 invoked when first instance of a given type is created and last one
2474 is destroyed respectively. This is need by CLIO.
2478 Description: Add lu_ref support to struct lu_device.
2479 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
2480 track leaked references.
2484 Description: Introduce lu_kmem_descr.
2485 Details : lu_kmem_descr and its companion interface allow to create
2486 and destroy a number of kmem caches at once.
2490 Description: Fix lu_object finalization race.
2491 Details : Fix a race between lu_object_find() finding an object and its
2492 concurrent finalization. This race is (most likely) not possible
2493 on the server, but might happen on the client.
2497 Description: Introduce lu_ref interface.
2498 Details : lu_ref is a debugging module allowing to track references to
2499 a given object. It is quite cpu expensive, and has to be
2500 explicitly enabled with --enable-lu_ref. See usage description
2505 Description: Factor lu_site procfs stats into a separate function.
2506 Details : Separate lu_site stats printing code into a separate function
2507 that can be reused on a client.
2511 Description: Constify instances of struct {lu,dt,md}_device_operations.
2512 Details : Constify instances of struct {lu,dt,md}_device_operations.
2516 Description: Introduce struct md_site and move meta-data specific parts of
2517 struct lu_site here.
2518 Details : Move md-specific fields out of struct lu_site into special struct
2519 md_site, so that lu_site can be used on a client.
2523 Description: Kill mdd_lov_destroy().
2524 Details : Remove unused mdd code.
2528 Description: Add st_block checking to multistat.c.
2529 Details : Add st_block checking to multistat.c.
2533 Description: Add lu_ref support to struct obd_device.
2534 Details : Add lu_ref tracking to obd_device.
2538 Description: Kill obd_set_fail_loc().
2539 Details : Remove unused code.
2543 Description: Add special type for ptlrpc_request interpret functions.
2544 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
2545 there. Actual usage will be in CLIO.
2549 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
2550 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
2551 doesn't work with lockdep.
2555 Description: Add rwv.c test program.
2556 Details : New testing program exercising readv(2) and writev(2) (Qian).
2560 Description: Add sendfile.c test program.
2561 Details : New testing program exercising sendfile(2) (Jay).
2565 Description: Ratelimit a message that can be very frequent.
2566 Details : Ratelimit a memory allocation failure message that can
2571 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2572 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2576 Description: Liblustre build fixes.
2577 Details : Liblustre build fixes.
2581 Description: libcfs: add cfs_{need,cond}_resched() interface.
2582 Details : libcfs: add cfs_{need,cond}_resched() definition and
2583 implementations for Linux, NT, and liblustre.
2585 Severity : enhancement
2587 Description: More exported tunables for mballoc
2588 Details : Add support for tunable preallocation window and new tunables for
2589 large/small requests
2593 Description: Detect corruption of block bitmap and checking for preallocations
2594 Details : Checks validity of on-disk block bitmap. Also it does better
2595 checking of number of applied preallocations. When corruption is
2596 found, it turns filesystem readonly to prevent further corruptions.
2600 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
2601 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
2602 window to 0. If it does hit cache miss, and read-pattern is still
2603 stride-io mode, does not reset the stride window, but also does not
2604 increase the stride window length in this case.
2608 Frequency : only for big-endian servers
2609 Description: Check if system is big-endian while mounting fs with extents feature
2610 Details : Mounting a filesystem with extents feature will fail on big-endian
2611 systems since ext3-based ldiskfs is not supported on big-endian
2612 systems. This can be over-riden with "bigendian_extents" mount option.
2614 Severity : enhancement
2616 Description: The root squash functionality
2617 Details : A security feature, which is to prevent users from being able
2618 to mount lustre on their desktop, run as root, and delete
2619 all of the files in the filesystem. The goal is accomplished by
2620 remapping user id (UID) and group id (GID) of the root user to
2621 a UID and GID specified by the system administartor via Lustre
2622 configuration management server (MGS). The functionality also
2623 allows to specify sets of clients for which the remapping does
2628 Description: Excessive recovery window
2629 Details : With AT enabled, the recovery window can be excessively long (6000+
2630 seconds). To address this problem, we no longer use
2631 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
2632 timeout no longer depends on the service time, it is set to
2633 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
2634 time via pb_service_time.
2638 Description: Watchdog triggered on MDS failover
2639 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
2640 the OSTs know that the MDS "UUID" can be reused for the same export
2641 from a different NID, so we do not need to wait for the export to be
2645 Frequency : rare, only if using MMP with Linux RAID
2647 Description: MMP doesn't work with Linux RAID
2648 Details : While using HA for Lustre servers with Linux RAID, it is possible
2649 that MMP will not detect multiple mounts. To make this work we
2650 need to unplug the device queue in RAID when the MMP block is being
2651 written. Also while reading the MMP block, we should read it from
2652 disk and not the cached one.
2654 Severity : enhancement
2656 Description: open file using fid
2657 Details : A file can be opened using just its fid, like
2658 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
2661 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
2664 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
2665 Details : add proper locking for creation of jbd_4k slab cache
2669 Description: MMP check in ext3_remount() fails without displaying any error
2670 Details : When multiple mount protection fails during remount, proper error
2673 Severity : enhancement
2675 Description: Allow stripe size to be up to 4G-64k
2676 Details : Fix math logic to allow large stripe sizes.
2680 Description: add check for >8TB ldiskfs filesystems
2681 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
2682 Don't allow >8TB ldiskfs filesystems to be mounted without
2683 force_over_8tb mount option
2688 Descriptoin: OSS complains "too many blocks" and crashes
2689 Details : In filter_commitrw_write():
2690 1. lquota_chkquota();
2691 2. obdo_from_inode();
2692 3. lquota_pending_commit();
2693 "2." has changed uid and gid in this case, then caused
2694 lquota_chkquota() and lquota_pending_commit() to use
2695 different uid and gid. That is the root of the bug.
2699 Descriptoin: getstripe doesn't report stripe count = -1 sanely
2700 Details : Detect and print out the 'stripe count = -1' case properly.
2704 Description: Out or order replies might be lost on replay
2705 Details : In ptlrpc_retain_replayable_request if we cannot find retained
2706 request with tid smaller then one currently being added, add it
2707 to the start, not end of the list.
2709 --------------------------------------------------------------------------------
2711 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
2713 * Support for kernels:
2714 2.6.5-7.283 (SLES 9),
2715 2.6.9-55.EL (RHEL 4),
2716 2.6.16.46-0.14 (SLES 10),
2717 2.6.18.8 vanilla (kernel.org)
2718 * Client support for unpatched kernels:
2719 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2720 2.6.16 - 2.6.22 vanilla (kernel.org)
2721 * Due to recently discovered recovery problems, we do not recommend
2722 using patchless RHEL 4 clients with this or any earlier release.
2723 * Recommended e2fsprogs version: 1.39.cfs8
2724 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2725 * Starting with this release, the ldiskfs backing filesystem required
2726 by Lustre is now in its own package, lustre-ldiskfs. This package
2727 should be installed. It is versioned separately from Lustre and
2728 may be released separately in future.
2732 Description: block reactivating mgc import until all deactivates complete
2733 Details : Fix race when failing back MDT/MGS to itself (testing)
2736 Frequency : at statup only
2738 Description: mds_lov_synchronize race leads to various problems
2739 Details : simultaneous MDT->OST connections at startup can cause the
2740 sync to abort, leaving the OSC in a bad state.
2742 Severity : enhancement
2744 Description: add optional extra BUILD_VERSION info
2745 Details : add a new environment variable (namely LUSTRE_VERS) which allows
2746 to override the lustre version.
2749 Frequency : 2.6.18 servers only
2751 Description: ll_kern_mount() doesn't release the module reference
2752 Details : The ldiskfs module reference count never drops down to 0
2753 because ll_kern_mount() doesn't release the module reference.
2758 Description: server LBUG when using old ost_num_threads parameter
2759 Details : Accept the old ost_num_threads parameter but warn that it
2760 is deprecated, and fix an off-by-one error that caused an LBUG.
2765 Description: Transient SCSI error results in persistent IO issue
2766 Details : iobuf->dr_error is not reinitialized to 0 between two
2770 Frequency : sometimes when underlying device returns I/O errors
2772 Description: OSTs not going read-only during write failures
2773 Details : OSTs are not remounted read-only when the journal commit threads
2774 get I/O errors because fsfilt_ext3 calls journal_start/stop()
2775 instead of the ext3 wrappers.
2779 Description: poor connect scaling with increasing client count
2780 Details : Don't run filter_grant_sanity_check for more than 100 exports
2781 to improve scaling for large numbers of clients.
2784 Frequency : SLES10 only
2786 Description: sanity-quota.sh quotacheck failed: rc = -22
2787 Details : Quotas cannot be enabled on SLES10.
2790 Frequency : liblustre clients only
2792 Description: getdirentries does not give error when run on compute nodes
2793 Details : getdirentries does not fail when the size specified as an argument
2794 is too small to contain at least one entry
2796 Severity : enhancement
2798 Description: Add LNET router traceability for debug purposes
2799 Details : If a checksum failure occurs with a router as part of the
2800 IO path, the NID of the last router that forwarded the bulk data
2801 is printed so it can be identified.
2806 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
2807 Details : Due to a race condition, liblustre clients were occasionally
2808 evicted incorrectly.
2810 Severity : enhancement
2812 Description: lfs setstripe use optional parameters instead of postional
2815 Severity : enhancement
2817 Description: Nanosecond timestamp support for ldiskfs
2818 Details : The on-disk ldiskfs filesystem has added support for nanosecond
2819 resolution timestamps. There is not yet support for this at
2820 the Lustre filesystem level.
2823 Frequency : during server recovery
2825 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
2826 Details : request with rq_no_resend flag not awake l_wait_event if they get
2830 Frequency : nfs export on patchless client
2832 Description: connectathon hang when test nfs export over patchless client
2833 Details : Disconnected dentry cannot be found with lookup, so we do not need
2834 to unhash it or make it invalid
2837 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
2838 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
2839 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
2844 Description: open req refcounting wrong on reconnect
2845 Details : If reconnect happened between getting open reply from server and
2846 call to mdc_set_replay_data in ll_file_open, we will schedule
2847 replay for unreferenced request that we are about to free.
2848 Subsequent close will crash in variety of ways.
2849 Check that request is still eligible for replay in
2850 mdc_set_replay_data().
2855 Description: disable writes to filesystem when reading health_check file
2856 Details : the default for reading the health_check proc file has changed
2857 to NOT do a journal transaction and write to disk, because this
2858 can cause reads of the /proc file to hang and block HA state
2859 checking on a healthy but otherwise heavily loaded system. It
2860 is possible to return to the previous behaviour during configure
2861 with --enable-health-write.
2863 Severity : enhancement
2865 Description: 64-bit inode version
2866 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
2867 to the inode. This will be required for version-based recovery.
2872 Description: MDS fails to start if a duplicate client export is detected
2873 Details : in some rare cases it was possible for a client to connect to
2874 an MDS multiple times. Upon recovery the MDS would detect this
2875 and fail during startup. Handle this more gracefully.
2877 Severity : enhancement
2879 Description: Add -o localflock option to simulate old noflock
2881 Details : This will achieve local-only flock/fcntl locks
2887 Description: log_commit_thread vs filter_destroy race leads to crash
2888 Details : Take import reference before releasing llog record semaphore
2893 Description: Wrong request locking in request set processing
2894 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
2895 imp_delayed_list, in this place should be used imp_lock.
2898 Frequency : when reconnection
2900 Description: Grant Leak when osc reconnect to OST
2901 Details : When osc reconnect ost, OST(filter) should check whether it
2902 should grant more space to client by comparing fed_grant and
2903 cl_avail_grant, and return the granted space to client instead
2904 of "new granted" space, because client will call osc_init_grant
2905 to update the client grant space info.
2908 Frequency : when client reconnect to OST
2910 Description: Grant Leak when osc do resend and replay bulk write
2911 Details : When osc reconnect to OST, OST(filter)should clear grant info of
2912 bulk write request, because the grant info while be sync between
2913 OSC and OST when reconnect, and we should ignore the grant info
2914 these of resend/replay write req.
2919 Description: Grant space more than avaiable left space sometimes.
2920 Details : When then OST is about to be full, if two bulk writing from
2921 different clients came to OST. Accord the avaliable space of the
2922 OST, the first req should be permitted, and the second one
2923 should be denied by ENOSPC. But if the seconde arrived before
2924 the first one is commited. The OST might wrongly permit second
2925 writing, which will cause grant space > avaiable space.
2928 Frequency : when client is evicted
2930 Description: Grant might be wrongly erased when osc is evicted by OST
2931 Details : when the import is evicted by server, it will fork another
2932 thread ptlrpc_invalidate_import_thread to invalidate the
2933 import, where the grant will be set to 0. While the original
2934 thread will update the grant it got when connecting. So if
2935 the former happened latter, the grant will be wrongly errased
2936 because of this race.
2941 Description: Checking Stale with correct fid
2942 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
2943 to check whether it is stale, because sometimes, we want the
2944 enqueue happened anyway, and op_data.fid2 will not be initialized.
2946 Severity : enhancement
2948 Description: update patchless client
2949 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
2952 Frequency : only with 2.4 kernel
2954 Description: random memory corruption
2955 Details : size of struct ll_inode_info is to big for union inode.u and this
2956 can be cause of random memory corruption.
2961 Description: Memory leak in recovery
2962 Details : Lov_mds_md was not free in an error handler in mds_create_object.
2963 It should also check obd_fail before fsfilt_start, otherwise if
2964 fsfilt_start return -EROFS,(failover mds during mds recovery).
2965 then the req will return with repmsg->transno = 0 and rc = EROFS.
2966 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
2967 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
2968 be freed no matter whether fsfilt_commit success or not.
2971 Frequency : only with huge count clients
2973 Description: Prevents from taking the superblock lock in llap_from_page for
2975 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
2976 for llap_from_page call in ll_removepage prevents from taking the
2977 superblock lock for a soon died page.
2982 Description: Not check open intent error before release open handle
2983 Details : in some rare cases, the open intent error is not checked before
2984 release open handle, which may cause
2985 ASSERTION(open_req->rq_transno != 0), because it tries to release
2986 the failed open handle.
2991 Description: Set cat log bitmap only after create log success.
2992 Details : in some rare cases, the cat log bitmap is set too early. and it
2993 should be set only after create log success.
2997 Description: Accessing a block bevice can re-enable I/O when Lustre is
2998 tearing down a device.
2999 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
3004 Description: service threads may hog cpus when there are a lot of requests
3005 Details : Insert cond_resched to give other threads a chance to use some CPU
3010 Description: the cat log was not initialized in recovery
3011 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
3012 unlink log on mds will not be initialized until mds post
3013 recovery. And also in mds post recovery, the unlink log will
3014 initialization will be done asynchronausly, so there will be race
3015 between add unlink log and unlink log initialization.
3019 Description: brw_stats were being printed incorrectly
3020 Details : brw_stats were being printed as log2 but all of them were not
3021 recorded as log2. Also remove some code duplication arising from
3022 filter_tally_{read,write}.
3026 Frequency : rare, only in recovery.
3027 Description: ASSERTION(req->rq_type != LI_POISON) failed
3028 Details : imp_lock should be held while iterating over imp_sending_list for
3029 prevent destroy request after get timeout in ptlrpc_queue_wait.
3033 Description: replay-single.sh test 52 fails
3034 Details : A lock's skiplist need to be cleanup when it being unlinked
3035 from its resource list.
3039 Description: Short directio read returns full requested size rather than
3041 Details : Direct I/O operations should return actual amount of bytes
3042 transferred rather than requested size.
3044 Severity : enhancement
3046 Description: metadata RPC reduction (e.g. for rm performance)
3047 Details : decrease the amount of synchronous RPC between clients and servers
3048 by canceling conflicing lock before the operation on the client
3049 and packing thier handles into the main operation RPC to server.
3051 Severity : enhancement
3053 Description: Async OSC create to avoid the blocking unnecessarily.
3054 Details : If a OST has no remain object, system will block on the creating
3055 when need to create a new object on this OST. Now, ways use
3056 pre-created objects when available, instead of blocking on an
3057 empty osc while others are not empty. If we must block, we block
3058 for the shortest possible period of time.
3062 Description: improve handling recoverable errors
3063 Details : if request processig with error which can be recoverable on server
3064 request should be resend, otherwise page released from cache and
3067 Severity : enhancement
3069 Description: refine locking for avoid write wrong info into lov_objid file
3070 Details : fix possible races with add new target and write/update data in
3073 --------------------------------------------------------------------------------
3075 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
3080 Frequency : on some architectures
3082 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
3083 Details : Uninitialized flags sometimes cause configuration commands to
3087 Frequency : patchless clients only
3089 Description: missing __iget() symbol export
3090 Details : The __iget() symbol export is missing. To avoid the need for
3091 this on patchless clients the deathrow inode reaper is turned
3092 off, and we depend on the VM to clean up old inodes. This
3093 dependency was during via the fix for bug 12181.
