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 2.6.32-573.26.1.el6 (RHEL6.7)
9 2.6.32-642.1.1.el6 (RHEL6.8)
10 3.10.0-327.22.2.el7 (RHEL7.2)
11 3.0.101-0.47.71 (SLES11 SP3)
12 3.0.101-77 (SLES11 SP4)
13 3.12.57-60.35 (SLES12 SP1)
14 vanilla linux 4.2.1 (ZFS only)
15 * Client known to build on unpatched kernels:
16 2.6.32-431.29.2.el6 (RHEL6.5)
17 2.6.32-504.30.3.el6 (RHEL6.6)
18 2.6.32-573.26.1.el6 (RHEL6.7)
19 2.6.32-642.1.1.el6 (RHEL6.8)
20 3.10.0-327.22.2.el7 (RHEL7.2)
21 3.0.101-0.47.71 (SLES11 SP3)
22 3.0.101-68 (SLES11 SP4)
23 3.0.101-77 (SLES11 SP4)
24 3.12.57-60.35 (SLES12 SP1)
26 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
27 * Recommended ZFS / SPL version: 0.6.5.7
28 * Tested with ZFS / SPL version: 0.6.5.7
29 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
30 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
31 stack overflow. For more information, please refer to bugzilla 17630.
32 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
33 the re-exporting nodes to be mounted with "32bitapi" mount option
35 --------------------------------------------------------------------------------
37 02-29-2016 Intel Corporation
39 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
40 for currently supported client and server kernel versions.
41 * Server known to build on patched kernels:
42 2.6.32-431.29.2.el6 (RHEL6.5)
43 2.6.32-504.30.3.el6 (RHEL6.6)
44 2.6.32-573.12.1.el6 (RHEL6.7)
45 3.10.0-327.3.1.el7 (RHEL7.2)
46 3.0.101-0.47.71 (SLES11 SP3)
47 3.0.101-68 (SLES11 SP4)
48 vanilla linux 4.2.1 (ZFS only)
49 * Client known to build on unpatched kernels:
50 2.6.32-431.29.2.el6 (RHEL6.5)
51 2.6.32-504.30.3.el6 (RHEL6.6)
52 2.6.32-573.12.1.el6 (RHEL6.7)
53 3.10.0-327.3.1.el7 (RHEL7.2)
54 3.0.101-0.47.71 (SLES11 SP3)
55 3.0.101-68 (SLES11 SP4)
58 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
59 * Recommended ZFS / SPL version: 0.6.4.2
60 * Tested with ZFS / SPL version: 0.6.4.2
61 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
62 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
63 stack overflow. For more information, please refer to bugzilla 17630.
64 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
65 the re-exporting nodes to be mounted with "32bitapi" mount option
67 --------------------------------------------------------------------------------
69 03-10-2015 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 2.6.32-431.29.2.el6 (RHEL6.5)
75 2.6.32-504.8.1.el6 (RHEL6.6)
76 3.0.101-0.46 (SLES11 SP3)
77 * Client known to build on unpatched kernels:
78 2.6.32-431.29.2.el6 (RHEL6.5)
79 2.6.32-504.8.1.el6 (RHEL6.6)
80 3.10.0-123.20.1.el7 (RHEL7)
81 3.0.101-0.46 (SLES11 SP3)
82 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
83 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
84 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
85 stack overflow. For more information, please refer to bugzilla 17630.
86 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
87 the re-exporting nodes to be mounted with "32bitapi" mount option
89 Severity : enhancement
91 Description: control OST-index in IDIF via ROCOMPAT flag.
92 Details : Introduce new flag OBD_ROCOMPAT_IDX_IN_IDIF that is stored in the
93 last_rcvd file. For new formatted OST device, it will be auto set;
94 for the case of upgrading from old OST device, you can enable it
95 via the lproc interface osd-ldiskfs.index_in_idif. With such flag
96 enabled, for new created OST-object, its IDIF-in-LMA will contain
97 the OST-index; for the existing OST-object, the OSD will convert
98 old format IDIF as new format IDIF with OST-index stored in the
99 LMA EA when accessing such OST-object or via OI scrub. Once such
100 flag is enabled, it cannot be reverted back, so the system cannot
101 be downgraded to the orignal incompatible version.
103 --------------------------------------------------------------------------------
105 07-30-2014 Intel Corporation
107 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
108 for currently supported client and server kernel versions.
109 * Server known to build on patched kernels:
110 2.6.32-431.20.3.el6 (RHEL6.5)
111 3.0.101-0.31 (SLES11 SP3)
112 * Client known to build on unpatched kernels:
113 2.6.32-431.20.3.el6 (RHEL6.5)
114 3.0.101-0.31 (SLES11 SP3)
117 * Recommended e2fsprogs version: 1.42.9.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 10-31-2013 Intel Corporation
128 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
129 * Server builds for kernels:
130 2.6.32-358.18.1.el6 (RHEL6.4)
131 * Client builds for unpatched kernels:
132 2.6.32-358.18.1.el6 (RHEL6.4)
133 3.0.92-0.8 (SLES11 SP3)
135 * Recommended e2fsprogs version: 1.42.7.wc2 or newer
136 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
137 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
138 stack overflow. For more information, please refer to bugzilla 17630.
139 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
140 the re-exporting nodes to be mounted with "32bitapi" mount option
142 --------------------------------------------------------------------------------
144 05-30-2013 Intel Corporation
146 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
147 * Server builds for kernels:
148 2.6.32-358.6.2.el6 (RHEL6.4)
149 * Client builds for unpatched kernels:
150 2.6.32-358.6.2.el6 (RHEL6.4)
151 3.0.74-0.6.6 (SLES11 SP2)
153 * Recommended e2fsprogs version: 1.42.7.wc1 or newer
154 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
155 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
156 stack overflow. For more information, please refer to bugzilla 17630.
