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 2.6.32-504.30.3.el6 (RHEL6.6)
8 2.6.32-573.8.1.el6 (RHEL6.7)
9 3.10.0-229.20.1.el7 (RHEL7.1)
10 3.0.101-0.47.67 (SLES11 SP3)
11 3.0.101-65 (SLES11 SP4)
12 vanilla linux 4.2.1 (ZFS only)
13 * Client known to build on unpatched kernels:
14 2.6.32-431.29.2.el6 (RHEL6.5)
15 2.6.32-504.30.3.el6 (RHEL6.6)
16 2.6.32-573.8.1.el6 (RHEL6.7)
17 3.10.0-229.20.1.el7 (RHEL7.1)
18 3.0.101-0.47.67 (SLES11 SP3)
19 3.0.101-65 (SLES11 SP4)
22 * Recommended e2fsprogs version: 1.42.13.wc3 or newer
23 * Recommended ZFS / SPL version: 0.6.5.3
24 * Tested with ZFS / SPL version: 0.6.5.3
25 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
26 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
27 stack overflow. For more information, please refer to bugzilla 17630.
28 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
29 the re-exporting nodes to be mounted with "32bitapi" mount option
31 --------------------------------------------------------------------------------
33 03-10-2015 Intel Corporation
35 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
36 for currently supported client and server kernel versions.
37 * Server known to build on patched kernels:
38 2.6.32-431.29.2.el6 (RHEL6.5)
39 2.6.32-504.8.1.el6 (RHEL6.6)
40 3.0.101-0.46 (SLES11 SP3)
41 * Client known to build on unpatched kernels:
42 2.6.32-431.29.2.el6 (RHEL6.5)
43 2.6.32-504.8.1.el6 (RHEL6.6)
44 3.10.0-123.20.1.el7 (RHEL7)
45 3.0.101-0.46 (SLES11 SP3)
46 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
47 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
48 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
49 stack overflow. For more information, please refer to bugzilla 17630.
50 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
51 the re-exporting nodes to be mounted with "32bitapi" mount option
53 Severity : enhancement
55 Description: control OST-index in IDIF via ROCOMPAT flag.
56 Details : Introduce new flag OBD_ROCOMPAT_IDX_IN_IDIF that is stored in the
57 last_rcvd file. For new formatted OST device, it will be auto set;
58 for the case of upgrading from old OST device, you can enable it
59 via the lproc interface osd-ldiskfs.index_in_idif. With such flag
60 enabled, for new created OST-object, its IDIF-in-LMA will contain
61 the OST-index; for the existing OST-object, the OSD will convert
62 old format IDIF as new format IDIF with OST-index stored in the
63 LMA EA when accessing such OST-object or via OI scrub. Once such
64 flag is enabled, it cannot be reverted back, so the system cannot
65 be downgraded to the orignal incompatible version.
67 --------------------------------------------------------------------------------
69 07-30-2014 Intel Corporation
71 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
72 for currently supported client and server kernel versions.
73 * Server known to build on patched kernels:
74 2.6.32-431.20.3.el6 (RHEL6.5)
75 3.0.101-0.31 (SLES11 SP3)
76 * Client known to build on unpatched kernels:
77 2.6.32-431.20.3.el6 (RHEL6.5)
78 3.0.101-0.31 (SLES11 SP3)
81 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
82 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
83 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
84 stack overflow. For more information, please refer to bugzilla 17630.
85 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
86 the re-exporting nodes to be mounted with "32bitapi" mount option
88 --------------------------------------------------------------------------------
90 10-31-2013 Intel Corporation
92 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
93 * Server builds for kernels:
94 2.6.32-358.18.1.el6 (RHEL6.4)
95 * Client builds for unpatched kernels:
96 2.6.32-358.18.1.el6 (RHEL6.4)
97 3.0.92-0.8 (SLES11 SP3)
99 * Recommended e2fsprogs version: 1.42.7.wc2 or newer
100 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
101 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
102 stack overflow. For more information, please refer to bugzilla 17630.
103 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
104 the re-exporting nodes to be mounted with "32bitapi" mount option
106 --------------------------------------------------------------------------------
108 05-30-2013 Intel Corporation
110 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
111 * Server builds for kernels:
112 2.6.32-358.6.2.el6 (RHEL6.4)
113 * Client builds for unpatched kernels:
114 2.6.32-358.6.2.el6 (RHEL6.4)
115 3.0.74-0.6.6 (SLES11 SP2)
117 * Recommended e2fsprogs version: 1.42.7.wc1 or newer
118 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
119 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
120 stack overflow. For more information, please refer to bugzilla 17630.
121 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
122 the re-exporting nodes to be mounted with "32bitapi" mount option
124 --------------------------------------------------------------------------------
126 9-30-2012 Whamcloud, Inc.
128 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
129 * Server support for kernels:
130 2.6.32-279.5.1.el6 (RHEL6)
131 * Client support for unpatched kernels:
132 2.6.32-279.5.1.el6 (RHEL6)
134 2.6.18-238.19.1.el5 (RHEL5)
135 2.6.16.60-0.69.1 (SLES10)
136 * Recommended e2fsprogs version: 1.41.90.wc4
137 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
138 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
139 stack overflow. For more information, please refer to bugzilla 17630.
140 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
141 the re-exporting nodes to be mounted with "32bitapi" mount option
143 Severity : enhancement
145 Description: Update RHEL6.3 kernel to 2.6.32-279.5.1.el6.
147 Severity : enhancement
149 Description: Update RHEL6.3 kernel to 2.6.32-279.2.1.el6.
151 Severity : enhancement
153 Description: Update RHEL6.2 kernel to 2.6.32-220.17.1.el6
155 Severity : enhancement
157 Description: Update RHEL6.2 kernel to 2.6.32-220.13.1.el6
159 Severity : enhancement
161 Description: Update RHEL6.2 kernel to 2.6.32-220.7.1.el6
163 --------------------------------------------------------------------------------
165 04-30-2012 Whamcloud, Inc.
167 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
168 * Server support for kernels:
169 2.6.32-220.4.2.el6 (RHEL6)
170 * Client support for unpatched kernels:
171 2.6.32-220.4.2.el6 (RHEL6)
173 2.6.18-238.19.1.el5 (RHEL5)
174 2.6.16.60-0.69.1 (SLES10)
175 * Recommended e2fsprogs version: 1.41.90.wc4
176 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
177 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
178 stack overflow. For more information, please refer to bugzilla 17630.
179 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
180 the re-exporting nodes to be mounted with "32bitapi" mount option
182 Severity : enhancement
184 Description: Update RHEL6.2 kernel to 2.6.32-220.4.2.el6
186 Severity : enhancement
188 Description: Update RHEL6.2 kernel to 2.6.32-220.4.1.el6
190 Severity : enhancement
192 Description: Update RHEL6.2 kernel to 2.6.32-220.el6
194 Severity : enhancement
196 Description: Update RHEL6.1 kernel to 2.6.32-131.17.1.el6
198 --------------------------------------------------------------------------------
200 09-30-2011 Whamcloud, Inc.
202 * Server support for kernels:
203 2.6.18-238.19.1.el5 (RHEL5)
204 2.6.32-131.6.1.el6 (RHEL6)
205 * Client support for unpatched kernels:
206 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
207 2.6.18-238.19.1.el5 (RHEL5)
208 2.6.32-131.6.1.el6 (RHEL6)
209 2.6.29.4-167.fc11 (FC11)
210 2.6.16.60-0.69.1 (SLES10)
212 * Recommended e2fsprogs version: 1.41.90.wc3
213 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
214 Lustre filesystem with 4K stack may cause a stack overflow. For more
215 information, please refer to bugzilla 17630.
216 * NFSv4 reexport requires Lustre client on the re-exporting nodes to be
217 mounted with "32bitapi" mount option
219 Severity : enhancement
220 Description: Update RHEL6.1 kernel to 2.6.32-131.6.1.el6
222 Severity : enhancement
223 Description: Update RHEL5.6 kernel to 2.6.18-238.19.1.el5
225 Severity : enhancement
226 Description: Update RHEL5.6 kernel to 2.6.18-238.9.1.el5
229 Frequency : only with SLES10
231 Description: Use OFED "KMP" provided by Novell
232 Details : SLES10 SP3 ships with OFED in a separate "KMP" package.
233 Lustre is now built against this package. That means you need to
234 install the ofed-kmp package from Novell for the patchless client.
235 Note that the ofed-kmp that Novell ships may not exactly match the
236 kernel version but should still be compatible.
238 Severity : enhancement
240 Description: Update RHEL5.5 kernel to 2.6.18-194.17.1.el5
241 Update OEL5.5 kernel to 2.6.18-194.17.1.0.1.el5.
243 --------------------------------------------------------------------------------
245 2010-07-15 Oracle, Inc.
247 * Server support for kernels:
248 2.6.18-164.11.1.el5 (RHEL 5)
249 2.6.18-164.11.1.0.1.el5 (OEL 5)
250 * Client support for unpatched kernels:
251 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
252 2.6.18-164.11.1.el5 (RHEL 5),
253 2.6.18-164.11.1.0.1.el5 (OEL 5)
254 2.6.16.60-0.42.8 (SLES 10),
256 2.6.29.4-167.fc11 (FC11)
257 * Recommended e2fsprogs version: 1.41.10-sun2
259 Severity : enhancement
261 Description: Distribute lfs_migrate script with lustre
262 Details : The lfs_migrate mentioned in the manual is now included in the
263 lustre tree in lustre/scripts and will distributed in lustre RPMs.
265 Severity : enhancement
267 Description: Failover nid list added to import proc file.
269 Severity : enhancement
271 Description: "som_preview" mount option required to use SOM
272 Details : The clients will need to use som_preview mount option
276 Description: Lustre does not do 1MB IOs to HW RAID
277 Details : Bump MAX_PHYS/HW_SEGMENTS and SG_ALL to 256 in the RHEL5 kernel.
279 Severity : enhancement
281 Description: lfs quota output is not very convenient for awk/sed-parsing
282 Details : Some positions in lfs quota output table could be empty or
283 non-empty which made it hard to parse it with scripts, now
284 a dash is put instead of space where there is not supposed
289 Description: Update to OFED 1.5.1
290 Details : OFED 1.5.1 will be used in environments that don't use
291 vendor supplied OFED.
295 Description: Remove lnet. prefix from lctl params display
296 Details : Also renamed lnet.memused to lnet_memused to
301 Description: Use conf_param -d to remove permanant settings.
305 Description: Add list_param -R
306 Details : lctl list_param now supports recursive mode.
309 Bugzilla : 21829, 22312
310 Description: Fix llobdstat.
311 Details : 1. Now updating the write_bytes counter
312 2. Port script fixes from 1.8.
316 Description: decrease the usage of memory on clients.
317 Details : 1. On clients, recycle dentries and inodes unused.
318 2. Delete the code related to ll_deathrow(att 6215 in bug 1443). It
323 Description: llobdstat fix and enhancement
324 Details : add a counter to set a limit to how many samples will be returned
325 fix a wildcard in the path to limit to obdfilter stats only
329 Description: lfs check only allowed for root
330 Details : Code cleanups and added a regresion test of non-root use of
335 Description: MDS Oopses in ptlrpc_dispatch_difficult_reply()/__list_add()
336 Details : Fix hr->hr_index overload.
340 Description: Fix FIEMAP functionality.
344 Description: FC11 patchless client support added
348 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and
349 OEL5.4 kernel to 2.6.18-164.11.1.0.1.el5.
353 Description: New acc-sm script and various test-framework enhancments.
357 Description: Previously landed patch was reverted.
361 Description: Update SLES10 kernel to 2.6.16.60-0.42.8 (patchless)
365 Description: Add ldiskfs maxdirsize mount option.
369 Description: Fix mkfs.lustre support for 16TB LUNs.
373 Description: Print mmp_check_interval
374 Details : print mmp_check_interval and make it possible to abort
375 mount operation in case it takes too long.
379 Description: Add cascading_rw to lustre mpi tests.
383 Description: Store SOM data in EA.
384 Details : Also several SOM related cleanups.
388 Description: Add refcount to md_open_data
392 Description: Use refcount on lov_request_set to prevent memory corruption.
396 Description: Statahead: use parent dir's i_mutex for synchronization.
400 Description: loadgen improvements.
404 Description: Enable peer health detection in ksocklnd and ko2iblnd.
408 Description: Control DCACHE_LUSTRE_INVALID flag with MDS_INODE_LOOKCUP lock
412 Description: Fix race on obd_devcie:md_stats
416 Description: Fix SLES 11 client builds.
417 Details : Resolved uninitalized pointer gcc warnings.
421 Description: Allow non-root access for "lfs check".
422 Details : Added a check in obd_class_ioctl() for OBD_IOC_PING_TARGET.
426 Description: Clear imp_force_reconnect correctly in ptlrpc_connect_interpret()
428 Severity : enhancement
430 Description: Update kernel to RHEL5.4 2.6.18-164.6.1.el5 and
431 OEL5 2.6.18-164.6.1.0.1.el5(Both in-kernel OFED enabled).
435 Description: Sanity test_27p fails.
436 Details : OBD_FAIL_OST_ENOSPC has been moved to filter_statfs().
437 Sanity tests 27 were cleaned up and improved.
441 Description: Unify req format on client/servers
442 Details : Use new req_capsule API [almost] everywhere instead of old PTLRPC
443 buffers and swabbers approach..
448 Description: drop join file stuff.
450 Severity : enhancement
452 Description: Add 2.6.27 and SLES11 (patchless client) support
456 Description: Fix in ptlrpc_expire_one_request() to print the signed time
459 Severity : enhancement
461 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
464 Severity : enhancement
466 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
468 Severity : enhancement
470 Description: correctly handle big reply message.
471 Details : send LNet event if reply is bigger than buffer and adjust this buffer
476 Description: Drop unnecessary __GFP_NOMEMALLOC flag from filter_get_page()
478 Severity : enhancement
480 Description: Add support for OEL5.
484 Description: Conf-sanity.sh 50g test - deactivated OST should not cause a panic.
486 Severity : enhancement
488 Description: "weak-modules" support
489 Details : Implement "weak-modules" support which enables kernel modules
490 to be used with any kernel that implements the same kABI. In
491 order to achieve this modules are now installed in
492 /lib/modules/$(uname -r)/updates/kernel on all distributions.
496 Description: client could not reconnect to OST because of an active request.
497 Details : The client now retries to reconnect to the same server, if a
498 connect request failed with EBUSY or -EAGAIN. Test case 35b
499 conf-sanity.sh has been created.
503 Description: Don't allow make backward step on assignin osc next id.
504 Details : race between allocation next id and ll_sync thread can be cause
505 of set wrong osc next id and can be kill valid ost objects.
509 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
513 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
517 Description: Error handling in osc_statfs_interpret() has been improved.
518 Details : Check in osc_statfs_interpret() for EBADR.
522 Description: Do not send statfs() requests to OSTs disabled by administrator.
523 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
525 Severity : enhancement
527 Description: Added sanity-quota test to for lfs(1) to work with numeric uid/gid.
529 Severity : enhancement
531 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
533 Severity : enhancement
535 Description: Enhancement for lfs(1) command to use numeric uid/gid.
539 Description: Improved cleanup procedure for conf-sanity test_46a.
543 Description: Do not start bulk transfer, if client tried to reconnect.
544 Details : Do not start bulk transfer, if client tried to reconnect,
545 but failed due to active rpcs. Check for exp_abort_active_req
546 has been implemented.
550 Description: Prevent inconsistences between linux and lustre mount structures.
551 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
552 Make the sleep interruptible.
554 Severity : enhancement
556 Description: provide server to client comms path
557 Details : server to client communications path via new LDLM_SET_INFO rpc
561 Description: truncate starts GFP_FS allocation under transaction causing deadlock
562 Details : ldiskfs_truncate calls grab_cache_page which may start page
563 allocation under an open transaction. This may lead to
564 calling prune_icache with consequent lustre reentrance.
568 Description: send correctly lsm on open replay
569 Details : MDS is trust to LSM size on replay open, but client can set wrong size
574 Description: enable client interface failover
575 Details : When a child reconnects from another NID, properly update export
576 nid hash position and ldlm reverse import.
578 Severity : enhancement
580 Description: Communicate OST degraded/readonly state via statfs to MDS
581 Details : Flags in the statfs returned from OSTs indicate whether the
582 OST is in a degraded RAID state, or if the filesystem has
583 turned read-only after a filesystem error is detected.
588 Description: don't panic if EPROTO was hit when reading symlink
589 Details : correctly handling request reference in error cases.
591 Severity : enhancement
593 Description: Add LustreNetLink, a kernel-userspace communcation path.
595 Severity : enhancement
597 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
602 Description: access to llog context before init.
603 Details : move handling CATALOGS file at osc layer and forbid access to llog
607 Frequency : rare, only when client is writing to OST on local node
609 Description: Avoid deadlock for local client writes
610 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
611 memory freeing context. This allows OST threads to set the
612 PF_MEMALLOC flag on task structures in order to allocate memory
613 from reserved pools and complete IO.
614 Use GFP_HIGHUSER for OST allocations for non-local client writes,
615 so that the OST threads generate memory pressure and allow
616 inactive pages to be reclaimed.
618 Severity : enhancement
620 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
625 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
626 Details : move ldlm namespace creation in setup phase to avoid grab
627 _lprocfs_lock with cli_sem held.
631 Description: Temporarily disable grant shrink.
632 Details : Disable the feature for debugging.
635 Frequency : only when formatting test filesystems
637 Description: Unable to run several mkfs.lustre on loop devices at the same time
638 Details : mkfs.lustre returns error 256 on the concurrent loop devices
639 formatting. The solution is to proper handle the error.
641 Severity : enhancement
643 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
645 Severity : enhancement
647 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
652 Description: resolve race between obd_disconnect and class_disconnect_exports
653 Details : if obd_disconnect will be called to already disconnected export he
654 forget release one reference and osc module can't unloaded.
656 Severity : enhancement
658 Description: Allow tuning service thread via /proc
659 Details : For each service a new
660 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
661 created that can be used to set min/max thread counts, and get the
662 current number of running threads.
667 Descriptoin: don't return error if have partially created objects for file.
668 Details : lov_update_create_set uses set->set_success as index for created
669 objects, so if some requests will be failed, they will have hole
670 at end of array and qos_shrink_lsm for allocate correct lsm.
672 Severity : enhancement
674 Description: Update OFED support to 1.4.1
679 Description: Reduce small size read RPC
680 Details : Set read-ahead limit for every file and only do read-ahead when
681 available read-ahead pages are bigger than 1M to avoid small size
684 Severity : enhancement
686 Description: move AT tunable parameters for more consistent usage
687 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
689 Severity : enhancement
691 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
692 Details : allow skip disconnected ost for send statfs request and hide error
698 Description: fix racy locking of mballoc block bitmaps causing BUG
699 Details : The locking of the mballoc buddy bitmap and the in-memory
700 block bitmap was using two different spin locks in some
701 cases. This made it possible to incorrectly access the
702 mballoc bitmap while another process was modifying it,
703 causing a sanity assertion to fail. While no on-disk corruption
704 was reported, there was some risk of this happening.
707 Frequency : rare, on llog test 6
709 Descriptoin: don't allow connect to already connected import
710 Details : allowing connect to already connected import is hide connecting problem.
715 Descriptoin: allow kill process which wait statahead result
716 Details : for some reasons 'ls' can stick in waiting result from statahead,
717 in this case need way for kill this process.
719 Severity : enhancement
721 Description: Add state history info file, enhance import info file
722 Details : Track import connection state changes in a new osc/mdc proc file;
723 add overview-type data to the osc/mdc import proc file.
725 Severity : enhancement
727 Description: MDS create should not wait for statfs RPC while holding DLM lock.
729 Severity : enhancement
731 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
736 Descriptoin: OST grant too much space to client even there are not enough space.
737 Details : Client will shrink its grant cache to OST if there are no write
738 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
739 this grant cache if there are already not enough avaible space
740 (left_space < total_clients * 32M).
743 Frequency : start MDS on uncleanly shutdowned MDS device
745 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
746 Details : stay in waiting mds<>ost recovery finished produce random bugs
747 due race between two ll_sync thread for one lov target. send
748 ACTIVATE event only if connect realy finished and import have
752 Frequency : rare, connect and disconnect target at same time
754 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
755 Details : don't call obd_disconnect under lov_lock. this long time
756 operation and can block ptlrpcd which answer to connect request.
761 Descriptoin: don't lose wakeup for imp_recovery_waitq
762 Details : recover_import_no_retry or invalidate_import and import_close can
763 both sleep on imp_recovery_waitq, but we was send only one wakeup
767 Frequency : always with long access acl
769 Descriptoin: mds can't pack reply with long acl.
770 Details : mds don't control size of acl but they limited by reint/getattr
773 Severity : enhancement
775 Description: Update to SLES10 kernel-2.6.16.60-0.33.
777 Severity : enhancement
779 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
782 Frequency : start MDS on uncleanly shutdowned MDS device
784 Descriptoin: aborting recovery hang on MDS
785 Details : don't throttle destroy RPCs for the MDT.
788 Frequency : on remount
790 Description: external journal device not working after the remount
791 Details : clear dev_rdonly flag for external journal devices in
797 Description: shutdown vs evict race
798 Details : client_disconnect_export vs connect request race.
799 if client will evicted at this time - we start invalidate
800 thread without referece to import and import can be freed
804 Frequency : rare, need acl's on inode.
806 Description: client can't handle ost additional correctly
807 Details : if ost was added after client connected to mds client can have
808 hit lnet_try_match_md ... to big messages to wide striped files.
809 in this case need teach client to handle config events about add
810 lov target and update client max ea size at that event.
812 Severity : enhancement
814 Description: Changelogs
815 Details : Changelogs are a lightweight mechanism to track filesystem
816 metadata and namespace changes. The changelog is recorded
817 permanently on the MDTs, and is periodically "consumed" / purged
818 when records are no longer needed.
820 Severity : enhancement
822 Description: compact fld format with extents
823 Details : Store range of seq rather than every seq in FLD. Seq
824 controller update FLD rather than clients. In Case of CMD, mdt0
825 has FLD, all other metadata server act as non persistent proxy
826 for FLD queries and cache fld entries in fld cache.
831 Description: don't skip ost target if they assigned to file
832 Details : Drop slow OSCs if we can, but not for requested start idx.
833 This means "if OSC is slow and it is not the requested
834 start OST, then it can be skipped, otherwise skip it only
835 if it is inactive/recovering/out-of-space.
839 Description: more cleanup in mds_lov
840 Details : not send LOV EA under replay, we can't know about they size at this
841 time. Don't allow client connect to mds before any ost connected,
842 for avoid problems with LOV EA size and returning EIO to client.
844 Severity : enhancement
846 Description: Interoperability at server side (Disk interoperability)
848 Severity : enhancement
850 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
852 Severity : enhancement
854 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
856 Severity : enhancement
858 Description: New client IO stack (CLIO).
860 Severity : enhancement
862 Description: Commit on sharing. Eliminate inter-client dependencies between
863 uncommitted transactions by doing transaction commits.
864 Thereby clients may recovery independently.
867 Frequency : Create a symlink file with a very long name
869 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
870 Details : If there is no extra space in the request for early cancels,
871 ldlm_req_handles_avail() returns 0 instead of a negative value.
873 Severity : enhancement
875 Description: Add /proc entry for import status
876 Details : The mdc, osc, and mgc import directories now have
877 an import directory that contains useful import data for debugging
880 Severity : enhancement
882 Description: Re-disable certain /proc logging
883 Details : Enable and disable client's offset_stats, extents_stats and
884 extents_stats_per_process stats logging on the fly.
887 Frequency : Only on FC kernels 2.6.22+
889 Description: oops in statahead
890 Details : Do not drop reference count for the dentry from VFS when lookup,
891 VFS will do that by itself.
893 Severity : enhancement
895 Description: Generic /proc file permissions
896 Details : Set /Proc file permissions in a more generic way to enable non-
897 root users operate on some /proc files.
901 Description: Hitting mdc_commit_close() ASSERTION
902 Details : Properly handle request reference release in
903 ll_release_openhandle().