3095 --------------------------------------------------------------------------------
3097 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
3099 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
3100 INTEROPERATE with older versions automatically. In many cases a
3101 special upgrade step is needed. Please read the
3102 user documentation before upgrading any part of a live system.
3103 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
3104 version will not interoperate with 1.6 betas before beta5 (1.5.95).
3105 * WARNING: Lustre configuration and startup changes are required with
3106 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
3110 Severity : enhancement
3112 Description: MountConf
3113 Details : Lustre configuration is now managed via mkfs and mount
3114 commands instead of lmc and lconf. New obd types (MGS, MGC)
3115 are added for dynamic configuration management. See
3116 https://mail.clusterfs.com/wikis/lustre/MountConf for
3119 Severity : enhancement
3121 Description: dynamic OST addition
3122 Details : OSTs can now be added to a live filesystem
3124 Severity : enhancement
3126 Description: startup order invariance
3127 Details : MDTs and OSTs can be started in any order. Clients only
3128 require the MDT to complete startup.
3130 Severity : enhancement
3132 Description: parallel, asynchronous orphan cleanup
3133 Details : orphan cleanup is now performed in separate threads for each
3134 OST, allowing parallel non-blocking operation.
3136 Severity : enhancement
3138 Description: optimized stripe assignment
3139 Details : stripe assignments are now made based on ost space available,
3140 ost previous usage, and OSS previous usage, in order to try
3141 to optimize storage space and networking resources.
3143 Severity : enhancement
3145 Description: Permanently set tunables
3146 Details : All writable /proc/fs/lustre tunables can now be permanently
3147 set on a per-server basis, at mkfs time or on a live system.
3149 Severity : enhancement
3151 Description: Lustre message v2
3152 Details : Add lustre message format v2.
3154 Severity : enhancement
3156 Description: client OST exclusion list
3157 Details : Clients can be started with a list of OSTs that should be
3158 declared "inactive" for known non-responsive OSTs.
3162 Description: SPEC SFS validation failure on NFS v2 over lustre.
3163 Details : Changes the blocksize for regular files to be 2x RPC size,
3164 and not depend on stripe size.
3166 Severity : enhancement
3168 Description: Multiple MD RPCs in flight.
3169 Details : Further unserialise some read-only MDS RPCs - learn about intents.
3170 To avoid overly-overloading MDS, introduce a limit on number of
3171 MDS RPCs in flight for a single client and add /proc controls
3172 to adjust this limit.
3174 Severity : enhancement
3176 Description: client read/write statistics
3177 Details : Add client read/write call usage stats for performance
3178 analysis of user processes.
3179 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
3180 file access. extents_stats shows chunk size distribution.
3181 extents_stats_per_process show chunk size distribution per
3184 Severity : enhancement
3186 Description: mds statistics
3187 Details : Add detailed mds operations statistics in
3188 /proc/fs/lustre/mds/*/stats.
3192 Description: Failure of copying files with lustre special EAs.
3193 Details : Client side always return success for setxattr call for lustre
3194 special xattr (currently only "trusted.lov").
3199 Description: Refcount LNET uuids
3200 Details : The global LNET uuid list grew linearly with every startup;
3201 refcount repeated list entries instead of always adding to
3204 Severity : enhancement
3206 Description: Dynamic service threads
3207 Details : Within a small range, start extra service threads
3208 automatically when the request queue builds up.
3211 Frequency : mixed-endian client/server environments
3213 Description: mixed-endian crashes
3214 Details : The new msg_v2 system had some failures in mixed-endian
3217 Severity : enhancement
3219 Description: Easy OST removal
3220 Details : OSTs can be permanently deactivated with e.g. 'lctl
3221 conf_param lustre-OST0001.osc.active=0'
3223 Severity : enhancement
3225 Description: MGS proc entries
3226 Details : Added basic proc entries for the MGS showing what filesystems
3229 Severity : enhancement
3231 Description: provide MGS failover
3232 Details : Added config lock reacquisition after MGS server failover.
3234 Severity : enhancement
3236 Description: add Linux 2.4 support
3237 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
3241 Description: a large application tries to do I/O to the same resource and dies
3242 in the middle of it.
3243 Details : Check the req->rq_arrival time after the call to
3244 ost_brw_lock_get(), but before we do anything about
3245 processing it & sending the BULK transfer request. This
3246 should help move old stale pending locks off the queue as
3247 quickly as obd_timeout.
3250 Frequency : when an incorrect nid is specified during startup
3252 Description: ptlrpc connect to non-existant node causes kernel crash
3253 Details : LNET can't be re-entered from an event callback, which
3254 happened when we expire a message after the export has been
3255 cleaned up. Instead, hand the zombie cleanup off to another
3258 Severity : enhancement
3260 Description: plain/inodebits lock performance improvement
3261 Details : Grouping plain/inodebits in granted list by their request modes
3262 and bits policy, thus improving the performance of search through
3266 Frequency : only if OST filesystem is corrupted
3268 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3269 Details : for a short time RPCs with bulk IO are in the replay list,
3270 but replay of bulk IOs is unimplemented. If the OST filesystem
3271 is corrupted due to disk cache incoherency and then replay is
3272 started it is possible to trip an assertion. Avoid putting
3273 committed RPCs into the replay list at all to avoid this issue.
3276 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3278 Description: System hang on startup
3279 Details : This bug allowed the liblustre (e.g. catamount) client to
3280 return to the app before handling all startup RPCs. This
3281 could leave the node unresponsive to lustre network traffic
3282 and manifested as a server ptllnd timeout.
3284 Severity : enhancement
3286 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3287 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
3288 boolean to control whether to print peer debug info when a
3289 client's RPC times out.
3292 Frequency : only for kernels with patches from Lustre below 1.4.3
3294 Description: Remove old rdonly API
3295 Details : Remove old rdonly API which unused from at least lustre 1.4.3
3298 Frequency : only for devices with external journals
3300 Description: Set external device read-only also
3301 Details : During a commanded failover stop, we set the disk device
3302 read-only while the server shuts down. We now also set any
3303 external journal device read-only at the same time.
3306 Frequency : when upgrading from 1.4 while trying to change parameters
3308 Description: The wrong (new) MDC name was used when setting parameters for
3309 upgraded MDT's. Also allows changing of OSC (and MDC)
3310 parameters if --writeconf is specified at tunefs upgrade time.
3313 Frequency : when setting specific ost indicies
3315 Description: QOS code breaks on skipped indicies
3316 Details : Add checks for missing OST indicies in the QOS code, so OSTs
3317 created with --index need not be sequential.
3319 Severity : enhancement
3321 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3322 Details : The uninit_groups feature works in conjunction with the kernel
3323 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3324 up the pass1 processing of e2fsck. This is a read-only feature
3325 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3326 mount filesystems that have had this feature enabled.
3328 Severity : enhancement
3330 Description: Improve multi-block allocation algorithm to avoid fragmentation
3331 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3332 allocation locality and avoid filesystem fragmentation.
3334 ------------------------------------------------------------------------------
3336 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
3338 * Support for kernels:
3339 2.4.21-47.0.1.EL (RHEL 3)
3340 2.6.5-7.283 (SLES 9)
3341 2.6.9-42.0.10.EL (RHEL 4)
3342 2.6.12.6 vanilla (kernel.org)
3343 2.6.16.27-0.9 (SLES 10)
3344 * Recommended e2fsprogs version: 1.39.cfs5
3346 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3350 Frequency : occasional, depends on client load and configuration
3351 Bugzilla : 12181, 12203
3352 Description: data loss for recently-modified files
3354 Details : In some cases it is possible that recently written or created
3355 files may not be written to disk in a timely manner (this should
3356 normally be within 30s unless client IO load is very high).
3357 The problem appears as zero-length files or files that are a
3358 multiple of 1MB in size after a client crash or client eviction
3359 that are missing data at the end of the file.
3361 This problem is more likely to be hit on clients where files are
3362 repeatedly created and unlinked in the same directory, clients
3363 have a large amount of RAM, have many CPUs, the filesystem has
3364 many OSTs, the clients are rebooted frequently, and/or the files
3365 are not accessed by other nodes after being written.
3367 The presence of the problem can be detected by looking at
3368 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
3369 smaller than the second (nr_unused) then dirty files will not
3370 be flushed automatically to disk. "sync; sleep 10" should be
3371 run several times on the node before unmounting it to update
3372 Lustre (this is also safe to run on nodes without this problem).
3374 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
3375 that can cause this same problem, so customers using that kernel
3376 also need to update the kernel in addition to Lustre. In order
3377 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
3379 It is normal that files written just before a client crash (less
3380 than 30s) may not yet have been flushed to disk, even for local
3384 Frequency : frequent on thin XT3 nodes
3386 Description: UUID collision on thin XT3 Linux nodes
3387 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
3388 randomly, since we relied on an insufficiently-seeded PRNG.
3393 Description: OSS hangs after "All ost request buffers busy"
3394 Details : A deadlock between quota and journal operations caused OSS
3395 hangs after printing "All ost request buffers busy."
3398 Frequency : always on liblustre builds
3400 Description: Cleanup compiler warnings on liblustre
3403 Frequency : always on liblustre builds on XT3
3405 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
3409 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
3410 > 32000 subdirectories
3411 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
3412 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
3413 subdirectory count crosses 32000. This will aid e2fsck to
3414 correctly handle more than 32000 subdirectories.
3417 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3419 Description: System hang on startup
3420 Details : This bug allowed the liblustre (e.g. catamount) client to
3421 return to the app before handling all startup RPCs. This
3422 could leave the node unresponsive to lustre network traffic
3423 and manifested as a server ptllnd timeout.
3425 Severity : enhancement
3427 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3428 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
3429 boolean to control whether to print peer debug info when a
3430 client's RPC times out.
3435 Description: make O_SYNC working on 2.6 kernels
3436 Details : 2.6 kernels use different method for mark pages for write,
3437 so need add a code to lustre for O_SYNC work.
3442 Description: Failure to close file and release space on NFS
3443 Details : Put inode details into lock acquired in ll_intent_file_open.
3444 Use mdc_intent_lock in ll_intent_open to properly
3445 detect all kind of errors unhandled by mdc_enqueue.
3450 Description: proc file read during shutdown sometimes raced obd removal,
3452 Details : Add lock to prevent obd access after proc file removal.
3455 Frequency : Only for files larger than 4GB on 32-bit clients.
3457 Description: improperly doing page alignment of locks
3458 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
3459 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3464 Description: RPCs being resent when they shouldn't be
3465 Details : Some RPCs that should not be resent are being resent. This
3466 can cause inconsistencies in the RPC state machine. Do not
3467 resend such requests.
3470 Frequency : rare, only with NFS export
3472 Description: Crash on NFS re-export node
3473 Details : under very unusual load conditions an assertion is hit in
3474 ll_intent_file_open()
3477 Frequency : only if OST filesystem is corrupted
3479 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3480 Details : for a short time RPCs with bulk IO are in the replay list,
3481 but replay of bulk IOs is unimplemented. If the OST filesystem
3482 is corrupted due to disk cache incoherency and then replay is
3483 started it is possible to trip an assertion. Avoid putting
3484 committed RPCs into the replay list at all to avoid this issue.
3489 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3490 Details : Large single O_DIRECT read and write calls can fail to allocate
3491 a sufficiently large buffer to process the request. In case of
3492 allocation failure the allocation is retried with a smaller
3493 buffer and broken into smaller requests.
3495 Severity : enhancement
3497 Description: Add -o localflock option to simulate old noflock behaviour.
3498 Details : This will achieve local-only flock/fcntl locks coherentness.
3503 Description: versioning check is incomplete
3504 Details : Checking the version difference of client vs. server, report
3505 error if the gap is too big.
3510 Description: add support PG_writeback bit
3511 Details : add support for PG_writeback bit for Lustre, for more carefull
3512 work with page cache in 2.6 kernel. This also fix some deadlocks
3513 and remove hack for work O_SYNC with 2.6 kernel.
3515 Severity : enhancement
3517 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3518 Details : The uninit_groups feature works in conjunction with the kernel
3519 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3520 up the pass1 processing of e2fsck. This is a read-only feature
3521 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3522 mount filesystems that have had this feature enabled.
3524 Severity : enhancement
3526 Description: Improve multi-block allocation algorithm to avoid fragmentation
3527 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3528 allocation locality and avoid filesystem fragmentation.
3530 ------------------------------------------------------------------------------
3532 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3534 * Support for kernels:
3535 2.6.9-42.0.3.EL (RHEL 4)
3536 2.6.5-7.276 (SLES 9)
3537 2.4.21-47.0.1.EL (RHEL 3)
3538 2.6.12.6 vanilla (kernel.org)
3539 2.6.16.21-0.8 (SLES10)
3540 * Recommended e2fsprogs version: 1.39.cfs2-0
3542 * The backwards-compatible /proc/sys/portals symlink has been removed
3543 in this release. Before upgrading, please ensure that you change
3544 any configuration scripts or /etc/sysctl.conf files that access
3545 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3546 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3547 in advance of the upgrade on any system running Lustre 1.4.6 or
3548 newer, since /proc/sys/lnet was added in that version.
3549 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3553 Frequency : only when quota is used
3555 Description: avoid scanning export list for quota master
3556 Details : Change the algorithms to avoid scanning export list in order
3557 to improve the efficiency.
3560 Frequency : MDS failover only, very rarely
3562 Description: "went back in time" messages on mds failover
3563 Details : The greatest transno may be lost when the current operation
3564 finishes with an error (transno==0) and the client's last_rcvd
3565 record is over-written. Save the greatest transno in the
3566 mds_last_transno for this case.
3569 Frequency : always for specific kernels and striping counts
3571 Description: client may get "Matching packet too big" without ACL support
3572 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
3573 when trying to access files in certain configurations. The
3574 clients should in fact be denied when mounting because they do
3575 not understand ACLs.
3578 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
3580 Description: many clients connecting with IO in progress causes connect timeouts
3581 Details : Avoid synchronous journal commits to avoid delays caused by many
3582 clients connecting/disconnecting when bulk IO is in progress.
3583 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
3584 OST_IO_PORTAL to avoid delays behind potentially many pending
3588 Frequency : occasionally with multiple writers to a single file
3590 Description: shared writes to file may result in wrong size reported by stat()
3591 Details : Allow growing of kms when extent lock is cancelled
3594 Frequency : always with random mmap IO to multi-striped file
3596 Description: mmap write might be lost if we are writing to a 'hole' in stripe
3597 Details : Only if the hole is at the end of OST object so that kms is too
3598 small. Fix is to increase kms accordingly in ll_nopage.
3601 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
3603 Description: writes to a missing object would leak memory on the OST
3604 Details : If there is an inconsistency between the MDS and OST filesystems,
3605 such that the MDS references an object that doesn't exist, writes
3606 to that object will leak memory due to incorrect cleanup in the
3607 error handling path, eventually running out of memory on the OST.
3612 Description: Creating too long symlink causes lustre errors
3613 Details : Check symlink and name lengths before sending requests to MDS.
3616 Frequency : only if flock is enabled (not on by default)
3618 Description: posix locks not released on fd closure on 2.6.9+
3619 Details : We failed to add posix locks to list of inode locks on 2.6.9+
3620 kernels, this caused such locks not to be released on fd close and
3621 then assertions on fs unmount about still used locks.
3624 Frequency : MDS failover only, very rarely
3626 Description: clients may get ASSERTION(granted_lock != NULL)
3627 Details : When request was taking a long time, and a client was resending
3628 a getattr by name lock request. The were multiple lock requests
3629 with the same client lock handle and
3630 mds_getattr_name->fixup_handle_for_resent_request found one of the
3631 lock handles but later failed with ASSERTION(granted_lock != NULL).
3636 Description: handle->h_buffer_credits > 0, assertion failure
3637 Details : h_buffer_credits is zero after truncate, causing assertion
3638 failure. This patch extends the transaction or creates a new
3642 Frequency : NFS re-export or patchless client
3643 Bugzilla : 11179, 10796
3644 Description: Crash on NFS re-export node (__d_move)
3645 Details : We do not want to hash the dentry if we don't have a lock.
3646 But if this dentry is later used in d_move, we'd hit uninitialised
3647 list head d_hash, so we just do this to init d_hash field but
3648 leave dentry unhashed.
3651 Frequency : NFS re-export or patchless client
3653 Description: NFS exports has problem with symbolic link
3654 Details : lustre client didn't properly install dentry when re-exported
3655 to NFS or running patchless client.