157 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
158 the re-exporting nodes to be mounted with "32bitapi" mount option
160 --------------------------------------------------------------------------------
162 9-30-2012 Whamcloud, Inc.
164 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
165 * Server support for kernels:
166 2.6.32-279.5.1.el6 (RHEL6)
167 * Client support for unpatched kernels:
168 2.6.32-279.5.1.el6 (RHEL6)
170 2.6.18-238.19.1.el5 (RHEL5)
171 2.6.16.60-0.69.1 (SLES10)
172 * Recommended e2fsprogs version: 1.41.90.wc4
173 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
174 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
175 stack overflow. For more information, please refer to bugzilla 17630.
176 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
177 the re-exporting nodes to be mounted with "32bitapi" mount option
179 Severity : enhancement
181 Description: Update RHEL6.3 kernel to 2.6.32-279.5.1.el6.
183 Severity : enhancement
185 Description: Update RHEL6.3 kernel to 2.6.32-279.2.1.el6.
187 Severity : enhancement
189 Description: Update RHEL6.2 kernel to 2.6.32-220.17.1.el6
191 Severity : enhancement
193 Description: Update RHEL6.2 kernel to 2.6.32-220.13.1.el6
195 Severity : enhancement
197 Description: Update RHEL6.2 kernel to 2.6.32-220.7.1.el6
199 --------------------------------------------------------------------------------
201 04-30-2012 Whamcloud, Inc.
203 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
204 * Server support for kernels:
205 2.6.32-220.4.2.el6 (RHEL6)
206 * Client support for unpatched kernels:
207 2.6.32-220.4.2.el6 (RHEL6)
209 2.6.18-238.19.1.el5 (RHEL5)
210 2.6.16.60-0.69.1 (SLES10)
211 * Recommended e2fsprogs version: 1.41.90.wc4
212 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
213 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
214 stack overflow. For more information, please refer to bugzilla 17630.
215 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
216 the re-exporting nodes to be mounted with "32bitapi" mount option
218 Severity : enhancement
220 Description: Update RHEL6.2 kernel to 2.6.32-220.4.2.el6
222 Severity : enhancement
224 Description: Update RHEL6.2 kernel to 2.6.32-220.4.1.el6
226 Severity : enhancement
228 Description: Update RHEL6.2 kernel to 2.6.32-220.el6
230 Severity : enhancement
232 Description: Update RHEL6.1 kernel to 2.6.32-131.17.1.el6
234 --------------------------------------------------------------------------------
236 09-30-2011 Whamcloud, Inc.
238 * Server support for kernels:
239 2.6.18-238.19.1.el5 (RHEL5)
240 2.6.32-131.6.1.el6 (RHEL6)
241 * Client support for unpatched kernels:
242 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
243 2.6.18-238.19.1.el5 (RHEL5)
244 2.6.32-131.6.1.el6 (RHEL6)
245 2.6.29.4-167.fc11 (FC11)
246 2.6.16.60-0.69.1 (SLES10)
248 * Recommended e2fsprogs version: 1.41.90.wc3
249 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
250 Lustre filesystem with 4K stack may cause a stack overflow. For more
251 information, please refer to bugzilla 17630.
252 * NFSv4 reexport requires Lustre client on the re-exporting nodes to be
253 mounted with "32bitapi" mount option
255 Severity : enhancement
256 Description: Update RHEL6.1 kernel to 2.6.32-131.6.1.el6
258 Severity : enhancement
259 Description: Update RHEL5.6 kernel to 2.6.18-238.19.1.el5
261 Severity : enhancement
262 Description: Update RHEL5.6 kernel to 2.6.18-238.9.1.el5
265 Frequency : only with SLES10
267 Description: Use OFED "KMP" provided by Novell
268 Details : SLES10 SP3 ships with OFED in a separate "KMP" package.
269 Lustre is now built against this package. That means you need to
270 install the ofed-kmp package from Novell for the patchless client.
271 Note that the ofed-kmp that Novell ships may not exactly match the
272 kernel version but should still be compatible.
274 Severity : enhancement
276 Description: Update RHEL5.5 kernel to 2.6.18-194.17.1.el5
277 Update OEL5.5 kernel to 2.6.18-194.17.1.0.1.el5.
279 --------------------------------------------------------------------------------
281 2010-07-15 Oracle, Inc.
283 * Server support for kernels:
284 2.6.18-164.11.1.el5 (RHEL 5)
285 2.6.18-164.11.1.0.1.el5 (OEL 5)
286 * Client support for unpatched kernels:
287 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
288 2.6.18-164.11.1.el5 (RHEL 5),
289 2.6.18-164.11.1.0.1.el5 (OEL 5)
290 2.6.16.60-0.42.8 (SLES 10),
292 2.6.29.4-167.fc11 (FC11)
293 * Recommended e2fsprogs version: 1.41.10-sun2
295 Severity : enhancement
297 Description: Distribute lfs_migrate script with lustre
298 Details : The lfs_migrate mentioned in the manual is now included in the
299 lustre tree in lustre/scripts and will distributed in lustre RPMs.