907 Description: quota recovery deadlock during mds failover
908 Details : This patch includes att18982, att18236, att18237 in bz14840.
910 1. fix osts hang when mds does failover with quotaon
911 2. prevent watchdog storm when osts threads wait for the
916 Frequency : only patchless client
917 Description: add workaround for race between add/remove dentry from hash
919 Severity : enhancement
921 Description: Allow OST glimpses to return PW locks
925 Description: LBUG when llog conf file is full
926 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
930 Description: Prevent import from entering FULL state when server in recovery
934 Description: service mount cannot take device name with ":"
935 Details : Only when device name contains ":/" will mount treat it as
941 Description: replace ptlrpcd with the statahead thread to interpret the async
942 statahead RPC callback
946 Frequency : on recovery
947 Description: I/O failures after umount during fail back
948 Details : if client reconnected to restarted server we need join to recovery
949 instead of find server handler is changed and process self
950 eviction with cancel all locks.
952 Severity : enhancement
954 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
956 Severity : enhancement
958 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
960 Severity : enhancement
962 Description: Upcall on Lustre log has been dumped
963 Details : Allow for a user mode script to be called once a Lustre log has
964 been dumped. It passes the filename of the dumped log to the
965 script, the location of the script can be specified via
966 /proc/sys/lnet/debug_log_upcall.
971 Description: avoid idr_remove called for id which is not allocated.
972 Details : Move assigment s_dev for clustered nfs to end of initialization,
973 to avoid problem with error handling.
978 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
979 Details : When connection is reused this not moved from CONN_UNUSED_HASH
980 into CONN_USED_HASH and this prodice warning when put connection
981 again in unused hash.
984 Severity : enhancement
986 Description: File striping can now be set to use an arbitrary pool of OSTs.
988 Severity : enhancement
990 Description: Export bytes_read/bytes_write count on OSC/OST.
994 Description: Early reply size mismatch, MGC loses connection
995 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
996 the connect flags are properly negotiated.
1001 Description: quotas are not honored with O_DIRECT
1002 Details : all writes with the flag O_DIRECT will use grants which leads to
1003 this problem. Now using OBD_BRW_SYNC to guard this.
1007 Description: add quota statistics
1008 Details : 1. sort out quota proc entries and proc code.
1009 2. add quota statistics
1011 Severity : enhancement
1013 Description: enable quota support for HEAD.
1017 Description: Properly propagate oinfo flags from lov to osc for statfs
1018 Details : restore missing copy oi_flags to lov requests.
1020 Severity : enhancement
1022 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
1023 : l_getgroups(8), lst(8), routerstat(8)
1024 Details : included man pages for llobdstat(8), llstat(8),
1025 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
1027 Severity : enhancement
1029 Description: configure's --enable-quota should check the
1030 : kernel .config for CONFIG_QUOTA
1031 Details : configure is terminated if --enable-quota is passed but
1032 : no quota support is in kernel
1036 Description: Remove portals compatibility
1037 Details : Remove portals compatibility, not interoperable with releases
1042 Description: Resolve device initialization race
1043 Details : Prevent proc handler from accessing devices added to the
1044 obd_devs array but yet be intialized.
1046 Severity : enhancement
1048 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
1050 Severity : enhancement
1052 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
1057 Description: Using wrong pointer in osc_brw_prep_request
1058 Details : Access to array[-1] can produce panic if kernel compiled with
1059 CONFIG_PAGE_ALLOC enabled
1063 Description: Client runs out of low memory
1064 Details : Consider only lowmem when counting initial number of llap pages
1068 Description: *optional* service tags registration
1069 Details : if the "service tags" package is installed on a Lustre node
1070 When the filesystem is mounted, a local-node service tag will
1071 be created. See http://inventory.sun.com/ for more information
1072 about the Service Tags asset management system.
1076 Description: Kernel BUG tries to release flock
1077 Details : Lustre does not destroy flock lock before last reference goes
1078 away. So always drop flock locks when client is evicted and
1079 perform unlock regardless of successfulness of speaking to MDS.
1083 Description: add refcount protection for osc callbacks, avoid panic on shutdown
1087 Description: sanity test 65a fails if stripecount of -1 is set
1088 Details : handle -1 striping on filesystem in ll_dirstripe_verify
1093 Description: ASSERTION(CheckWriteback(page,cmd)) failed
1094 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
1097 Severity : enhancement
1099 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
1103 Description: do not process already freed flock
1104 Details : flock can possibly be freed by another thread before it reaches
1105 to ldlm_flock_completion_ast.
1109 Description: LBUG during stress test
1110 Details : Need properly lock accesses the flock deadlock detection list.
1114 Description: oops in page fault handler
1115 Details : kernel page fault handler can return two special 'pages' in error
1116 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
1120 Description: timeout with invalidate import.
1121 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
1122 should be handled by ptlrpcd. This produce long age waiting and
1123 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
1125 Severity : enhancement
1127 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
1131 Description: enable MGS and MDT services start separately
1132 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
1133 a co-located MGS without starting the MGS, which is a complement
1134 to 'nosvc' mount option.
1138 Frequency : after recovery
1139 Description: precreate to many object's after del orphan.
1140 Details : del orphan st in oscc last_id == next_id and this triger growing
1141 count of precreated objects. Set flag LOW to skip increase count
1142 of precreated objects.
1146 Frequency : rare, on clear nid stats
1147 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
1148 Details : when clean nid stats sometimes try destroy live entry,
1149 and this produce panic in free.
1153 Description: Stack overflow during MDS log replay
1154 ease stack pressure by using a thread dealing llog_process.
1158 Description: wait until IO finished before start new when do lock cancel.
1159 Details : VM protocol want old IO finished before start new, in this case
1160 need wait until PG_writeback is cleared until check dirty flag and
1161 call writepages in lock cancel callback.
1163 Severity : enhancement
1165 Description: using special macro for print time and cleanup in includes.
1169 Description: mds_mfd_close() ASSERTION(rc == 0)
1170 Details : In mds_mfd_close(), we need protect inode's writecount change
1171 within its orphan write semaphore to prevent possible races.
1175 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
1179 Frequency : rare, on shutdown ost
1180 Description: don't hit live lock with umount ost.
1181 Details : shrink_dcache_parent can be in long loop with destroy dentries,
1182 use shrink_dcache_sb instead.
1186 Description: don't panic with use echo client
1187 Details : echo client pass NULL as client nid pointer and this produce null
1188 pointer dereference.
1192 Description: fix build on ppc32
1193 Details : compile code with -m64 flag produce wrong object file for ppc32.
1197 Description: add message levels for liblustreapi
1201 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
1202 Details : Move the 'good_osts' check before the 'total_bavail' check. This
1203 will result in an -EAGAIN and in the exit call path we call
1204 alloc_rr() which will with increasing aggressiveness attempt to
1205 aquire precreated objects on the minimum number of required OSCs.
1209 Description: Use old size assignment to avoid deadlock
1210 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
1211 the scheduling while holding a spinlock. We do not need locking
1212 for size in ll_update_inode() because size is only updated from
1213 the MDS for directories or files without objects, so there is no
1214 other place to do the update, and concurrent access to such inodes
1215 are protected by the inode lock.
1219 Description: resolve "_IOWR redefined" build error on SLES10
1223 Description: dump the memory debugging after all modules are unloaded to
1224 suppress false negative in conf_sanity test 39
1226 Severity : enhancement
1228 Description: build kernel-ib packages for OFED 1.3 in our release cycle
1233 Description: fix SLES kernel versioning
1234 Details : the kernel version for our SLES 10 kernel did not include a "-"
1235 before the "smp" at the end. while this was not a problem in
1236 general, it did mean that software trying to use the kernel
1237 version to try to detect a vendor specific kernel would fail.
1238 this was most evident by the OFED build scripts.
1242 Description: Don't update lov_desc members until making sure they are valid
1243 Details : When updating lov_desc members via proc fs, need fix their
1244 validities before doing the real update.
1248 Description: don't put request into delay list while invalidate in flight.
1249 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
1250 import in flight. this produce timeout for invalidate and sometimes
1251 can cause stale data.
1255 Frequency : on ppc only
1256 Description: not convert ost objects for directory because it's not exist.
1257 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
1261 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
1263 Severity : enhancement
1265 Description: Update to support 2.6.22.14 vanilla kernel.
1269 Frequency : rare, on recovery
1270 Description: read procfs can produce deadlock in some situation
1271 Details : Holding lprocfs lock which send rpc can produce block for destroy
1272 obd objects and this also block reconnect with -EALREADY. This
1273 isn't fix all lprocfs bugs - but make it rare.
1275 Severity : enhancement
1277 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
1280 Frequency : frequent on X2 node
1282 Description: mdc_set_open_replay_data LBUG
1283 Details : Set replay data for requests that are eligible for replay.
1287 Description: lustre_mgs: operation 101 on unconnected MGS
1288 Details : When MGC is disconnected from MGS long enough, MGS will evict the
1289 MGC, and late on MGC cannot successfully connect to MGS and a lot
1290 of the error messages complaining that MGS is not connected.
1294 Frequency : on network error
1295 Description: panic with double free request if network error
1296 Details : mdc_finish_enqueue is finish request if any network error ocuring,
1297 but it's true only for synchronus enqueue, for async enqueue
1298 (via ptlrpcd) this incorrect and ptlrpcd want finish request
1301 Severity : enhancement
1303 Description: client-side metadata stat-ahead during readdir(directory readahead)
1304 Details : perform client-side metadata stat-ahead when the client detects
1305 readdir and sequential stat of dir entries therein
1308 Frequency : on start mds
1310 Description: Implement get_info(last_id) in obdfilter.
1313 Frequency : occasional
1315 Description: Correctly check stale fid, not start epoch if ost not support SOM
1316 Details : open with flag O_CREATE need set old fid in op_fid3 because
1317 op_fid2 was overwritten with new generated fid, but mds can answer
1318 with one of these two fids and both is not stale. Setattr
1319 incorrectly started an epoch and assume will be called
1320 done_writing, but without SOM done_writing ever being called.
1323 Frequency : rare, depends on device drivers and load
1325 Description: MDS or OSS nodes crash due to stack overflow
1326 Details : Code changes in 1.8.0 increased the stack usage of some functions.
1327 In some cases, in conjunction with device drivers that use a lot
1328 of stack the MDS (or possibly OSS) service threads could overflow
1329 the stack. One change which was identified to consume additional
1330 stack has been reworked to avoid the extra stack usage.
1333 Frequency : occasional
1335 Description: Do not fail import if osc_interpret_create gets -EAGAIN
1336 Details : If osc_interpret_create got -EAGAIN it immediately exits and
1337 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
1338 oscc_has_objects and see OSC has no objests and call
1339 oscc_internal_create to resend create request.
1341 Severity : enhancement
1343 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
1345 Severity : enhancement
1347 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
1350 Frequency : very rare
1352 Description: Fix replay if there is an un-replied request and open
1353 Details : In some cases, older replay request will revert the
1354 mcd->mcd_last_xid on MDS which is used to record the client's
1355 latest sent request.
1357 Severity : enhancement
1359 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
1361 Severity : enhancement
1363 Description: Add rhel5 support to HEAD.
1365 Serverity : enhancement
1367 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
1372 Description: Don't allow skipping OSTs if index has been specified.
1373 Details : Don't allow skipping OSTs if index has been specified, make
1374 locking in internal create lots better.
1378 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
1379 Details : access to bitfield in structure is always rounded to long
1380 and this produce problem with not atomic change any bit.
1384 Description: Lustre make rpms failed.
1385 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
1386 builds Lustre rpms from tarball.
1389 Frequency : rare on shutdown ost
1391 Description: If llog cancel was not send before clean_exports phase, this can
1392 produce deadlock in llog code.
1393 Details : If llog thread has last reference to obd and call class_import_put
1394 this produce deadlock because llog_cleanup_commit_master wait when
1395 last llog_commit_thread exited, but this never success because was
1396 called from llog_commit_thread.
1400 Description: allow userland application know is lost one of stripes.
1401 Details : fill lvb_blocks with error code on ost and return it to
1402 application if error flag found.
1406 Description: NULL lov_tgts causing MDS oops
1407 Details : more safe checks for NULL lov_tgts for avoid oops.
1409 Severity : enhancement
1411 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
1415 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
1416 Details : If an OST is down the MDS will hang indefinitely in
1417 obd_statfs_rqset() waiting for the statfs data. While for
1418 MDS QOS usage of statfs, it should not stuck in waiting.
1420 Severity : enhancement
1422 Description: remote_acl support
1423 Details : Support ACL-based permission check for remote user.
1424 Support setfacl/getfacl for remote user with the utils
1425 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
1426 the system "{s,g}etfacl" utils.
1428 Severity : enhancement
1430 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
1432 Severity : enhancement
1434 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
1438 Description: make mgs_setparam() handle fsname containing dash
1439 Details : fsname containing a dash does not work with lctl conf_param
1441 Severity : enhancement
1443 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
1445 Severity : enhancement
1447 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
1451 Frequency : rare, at shutdown
1452 Description: access already free / zero obd_namespace.
1453 Details : if client_disconnect_export was called without force flag set,
1454 and exist connect request in flight, this can produce access to
1455 NULL pointer (or already free pointer) when connect_interpret
1456 store ocd flags in obd_namespace.
1460 Frequency : only at startup
1461 Description: not alloc memory with spinlock held.
1462 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
1463 if any spinlock held.
1465 Severity : enhancement
1467 Description: make lustre randomly fail allocating memory
1468 Details : Make lustre randomly failed allocating memory for testing purpose.
1470 Severity : enhancement
1472 Description: lost problems with lov objid file
1473 Details : Fixes some scability and access to not inited memory problems
1474 in work with lov objdid file.
1479 Description: lfs find does not continue on file error
1480 Details : Continue other files processing when a file/dir is absent.
1484 Description: Inconsistent usage of lustre_pack_reply()
1485 Details : Standardize the usage of lustre_pack_reply() such that it
1486 always generate a CERROR on failure.
1489 Frequency : occasional
1491 Description: MDS hang or stay in waiting lock
1492 Details : If client receive lock with CBPENDING flag ldlm need send lock
1493 cancel as separate rpc, to avoid situation when cancel request
1494 can't processed due all i/o threads stay in wait lock.
1498 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1502 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1506 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1510 Description: Update to RHEL4 latest kernel.
1512 Severity : enhancement
1514 Description: Build SLES10 patchless client fails
1515 Details : The configure was broken by run ./configure with
1516 --with-linux-obj=.... argument for patchless client. When the
1517 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1518 can't search header adequately. Use absolute path such as
1519 -I($LINUX)/include instead.
1523 Description: interrupt oig_wait produce painc on resend.
1524 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1525 private set, and this produce situation when rq_ptlrpcd_data not
1526 copyed to new allocated request and triggered LBUG on assert
1527 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1528 ptlrpcd set if request is changed and can be safety changed to use
1533 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1534 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1535 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1536 these macros and replace them with their difinitions instead.
1540 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1541 Details : Update which_patch & target file for SLES10 latest kernel.
1544 Description: Remove client patches from SLES 10 kernel.
1545 Details : This causes SLES 10 clients to behave as patchless clients
1546 even on a Lustre-patched (server) kernel.
1548 Severity : enhancement
1550 Description: self-adjustable client's lru lists
1551 Details : use adaptive algorithm for managing client cached locks lru
1552 lists according to current server load, other client's work
1553 pattern, memory activities, etc. Both, server and client
1554 side namespaces provide number of proc tunables for controlling
1557 Severity : enhancement
1559 Description: light-weight GSS support
1560 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
1561 and reduce performance overhead.
1563 Severity : enhancement
1565 Description: Linux keyring support
1566 Details : Support using service of Linux keyring for Lustre GSS internal
1567 context refresh/cache mechanism.
1571 Description: Fix errors in lfs documentation
1572 Details : Fixes man pages
1576 Description: don't use GFP_* in generic Lustre code.
1577 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
1581 Description: obdclass is limited by single OBD_ALLOC(idarray)
1582 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
1586 Description: warnings with build patchless client with vanila 2.6.19 and up
1587 Details : change old ctl_table style and replace ctl_table/ctl_table_header
1588 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
1592 Frequency : Only for SLES
1593 Description: sanity_quota fail test_1
1594 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
1595 file, which makes TSTID[2] inunique.
1599 Frequency : Only for SLES9
1600 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
1601 Details : Update target/ChangeLog/which_patch .
1605 Frequency : Only for test_10 in sanity-quota.sh
1606 Description: a bug in quota test code
1607 Details : checking a test flag in a wrong place
1611 Frequency : Only for test_13 in sanity-quota.sh
1612 Description: a bug in quota test script
1613 Details : shouldn't check overfull allocated quota
1617 Description: jbd statistics
1618 Details : Port older jbd statistics patch for sles10
1620 Severity : enhancement
1622 Description: Add printing inode info into message about error in writepage.
1627 Description: reply_lock_interpret crash due to race with it and lock cancel.
1628 Details : Do not replay locks that are being cancelled. Do not reference
1629 locks by their address during replay, just by their handle.
1633 Frequency : When flocks are used.
1634 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
1635 Details : Flock locks might destroy just granted lock if it could be merged
1636 with another existing flock, this is done in completion handler,
1637 so teach ldlm_cli_enquque_fini that this is a valid case for
1643 Description: Oops in read and write path when failing to allocate lock.
1644 Details : Check if lock allocation failed and return error back.
1648 Description: lstripe command fails for valid OST index
1649 Details : The stripe offset is compared to lov->desc.ld_tgt_count
1650 instead of lov->desc.ld_active_tgt_count.
1654 Description: sanity.sh failed test 103
1655 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
1660 Description: df doesn't work properly if diskfs blocksize != 4K
1661 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
1665 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
1666 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
1668 Severity : enhancement
1670 Description: Adaptive timeouts
1671 Details : RPC timeouts adapt to changing server load and network
1672 conditions to reduce resend attempts and improve recovery time.
1676 Description: llapi_file_create() does not allow some changes
1677 Details : add llapi_file_open() that allows specifying the mode and
1678 open flags, and also returns an open file handle.
1682 Description: merge and cleanup kernel patches.
1683 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
1687 Description: Add journal checksum support.(Kernel part)
1688 Details : The journal checksum feature adds two new flags i.e
1689 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
1690 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
1691 indicates that the commit block contains the checksum for
1692 the blocks described by the descriptor blocks. Now commit
1693 record can be sent to disk without waiting for descriptor
1694 blocks to be written to disk. This behavior is controlled
1695 using JBD2_FEATURE_ASYNC_COMMIT flag.
1699 Description: OSS needs mutliple precreate threads
1700 Details : Add ability to start more than one create thread per OSS.
1704 Description: Sanity.sh test_65k failed: setstripe should have succeeded
1705 Details : execute lfs setstripe on client
1709 Description: mds_obd_create error creating tmp object
1710 Details : When the user sets quota on root, llog will be affected and can't
1711 create files and write files.
1714 Frequency : Always on ia64 patchless client, and possibly others.
1716 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
1717 Details : This allows the patchless client to be loaded on architectures
1718 without this export.
1722 Description: RedHat Update kernel for RHEL5
1723 Details : Add the kernel config file for RHEL5.
1727 Description: RedHat Update kernel for RHEL5
1728 Details : Modify the kernel config file more closer RHEL5.
1732 Description: Build failure against Centos5 (RHEL5)
1733 Details : Define PAGE_SIZE when it isn't present.
1737 Description: test_7 on sanity-quota.sh can't be used on separate machines
1738 Details : add facet to handle it
1742 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
1743 Details : with NFS, the anon dentry's parent was set to itself in
1744 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
1745 corresponding dentry other than use rec->ur_name.
1747 Severity : enhancement
1749 Description: lfs setstripe enhancement
1750 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
1754 Description: enable data checksumming by default
1755 Details : enable checksum by default, allow --disable-checksum
1756 configure option and "-o nochecksum" mount option. Checksums
1757 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
1761 Description: lustre support for RHEL5
1762 Details : Add support for RHEL5.
1766 Description: Client eviction due to failover config
1767 Details : after a connection loss, the lustre client should attempt to
1768 reconnect to the last active server first before trying the
1769 other potential connections.
1773 Description: when mds and osts use different quota unit(32bit and 64bit),
1774 quota will be released repeatly.
1775 Details : void sending multiple quota reqs to mds, which will keep the
1776 status between the reqs.
1780 Description: osts not allocated evenly to files
1781 Details : change the condition to increase offset_idx
1784 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
1786 Description: Data corruption for OSTs that are formatted larger than 2TB
1788 Details : When generating the bio request for lustre file writes the
1789 sector number would overflow a temporary variable before being
1790 used for the IO. The data reads correctly from Lustre (which
1791 will overflow in a similar manner) but other file data or
1792 filesystem metadata may be corrupted in some cases.
1796 Description: Tune the kernel for good SCSI performance.
1797 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
1798 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
1803 Description: rewrite ext2-derived code in obdclass/uuid.c
1804 Details : rewrite inherited code (uuid parsing code from ext2 utils)
1805 from scratch preserving functionality.
1809 Description: "lfs find -obd UUID" prints directories
1810 Details : "lfs find -obd UUID" will return all directory names instead
1811 of just file names. It is incorrect because the directories
1812 do not reside on the OSTs.
1816 Description: use i_size_read and i_size_write in 2.6 port
1817 Details : replace inode->i_size access with i_size_read/write()
1820 Frequency : when removing large files
1822 Description: scheduling issue during removal of large Lustre files
1823 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
1824 It causes scheduling issues when removing large files (17TB in the
1828 Frequency : only with liblustre clients on XT3
1830 Description: evictions taking too long
1831 Details : allow llrd to evict clients directly on OSTs
1834 Frequency : only on ppc
1836 Description: /proc/fs/lustre/devices broken on ppc
1837 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
1838 We should make sure the type of 'index' is loff_t and then cast
1839 explicitly as needed below. Do not assign an explicitly cast
1843 Frequency : only for rhel5
1845 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
1846 Details : Modify the target file & which_kernel.
1848 Severity : enhancement
1850 Description: omit set fsid for export NFS
1851 Details : fix set/restore device id for avoid EMFILE error and mark lustre
1852 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
1856 Frequency : Always, for kernels after 2.6.16
1857 Description: Fix warning idr_remove called for id=.. which is not allocated.
1858 Details : Last kernels save old s_dev before kill super and not allow
1859 to restore from callback - restore it before call kill_anon_super.
1863 Description: buffer overruns could theoretically occur
1864 Details : llapi_semantic_traverse() modifies the "path" argument by
1865 appending values to the end of the origin string, and a
1866 overrun may occur. Adding buffer overrun check in liblustreapi.
1870 Description: Fix error on 'ls .' at the top of the Lustre mount.
1871 Details : Don't revalidate dentry if it is a root dentry.
1875 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
1876 Details : Modify vm-tunables-rhel4.patch.
1880 Description: Kernel config for 2.6.18-vanilla.
1881 Details : Modify targets/2.6-vanilla.target.in.
1882 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
1883 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
1884 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
1885 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
1889 Description: improve handling recoverable errors
1890 Details : if request processig with error which can be recoverable on server
1891 request should be resend, otherwise page released from cache and
1896 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
1897 Details : Modify target file & which_patch.
1898 A flaw was found in the IA32 system call emulation provided
1899 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
1900 value could be stored in the %RAX register, which could trigger an
1901 out-of-bounds system call table access. An untrusted local user
1902 could exploit this flaw to run code in the kernel
1903 (ie a root privilege escalation). (CVE-2007-4573).
1907 Description: O_DIRECT bypasses client statistics.
1908 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
1909 read_ahead_stats were updated. copied the stats section from
1910 osc_send_oap_rpc() into async_internal().
1914 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
1918 Description: change order of libsysio includes
1919 Details : '#include sysio.h' should always come before '#include xtio.h'
1923 Description: handle "serious error: objid * already exists" more gracefully
1924 Details : If LAST_ID value on disk is smaller than the objects existing in
1925 the O/0/d* directories, it indicates disk corruption and causes an
1926 LBUG(). If the object is 0-length, then we should use the existing
1927 object. This will help to avoid a full fsck in most cases.
1929 Severity : enhancement
1931 Description: adapt the lustre_config script to support the upgrade case
1932 Details : Add "-u" option for lustre_config script to support upgrading 1.4
1933 server targets to 1.6 in parallel.
1937 Description: To avoid grant space > avaible space when the disk is almost
1938 full. Without this patch you might see the error "grant XXXX >
1939 available" or some LBUG about grant, when the disk is almost
1941 Details : In filter_check_grant, for non_grant cache write, we should
1942 check the left space by if (*left > ungranted + bytes), instead
1943 of (*left > ungranted), because only we are sure the left space
1944 is enough for another "bytes", then the ungrant space should be
1945 increase. In client, we should update cl_avail_grant only there
1946 is OBD_MD_FLGRANT in the reply.
1950 Description: Update RHEL 4 kernel to fix local root privilege escalation.
1951 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
1952 described in CVE-2007-4573. This problem could allow untrusted
1953 local users to gain root access.