3658 Frequency : NFS re-export or patchless client
3660 Description: Various nfs/patchless fixes.
3661 Details : fixes reuse disconected alias for lookup process - this fixes
3662 warning "find_exported_dentry: npd != pd",
3663 fix permission error with open files at nfs.
3664 fix apply umask when do revalidate.
3667 Frequency : occasional
3669 Description: Crash on NFS re-export node
3670 Details : calling clear_page() on the wrong pointer triggered oops in
3671 generic_mapping_read().
3674 Frequency : rarely, using O_DIRECT IO
3676 Description: unaligned directio crashes client with LASSERT
3677 Details : check for unaligned buffers before trying any requests.
3680 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
3682 Description: stale data returned from RAID cache
3683 Details : If only a small amount of IO is done to the RAID device before
3684 reading it again it is possible to get stale data from the RAID
3685 cache instead of reading it from disk.
3688 Frequency : always for sles10 kernel
3690 Description: sles10 support
3691 Details : ll_follow_link: compile fixes and using of nd_set_link
3692 under newer kernels.
3695 Frequency : depends on arch, kernel and compiler version, always on sles10
3698 Description: recursive or deep enough symlinks cause stack overflow
3699 Details : getting rid of large stack-allocated variable in
3703 Frequency : depends on hardware
3705 Description: lustre write performance loss in the SLES10 kernel
3706 Details : the performance loss is caused by using of write barriers in the
3707 ext3 code. The SLES10 kernel turns barrier support on by
3708 default. The fix is to undo that change for ldiskfs.
3710 ------------------------------------------------------------------------------
3712 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
3714 * Support for kernels:
3715 2.6.9-42.0.3EL (RHEL 4)
3716 2.6.5-7.276 (SLES 9)
3717 2.4.21-47.0.1.EL (RHEL 3)
3718 2.6.12.6 vanilla (kernel.org)
3722 Frequency : quota enabled and large files being deleted
3724 Description: releasing more than 4GB of quota at once hangs OST
3725 Details : If a user deletes more than 4GB of files on a single OST it
3726 will cause the OST to spin in an infinite loop. Release
3727 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3732 Description: statfs data retrieved from /proc may be stale or zero
3733 Details : When reading per-device statfs data from /proc, in the
3734 {kbytes,files}_{total,free,avail} files, it may appear
3735 as zero or be out of date.
3738 Frequency : systems with MD RAID1 external journal devices
3740 Description: lconf's call to blkid is confused by RAID1 journal devices
3741 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3742 of returning all block devices that match the journal UUID.
3745 Frequency : always, for aggregate stripe size over 4GB
3747 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
3748 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
3749 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
3752 Frequency : always if "lfs find" used on a local file/directory
3754 Description: "lfs find" segfaults if used on a local file/directory
3755 Details : The case where a directory component was not specified wasn't
3756 handled correctly. Handle this properly.
3759 Frequency : always on ppc64
3761 Description: the write to an ext3 filesystem mounted with mballoc got stuck
3762 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3763 perform endianness conversion.
3766 Frequency : rarely (truncate to non-zero file size after write under load)
3767 Bugzilla : 10730, 10687
3768 Description: Files padded with zeros to next 4K multiple
3769 Details : With filesystems mounted using the "extents" option (2.6 kernels)
3770 it is possible that files that are truncated to a non-zero size
3771 immediately after being written are filled with zero bytes beyond
3772 the truncated size. No file data is lost.
3774 Severity : enhancement
3776 Description: Allow recovery/failover for liblustre clients.
3777 Details : liblustre clients were unaware of failover configurations until
3780 Severity : enhancement
3782 Description: user file locks should fail when not mounting with flock option
3783 Details : Set up an error-returning stub in ll_file_operations.lock field
3784 to prevent incorrect behaviour when client is mounted without
3785 flock option. Also, set up properly f_op->flock field for
3789 Frequency : always on ia64
3791 Description: "lfs df" loops on printing out MDS statfs information
3792 Details : The obd_ioctl_data was not initialized and in some systems
3793 this caused a failure during the ioctl that did not return
3794 an error. Initialize the struct and return an error on failure.
3797 Frequency : SLES 9 only
3799 Description: Error of copying files with lustre special EAs as root
3800 Details : Client side always return success for setxattr call for lustre
3801 special xattr (currently only "trusted.lov").
3804 Frequency : rarely on clusters with both ia64+i386 clients
3806 Description: ia64+i686 clients doing shared IO on the same file may LBUG
3807 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
3808 clients are doing concurrent writes to the same file it is
3809 possible that the ia64 clients may LASSERT because the OST
3810 extent locks are not PAGE_SIZE aligned. Ensure that grown
3811 locks are always aligned on the request boundary.
3814 Frequency : specific use, occasional
3816 Description: Overwriting in use executable truncates on-disk binary image
3817 Details : If one node attempts to overwrite an executable in use by
3818 another node, we now correctly return ETXTBSY instead of
3819 truncating the file.
3821 Severity : enhancement
3823 Description: Async OSC create to avoid the blocking unnecessarily.
3824 Details : If a OST has no remain object, system will block on the creating
3825 when need to create a new object on this OST. Now, ways use
3826 pre-created objects when available, instead of blocking on an
3827 empty osc while others are not empty. If we must block, we block
3828 for the shortest possible period of time.
3833 Description: chmod on Lustre root is propagated to other clients
3834 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
3835 invalid by the follow_mount time.
3838 Frequency : liblustre clients only
3840 Description: Race in 'instant cancel' lock handling could lead to such locks
3841 never to be granted in case of SMP MDS
3842 Details : Do not destroy not yet granted but cbpending locks in
3846 Frequency : replay/resend of open
3848 Description: non null lock assetion failure in mds_intent_policy
3849 Details : Trying to replay/resend lockless open requests resulted in
3850 mds_open() returning 0 with no lock. Now it sets a flag if
3851 a lock is going to be returned.
3853 Severity : enhancement
3855 Description: Checksum enhancements
3856 Details : New checksum enhancements allow for resending RPCs that failed
3859 Severity : enhancement
3861 Description: Tunables on number of dirty pages in cacche
3862 Details : Allow to set limit on number of dirty pages cached.
3867 Description: client crash on unmount - lock still has references
3868 Details : In some error handling cases it was possible to leak a lock
3869 reference on a client while accessing a file. This was not
3870 harmful to the client during operation, but would cause the
3871 client to crash when the filesystem is unmounted.
3874 Frequency : specific case, rare
3876 Description: ETXTBSY on mds though file not in use by client
3877 Details : ETXTBSY is no longer incorrectly returned when attempting to
3878 chmod or chown a directory that the user previously tried to
3879 execute or a currently-executing binary.
3882 Frequency : extremely rare except on liblustre-based clients
3884 Description: Lustre space not freed when files are deleted
3885 Details : Clean up open-unlinked files after client eviction. Previously
3886 the unlink was skipped and the files remained as orphans.
3891 Description: OST failure "would be an LBUG" in waiting_locks_callback()
3892 Details : In some cases it was possible to send a blocking callback to
3893 a client doing a glimpse, even though that client didn't get
3894 a lock granted. When the glimpse lock is cancelled on the OST
3895 the freed lock is left on the waiting list and corrupted the list.
3898 Frequency : all core dumps
3900 Description: Broke core dumps to lustre
3901 Details : Negative dentry may be unhashed if parent does not have UPDATE
3902 lock, but some callers, e.g. do_coredump, expect dentry to be
3903 hashed after successful create, hash it in ll_create_it.
3905 ------------------------------------------------------------------------------
3907 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
3909 * Support for kernels:
3910 2.6.9-42.0.2.EL (RHEL 4)
3911 2.6.5-7.276 (SLES 9)
3912 2.4.21-40.EL (RHEL 3)
3913 2.6.12.6 vanilla (kernel.org)
3917 Frequency : always on RHEL 3
3919 Description: Number of open files grows over time
3920 Details : The number of open files grows over time, whether or not
3921 Lustre is started. This was due to a filp leak introduced
3922 by one of our kernel patches.
3924 ------------------------------------------------------------------------------
3926 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
3928 * Support for kernels:
3929 2.6.9-42.EL (RHEL 4)
3930 2.6.5-7.276 (SLES 9)
3931 2.4.21-40.EL (RHEL 3)
3932 2.6.12.6 vanilla (kernel.org)
3937 Bugzilla : 5719, 9635, 9792, 9684
3938 Description: OST (or MDS) trips assertions in (re)connection under heavy load
3939 Details : If a server is under heavy load and cannot reply to new
3940 connection requests before the client resends the (re)connect,
3941 the connection handling code can behave badly if two service
3942 threads are concurrently handing separate (re)connections from
3943 the same client. Add better locking to the connection handling
3944 code, and ensure that only a single connection will be processed
3945 for a given client UUID, even if the lock is dropped.
3947 Severity : enhancement
3949 Description: add TCP zero-copy support to kernel
3950 Details : Add support to the kernel TCP stack to allow zero-copy bulk
3951 sends if the hardware supports scatter-gather and checksumming.
3952 This allows socklnd to do client-write and server-read more
3953 efficiently and reduce CPU utilization from skbuf copying.
3956 Frequency : only if NFS exporting from client
3958 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
3959 Details : If setting attributes on a file created under NFS that had
3960 never been opened it would be possible to oops the client
3961 if the file had no objects.
3964 Frequency : always for liblustre
3966 Description: liblustre client does MDS+OSTs setattr RPC for each write
3967 Details : When doing a write from a liblustre client, the client
3968 incorrectly issued an RPC to the MDS and each OST the file was
3969 striped over in order to update the timestamps. When writing
3970 with small chunks and many clients this could overwhelm the MDS
3971 with RPCs. In all cases it would slow down the write because
3972 these RPCs are unnecessary.
3974 Severity : enhancement
3976 Description: allow number of MDS service threads to be changed at module load
3977 Details : It is now possible to change the number of MDS service threads
3978 running. Adding "options mds mds_num_threads={N}" to the MDS's
3979 /etc/modprobe.conf will set the number of threads for the next
3980 time Lustre is restarted (assuming the "mds" module is also
3981 reloaded at that time). The default number of threads will
3982 stay the same, 32 for most systems.
3987 Description: OST crash if filesystem is unformatted or corrupt
3988 Details : If an OST is started on a device that has never been formatted
3989 or if the filesystem is corrupt and cannot even mount then the
3990 error handling cleanup routines would dereference a NULL pointer.
3995 Description: NULL pointer deref in llap_from_page.
3996 Details : get_cache_page_nowait can return a page with NULL (or otherwise
3997 incorrect) mapping if the page was truncated/reclaimed while it was
3998 searched for. Check for this condition and skip such pages when
3999 doing readahead. Introduce extra check to llap_from_page() to
4000 verify page->mapping->host is non-NULL (so page is not anonymous).
4003 Frequency : Sometimes when using sys_sendfile
4005 Description: "page not covered by a lock" warnings from ll_readpage
4006 Details : sendfile called ll_readpage without right page locks present.
4007 Now we introduced ll_file_sendfile that does necessary locking
4008 around call to generic_file_sendfile() much like we do in
4012 Frequency : with certain MDS communication failures at client mount time
4014 Description: NULL pointer deref after failed client mount
4015 Details : a client connection request may delayed by the network layer
4016 and not be sent until after the PTLRPC layer has timed out the
4017 request. If the client fails the mount immediately it will try
4018 to clean up before the network times out the request. Add a
4019 reference from the request import to the obd device and delay
4020 the cleanup until the network drops the request.
4023 Frequency : occasionally during client (re)connect
4025 Description: assertion failure during client (re)connect
4026 Details : processing a client connection request may be delayed by the
4027 client or server longer than the client connect timeout. This
4028 causes the client to resend the connection request. If the
4029 original connection request is replied in this interval, the
4030 client may trip an assertion failure in ptlrpc_connect_interpret()
4031 which thought it would be the only running connect process.
4034 Frequency : only with obd_echo servers and clients that are rebooted
4036 Description: kernel BUG accessing uninitialized data structure
4037 Details : When running an obd_echo server it did not start the ping_evictor
4038 thread, and when a client was evicted an uninitialized data
4039 structure was accessed. Start the ping_evictor in the RPC
4040 service startup instead of the OBD startup.
4042 Severity : enhancement
4043 Bugzilla : 10193 (patchless)
4044 Description: Remove dependency on various unexported kernel interfaces.
4045 Details : No longer need reparent_to_init, exit_mm, exit_files,
4046 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
4049 Frequency : rare (only users of deprecated and unsupported LDAP config)
4051 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
4052 Details : LDAP apparently contains 'lustreName' attributes instead of
4053 'name'. A simple remapping of the name is sufficient.
4056 Frequency : rare (only with non-default dump_on_timeout debug enabled)
4058 Description: waiting_locks_callback trips kernel BUG if client is evicted
4059 Details : Running with the dump_on_timeout debug flag turned on makes
4060 it possible that the waiting_locks_callback() can try to dump
4061 the Lustre kernel debug logs from an interrupt handler. Defer
4062 this log dumping to the expired_lock_main() thread.
4064 Severity : enhancement
4066 Description: Support NFS exporting on 2.6 kernels.
4067 Details : Implement non-rawops metadata methods for NFS server to use without
4068 changing NFS server code.
4071 Frequency : very rare (synthetic metadata workload only)
4073 Description: two racing renames might cause an MDS thread to deadlock
4074 Details : Running the "racer" program may cause one MDS thread to rename
4075 a file from being the source of a rename to being the target of
4076 a rename at exactly the same time that another thread is doing
4077 so, and the second thread has already enqueued these locks after
4078 doing a lookup of the target and is trying to relock them in
4079 order. Ensure that we don't try to re-lock the same resource.
4082 Frequency : only very large systems with liblustre clients
4084 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
4085 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
4086 all OSTs in parallel. This allows the eviction of stale liblustre
4087 clients to proceed much faster than if they were done in series,
4088 and also offers similar improvements for other set_info RPCs.
4093 Description: excessive CPU usage during initial read phase on client
4094 Details : During the initial read phase on a client, it would agressively
4095 retry readahead on the file, consuming too much CPU and impacting
4096 performance (since 1.4.5.8). Improve the readahead algorithm
4097 to avoid this, and also improve some other common cases (read
4098 of small files in particular, where "small" is files smaller than
4099 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
4104 Description: MDS crash when receiving packet with unknown intent.
4105 Details : Do not LBUG in unknown intent case, just return -EFAULT
4107 Severity : enhancement
4108 Bugzilla : 9293, 9385
4109 Description: MDS RPCs are serialised on client. This is unnecessary for some.
4110 Details : Do not serialize getattr (non-intent version) and statfs.
4113 Frequency : occasional, when OST network is overloaded/intermittent
4115 Description: client evicted by OST after bulk IO timeout
4116 Details : If a client sends a bulk IO request (read or write) the OST
4117 may evict the client if it is unresposive to its data GET/PUT
4118 request. This is incorrect if the network is overloaded (takes
4119 too long to transfer the RPC data) or dropped the OST GET/PUT
4120 request. There is no need to evict the client at all, since
4121 the pinger and/or lock callbacks will handle this, and the
4122 client can restart the bulk request.
4125 Frequency : Always when mmapping file with no objects
4127 Description: client crashes when mmapping file with no objects
4128 Details : Check that we actually have objects in a file before doing any
4129 operations on objects in ll_vm_open, ll_vm_close and
4135 Description: Request leak when working with deleted CWD
4136 Details : Introduce advanced request refcount tracking for requests
4137 referenced from lustre intent.
4139 Severity : Enhancement
4141 Description: Cache open file handles on client.
4142 Details : MDS now will return special lock along with openhandle, if
4143 requested and client is allowed to hold openhandle, even if unused,
4144 until such a lock is revoked. Helps NFS a lot, since NFS is opening
4145 closing files for every read/write openration.
4147 Severity : Enhancement
4149 Description: Cache open negative dentries on client when possible.
4150 Details : Guard negative dentries with UPDATE lock on parent dir, drop
4151 negative dentries on lock revocation.
4156 Description: Remounting a client read-only wasn't possible with a zconf mount
4157 Details : It wasn't possible to remount a client read-only with llmount.
4159 Severity : enhancement
4160 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
4161 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
4162 MPICH is included to add Lustre-specific ADIO interfaces.
4163 This is based closely on the UFS ADIO layer and only differs
4164 in file creation, in order to allow the OST striping to be set.
4165 This is user-contributed code and not supported by CFS.
4170 Description: extended inode attributes (immutable, append-only) work improperly
4171 when 2.4 and 2.6 kernels are used on client/server or vice versa
4172 Details : Introduce kernel-independent values for these flags.
4174 Severity : enhancement
4177 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
4178 Details : Many of the /proc/ tunables can only be tuned at a megabyte
4179 granularity. Now, Fractional MB granularity is be supported,
4180 this is very useful for low memory system.