301 Severity : enhancement
303 Description: Failover nid list added to import proc file.
305 Severity : enhancement
307 Description: "som_preview" mount option required to use SOM
308 Details : The clients will need to use som_preview mount option
312 Description: Lustre does not do 1MB IOs to HW RAID
313 Details : Bump MAX_PHYS/HW_SEGMENTS and SG_ALL to 256 in the RHEL5 kernel.
315 Severity : enhancement
317 Description: lfs quota output is not very convenient for awk/sed-parsing
318 Details : Some positions in lfs quota output table could be empty or
319 non-empty which made it hard to parse it with scripts, now
320 a dash is put instead of space where there is not supposed
325 Description: Update to OFED 1.5.1
326 Details : OFED 1.5.1 will be used in environments that don't use
327 vendor supplied OFED.
331 Description: Remove lnet. prefix from lctl params display
332 Details : Also renamed lnet.memused to lnet_memused to
337 Description: Use conf_param -d to remove permanant settings.
341 Description: Add list_param -R
342 Details : lctl list_param now supports recursive mode.
345 Bugzilla : 21829, 22312
346 Description: Fix llobdstat.
347 Details : 1. Now updating the write_bytes counter
348 2. Port script fixes from 1.8.
352 Description: decrease the usage of memory on clients.
353 Details : 1. On clients, recycle dentries and inodes unused.
354 2. Delete the code related to ll_deathrow(att 6215 in bug 1443). It
359 Description: llobdstat fix and enhancement
360 Details : add a counter to set a limit to how many samples will be returned
361 fix a wildcard in the path to limit to obdfilter stats only
365 Description: lfs check only allowed for root
366 Details : Code cleanups and added a regresion test of non-root use of
371 Description: MDS Oopses in ptlrpc_dispatch_difficult_reply()/__list_add()
372 Details : Fix hr->hr_index overload.
376 Description: Fix FIEMAP functionality.
380 Description: FC11 patchless client support added
384 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and
385 OEL5.4 kernel to 2.6.18-164.11.1.0.1.el5.
389 Description: New acc-sm script and various test-framework enhancments.
393 Description: Previously landed patch was reverted.
397 Description: Update SLES10 kernel to 2.6.16.60-0.42.8 (patchless)
401 Description: Add ldiskfs maxdirsize mount option.
405 Description: Fix mkfs.lustre support for 16TB LUNs.
409 Description: Print mmp_check_interval
410 Details : print mmp_check_interval and make it possible to abort
411 mount operation in case it takes too long.
415 Description: Add cascading_rw to lustre mpi tests.
419 Description: Store SOM data in EA.
420 Details : Also several SOM related cleanups.
424 Description: Add refcount to md_open_data
428 Description: Use refcount on lov_request_set to prevent memory corruption.
432 Description: Statahead: use parent dir's i_mutex for synchronization.
436 Description: loadgen improvements.
440 Description: Enable peer health detection in ksocklnd and ko2iblnd.
444 Description: Control DCACHE_LUSTRE_INVALID flag with MDS_INODE_LOOKCUP lock
448 Description: Fix race on obd_devcie:md_stats
452 Description: Fix SLES 11 client builds.
453 Details : Resolved uninitalized pointer gcc warnings.
457 Description: Allow non-root access for "lfs check".
458 Details : Added a check in obd_class_ioctl() for OBD_IOC_PING_TARGET.
462 Description: Clear imp_force_reconnect correctly in ptlrpc_connect_interpret()
464 Severity : enhancement
466 Description: Update kernel to RHEL5.4 2.6.18-164.6.1.el5 and
467 OEL5 2.6.18-164.6.1.0.1.el5(Both in-kernel OFED enabled).
471 Description: Sanity test_27p fails.
472 Details : OBD_FAIL_OST_ENOSPC has been moved to filter_statfs().
473 Sanity tests 27 were cleaned up and improved.
477 Description: Unify req format on client/servers
478 Details : Use new req_capsule API [almost] everywhere instead of old PTLRPC
479 buffers and swabbers approach..
484 Description: drop join file stuff.
486 Severity : enhancement
488 Description: Add 2.6.27 and SLES11 (patchless client) support
492 Description: Fix in ptlrpc_expire_one_request() to print the signed time
495 Severity : enhancement
497 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
500 Severity : enhancement
502 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
504 Severity : enhancement
506 Description: correctly handle big reply message.
507 Details : send LNet event if reply is bigger than buffer and adjust this buffer
512 Description: Drop unnecessary __GFP_NOMEMALLOC flag from filter_get_page()
514 Severity : enhancement
516 Description: Add support for OEL5.
520 Description: Conf-sanity.sh 50g test - deactivated OST should not cause a panic.
522 Severity : enhancement
524 Description: "weak-modules" support
525 Details : Implement "weak-modules" support which enables kernel modules
526 to be used with any kernel that implements the same kABI. In
527 order to achieve this modules are now installed in
528 /lib/modules/$(uname -r)/updates/kernel on all distributions.
532 Description: client could not reconnect to OST because of an active request.
533 Details : The client now retries to reconnect to the same server, if a
534 connect request failed with EBUSY or -EAGAIN. Test case 35b
535 conf-sanity.sh has been created.
539 Description: Don't allow make backward step on assignin osc next id.
540 Details : race between allocation next id and ll_sync thread can be cause
541 of set wrong osc next id and can be kill valid ost objects.
545 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
549 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
553 Description: Error handling in osc_statfs_interpret() has been improved.
554 Details : Check in osc_statfs_interpret() for EBADR.
558 Description: Do not send statfs() requests to OSTs disabled by administrator.
559 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
561 Severity : enhancement
563 Description: Added sanity-quota test to for lfs(1) to work with numeric uid/gid.
565 Severity : enhancement
567 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
569 Severity : enhancement
571 Description: Enhancement for lfs(1) command to use numeric uid/gid.
575 Description: Improved cleanup procedure for conf-sanity test_46a.
579 Description: Do not start bulk transfer, if client tried to reconnect.
580 Details : Do not start bulk transfer, if client tried to reconnect,
581 but failed due to active rpcs. Check for exp_abort_active_req
582 has been implemented.