1956 Frequency : when using O_DIRECT and quotas
1958 Description: Incorrect file ownership on O_DIRECT output files
1959 Details : block usage reported by 'lfs quota' does not take into account
1960 files that have been written with O_DIRECT.
1965 Description: touch file failed when fs is not full
1966 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
1967 otherwise we can get ENOSP while fs is not full.
1971 Description: parallel lock callbacks
1972 Details : Instead of sending blocking and completion callbacks as separated
1973 requests, adding them to a set and sending in parallel.
1976 Frequency : only for Cray XT3
1977 Bugzilla : 12829/13455
1978 Description: Changing primary group doesn't change the group lustre assigns to
1980 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
1981 overridden with the primary group provided by the group upcall,
1982 whereas the supplied fsgid can be trusted if it is in the list of
1983 supplementary groups returned by the group upcall.
1985 Severity : enhancement
1987 Description: Allow masking D_WARNING, D_ERROR messages from console
1988 Details : Console messages can now be disabled via lnet.printk.
1992 Description: User code with malformed file open parameter crashes client node
1993 Details : Before packing join_file req, all the related reference should be
1994 checked carefully in case some malformed flags cause fake
1995 join_file req on client.
1999 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
2000 Details : ldlm_completion_ast() assumes that a lock is granted when the req
2001 mode is equal to the granted mode. However, it should also check
2002 that LDLM_FL_CP_REQD is not set.
2006 Description: Heavy nfs access might result in deadlocks
2007 Details : After ELC code landed, it is now improper to enqueue any mds
2008 locks under och_sem, because enqueue might want to decide to
2009 cancel open locks for same inode we are holding och_sem for.
2013 Description: Client eviction while running blogbench
2014 Details : A lot of unlink operations with concurrent I/O can lead to a
2015 deadlock causing evictions. To address the problem, the number of
2016 oustanding OST_DESTROY requests is now throttled to
2017 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
2018 callbacks are processed in priority.
2022 Description: enable ACLs on MDS by default
2023 Details : ACLs must be enabled on MDS by default.
2026 Frequency : PPC/PPC64 only
2028 Description: conflicts between asm-ppc64/types.h and lustre_types.h
2029 Details : fix duplicated definitions between asm-ppc64/types.h and
2030 lustre_types.h on PPC.
2033 Frequency : PPC/PPC64 only
2035 Description: asm-ppc/segment.h does not exist
2036 Details : fix compile issue on PPC.
2040 Description: better handle error messages in extents code
2043 Frequency : RHEL4 only
2045 Description: mkfs is very slow on IA64/RHEL4
2046 Details : A performance regression has been discovered in the MPT Fusion
2047 driver between versions 3.02.73rh and 3.02.99.00rh. As a
2048 consequence, we have downgraded the MPT Fusion driver in the RHEL4
2049 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
2051 Severity : enhancement
2053 Description: SNMP support enhancement
2054 Details : Adding total number of sampled request for an MDS node in snmp
2057 Severity : enhancement
2059 Description: Optimize ldlm waiting list processing for PR extent locks
2060 Details : When processing waiting list for read extent lock and meeting read
2061 lock that is same or wider to it that is not contended, skip
2062 processing rest of the list and immediatelly return current
2063 status of conflictness, since we are guaranteed there are no
2064 conflicting locks in the rest of the list.
2068 Description: Time out and refuse to reconnect
2069 Details : When the failover node is the primary node, it is possible
2070 to have two identical connections in imp_conn_list. We must
2071 compare not conn's pointers but NIDs, otherwise we can defeat
2072 connection throttling.
2076 Description: port llog fixes from b1_6 into HEAD
2077 Details : Port llog reference couting and some llog cleanups from b1_6
2078 (bug 10800) into HEAD, for protect from panic and access to already
2079 free llog structures.
2083 Description: Detect stride IO mode in read-ahead
2084 Details : When a client does stride read, read-ahead should detect that and
2085 read-ahead pages according to the detected stride pattern.
2089 Description: data checksumming impacts single node performance
2090 Details : add support for several checksum algorithm. Currently, only CRC32
2091 and Adler-32 are supported. The checksum type can be changed on
2092 the fly via /proc/fs/lustre/osc/*/checksum_type.
2096 Description: use adler32 for page checksums
2097 Details : when available, use the Adler-32 algorithm instead of CRC32 for
2102 Description: build for x2 fails
2103 Details : fix compile issue on Cray systems.
2107 Description: Properly match for duplicate locks
2108 Details : Due to different lock order from skiplists code, we need to
2109 traverse entire list for now
2112 Frequency : only on PPC/SLES10
2114 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
2115 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
2116 be defined. Add a hack in mkfs_lustre.c to work around this compile
2121 Description: LASSERT on MDS when client holding flock lock dies
2122 Details : ldlm pool logic depends on number of granted locks equal to
2123 number of released locks which is not true for flock locks, so
2124 just exclude such locks from consideration.
2128 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
2129 Details : Use fsfilt_sync() for both the whole filesystem sync and
2130 individual file sync to eliminate dangerous inode locking
2131 with I_LOCK that can lead to a deadlock.
2135 Description: performance in 1.6.3
2136 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
2137 sure that 1MB requests are not fragmented by the block layer.
2139 Severity : enhancement
2141 Description: organize the server-side client stats on per-nid basis
2142 Details : Change the structure of stats under obdfilter and mds to
2150 The "uuid"s file would list the uuids of _active_ exports.
2151 And the clear entry is to clear all stats and stale nids.
2153 Severity : enhancement
2155 Description: eliminate client locks in face of contention
2156 Details : file contention detection and lockless i/o implementation
2157 for contended files.
2161 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
2164 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
2166 Description: MDS may lose file striping (and hence file data) in some cases
2167 Details : If there are additional extended attributes stored on the MDS,
2168 in particular ACLs, SELinux, or user attributes (if user_xattr
2169 is specified for the client mount options) then there is a risk
2170 of attribute loss. Additionally, the Lustre file striping
2171 needs to be larger than default (e.g. striped over all OSTs),
2172 and an additional attribute must be stored initially in the
2173 inode and then increase in size enough to be moved to the
2174 external attribute block (e.g. ACL growing in size) for file
2179 Description: skiplist implementation simplification
2180 Details : skiplists are used to group compatible locks on granted list
2181 that was implemented as tracking first and last lock of each
2182 lock group the patch changes that to using doubly linked lists
2186 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
2187 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
2188 because LASSERT on both the data supplied by a client, and the
2189 data on disk is dangerous and incorrect.
2191 Severity : enhancement
2193 Description: Slow truncate/writes to huge files at high offsets.
2194 Details : Directly associate cached pages to lock that protect those pages,
2195 this allows us to quickly find what pages to write and remove
2196 once lock callback is received.
2200 Description: more ldlm soft lockups
2201 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
2202 starve other threads from the resource lock for a long time in
2203 case of long waiting queue, so change the debug level from
2204 D_OTHER to the less frequently used D_INFO.
2206 Severity : enhancement
2208 Description: add -gid, -group, -uid, -user options to lfs find
2212 Description: Hung threads in invalidate_inode_pages2_range
2213 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
2214 once one is completed. As a result, some RPCs are stuck in the
2215 queue and are never sent.
2219 Description: filter threads hungs on waiting journal commit
2220 Details : Cleanup filter group llog code, then only filter group llog will
2221 be only created in the MDS/OST syncing process.
2225 Description: Procfs and llog threads access destoryed import sometimes.
2226 Details : Sync the import destoryed process with procfs and llog threads by
2227 the import refcount and semaphore.
2229 Severity : enhancement
2231 Description: openlock cache of b1_6 port to HEAD
2236 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
2237 Details : If insertion of an extent fails, then discard the inode
2238 preallocation and free data blocks else it can lead to duplicate
2243 Description: don't always update ctime in ext3_xattr_set_handle()
2244 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
2245 In some cases the ctime should not be updated, for example for
2246 2.0->1.8 compatibility it is necessary to delete an xattr and it
2247 should not update the ctime.
2251 Bugzilla : 15713/16362
2252 Description: Assertion in iopen_connect_dentry in 1.6.3
2253 Details : looking up an inode via iopen with the wrong generation number can
2254 populate the dcache with a disconneced dentry while the inode
2255 number is in the process of being reallocated. This causes an
2256 assertion failure in iopen since the inode's dentry list contains
2257 both a connected and disconnected dentry.
2261 Description: assertion failure in ldlm_handle2lock()
2262 Details : fix a race between class_handle_unhash() and class_handle2object()
2263 introduced in lustre 1.6.5 by bug 13622.
2268 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
2269 Details : remove the limit of 256 scsi disks in the sd_iostat patch
2274 Description: oops in sd_iostats_seq_show()
2275 Details : unloading/reloading the scsi low level driver triggers a kernel
2276 bug when trying to access the sd iostat file.
2281 Description: Kernel panics during QLogic driver reload
2282 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
2283 patch, causing memory corruption.
2288 Description: journal_dev option does not work in b1_6
2289 Details : pass mount option during pre-mount.
2291 Severity : enhancement
2293 Description: Add a FIEMAP(FIle Extent MAP) ioctl
2294 Details : FIEMAP ioctl will allow an application to efficiently fetch the
2295 extent information of a file. It can be used to map logical blocks
2296 in a file to physical blocks in the block device.
2300 Description: LDLM soft lockups - improvement
2301 Details : It is be possible to send the lock handle along with each read
2302 or write request because the client is already doing a lock match
2303 itself so there isn't any reason the OST should have to re-do that
2307 Frequency : only on Cray X2
2309 Description: X2 build failures
2310 Details : fix build failures on Cray X2.
2314 Description: xid & resent requests
2315 Details : Initialize RPC XID from clock at startup (randomly if clock is
2318 Severity : enhancement
2320 Description: Add lustre_start utility to start or stop multiple Lustre servers
2325 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
2326 Details : In case of memory pressure, list_del() can be called twice on
2327 req->rq_history_list, causing a kernel oops.
2331 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
2332 Details : fix stack overflow in the distributed lock manager by defering
2333 export eviction after a failed AST to the elt thread instead of
2334 handling it in the dlm interpret routine.
2338 Description: Convert some comments to new format.
2339 Details : Update documenting comments to match doxygen conventions.
2343 Description: Grammar fixes.
2344 Details : A couple of trivial spelling fixes.
2348 Description: OSD_COUNTERS-mandatory
2349 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
2350 to simplify the code.
2354 Description: simplify cmm_device freeing logic.
2355 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
2359 Description: Add lockdep support to dt_object_operations locking interface.
2360 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
2361 parameter indicating lock ordering. Update mdd code to use new
2366 Description: Introduce failloc constants for lockless IO tests.
2367 Details : Add two new failloc constants to test lockless IO. Only one of
2368 them in implemented---another is checked in yet to be landed
2373 Description: Add lockdep support for inode mutex.
2374 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
2375 in the situations where Lustre has to lock more than one inode
2380 Description: Add optional invariants checking support.
2381 Details : Add new LINVRNT() macro, optional on new --enable-invariants
2382 configure switch. This macro is to be used for consistency and
2383 sanity checks that are too expensive to be left in `production'
2388 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
2389 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
2394 Description: Add lockdep support for ldlm_lock and ldlm_resource.
2395 Details : Use spin_lock_nested() in (the only) situation where more than
2396 one ldlm_lock is locked simultaneously. Also, fix possible
2397 dead-lock in ldlm_lock_change_resource() by enforcing particular
2402 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
2403 Details : Instead of specifying each ldlm_lock call-back through separate
2404 parameter, wrap them into struct ldlm_callback_suite.
2408 Description: Kill join_lru obd method and its callers.
2409 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
2410 in memory---a requirement implemented through ->o_join_lru() obd
2411 method in HEAD. Remove this method and its users.
2415 Description: Add asynchronous ldlm ENQUEUE completion handler.
2416 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
2417 case that stock ldlm_completion_ast() cannot handle as it waits
2418 until lock is granted. Introduce new ldlm_completion_ast_async()
2419 for this. Also comment ldlm_completion_ast().
2423 Description: ldlm_error <-> errno conversion.
2424 Details : Add functions to map (rather arbitrary) between LDLM error codes
2425 and standard errno values. CLIO needs this to prevent LDLM specific
2426 constants from escaping ldlm and osc.
2430 Description: Kill unused ldlm_handle2lock_ns() function.
2431 Details : Kill unused ldlm_handle2lock_ns() function.
2435 Description: Add lu_ref support to ldlm_lock
2436 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
2437 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
2438 and ldlm_resource. LDLM interface has to be changed, because code that
2439 releases a reference on a lock, has to "know" what reference this is.
2440 In the most frequent case
2442 lock = ldlm_handle2lock(handle);
2444 LDLM_LOCK_PUT(lock);
2446 no changes are required. When any other reference (received _not_ from
2447 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
2448 instead of LDLM_LOCK_PUT().
2450 Arguably, changes are pervasive, and interface requires some discipline
2451 for proper use. On the other hand, it was very instrumental in finding
2452 a few leaked lock references.
2456 Description: Add ldlm_lock_addref_try().
2457 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
2458 attempts to addref a lock that might be being canceled
2463 Description: Add ldlm_weigh_callback().
2464 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
2465 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
2466 hard-coded `number of pages' logic.
2470 Description: Add lockdep annotations to llog code.
2471 Details : Use appropriately tagged _nested() locking calls in the places
2472 where llog takes more than one ->lgh_lock lock.
2476 Description: Add loi_kms_set().
2477 Details : Wrap kms updates into a helper function.
2481 Description: Constify instances of struct lsm_operations.
2482 Details : Constify instances of struct lsm_operations.
2486 Description: lu_conf support.
2487 Details : On a server, a file system object is uniquely identified
2488 by a fid, which is sufficient to locate and load all object
2489 state (inode). On a client, on the other hand, more data are
2490 necessary instantiate an object. Change lu_object_find() and
2491 friends to take additional `lu_conf' argument describing object.
2492 Typically this includes layout information.
2496 Description: lu_context fixes.
2497 Details : Introduce new lu_context functions that are needed on the client
2498 side, where some system threads (ptlrpcd) are shared by multiple
2499 modules, and so cannot be stopped during module shutdown.
2503 Description: Add start and stop methods to lu_device_type_operations.
2504 Details : Introduce two new methods in lu_device_type_operations, that are
2505 invoked when first instance of a given type is created and last one
2506 is destroyed respectively. This is need by CLIO.
2510 Description: Add lu_ref support to struct lu_device.
2511 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
2512 track leaked references.
2516 Description: Introduce lu_kmem_descr.
2517 Details : lu_kmem_descr and its companion interface allow to create
2518 and destroy a number of kmem caches at once.
2522 Description: Fix lu_object finalization race.
2523 Details : Fix a race between lu_object_find() finding an object and its
2524 concurrent finalization. This race is (most likely) not possible
2525 on the server, but might happen on the client.
2529 Description: Introduce lu_ref interface.
2530 Details : lu_ref is a debugging module allowing to track references to
2531 a given object. It is quite cpu expensive, and has to be
2532 explicitly enabled with --enable-lu_ref. See usage description
2537 Description: Factor lu_site procfs stats into a separate function.
2538 Details : Separate lu_site stats printing code into a separate function
2539 that can be reused on a client.
2543 Description: Constify instances of struct {lu,dt,md}_device_operations.
2544 Details : Constify instances of struct {lu,dt,md}_device_operations.
2548 Description: Introduce struct md_site and move meta-data specific parts of
2549 struct lu_site here.
2550 Details : Move md-specific fields out of struct lu_site into special struct
2551 md_site, so that lu_site can be used on a client.
2555 Description: Kill mdd_lov_destroy().
2556 Details : Remove unused mdd code.
2560 Description: Add st_block checking to multistat.c.
2561 Details : Add st_block checking to multistat.c.
2565 Description: Add lu_ref support to struct obd_device.
2566 Details : Add lu_ref tracking to obd_device.
2570 Description: Kill obd_set_fail_loc().
2571 Details : Remove unused code.
2575 Description: Add special type for ptlrpc_request interpret functions.
2576 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
2577 there. Actual usage will be in CLIO.
2581 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
2582 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
2583 doesn't work with lockdep.
2587 Description: Add rwv.c test program.
2588 Details : New testing program exercising readv(2) and writev(2) (Qian).
2592 Description: Add sendfile.c test program.
2593 Details : New testing program exercising sendfile(2) (Jay).
2597 Description: Ratelimit a message that can be very frequent.
2598 Details : Ratelimit a memory allocation failure message that can
2603 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2604 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2608 Description: Liblustre build fixes.
2609 Details : Liblustre build fixes.
2613 Description: libcfs: add cfs_{need,cond}_resched() interface.
2614 Details : libcfs: add cfs_{need,cond}_resched() definition and
2615 implementations for Linux, NT, and liblustre.
2617 Severity : enhancement
2619 Description: More exported tunables for mballoc
2620 Details : Add support for tunable preallocation window and new tunables for
2621 large/small requests
2625 Description: Detect corruption of block bitmap and checking for preallocations
2626 Details : Checks validity of on-disk block bitmap. Also it does better
2627 checking of number of applied preallocations. When corruption is
2628 found, it turns filesystem readonly to prevent further corruptions.
2632 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
2633 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
2634 window to 0. If it does hit cache miss, and read-pattern is still
2635 stride-io mode, does not reset the stride window, but also does not
2636 increase the stride window length in this case.
2640 Frequency : only for big-endian servers
2641 Description: Check if system is big-endian while mounting fs with extents feature
2642 Details : Mounting a filesystem with extents feature will fail on big-endian
2643 systems since ext3-based ldiskfs is not supported on big-endian
2644 systems. This can be over-riden with "bigendian_extents" mount option.
2646 Severity : enhancement
2648 Description: The root squash functionality
2649 Details : A security feature, which is to prevent users from being able
2650 to mount lustre on their desktop, run as root, and delete
2651 all of the files in the filesystem. The goal is accomplished by
2652 remapping user id (UID) and group id (GID) of the root user to
2653 a UID and GID specified by the system administartor via Lustre
2654 configuration management server (MGS). The functionality also
2655 allows to specify sets of clients for which the remapping does
2660 Description: Excessive recovery window
2661 Details : With AT enabled, the recovery window can be excessively long (6000+
2662 seconds). To address this problem, we no longer use
2663 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
2664 timeout no longer depends on the service time, it is set to
2665 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
2666 time via pb_service_time.
2670 Description: Watchdog triggered on MDS failover
2671 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
2672 the OSTs know that the MDS "UUID" can be reused for the same export
2673 from a different NID, so we do not need to wait for the export to be
2677 Frequency : rare, only if using MMP with Linux RAID
2679 Description: MMP doesn't work with Linux RAID
2680 Details : While using HA for Lustre servers with Linux RAID, it is possible
2681 that MMP will not detect multiple mounts. To make this work we
2682 need to unplug the device queue in RAID when the MMP block is being
2683 written. Also while reading the MMP block, we should read it from
2684 disk and not the cached one.
2686 Severity : enhancement
2688 Description: open file using fid
2689 Details : A file can be opened using just its fid, like
2690 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
2693 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
2696 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
2697 Details : add proper locking for creation of jbd_4k slab cache
2701 Description: MMP check in ext3_remount() fails without displaying any error
2702 Details : When multiple mount protection fails during remount, proper error
2705 Severity : enhancement
2707 Description: Allow stripe size to be up to 4G-64k
2708 Details : Fix math logic to allow large stripe sizes.
2712 Description: add check for >8TB ldiskfs filesystems
2713 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
2714 Don't allow >8TB ldiskfs filesystems to be mounted without
2715 force_over_8tb mount option
2720 Descriptoin: OSS complains "too many blocks" and crashes
2721 Details : In filter_commitrw_write():
2722 1. lquota_chkquota();
2723 2. obdo_from_inode();
2724 3. lquota_pending_commit();
2725 "2." has changed uid and gid in this case, then caused
2726 lquota_chkquota() and lquota_pending_commit() to use
2727 different uid and gid. That is the root of the bug.
2731 Descriptoin: getstripe doesn't report stripe count = -1 sanely
2732 Details : Detect and print out the 'stripe count = -1' case properly.
2736 Description: Out or order replies might be lost on replay
2737 Details : In ptlrpc_retain_replayable_request if we cannot find retained
2738 request with tid smaller than one currently being added, add it
2739 to the start, not end of the list.
2741 --------------------------------------------------------------------------------
2743 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
2745 * Support for kernels:
2746 2.6.5-7.283 (SLES 9),
2747 2.6.9-55.EL (RHEL 4),
2748 2.6.16.46-0.14 (SLES 10),
2749 2.6.18.8 vanilla (kernel.org)
2750 * Client support for unpatched kernels:
2751 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2752 2.6.16 - 2.6.22 vanilla (kernel.org)
2753 * Due to recently discovered recovery problems, we do not recommend
2754 using patchless RHEL 4 clients with this or any earlier release.
2755 * Recommended e2fsprogs version: 1.39.cfs8
2756 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2757 * Starting with this release, the ldiskfs backing filesystem required
2758 by Lustre is now in its own package, lustre-ldiskfs. This package
2759 should be installed. It is versioned separately from Lustre and
2760 may be released separately in future.
2764 Description: block reactivating mgc import until all deactivates complete
2765 Details : Fix race when failing back MDT/MGS to itself (testing)
2768 Frequency : at statup only
2770 Description: mds_lov_synchronize race leads to various problems
2771 Details : simultaneous MDT->OST connections at startup can cause the
2772 sync to abort, leaving the OSC in a bad state.
2774 Severity : enhancement
2776 Description: add optional extra BUILD_VERSION info
2777 Details : add a new environment variable (namely LUSTRE_VERS) which allows
2778 to override the lustre version.
2781 Frequency : 2.6.18 servers only
2783 Description: ll_kern_mount() doesn't release the module reference
2784 Details : The ldiskfs module reference count never drops down to 0
2785 because ll_kern_mount() doesn't release the module reference.
2790 Description: server LBUG when using old ost_num_threads parameter
2791 Details : Accept the old ost_num_threads parameter but warn that it
2792 is deprecated, and fix an off-by-one error that caused an LBUG.
2797 Description: Transient SCSI error results in persistent IO issue
2798 Details : iobuf->dr_error is not reinitialized to 0 between two
2802 Frequency : sometimes when underlying device returns I/O errors
2804 Description: OSTs not going read-only during write failures
2805 Details : OSTs are not remounted read-only when the journal commit threads
2806 get I/O errors because fsfilt_ext3 calls journal_start/stop()
2807 instead of the ext3 wrappers.
2811 Description: poor connect scaling with increasing client count
2812 Details : Don't run filter_grant_sanity_check for more than 100 exports
2813 to improve scaling for large numbers of clients.
2816 Frequency : SLES10 only
2818 Description: sanity-quota.sh quotacheck failed: rc = -22
2819 Details : Quotas cannot be enabled on SLES10.
2822 Frequency : liblustre clients only
2824 Description: getdirentries does not give error when run on compute nodes
2825 Details : getdirentries does not fail when the size specified as an argument
2826 is too small to contain at least one entry
2828 Severity : enhancement
2830 Description: Add LNET router traceability for debug purposes
2831 Details : If a checksum failure occurs with a router as part of the
2832 IO path, the NID of the last router that forwarded the bulk data
2833 is printed so it can be identified.
2838 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
2839 Details : Due to a race condition, liblustre clients were occasionally
2840 evicted incorrectly.
2842 Severity : enhancement
2844 Description: lfs setstripe use optional parameters instead of postional
2847 Severity : enhancement
2849 Description: Nanosecond timestamp support for ldiskfs
2850 Details : The on-disk ldiskfs filesystem has added support for nanosecond
2851 resolution timestamps. There is not yet support for this at
2852 the Lustre filesystem level.
2855 Frequency : during server recovery
2857 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
2858 Details : request with rq_no_resend flag not awake l_wait_event if they get
2862 Frequency : nfs export on patchless client
2864 Description: connectathon hang when test nfs export over patchless client
2865 Details : Disconnected dentry cannot be found with lookup, so we do not need
2866 to unhash it or make it invalid
2869 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
2870 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
2871 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
2876 Description: open req refcounting wrong on reconnect
2877 Details : If reconnect happened between getting open reply from server and
2878 call to mdc_set_replay_data in ll_file_open, we will schedule
2879 replay for unreferenced request that we are about to free.
2880 Subsequent close will crash in variety of ways.
2881 Check that request is still eligible for replay in
2882 mdc_set_replay_data().