4182 Severity : enhancement
4184 Description: Getattr by fid
4185 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
4186 locks, avoids extra getattr rpc requests to MDS, allows '/' to
4187 have locks and avoids getattr rpc requests for it on every stat.
4190 Frequency : Always, for filesystems larger than 2TB
4192 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4193 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
4194 filesystems larger than 2TB to be mounted with mballoc enabled.
4197 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
4199 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4200 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
4201 filesystems larger than 2TB is mounted then the kernel will
4202 silently corrupt the start of the filesystem. CONFIG_LBD is
4203 enabled for all CFS-supported kernels, but the possibility of
4204 this happening with a modified kernel config exists.
4206 Severity : enhancement
4208 Description: add client O_DIRECT support for 2.6 kernels
4209 Details : It is now possible to do O_DIRECT reads and writes to files
4210 in the Lustre client mountpoint on 2.6 kernel clients.
4212 Severity : enhancement
4214 Description: parallel glimpse, setattr, statfs, punch, destroy requests
4215 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
4216 parallel, not waiting for response from every OST before sending
4217 a rpc to the next OST.
4222 Description: setattr vs write race when updating file timestamps
4223 Details : Client processes that update a file timestamp into the past
4224 right after writing to the file (e.g. tar) it is possible that
4225 the updated file modification time can be reset to the current
4226 time due to a race between processing the setattr and write RPC.
4228 Severity : enhancement
4230 Description: Bring 'lfs find' closer in line with regular Linux find.
4231 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
4232 -print0 options and obtains all the needed info through the lustre
4235 Severity : enhancement
4237 Description: support up to 1024 configured devices on one node
4238 Details : change obd_dev array from statically allocated to dynamically
4239 allocated structs as they are first used to reduce memory usage
4244 Description: Flush dirty partially truncated pages during truncate
4245 Details : Immediatelly flush partially truncated pages in filter_setattr,
4246 this way we completely avoid having any pages in page cache on OST
4247 and can retire ugly workarounds during writes to flush such pages.
4252 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
4253 Details : protect inode from truncation within vfs_unlink() context
4254 just take a reference before calling vfs_unlink() and release it
4255 when parent's i_sem is free.
4260 Description: last_id value checked outside lock on OST caused LASSERT failure
4261 Details : If there were multiple MDS->OST object precreate requests in
4262 flight, it was possible that the OST's last object id was checked
4263 outside a lock and incorrectly tripped an assertion. Move checks
4264 inside locks, and discard old precreate requests.
4267 Frequency : always, if extents are used on OSTs
4269 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
4270 Details : OSTs using the extents format would not zero the high 16 bits of
4271 the index physical block number. This is not a problem for any
4272 OST filesystems smaller than 16TB, and no kernels support ext3
4273 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
4274 new/modified files) and can be fixed for existing filesystems
4275 with e2fsprogs-1.39-cfs1.
4280 Description: import connection selection may be incorrect if timer wraps
4281 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
4282 import connections to be ignored if the 32-bit jiffies counter
4283 wraps. Use a 64-bit jiffies counter.
4286 Frequency : very large clusters immediately after boot
4288 Description: LNET request buffers exhausted under heavy short-term load
4289 Details : If a large number of client requests are generated on a service
4290 that has previously never seen so many requests it is possible
4291 that the request buffer growth cannot keep up with the spike in
4292 demand. Instead of dropping incoming requests, they are held in
4293 the LND until the RPC service can accept more requests.
4296 Frequency : Sometimes during replay
4298 Description: Assertion failure in ll_local_open after replay.
4299 Details : If replay happened on an open request reply before we were able
4300 to set replay handler, reply will become not swabbed tripping the
4301 assertion in ll_local_open. Now we set the handler right after
4302 recognising of open request
4305 Frequency : very rare
4307 Description: kernel reports "badness in vsnprintf"
4308 Details : Reading from the "recovery_status" /proc file in small chunks
4309 may cause a negative length in lprocfs_obd_rd_recovery_status()
4310 call to vsnprintf() (which is otherwise harmless). Exit early
4311 if there is no more space in the output buffer.
4313 Severity : enhancement
4315 Description: clear OBD RPC statistics by writing to them
4316 Details : It is now possible to clear the OBD RPC statistics by writing
4317 to the "stats" file.
4322 Description: Client mtime is not the same on different clients after utimes
4323 Details : In some cases, the client was using the utimes() syscall on
4324 a file cached on another node. The clients now validate the
4325 ctime from the MDS + OSTs to determine which one is right.
4330 Description: Inability to activate failout mode
4331 Details : lconf script incorrectly assumed that in pythong string's numeric
4332 value is used in comparisons.
4335 Frequency : always with multiple stripes per file
4337 Description: Inefficient object allocation for mutli-stripe files
4338 Details : When selecting which OSTs to stripe files over, for files with
4339 a stripe count that divides evenly into the number of OSTs,
4340 the MDS is always picking the same starting OST for each file.
4341 Return the OST selection heuristic to the original design.
4346 Description: mount failures may take full timeout to return an error
4347 Details : Under some heavy load conditions it is possible that a
4348 failed mount can wait for the full obd_timeout interval,
4349 possibly several minutes, before reporting an error.
4350 Instead return an error as soon as the status is known.
4352 Frequency : quota enabled and large files being deleted
4354 Description: releasing more than 4GB of quota at once hangs OST
4355 Details : If a user deletes more than 4GB of files on a single OST it
4356 will cause the OST to spin in an infinite loop. Release
4357 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4362 Description: statfs data retrieved from /proc may be stale or zero
4363 Details : When reading per-device statfs data from /proc, in the
4364 {kbytes,files}_{total,free,avail} files, it may appear
4365 as zero or be out of date.
4368 Frequency : systems with MD RAID1 external journal devices
4370 Description: lconf's call to blkid is confused by RAID1 journal devices
4371 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4372 of returning all block devices that match the journal UUID.
4375 Frequency : always, for aggregate stripe size over 4GB
4377 Description: assertion fails when trying to use 4GB stripe size
4378 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
4379 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
4382 Frequency : always on ppc64
4384 Description: the first write on an ext3 filesystem with mballoc got stuck
4385 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4386 perform endianness conversion.
4388 ------------------------------------------------------------------------------
4390 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4392 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
4393 INTEROPERATE with older versions automatically. Please read the
4394 user documentation before upgrading any part of a live system.
4395 * WARNING: Lustre networking configuration changes are required with
4396 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
4399 * Support for kernels:
4400 2.6.9-22.0.2.EL (RHEL 4)
4401 2.6.5-7.244 (SLES 9)
4402 2.6.12.6 vanilla (kernel.org)
4405 Severity : enhancement
4406 Bugzilla : 7981/8208
4407 Description: Introduced Lustre Networking (LNET)
4408 Details : LNET is new networking infrastructure for Lustre, it includes
4409 a reorganized network configuration mode (see the user
4410 documentation for full details) as well as support for routing
4411 between different network fabrics. Lustre Networking Devices
4412 (LNDs) for the supported network fabrics have also been
4413 created for this new infrastructure.
4415 Severity : enhancement
4416 Description: Introduced Access control lists
4417 Details : clients can set ACLs on files and directories in order to have
4418 more fine-grained permissions than the standard Unix UGO+RWX.
4419 The MDS must be started with the "-o acl" mount option.
4421 Severity : enhancement
4422 Description: Introduced filesystem quotas
4423 Details : Administrators may now establish per-user quotas on the
4426 Severity : enhancement
4428 Description: Configuration change for the XT3
4429 The PTLLND is now used to run Lustre over Portals on the XT3
4430 The configure option(s) --with-cray-portals are no longer used.
4431 Rather --with-portals=<path-to-portals-includes> is used to
4432 enable building on the XT3. In addition to enable XT3 specific
4433 features the option --enable-cray-xt3 must be used.
4438 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
4439 Details : A race between allocating a new llcd and re-getting the llcd_lock
4440 allowed another thread to grab newly-allocated llcd.
4442 Severity : enhancement
4444 Description: 2.6 OST async journal commit and locking fix to improve performance
4445 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
4446 2.6 kernels are now async as they already were in 2.4 kernels so
4447 that they can commit concurrently with the network bulk transfer.
4448 For block-allocated files the filter allocation semaphore is held
4449 to avoid filesystem fragmentation during allocation. BKL lock
4450 removed for 2.6 xattr operations where it is no longer needed.
4455 Description: lconf incorrectly determined whether two IP networks could talk
4456 Details : In some more complicated routing and multiple-network
4457 configurations, lconf will avoid trying to make a network
4458 connection to a disjoint part of the IP space. It was doing the
4459 math incorrectly for one set of cases.
4464 Description: Fix for potential infinite loop processing records in an llog.
4465 Details : If an llog record is corrupted/zeroed, it is possible to loop
4466 forever in llog_process(). Validate the llog record length
4467 and skip the remainder of the block on error.
4470 Frequency : occasional (liblustre only)
4472 Description: liblustre could not open files whose last component is a symlink
4473 Details : sysio_path_walk() would incorrectly pass the open intent to
4474 intermediate path components.
4477 Frequency : rare (liblustre only with non-standard tuning)
4478 Bugzilla : 7201 (7350)
4479 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4480 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4481 after it has been released, only do this for referenced locks
4483 Severity : enhancement
4485 Description: specifying an (invalid) directory default stripe_size of -1
4486 would reset the directory default striping
4487 Details : stripe_size -1 was used internally to signal directory stripe
4488 removal, now use "all default" to signal dir stripe removal
4489 as a directory striping of "all default" is not useful
4492 Frequency : common for large clusters running liblustre clients
4494 Description: doing an ls when liblustre clients are running is slow
4495 Details : sending a glimpse AST to a liblustre client waits for every AST
4496 to time out, as liblustre clients will not respond. Since they
4497 cannot cache data we refresh the OST lock LVB from disk instead.
4499 Severity : enhancement
4501 Description: doing an ls at the same time as file IO can be slow
4502 Details : enqueue and other "small" requests can be blocked behind many
4503 large IO requests. Create a new OST IO portal for non-IO
4504 requests so they can be processed faster.
4507 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4509 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4510 created file with mode 000
4511 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4512 allow an ftruncate() as a non-root user to a file with mode 000.
4513 Lustre now respects this flag to disable mode checks when
4514 truncating a file owned by the user
4517 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4520 Description: Revoking locks from clients that went dead or catatonic might take
4522 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4523 cancellation of such locks instant on servers without waiting for
4524 any reply from clients. Clients drops these locks when cancel
4525 notification from server is received without replying.
4528 Frequency : liblustre-only, when liblustre client dies or becomes busy
4530 Description: Doing ls on Linux clients can take a long time with active
4532 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4533 granting such locks to it in the first place if possible. Locks
4534 are taken by proxy on the OST during the read or write and
4535 dropped immediately afterward. Add connect flags handling, do
4536 not grant locks to liblustre clients for glimpse ASTs.
4538 Severity : enhancement
4540 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4541 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4542 sequential reads and behaves badly with applications doing random
4543 reads. Improve it by reading ahead at least read region, and
4544 avoiding excessive large RPC for small reads.
4546 Severity : enhancement
4548 Description: Creating more than 1000 files for a single job may cause a load
4549 imbalance on the OSTs if there are also a large number of OSTs.
4550 Details : qos_prep_create() uses an OST index reseed value that is an
4551 even multiple of the number of available OSTs so that if the
4552 reseed happens in the middle of the object allocation it will
4553 still utilize the OSTs as uniformly as possible.
4558 Description: OST or MDS may oops in ping_evictor_main()
4559 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
4560 but doesn't restart at beginning of obd_exports_timed list
4563 Severity : enhancement
4565 Description: improve by-nid export eviction on the MDS and OST
4566 Details : allow multiple exports with the same NID to be evicted at one
4567 time without re-searching the exports list.
4570 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
4572 Description: MDS may oops in groups_free()
4573 Details : in rare race conditions a newly allocated group_info struct is
4574 freed again, and this can be NULL. The 2.4 compatibility code
4575 for groups_free() checked for a NULL pointer, but 2.6 did not.
4578 Frequency : common for liblustre clients doing little filesystem IO
4579 Bugzilla : 9352, 7313
4580 Description: server may evict liblustre clients accessing contended locks
4581 Details : if a client is granted a lock or receives a completion AST
4582 with a blocking AST already set it would not reply to the AST
4583 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
4586 Frequency : lfs setstripe, only systems with more than 160 OSTs
4588 Description: unable to set striping with a starting offset beyond OST 160
4589 Details : llapi_create_file() incorrectly limited the starting stripe
4590 index to the maximum single-file stripe count.
4593 Frequency : LDAP users only
4595 Description: lconf did not handle in-kernel recovery with LDAP properly
4596 Details : lconf/LustreDB get_refs() is searching the wrong namespace
4598 Severity : enhancement
4600 Description: bind OST threads to NUMA nodes to improve performance
4601 Details : all OST threads are uniformly bound to CPUs on a single NUMA
4602 node and do their allocations there to localize memory access
4604 Severity : enhancement
4606 Description: llmount can determine client NID directly from Myrinet (GM)
4607 Details : the client NID code from gmnalnid was moved directly into
4608 llmount, removing the need to use this or specifying the
4609 client NID explicitly when mounting GM clients with zeroconf
4612 Frequency : if client is started with down MDS
4614 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
4615 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
4616 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
4617 ptlrpc_import_delayed_req() because we didn't check if the
4618 request was interrupted, and we also didn't break out of the
4619 event loop if there was no timeout
4624 Description: data loss during non-page-aligned writes to a single file from
4625 both multiple nodes and multiple threads on one node at same time
4626 Details : updates to KMS and lsm weren't protected by common lock. Resulting
4627 inconsistency led to false short-reads, that were cached and later
4628 used by ->prepare_write() to fill in partially written page,
4629 leading to data loss.
4632 Frequency : always, if lconf --abort_recovery used
4634 Description: lconf --abort_recovery fails with 'Operation not supported'
4635 Details : lconf was attempting to abort recovery on the MDT device and not
4638 Severity : enhancement
4640 Description: remove cleanup logs
4641 Details : replace lconf-generated cleanup logs with lustre internal
4642 cleanup routines. Eliminates the need for client-cleanup and
4645 Severity : enhancement
4647 Description: add support for EAs (user and system) on lustre filesystems
4648 Details : it is now possible to store extended attributes in the Lustre
4649 client filesystem, and with the user_xattr mount option it
4650 is possible to allow users to store EAs on their files also
4652 Severity : enhancement
4654 Description: Add possibility (config option) to show minimal available OST free
4656 Details : When compiled with --enable-mindf configure option, statfs(2)
4657 (and so, df) will return least minimal free space available from
4658 all OSTs as amount of free space on FS, instead of summary of
4659 free spaces of all OSTs.
4661 Severity : enhancement
4663 Description: do not expand extent locks acquired on OST-side
4664 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
4665 by server: they are not cached anyway.
4668 Frequency : when mmap is used/binaries executed from Lustre
4670 Description: Unmmap pages before throwing them away from read cache.
4671 Details : llap_shrink cache now attempts to unmap pages before discarding
4672 them (if unmapping failed - do not discard). SLES9 kernel has
4673 extra checks that trigger if this unmapping is not done first.
4678 Description: lconf didn't resolve symlinks before checking to see whether a
4679 given mountpoint was already in use
4682 Frequency : when migrating failover services
4683 Bugzilla : 6395, 9514
4684 Description: When migrating a subset of services from a node (e.g. failback
4685 from a failover service node) the remaining services would
4686 time out and evict clients.
4687 Details : lconf --force (implied by --failover) sets the global obd_timeout
4688 to 5 seconds in order to quickly disconnect, but this caused
4689 other RPCs to time out too quickly. Do not change the global
4690 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
4692 Severity : enhancement
4693 Frequency : if MDS is started with down OST
4694 Bugzilla : 9439,5706
4695 Description: Allow startup/shutdown of an MDS without depending on the
4696 availability of the OSTs.
4697 Details : Asynchronously call mds_lov_synchronize during MDS startup.
4698 Add appropriate locking and lov-osc refcounts for safe
4699 cleaning. Add osc abort_inflight calls in case the
4700 synchronize never started.
4703 Frequency : occasional (Cray XT3 only)
4705 Description: root not authorized to access files in CRAY_PORTALS environment
4706 Details : The client process capabilities were not honoured on the MDS in
4707 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
4708 been accessed by an authorized user then root was able to access
4709 the file on the local client also. The root user capabilities
4710 are now allowed on the MDS, as this environment has secure UID.
4713 Frequency : occasional
4715 Description: ldiskfs "too long searching" message happens too often
4716 Details : A debugging message (otherwise harmless) prints too often on
4717 the OST console. This has been reduced to only happen when
4718 there are fragmentation problems on the filesystem.