586 Description: Prevent inconsistences between linux and lustre mount structures.
587 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
588 Make the sleep interruptible.
590 Severity : enhancement
592 Description: provide server to client comms path
593 Details : server to client communications path via new LDLM_SET_INFO rpc
597 Description: truncate starts GFP_FS allocation under transaction causing deadlock
598 Details : ldiskfs_truncate calls grab_cache_page which may start page
599 allocation under an open transaction. This may lead to
600 calling prune_icache with consequent lustre reentrance.
604 Description: send correctly lsm on open replay
605 Details : MDS is trust to LSM size on replay open, but client can set wrong size
610 Description: enable client interface failover
611 Details : When a child reconnects from another NID, properly update export
612 nid hash position and ldlm reverse import.
614 Severity : enhancement
616 Description: Communicate OST degraded/readonly state via statfs to MDS
617 Details : Flags in the statfs returned from OSTs indicate whether the
618 OST is in a degraded RAID state, or if the filesystem has
619 turned read-only after a filesystem error is detected.
624 Description: don't panic if EPROTO was hit when reading symlink
625 Details : correctly handling request reference in error cases.
627 Severity : enhancement
629 Description: Add LustreNetLink, a kernel-userspace communcation path.
631 Severity : enhancement
633 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
638 Description: access to llog context before init.
639 Details : move handling CATALOGS file at osc layer and forbid access to llog
643 Frequency : rare, only when client is writing to OST on local node
645 Description: Avoid deadlock for local client writes
646 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
647 memory freeing context. This allows OST threads to set the
648 PF_MEMALLOC flag on task structures in order to allocate memory
649 from reserved pools and complete IO.
650 Use GFP_HIGHUSER for OST allocations for non-local client writes,
651 so that the OST threads generate memory pressure and allow
652 inactive pages to be reclaimed.
654 Severity : enhancement
656 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
661 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
662 Details : move ldlm namespace creation in setup phase to avoid grab
663 _lprocfs_lock with cli_sem held.
667 Description: Temporarily disable grant shrink.
668 Details : Disable the feature for debugging.
671 Frequency : only when formatting test filesystems
673 Description: Unable to run several mkfs.lustre on loop devices at the same time
674 Details : mkfs.lustre returns error 256 on the concurrent loop devices
675 formatting. The solution is to proper handle the error.
677 Severity : enhancement
679 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
681 Severity : enhancement
683 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
688 Description: resolve race between obd_disconnect and class_disconnect_exports
689 Details : if obd_disconnect will be called to already disconnected export he
690 forget release one reference and osc module can't unloaded.
692 Severity : enhancement
694 Description: Allow tuning service thread via /proc
695 Details : For each service a new
696 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
697 created that can be used to set min/max thread counts, and get the
698 current number of running threads.
703 Descriptoin: don't return error if have partially created objects for file.
704 Details : lov_update_create_set uses set->set_success as index for created
705 objects, so if some requests will be failed, they will have hole
706 at end of array and qos_shrink_lsm for allocate correct lsm.
708 Severity : enhancement
710 Description: Update OFED support to 1.4.1
715 Description: Reduce small size read RPC
716 Details : Set read-ahead limit for every file and only do read-ahead when
717 available read-ahead pages are bigger than 1M to avoid small size
720 Severity : enhancement
722 Description: move AT tunable parameters for more consistent usage
723 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
725 Severity : enhancement
727 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
728 Details : allow skip disconnected ost for send statfs request and hide error
734 Description: fix racy locking of mballoc block bitmaps causing BUG
735 Details : The locking of the mballoc buddy bitmap and the in-memory
736 block bitmap was using two different spin locks in some
737 cases. This made it possible to incorrectly access the
738 mballoc bitmap while another process was modifying it,
739 causing a sanity assertion to fail. While no on-disk corruption
740 was reported, there was some risk of this happening.
743 Frequency : rare, on llog test 6
745 Descriptoin: don't allow connect to already connected import
746 Details : allowing connect to already connected import is hide connecting problem.
751 Descriptoin: allow kill process which wait statahead result
752 Details : for some reasons 'ls' can stick in waiting result from statahead,
753 in this case need way for kill this process.
755 Severity : enhancement
757 Description: Add state history info file, enhance import info file
758 Details : Track import connection state changes in a new osc/mdc proc file;
759 add overview-type data to the osc/mdc import proc file.
761 Severity : enhancement
763 Description: MDS create should not wait for statfs RPC while holding DLM lock.
765 Severity : enhancement
767 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
772 Descriptoin: OST grant too much space to client even there are not enough space.
773 Details : Client will shrink its grant cache to OST if there are no write
774 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
775 this grant cache if there are already not enough avaible space
776 (left_space < total_clients * 32M).
779 Frequency : start MDS on uncleanly shutdowned MDS device
781 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
782 Details : stay in waiting mds<>ost recovery finished produce random bugs
783 due race between two ll_sync thread for one lov target. send
784 ACTIVATE event only if connect realy finished and import have
788 Frequency : rare, connect and disconnect target at same time
790 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
791 Details : don't call obd_disconnect under lov_lock. this long time
792 operation and can block ptlrpcd which answer to connect request.