2887 Description: disable writes to filesystem when reading health_check file
2888 Details : the default for reading the health_check proc file has changed
2889 to NOT do a journal transaction and write to disk, because this
2890 can cause reads of the /proc file to hang and block HA state
2891 checking on a healthy but otherwise heavily loaded system. It
2892 is possible to return to the previous behaviour during configure
2893 with --enable-health-write.
2895 Severity : enhancement
2897 Description: 64-bit inode version
2898 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
2899 to the inode. This will be required for version-based recovery.
2904 Description: MDS fails to start if a duplicate client export is detected
2905 Details : in some rare cases it was possible for a client to connect to
2906 an MDS multiple times. Upon recovery the MDS would detect this
2907 and fail during startup. Handle this more gracefully.
2909 Severity : enhancement
2911 Description: Add -o localflock option to simulate old noflock
2913 Details : This will achieve local-only flock/fcntl locks
2919 Description: log_commit_thread vs filter_destroy race leads to crash
2920 Details : Take import reference before releasing llog record semaphore
2925 Description: Wrong request locking in request set processing
2926 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
2927 imp_delayed_list, in this place should be used imp_lock.
2930 Frequency : when reconnection
2932 Description: Grant Leak when osc reconnect to OST
2933 Details : When osc reconnect ost, OST(filter) should check whether it
2934 should grant more space to client by comparing fed_grant and
2935 cl_avail_grant, and return the granted space to client instead
2936 of "new granted" space, because client will call osc_init_grant
2937 to update the client grant space info.
2940 Frequency : when client reconnect to OST
2942 Description: Grant Leak when osc do resend and replay bulk write
2943 Details : When osc reconnect to OST, OST(filter)should clear grant info of
2944 bulk write request, because the grant info while be sync between
2945 OSC and OST when reconnect, and we should ignore the grant info
2946 these of resend/replay write req.
2951 Description: Grant space more than available left space sometimes.
2952 Details : When then OST is about to be full, if two bulk writing from
2953 different clients came to OST. Accord the avaliable space of the
2954 OST, the first req should be permitted, and the second one
2955 should be denied by ENOSPC. But if the seconde arrived before
2956 the first one is commited. The OST might wrongly permit second
2957 writing, which will cause grant space > available space.
2960 Frequency : when client is evicted
2962 Description: Grant might be wrongly erased when osc is evicted by OST
2963 Details : when the import is evicted by server, it will fork another
2964 thread ptlrpc_invalidate_import_thread to invalidate the
2965 import, where the grant will be set to 0. While the original
2966 thread will update the grant it got when connecting. So if
2967 the former happened latter, the grant will be wrongly errased
2968 because of this race.
2973 Description: Checking Stale with correct fid
2974 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
2975 to check whether it is stale, because sometimes, we want the
2976 enqueue happened anyway, and op_data.fid2 will not be initialized.
2978 Severity : enhancement
2980 Description: update patchless client
2981 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
2984 Frequency : only with 2.4 kernel
2986 Description: random memory corruption
2987 Details : size of struct ll_inode_info is to big for union inode.u and this
2988 can be cause of random memory corruption.
2993 Description: Memory leak in recovery
2994 Details : Lov_mds_md was not free in an error handler in mds_create_object.
2995 It should also check obd_fail before fsfilt_start, otherwise if
2996 fsfilt_start return -EROFS,(failover mds during mds recovery).
2997 then the req will return with repmsg->transno = 0 and rc = EROFS.
2998 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
2999 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
3000 be freed no matter whether fsfilt_commit success or not.
3003 Frequency : only with huge count clients
3005 Description: Prevents from taking the superblock lock in llap_from_page for
3007 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
3008 for llap_from_page call in ll_removepage prevents from taking the
3009 superblock lock for a soon died page.
3014 Description: Not check open intent error before release open handle
3015 Details : in some rare cases, the open intent error is not checked before
3016 release open handle, which may cause
3017 ASSERTION(open_req->rq_transno != 0), because it tries to release
3018 the failed open handle.
3023 Description: Set cat log bitmap only after create log success.
3024 Details : in some rare cases, the cat log bitmap is set too early. and it
3025 should be set only after create log success.
3029 Description: Accessing a block bevice can re-enable I/O when Lustre is
3030 tearing down a device.
3031 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
3036 Description: service threads may hog cpus when there are a lot of requests
3037 Details : Insert cond_resched to give other threads a chance to use some CPU
3042 Description: the cat log was not initialized in recovery
3043 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
3044 unlink log on mds will not be initialized until mds post
3045 recovery. And also in mds post recovery, the unlink log will
3046 initialization will be done asynchronausly, so there will be race
3047 between add unlink log and unlink log initialization.
3051 Description: brw_stats were being printed incorrectly
3052 Details : brw_stats were being printed as log2 but all of them were not
3053 recorded as log2. Also remove some code duplication arising from
3054 filter_tally_{read,write}.
3058 Frequency : rare, only in recovery.
3059 Description: ASSERTION(req->rq_type != LI_POISON) failed
3060 Details : imp_lock should be held while iterating over imp_sending_list for
3061 prevent destroy request after get timeout in ptlrpc_queue_wait.
3065 Description: replay-single.sh test 52 fails
3066 Details : A lock's skiplist need to be cleanup when it being unlinked
3067 from its resource list.
3071 Description: Short directio read returns full requested size rather than
3073 Details : Direct I/O operations should return actual amount of bytes
3074 transferred rather than requested size.
3076 Severity : enhancement
3078 Description: metadata RPC reduction (e.g. for rm performance)
3079 Details : decrease the amount of synchronous RPC between clients and servers
3080 by canceling conflicing lock before the operation on the client
3081 and packing thier handles into the main operation RPC to server.
3083 Severity : enhancement
3085 Description: Async OSC create to avoid the blocking unnecessarily.
3086 Details : If an OST has no remaining object, system will block on the
3087 creation when it needs to create a new object on this OST. Now,
3088 ways use pre-created objects when available, instead of blocking on
3089 an empty osc while others are not empty. If we must block, we
3090 block for the shortest possible period of time.
3094 Description: improve handling recoverable errors
3095 Details : if request processig with error which can be recoverable on server
3096 request should be resend, otherwise page released from cache and
3099 Severity : enhancement
3101 Description: refine locking for avoid write wrong info into lov_objid file
3102 Details : fix possible races with add new target and write/update data in
3105 --------------------------------------------------------------------------------
3107 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
3112 Frequency : on some architectures
3114 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
3115 Details : Uninitialized flags sometimes cause configuration commands to
3119 Frequency : patchless clients only
3121 Description: missing __iget() symbol export
3122 Details : The __iget() symbol export is missing. To avoid the need for
3123 this on patchless clients the deathrow inode reaper is turned
3124 off, and we depend on the VM to clean up old inodes. This
3125 dependency was during via the fix for bug 12181.
3127 --------------------------------------------------------------------------------
3129 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
3131 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
3132 INTEROPERATE with older versions automatically. In many cases a
3133 special upgrade step is needed. Please read the
3134 user documentation before upgrading any part of a live system.
3135 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
3136 version will not interoperate with 1.6 betas before beta5 (1.5.95).
3137 * WARNING: Lustre configuration and startup changes are required with
3138 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
3142 Severity : enhancement
3144 Description: MountConf
3145 Details : Lustre configuration is now managed via mkfs and mount
3146 commands instead of lmc and lconf. New obd types (MGS, MGC)
3147 are added for dynamic configuration management. See
3148 https://mail.clusterfs.com/wikis/lustre/MountConf for
3151 Severity : enhancement
3153 Description: dynamic OST addition
3154 Details : OSTs can now be added to a live filesystem
3156 Severity : enhancement
3158 Description: startup order invariance
3159 Details : MDTs and OSTs can be started in any order. Clients only
3160 require the MDT to complete startup.
3162 Severity : enhancement
3164 Description: parallel, asynchronous orphan cleanup
3165 Details : orphan cleanup is now performed in separate threads for each
3166 OST, allowing parallel non-blocking operation.
3168 Severity : enhancement
3170 Description: optimized stripe assignment
3171 Details : stripe assignments are now made based on ost space available,
3172 ost previous usage, and OSS previous usage, in order to try
3173 to optimize storage space and networking resources.
3175 Severity : enhancement
3177 Description: Permanently set tunables
3178 Details : All writable /proc/fs/lustre tunables can now be permanently
3179 set on a per-server basis, at mkfs time or on a live system.
3181 Severity : enhancement
3183 Description: Lustre message v2
3184 Details : Add lustre message format v2.
3186 Severity : enhancement
3188 Description: client OST exclusion list
3189 Details : Clients can be started with a list of OSTs that should be
3190 declared "inactive" for known non-responsive OSTs.
3194 Description: SPEC SFS validation failure on NFS v2 over lustre.
3195 Details : Changes the blocksize for regular files to be 2x RPC size,
3196 and not depend on stripe size.
3198 Severity : enhancement
3200 Description: Multiple MD RPCs in flight.
3201 Details : Further unserialise some read-only MDS RPCs - learn about intents.
3202 To avoid overly-overloading MDS, introduce a limit on number of
3203 MDS RPCs in flight for a single client and add /proc controls
3204 to adjust this limit.
3206 Severity : enhancement
3208 Description: client read/write statistics
3209 Details : Add client read/write call usage stats for performance
3210 analysis of user processes.
3211 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
3212 file access. extents_stats shows chunk size distribution.
3213 extents_stats_per_process show chunk size distribution per
3216 Severity : enhancement
3218 Description: mds statistics
3219 Details : Add detailed mds operations statistics in
3220 /proc/fs/lustre/mds/*/stats.
3224 Description: Failure of copying files with lustre special EAs.
3225 Details : Client side always return success for setxattr call for lustre
3226 special xattr (currently only "trusted.lov").
3231 Description: Refcount LNET uuids
3232 Details : The global LNET uuid list grew linearly with every startup;
3233 refcount repeated list entries instead of always adding to
3236 Severity : enhancement
3238 Description: Dynamic service threads
3239 Details : Within a small range, start extra service threads
3240 automatically when the request queue builds up.
3243 Frequency : mixed-endian client/server environments
3245 Description: mixed-endian crashes
3246 Details : The new msg_v2 system had some failures in mixed-endian
3249 Severity : enhancement
3251 Description: Easy OST removal
3252 Details : OSTs can be permanently deactivated with e.g. 'lctl
3253 conf_param lustre-OST0001.osc.active=0'
3255 Severity : enhancement
3257 Description: MGS proc entries
3258 Details : Added basic proc entries for the MGS showing what filesystems
3261 Severity : enhancement
3263 Description: provide MGS failover
3264 Details : Added config lock reacquisition after MGS server failover.
3266 Severity : enhancement
3268 Description: add Linux 2.4 support
3269 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
3273 Description: a large application tries to do I/O to the same resource and dies
3274 in the middle of it.
3275 Details : Check the req->rq_arrival time after the call to
3276 ost_brw_lock_get(), but before we do anything about
3277 processing it & sending the BULK transfer request. This
3278 should help move old stale pending locks off the queue as
3279 quickly as obd_timeout.
3282 Frequency : when an incorrect nid is specified during startup
3284 Description: ptlrpc connect to non-existant node causes kernel crash
3285 Details : LNET can't be re-entered from an event callback, which
3286 happened when we expire a message after the export has been
3287 cleaned up. Instead, hand the zombie cleanup off to another
3290 Severity : enhancement
3292 Description: plain/inodebits lock performance improvement
3293 Details : Grouping plain/inodebits in granted list by their request modes
3294 and bits policy, thus improving the performance of search through
3298 Frequency : only if OST filesystem is corrupted
3300 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3301 Details : for a short time RPCs with bulk IO are in the replay list,
3302 but replay of bulk IOs is unimplemented. If the OST filesystem
3303 is corrupted due to disk cache incoherency and then replay is
3304 started it is possible to trip an assertion. Avoid putting
3305 committed RPCs into the replay list at all to avoid this issue.
3308 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3310 Description: System hang on startup
3311 Details : This bug allowed the liblustre (e.g. catamount) client to
3312 return to the app before handling all startup RPCs. This
3313 could leave the node unresponsive to lustre network traffic
3314 and manifested as a server ptllnd timeout.
3316 Severity : enhancement
3318 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3319 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
3320 boolean to control whether to print peer debug info when a
3321 client's RPC times out.
3324 Frequency : only for kernels with patches from Lustre below 1.4.3
3326 Description: Remove old rdonly API
3327 Details : Remove old rdonly API which unused from at least lustre 1.4.3
3330 Frequency : only for devices with external journals
3332 Description: Set external device read-only also
3333 Details : During a commanded failover stop, we set the disk device
3334 read-only while the server shuts down. We now also set any
3335 external journal device read-only at the same time.
3338 Frequency : when upgrading from 1.4 while trying to change parameters
3340 Description: The wrong (new) MDC name was used when setting parameters for
3341 upgraded MDT's. Also allows changing of OSC (and MDC)
3342 parameters if --writeconf is specified at tunefs upgrade time.
3345 Frequency : when setting specific ost indicies
3347 Description: QOS code breaks on skipped indicies
3348 Details : Add checks for missing OST indicies in the QOS code, so OSTs
3349 created with --index need not be sequential.
3351 Severity : enhancement
3353 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3354 Details : The uninit_groups feature works in conjunction with the kernel
3355 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3356 up the pass1 processing of e2fsck. This is a read-only feature
3357 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3358 mount filesystems that have had this feature enabled.
3360 Severity : enhancement
3362 Description: Improve multi-block allocation algorithm to avoid fragmentation
3363 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3364 allocation locality and avoid filesystem fragmentation.
3366 ------------------------------------------------------------------------------
3368 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
3370 * Support for kernels:
3371 2.4.21-47.0.1.EL (RHEL 3)
3372 2.6.5-7.283 (SLES 9)
3373 2.6.9-42.0.10.EL (RHEL 4)
3374 2.6.12.6 vanilla (kernel.org)
3375 2.6.16.27-0.9 (SLES 10)
3376 * Recommended e2fsprogs version: 1.39.cfs5
3378 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3382 Frequency : occasional, depends on client load and configuration
3383 Bugzilla : 12181, 12203
3384 Description: data loss for recently-modified files
3386 Details : In some cases it is possible that recently written or created
3387 files may not be written to disk in a timely manner (this should
3388 normally be within 30s unless client IO load is very high).
3389 The problem appears as zero-length files or files that are a
3390 multiple of 1MB in size after a client crash or client eviction
3391 that are missing data at the end of the file.
3393 This problem is more likely to be hit on clients where files are
3394 repeatedly created and unlinked in the same directory, clients
3395 have a large amount of RAM, have many CPUs, the filesystem has
3396 many OSTs, the clients are rebooted frequently, and/or the files
3397 are not accessed by other nodes after being written.
3399 The presence of the problem can be detected by looking at
3400 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
3401 smaller than the second (nr_unused) then dirty files will not
3402 be flushed automatically to disk. "sync; sleep 10" should be
3403 run several times on the node before unmounting it to update
3404 Lustre (this is also safe to run on nodes without this problem).
3406 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
3407 that can cause this same problem, so customers using that kernel
3408 also need to update the kernel in addition to Lustre. In order
3409 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
3411 It is normal that files written just before a client crash (less
3412 than 30s) may not yet have been flushed to disk, even for local
3416 Frequency : frequent on thin XT3 nodes
3418 Description: UUID collision on thin XT3 Linux nodes
3419 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
3420 randomly, since we relied on an insufficiently-seeded PRNG.
3425 Description: OSS hangs after "All ost request buffers busy"
3426 Details : A deadlock between quota and journal operations caused OSS
3427 hangs after printing "All ost request buffers busy."
3430 Frequency : always on liblustre builds
3432 Description: Cleanup compiler warnings on liblustre
3435 Frequency : always on liblustre builds on XT3
3437 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
3441 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
3442 > 32000 subdirectories
3443 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
3444 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
3445 subdirectory count crosses 32000. This will aid e2fsck to
3446 correctly handle more than 32000 subdirectories.
3449 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3451 Description: System hang on startup
3452 Details : This bug allowed the liblustre (e.g. catamount) client to
3453 return to the app before handling all startup RPCs. This
3454 could leave the node unresponsive to lustre network traffic
3455 and manifested as a server ptllnd timeout.
3457 Severity : enhancement
3459 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3460 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
3461 boolean to control whether to print peer debug info when a
3462 client's RPC times out.
3467 Description: make O_SYNC working on 2.6 kernels
3468 Details : 2.6 kernels use different method for mark pages for write,
3469 so need add a code to lustre for O_SYNC work.
3474 Description: Failure to close file and release space on NFS
3475 Details : Put inode details into lock acquired in ll_intent_file_open.
3476 Use mdc_intent_lock in ll_intent_open to properly
3477 detect all kind of errors unhandled by mdc_enqueue.
3482 Description: proc file read during shutdown sometimes raced obd removal,
3484 Details : Add lock to prevent obd access after proc file removal.
3487 Frequency : Only for files larger than 4GB on 32-bit clients.
3489 Description: improperly doing page alignment of locks
3490 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
3491 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3496 Description: RPCs being resent when they shouldn't be
3497 Details : Some RPCs that should not be resent are being resent. This
3498 can cause inconsistencies in the RPC state machine. Do not
3499 resend such requests.
3502 Frequency : rare, only with NFS export
3504 Description: Crash on NFS re-export node
3505 Details : under very unusual load conditions an assertion is hit in
3506 ll_intent_file_open()
3509 Frequency : only if OST filesystem is corrupted
3511 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3512 Details : for a short time RPCs with bulk IO are in the replay list,
3513 but replay of bulk IOs is unimplemented. If the OST filesystem
3514 is corrupted due to disk cache incoherency and then replay is
3515 started it is possible to trip an assertion. Avoid putting
3516 committed RPCs into the replay list at all to avoid this issue.
3521 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3522 Details : Large single O_DIRECT read and write calls can fail to allocate
3523 a sufficiently large buffer to process the request. In case of
3524 allocation failure the allocation is retried with a smaller
3525 buffer and broken into smaller requests.
3527 Severity : enhancement
3529 Description: Add -o localflock option to simulate old noflock behaviour.
3530 Details : This will achieve local-only flock/fcntl locks coherentness.
3535 Description: versioning check is incomplete
3536 Details : Checking the version difference of client vs. server, report
3537 error if the gap is too big.
3542 Description: add support PG_writeback bit
3543 Details : add support for PG_writeback bit for Lustre, for more carefull
3544 work with page cache in 2.6 kernel. This also fix some deadlocks
3545 and remove hack for work O_SYNC with 2.6 kernel.
3547 Severity : enhancement
3549 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3550 Details : The uninit_groups feature works in conjunction with the kernel
3551 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3552 up the pass1 processing of e2fsck. This is a read-only feature
3553 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3554 mount filesystems that have had this feature enabled.
3556 Severity : enhancement
3558 Description: Improve multi-block allocation algorithm to avoid fragmentation
3559 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3560 allocation locality and avoid filesystem fragmentation.
3562 ------------------------------------------------------------------------------
3564 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3566 * Support for kernels:
3567 2.6.9-42.0.3.EL (RHEL 4)
3568 2.6.5-7.276 (SLES 9)
3569 2.4.21-47.0.1.EL (RHEL 3)
3570 2.6.12.6 vanilla (kernel.org)
3571 2.6.16.21-0.8 (SLES10)
3572 * Recommended e2fsprogs version: 1.39.cfs2-0
3574 * The backwards-compatible /proc/sys/portals symlink has been removed
3575 in this release. Before upgrading, please ensure that you change
3576 any configuration scripts or /etc/sysctl.conf files that access
3577 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3578 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3579 in advance of the upgrade on any system running Lustre 1.4.6 or
3580 newer, since /proc/sys/lnet was added in that version.
3581 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3585 Frequency : only when quota is used
3587 Description: avoid scanning export list for quota master
3588 Details : Change the algorithms to avoid scanning export list in order
3589 to improve the efficiency.
3592 Frequency : MDS failover only, very rarely
3594 Description: "went back in time" messages on mds failover
3595 Details : The greatest transno may be lost when the current operation
3596 finishes with an error (transno==0) and the client's last_rcvd
3597 record is over-written. Save the greatest transno in the
3598 mds_last_transno for this case.
3601 Frequency : always for specific kernels and striping counts
3603 Description: client may get "Matching packet too big" without ACL support
3604 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
3605 when trying to access files in certain configurations. The
3606 clients should in fact be denied when mounting because they do
3607 not understand ACLs.
3610 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
3612 Description: many clients connecting with IO in progress causes connect timeouts
3613 Details : Avoid synchronous journal commits to avoid delays caused by many
3614 clients connecting/disconnecting when bulk IO is in progress.
3615 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
3616 OST_IO_PORTAL to avoid delays behind potentially many pending
3620 Frequency : occasionally with multiple writers to a single file
3622 Description: shared writes to file may result in wrong size reported by stat()
3623 Details : Allow growing of kms when extent lock is cancelled
3626 Frequency : always with random mmap IO to multi-striped file
3628 Description: mmap write might be lost if we are writing to a 'hole' in stripe
3629 Details : Only if the hole is at the end of OST object so that kms is too
3630 small. Fix is to increase kms accordingly in ll_nopage.
3633 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
3635 Description: writes to a missing object would leak memory on the OST
3636 Details : If there is an inconsistency between the MDS and OST filesystems,
3637 such that the MDS references an object that doesn't exist, writes
3638 to that object will leak memory due to incorrect cleanup in the
3639 error handling path, eventually running out of memory on the OST.
3644 Description: Creating too long symlink causes lustre errors
3645 Details : Check symlink and name lengths before sending requests to MDS.
3648 Frequency : only if flock is enabled (not on by default)
3650 Description: posix locks not released on fd closure on 2.6.9+
3651 Details : We failed to add posix locks to list of inode locks on 2.6.9+
3652 kernels, this caused such locks not to be released on fd close and
3653 then assertions on fs unmount about still used locks.
3656 Frequency : MDS failover only, very rarely
3658 Description: clients may get ASSERTION(granted_lock != NULL)
3659 Details : When request was taking a long time, and a client was resending
3660 a getattr by name lock request. The were multiple lock requests
3661 with the same client lock handle and
3662 mds_getattr_name->fixup_handle_for_resent_request found one of the
3663 lock handles but later failed with ASSERTION(granted_lock != NULL).
3668 Description: handle->h_buffer_credits > 0, assertion failure
3669 Details : h_buffer_credits is zero after truncate, causing assertion
3670 failure. This patch extends the transaction or creates a new
3674 Frequency : NFS re-export or patchless client
3675 Bugzilla : 11179, 10796
3676 Description: Crash on NFS re-export node (__d_move)
3677 Details : We do not want to hash the dentry if we don't have a lock.
3678 But if this dentry is later used in d_move, we'd hit uninitialised
3679 list head d_hash, so we just do this to init d_hash field but
3680 leave dentry unhashed.
3683 Frequency : NFS re-export or patchless client
3685 Description: NFS exports has problem with symbolic link
3686 Details : lustre client didn't properly install dentry when re-exported
3687 to NFS or running patchless client.
3690 Frequency : NFS re-export or patchless client
3692 Description: Various nfs/patchless fixes.
3693 Details : fixes reuse disconected alias for lookup process - this fixes
3694 warning "find_exported_dentry: npd != pd",
3695 fix permission error with open files at nfs.
3696 fix apply umask when do revalidate.
3699 Frequency : occasional
3701 Description: Crash on NFS re-export node
3702 Details : calling clear_page() on the wrong pointer triggered oops in
3703 generic_mapping_read().
3706 Frequency : rarely, using O_DIRECT IO
3708 Description: unaligned directio crashes client with LASSERT
3709 Details : check for unaligned buffers before trying any requests.
3712 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
3714 Description: stale data returned from RAID cache
3715 Details : If only a small amount of IO is done to the RAID device before
3716 reading it again it is possible to get stale data from the RAID
3717 cache instead of reading it from disk.
3720 Frequency : always for sles10 kernel
3722 Description: sles10 support
3723 Details : ll_follow_link: compile fixes and using of nd_set_link
3724 under newer kernels.
3727 Frequency : depends on arch, kernel and compiler version, always on sles10
3730 Description: recursive or deep enough symlinks cause stack overflow
3731 Details : getting rid of large stack-allocated variable in
3735 Frequency : depends on hardware
3737 Description: lustre write performance loss in the SLES10 kernel
3738 Details : the performance loss is caused by using of write barriers in the
3739 ext3 code. The SLES10 kernel turns barrier support on by
3740 default. The fix is to undo that change for ldiskfs.