4723 Description: Division by zero in statfs when all OSCs are inactive
4724 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
4725 lov_statfs divides by value returned by lov_get_stripecnt().
4729 Bugzilla : 9489, 3273
4730 Description: First write from each client to each OST was only 4kB in size,
4731 to initialize client writeback cache, which caused sub-optimal
4732 RPCs and poor layout on disk for the first writen file.
4733 Details : Clients now request an initial cache grant at (re)connect time
4734 and so that they can start streaming writes to the cache right
4735 away and always do full-sized RPCs if there is enough data.
4736 If the OST is rebooted the client also re-establishes its grant
4737 so that client cached writes will be honoured under the grant.
4742 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
4743 Details : Now I/O RPCs go to different portal number and (presumably) fast
4744 lock requests (and glimses) and other RPCs get their own service
4745 threads pool that should be able to service those RPCs
4748 Severity : enhancement
4750 Description: Ability to exchange lustre version between client and servers and
4751 issue warnings at client side if client is too old. Also for
4752 liblustre clients there is ability to refuse connection of too old
4754 Details : New 'version' field is added to connect data structure that is
4755 filled with version info. That info is later checked by server and
4759 Frequency : rare, liblustre only.
4760 Bugzilla : 9296, 9581
4761 Description: Two simultaneous writes from liblustre at offset within same page
4762 might proceed at the same time overwriting eachother with stale
4764 Details : I/O lock withing llu_file_prwv was released too early, before data
4765 actually was hitting the wire. Extended lock-holding time until
4766 server acknowledges receiving data.
4769 Frequency : extremely rare. Never observed in practice.
4771 Description: avoid generating lustre_handle cookie of 0.
4772 Details : class_handle_hash() generates handle cookies by incrementing
4773 global counter, and can hit 0 occasionaly (this is unlikely, but
4774 not impossible, because initial value of cookie counter is
4775 selected randonly). Value of 0 is used as a sentinel meaning
4776 "unassigned handle" --- avoid it. Also coalesce two critical
4777 sections in this function into one.
4779 Severity : enhancement
4781 Description: allow liblustre clients to delegate truncate locking to OST
4782 Details : To avoid overhead of locking, liblustre client instructs OST to
4783 take extent lock in ost_punch() on client's behalf. New connection
4784 flag is added to handle backward compatibility.
4786 Severity : enhancement
4787 Bugzilla : 4928, 7341, 9758
4788 Description: allow number of OST service threads to be specified
4789 Details : a module parameter allows the number of OST service threads
4790 to be specified via "options ost ost_num_threads={N}" in the
4791 OSS's /etc/modules.conf or /etc/modprobe.conf.
4795 Bugzilla : 6146, 9635, 9895
4796 Description: servers crash with bad pointer in target_handle_connect()
4797 Details : In rare cases when a client is reconnecting it was possible that
4798 the connection request was the last reference for that export.
4799 We would temporarily drop the export reference and get a new
4800 one, but this may have been the last reference and the export
4801 was just destroyed. Get new reference before dropping old one.
4803 Severity : enhancement
4804 Frequency : if client is started with failover MDS
4806 Description: Allow multiple MDS hostnames in the mount command
4807 Details : Try to read the configuration from all specified MDS
4808 hostnames during a client mount in case the "primary"
4811 Severity : enhancement
4813 Description: Stop sending data to evicted clients as soon as possible.
4814 Details : Check if the client we are about to send or are sending data to
4815 was evicted already. (Check is done every second of waiting,
4816 for which l_wait_event interface was extended to allow checking
4817 of exit condition at specified intervals).
4820 Frequency : rare, normally only when NFS exporting is done from client
4822 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
4824 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
4827 Frequency : always, if srand() is called before liblustre initialization
4829 Description: Liblustre uses system PRNG disturbing its usage by user application
4830 Details : Introduce internal to lustre fast and high-quality PRNG for
4831 lustre usage and make liblustre and some other places in generic
4832 lustre code to use it.
4834 Severity : enhancement
4835 Bugzilla : 9477, 9557, 9870
4836 Description: Verify that the MDS configuration logs are updated when xml is
4837 Details : Check if the .xml configuration logs are newer than the config
4838 logs stored on the MDS and report an error if this is the case.
4839 Request --write-conf, or allow starting with --old_conf.
4841 Severity : enhancement
4843 Description: Handle symlinks in the path when checking if Lustre is mounted.
4844 Details : Resolve intermediate symlinks when checking if a client has
4845 mounted a filesystem to avoid duplicate client mounts.
4850 Description: lconf can hit an error exception but still return success.
4851 Details : The lconf command catches the Command error exception at the top
4852 level script context and will exit with the associated exit
4853 status, but doesn't ensure that this exit status is non-zero.
4858 Description: failure of ptlrpc thread startup can cause oops
4859 Details : Starting a ptlrpc service thread can fail if there are a large
4860 number of threads or the server memory is very fragmented.
4861 Handle this without oopsing.
4864 Frequency : always, only if liblustre and non-default acceptor port was used
4866 Description: liblustre cannot connect to servers with non-default acceptor port
4867 Details : tcpnal_set_default_params() was not called and was therefore
4868 ignoring the environment varaible TCPNAL_PORT, as well as other
4869 TCPNAL_ environment variables
4874 Description: two objects could be created on the same OST for a single file
4875 Details : If an OST is down, in some cases it was possible to create two
4876 objects on a single OST for a single file. No problems other
4877 than potential performance impact and spurious error messages.
4881 Bugzilla : 5681, 9562
4882 Description: Client may oops in ll_unhash_aliases
4883 Details : Client dcache may become inconsistent in race condition.
4884 In some cases "getcwd" can fail if the current directory is
4890 Description: Inode refcounting problems in NFS export code
4891 Details : link_raw functions used to call d_instantiate without obtaining
4892 extra inode reference first.
4896 Bugzilla : 9942, 9903
4897 Description: Referencing freed requests leading to crash, memleaks with NFS.
4898 Details : We used to require that call to ll_revalidate_it was always
4899 followed by ll_lookup_it. Also with revalidate_special() it is
4900 possible to call ll_revalidate_it() twice for the same dentry
4901 even if first occurence returned success. This fix changes semantic
4902 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
4903 reference on a request referred from the intent.
4904 ll_intent_release() then releases such a request.
4907 Frequency : rare, normally benchmark loads only
4909 Description: unlinked inodes were kept in memory on the client
4910 Details : If a client is repeatedly creating and unlinking files it
4911 can accumulate a lot of stale inodes in the inode slab cache.
4912 If there is no other client load running this can cause the
4913 client node to run out of memory. Instead flush old inodes
4914 from client cache that have the same inode number as a new inode.
4917 Frequency : SLES9 2.6.5 kernel and long filenames only
4918 Bugzilla : 9969, 10379
4919 Description: utime reports stale NFS file handle
4920 Details : SLES9 uses out-of-dentry names in some cases, which confused
4921 the lustre dentry revalidation. Change it to always use the
4925 Frequency : rare, unless heavy write-truncate concurrency is continuous
4926 Bugzilla : 4180, 6984, 7171, 9963, 9331
4927 Description: OST becomes very slow and/or deadlocked during object unlink
4928 Details : filter_destroy() was holding onto the parent directory lock
4929 while truncating+unlinking objects. For very large objects this
4930 may block other threads for a long time and slow overall OST
4931 responsiveness. It may also be possible to get a lock ordering
4932 deadlock in this case, or run out of journal credits because of
4933 the combined truncate+unlink. Solution is to do object truncate
4934 first in one transaction without parent lock, and then do the
4935 final unlink in a new transaction with the parent lock. This
4936 reduces the lock hold time dramatically.
4939 Frequency : rare, 2.4 kernels only
4941 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
4942 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
4943 before unmounting their filesystem in order to avoid deadlock.
4944 The kernel_locked() function in 2.4 kernels only checks whether
4945 the kernel lock is held, not whether it is this process that is
4946 holding it as 2.6 kernels do.
4951 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
4952 Details : The client ptlrpc code may be trying to reconnect to a down
4953 server before a previous connection attempt has timed out.
4954 Increase the reconnect interval to be longer than the connection
4955 timeout interval to avoid sending duplicate connections to
4959 Frequency : echo_client brw_test command
4961 Description: fix echo_client to work with OST preallocated code
4962 Details : OST preallocation code (5137) didn't take echo_client IO path
4963 into account: echo_client calls filter methods outside of any
4964 OST thread and, hence, there is no per-thread preallocated
4965 pages and buffers to use. Solution: hijack pga pages for IO. As
4966 a byproduct, this avoids unnecessary data copying.
4970 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
4971 Description: Client can oops in mdc_commit_close() after open replay
4972 Details : It was possible for the MDS to return an open request with no
4973 transaction number in mds_finish_transno() if the client was
4974 evicted, but without actually returning an error. Clients
4975 would later try to replay that open and may trip an assertion
4976 Simplify the client close codepath, and always return an error
4977 from the MDS in case the open is not successful.
4980 Frequency : rare, 2.6 OSTs only
4982 Description: OST may deadlock under high load on fragmented files
4983 Details : If there was a heavy load and highly-fragmented OST filesystems
4984 it was possible to have all the OST threads deadlock waiting on
4985 allocation of biovecs, because the biovecs were not released
4986 until the entire RPC IO was completed. Instead, release biovecs
4987 as soon as they are complete to ensure forward IO progress.
4989 Severity : enhancement
4991 Description: Support for specifying external journal device at mount
4992 Details : If an OST or MDS device is formatted with an external journal
4993 device, this device major/minor is stored in the ext3 superblock
4994 and may not be valid for failover. Allow detecting and
4995 specifying the external journal at mount time.
5000 Description: Mounting an MDS with pending unlinked files may cause oops
5001 Details : target_finish_recovery() calls mds_postrecov() which returned
5002 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
5003 considers this an error and immediately begins cleaning up the
5004 lov, just after starting the mds_lov process
5006 Severity : enhancement
5008 Description: Implement 'lfs df' to report actual free space on per-OST basis
5009 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
5010 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
5011 usage of objects; '-h' to report in human readable format.
5013 ------------------------------------------------------------------------------
5015 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
5022 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
5023 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
5024 at startup during orphan unlinking. Instead, skip buddy update
5025 before setup, as it will be regenerated after recovery is complete.
5030 Description: If an OST is inactive, its locks might reference stale inodes.
5031 Details : lov_change_cbdata() must iterate over all namespaces, even if
5032 they are inactive to clear inode references from the lock.
5034 Severity : enhancement
5035 Frequency : occasional, if non-standard max_dirty_mb used
5037 Description: Client will block write RPCs if not enough grant
5038 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
5039 then the client will block writes while waiting for another RPC
5040 to complete instead of consuming its dirty limit. With change
5041 we get improved performance when max_dirty_mb is small.
5043 Severity : enhancement
5044 Bugzilla : 3389, 6253
5045 Description: Add support for supplementary groups on the MDS.
5046 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
5047 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
5048 lookups for user supplementary groups into a cache.
5052 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
5053 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
5054 for liblustre O_EXCL is enforced by MDS.
5058 Description: readdir never returns NULL in liblustre.
5059 Details : Corrected llu_iop_getdirentries logic, to return offset of next
5060 dentry in struct dirent.
5064 Frequency : liblustre only, depends on application IO pattern
5065 Description: liblustre clients evicted if not contacting servers
5066 Details : Don't put liblustre clients into the ping_evictor list, so
5067 they will not be evicted by the pinger ever.
5069 Severity : enhancement
5071 Description: Add ability to evict clients by NID from MDS.
5072 Details : By echoing "nid:$NID" string into
5073 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
5074 $NID would be instantly evicted from this MDS and from all active
5075 OSTs connected to it.
5079 Description: Do not query file size twice, somewhat slowing stat(2) calls.
5080 Details : lookup_it_finish() used to query file size from OSTs that was not
5085 Description: service threads change working directory to that of init
5086 Details : Starting lustre service threads may pin the working directory
5087 of the parent thread, making that filesystem busy. Threads
5088 now change to the working directory of init to avoid this.
5092 Frequency : during shutdown only
5093 Description: shutdown with a failed MDS or OST can cause unmount to hang
5094 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
5099 Frequency : 2.6 only
5100 Description: chmod/chown may include an extra supplementary group
5101 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
5102 supplementary group and if none is specified this is used.
5105 Bugzilla : 5479 (6816)
5107 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
5108 Details : If lookup is in progress on a file that is unlinked we might try
5109 to revalidate the inode and fail in revalidate after lookup is
5110 complete and ll_file_open() enqueues the open again but
5111 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
5114 Frequency : always, if lconf --abort_recovery used
5116 Description: lconf --abort_recovery fails with 'Operation not supported'
5117 Details : lconf was attempting to abort recovery on the MDT device and not
5120 ------------------------------------------------------------------------------
5122 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
5127 Frequency : rare (only unsupported configurations with a node running as an
5129 Bugzilla : 6514, 5137
5130 Description: Mounting a Lustre file system on a node running as an OST could
5132 Details : OSTs now preallocates memory needed to write out data at
5133 startup, instead of when needed, to avoid having to
5134 allocate memory in possibly low memory situations.
5135 Specifically, if the file system is mounted on on OST,
5136 memory pressure could force it to try to write out data,
5137 which it needed to allocate memory to do. Due to the low
5138 memory, it would be unable to do so and the node would
5139 become unresponsive.
5141 Severity : enhancement
5143 Description: Addition of lconf --service command line option
5144 Details : lconf now accepts a '--service <arg>' option, which is
5145 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
5147 Severity : enhancement
5149 Description: Failover mode is now the default for OSTs.
5150 Details : By default, OSTs will now run in failover mode. To return to
5151 the old behaviour, add '--failout' to the lmc line for OSTs.
5153 Severity : enhancement
5155 Description: Health checks are now provided for MDS and OSTs
5156 Details : Additional detailed health check information on MSD and OSTs
5157 is now provided through the procfs health_check value.
5160 Frequency : occasional, depends on IO load
5162 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
5163 after numerous create/delete cycles
5164 Details : The ext3 inode allocation policy would not allocate new inodes
5165 very well on the OSTs because there are no new directories
5166 being created. Instead we look for groups with free space if
5167 the parent directories are nearly full.
5172 Description: Network or server problems during mount may cause partially
5173 mounted clients instead of returning an error.
5174 Details : The config llog parsing code may overwrite the error return
5175 code during mount error handling, returning success instead
5181 Description: MDS can fail to allocate large reply buffers
5182 Details : After long uptimes the MDS can fail to allocate large reply
5183 buffers (e.g. zconf client mount config records) due to memory
5184 fragmentation or consumption by the buffer cache. Preallocate
5185 some large reply buffers so that these replies can be sent even
5186 under memory pressure.
5190 Frequency : rare (liblustre)
5191 Description: fsx running with liblustre complained that using truncate() to
5192 extend the file doesn't work. This patch corrects that issue.
5193 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
5194 ATTR_SIZE and lsm use in llu_setattr_raw.
5198 Frequency : rare, only 2.6 kernels
5199 Description: Unusual file access patterns on the MDS may result in inode
5200 data being lost in very rare circumstances.
5201 Details : Bad interaction between the ea-in-inode patch and the "no-read"
5202 code in the 2.6 kernel caused the inode and/or EA data not to
5203 be read from disk, causing single-file corruption.
5207 Frequency : rare, only 2.6 filesystems using extents
5208 Description: Heavy concurrent write and delete load may cause data corruption.
5209 Details : It was possible under high-load situations to have an extent
5210 metadata block in the block device cache from a just-unlinked
5211 file overwrite a newly-allocated data block. We now unmap any
5212 metadata buffers that alias just-allocated data blocks.
5216 Frequency : filesystems with default stripe_count larger than 77
5217 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
5218 Details : lconf specifies an inode size of 4096 bytes when the default
5219 stripe_count is larger than 77. This conflicts with the default
5220 inode density of 1 per 4096 bytes. Allocate smaller inodes in
5221 this case to avoid pinning too much memory for large EAs.
5223 ------------------------------------------------------------------------------
5225 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
5230 Frequency : rare (extremely heavy IO load with hundreds of clients)
5232 Description: Client is evicted, gets IO error writing to file
5233 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
5234 caused clients to be evicted for AST timeouts. The fixes in
5235 bug 5192 mean we no longer need to have such short AST timeouts
5236 so ldlm_timeout has been increased.