797 Descriptoin: don't lose wakeup for imp_recovery_waitq
798 Details : recover_import_no_retry or invalidate_import and import_close can
799 both sleep on imp_recovery_waitq, but we was send only one wakeup
803 Frequency : always with long access acl
805 Descriptoin: mds can't pack reply with long acl.
806 Details : mds don't control size of acl but they limited by reint/getattr
809 Severity : enhancement
811 Description: Update to SLES10 kernel-2.6.16.60-0.33.
813 Severity : enhancement
815 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
818 Frequency : start MDS on uncleanly shutdowned MDS device
820 Descriptoin: aborting recovery hang on MDS
821 Details : don't throttle destroy RPCs for the MDT.
824 Frequency : on remount
826 Description: external journal device not working after the remount
827 Details : clear dev_rdonly flag for external journal devices in
833 Description: shutdown vs evict race
834 Details : client_disconnect_export vs connect request race.
835 if client will evicted at this time - we start invalidate
836 thread without referece to import and import can be freed
840 Frequency : rare, need acl's on inode.
842 Description: client can't handle ost additional correctly
843 Details : if ost was added after client connected to mds client can have
844 hit lnet_try_match_md ... to big messages to wide striped files.
845 in this case need teach client to handle config events about add
846 lov target and update client max ea size at that event.
848 Severity : enhancement
850 Description: Changelogs
851 Details : Changelogs are a lightweight mechanism to track filesystem
852 metadata and namespace changes. The changelog is recorded
853 permanently on the MDTs, and is periodically "consumed" / purged
854 when records are no longer needed.
856 Severity : enhancement
858 Description: compact fld format with extents
859 Details : Store range of seq rather than every seq in FLD. Seq
860 controller update FLD rather than clients. In Case of CMD, mdt0
861 has FLD, all other metadata server act as non persistent proxy
862 for FLD queries and cache fld entries in fld cache.
867 Description: don't skip ost target if they assigned to file
868 Details : Drop slow OSCs if we can, but not for requested start idx.
869 This means "if OSC is slow and it is not the requested
870 start OST, then it can be skipped, otherwise skip it only
871 if it is inactive/recovering/out-of-space.
875 Description: more cleanup in mds_lov
876 Details : not send LOV EA under replay, we can't know about they size at this
877 time. Don't allow client connect to mds before any ost connected,
878 for avoid problems with LOV EA size and returning EIO to client.
880 Severity : enhancement
882 Description: Interoperability at server side (Disk interoperability)
884 Severity : enhancement
886 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
888 Severity : enhancement
890 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
892 Severity : enhancement
894 Description: New client IO stack (CLIO).
896 Severity : enhancement
898 Description: Commit on sharing. Eliminate inter-client dependencies between
899 uncommitted transactions by doing transaction commits.
900 Thereby clients may recovery independently.
903 Frequency : Create a symlink file with a very long name
905 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
906 Details : If there is no extra space in the request for early cancels,
907 ldlm_req_handles_avail() returns 0 instead of a negative value.
909 Severity : enhancement
911 Description: Add /proc entry for import status
912 Details : The mdc, osc, and mgc import directories now have
913 an import directory that contains useful import data for debugging
916 Severity : enhancement
918 Description: Re-disable certain /proc logging
919 Details : Enable and disable client's offset_stats, extents_stats and
920 extents_stats_per_process stats logging on the fly.
923 Frequency : Only on FC kernels 2.6.22+
925 Description: oops in statahead
926 Details : Do not drop reference count for the dentry from VFS when lookup,
927 VFS will do that by itself.
929 Severity : enhancement
931 Description: Generic /proc file permissions
932 Details : Set /Proc file permissions in a more generic way to enable non-
933 root users operate on some /proc files.
937 Description: Hitting mdc_commit_close() ASSERTION
938 Details : Properly handle request reference release in
939 ll_release_openhandle().
943 Description: quota recovery deadlock during mds failover
944 Details : This patch includes att18982, att18236, att18237 in bz14840.
946 1. fix osts hang when mds does failover with quotaon
947 2. prevent watchdog storm when osts threads wait for the
952 Frequency : only patchless client
953 Description: add workaround for race between add/remove dentry from hash
955 Severity : enhancement
957 Description: Allow OST glimpses to return PW locks
961 Description: LBUG when llog conf file is full
962 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
966 Description: Prevent import from entering FULL state when server in recovery
970 Description: service mount cannot take device name with ":"
971 Details : Only when device name contains ":/" will mount treat it as
977 Description: replace ptlrpcd with the statahead thread to interpret the async
978 statahead RPC callback
982 Frequency : on recovery
983 Description: I/O failures after umount during fail back
984 Details : if client reconnected to restarted server we need join to recovery
985 instead of find server handler is changed and process self
986 eviction with cancel all locks.
988 Severity : enhancement
990 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
992 Severity : enhancement
994 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
996 Severity : enhancement
998 Description: Upcall on Lustre log has been dumped
999 Details : Allow for a user mode script to be called once a Lustre log has
1000 been dumped. It passes the filename of the dumped log to the
1001 script, the location of the script can be specified via
1002 /proc/sys/lnet/debug_log_upcall.
1007 Description: avoid idr_remove called for id which is not allocated.
1008 Details : Move assigment s_dev for clustered nfs to end of initialization,
1009 to avoid problem with error handling.
1014 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
1015 Details : When connection is reused this not moved from CONN_UNUSED_HASH
1016 into CONN_USED_HASH and this prodice warning when put connection
1017 again in unused hash.