3742 ------------------------------------------------------------------------------
3744 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
3746 * Support for kernels:
3747 2.6.9-42.0.3EL (RHEL 4)
3748 2.6.5-7.276 (SLES 9)
3749 2.4.21-47.0.1.EL (RHEL 3)
3750 2.6.12.6 vanilla (kernel.org)
3754 Frequency : quota enabled and large files being deleted
3756 Description: releasing more than 4GB of quota at once hangs OST
3757 Details : If a user deletes more than 4GB of files on a single OST it
3758 will cause the OST to spin in an infinite loop. Release
3759 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3764 Description: statfs data retrieved from /proc may be stale or zero
3765 Details : When reading per-device statfs data from /proc, in the
3766 {kbytes,files}_{total,free,avail} files, it may appear
3767 as zero or be out of date.
3770 Frequency : systems with MD RAID1 external journal devices
3772 Description: lconf's call to blkid is confused by RAID1 journal devices
3773 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3774 of returning all block devices that match the journal UUID.
3777 Frequency : always, for aggregate stripe size over 4GB
3779 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
3780 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
3781 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
3784 Frequency : always if "lfs find" used on a local file/directory
3786 Description: "lfs find" segfaults if used on a local file/directory
3787 Details : The case where a directory component was not specified wasn't
3788 handled correctly. Handle this properly.
3791 Frequency : always on ppc64
3793 Description: the write to an ext3 filesystem mounted with mballoc got stuck
3794 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3795 perform endianness conversion.
3798 Frequency : rarely (truncate to non-zero file size after write under load)
3799 Bugzilla : 10730, 10687
3800 Description: Files padded with zeros to next 4K multiple
3801 Details : With filesystems mounted using the "extents" option (2.6 kernels)
3802 it is possible that files that are truncated to a non-zero size
3803 immediately after being written are filled with zero bytes beyond
3804 the truncated size. No file data is lost.
3806 Severity : enhancement
3808 Description: Allow recovery/failover for liblustre clients.
3809 Details : liblustre clients were unaware of failover configurations until
3812 Severity : enhancement
3814 Description: user file locks should fail when not mounting with flock option
3815 Details : Set up an error-returning stub in ll_file_operations.lock field
3816 to prevent incorrect behaviour when client is mounted without
3817 flock option. Also, set up properly f_op->flock field for
3821 Frequency : always on ia64
3823 Description: "lfs df" loops on printing out MDS statfs information
3824 Details : The obd_ioctl_data was not initialized and in some systems
3825 this caused a failure during the ioctl that did not return
3826 an error. Initialize the struct and return an error on failure.
3829 Frequency : SLES 9 only
3831 Description: Error of copying files with lustre special EAs as root
3832 Details : Client side always return success for setxattr call for lustre
3833 special xattr (currently only "trusted.lov").
3836 Frequency : rarely on clusters with both ia64+i386 clients
3838 Description: ia64+i686 clients doing shared IO on the same file may LBUG
3839 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
3840 clients are doing concurrent writes to the same file it is
3841 possible that the ia64 clients may LASSERT because the OST
3842 extent locks are not PAGE_SIZE aligned. Ensure that grown
3843 locks are always aligned on the request boundary.
3846 Frequency : specific use, occasional
3848 Description: Overwriting in use executable truncates on-disk binary image
3849 Details : If one node attempts to overwrite an executable in use by
3850 another node, we now correctly return ETXTBSY instead of
3851 truncating the file.
3853 Severity : enhancement
3855 Description: Async OSC create to avoid the blocking unnecessarily.
3856 Details : If an OST has no remaining object, system will block on the creating
3857 when need to create a new object on this OST. Now, ways use
3858 pre-created objects when available, instead of blocking on an
3859 empty osc while others are not empty. If we must block, we block
3860 for the shortest possible period of time.
3865 Description: chmod on Lustre root is propagated to other clients
3866 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
3867 invalid by the follow_mount time.
3870 Frequency : liblustre clients only
3872 Description: Race in 'instant cancel' lock handling could lead to such locks
3873 never to be granted in case of SMP MDS
3874 Details : Do not destroy not yet granted but cbpending locks in
3878 Frequency : replay/resend of open
3880 Description: non null lock assetion failure in mds_intent_policy
3881 Details : Trying to replay/resend lockless open requests resulted in
3882 mds_open() returning 0 with no lock. Now it sets a flag if
3883 a lock is going to be returned.
3885 Severity : enhancement
3887 Description: Checksum enhancements
3888 Details : New checksum enhancements allow for resending RPCs that failed
3891 Severity : enhancement
3893 Description: Tunables on number of dirty pages in cacche
3894 Details : Allow to set limit on number of dirty pages cached.
3899 Description: client crash on unmount - lock still has references
3900 Details : In some error handling cases it was possible to leak a lock
3901 reference on a client while accessing a file. This was not
3902 harmful to the client during operation, but would cause the
3903 client to crash when the filesystem is unmounted.
3906 Frequency : specific case, rare
3908 Description: ETXTBSY on mds though file not in use by client
3909 Details : ETXTBSY is no longer incorrectly returned when attempting to
3910 chmod or chown a directory that the user previously tried to
3911 execute or a currently-executing binary.
3914 Frequency : extremely rare except on liblustre-based clients
3916 Description: Lustre space not freed when files are deleted
3917 Details : Clean up open-unlinked files after client eviction. Previously
3918 the unlink was skipped and the files remained as orphans.
3923 Description: OST failure "would be an LBUG" in waiting_locks_callback()
3924 Details : In some cases it was possible to send a blocking callback to
3925 a client doing a glimpse, even though that client didn't get
3926 a lock granted. When the glimpse lock is cancelled on the OST
3927 the freed lock is left on the waiting list and corrupted the list.
3930 Frequency : all core dumps
3932 Description: Broke core dumps to lustre
3933 Details : Negative dentry may be unhashed if parent does not have UPDATE
3934 lock, but some callers, e.g. do_coredump, expect dentry to be
3935 hashed after successful create, hash it in ll_create_it.
3937 ------------------------------------------------------------------------------
3939 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
3941 * Support for kernels:
3942 2.6.9-42.0.2.EL (RHEL 4)
3943 2.6.5-7.276 (SLES 9)
3944 2.4.21-40.EL (RHEL 3)
3945 2.6.12.6 vanilla (kernel.org)
3949 Frequency : always on RHEL 3
3951 Description: Number of open files grows over time
3952 Details : The number of open files grows over time, whether or not
3953 Lustre is started. This was due to a filp leak introduced
3954 by one of our kernel patches.
3956 ------------------------------------------------------------------------------
3958 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
3960 * Support for kernels:
3961 2.6.9-42.EL (RHEL 4)
3962 2.6.5-7.276 (SLES 9)
3963 2.4.21-40.EL (RHEL 3)
3964 2.6.12.6 vanilla (kernel.org)
3969 Bugzilla : 5719, 9635, 9792, 9684
3970 Description: OST (or MDS) trips assertions in (re)connection under heavy load
3971 Details : If a server is under heavy load and cannot reply to new
3972 connection requests before the client resends the (re)connect,
3973 the connection handling code can behave badly if two service
3974 threads are concurrently handing separate (re)connections from
3975 the same client. Add better locking to the connection handling
3976 code, and ensure that only a single connection will be processed
3977 for a given client UUID, even if the lock is dropped.
3979 Severity : enhancement
3981 Description: add TCP zero-copy support to kernel
3982 Details : Add support to the kernel TCP stack to allow zero-copy bulk
3983 sends if the hardware supports scatter-gather and checksumming.
3984 This allows socklnd to do client-write and server-read more
3985 efficiently and reduce CPU utilization from skbuf copying.
3988 Frequency : only if NFS exporting from client
3990 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
3991 Details : If setting attributes on a file created under NFS that had
3992 never been opened it would be possible to oops the client
3993 if the file had no objects.
3996 Frequency : always for liblustre
3998 Description: liblustre client does MDS+OSTs setattr RPC for each write
3999 Details : When doing a write from a liblustre client, the client
4000 incorrectly issued an RPC to the MDS and each OST the file was
4001 striped over in order to update the timestamps. When writing
4002 with small chunks and many clients this could overwhelm the MDS
4003 with RPCs. In all cases it would slow down the write because
4004 these RPCs are unnecessary.
4006 Severity : enhancement
4008 Description: allow number of MDS service threads to be changed at module load
4009 Details : It is now possible to change the number of MDS service threads
4010 running. Adding "options mds mds_num_threads={N}" to the MDS's
4011 /etc/modprobe.conf will set the number of threads for the next
4012 time Lustre is restarted (assuming the "mds" module is also
4013 reloaded at that time). The default number of threads will
4014 stay the same, 32 for most systems.
4019 Description: OST crash if filesystem is unformatted or corrupt
4020 Details : If an OST is started on a device that has never been formatted
4021 or if the filesystem is corrupt and cannot even mount then the
4022 error handling cleanup routines would dereference a NULL pointer.
4027 Description: NULL pointer deref in llap_from_page.
4028 Details : get_cache_page_nowait can return a page with NULL (or otherwise
4029 incorrect) mapping if the page was truncated/reclaimed while it was
4030 searched for. Check for this condition and skip such pages when
4031 doing readahead. Introduce extra check to llap_from_page() to
4032 verify page->mapping->host is non-NULL (so page is not anonymous).
4035 Frequency : Sometimes when using sys_sendfile
4037 Description: "page not covered by a lock" warnings from ll_readpage
4038 Details : sendfile called ll_readpage without right page locks present.
4039 Now we introduced ll_file_sendfile that does necessary locking
4040 around call to generic_file_sendfile() much like we do in
4044 Frequency : with certain MDS communication failures at client mount time
4046 Description: NULL pointer deref after failed client mount
4047 Details : a client connection request may delayed by the network layer
4048 and not be sent until after the PTLRPC layer has timed out the
4049 request. If the client fails the mount immediately it will try
4050 to clean up before the network times out the request. Add a
4051 reference from the request import to the obd device and delay
4052 the cleanup until the network drops the request.
4055 Frequency : occasionally during client (re)connect
4057 Description: assertion failure during client (re)connect
4058 Details : processing a client connection request may be delayed by the
4059 client or server longer than the client connect timeout. This
4060 causes the client to resend the connection request. If the
4061 original connection request is replied in this interval, the
4062 client may trip an assertion failure in ptlrpc_connect_interpret()
4063 which thought it would be the only running connect process.
4066 Frequency : only with obd_echo servers and clients that are rebooted
4068 Description: kernel BUG accessing uninitialized data structure
4069 Details : When running an obd_echo server it did not start the ping_evictor
4070 thread, and when a client was evicted an uninitialized data
4071 structure was accessed. Start the ping_evictor in the RPC
4072 service startup instead of the OBD startup.
4074 Severity : enhancement
4075 Bugzilla : 10193 (patchless)
4076 Description: Remove dependency on various unexported kernel interfaces.
4077 Details : No longer need reparent_to_init, exit_mm, exit_files,
4078 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
4081 Frequency : rare (only users of deprecated and unsupported LDAP config)
4083 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
4084 Details : LDAP apparently contains 'lustreName' attributes instead of
4085 'name'. A simple remapping of the name is sufficient.
4088 Frequency : rare (only with non-default dump_on_timeout debug enabled)
4090 Description: waiting_locks_callback trips kernel BUG if client is evicted
4091 Details : Running with the dump_on_timeout debug flag turned on makes
4092 it possible that the waiting_locks_callback() can try to dump
4093 the Lustre kernel debug logs from an interrupt handler. Defer
4094 this log dumping to the expired_lock_main() thread.
4096 Severity : enhancement
4098 Description: Support NFS exporting on 2.6 kernels.
4099 Details : Implement non-rawops metadata methods for NFS server to use without
4100 changing NFS server code.
4103 Frequency : very rare (synthetic metadata workload only)
4105 Description: two racing renames might cause an MDS thread to deadlock
4106 Details : Running the "racer" program may cause one MDS thread to rename
4107 a file from being the source of a rename to being the target of
4108 a rename at exactly the same time that another thread is doing
4109 so, and the second thread has already enqueued these locks after
4110 doing a lookup of the target and is trying to relock them in
4111 order. Ensure that we don't try to re-lock the same resource.
4114 Frequency : only very large systems with liblustre clients
4116 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
4117 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
4118 all OSTs in parallel. This allows the eviction of stale liblustre
4119 clients to proceed much faster than if they were done in series,
4120 and also offers similar improvements for other set_info RPCs.
4125 Description: excessive CPU usage during initial read phase on client
4126 Details : During the initial read phase on a client, it would agressively
4127 retry readahead on the file, consuming too much CPU and impacting
4128 performance (since 1.4.5.8). Improve the readahead algorithm
4129 to avoid this, and also improve some other common cases (read
4130 of small files in particular, where "small" is files smaller than
4131 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
4136 Description: MDS crash when receiving packet with unknown intent.
4137 Details : Do not LBUG in unknown intent case, just return -EFAULT
4139 Severity : enhancement
4140 Bugzilla : 9293, 9385
4141 Description: MDS RPCs are serialised on client. This is unnecessary for some.
4142 Details : Do not serialize getattr (non-intent version) and statfs.
4145 Frequency : occasional, when OST network is overloaded/intermittent
4147 Description: client evicted by OST after bulk IO timeout
4148 Details : If a client sends a bulk IO request (read or write) the OST
4149 may evict the client if it is unresposive to its data GET/PUT
4150 request. This is incorrect if the network is overloaded (takes
4151 too long to transfer the RPC data) or dropped the OST GET/PUT
4152 request. There is no need to evict the client at all, since
4153 the pinger and/or lock callbacks will handle this, and the
4154 client can restart the bulk request.
4157 Frequency : Always when mmapping file with no objects
4159 Description: client crashes when mmapping file with no objects
4160 Details : Check that we actually have objects in a file before doing any
4161 operations on objects in ll_vm_open, ll_vm_close and
4167 Description: Request leak when working with deleted CWD
4168 Details : Introduce advanced request refcount tracking for requests
4169 referenced from lustre intent.
4171 Severity : Enhancement
4173 Description: Cache open file handles on client.
4174 Details : MDS now will return special lock along with openhandle, if
4175 requested and client is allowed to hold openhandle, even if unused,
4176 until such a lock is revoked. Helps NFS a lot, since NFS is opening
4177 closing files for every read/write openration.
4179 Severity : Enhancement
4181 Description: Cache open negative dentries on client when possible.
4182 Details : Guard negative dentries with UPDATE lock on parent dir, drop
4183 negative dentries on lock revocation.
4188 Description: Remounting a client read-only wasn't possible with a zconf mount
4189 Details : It wasn't possible to remount a client read-only with llmount.
4191 Severity : enhancement
4192 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
4193 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
4194 MPICH is included to add Lustre-specific ADIO interfaces.
4195 This is based closely on the UFS ADIO layer and only differs
4196 in file creation, in order to allow the OST striping to be set.
4197 This is user-contributed code and not supported by CFS.
4202 Description: extended inode attributes (immutable, append-only) work improperly
4203 when 2.4 and 2.6 kernels are used on client/server or vice versa
4204 Details : Introduce kernel-independent values for these flags.
4206 Severity : enhancement
4209 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
4210 Details : Many of the /proc/ tunables can only be tuned at a megabyte
4211 granularity. Now, Fractional MB granularity is be supported,
4212 this is very useful for low memory system.
4214 Severity : enhancement
4216 Description: Getattr by fid
4217 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
4218 locks, avoids extra getattr rpc requests to MDS, allows '/' to
4219 have locks and avoids getattr rpc requests for it on every stat.
4222 Frequency : Always, for filesystems larger than 2TB
4224 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4225 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
4226 filesystems larger than 2TB to be mounted with mballoc enabled.
4229 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
4231 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4232 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
4233 filesystems larger than 2TB is mounted then the kernel will
4234 silently corrupt the start of the filesystem. CONFIG_LBD is
4235 enabled for all CFS-supported kernels, but the possibility of
4236 this happening with a modified kernel config exists.
4238 Severity : enhancement
4240 Description: add client O_DIRECT support for 2.6 kernels
4241 Details : It is now possible to do O_DIRECT reads and writes to files
4242 in the Lustre client mountpoint on 2.6 kernel clients.
4244 Severity : enhancement
4246 Description: parallel glimpse, setattr, statfs, punch, destroy requests
4247 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
4248 parallel, not waiting for response from every OST before sending
4249 a rpc to the next OST.
4254 Description: setattr vs write race when updating file timestamps
4255 Details : Client processes that update a file timestamp into the past
4256 right after writing to the file (e.g. tar) it is possible that
4257 the updated file modification time can be reset to the current
4258 time due to a race between processing the setattr and write RPC.
4260 Severity : enhancement
4262 Description: Bring 'lfs find' closer in line with regular Linux find.
4263 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
4264 -print0 options and obtains all the needed info through the lustre
4267 Severity : enhancement
4269 Description: support up to 1024 configured devices on one node
4270 Details : change obd_dev array from statically allocated to dynamically
4271 allocated structs as they are first used to reduce memory usage
4276 Description: Flush dirty partially truncated pages during truncate
4277 Details : Immediatelly flush partially truncated pages in filter_setattr,
4278 this way we completely avoid having any pages in page cache on OST
4279 and can retire ugly workarounds during writes to flush such pages.
4284 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
4285 Details : protect inode from truncation within vfs_unlink() context
4286 just take a reference before calling vfs_unlink() and release it
4287 when parent's i_sem is free.
4292 Description: last_id value checked outside lock on OST caused LASSERT failure
4293 Details : If there were multiple MDS->OST object precreate requests in
4294 flight, it was possible that the OST's last object id was checked
4295 outside a lock and incorrectly tripped an assertion. Move checks
4296 inside locks, and discard old precreate requests.
4299 Frequency : always, if extents are used on OSTs
4301 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
4302 Details : OSTs using the extents format would not zero the high 16 bits of
4303 the index physical block number. This is not a problem for any
4304 OST filesystems smaller than 16TB, and no kernels support ext3
4305 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
4306 new/modified files) and can be fixed for existing filesystems
4307 with e2fsprogs-1.39-cfs1.
4312 Description: import connection selection may be incorrect if timer wraps
4313 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
4314 import connections to be ignored if the 32-bit jiffies counter
4315 wraps. Use a 64-bit jiffies counter.
4318 Frequency : very large clusters immediately after boot
4320 Description: LNET request buffers exhausted under heavy short-term load
4321 Details : If a large number of client requests are generated on a service
4322 that has previously never seen so many requests it is possible
4323 that the request buffer growth cannot keep up with the spike in
4324 demand. Instead of dropping incoming requests, they are held in
4325 the LND until the RPC service can accept more requests.
4328 Frequency : Sometimes during replay
4330 Description: Assertion failure in ll_local_open after replay.
4331 Details : If replay happened on an open request reply before we were able
4332 to set replay handler, reply will become not swabbed tripping the
4333 assertion in ll_local_open. Now we set the handler right after
4334 recognising of open request
4337 Frequency : very rare
4339 Description: kernel reports "badness in vsnprintf"
4340 Details : Reading from the "recovery_status" /proc file in small chunks
4341 may cause a negative length in lprocfs_obd_rd_recovery_status()
4342 call to vsnprintf() (which is otherwise harmless). Exit early
4343 if there is no more space in the output buffer.
4345 Severity : enhancement
4347 Description: clear OBD RPC statistics by writing to them
4348 Details : It is now possible to clear the OBD RPC statistics by writing
4349 to the "stats" file.
4354 Description: Client mtime is not the same on different clients after utimes
4355 Details : In some cases, the client was using the utimes() syscall on
4356 a file cached on another node. The clients now validate the
4357 ctime from the MDS + OSTs to determine which one is right.
4362 Description: Inability to activate failout mode
4363 Details : lconf script incorrectly assumed that in pythong string's numeric
4364 value is used in comparisons.
4367 Frequency : always with multiple stripes per file
4369 Description: Inefficient object allocation for mutli-stripe files
4370 Details : When selecting which OSTs to stripe files over, for files with
4371 a stripe count that divides evenly into the number of OSTs,
4372 the MDS is always picking the same starting OST for each file.
4373 Return the OST selection heuristic to the original design.
4378 Description: mount failures may take full timeout to return an error
4379 Details : Under some heavy load conditions it is possible that a
4380 failed mount can wait for the full obd_timeout interval,
4381 possibly several minutes, before reporting an error.
4382 Instead return an error as soon as the status is known.
4384 Frequency : quota enabled and large files being deleted
4386 Description: releasing more than 4GB of quota at once hangs OST
4387 Details : If a user deletes more than 4GB of files on a single OST it
4388 will cause the OST to spin in an infinite loop. Release
4389 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4394 Description: statfs data retrieved from /proc may be stale or zero
4395 Details : When reading per-device statfs data from /proc, in the
4396 {kbytes,files}_{total,free,avail} files, it may appear
4397 as zero or be out of date.
4400 Frequency : systems with MD RAID1 external journal devices
4402 Description: lconf's call to blkid is confused by RAID1 journal devices
4403 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4404 of returning all block devices that match the journal UUID.
4407 Frequency : always, for aggregate stripe size over 4GB
4409 Description: assertion fails when trying to use 4GB stripe size
4410 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
4411 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
4414 Frequency : always on ppc64
4416 Description: the first write on an ext3 filesystem with mballoc got stuck
4417 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4418 perform endianness conversion.
4420 ------------------------------------------------------------------------------
4422 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4424 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
4425 INTEROPERATE with older versions automatically. Please read the
4426 user documentation before upgrading any part of a live system.
4427 * WARNING: Lustre networking configuration changes are required with
4428 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
4431 * Support for kernels:
4432 2.6.9-22.0.2.EL (RHEL 4)
4433 2.6.5-7.244 (SLES 9)
4434 2.6.12.6 vanilla (kernel.org)
4437 Severity : enhancement
4438 Bugzilla : 7981/8208
4439 Description: Introduced Lustre Networking (LNET)
4440 Details : LNET is new networking infrastructure for Lustre, it includes
4441 a reorganized network configuration mode (see the user
4442 documentation for full details) as well as support for routing
4443 between different network fabrics. Lustre Networking Devices
4444 (LNDs) for the supported network fabrics have also been
4445 created for this new infrastructure.
4447 Severity : enhancement
4448 Description: Introduced Access control lists
4449 Details : clients can set ACLs on files and directories in order to have
4450 more fine-grained permissions than the standard Unix UGO+RWX.
4451 The MDS must be started with the "-o acl" mount option.
4453 Severity : enhancement
4454 Description: Introduced filesystem quotas
4455 Details : Administrators may now establish per-user quotas on the
4458 Severity : enhancement
4460 Description: Configuration change for the XT3
4461 The PTLLND is now used to run Lustre over Portals on the XT3
4462 The configure option(s) --with-cray-portals are no longer used.
4463 Rather --with-portals=<path-to-portals-includes> is used to
4464 enable building on the XT3. In addition to enable XT3 specific
4465 features the option --enable-cray-xt3 must be used.
4470 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
4471 Details : A race between allocating a new llcd and re-getting the llcd_lock
4472 allowed another thread to grab newly-allocated llcd.
4474 Severity : enhancement
4476 Description: 2.6 OST async journal commit and locking fix to improve performance
4477 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
4478 2.6 kernels are now async as they already were in 2.4 kernels so
4479 that they can commit concurrently with the network bulk transfer.
4480 For block-allocated files the filter allocation semaphore is held
4481 to avoid filesystem fragmentation during allocation. BKL lock
4482 removed for 2.6 xattr operations where it is no longer needed.
4487 Description: lconf incorrectly determined whether two IP networks could talk
4488 Details : In some more complicated routing and multiple-network
4489 configurations, lconf will avoid trying to make a network
4490 connection to a disjoint part of the IP space. It was doing the
4491 math incorrectly for one set of cases.
4496 Description: Fix for potential infinite loop processing records in an llog.
4497 Details : If an llog record is corrupted/zeroed, it is possible to loop
4498 forever in llog_process(). Validate the llog record length
4499 and skip the remainder of the block on error.
4502 Frequency : occasional (liblustre only)
4504 Description: liblustre could not open files whose last component is a symlink
4505 Details : sysio_path_walk() would incorrectly pass the open intent to
4506 intermediate path components.
4509 Frequency : rare (liblustre only with non-standard tuning)
4510 Bugzilla : 7201 (7350)
4511 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4512 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4513 after it has been released, only do this for referenced locks
4515 Severity : enhancement
4517 Description: specifying an (invalid) directory default stripe_size of -1
4518 would reset the directory default striping
4519 Details : stripe_size -1 was used internally to signal directory stripe
4520 removal, now use "all default" to signal dir stripe removal
4521 as a directory striping of "all default" is not useful
4524 Frequency : common for large clusters running liblustre clients
4526 Description: doing an ls when liblustre clients are running is slow
4527 Details : sending a glimpse AST to a liblustre client waits for every AST
4528 to time out, as liblustre clients will not respond. Since they
4529 cannot cache data we refresh the OST lock LVB from disk instead.