5239 Frequency : occasional during --force or --failover shutdown under load
5240 Bugzilla : 5949, 4834
5241 Description: Server oops/LBUG if stopped with --force or --failover under load
5242 Details : a collection of import/export refcount and cleanup ordering
5243 issues fixed for safer force cleanup
5246 Frequency : only filesystems larger than 120 OSTs
5247 Bugzilla : 5990, 6223
5248 Description: lfs getstripe would oops on a very large filesystem
5249 Details : lov_getconfig used kfree on vmalloc'd memory
5252 Frequency : only filesystems exporting via NFS to Solaris 10 clients
5253 Bugzilla : 6242, 6243
5254 Description: reading from files that had been truncated to a non-zero size
5255 but never opened returned no data
5256 Details : ll_file_read() reads zeros from no-object files to EOF
5261 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
5262 Details : There's little harm in aborting MDS/OSS recovery and letting it
5263 try again, so I removed the LASSERT and return an error instead.
5265 Severity : enhancement
5267 Description: New debugging infrastructure for tracking down data corruption
5268 Details : The I/O checksum code was replaced to: (a) control it at runtime,
5269 (b) cover more of the client-side code path, and (c) try to narrow
5270 down where problems occurred
5274 Bugzilla : 3819, 4364, 4397, 6313
5275 Description: Racing close and eviction MDS could cause assertion in mds_close
5276 Details : It was possible to get multiple mfd references during close and
5277 client eviction, leading to one thread referencing a freed mfd.
5279 Severity: : enhancement
5280 Bugzilla : 3262, 6359
5281 Description: Attempts to reconnect to servers are now more aggressive.
5282 Details : This builds on the enhanced upcall-less recovery that was added
5283 in 1.4.2. When trying to reconnect to servers, clients will
5284 now try each server in the failover group every 10 seconds. By
5285 default, clients would previously try one server every 25 seconds.
5290 Description: After recovery, certain operations trigger a failed
5291 assertion on a client.
5292 Details : Failing over an mds, using lconf -d --failover, while a
5293 client was doing a readdir() call would cause the client to
5294 LBUG after recovery completed and the readdir() was resent.
5296 Severity : enhancement
5298 Description: Default groups are now added by lconf
5299 Details : You can now run lconf --group <servicename> without having to
5300 manually add groups with lmc.
5303 Frequency : occasional
5305 Description: Nodes with an elan id of 0 trigger a failed assertion
5308 Frequency : always when accessing e.g. tty/console device nodes
5310 Description: tty and some other devices nodes cannot be used on lustre
5311 Details : file's private_data field is used by device data and lustre
5312 values in there got lost. New field was added to struct file to
5313 store fs-specific private data.
5316 Frequency : when exporting Lustre via NFS
5318 Description: NFSD failed occasionally when looking up a path component
5319 Details : NFSD is looking up ".." which was broken in ext3 directories
5320 that had grown large enough to become hashed.
5323 Frequency : Clusters with multiple interfaces not on the same subnet
5325 Description: Nodes will repeatedly try to reconnect to an interface which it
5326 cannot reach and report an error to the log.
5327 Details : Extra peer list entries will be created by lconf with some peers
5328 unreachable. lconf now validates the peer before adding it.
5331 Frequency : Only if a default stripe is set on the filesystem root.
5333 Description: Setting a default stripe on the filesystem root prevented the
5334 filesystem from being remounted.
5335 Details : The client was sending extra request flags in the root getattr
5336 request and did not allocate a reply buffer for the dir EA.
5339 Frequency : occasional, higher if lots of files are accessed by one client
5340 Bugzilla : 6159, 6097
5341 Description: Client trips assertion regarding lsm mismatch/magic
5342 Details : While revalidating inodes the VFS looks up inodes with ifind()
5343 and in rare cases can find an inode that is being freed.
5344 The ll_test_inode() code will free the lsm during ifind()
5345 when it finds an existing inode and then the VFS later attaches
5346 this free lsm to a new inode.
5350 Bugzilla : 6422, 7030
5351 Description: MDS deadlock between mkdir and client eviction
5352 Details : Creating a new file via mkdir or mknod (starting a transaction
5353 and getting the ns lock) can deadlock with client eviction
5354 (gets ns lock and trying to finish a synchronous transaction).
5357 Frequency : occasional
5358 Description: While starting a server, the fsfilt_ext3 module could not be
5360 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
5361 kernels. Previously, lconf would still use the ext3 name
5362 when trying to load modules. Now, it will correctly use
5363 ext3 on 2.4 and ldiskfs on 2.6.
5365 Severity : enhancement
5366 Description: The default stripe count has been changed to 1
5367 Details : The interpretation of the default stripe count (0, to lfs
5368 or lmc) has been changed to mean striping across a single
5369 OST, rather than all available. For general usage we have
5370 found a stripe count of 1 or 2 works best.
5372 Severity : enhancement
5373 Description: Add support for compiling against Cray portals.
5374 Details : Conditional compiling for some areas that are different
5378 Frequency : occasional
5379 Bugzilla : 6409, 6834
5380 Description: Creating files with an explicit stripe count may lead to
5381 a failed assertion on the MDS
5382 Details : If some OSTs are full or unavailable, creating files may
5383 trigger a failed assertion on the MDS. Now, Lustre will
5384 try to use other servers or return an error to the
5388 Frequency : occasional
5390 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
5391 caused lock timeout during readahead (since 1.4.2).
5392 Details : Processes doing readahead might match a lock that hasn't been
5393 granted yet if there are overlapping and conflicting lock
5394 requests. The readahead process waits on ungranted lock
5395 (original lock is CBPENDING), while OST waits for that process
5396 to cancel CBPENDING read lock and eventually evicts client.
5398 Severity : enhancement
5400 Description: Initial enabling of flock support for clients
5401 Details : Implements fcntl advisory locking and file status functions.
5402 This feature is provided as an optional mount flag (default
5403 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
5404 locking are implemented yet, and those that are are not guaranteed
5405 to be completely correct in production environments.
5406 mount -t lustre -o [flock|noflock] ...
5409 Frequency : occasional
5411 Description: OSTs running 2.4 kernels but with extents enabled might trip an
5412 assertion in the ext3 JBD (journaling) layer.
5413 Details : The b_committed_data struct is protected by the big kernel lock
5414 in 2.4 kernels, serializing journal_commit_transaction() and
5415 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
5416 access to this struct. In 2.6 kernels there is finer grained
5417 locking to improve SMP performance of the JBD layer.
5421 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
5426 Description: merge and cleanup kernel patches.
5429 -----------------------------------------------------------------------------
5431 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
5433 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
5434 versions of Lustre. Please update all servers and clients to
5435 version 1.4.2 or later at the same time. You must also run
5436 "lconf --write-conf {config}.xml" on the MDS while it is stopped
5437 to update the configuration logs.
5439 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
5440 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
5441 - swab llog records of type '0' so we get proper header size/idx (5861)
5442 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
5443 - fix rename of one directory over another leaking an inode (5953)
5444 - avoid SetPageDirty on 2.6 (5981)
5445 - don't re-add just-being-destroyed locks to the waiting list (5653)
5446 - when creating new directories, inherit the parent's custom
5447 striping settings if present parent (3048)
5448 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
5449 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
5450 - don't hold client locks on temporary worklist from l_lru (5666)
5451 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
5452 - automatically evict dead clients (5921)
5453 - Update file size properly in create+truncate+fstat case (6196)
5454 - Do not unhash mountpoint dentries, do not allow removal of
5456 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
5457 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
5458 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
5459 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
5460 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
5461 - Correctly handle reads of files with no objects (6243)
5462 - lctl recover will also mark a device active if deactivate used (5933)
5464 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
5465 - allow --write-conf on an MDS with different nettype than client (5619)
5466 - don't write config llogs to MDS for mounts not from that MDS (5617)
5467 - lconf should create multiple TCP connections from a client (5201)
5468 - init scripts are now turned off by default; run chkconfig --on
5469 lustre and chkconfig --on lustrefs to use them
5470 - upcalls are no longer needed for clients to recover to failover
5472 - add --abort-recovery option to lconf to abort recovery on device
5474 - add support for an arbitrary number of OSTs (3026)
5475 - Quota support protocol changes.
5476 - forward compatibility changes to wire structs (6007)
5477 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5478 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5479 - improved "lustre status" script
5480 - initialize blocksize for non-regular files (6062)
5481 - added --disable-server and --disable-client configure options (5782)
5482 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5483 - Vanilla 2.4.29 support
5484 - increase maximum number of obd devices to 520 (6242)
5485 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5486 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5487 SLES 9 2.6.5 kernels
5488 - limit stripes per file to 160 (the maximum EA size) (6093)
5490 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5493 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5494 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5495 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5496 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5497 - small ext3 extents cleanups and fixes (5733)
5498 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5499 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5500 - avoid needless client->OST connect, fix handle mismatch (5317)
5501 - fix DLM error path that led to out-of-sync client, long delays (5779)
5502 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5503 - fix several locking issues related to i_size (5492,5624,5654,5672)
5504 - don't move pending lock onto export if it is already evicted (5683)
5505 - fix kernel oops when creating .foo in unlinked directory (5548)
5506 - fix deadlock in obdfilter statistics vs. object create (5811)
5507 - use time_{before,after} to avoid timer jiffies wrap (5882)
5508 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5509 - Do not leak request if server was not able to process it (5154)
5510 - If mds_open unable to find parent dir, make that negative lookup(5154)
5511 - don't create new directories with extent-mapping (5909, 5936)
5513 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5514 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5515 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5516 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5517 - support for RHEL4 kernel on i686 (5773)
5518 - provide error messages when incompatible logs are encountered (5898)
5520 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5521 * version 1.4.0.10 (1.4.1 release candidate 1)
5523 - don't keep a lock reference when lock is not granted (4238)
5524 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5525 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5526 - drop import inflight refcount on signal_completed_replay error (5255)
5527 - unlock page after async write error during send (3677)
5528 - handle missing objects in filter_preprw_read properly (5265)
5529 - no transno return for symlink open, don't save no-trasno open (3440)
5530 - don't try to complete elan receive that already failed (4012)
5531 - free RPC server reply state on error (5406)
5532 - clean up thread from ptlrpc_start_thread() on error (5160)
5533 - readahead could read extra page into cache that wasn't ejected (5388)
5534 - prevent races in class_attach/setup/cleanup/detach (5260)
5535 - don't dereference de->d_inode after l_dput of de (5458)
5536 - use "int" for stripe value returned from lock_to_stripe (5544)
5537 - mballoc allocation and error-checking fixes in 2.6 (5504)
5538 - block device patches to fix I/O request sizes in 2.6 (5482)
5539 - look up hostnames for IB nals (5602)
5540 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5541 - don't start multiple acceptors for the same port (5277)
5542 - fix incorrect LASSERT in mds_getattr_name (5635)
5543 - export a proc file for general "ping" checking (5628)
5544 - fix "lfs check" to not block when the MDS is down (5628)
5546 - service request history (4965)
5547 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5548 - create an "evict_client" /proc entry on OSTs, like the MDS has
5549 - fix mount usage message, return errors per mount(8) (5168)
5550 - change grep [] to grep "[]" in tests so they work in more UMLs
5551 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5552 - remove ancient LOV_MAGIC_V0 EA support (5047)
5553 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5554 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5555 - allow readahead allocations to fail when low on memory (5383)
5556 - mmap locking landed again, after considerable improvement (2828)
5557 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
5559 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
5562 - send OST transaction number in read/write reply to free req (4966)
5563 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5564 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5565 - the watchdog thread now runs as interruptible (5246)
5566 - flock/lockf fixes (but it's still disabled, pending 5135)
5567 - don't use EXT3 constants in llite code (5094)
5568 - memory shortage at startup could cause assertion (5176)
5570 - reorganization of lov code
5571 - single portals codebase
5573 - add extents/mballoc support (5025)
5574 - direct I/O reads in the obdfilter (4048)
5575 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
5577 tbd Cluster File Systems, Inc. <info@clusterfs.com>
5580 - send OST transaction number in read/write reply to free req (4966)
5581 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5582 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5583 - the watchdog thread now runs as interruptible (5246)
5584 - handle missing objects in filter_preprw_read properly (5265)
5585 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5586 - drop import inflight refcount on signal_completed_replay error (5255)
5587 - unlock page after async write error during send (3677)
5588 - return original error code on reconstructed replies (3761)
5589 - no transno return for symlink open, don't save no-trasno open (3440)
5591 - add pid to ldlm debugging output (4922)
5592 - bump the watchdog timeouts -- we can't handle 30sec yet
5593 - extra debugging for orphan dentry/inode bug (5259)
5595 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
5598 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
5599 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
5600 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
5601 - fix NULL dereference in /proc/sys/portals/routes (4827)
5602 - allow failed mdc_close() operations to be interrupted (4561)
5603 - stop precreate on OST before MDS would time out on it (4778)
5604 - don't send partial-page writes before EOF from client (4410)
5605 - discard client grant for sub-page writes on large-page clients (4520)
5606 - don't free dentries not owned by NFS code, check generation (4806)
5607 - fix lsm leak if mds_create_objects() fails (4801)
5608 - limit debug_daemon file size, always print CERROR messages (4789)
5609 - use transno after validating reply (3892)
5610 - process timed out requests if import state changes (3754)
5611 - update mtime on OST during writes, return in glimpse (4829)
5612 - add mkfsoptions to LDAP (4679)
5613 - use ->max_readahead method instead of zapping global ra (5039)
5614 - don't interrupt __l_wait_event() during strace
5616 - add software watchdogs to catch hung threads quickly (4941)
5617 - make lustrefs init script start after nfs is mounted
5618 - fix CWARN/ERROR duplication (4930)
5619 - return async write errors to application if possible (2248)
5620 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
5621 - print NAL number in %x format (4645)
5622 - update barely-supported suse-2.4.21-171 series (4842)
5623 - support for sles 9 %post scripts
5624 - support for building 2.6 kernel-source packages
5625 - support for sles km_* packages
5627 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
5630 - ignore -ENOENT errors in osc_destroy (3639)
5631 - notify osc create thread that OSC is being cleaned up (4600)
5632 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
5633 - reconstruct ost_handle() like mds_handle() (4657)
5634 - create a new thread to do import eviction to avoid deadlock (3969)
5635 - let lconf resolve symlinked-to devices (4629)
5636 - don't unlink "objects" from directory with default EA (4554)
5637 - hold socknal file ref over connect in case target is down (4394)
5638 - allow more than 32000 subdirectories in a single directory (3244)
5639 - fix blocks count for O_DIRECT writes (3751)
5640 - OST returns ENOSPC from object create when no space left (4539)
5641 - don't send truncate RPC if file size isn't changing (4410)
5642 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
5643 - bind to privileged port in socknal and tcpnal (3689)
5645 - rate limit CERROR/CWARN console message to avoid overload (4519)
5646 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
5647 - basic mmap support (3918)
5648 - kernel patch series update from b1_4 (4711)
5650 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
5653 - avoid crash during MDS cleanup with OST shut down (2775)
5654 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
5655 - don't use bad inodes on the MDS (3744)
5656 - dynamic object preallocation to improve recovery speed (4236)
5657 - don't hold spinlock over lock dumping or change debug flags (4401)
5658 - don't zero obd_dev when it is force cleaned (3651)
5659 - print grants to console if they go negative (4431)
5660 - "lctl deactivate" will stop automatic recovery attempts (3406)
5661 - look for existing locks in ldlm_handle_enqueue() (3764)
5662 - don't resolve lock handle twice in recovery avoiding race (4401)
5663 - revalidate should check working dir is a directory (4134)
5665 - don't always mark "slow" obdfilter messages as errors (4418)
5667 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
5670 - don't close LustreDB during write_conf until it is done (3860)
5671 - fix typo in lconf for_each_profile (3821)
5672 - allow dumping logs from multiple threads at one time (3820)
5673 - don't allow multiple threads in OSC recovery (3812)
5674 - fix debug_size parameters (3864)
5675 - fix mds_postrecov to initialize import for llog ctxt (3121)
5676 - replace config semaphore with spinlock (3306)
5677 - be sure to send a reply for a CANCEL rpc with bad export (3863)
5678 - don't allow enqueue to complete on a destroyed export (3822)
5679 - down write_lock before checking llog header bitmap (3825)
5680 - recover from lock replay timeout (3764)
5681 - up llog sem before sending rpc (3652)
5682 - reduce ns lock hold times when setting kms (3267)
5683 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
5684 - fix NULL deref and obd_dev leak on setup error (3312)
5685 - replace some LBUG about llog ops with error handling (3841)
5686 - don't match INVALID dentries from d_lookup and spin (3784)
5687 - hold dcache_lock while marking dentries INVALID and hashing (4255)
5688 - fix invalid assertion in ptlrpc_set_wait (3880)
5690 - add libwrap support for the TCP acceptor (3996)
5691 - add /proc/sys/portals/routes for non-root route listing (3994)
5692 - allow setting MDS UUID in .xml (2580)
5693 - print the stack of a process that LBUGs (4228)
5695 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
5698 - don't cleanup request in ll_file_open() on failed MDS open (3430)
5699 - make sure to unset replay flag from failed open requests (3440)
5700 - if default stripe count is 0, use OST count for inode size (3636)
5701 - update parent mtime/ctime on client for create/unlink (2611)
5702 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
5703 - free recovery state on server during a forced cleanup (3571)
5704 - unregister_reply for resent reqs (3063)
5705 - loop back devices mounting and status check on 2.6 (3563)
5706 - fix resource-creation race that can provoke i_size == 0 (3513)
5707 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
5708 - more debugging for page-accounting assertion (3746)
5709 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
5710 - avoid deadlock after precreation failure (3758)
5711 - fix race and lock order deadlock in orphan handling (3450, 3750)
5712 - add validity checks when grabbing inodes from l_ast_data (3599)
5714 - add /proc/.../recovery_status to obdfilter (3428)
5715 - lightweight CDEBUG infrastructure, debug daemon (3668)
5716 - change default OSC RPC parameters to be better on small clusters
5717 - turn off OST read cache for files smaller than 32MB
5718 - install man pages and include them in rpms (3100)
5719 - add new init script for (un)mounting lustre filesystems (2593)
5720 - run chkconfig in %post for init scripts (3701)
5721 - drop scimac NAL (unmaintained)
5723 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
5726 - clean kiobufs before and after use (3485)
5727 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
5728 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
5729 - change init script to fail more clearly if not run as root (1528)