1020 Severity : enhancement
1022 Description: File striping can now be set to use an arbitrary pool of OSTs.
1024 Severity : enhancement
1026 Description: Export bytes_read/bytes_write count on OSC/OST.
1030 Description: Early reply size mismatch, MGC loses connection
1031 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
1032 the connect flags are properly negotiated.
1037 Description: quotas are not honored with O_DIRECT
1038 Details : all writes with the flag O_DIRECT will use grants which leads to
1039 this problem. Now using OBD_BRW_SYNC to guard this.
1043 Description: add quota statistics
1044 Details : 1. sort out quota proc entries and proc code.
1045 2. add quota statistics
1047 Severity : enhancement
1049 Description: enable quota support for HEAD.
1053 Description: Properly propagate oinfo flags from lov to osc for statfs
1054 Details : restore missing copy oi_flags to lov requests.
1056 Severity : enhancement
1058 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
1059 : l_getgroups(8), lst(8), routerstat(8)
1060 Details : included man pages for llobdstat(8), llstat(8),
1061 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
1063 Severity : enhancement
1065 Description: configure's --enable-quota should check the
1066 : kernel .config for CONFIG_QUOTA
1067 Details : configure is terminated if --enable-quota is passed but
1068 : no quota support is in kernel
1072 Description: Remove portals compatibility
1073 Details : Remove portals compatibility, not interoperable with releases
1078 Description: Resolve device initialization race
1079 Details : Prevent proc handler from accessing devices added to the
1080 obd_devs array but yet be intialized.
1082 Severity : enhancement
1084 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
1086 Severity : enhancement
1088 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
1093 Description: Using wrong pointer in osc_brw_prep_request
1094 Details : Access to array[-1] can produce panic if kernel compiled with
1095 CONFIG_PAGE_ALLOC enabled
1099 Description: Client runs out of low memory
1100 Details : Consider only lowmem when counting initial number of llap pages
1104 Description: *optional* service tags registration
1105 Details : if the "service tags" package is installed on a Lustre node
1106 When the filesystem is mounted, a local-node service tag will
1107 be created. See http://inventory.sun.com/ for more information
1108 about the Service Tags asset management system.
1112 Description: Kernel BUG tries to release flock
1113 Details : Lustre does not destroy flock lock before last reference goes
1114 away. So always drop flock locks when client is evicted and
1115 perform unlock regardless of successfulness of speaking to MDS.
1119 Description: add refcount protection for osc callbacks, avoid panic on shutdown
1123 Description: sanity test 65a fails if stripecount of -1 is set
1124 Details : handle -1 striping on filesystem in ll_dirstripe_verify
1129 Description: ASSERTION(CheckWriteback(page,cmd)) failed
1130 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
1133 Severity : enhancement
1135 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
1139 Description: do not process already freed flock
1140 Details : flock can possibly be freed by another thread before it reaches
1141 to ldlm_flock_completion_ast.
1145 Description: LBUG during stress test
1146 Details : Need properly lock accesses the flock deadlock detection list.
1150 Description: oops in page fault handler
1151 Details : kernel page fault handler can return two special 'pages' in error
1152 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
1156 Description: timeout with invalidate import.
1157 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
1158 should be handled by ptlrpcd. This produce long age waiting and
1159 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
1161 Severity : enhancement
1163 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
1167 Description: enable MGS and MDT services start separately
1168 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
1169 a co-located MGS without starting the MGS, which is a complement
1170 to 'nosvc' mount option.
1174 Frequency : after recovery
1175 Description: precreate to many object's after del orphan.
1176 Details : del orphan st in oscc last_id == next_id and this triger growing
1177 count of precreated objects. Set flag LOW to skip increase count
1178 of precreated objects.
1182 Frequency : rare, on clear nid stats
1183 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
1184 Details : when clean nid stats sometimes try destroy live entry,
1185 and this produce panic in free.
1189 Description: Stack overflow during MDS log replay
1190 ease stack pressure by using a thread dealing llog_process.
1194 Description: wait until IO finished before start new when do lock cancel.
1195 Details : VM protocol want old IO finished before start new, in this case
1196 need wait until PG_writeback is cleared until check dirty flag and
1197 call writepages in lock cancel callback.
1199 Severity : enhancement
1201 Description: using special macro for print time and cleanup in includes.
1205 Description: mds_mfd_close() ASSERTION(rc == 0)
1206 Details : In mds_mfd_close(), we need protect inode's writecount change
1207 within its orphan write semaphore to prevent possible races.
1211 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
1215 Frequency : rare, on shutdown ost
1216 Description: don't hit live lock with umount ost.
1217 Details : shrink_dcache_parent can be in long loop with destroy dentries,
1218 use shrink_dcache_sb instead.
1222 Description: don't panic with use echo client
1223 Details : echo client pass NULL as client nid pointer and this produce null
1224 pointer dereference.
1228 Description: fix build on ppc32
1229 Details : compile code with -m64 flag produce wrong object file for ppc32.
1233 Description: add message levels for liblustreapi
1237 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
1238 Details : Move the 'good_osts' check before the 'total_bavail' check. This
1239 will result in an -EAGAIN and in the exit call path we call
1240 alloc_rr() which will with increasing aggressiveness attempt to
1241 aquire precreated objects on the minimum number of required OSCs.
1245 Description: Use old size assignment to avoid deadlock
1246 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
1247 the scheduling while holding a spinlock. We do not need locking
1248 for size in ll_update_inode() because size is only updated from
1249 the MDS for directories or files without objects, so there is no
1250 other place to do the update, and concurrent access to such inodes
1251 are protected by the inode lock.