4531 Severity : enhancement
4533 Description: doing an ls at the same time as file IO can be slow
4534 Details : enqueue and other "small" requests can be blocked behind many
4535 large IO requests. Create a new OST IO portal for non-IO
4536 requests so they can be processed faster.
4539 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4541 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4542 created file with mode 000
4543 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4544 allow an ftruncate() as a non-root user to a file with mode 000.
4545 Lustre now respects this flag to disable mode checks when
4546 truncating a file owned by the user
4549 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4552 Description: Revoking locks from clients that went dead or catatonic might take
4554 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4555 cancellation of such locks instant on servers without waiting for
4556 any reply from clients. Clients drops these locks when cancel
4557 notification from server is received without replying.
4560 Frequency : liblustre-only, when liblustre client dies or becomes busy
4562 Description: Doing ls on Linux clients can take a long time with active
4564 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4565 granting such locks to it in the first place if possible. Locks
4566 are taken by proxy on the OST during the read or write and
4567 dropped immediately afterward. Add connect flags handling, do
4568 not grant locks to liblustre clients for glimpse ASTs.
4570 Severity : enhancement
4572 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4573 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4574 sequential reads and behaves badly with applications doing random
4575 reads. Improve it by reading ahead at least read region, and
4576 avoiding excessive large RPC for small reads.
4578 Severity : enhancement
4580 Description: Creating more than 1000 files for a single job may cause a load
4581 imbalance on the OSTs if there are also a large number of OSTs.
4582 Details : qos_prep_create() uses an OST index reseed value that is an
4583 even multiple of the number of available OSTs so that if the
4584 reseed happens in the middle of the object allocation it will
4585 still utilize the OSTs as uniformly as possible.
4590 Description: OST or MDS may oops in ping_evictor_main()
4591 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
4592 but doesn't restart at beginning of obd_exports_timed list
4595 Severity : enhancement
4597 Description: improve by-nid export eviction on the MDS and OST
4598 Details : allow multiple exports with the same NID to be evicted at one
4599 time without re-searching the exports list.
4602 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
4604 Description: MDS may oops in groups_free()
4605 Details : in rare race conditions a newly allocated group_info struct is
4606 freed again, and this can be NULL. The 2.4 compatibility code
4607 for groups_free() checked for a NULL pointer, but 2.6 did not.
4610 Frequency : common for liblustre clients doing little filesystem IO
4611 Bugzilla : 9352, 7313
4612 Description: server may evict liblustre clients accessing contended locks
4613 Details : if a client is granted a lock or receives a completion AST
4614 with a blocking AST already set it would not reply to the AST
4615 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
4618 Frequency : lfs setstripe, only systems with more than 160 OSTs
4620 Description: unable to set striping with a starting offset beyond OST 160
4621 Details : llapi_create_file() incorrectly limited the starting stripe
4622 index to the maximum single-file stripe count.
4625 Frequency : LDAP users only
4627 Description: lconf did not handle in-kernel recovery with LDAP properly
4628 Details : lconf/LustreDB get_refs() is searching the wrong namespace
4630 Severity : enhancement
4632 Description: bind OST threads to NUMA nodes to improve performance
4633 Details : all OST threads are uniformly bound to CPUs on a single NUMA
4634 node and do their allocations there to localize memory access
4636 Severity : enhancement
4638 Description: llmount can determine client NID directly from Myrinet (GM)
4639 Details : the client NID code from gmnalnid was moved directly into
4640 llmount, removing the need to use this or specifying the
4641 client NID explicitly when mounting GM clients with zeroconf
4644 Frequency : if client is started with down MDS
4646 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
4647 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
4648 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
4649 ptlrpc_import_delayed_req() because we didn't check if the
4650 request was interrupted, and we also didn't break out of the
4651 event loop if there was no timeout
4656 Description: data loss during non-page-aligned writes to a single file from
4657 both multiple nodes and multiple threads on one node at same time
4658 Details : updates to KMS and lsm weren't protected by common lock. Resulting
4659 inconsistency led to false short-reads, that were cached and later
4660 used by ->prepare_write() to fill in partially written page,
4661 leading to data loss.
4664 Frequency : always, if lconf --abort_recovery used
4666 Description: lconf --abort_recovery fails with 'Operation not supported'
4667 Details : lconf was attempting to abort recovery on the MDT device and not
4670 Severity : enhancement
4672 Description: remove cleanup logs
4673 Details : replace lconf-generated cleanup logs with lustre internal
4674 cleanup routines. Eliminates the need for client-cleanup and
4677 Severity : enhancement
4679 Description: add support for EAs (user and system) on lustre filesystems
4680 Details : it is now possible to store extended attributes in the Lustre
4681 client filesystem, and with the user_xattr mount option it
4682 is possible to allow users to store EAs on their files also
4684 Severity : enhancement
4686 Description: Add possibility (config option) to show minimal available OST free
4688 Details : When compiled with --enable-mindf configure option, statfs(2)
4689 (and so, df) will return least minimal free space available from
4690 all OSTs as amount of free space on FS, instead of summary of
4691 free spaces of all OSTs.
4693 Severity : enhancement
4695 Description: do not expand extent locks acquired on OST-side
4696 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
4697 by server: they are not cached anyway.
4700 Frequency : when mmap is used/binaries executed from Lustre
4702 Description: Unmmap pages before throwing them away from read cache.
4703 Details : llap_shrink cache now attempts to unmap pages before discarding
4704 them (if unmapping failed - do not discard). SLES9 kernel has
4705 extra checks that trigger if this unmapping is not done first.
4710 Description: lconf didn't resolve symlinks before checking to see whether a
4711 given mountpoint was already in use
4714 Frequency : when migrating failover services
4715 Bugzilla : 6395, 9514
4716 Description: When migrating a subset of services from a node (e.g. failback
4717 from a failover service node) the remaining services would
4718 time out and evict clients.
4719 Details : lconf --force (implied by --failover) sets the global obd_timeout
4720 to 5 seconds in order to quickly disconnect, but this caused
4721 other RPCs to time out too quickly. Do not change the global
4722 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
4724 Severity : enhancement
4725 Frequency : if MDS is started with down OST
4726 Bugzilla : 9439,5706
4727 Description: Allow startup/shutdown of an MDS without depending on the
4728 availability of the OSTs.
4729 Details : Asynchronously call mds_lov_synchronize during MDS startup.
4730 Add appropriate locking and lov-osc refcounts for safe
4731 cleaning. Add osc abort_inflight calls in case the
4732 synchronize never started.
4735 Frequency : occasional (Cray XT3 only)
4737 Description: root not authorized to access files in CRAY_PORTALS environment
4738 Details : The client process capabilities were not honoured on the MDS in
4739 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
4740 been accessed by an authorized user then root was able to access
4741 the file on the local client also. The root user capabilities
4742 are now allowed on the MDS, as this environment has secure UID.
4745 Frequency : occasional
4747 Description: ldiskfs "too long searching" message happens too often
4748 Details : A debugging message (otherwise harmless) prints too often on
4749 the OST console. This has been reduced to only happen when
4750 there are fragmentation problems on the filesystem.
4755 Description: Division by zero in statfs when all OSCs are inactive
4756 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
4757 lov_statfs divides by value returned by lov_get_stripecnt().
4761 Bugzilla : 9489, 3273
4762 Description: First write from each client to each OST was only 4kB in size,
4763 to initialize client writeback cache, which caused sub-optimal
4764 RPCs and poor layout on disk for the first written file.
4765 Details : Clients now request an initial cache grant at (re)connect time
4766 and so that they can start streaming writes to the cache right
4767 away and always do full-sized RPCs if there is enough data.
4768 If the OST is rebooted the client also re-establishes its grant
4769 so that client cached writes will be honoured under the grant.
4774 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
4775 Details : Now I/O RPCs go to different portal number and (presumably) fast
4776 lock requests (and glimses) and other RPCs get their own service
4777 threads pool that should be able to service those RPCs
4780 Severity : enhancement
4782 Description: Ability to exchange lustre version between client and servers and
4783 issue warnings at client side if client is too old. Also for
4784 liblustre clients there is ability to refuse connection of too old
4786 Details : New 'version' field is added to connect data structure that is
4787 filled with version info. That info is later checked by server and
4791 Frequency : rare, liblustre only.
4792 Bugzilla : 9296, 9581
4793 Description: Two simultaneous writes from liblustre at offset within same page
4794 might proceed at the same time overwriting eachother with stale
4796 Details : I/O lock withing llu_file_prwv was released too early, before data
4797 actually was hitting the wire. Extended lock-holding time until
4798 server acknowledges receiving data.
4801 Frequency : extremely rare. Never observed in practice.
4803 Description: avoid generating lustre_handle cookie of 0.
4804 Details : class_handle_hash() generates handle cookies by incrementing
4805 global counter, and can hit 0 occasionaly (this is unlikely, but
4806 not impossible, because initial value of cookie counter is
4807 selected randonly). Value of 0 is used as a sentinel meaning
4808 "unassigned handle" --- avoid it. Also coalesce two critical
4809 sections in this function into one.
4811 Severity : enhancement
4813 Description: allow liblustre clients to delegate truncate locking to OST
4814 Details : To avoid overhead of locking, liblustre client instructs OST to
4815 take extent lock in ost_punch() on client's behalf. New connection
4816 flag is added to handle backward compatibility.
4818 Severity : enhancement
4819 Bugzilla : 4928, 7341, 9758
4820 Description: allow number of OST service threads to be specified
4821 Details : a module parameter allows the number of OST service threads
4822 to be specified via "options ost ost_num_threads={N}" in the
4823 OSS's /etc/modules.conf or /etc/modprobe.conf.
4827 Bugzilla : 6146, 9635, 9895
4828 Description: servers crash with bad pointer in target_handle_connect()
4829 Details : In rare cases when a client is reconnecting it was possible that
4830 the connection request was the last reference for that export.
4831 We would temporarily drop the export reference and get a new
4832 one, but this may have been the last reference and the export
4833 was just destroyed. Get new reference before dropping old one.
4835 Severity : enhancement
4836 Frequency : if client is started with failover MDS
4838 Description: Allow multiple MDS hostnames in the mount command
4839 Details : Try to read the configuration from all specified MDS
4840 hostnames during a client mount in case the "primary"
4843 Severity : enhancement
4845 Description: Stop sending data to evicted clients as soon as possible.
4846 Details : Check if the client we are about to send or are sending data to
4847 was evicted already. (Check is done every second of waiting,
4848 for which l_wait_event interface was extended to allow checking
4849 of exit condition at specified intervals).
4852 Frequency : rare, normally only when NFS exporting is done from client
4854 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
4856 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
4859 Frequency : always, if srand() is called before liblustre initialization
4861 Description: Liblustre uses system PRNG disturbing its usage by user application
4862 Details : Introduce internal to lustre fast and high-quality PRNG for
4863 lustre usage and make liblustre and some other places in generic
4864 lustre code to use it.
4866 Severity : enhancement
4867 Bugzilla : 9477, 9557, 9870
4868 Description: Verify that the MDS configuration logs are updated when xml is
4869 Details : Check if the .xml configuration logs are newer than the config
4870 logs stored on the MDS and report an error if this is the case.
4871 Request --write-conf, or allow starting with --old_conf.
4873 Severity : enhancement
4875 Description: Handle symlinks in the path when checking if Lustre is mounted.
4876 Details : Resolve intermediate symlinks when checking if a client has
4877 mounted a filesystem to avoid duplicate client mounts.
4882 Description: lconf can hit an error exception but still return success.
4883 Details : The lconf command catches the Command error exception at the top
4884 level script context and will exit with the associated exit
4885 status, but doesn't ensure that this exit status is non-zero.
4890 Description: failure of ptlrpc thread startup can cause oops
4891 Details : Starting a ptlrpc service thread can fail if there are a large
4892 number of threads or the server memory is very fragmented.
4893 Handle this without oopsing.
4896 Frequency : always, only if liblustre and non-default acceptor port was used
4898 Description: liblustre cannot connect to servers with non-default acceptor port
4899 Details : tcpnal_set_default_params() was not called and was therefore
4900 ignoring the environment varaible TCPNAL_PORT, as well as other
4901 TCPNAL_ environment variables
4906 Description: two objects could be created on the same OST for a single file
4907 Details : If an OST is down, in some cases it was possible to create two
4908 objects on a single OST for a single file. No problems other
4909 than potential performance impact and spurious error messages.
4913 Bugzilla : 5681, 9562
4914 Description: Client may oops in ll_unhash_aliases
4915 Details : Client dcache may become inconsistent in race condition.
4916 In some cases "getcwd" can fail if the current directory is
4922 Description: Inode refcounting problems in NFS export code
4923 Details : link_raw functions used to call d_instantiate without obtaining
4924 extra inode reference first.
4928 Bugzilla : 9942, 9903
4929 Description: Referencing freed requests leading to crash, memleaks with NFS.
4930 Details : We used to require that call to ll_revalidate_it was always
4931 followed by ll_lookup_it. Also with revalidate_special() it is
4932 possible to call ll_revalidate_it() twice for the same dentry
4933 even if first occurence returned success. This fix changes semantic
4934 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
4935 reference on a request referred from the intent.
4936 ll_intent_release() then releases such a request.
4939 Frequency : rare, normally benchmark loads only
4941 Description: unlinked inodes were kept in memory on the client
4942 Details : If a client is repeatedly creating and unlinking files it
4943 can accumulate a lot of stale inodes in the inode slab cache.
4944 If there is no other client load running this can cause the
4945 client node to run out of memory. Instead flush old inodes
4946 from client cache that have the same inode number as a new inode.
4949 Frequency : SLES9 2.6.5 kernel and long filenames only
4950 Bugzilla : 9969, 10379
4951 Description: utime reports stale NFS file handle
4952 Details : SLES9 uses out-of-dentry names in some cases, which confused
4953 the lustre dentry revalidation. Change it to always use the
4957 Frequency : rare, unless heavy write-truncate concurrency is continuous
4958 Bugzilla : 4180, 6984, 7171, 9963, 9331
4959 Description: OST becomes very slow and/or deadlocked during object unlink
4960 Details : filter_destroy() was holding onto the parent directory lock
4961 while truncating+unlinking objects. For very large objects this
4962 may block other threads for a long time and slow overall OST
4963 responsiveness. It may also be possible to get a lock ordering
4964 deadlock in this case, or run out of journal credits because of
4965 the combined truncate+unlink. Solution is to do object truncate
4966 first in one transaction without parent lock, and then do the
4967 final unlink in a new transaction with the parent lock. This
4968 reduces the lock hold time dramatically.
4971 Frequency : rare, 2.4 kernels only
4973 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
4974 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
4975 before unmounting their filesystem in order to avoid deadlock.
4976 The kernel_locked() function in 2.4 kernels only checks whether
4977 the kernel lock is held, not whether it is this process that is
4978 holding it as 2.6 kernels do.
4983 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
4984 Details : The client ptlrpc code may be trying to reconnect to a down
4985 server before a previous connection attempt has timed out.
4986 Increase the reconnect interval to be longer than the connection
4987 timeout interval to avoid sending duplicate connections to
4991 Frequency : echo_client brw_test command
4993 Description: fix echo_client to work with OST preallocated code
4994 Details : OST preallocation code (5137) didn't take echo_client IO path
4995 into account: echo_client calls filter methods outside of any
4996 OST thread and, hence, there is no per-thread preallocated
4997 pages and buffers to use. Solution: hijack pga pages for IO. As
4998 a byproduct, this avoids unnecessary data copying.
5002 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
5003 Description: Client can oops in mdc_commit_close() after open replay
5004 Details : It was possible for the MDS to return an open request with no
5005 transaction number in mds_finish_transno() if the client was
5006 evicted, but without actually returning an error. Clients
5007 would later try to replay that open and may trip an assertion
5008 Simplify the client close codepath, and always return an error
5009 from the MDS in case the open is not successful.
5012 Frequency : rare, 2.6 OSTs only
5014 Description: OST may deadlock under high load on fragmented files
5015 Details : If there was a heavy load and highly-fragmented OST filesystems
5016 it was possible to have all the OST threads deadlock waiting on
5017 allocation of biovecs, because the biovecs were not released
5018 until the entire RPC IO was completed. Instead, release biovecs
5019 as soon as they are complete to ensure forward IO progress.
5021 Severity : enhancement
5023 Description: Support for specifying external journal device at mount
5024 Details : If an OST or MDS device is formatted with an external journal
5025 device, this device major/minor is stored in the ext3 superblock
5026 and may not be valid for failover. Allow detecting and
5027 specifying the external journal at mount time.
5032 Description: Mounting an MDS with pending unlinked files may cause oops
5033 Details : target_finish_recovery() calls mds_postrecov() which returned
5034 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
5035 considers this an error and immediately begins cleaning up the
5036 lov, just after starting the mds_lov process
5038 Severity : enhancement
5040 Description: Implement 'lfs df' to report actual free space on per-OST basis
5041 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
5042 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
5043 usage of objects; '-h' to report in human readable format.
5045 ------------------------------------------------------------------------------
5047 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
5054 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
5055 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
5056 at startup during orphan unlinking. Instead, skip buddy update
5057 before setup, as it will be regenerated after recovery is complete.
5062 Description: If an OST is inactive, its locks might reference stale inodes.
5063 Details : lov_change_cbdata() must iterate over all namespaces, even if
5064 they are inactive to clear inode references from the lock.
5066 Severity : enhancement
5067 Frequency : occasional, if non-standard max_dirty_mb used
5069 Description: Client will block write RPCs if not enough grant
5070 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
5071 then the client will block writes while waiting for another RPC
5072 to complete instead of consuming its dirty limit. With change
5073 we get improved performance when max_dirty_mb is small.
5075 Severity : enhancement
5076 Bugzilla : 3389, 6253
5077 Description: Add support for supplementary groups on the MDS.
5078 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
5079 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
5080 lookups for user supplementary groups into a cache.
5084 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
5085 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
5086 for liblustre O_EXCL is enforced by MDS.
5090 Description: readdir never returns NULL in liblustre.
5091 Details : Corrected llu_iop_getdirentries logic, to return offset of next
5092 dentry in struct dirent.
5096 Frequency : liblustre only, depends on application IO pattern
5097 Description: liblustre clients evicted if not contacting servers
5098 Details : Don't put liblustre clients into the ping_evictor list, so
5099 they will not be evicted by the pinger ever.
5101 Severity : enhancement
5103 Description: Add ability to evict clients by NID from MDS.
5104 Details : By echoing "nid:$NID" string into
5105 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
5106 $NID would be instantly evicted from this MDS and from all active
5107 OSTs connected to it.
5111 Description: Do not query file size twice, somewhat slowing stat(2) calls.
5112 Details : lookup_it_finish() used to query file size from OSTs that was not
5117 Description: service threads change working directory to that of init
5118 Details : Starting lustre service threads may pin the working directory
5119 of the parent thread, making that filesystem busy. Threads
5120 now change to the working directory of init to avoid this.
5124 Frequency : during shutdown only
5125 Description: shutdown with a failed MDS or OST can cause unmount to hang
5126 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
5131 Frequency : 2.6 only
5132 Description: chmod/chown may include an extra supplementary group
5133 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
5134 supplementary group and if none is specified this is used.
5137 Bugzilla : 5479 (6816)
5139 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
5140 Details : If lookup is in progress on a file that is unlinked we might try
5141 to revalidate the inode and fail in revalidate after lookup is
5142 complete and ll_file_open() enqueues the open again but
5143 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
5146 Frequency : always, if lconf --abort_recovery used
5148 Description: lconf --abort_recovery fails with 'Operation not supported'
5149 Details : lconf was attempting to abort recovery on the MDT device and not
5152 ------------------------------------------------------------------------------
5154 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
5159 Frequency : rare (only unsupported configurations with a node running as an
5161 Bugzilla : 6514, 5137
5162 Description: Mounting a Lustre file system on a node running as an OST could
5164 Details : OSTs now preallocates memory needed to write out data at
5165 startup, instead of when needed, to avoid having to
5166 allocate memory in possibly low memory situations.
5167 Specifically, if the file system is mounted on on OST,
5168 memory pressure could force it to try to write out data,
5169 which it needed to allocate memory to do. Due to the low
5170 memory, it would be unable to do so and the node would
5171 become unresponsive.
5173 Severity : enhancement
5175 Description: Addition of lconf --service command line option
5176 Details : lconf now accepts a '--service <arg>' option, which is
5177 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
5179 Severity : enhancement
5181 Description: Failover mode is now the default for OSTs.
5182 Details : By default, OSTs will now run in failover mode. To return to
5183 the old behaviour, add '--failout' to the lmc line for OSTs.
5185 Severity : enhancement
5187 Description: Health checks are now provided for MDS and OSTs
5188 Details : Additional detailed health check information on MSD and OSTs
5189 is now provided through the procfs health_check value.
5192 Frequency : occasional, depends on IO load
5194 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
5195 after numerous create/delete cycles
5196 Details : The ext3 inode allocation policy would not allocate new inodes
5197 very well on the OSTs because there are no new directories
5198 being created. Instead we look for groups with free space if
5199 the parent directories are nearly full.
5204 Description: Network or server problems during mount may cause partially
5205 mounted clients instead of returning an error.
5206 Details : The config llog parsing code may overwrite the error return
5207 code during mount error handling, returning success instead
5213 Description: MDS can fail to allocate large reply buffers
5214 Details : After long uptimes the MDS can fail to allocate large reply
5215 buffers (e.g. zconf client mount config records) due to memory
5216 fragmentation or consumption by the buffer cache. Preallocate
5217 some large reply buffers so that these replies can be sent even
5218 under memory pressure.
5222 Frequency : rare (liblustre)
5223 Description: fsx running with liblustre complained that using truncate() to
5224 extend the file doesn't work. This patch corrects that issue.
5225 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
5226 ATTR_SIZE and lsm use in llu_setattr_raw.
5230 Frequency : rare, only 2.6 kernels
5231 Description: Unusual file access patterns on the MDS may result in inode
5232 data being lost in very rare circumstances.
5233 Details : Bad interaction between the ea-in-inode patch and the "no-read"
5234 code in the 2.6 kernel caused the inode and/or EA data not to
5235 be read from disk, causing single-file corruption.
5239 Frequency : rare, only 2.6 filesystems using extents
5240 Description: Heavy concurrent write and delete load may cause data corruption.
5241 Details : It was possible under high-load situations to have an extent
5242 metadata block in the block device cache from a just-unlinked
5243 file overwrite a newly-allocated data block. We now unmap any
5244 metadata buffers that alias just-allocated data blocks.
5248 Frequency : filesystems with default stripe_count larger than 77
5249 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
5250 Details : lconf specifies an inode size of 4096 bytes when the default
5251 stripe_count is larger than 77. This conflicts with the default
5252 inode density of 1 per 4096 bytes. Allocate smaller inodes in
5253 this case to avoid pinning too much memory for large EAs.
5255 ------------------------------------------------------------------------------
5257 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
5262 Frequency : rare (extremely heavy IO load with hundreds of clients)
5264 Description: Client is evicted, gets IO error writing to file
5265 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
5266 caused clients to be evicted for AST timeouts. The fixes in
5267 bug 5192 mean we no longer need to have such short AST timeouts
5268 so ldlm_timeout has been increased.
5271 Frequency : occasional during --force or --failover shutdown under load
5272 Bugzilla : 5949, 4834
5273 Description: Server oops/LBUG if stopped with --force or --failover under load
5274 Details : a collection of import/export refcount and cleanup ordering
5275 issues fixed for safer force cleanup
5278 Frequency : only filesystems larger than 120 OSTs
5279 Bugzilla : 5990, 6223
5280 Description: lfs getstripe would oops on a very large filesystem
5281 Details : lov_getconfig used kfree on vmalloc'd memory
5284 Frequency : only filesystems exporting via NFS to Solaris 10 clients
5285 Bugzilla : 6242, 6243
5286 Description: reading from files that had been truncated to a non-zero size
5287 but never opened returned no data
5288 Details : ll_file_read() reads zeros from no-object files to EOF
5293 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
5294 Details : There's little harm in aborting MDS/OSS recovery and letting it
5295 try again, so I removed the LASSERT and return an error instead.