5730 - allow clients to reconnect during replay (1742)
5731 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
5732 - don't do DNS lookups on NIDs too small for IP addresses (3442)
5733 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
5734 - fix cond_resched (3554)
5735 - only evict unfinished clients after recovery (3515)
5736 - allow bulk resend, prevent data loss (3570)
5737 - dynamic ptlrpc request buffer allocation (2102)
5738 - don't allow unlinking open directory if it isn't empty (2904)
5739 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
5740 - remove extraneous obd dereference causing LASSERT failure (3334)
5741 - don't use get_cycles() when creating temp. files on the mds (3156)
5742 - hold i_sem when setting i_size in ll_extent_lock() (3564)
5743 - handle EEXIST for set-stripe, set proper directory name (3336)
5745 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
5746 - fix ksocknal_fmb_callback() error messages (2918)
5748 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
5751 - don't copy lvb into (possibly NULL) reply on error (2983)
5752 - don't deref dentry after dput, don't free lvb on error (2922)
5753 - use the kms to determine writeback rpc length (2947)
5754 - increment oti_logcookies when osc is inactive (2948)
5755 - update client's i_blocks count via lvb messages (2543)
5756 - handle intent open/close of special files properly (1557)
5757 - mount MDS with errors=remount-ro, like obdfilter (2009)
5758 - initialize lock handle to avoid ASSERT on error cleanup (3057)
5759 - don't use cancelling-locks' kms values (2947)
5760 - use highest lock extent for kms, not last one (2925)
5761 - don't dereference ERR_PTR() dentry in error handling path (3107)
5762 - fix thread race in portals_debug_dumplog() (3122)
5763 - create lprocfs device entries at setup instead of at attach (1519)
5764 - common AST error handler, don't evict client on completion race (3145)
5765 - zero nameidata in detach_mnt in 2.6 (3118)
5766 - verify d_inode after revalidate_special is valid in 2.6 (3116)
5767 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
5768 - initialize RPC timeout timer earlier for 2.6 (3219)
5769 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
5770 - print nal/nid for unknown nid (3258)
5771 - additional checks for oscc recovery before doing precreate (3284)
5772 - fix ll_extent_lock() error return code for 64-bit systems (3043)
5773 - don't crash in mdc_close for bad permissions on open (3285)
5774 - zero i_rdev for non-device files (3147)
5775 - clear page->private before handing to FS, better assertion (3119)
5776 - tune the read pipeline (3236)
5777 - fix incorrect decref of invalidated dentry (2350)
5778 - provide read-ahead stats and refine rpc in flight stats (3328)
5779 - don't hold journal transaction open across create RPC (3313)
5780 - update atime on MDS at close time (3265)
5781 - close LDAP connection when recovering to avoid server load (3315)
5782 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
5783 - don't leak open file on MDS after open resend (3325)
5784 - serialize filter_precreate and filter_destroy_precreated (3329)
5785 - loop device shouldn't call sync_dev() for nul device (3092)
5786 - clear page cache after eviction (2766)
5787 - resynchronize MDS->OST in background (2824)
5788 - refuse to mount the same filesystem twice on same mountpoint (3394)
5789 - allow llmount to create routes for mounting behind routers (3320)
5790 - push lock cancellation to blocking thread for glimpse ASTs (3409)
5791 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
5792 - fix rare problem with rename on htree directories (3417)
5794 - allow default OST striping configuration per directory (1414)
5795 - fix compilation for qswnal for 2.6 kernels (3125)
5796 - increase maximum number of MDS request buffers for large systems
5797 - change liblustreapi to be useful for external progs like lfsck (3098)
5798 - increase local configuration timeout for slow disks (3353)
5799 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
5801 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5804 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
5805 - don't overwrite extent policy data in reply if lock was blocked (2901)
5806 - drop filter export grants atomically with removal from device (2663)
5807 - del obd_self_export from work_list in class_disconnect_exports (2908)
5808 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
5809 - swab reply message in mdc_close, other PPC fixes (2464)
5810 - fix destroying of named logs (2325)
5811 - overwrite old logs when running lconf --write_conf (2264)
5812 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
5813 - fix race in target_handle_connect (2898)
5814 - mds_reint_create() should take same inode create lock (2926)
5815 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
5816 - don't close files for self_export to avoid uninitialized obd (2936)
5817 - allow MDS with the same name as client node (2939)
5818 - hold dentry reference for closed log files for unlink (2325)
5819 - reserve space for all logs during transactions (2059)
5820 - don't evict page beyond end of stripe extent (2925)
5821 - don't oops on a deleted current working directory (2399)
5822 - handle hard links to targets without a parent properly (2517)
5823 - don't dereference NULL lock when racing during eviction (2867)
5824 - don't grow lock extents when lots of conflicting locks (2919)
5826 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
5829 - account for cache space usage on clients to avoid data loss (974)
5830 - lfsck support in lustre kernel code (2349)
5831 - reduce journal credits needed for BRW writes (2370)
5832 - orphan handling to avoid losing space on client/server crashes
5833 - ptlrpcd can be blocked, stopping ALL progress (2477)
5834 - use lock value blocks to assist in proper KMS, faster stat (1021)
5835 - takes i_sem instead of DLM locks internally on obdfilter (2720)
5836 - recovery for initial connections (2355)
5837 - fixes for mds_cleanup_orphans (1934)
5838 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
5839 - block all file creations until orphan recovery completes (1901)
5840 - client remove rq_connection from request struct (2423)
5841 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
5842 - recovery timer race (2670)
5843 - mdc_close recovey bug (2532)
5844 - ptlrpc cleanup bug (2710)
5845 - mds timeout on local locks (2588)
5846 - namespace lock held during RPCs (2431)
5847 - handle interrupted sync write properly (2503)
5848 - don't try to handle a message that hasn't been replied to (2699)
5849 - client assert failure during cleanup after abort recovery (2701)
5850 - leak mdc device after failed mount (2712)
5851 - ptlrpc_check_set allows timedout requests to complete (2714)
5852 - wait for inflight reqs when ptlrpcd finishes (2710)
5853 - make sure unregistered services are removed from the srv_list
5854 - reset bulk XID's when resending them (caught by 1138 test)
5855 - unregister_bulk after timeout
5856 - fix lconf error (2694)
5857 - handle write after unfinished setstripe, stripe-only getstripe (2388)
5858 - readahead locks pages, leaves pending causing memory pressure (2673)
5859 - increase OST request buffers to 4096 on large machines (2729)
5860 - fix up permission of existing directories in simple_mkdir (2661)
5861 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
5862 - don't assign transno to errored transactions (2742)
5863 - don't delete objects on OST if given a bogus objid from MDS (2751)
5864 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
5865 - if rq_no_resend, then timeout request after recovery (2432)
5866 - fix MDS llog_logid record size, 64-bit array alignment (2733)
5867 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
5868 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
5869 - MDS recovery shouldn't delete objects that it has given out (2730)
5870 - if enqueue arrives after completion, don't clobber LVB (2819)
5871 - don't unlock pages twice when trigger_group_io returns error (2814)
5872 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
5873 - don't write pages to disk if there was an error (1450)
5874 - don't ping imports that have recovery disabled (2676)
5875 - take buffered bytes into account when balancing socknal conn (2817)
5876 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
5877 - reconnect unlink llog connection after MDS reconnects to OST (2816)
5878 - remove little-endian swabbing of llog records (1987)
5879 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
5880 - retry reposting request buffers if they fail (1191)
5881 - grow extent at grant time to avoid granting a revoked lock (2809)
5882 - lock revoke doesn't evict page if covered by a second lock (2765)
5883 - disable VM readahead to avoid reading outside lock extents (2805)
5885 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
5886 - updated kernel patches for hp-2.4.20 kernel (2681)
5888 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
5891 - fix truncated write corruption (2366)
5892 - fix for failed assertion in iopen_connect_dentry (1792,2517)
5894 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
5895 - ensure len(uuid) < 37 in lmc (1171)
5896 - fix ia64 OOPS in llog_test (2255)
5897 - zero end of page at obdfilter for partial page writes (2648)
5898 - don't leave stale dentries around after renames (bug 2428)
5899 - fix timeouts when evicting a client with a single lock held (2642)
5900 - set deadline for the initial HELLO message to drain (2634)
5901 - print out dotted-quad IP addresses in the socknal (2302)
5903 - additional debugging for MDS client eviction problem (2443)
5904 - fix mkfsoptions support for osts (2603, 2604)
5906 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
5909 - add series for the vanilla 2.6.0 kernel
5910 - add series for the vanilla 2.4.24 kernel
5911 - add series for a cray x86/64 UL kernel drop
5912 - fix xattr patches for the vanilla 2.4.19 series
5914 - generate true UUIDs in lmc (1171)
5915 - have portals stack dumping break in UML (2466)
5916 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
5917 - allocate with _NOFS in ldlm to avoid deadlock (1933)
5918 - wake callback waiting threads on client eviction (2460)
5919 - Add --ptldebug and --subsystem to lmc (1719)
5920 - update assertion to allow safe interrupt allocation
5921 - set rq_no_resend for cancel requests (2432)
5922 - recalculate ptlrpcd timeout after resend (2494)
5923 - call vfs_rmdir when removing pending directories (2368)
5924 - fix renaming a file to itself (2429)
5925 - lmc creates a default one-stripe lov (2454)
5926 - expand procfs space to handle large clusters (2326)
5927 - increase UML stack to avoid overflow
5928 - update lconf's list of debug and subsystem masks
5929 - fix lfs find --obd (2510)
5930 - /proc tunable for disabling filter read caching (2591)
5931 - stop rpm packages from altering slapd.conf (2301)
5932 - disable nagle in the socknal under 0conf (2578)
5933 - choose mds inode size based on stripe count (2572)
5934 - fix kernel-source rpm problems (2516)
5936 - add --disable-doc to avoid pdf generation (2421)
5937 - update documentation, tests, type-os, comments
5938 - avoid format warnings on ia64
5939 - remove the TOE NAL
5940 - tiny code cleanups by removing unused fields
5942 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
5945 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
5946 - remove the most blatant lies from BUILDING file (2371)
5947 - change default debug level to reasonable production setting
5948 - reduce client side cache size to reduce cache flush time
5949 - reduce max RPCs in flight to avoid unnecessary file fragmentation
5950 - make TCP zerocopy and pinger support enabled by default (2476)
5951 - sync writes completed after process exits caused crashes (2319)
5952 - maintain correct mount count on the MDS (2356)
5953 - backout 1557, because 2316 wasn't really fixed
5954 - better file I/O statistics gathering in /proc
5955 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
5956 - another kernel patch to fix zero-copy TCP function export
5957 - don't take duplicate lock when processing re-sent getattr (2420)
5958 - lctl uses obd_self_export instead of creating new conn (2353)
5959 - MDS/OST recovery case which requires object creation asserted (2425)
5960 - move lfs from /usr/sbin to /usr/bin in packages
5961 - fix race between mds_client_add and mds_client_free (2417)
5962 - use kmalloc instead of slabs in portals (2430)
5963 - don't create duplicate records when a failover MDS is present (2442)
5964 - remove unnecessary mount age check (2332)
5965 - don't remove directory inodes from locks prematurely (2451)
5966 - don't break if MDS service name is the same as hostname (2103)
5967 - fix races in client write RPC generation when cache full (2482)
5969 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
5972 - remove now-unused request->rq_obd (278)
5973 - if an allocation fails, print out how much memory we've used (1933)
5974 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
5975 - add the "configurable stack size" patch to most series files (1256)
5976 - ability to write large log records, for 100+ OST configs (2306)
5977 - fix NULL deref when filter_prep fails (2314)
5978 - fix operator precedence error in filter_sync
5979 - dynamic allocation of socknal TX descriptors (2315)
5980 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
5981 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
5982 - ensure that CWARN messages in Portals always get to the syslog
5983 - __init/__exit are not for prototype decls (ldlm_init/exit)
5984 - x86-64 compile warning fixes
5985 - fix gateway LMC keyword conflict (2318)
5986 - fix MDS lock inversions in getattr/reint paths (1844)
5987 - fix a rare lock re-ordering bug, which caused deadlock (2322)
5988 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
5989 - DLM race condition prevented some lock evictions (2328)
5990 - ENOMEM detection and retry on socknal sends (2230)
5991 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
5992 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
5993 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
5994 - better error messages when a client is rejected during recovery (1505)
5995 - avoid cancelling locks which were never granted, after failure (2330)
5996 - fix i_sem/journal inversion in mds_client_add (2333)
5997 - fix truncate/getattr lock cycle deadlock (2334)
5998 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
5999 - fix two rare exit paths which could leak an l_lock() ref (2321)
6000 - fencepost error in MDS/OST orphan recovery (2226)
6001 - make log record alignment 8 bytes (1988)
6002 - lstripe now fails when requested offset > ost_count (2237)
6003 - ensure that all kernel series have a complete list.h (1607)
6004 - fix crashes in special-file operations (2316)
6005 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
6007 - allow configurable automake binary, for testing new versions
6008 - small update to the lfs documentation
6010 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
6012 * fix negative export reference count in fsfilt_sync (2312)
6014 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
6015 * release candidate 0.9.1
6017 - orphans are moved into the PENDING directory for possible recovery
6018 - replayed opens now open by fid for orphan/rename safety (1042)
6019 - last close of an orphan inode generates a transno (683)
6020 - chdir() and mount() now pin the directory entry (1020)
6021 - avoid CERROR in normal ll_setattr_raw() error case (1500)
6022 - discard very old requests without processing them (1502)
6023 - remove some common, well-understood CERRORs (1505)
6024 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
6025 - clear "grant" flags in OST replies until OST grant code lands (1644)
6026 - fix read performance by not clobbering i_blksize on client (1598)
6027 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
6028 - make LRU size a /proc tunable, clears locks when reduced (707)
6029 - fix some lprocfs rot that prevented ptlbd from loading (1732)
6030 - server locks take references on exports now (1558)
6031 - build fixes for 2.4.20-rh trees (1663)
6032 - return an error from lov_create if all OSCs are inactive (1751)
6033 - fix import levels when a reconnect happens without a timeout (1597)
6034 - exit early from mds_open if we get a lookup error (1749)
6035 - partial page read at EOF wouldn't wait for disk before sending (1642)
6036 - avoid NULL deref in obdfilter when reading page past EOF (1592)
6037 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
6038 - fix LBUG in ll_it_open_error with rc = -2 (1861)
6039 - write/truncate lock inversion (1639)
6040 - Don't auto-load obdclass, portals modules during cleanup (1495)
6041 - fix timestamps from jumping to "now" (1763)
6042 - extra journal assertions (1648)
6043 - add an extra multiunlink test (1771)
6044 - fix read_record/write_record API (1776)
6045 - fix leak of offset_extent, possible incorrect i_size later (1772)
6046 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
6047 - Debugging for the kqswnal_get_idle_tx problems (1820)
6048 - Allow recovery to be attempted multiple times (1536)
6049 - Write out MDS last_rcvd file after it is first created (1600)
6050 - Fix tx_descriptor leak in failed transmit situations (1827)
6051 - ext3 journaling fixes for assertion failure after IO error (1871)
6052 - class_export_put() on freed export after completion AST error (1896)
6053 - Fix revalidate looping in VFS (1322)
6054 - Don't access a freed export during MDS_REINT timeout (1521)
6055 - Add open-unlink recovery support on the MDS (1673,1764)
6056 - Return an error if no MDS data was read from last_rcvd (1946)
6057 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
6058 - Don't setup a disk device that doesn't match exported UUID (317)
6059 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
6060 - avoid committing NULL handle in force close
6061 - local.sh is now a one-stripe LOV configuration
6062 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
6063 - fix ext3 htree duplicate directory entry corruption (1516)
6064 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
6065 - update to new LOV EA format (2097)
6066 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
6068 2003-06-15 Phil Schwan <phil@clusterfs.com>
6071 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
6072 - per-import/export recovery handling (958, 931, 959)
6073 - multiple last-rcvd slots, for serving multiple FSes (949)
6074 - connections are again shared between multiple imp/exports (963, 964)
6075 - "umount -f" would hang if any requests needed to be sent (393, 978)
6076 - avoid pinning large req buffer by copying for queued messages (989)
6077 - add "uuid" to "lctl device" command to help upcalls (991)
6078 - "open" RPCs with transnos would confuse recovery counters (1037)
6079 - do proper endian conversion of all wire messages (288, 340, 891)
6080 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
6081 - call waiting locks callback from LDLM recovery thread (1127, 1151)
6082 - fix ptlrpc_connection leak in target_handle_connect (1174)
6083 - fix import refcounting bug in OST and MDS cleanup (1134)
6084 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
6085 - fix very unlikely obd_types race condition (501)
6086 - remove osc_open hack for echo_client (1187)
6087 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
6088 - a failure in read_inode2 leads to deadlock (1139)
6089 - cancel ack-locks as soon as transaction is committed (1072)
6090 - fix major leaks and crashes in the bulk I/O path (937, 1057)
6091 - make sure to commitrw after any preprw to avoid deadlock (1162)
6092 - failing to execute a file in a lustre FS would lock inode (1203)
6093 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
6094 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
6095 - obd_uuid2tgt would walk off the end of the list (1255)
6096 - on IA64 the osc would give portals incorrect bulk size (1258)
6097 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
6098 - fix lock inversion caused by new llite matching code (1282)
6099 - limit the number of dirty pages on a client to 10MB (1286)
6100 - timed out locks were not being corrected cancelled (1289)
6101 - fix O_DIRECT above 4GB on IA-32 (1292)
6102 * major user-visible changes
6103 - fail out/fail over policy now controlled by the upcall (993)
6105 - add OBD_PING to check server availability and failure (954)
6106 - lustre messages are now sent in sending host order (288, 340, 891)
6107 - add eadatalen to MDS getattr reply (340)
6108 - OST read replies may contain second buffer, with per-page status (593)
6110 2003-03-11 Phil Schwan <phil@clusterfs.com>
6113 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
6114 - failed open()s could cause deadlock; fixed (867, 869)
6115 - stop cancelling OST locks when files are closed (481)
6116 - overlapping XID spaces caused network corruption (851, 853)
6117 - fix unsafe fsfilt counter arithmetic; change to atomic_t
6118 - setattr_raw added, to do single-RPC, server-side setattrs
6119 - lmc/lconf syntax change for OST UUIDs
6120 - fix crashy race condition between ptlrpc_free_req and osc_close
6121 - don't use request in mdc_enqueue if we hit a timeout (889)
6122 - don't set the inode i_size for regular files from the MDS (896)
6123 - handle out of order completion AST (842)
6124 - don't LBUG if a lock request times out after receiving AST (913)
6125 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
6126 - if a bad lock AST arrives, send an error instead of dropping entirely
6127 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
6128 - fix leak in bulk IO when only partially completed (899, 900, 926)
6129 - fix O_DIRECT for ia64 (55)
6130 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
6131 - C-z after timeout could hang a process forever; fixed (977)
6133 - client-side I/O cache (678, 924, 929, 941, 970)
6135 - READPAGE and SETATTRs which don't take server-side locks get
6138 2003-02-11 Phil Schwan <phil@clusterfs.com>
6141 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
6142 - Fix fsfilt_extN_readpage() to read a full page of directory
6143 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
6144 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
6145 - Fix imbalanced LOV object allocation and out-of-bound access (469)
6146 - Most intent operations were removed, in favour of a new RPC mode
6147 that does a single RPC to the server and bypasses most of the VFS
6148 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
6149 - Aggressively cancel local locks on DLM servers
6150 - mds_reint_unlink sends EA to the client if it's the last nlink.