1255 Description: resolve "_IOWR redefined" build error on SLES10
1259 Description: dump the memory debugging after all modules are unloaded to
1260 suppress false negative in conf_sanity test 39
1262 Severity : enhancement
1264 Description: build kernel-ib packages for OFED 1.3 in our release cycle
1269 Description: fix SLES kernel versioning
1270 Details : the kernel version for our SLES 10 kernel did not include a "-"
1271 before the "smp" at the end. while this was not a problem in
1272 general, it did mean that software trying to use the kernel
1273 version to try to detect a vendor specific kernel would fail.
1274 this was most evident by the OFED build scripts.
1278 Description: Don't update lov_desc members until making sure they are valid
1279 Details : When updating lov_desc members via proc fs, need fix their
1280 validities before doing the real update.
1284 Description: don't put request into delay list while invalidate in flight.
1285 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
1286 import in flight. this produce timeout for invalidate and sometimes
1287 can cause stale data.
1291 Frequency : on ppc only
1292 Description: not convert ost objects for directory because it's not exist.
1293 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
1297 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
1299 Severity : enhancement
1301 Description: Update to support 2.6.22.14 vanilla kernel.
1305 Frequency : rare, on recovery
1306 Description: read procfs can produce deadlock in some situation
1307 Details : Holding lprocfs lock which send rpc can produce block for destroy
1308 obd objects and this also block reconnect with -EALREADY. This
1309 isn't fix all lprocfs bugs - but make it rare.
1311 Severity : enhancement
1313 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
1316 Frequency : frequent on X2 node
1318 Description: mdc_set_open_replay_data LBUG
1319 Details : Set replay data for requests that are eligible for replay.
1323 Description: lustre_mgs: operation 101 on unconnected MGS
1324 Details : When MGC is disconnected from MGS long enough, MGS will evict the
1325 MGC, and late on MGC cannot successfully connect to MGS and a lot
1326 of the error messages complaining that MGS is not connected.
1330 Frequency : on network error
1331 Description: panic with double free request if network error
1332 Details : mdc_finish_enqueue is finish request if any network error ocuring,
1333 but it's true only for synchronus enqueue, for async enqueue
1334 (via ptlrpcd) this incorrect and ptlrpcd want finish request
1337 Severity : enhancement
1339 Description: client-side metadata stat-ahead during readdir(directory readahead)
1340 Details : perform client-side metadata stat-ahead when the client detects
1341 readdir and sequential stat of dir entries therein
1344 Frequency : on start mds
1346 Description: Implement get_info(last_id) in obdfilter.
1349 Frequency : occasional
1351 Description: Correctly check stale fid, not start epoch if ost not support SOM
1352 Details : open with flag O_CREATE need set old fid in op_fid3 because
1353 op_fid2 was overwritten with new generated fid, but mds can answer
1354 with one of these two fids and both is not stale. Setattr
1355 incorrectly started an epoch and assume will be called
1356 done_writing, but without SOM done_writing ever being called.
1359 Frequency : rare, depends on device drivers and load
1361 Description: MDS or OSS nodes crash due to stack overflow
1362 Details : Code changes in 1.8.0 increased the stack usage of some functions.
1363 In some cases, in conjunction with device drivers that use a lot
1364 of stack the MDS (or possibly OSS) service threads could overflow
1365 the stack. One change which was identified to consume additional
1366 stack has been reworked to avoid the extra stack usage.
1369 Frequency : occasional
1371 Description: Do not fail import if osc_interpret_create gets -EAGAIN
1372 Details : If osc_interpret_create got -EAGAIN it immediately exits and
1373 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
1374 oscc_has_objects and see OSC has no objests and call
1375 oscc_internal_create to resend create request.
1377 Severity : enhancement
1379 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
1381 Severity : enhancement
1383 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
1386 Frequency : very rare
1388 Description: Fix replay if there is an un-replied request and open
1389 Details : In some cases, older replay request will revert the
1390 mcd->mcd_last_xid on MDS which is used to record the client's
1391 latest sent request.
1393 Severity : enhancement
1395 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
1397 Severity : enhancement
1399 Description: Add rhel5 support to HEAD.
1401 Serverity : enhancement
1403 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
1408 Description: Don't allow skipping OSTs if index has been specified.
1409 Details : Don't allow skipping OSTs if index has been specified, make
1410 locking in internal create lots better.
1414 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
1415 Details : access to bitfield in structure is always rounded to long
1416 and this produce problem with not atomic change any bit.
1420 Description: Lustre make rpms failed.
1421 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
1422 builds Lustre rpms from tarball.
1425 Frequency : rare on shutdown ost
1427 Description: If llog cancel was not send before clean_exports phase, this can
1428 produce deadlock in llog code.
1429 Details : If llog thread has last reference to obd and call class_import_put
1430 this produce deadlock because llog_cleanup_commit_master wait when
1431 last llog_commit_thread exited, but this never success because was
1432 called from llog_commit_thread.
1436 Description: allow userland application know is lost one of stripes.
1437 Details : fill lvb_blocks with error code on ost and return it to
1438 application if error flag found.
1442 Description: NULL lov_tgts causing MDS oops
1443 Details : more safe checks for NULL lov_tgts for avoid oops.
1445 Severity : enhancement
1447 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
1451 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
1452 Details : If an OST is down the MDS will hang indefinitely in
1453 obd_statfs_rqset() waiting for the statfs data. While for
1454 MDS QOS usage of statfs, it should not stuck in waiting.
1456 Severity : enhancement
1458 Description: remote_acl support
1459 Details : Support ACL-based permission check for remote user.
1460 Support setfacl/getfacl for remote user with the utils
1461 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
1462 the system "{s,g}etfacl" utils.