5297 Severity : enhancement
5299 Description: New debugging infrastructure for tracking down data corruption
5300 Details : The I/O checksum code was replaced to: (a) control it at runtime,
5301 (b) cover more of the client-side code path, and (c) try to narrow
5302 down where problems occurred
5306 Bugzilla : 3819, 4364, 4397, 6313
5307 Description: Racing close and eviction MDS could cause assertion in mds_close
5308 Details : It was possible to get multiple mfd references during close and
5309 client eviction, leading to one thread referencing a freed mfd.
5311 Severity: : enhancement
5312 Bugzilla : 3262, 6359
5313 Description: Attempts to reconnect to servers are now more aggressive.
5314 Details : This builds on the enhanced upcall-less recovery that was added
5315 in 1.4.2. When trying to reconnect to servers, clients will
5316 now try each server in the failover group every 10 seconds. By
5317 default, clients would previously try one server every 25 seconds.
5322 Description: After recovery, certain operations trigger a failed
5323 assertion on a client.
5324 Details : Failing over an mds, using lconf -d --failover, while a
5325 client was doing a readdir() call would cause the client to
5326 LBUG after recovery completed and the readdir() was resent.
5328 Severity : enhancement
5330 Description: Default groups are now added by lconf
5331 Details : You can now run lconf --group <servicename> without having to
5332 manually add groups with lmc.
5335 Frequency : occasional
5337 Description: Nodes with an elan id of 0 trigger a failed assertion
5340 Frequency : always when accessing e.g. tty/console device nodes
5342 Description: tty and some other devices nodes cannot be used on lustre
5343 Details : file's private_data field is used by device data and lustre
5344 values in there got lost. New field was added to struct file to
5345 store fs-specific private data.
5348 Frequency : when exporting Lustre via NFS
5350 Description: NFSD failed occasionally when looking up a path component
5351 Details : NFSD is looking up ".." which was broken in ext3 directories
5352 that had grown large enough to become hashed.
5355 Frequency : Clusters with multiple interfaces not on the same subnet
5357 Description: Nodes will repeatedly try to reconnect to an interface which it
5358 cannot reach and report an error to the log.
5359 Details : Extra peer list entries will be created by lconf with some peers
5360 unreachable. lconf now validates the peer before adding it.
5363 Frequency : Only if a default stripe is set on the filesystem root.
5365 Description: Setting a default stripe on the filesystem root prevented the
5366 filesystem from being remounted.
5367 Details : The client was sending extra request flags in the root getattr
5368 request and did not allocate a reply buffer for the dir EA.
5371 Frequency : occasional, higher if lots of files are accessed by one client
5372 Bugzilla : 6159, 6097
5373 Description: Client trips assertion regarding lsm mismatch/magic
5374 Details : While revalidating inodes the VFS looks up inodes with ifind()
5375 and in rare cases can find an inode that is being freed.
5376 The ll_test_inode() code will free the lsm during ifind()
5377 when it finds an existing inode and then the VFS later attaches
5378 this free lsm to a new inode.
5382 Bugzilla : 6422, 7030
5383 Description: MDS deadlock between mkdir and client eviction
5384 Details : Creating a new file via mkdir or mknod (starting a transaction
5385 and getting the ns lock) can deadlock with client eviction
5386 (gets ns lock and trying to finish a synchronous transaction).
5389 Frequency : occasional
5390 Description: While starting a server, the fsfilt_ext3 module could not be
5392 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
5393 kernels. Previously, lconf would still use the ext3 name
5394 when trying to load modules. Now, it will correctly use
5395 ext3 on 2.4 and ldiskfs on 2.6.
5397 Severity : enhancement
5398 Description: The default stripe count has been changed to 1
5399 Details : The interpretation of the default stripe count (0, to lfs
5400 or lmc) has been changed to mean striping across a single
5401 OST, rather than all available. For general usage we have
5402 found a stripe count of 1 or 2 works best.
5404 Severity : enhancement
5405 Description: Add support for compiling against Cray portals.
5406 Details : Conditional compiling for some areas that are different
5410 Frequency : occasional
5411 Bugzilla : 6409, 6834
5412 Description: Creating files with an explicit stripe count may lead to
5413 a failed assertion on the MDS
5414 Details : If some OSTs are full or unavailable, creating files may
5415 trigger a failed assertion on the MDS. Now, Lustre will
5416 try to use other servers or return an error to the
5420 Frequency : occasional
5422 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
5423 caused lock timeout during readahead (since 1.4.2).
5424 Details : Processes doing readahead might match a lock that hasn't been
5425 granted yet if there are overlapping and conflicting lock
5426 requests. The readahead process waits on ungranted lock
5427 (original lock is CBPENDING), while OST waits for that process
5428 to cancel CBPENDING read lock and eventually evicts client.
5430 Severity : enhancement
5432 Description: Initial enabling of flock support for clients
5433 Details : Implements fcntl advisory locking and file status functions.
5434 This feature is provided as an optional mount flag (default
5435 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
5436 locking are implemented yet, and those that are are not guaranteed
5437 to be completely correct in production environments.
5438 mount -t lustre -o [flock|noflock] ...
5441 Frequency : occasional
5443 Description: OSTs running 2.4 kernels but with extents enabled might trip an
5444 assertion in the ext3 JBD (journaling) layer.
5445 Details : The b_committed_data struct is protected by the big kernel lock
5446 in 2.4 kernels, serializing journal_commit_transaction() and
5447 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
5448 access to this struct. In 2.6 kernels there is finer grained
5449 locking to improve SMP performance of the JBD layer.
5453 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
5458 Description: merge and cleanup kernel patches.
5461 -----------------------------------------------------------------------------
5463 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
5465 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
5466 versions of Lustre. Please update all servers and clients to
5467 version 1.4.2 or later at the same time. You must also run
5468 "lconf --write-conf {config}.xml" on the MDS while it is stopped
5469 to update the configuration logs.
5471 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
5472 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
5473 - swab llog records of type '0' so we get proper header size/idx (5861)
5474 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
5475 - fix rename of one directory over another leaking an inode (5953)
5476 - avoid SetPageDirty on 2.6 (5981)
5477 - don't re-add just-being-destroyed locks to the waiting list (5653)
5478 - when creating new directories, inherit the parent's custom
5479 striping settings if present parent (3048)
5480 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
5481 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
5482 - don't hold client locks on temporary worklist from l_lru (5666)
5483 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
5484 - automatically evict dead clients (5921)
5485 - Update file size properly in create+truncate+fstat case (6196)
5486 - Do not unhash mountpoint dentries, do not allow removal of
5488 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
5489 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
5490 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
5491 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
5492 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
5493 - Correctly handle reads of files with no objects (6243)
5494 - lctl recover will also mark a device active if deactivate used (5933)
5496 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
5497 - allow --write-conf on an MDS with different nettype than client (5619)
5498 - don't write config llogs to MDS for mounts not from that MDS (5617)
5499 - lconf should create multiple TCP connections from a client (5201)
5500 - init scripts are now turned off by default; run chkconfig --on
5501 lustre and chkconfig --on lustrefs to use them
5502 - upcalls are no longer needed for clients to recover to failover
5504 - add --abort-recovery option to lconf to abort recovery on device
5506 - add support for an arbitrary number of OSTs (3026)
5507 - Quota support protocol changes.
5508 - forward compatibility changes to wire structs (6007)
5509 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5510 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5511 - improved "lustre status" script
5512 - initialize blocksize for non-regular files (6062)
5513 - added --disable-server and --disable-client configure options (5782)
5514 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5515 - Vanilla 2.4.29 support
5516 - increase maximum number of obd devices to 520 (6242)
5517 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5518 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5519 SLES 9 2.6.5 kernels
5520 - limit stripes per file to 160 (the maximum EA size) (6093)
5522 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5525 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5526 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5527 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5528 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5529 - small ext3 extents cleanups and fixes (5733)
5530 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5531 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5532 - avoid needless client->OST connect, fix handle mismatch (5317)
5533 - fix DLM error path that led to out-of-sync client, long delays (5779)
5534 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5535 - fix several locking issues related to i_size (5492,5624,5654,5672)
5536 - don't move pending lock onto export if it is already evicted (5683)
5537 - fix kernel oops when creating .foo in unlinked directory (5548)
5538 - fix deadlock in obdfilter statistics vs. object create (5811)
5539 - use time_{before,after} to avoid timer jiffies wrap (5882)
5540 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5541 - Do not leak request if server was not able to process it (5154)
5542 - If mds_open unable to find parent dir, make that negative lookup(5154)
5543 - don't create new directories with extent-mapping (5909, 5936)
5545 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5546 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5547 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5548 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5549 - support for RHEL4 kernel on i686 (5773)
5550 - provide error messages when incompatible logs are encountered (5898)
5552 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5553 * version 1.4.0.10 (1.4.1 release candidate 1)
5555 - don't keep a lock reference when lock is not granted (4238)
5556 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5557 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5558 - drop import inflight refcount on signal_completed_replay error (5255)
5559 - unlock page after async write error during send (3677)
5560 - handle missing objects in filter_preprw_read properly (5265)
5561 - no transno return for symlink open, don't save no-trasno open (3440)
5562 - don't try to complete elan receive that already failed (4012)
5563 - free RPC server reply state on error (5406)
5564 - clean up thread from ptlrpc_start_thread() on error (5160)
5565 - readahead could read extra page into cache that wasn't ejected (5388)
5566 - prevent races in class_attach/setup/cleanup/detach (5260)
5567 - don't dereference de->d_inode after l_dput of de (5458)
5568 - use "int" for stripe value returned from lock_to_stripe (5544)
5569 - mballoc allocation and error-checking fixes in 2.6 (5504)
5570 - block device patches to fix I/O request sizes in 2.6 (5482)
5571 - look up hostnames for IB nals (5602)
5572 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5573 - don't start multiple acceptors for the same port (5277)
5574 - fix incorrect LASSERT in mds_getattr_name (5635)
5575 - export a proc file for general "ping" checking (5628)
5576 - fix "lfs check" to not block when the MDS is down (5628)
5578 - service request history (4965)
5579 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5580 - create an "evict_client" /proc entry on OSTs, like the MDS has
5581 - fix mount usage message, return errors per mount(8) (5168)
5582 - change grep [] to grep "[]" in tests so they work in more UMLs
5583 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5584 - remove ancient LOV_MAGIC_V0 EA support (5047)
5585 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5586 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5587 - allow readahead allocations to fail when low on memory (5383)
5588 - mmap locking landed again, after considerable improvement (2828)
5589 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
5591 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
5594 - send OST transaction number in read/write reply to free req (4966)
5595 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5596 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5597 - the watchdog thread now runs as interruptible (5246)
5598 - flock/lockf fixes (but it's still disabled, pending 5135)
5599 - don't use EXT3 constants in llite code (5094)
5600 - memory shortage at startup could cause assertion (5176)
5602 - reorganization of lov code
5603 - single portals codebase
5605 - add extents/mballoc support (5025)
5606 - direct I/O reads in the obdfilter (4048)
5607 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
5609 tbd Cluster File Systems, Inc. <info@clusterfs.com>
5612 - send OST transaction number in read/write reply to free req (4966)
5613 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5614 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5615 - the watchdog thread now runs as interruptible (5246)
5616 - handle missing objects in filter_preprw_read properly (5265)
5617 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5618 - drop import inflight refcount on signal_completed_replay error (5255)
5619 - unlock page after async write error during send (3677)
5620 - return original error code on reconstructed replies (3761)
5621 - no transno return for symlink open, don't save no-trasno open (3440)
5623 - add pid to ldlm debugging output (4922)
5624 - bump the watchdog timeouts -- we can't handle 30sec yet
5625 - extra debugging for orphan dentry/inode bug (5259)
5627 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
5630 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
5631 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
5632 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
5633 - fix NULL dereference in /proc/sys/portals/routes (4827)
5634 - allow failed mdc_close() operations to be interrupted (4561)
5635 - stop precreate on OST before MDS would time out on it (4778)
5636 - don't send partial-page writes before EOF from client (4410)
5637 - discard client grant for sub-page writes on large-page clients (4520)
5638 - don't free dentries not owned by NFS code, check generation (4806)
5639 - fix lsm leak if mds_create_objects() fails (4801)
5640 - limit debug_daemon file size, always print CERROR messages (4789)
5641 - use transno after validating reply (3892)
5642 - process timed out requests if import state changes (3754)
5643 - update mtime on OST during writes, return in glimpse (4829)
5644 - add mkfsoptions to LDAP (4679)
5645 - use ->max_readahead method instead of zapping global ra (5039)
5646 - don't interrupt __l_wait_event() during strace
5648 - add software watchdogs to catch hung threads quickly (4941)
5649 - make lustrefs init script start after nfs is mounted
5650 - fix CWARN/ERROR duplication (4930)
5651 - return async write errors to application if possible (2248)
5652 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
5653 - print NAL number in %x format (4645)
5654 - update barely-supported suse-2.4.21-171 series (4842)
5655 - support for sles 9 %post scripts
5656 - support for building 2.6 kernel-source packages
5657 - support for sles km_* packages
5659 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
5662 - ignore -ENOENT errors in osc_destroy (3639)
5663 - notify osc create thread that OSC is being cleaned up (4600)
5664 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
5665 - reconstruct ost_handle() like mds_handle() (4657)
5666 - create a new thread to do import eviction to avoid deadlock (3969)
5667 - let lconf resolve symlinked-to devices (4629)
5668 - don't unlink "objects" from directory with default EA (4554)
5669 - hold socknal file ref over connect in case target is down (4394)
5670 - allow more than 32000 subdirectories in a single directory (3244)
5671 - fix blocks count for O_DIRECT writes (3751)
5672 - OST returns ENOSPC from object create when no space left (4539)
5673 - don't send truncate RPC if file size isn't changing (4410)
5674 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
5675 - bind to privileged port in socknal and tcpnal (3689)
5677 - rate limit CERROR/CWARN console message to avoid overload (4519)
5678 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
5679 - basic mmap support (3918)
5680 - kernel patch series update from b1_4 (4711)
5682 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
5685 - avoid crash during MDS cleanup with OST shut down (2775)
5686 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
5687 - don't use bad inodes on the MDS (3744)
5688 - dynamic object preallocation to improve recovery speed (4236)
5689 - don't hold spinlock over lock dumping or change debug flags (4401)
5690 - don't zero obd_dev when it is force cleaned (3651)
5691 - print grants to console if they go negative (4431)
5692 - "lctl deactivate" will stop automatic recovery attempts (3406)
5693 - look for existing locks in ldlm_handle_enqueue() (3764)
5694 - don't resolve lock handle twice in recovery avoiding race (4401)
5695 - revalidate should check working dir is a directory (4134)
5697 - don't always mark "slow" obdfilter messages as errors (4418)
5699 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
5702 - don't close LustreDB during write_conf until it is done (3860)
5703 - fix typo in lconf for_each_profile (3821)
5704 - allow dumping logs from multiple threads at one time (3820)
5705 - don't allow multiple threads in OSC recovery (3812)
5706 - fix debug_size parameters (3864)
5707 - fix mds_postrecov to initialize import for llog ctxt (3121)
5708 - replace config semaphore with spinlock (3306)
5709 - be sure to send a reply for a CANCEL rpc with bad export (3863)
5710 - don't allow enqueue to complete on a destroyed export (3822)
5711 - down write_lock before checking llog header bitmap (3825)
5712 - recover from lock replay timeout (3764)
5713 - up llog sem before sending rpc (3652)
5714 - reduce ns lock hold times when setting kms (3267)
5715 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
5716 - fix NULL deref and obd_dev leak on setup error (3312)
5717 - replace some LBUG about llog ops with error handling (3841)
5718 - don't match INVALID dentries from d_lookup and spin (3784)
5719 - hold dcache_lock while marking dentries INVALID and hashing (4255)
5720 - fix invalid assertion in ptlrpc_set_wait (3880)
5722 - add libwrap support for the TCP acceptor (3996)
5723 - add /proc/sys/portals/routes for non-root route listing (3994)
5724 - allow setting MDS UUID in .xml (2580)
5725 - print the stack of a process that LBUGs (4228)
5727 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
5730 - don't cleanup request in ll_file_open() on failed MDS open (3430)
5731 - make sure to unset replay flag from failed open requests (3440)
5732 - if default stripe count is 0, use OST count for inode size (3636)
5733 - update parent mtime/ctime on client for create/unlink (2611)
5734 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
5735 - free recovery state on server during a forced cleanup (3571)
5736 - unregister_reply for resent reqs (3063)
5737 - loop back devices mounting and status check on 2.6 (3563)
5738 - fix resource-creation race that can provoke i_size == 0 (3513)
5739 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
5740 - more debugging for page-accounting assertion (3746)
5741 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
5742 - avoid deadlock after precreation failure (3758)
5743 - fix race and lock order deadlock in orphan handling (3450, 3750)
5744 - add validity checks when grabbing inodes from l_ast_data (3599)
5746 - add /proc/.../recovery_status to obdfilter (3428)
5747 - lightweight CDEBUG infrastructure, debug daemon (3668)
5748 - change default OSC RPC parameters to be better on small clusters
5749 - turn off OST read cache for files smaller than 32MB
5750 - install man pages and include them in rpms (3100)
5751 - add new init script for (un)mounting lustre filesystems (2593)
5752 - run chkconfig in %post for init scripts (3701)
5753 - drop scimac NAL (unmaintained)
5755 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
5758 - clean kiobufs before and after use (3485)
5759 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
5760 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
5761 - change init script to fail more clearly if not run as root (1528)
5762 - allow clients to reconnect during replay (1742)
5763 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
5764 - don't do DNS lookups on NIDs too small for IP addresses (3442)
5765 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
5766 - fix cond_resched (3554)
5767 - only evict unfinished clients after recovery (3515)
5768 - allow bulk resend, prevent data loss (3570)
5769 - dynamic ptlrpc request buffer allocation (2102)
5770 - don't allow unlinking open directory if it isn't empty (2904)
5771 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
5772 - remove extraneous obd dereference causing LASSERT failure (3334)
5773 - don't use get_cycles() when creating temp. files on the mds (3156)
5774 - hold i_sem when setting i_size in ll_extent_lock() (3564)
5775 - handle EEXIST for set-stripe, set proper directory name (3336)
5777 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
5778 - fix ksocknal_fmb_callback() error messages (2918)
5780 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
5783 - don't copy lvb into (possibly NULL) reply on error (2983)
5784 - don't deref dentry after dput, don't free lvb on error (2922)
5785 - use the kms to determine writeback rpc length (2947)
5786 - increment oti_logcookies when osc is inactive (2948)
5787 - update client's i_blocks count via lvb messages (2543)
5788 - handle intent open/close of special files properly (1557)
5789 - mount MDS with errors=remount-ro, like obdfilter (2009)
5790 - initialize lock handle to avoid ASSERT on error cleanup (3057)
5791 - don't use cancelling-locks' kms values (2947)
5792 - use highest lock extent for kms, not last one (2925)
5793 - don't dereference ERR_PTR() dentry in error handling path (3107)
5794 - fix thread race in portals_debug_dumplog() (3122)
5795 - create lprocfs device entries at setup instead of at attach (1519)
5796 - common AST error handler, don't evict client on completion race (3145)
5797 - zero nameidata in detach_mnt in 2.6 (3118)
5798 - verify d_inode after revalidate_special is valid in 2.6 (3116)
5799 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
5800 - initialize RPC timeout timer earlier for 2.6 (3219)
5801 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
5802 - print nal/nid for unknown nid (3258)
5803 - additional checks for oscc recovery before doing precreate (3284)
5804 - fix ll_extent_lock() error return code for 64-bit systems (3043)
5805 - don't crash in mdc_close for bad permissions on open (3285)
5806 - zero i_rdev for non-device files (3147)
5807 - clear page->private before handing to FS, better assertion (3119)
5808 - tune the read pipeline (3236)
5809 - fix incorrect decref of invalidated dentry (2350)
5810 - provide read-ahead stats and refine rpc in flight stats (3328)
5811 - don't hold journal transaction open across create RPC (3313)
5812 - update atime on MDS at close time (3265)
5813 - close LDAP connection when recovering to avoid server load (3315)
5814 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
5815 - don't leak open file on MDS after open resend (3325)
5816 - serialize filter_precreate and filter_destroy_precreated (3329)
5817 - loop device shouldn't call sync_dev() for nul device (3092)
5818 - clear page cache after eviction (2766)
5819 - resynchronize MDS->OST in background (2824)
5820 - refuse to mount the same filesystem twice on same mountpoint (3394)
5821 - allow llmount to create routes for mounting behind routers (3320)
5822 - push lock cancellation to blocking thread for glimpse ASTs (3409)
5823 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
5824 - fix rare problem with rename on htree directories (3417)
5826 - allow default OST striping configuration per directory (1414)
5827 - fix compilation for qswnal for 2.6 kernels (3125)
5828 - increase maximum number of MDS request buffers for large systems
5829 - change liblustreapi to be useful for external progs like lfsck (3098)
5830 - increase local configuration timeout for slow disks (3353)
5831 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
5833 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5836 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
5837 - don't overwrite extent policy data in reply if lock was blocked (2901)
5838 - drop filter export grants atomically with removal from device (2663)
5839 - del obd_self_export from work_list in class_disconnect_exports (2908)
5840 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
5841 - swab reply message in mdc_close, other PPC fixes (2464)
5842 - fix destroying of named logs (2325)
5843 - overwrite old logs when running lconf --write_conf (2264)
5844 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
5845 - fix race in target_handle_connect (2898)
5846 - mds_reint_create() should take same inode create lock (2926)
5847 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
5848 - don't close files for self_export to avoid uninitialized obd (2936)
5849 - allow MDS with the same name as client node (2939)
5850 - hold dentry reference for closed log files for unlink (2325)
5851 - reserve space for all logs during transactions (2059)
5852 - don't evict page beyond end of stripe extent (2925)
5853 - don't oops on a deleted current working directory (2399)
5854 - handle hard links to targets without a parent properly (2517)
5855 - don't dereference NULL lock when racing during eviction (2867)
5856 - don't grow lock extents when lots of conflicting locks (2919)
5858 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
5861 - account for cache space usage on clients to avoid data loss (974)
5862 - lfsck support in lustre kernel code (2349)
5863 - reduce journal credits needed for BRW writes (2370)
5864 - orphan handling to avoid losing space on client/server crashes
5865 - ptlrpcd can be blocked, stopping ALL progress (2477)
5866 - use lock value blocks to assist in proper KMS, faster stat (1021)
5867 - takes i_sem instead of DLM locks internally on obdfilter (2720)
5868 - recovery for initial connections (2355)
5869 - fixes for mds_cleanup_orphans (1934)
5870 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
5871 - block all file creations until orphan recovery completes (1901)
5872 - client remove rq_connection from request struct (2423)
5873 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
5874 - recovery timer race (2670)
5875 - mdc_close recovey bug (2532)
5876 - ptlrpc cleanup bug (2710)
5877 - mds timeout on local locks (2588)
5878 - namespace lock held during RPCs (2431)
5879 - handle interrupted sync write properly (2503)
5880 - don't try to handle a message that hasn't been replied to (2699)
5881 - client assert failure during cleanup after abort recovery (2701)
5882 - leak mdc device after failed mount (2712)
5883 - ptlrpc_check_set allows timedout requests to complete (2714)
5884 - wait for inflight reqs when ptlrpcd finishes (2710)
5885 - make sure unregistered services are removed from the srv_list
5886 - reset bulk XID's when resending them (caught by 1138 test)
5887 - unregister_bulk after timeout
5888 - fix lconf error (2694)
5889 - handle write after unfinished setstripe, stripe-only getstripe (2388)
5890 - readahead locks pages, leaves pending causing memory pressure (2673)
5891 - increase OST request buffers to 4096 on large machines (2729)
5892 - fix up permission of existing directories in simple_mkdir (2661)