6151 client uses that EA to unlink OST objects.
6152 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
6153 - recursive symlinks were fixed (439)
6154 - fixed NULL deref in DEBUG_REQ
6155 - filter_update_lastobjid no longer calls sync, which annoyed extN
6156 - fixed multi-client small-writes to a single file problem (445)
6157 - fixed mtime updates during file writes (607)
6158 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
6159 - fixed bug in obd_brw_read/write() (under guise of testing 367)
6160 - fixed Linux OST size reporting problem (444, 656)
6161 - OST now updates object mtime with writes or setattr (607, 619)
6162 - client verifies file size before zeroing page past EOF (445)
6163 - OST now writes last allocated objid to disk with allocation (108)
6164 - LOV on echo now works (409)
6166 - mds_reint_unlink sends a new buffer, with the EA included. this
6167 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
6168 set if a regular file was being unlinked, and it was the last link
6169 - use PtlGet from the target for bulk writes (315)
6170 - OST now updates object mtime with writes or setattr (607, 619)
6171 - LDLM now has a grant-time callback to revalidate locked items, if
6173 - Many MDS operations were reorganized to combat race conditions
6175 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
6176 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
6178 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
6181 - Fully reactivate OST imports after reconnection (512, others)
6182 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
6183 - More graceful error handling for truncating on dead OST (515)
6184 - Don't error out unless we're actually accessing dead stripes (474)
6185 - Fix garbage sizes when stripes are missing (410)
6186 - LRU counters were broken, causing constant lock purge (433, 432)
6187 - garbage on read from stripes with failed OSTs (441)
6188 - mark OSCs as active before reconnecting during recovery (438)
6189 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
6190 - lfind did not preserve OST order in output (443)
6191 - symlinks cause hung clients, incorrect data (439)
6192 - stop dereferencing request after dropping refcount (457)
6193 - don't LASSERT(spin_is_locked) on non-SMP (455)
6194 - fixes for many rename() bugs
6195 - fstat didn't correctly synchronize attributes (399)
6196 - server must handle lock cancellation during blocking AST prep (487)
6197 - bulk descriptors were free()d too soon (511)
6198 - fix paths in lconf, which would load incorrect modules (451, 507)
6199 - fix confusing lconf 'host not found' error message (386)
6200 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
6201 - fix race condition in mdc_blocking_ast() for inode access (526)
6202 - fix lov_unpackmd() unpacking wrong number of stripes (537)
6203 - fix lov_set_osc_active() marking wrong OSC inactive (440)
6204 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
6205 - fix multiple writes of stripe MD to MDS (358, maybe 519)
6206 - fix lstripe in several ways (kernel side) (527)
6207 - fix request leak in ldlm_cli_enqueue (262)
6208 - incorrect OSC was marked inactive after OST failure
6209 - call mds_fs_cleanup before unmounting filesystem (524)
6210 - fix races between taking ns_lock and ldlm_lock_change_resource
6211 - fix races updating LOV export open file list
6212 - fix lov_enqueue error path, avoid decref-ing bad lock handle
6213 - fix recovery NULL deref in ldlm_cli_cancel_unused
6214 - fix some DLM races by using new hash table for lock handles (419)
6215 - permit the client to specify desired inodes, at replay
6216 - duplicate requests when we queue them for replay reintegration
6217 - fix last_rcvd offset calculation
6218 - sync after each recovered transaction, so we always make progress
6219 - never, not always, ERESTART requests without transnos
6220 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
6222 - skip replay if the MDS says that the client is already connected
6223 - don't check for a recovery-enabled export to match lctl's UUID
6224 - don't INC_USE_COUNT for phantom exports
6225 - don't crash when cleaning up phantom exports (567)
6226 - don't double-finish or set replay data for errored mdc_open requests
6227 - abort requests when they time out, so we don't get old replies
6228 - send/receive replies for AST messages again
6229 - if the client says that it doesn't have the lock, cancel it on the
6231 - if we timeout during I/O, don't try to cancel an in-use lock; instead
6232 mark it as destroyed, it will all work out when decref is called
6233 - fix module use counts (22, 581)
6235 - ASTs now expect a reply (server cancels lock on error reply)
6237 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
6240 - fix many simultaneous client startup (392)
6241 - fix dentry->d_it clobbering
6242 - credentials weren't being shipped for readdir/getattr operations
6243 - remove invalid assertions triggered during some concurrent MD
6245 - proper Lustre versions added (336, 389)
6246 - fix memory leak for create error case (398)
6247 - fix LOV locking bug that would get cli/srv out of sync
6248 - fix echo client over LOV (409)
6249 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
6250 - fix double-O_EXCL intent crash (424)
6251 - avoid sending multiple lock CANCELs (352)
6253 - MDS can do multi-client recovery (modulo bugs in new code)
6255 - many updates, edits, cleanups
6257 2002-11-18 Phil Schwan <phil@clusterfs.com>
6260 - fix null d_it dereference (346)
6261 - fix full OST/dbench hang (333)
6262 - fix permission problem with file removal (286)
6263 - fix removal of OSCs from LOV when they fail
6264 - fix NULL deref during bulk timeout (214)
6265 - fix problems related to multiple filesystems on one MDS (241)
6266 - fixed serious subtle metadata locking bugs
6267 - free locks on clients when inodes are removed due to memory
6269 - fix inode pointer in lock data (285)
6270 - partial support for multiple MDS on a single host (241)
6271 - data locks weren't cancelled at clear_inode time (290, 311)
6272 - intent locks could lead to unbounded lock growth (205)
6273 - added a maximum lock count, an LRU list, and a flusher
6274 - fix multiple rename (365)
6275 - properly abstracted the echo client
6276 - OSC locked 1 byte too many; fixed
6277 - rewrote brw callback code:
6278 - fixed recovery bugs related to LOVs (306)
6279 - fixed too-many-pages-in-one-write crash (191)
6280 - fixed (again) crash in sync_io_timeout (214)
6281 - probably fixed callback-related race (385)
6283 - Add capability to MDS protocol
6284 - LDLM cancellations and callbacks on different portals
6286 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
6289 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
6290 - defer open object destruction to close time (601981, 138)
6291 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
6292 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
6293 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
6294 - don't call obd_setattr in ll_file_release for destroyed objects
6295 * protocol change to lustre_msg: move |version| and add |flags|
6296 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
6297 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
6298 * added replay of create, unlink, link and rename operations during
6299 MDS failover; recovery should be much more robust now
6300 * remove failed OSCs from LOVs (only lov_create uses this so far)
6301 * the lustre-HOWTO was brought (more) up to date (582544)
6303 2002-10-23 Phil Schwan <phil@clusterfs.com>
6306 - in-use dentries weren't being reused properly (617851)
6307 - prevent multiple LDLM setup (599178)
6308 - fix LOV size calculations for truncate (617853)
6309 - fix client handling of MDS intent errors (POSIX)
6310 - fix permission bug in lovstripe.c test (624321)
6311 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
6312 - truncate past end of file could corrupt data
6313 - proper cleanup after timeouts, crashes, etc (592524, 550815)
6314 - a race in recovery could return ETIMEDOUT to apps (623947)
6315 - building outside the source directory was fixed
6316 * the lustre-HOWTO was brought (more) up to date (582544)
6317 * major progress was made on recovery functionality
6319 2002-10-10 Phil Schwan <phil@clusterfs.com>
6322 - recovery deadlock fix
6323 - rm -rf causes LBUG fix (617817)
6324 - file open by multiple tasks fix (618962)
6325 - directory permissions bugs (602707 and 620007)
6326 - journal_stop fixed with locking (611313)
6327 - O_APPEND failures resolved (618273, perhaps 614459)
6328 - lconf PATH fix (619770)
6329 - IA64 build fix (621450)
6330 - RPC buffer sizes scale with amount of memory
6332 2002-10-01 Phil Schwan <phil@clusterfs.com>
6335 - locks would be cancelled without throwing away data pages,
6336 resulting in inconsistent data (605627)
6337 - inode attributes were not always being refreshed (605627, 612449)
6338 - lconf now continues to cleanup after lctl reports an error
6339 - MDS now enforces user permissions (602707)
6340 - lprocfs cleanup fixed, but not yet enabled (614157)
6341 - fixed infinite server hang, should a client not respond to an AST
6342 - avoid going into recovery if user calls readlink() with a buffer
6343 that's too small (613941)
6344 - AST RPCs no longer require replies (614867) -- this may be changed
6345 - don't crash server if client sends an IOV that's too big (611336)
6346 - fixed lock conversion deadlock (611892)
6347 - fixed the following of symlinks (614622)
6348 * recovery: the server can remove locks from a client that dies, other
6349 clients can make progress
6350 * more extN patch fixes
6351 * compile-time configurable ptlrpc buffer allocations
6353 - collaborative read cache document
6354 - Lustre Lite Performance CDR document-in-progress
6356 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6359 - fix typo in patch-2.4.18
6361 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6364 - clear ptlrpc request each time in handle_incoming_request()
6365 - unlink of files now destroys the object on the OST
6367 2002-09-19 Peter Braam <braam@clusterfs.com>
6369 * add hard link support
6370 * change obdfile creation method
6371 * kernel patch changed
6373 2002-09-19 Peter Braam <braam@clusterfs.com>
6376 - stack overflow bug in extN fixed
6378 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
6380 * documentation updates
6381 - add man pages for config tools
6382 - update tests/README to describe testing with new config tools
6383 - finish metadata API descriptions
6384 * bug fixes and cleanups
6385 - statfs workaround for 16TB limit
6386 - LOV stripe allocation improved, can stripe on subset of OSTs
6387 - LOV file size/IO offset was wrong for files > 4GB in size
6388 - object EA data was being dropped, caused files to be unreadable
6389 - memory overflow with non-LOV OST caused memory corruption
6390 - fixed regression tests to work with new config tools, obdfilter
6391 - fixed bug when directory size became larger than 1 block
6392 - fixed bug (for single client case) when PWD was deleted
6393 - invalidate local directory pages when doing intent-based ops
6394 - avoid LDLM oops when lock callback contained bad data
6396 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
6398 * documentation updates
6399 * bug fixes and cleanups
6400 - configuration tools
6403 - 64-bit compile warnings
6404 - 64-bit internal statfs data
6406 * test_brw on persistent OST devices
6408 * lprocfs (disabled)
6410 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
6412 * documentation updates
6413 * bug fixes and cleanups
6414 * configuration tools
6416 2002-08-30 Peter J. Braam <braam@clusterfs.com>
6419 * many small fixes to 0_5_4
6420 * io/network handling
6421 * thinkos in MDS operations
6423 2002-08-24 Peter J. Braam <braam@clusterfs.com>
6426 * crucial basic fixes to 0.5.3
6427 * IOR, Iozone work over Elan
6430 2002-08-07 Phil Schwan <phil@clusterfs.com>
6431 * version 0_5_3, our first alpha
6432 * we use the new Portals iovs
6433 * documentation updates
6434 * bug fixes and cleanups
6435 * small changes in the DLM wire protocol
6437 2002-07-25 Peter J. Braam <braam@clusterfs.com>
6438 * version 0_5_1 with some initial stability,
6439 * locking on MD and file I/O.
6440 * documentation updates
6441 * several bug fixes since 0.5.0
6442 * small changes in wire protocol
6444 2002-07-18 Phil Schwan <phil@clusterfs.com>
6446 * delivered as Lustre Light Alpha
6447 * fixed a crash after handling invalid MDS requests
6448 * fixed directory pages for architectures with non-4k pages sizes
6450 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
6451 * release version v0_4_4
6452 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
6454 * Fixes a number of interruption problems with OST operations.
6455 * Update documentation for portals header changes
6456 * Move all wire protocol structs/defines to lustre_idl.h
6457 * Fixes symlink length bug.
6458 * Add tcpdump to repository.
6460 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
6461 * release version v0_4_3
6462 * Fixes statfs for inodes on extN.
6463 * Fixes bug in runtests which would delete /etc/hosts.
6464 * Use 64-bit object IDs wherever possible (not into VFS though)
6465 Remove ost_get_info, which is unused by lustre, and out of date.
6467 2002-07-03 Peter Braam <braam@clusterfs.com>
6468 * release version v0_4_2 Fixes a lookup error (type not passed)
6469 * move forward to head of Portals
6470 * move forward to latest Lustre kernel
6472 2002-06-25 Peter Braam <braam@clusterfs.com>
6473 * release version v0_4_1. Hopefully stable on single node use.