5893 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
5894 - don't assign transno to errored transactions (2742)
5895 - don't delete objects on OST if given a bogus objid from MDS (2751)
5896 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
5897 - if rq_no_resend, then timeout request after recovery (2432)
5898 - fix MDS llog_logid record size, 64-bit array alignment (2733)
5899 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
5900 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
5901 - MDS recovery shouldn't delete objects that it has given out (2730)
5902 - if enqueue arrives after completion, don't clobber LVB (2819)
5903 - don't unlock pages twice when trigger_group_io returns error (2814)
5904 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
5905 - don't write pages to disk if there was an error (1450)
5906 - don't ping imports that have recovery disabled (2676)
5907 - take buffered bytes into account when balancing socknal conn (2817)
5908 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
5909 - reconnect unlink llog connection after MDS reconnects to OST (2816)
5910 - remove little-endian swabbing of llog records (1987)
5911 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
5912 - retry reposting request buffers if they fail (1191)
5913 - grow extent at grant time to avoid granting a revoked lock (2809)
5914 - lock revoke doesn't evict page if covered by a second lock (2765)
5915 - disable VM readahead to avoid reading outside lock extents (2805)
5917 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
5918 - updated kernel patches for hp-2.4.20 kernel (2681)
5920 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
5923 - fix truncated write corruption (2366)
5924 - fix for failed assertion in iopen_connect_dentry (1792,2517)
5926 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
5927 - ensure len(uuid) < 37 in lmc (1171)
5928 - fix ia64 OOPS in llog_test (2255)
5929 - zero end of page at obdfilter for partial page writes (2648)
5930 - don't leave stale dentries around after renames (bug 2428)
5931 - fix timeouts when evicting a client with a single lock held (2642)
5932 - set deadline for the initial HELLO message to drain (2634)
5933 - print out dotted-quad IP addresses in the socknal (2302)
5935 - additional debugging for MDS client eviction problem (2443)
5936 - fix mkfsoptions support for osts (2603, 2604)
5938 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
5941 - add series for the vanilla 2.6.0 kernel
5942 - add series for the vanilla 2.4.24 kernel
5943 - add series for a cray x86/64 UL kernel drop
5944 - fix xattr patches for the vanilla 2.4.19 series
5946 - generate true UUIDs in lmc (1171)
5947 - have portals stack dumping break in UML (2466)
5948 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
5949 - allocate with _NOFS in ldlm to avoid deadlock (1933)
5950 - wake callback waiting threads on client eviction (2460)
5951 - Add --ptldebug and --subsystem to lmc (1719)
5952 - update assertion to allow safe interrupt allocation
5953 - set rq_no_resend for cancel requests (2432)
5954 - recalculate ptlrpcd timeout after resend (2494)
5955 - call vfs_rmdir when removing pending directories (2368)
5956 - fix renaming a file to itself (2429)
5957 - lmc creates a default one-stripe lov (2454)
5958 - expand procfs space to handle large clusters (2326)
5959 - increase UML stack to avoid overflow
5960 - update lconf's list of debug and subsystem masks
5961 - fix lfs find --obd (2510)
5962 - /proc tunable for disabling filter read caching (2591)
5963 - stop rpm packages from altering slapd.conf (2301)
5964 - disable nagle in the socknal under 0conf (2578)
5965 - choose mds inode size based on stripe count (2572)
5966 - fix kernel-source rpm problems (2516)
5968 - add --disable-doc to avoid pdf generation (2421)
5969 - update documentation, tests, type-os, comments
5970 - avoid format warnings on ia64
5971 - remove the TOE NAL
5972 - tiny code cleanups by removing unused fields
5974 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
5977 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
5978 - remove the most blatant lies from BUILDING file (2371)
5979 - change default debug level to reasonable production setting
5980 - reduce client side cache size to reduce cache flush time
5981 - reduce max RPCs in flight to avoid unnecessary file fragmentation
5982 - make TCP zerocopy and pinger support enabled by default (2476)
5983 - sync writes completed after process exits caused crashes (2319)
5984 - maintain correct mount count on the MDS (2356)
5985 - backout 1557, because 2316 wasn't really fixed
5986 - better file I/O statistics gathering in /proc
5987 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
5988 - another kernel patch to fix zero-copy TCP function export
5989 - don't take duplicate lock when processing re-sent getattr (2420)
5990 - lctl uses obd_self_export instead of creating new conn (2353)
5991 - MDS/OST recovery case which requires object creation asserted (2425)
5992 - move lfs from /usr/sbin to /usr/bin in packages
5993 - fix race between mds_client_add and mds_client_free (2417)
5994 - use kmalloc instead of slabs in portals (2430)
5995 - don't create duplicate records when a failover MDS is present (2442)
5996 - remove unnecessary mount age check (2332)
5997 - don't remove directory inodes from locks prematurely (2451)
5998 - don't break if MDS service name is the same as hostname (2103)
5999 - fix races in client write RPC generation when cache full (2482)
6001 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
6004 - remove now-unused request->rq_obd (278)
6005 - if an allocation fails, print out how much memory we've used (1933)
6006 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
6007 - add the "configurable stack size" patch to most series files (1256)
6008 - ability to write large log records, for 100+ OST configs (2306)
6009 - fix NULL deref when filter_prep fails (2314)
6010 - fix operator precedence error in filter_sync
6011 - dynamic allocation of socknal TX descriptors (2315)
6012 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
6013 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
6014 - ensure that CWARN messages in Portals always get to the syslog
6015 - __init/__exit are not for prototype decls (ldlm_init/exit)
6016 - x86-64 compile warning fixes
6017 - fix gateway LMC keyword conflict (2318)
6018 - fix MDS lock inversions in getattr/reint paths (1844)
6019 - fix a rare lock re-ordering bug, which caused deadlock (2322)
6020 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
6021 - DLM race condition prevented some lock evictions (2328)
6022 - ENOMEM detection and retry on socknal sends (2230)
6023 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
6024 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
6025 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
6026 - better error messages when a client is rejected during recovery (1505)
6027 - avoid cancelling locks which were never granted, after failure (2330)
6028 - fix i_sem/journal inversion in mds_client_add (2333)
6029 - fix truncate/getattr lock cycle deadlock (2334)
6030 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
6031 - fix two rare exit paths which could leak an l_lock() ref (2321)
6032 - fencepost error in MDS/OST orphan recovery (2226)
6033 - make log record alignment 8 bytes (1988)
6034 - lstripe now fails when requested offset > ost_count (2237)
6035 - ensure that all kernel series have a complete list.h (1607)
6036 - fix crashes in special-file operations (2316)
6037 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
6039 - allow configurable automake binary, for testing new versions
6040 - small update to the lfs documentation
6042 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
6044 * fix negative export reference count in fsfilt_sync (2312)
6046 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
6047 * release candidate 0.9.1
6049 - orphans are moved into the PENDING directory for possible recovery
6050 - replayed opens now open by fid for orphan/rename safety (1042)
6051 - last close of an orphan inode generates a transno (683)
6052 - chdir() and mount() now pin the directory entry (1020)
6053 - avoid CERROR in normal ll_setattr_raw() error case (1500)
6054 - discard very old requests without processing them (1502)
6055 - remove some common, well-understood CERRORs (1505)
6056 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
6057 - clear "grant" flags in OST replies until OST grant code lands (1644)
6058 - fix read performance by not clobbering i_blksize on client (1598)
6059 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
6060 - make LRU size a /proc tunable, clears locks when reduced (707)
6061 - fix some lprocfs rot that prevented ptlbd from loading (1732)
6062 - server locks take references on exports now (1558)
6063 - build fixes for 2.4.20-rh trees (1663)
6064 - return an error from lov_create if all OSCs are inactive (1751)
6065 - fix import levels when a reconnect happens without a timeout (1597)
6066 - exit early from mds_open if we get a lookup error (1749)
6067 - partial page read at EOF wouldn't wait for disk before sending (1642)
6068 - avoid NULL deref in obdfilter when reading page past EOF (1592)
6069 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
6070 - fix LBUG in ll_it_open_error with rc = -2 (1861)
6071 - write/truncate lock inversion (1639)
6072 - Don't auto-load obdclass, portals modules during cleanup (1495)
6073 - fix timestamps from jumping to "now" (1763)
6074 - extra journal assertions (1648)
6075 - add an extra multiunlink test (1771)
6076 - fix read_record/write_record API (1776)
6077 - fix leak of offset_extent, possible incorrect i_size later (1772)
6078 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
6079 - Debugging for the kqswnal_get_idle_tx problems (1820)
6080 - Allow recovery to be attempted multiple times (1536)
6081 - Write out MDS last_rcvd file after it is first created (1600)
6082 - Fix tx_descriptor leak in failed transmit situations (1827)
6083 - ext3 journaling fixes for assertion failure after IO error (1871)
6084 - class_export_put() on freed export after completion AST error (1896)
6085 - Fix revalidate looping in VFS (1322)
6086 - Don't access a freed export during MDS_REINT timeout (1521)
6087 - Add open-unlink recovery support on the MDS (1673,1764)
6088 - Return an error if no MDS data was read from last_rcvd (1946)
6089 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
6090 - Don't setup a disk device that doesn't match exported UUID (317)
6091 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
6092 - avoid committing NULL handle in force close
6093 - local.sh is now a one-stripe LOV configuration
6094 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
6095 - fix ext3 htree duplicate directory entry corruption (1516)
6096 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
6097 - update to new LOV EA format (2097)
6098 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
6100 2003-06-15 Phil Schwan <phil@clusterfs.com>
6103 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
6104 - per-import/export recovery handling (958, 931, 959)
6105 - multiple last-rcvd slots, for serving multiple FSes (949)
6106 - connections are again shared between multiple imp/exports (963, 964)
6107 - "umount -f" would hang if any requests needed to be sent (393, 978)
6108 - avoid pinning large req buffer by copying for queued messages (989)
6109 - add "uuid" to "lctl device" command to help upcalls (991)
6110 - "open" RPCs with transnos would confuse recovery counters (1037)
6111 - do proper endian conversion of all wire messages (288, 340, 891)
6112 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
6113 - call waiting locks callback from LDLM recovery thread (1127, 1151)
6114 - fix ptlrpc_connection leak in target_handle_connect (1174)
6115 - fix import refcounting bug in OST and MDS cleanup (1134)
6116 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
6117 - fix very unlikely obd_types race condition (501)
6118 - remove osc_open hack for echo_client (1187)
6119 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
6120 - a failure in read_inode2 leads to deadlock (1139)
6121 - cancel ack-locks as soon as transaction is committed (1072)
6122 - fix major leaks and crashes in the bulk I/O path (937, 1057)
6123 - make sure to commitrw after any preprw to avoid deadlock (1162)
6124 - failing to execute a file in a lustre FS would lock inode (1203)
6125 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
6126 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
6127 - obd_uuid2tgt would walk off the end of the list (1255)
6128 - on IA64 the osc would give portals incorrect bulk size (1258)
6129 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
6130 - fix lock inversion caused by new llite matching code (1282)
6131 - limit the number of dirty pages on a client to 10MB (1286)
6132 - timed out locks were not being corrected cancelled (1289)
6133 - fix O_DIRECT above 4GB on IA-32 (1292)
6134 * major user-visible changes
6135 - fail out/fail over policy now controlled by the upcall (993)
6137 - add OBD_PING to check server availability and failure (954)
6138 - lustre messages are now sent in sending host order (288, 340, 891)
6139 - add eadatalen to MDS getattr reply (340)
6140 - OST read replies may contain second buffer, with per-page status (593)
6142 2003-03-11 Phil Schwan <phil@clusterfs.com>
6145 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
6146 - failed open()s could cause deadlock; fixed (867, 869)
6147 - stop cancelling OST locks when files are closed (481)
6148 - overlapping XID spaces caused network corruption (851, 853)
6149 - fix unsafe fsfilt counter arithmetic; change to atomic_t
6150 - setattr_raw added, to do single-RPC, server-side setattrs
6151 - lmc/lconf syntax change for OST UUIDs
6152 - fix crashy race condition between ptlrpc_free_req and osc_close
6153 - don't use request in mdc_enqueue if we hit a timeout (889)
6154 - don't set the inode i_size for regular files from the MDS (896)
6155 - handle out of order completion AST (842)
6156 - don't LBUG if a lock request times out after receiving AST (913)
6157 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
6158 - if a bad lock AST arrives, send an error instead of dropping entirely
6159 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
6160 - fix leak in bulk IO when only partially completed (899, 900, 926)
6161 - fix O_DIRECT for ia64 (55)
6162 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
6163 - C-z after timeout could hang a process forever; fixed (977)
6165 - client-side I/O cache (678, 924, 929, 941, 970)
6167 - READPAGE and SETATTRs which don't take server-side locks get
6170 2003-02-11 Phil Schwan <phil@clusterfs.com>
6173 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
6174 - Fix fsfilt_extN_readpage() to read a full page of directory
6175 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
6176 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
6177 - Fix imbalanced LOV object allocation and out-of-bound access (469)
6178 - Most intent operations were removed, in favour of a new RPC mode
6179 that does a single RPC to the server and bypasses most of the VFS
6180 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
6181 - Aggressively cancel local locks on DLM servers
6182 - mds_reint_unlink sends EA to the client if it's the last nlink.
6183 client uses that EA to unlink OST objects.
6184 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
6185 - recursive symlinks were fixed (439)
6186 - fixed NULL deref in DEBUG_REQ
6187 - filter_update_lastobjid no longer calls sync, which annoyed extN
6188 - fixed multi-client small-writes to a single file problem (445)
6189 - fixed mtime updates during file writes (607)
6190 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
6191 - fixed bug in obd_brw_read/write() (under guise of testing 367)
6192 - fixed Linux OST size reporting problem (444, 656)
6193 - OST now updates object mtime with writes or setattr (607, 619)
6194 - client verifies file size before zeroing page past EOF (445)
6195 - OST now writes last allocated objid to disk with allocation (108)
6196 - LOV on echo now works (409)
6198 - mds_reint_unlink sends a new buffer, with the EA included. this
6199 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
6200 set if a regular file was being unlinked, and it was the last link
6201 - use PtlGet from the target for bulk writes (315)
6202 - OST now updates object mtime with writes or setattr (607, 619)
6203 - LDLM now has a grant-time callback to revalidate locked items, if
6205 - Many MDS operations were reorganized to combat race conditions
6207 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
6208 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
6210 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
6213 - Fully reactivate OST imports after reconnection (512, others)
6214 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
6215 - More graceful error handling for truncating on dead OST (515)
6216 - Don't error out unless we're actually accessing dead stripes (474)
6217 - Fix garbage sizes when stripes are missing (410)
6218 - LRU counters were broken, causing constant lock purge (433, 432)
6219 - garbage on read from stripes with failed OSTs (441)
6220 - mark OSCs as active before reconnecting during recovery (438)
6221 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
6222 - lfind did not preserve OST order in output (443)
6223 - symlinks cause hung clients, incorrect data (439)
6224 - stop dereferencing request after dropping refcount (457)
6225 - don't LASSERT(spin_is_locked) on non-SMP (455)
6226 - fixes for many rename() bugs
6227 - fstat didn't correctly synchronize attributes (399)
6228 - server must handle lock cancellation during blocking AST prep (487)
6229 - bulk descriptors were free()d too soon (511)
6230 - fix paths in lconf, which would load incorrect modules (451, 507)
6231 - fix confusing lconf 'host not found' error message (386)
6232 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
6233 - fix race condition in mdc_blocking_ast() for inode access (526)
6234 - fix lov_unpackmd() unpacking wrong number of stripes (537)
6235 - fix lov_set_osc_active() marking wrong OSC inactive (440)
6236 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
6237 - fix multiple writes of stripe MD to MDS (358, maybe 519)
6238 - fix lstripe in several ways (kernel side) (527)
6239 - fix request leak in ldlm_cli_enqueue (262)
6240 - incorrect OSC was marked inactive after OST failure
6241 - call mds_fs_cleanup before unmounting filesystem (524)
6242 - fix races between taking ns_lock and ldlm_lock_change_resource
6243 - fix races updating LOV export open file list
6244 - fix lov_enqueue error path, avoid decref-ing bad lock handle
6245 - fix recovery NULL deref in ldlm_cli_cancel_unused
6246 - fix some DLM races by using new hash table for lock handles (419)
6247 - permit the client to specify desired inodes, at replay
6248 - duplicate requests when we queue them for replay reintegration
6249 - fix last_rcvd offset calculation
6250 - sync after each recovered transaction, so we always make progress
6251 - never, not always, ERESTART requests without transnos
6252 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
6254 - skip replay if the MDS says that the client is already connected
6255 - don't check for a recovery-enabled export to match lctl's UUID
6256 - don't INC_USE_COUNT for phantom exports
6257 - don't crash when cleaning up phantom exports (567)
6258 - don't double-finish or set replay data for errored mdc_open requests
6259 - abort requests when they time out, so we don't get old replies
6260 - send/receive replies for AST messages again
6261 - if the client says that it doesn't have the lock, cancel it on the
6263 - if we timeout during I/O, don't try to cancel an in-use lock; instead
6264 mark it as destroyed, it will all work out when decref is called
6265 - fix module use counts (22, 581)
6267 - ASTs now expect a reply (server cancels lock on error reply)
6269 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
6272 - fix many simultaneous client startup (392)
6273 - fix dentry->d_it clobbering
6274 - credentials weren't being shipped for readdir/getattr operations
6275 - remove invalid assertions triggered during some concurrent MD
6277 - proper Lustre versions added (336, 389)
6278 - fix memory leak for create error case (398)
6279 - fix LOV locking bug that would get cli/srv out of sync
6280 - fix echo client over LOV (409)
6281 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
6282 - fix double-O_EXCL intent crash (424)
6283 - avoid sending multiple lock CANCELs (352)
6285 - MDS can do multi-client recovery (modulo bugs in new code)
6287 - many updates, edits, cleanups
6289 2002-11-18 Phil Schwan <phil@clusterfs.com>
6292 - fix null d_it dereference (346)
6293 - fix full OST/dbench hang (333)
6294 - fix permission problem with file removal (286)
6295 - fix removal of OSCs from LOV when they fail
6296 - fix NULL deref during bulk timeout (214)
6297 - fix problems related to multiple filesystems on one MDS (241)
6298 - fixed serious subtle metadata locking bugs
6299 - free locks on clients when inodes are removed due to memory
6301 - fix inode pointer in lock data (285)
6302 - partial support for multiple MDS on a single host (241)
6303 - data locks weren't cancelled at clear_inode time (290, 311)
6304 - intent locks could lead to unbounded lock growth (205)
6305 - added a maximum lock count, an LRU list, and a flusher
6306 - fix multiple rename (365)
6307 - properly abstracted the echo client
6308 - OSC locked 1 byte too many; fixed
6309 - rewrote brw callback code:
6310 - fixed recovery bugs related to LOVs (306)
6311 - fixed too-many-pages-in-one-write crash (191)
6312 - fixed (again) crash in sync_io_timeout (214)
6313 - probably fixed callback-related race (385)
6315 - Add capability to MDS protocol
6316 - LDLM cancellations and callbacks on different portals
6318 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
6321 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
6322 - defer open object destruction to close time (601981, 138)
6323 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
6324 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
6325 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
6326 - don't call obd_setattr in ll_file_release for destroyed objects
6327 * protocol change to lustre_msg: move |version| and add |flags|
6328 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
6329 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
6330 * added replay of create, unlink, link and rename operations during
6331 MDS failover; recovery should be much more robust now
6332 * remove failed OSCs from LOVs (only lov_create uses this so far)
6333 * the lustre-HOWTO was brought (more) up to date (582544)
6335 2002-10-23 Phil Schwan <phil@clusterfs.com>
6338 - in-use dentries weren't being reused properly (617851)
6339 - prevent multiple LDLM setup (599178)
6340 - fix LOV size calculations for truncate (617853)
6341 - fix client handling of MDS intent errors (POSIX)
6342 - fix permission bug in lovstripe.c test (624321)
6343 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
6344 - truncate past end of file could corrupt data
6345 - proper cleanup after timeouts, crashes, etc (592524, 550815)
6346 - a race in recovery could return ETIMEDOUT to apps (623947)
6347 - building outside the source directory was fixed
6348 * the lustre-HOWTO was brought (more) up to date (582544)
6349 * major progress was made on recovery functionality
6351 2002-10-10 Phil Schwan <phil@clusterfs.com>
6354 - recovery deadlock fix
6355 - rm -rf causes LBUG fix (617817)
6356 - file open by multiple tasks fix (618962)
6357 - directory permissions bugs (602707 and 620007)
6358 - journal_stop fixed with locking (611313)
6359 - O_APPEND failures resolved (618273, perhaps 614459)
6360 - lconf PATH fix (619770)
6361 - IA64 build fix (621450)
6362 - RPC buffer sizes scale with amount of memory
6364 2002-10-01 Phil Schwan <phil@clusterfs.com>
6367 - locks would be cancelled without throwing away data pages,
6368 resulting in inconsistent data (605627)
6369 - inode attributes were not always being refreshed (605627, 612449)
6370 - lconf now continues to cleanup after lctl reports an error
6371 - MDS now enforces user permissions (602707)
6372 - lprocfs cleanup fixed, but not yet enabled (614157)
6373 - fixed infinite server hang, should a client not respond to an AST
6374 - avoid going into recovery if user calls readlink() with a buffer
6375 that's too small (613941)
6376 - AST RPCs no longer require replies (614867) -- this may be changed
6377 - don't crash server if client sends an IOV that's too big (611336)
6378 - fixed lock conversion deadlock (611892)
6379 - fixed the following of symlinks (614622)
6380 * recovery: the server can remove locks from a client that dies, other
6381 clients can make progress
6382 * more extN patch fixes
6383 * compile-time configurable ptlrpc buffer allocations
6385 - collaborative read cache document
6386 - Lustre Lite Performance CDR document-in-progress
6388 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6391 - fix typo in patch-2.4.18
6393 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6396 - clear ptlrpc request each time in handle_incoming_request()
6397 - unlink of files now destroys the object on the OST
6399 2002-09-19 Peter Braam <braam@clusterfs.com>
6401 * add hard link support
6402 * change obdfile creation method
6403 * kernel patch changed
6405 2002-09-19 Peter Braam <braam@clusterfs.com>
6408 - stack overflow bug in extN fixed
6410 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
6412 * documentation updates
6413 - add man pages for config tools
6414 - update tests/README to describe testing with new config tools
6415 - finish metadata API descriptions
6416 * bug fixes and cleanups
6417 - statfs workaround for 16TB limit
6418 - LOV stripe allocation improved, can stripe on subset of OSTs
6419 - LOV file size/IO offset was wrong for files > 4GB in size
6420 - object EA data was being dropped, caused files to be unreadable
6421 - memory overflow with non-LOV OST caused memory corruption
6422 - fixed regression tests to work with new config tools, obdfilter
6423 - fixed bug when directory size became larger than 1 block
6424 - fixed bug (for single client case) when PWD was deleted
6425 - invalidate local directory pages when doing intent-based ops
6426 - avoid LDLM oops when lock callback contained bad data
6428 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
6430 * documentation updates
6431 * bug fixes and cleanups
6432 - configuration tools
6435 - 64-bit compile warnings
6436 - 64-bit internal statfs data
6438 * test_brw on persistent OST devices
6440 * lprocfs (disabled)
6442 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
6444 * documentation updates
6445 * bug fixes and cleanups
6446 * configuration tools
6448 2002-08-30 Peter J. Braam <braam@clusterfs.com>
6451 * many small fixes to 0_5_4
6452 * io/network handling
6453 * thinkos in MDS operations
6455 2002-08-24 Peter J. Braam <braam@clusterfs.com>
6458 * crucial basic fixes to 0.5.3
6459 * IOR, Iozone work over Elan
6462 2002-08-07 Phil Schwan <phil@clusterfs.com>
6463 * version 0_5_3, our first alpha
6464 * we use the new Portals iovs
6465 * documentation updates
6466 * bug fixes and cleanups
6467 * small changes in the DLM wire protocol
6469 2002-07-25 Peter J. Braam <braam@clusterfs.com>
6470 * version 0_5_1 with some initial stability,
6471 * locking on MD and file I/O.
6472 * documentation updates
6473 * several bug fixes since 0.5.0
6474 * small changes in wire protocol
6476 2002-07-18 Phil Schwan <phil@clusterfs.com>
6478 * delivered as Lustre Light Alpha
6479 * fixed a crash after handling invalid MDS requests
6480 * fixed directory pages for architectures with non-4k pages sizes
6482 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
6483 * release version v0_4_4
6484 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
6486 * Fixes a number of interruption problems with OST operations.
6487 * Update documentation for portals header changes
6488 * Move all wire protocol structs/defines to lustre_idl.h
6489 * Fixes symlink length bug.
6490 * Add tcpdump to repository.
6492 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
6493 * release version v0_4_3
6494 * Fixes statfs for inodes on extN.
6495 * Fixes bug in runtests which would delete /etc/hosts.
6496 * Use 64-bit object IDs wherever possible (not into VFS though)
6497 Remove ost_get_info, which is unused by lustre, and out of date.
6499 2002-07-03 Peter Braam <braam@clusterfs.com>
6500 * release version v0_4_2 Fixes a lookup error (type not passed)
6501 * move forward to head of Portals
6502 * move forward to latest Lustre kernel
6504 2002-06-25 Peter Braam <braam@clusterfs.com>
6505 * release version v0_4_1. Hopefully stable on single node use.