1464 Severity : enhancement
1466 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
1468 Severity : enhancement
1470 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
1474 Description: make mgs_setparam() handle fsname containing dash
1475 Details : fsname containing a dash does not work with lctl conf_param
1477 Severity : enhancement
1479 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
1481 Severity : enhancement
1483 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
1487 Frequency : rare, at shutdown
1488 Description: access already free / zero obd_namespace.
1489 Details : if client_disconnect_export was called without force flag set,
1490 and exist connect request in flight, this can produce access to
1491 NULL pointer (or already free pointer) when connect_interpret
1492 store ocd flags in obd_namespace.
1496 Frequency : only at startup
1497 Description: not alloc memory with spinlock held.
1498 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
1499 if any spinlock held.
1501 Severity : enhancement
1503 Description: make lustre randomly fail allocating memory
1504 Details : Make lustre randomly failed allocating memory for testing purpose.
1506 Severity : enhancement
1508 Description: lost problems with lov objid file
1509 Details : Fixes some scability and access to not inited memory problems
1510 in work with lov objdid file.
1515 Description: lfs find does not continue on file error
1516 Details : Continue other files processing when a file/dir is absent.
1520 Description: Inconsistent usage of lustre_pack_reply()
1521 Details : Standardize the usage of lustre_pack_reply() such that it
1522 always generate a CERROR on failure.
1525 Frequency : occasional
1527 Description: MDS hang or stay in waiting lock
1528 Details : If client receive lock with CBPENDING flag ldlm need send lock
1529 cancel as separate rpc, to avoid situation when cancel request
1530 can't processed due all i/o threads stay in wait lock.
1534 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1538 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1542 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1546 Description: Update to RHEL4 latest kernel.
1548 Severity : enhancement
1550 Description: Build SLES10 patchless client fails
1551 Details : The configure was broken by run ./configure with
1552 --with-linux-obj=.... argument for patchless client. When the
1553 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1554 can't search header adequately. Use absolute path such as
1555 -I($LINUX)/include instead.
1559 Description: interrupt oig_wait produce painc on resend.
1560 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1561 private set, and this produce situation when rq_ptlrpcd_data not
1562 copyed to new allocated request and triggered LBUG on assert
1563 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1564 ptlrpcd set if request is changed and can be safety changed to use
1569 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1570 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1571 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1572 these macros and replace them with their difinitions instead.
1576 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1577 Details : Update which_patch & target file for SLES10 latest kernel.
1580 Description: Remove client patches from SLES 10 kernel.
1581 Details : This causes SLES 10 clients to behave as patchless clients
1582 even on a Lustre-patched (server) kernel.
1584 Severity : enhancement
1586 Description: self-adjustable client's lru lists
1587 Details : use adaptive algorithm for managing client cached locks lru
1588 lists according to current server load, other client's work
1589 pattern, memory activities, etc. Both, server and client
1590 side namespaces provide number of proc tunables for controlling
1593 Severity : enhancement
1595 Description: light-weight GSS support
1596 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
1597 and reduce performance overhead.
1599 Severity : enhancement
1601 Description: Linux keyring support
1602 Details : Support using service of Linux keyring for Lustre GSS internal
1603 context refresh/cache mechanism.
1607 Description: Fix errors in lfs documentation
1608 Details : Fixes man pages
1612 Description: don't use GFP_* in generic Lustre code.
1613 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
1617 Description: obdclass is limited by single OBD_ALLOC(idarray)
1618 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
1622 Description: warnings with build patchless client with vanila 2.6.19 and up
1623 Details : change old ctl_table style and replace ctl_table/ctl_table_header
1624 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
1628 Frequency : Only for SLES
1629 Description: sanity_quota fail test_1
1630 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
1631 file, which makes TSTID[2] inunique.
1635 Frequency : Only for SLES9
1636 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
1637 Details : Update target/ChangeLog/which_patch .
1641 Frequency : Only for test_10 in sanity-quota.sh
1642 Description: a bug in quota test code
1643 Details : checking a test flag in a wrong place
1647 Frequency : Only for test_13 in sanity-quota.sh
1648 Description: a bug in quota test script
1649 Details : shouldn't check overfull allocated quota
1653 Description: jbd statistics
1654 Details : Port older jbd statistics patch for sles10
1656 Severity : enhancement
1658 Description: Add printing inode info into message about error in writepage.
1663 Description: reply_lock_interpret crash due to race with it and lock cancel.
1664 Details : Do not replay locks that are being cancelled. Do not reference
1665 locks by their address during replay, just by their handle.
1669 Frequency : When flocks are used.
1670 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
1671 Details : Flock locks might destroy just granted lock if it could be merged
1672 with another existing flock, this is done in completion handler,
1673 so teach ldlm_cli_enquque_fini that this is a valid case for
1679 Description: Oops in read and write path when failing to allocate lock.
1680 Details : Check if lock allocation failed and return error back.
1684 Description: lstripe command fails for valid OST index
1685 Details : The stripe offset is compared to lov->desc.ld_tgt_count
1686 instead of lov->desc.ld_active_tgt_count.
1690 Description: sanity.sh failed test 103
1691 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
1696 Description: df doesn't work properly if diskfs blocksize != 4K
1697 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
1701 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
1702 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
1704 Severity : enhancement
1706 Description: Adaptive timeouts
1707 Details : RPC timeouts adapt to changing server load and network
1708 conditions to reduce resend attempts and improve recovery time.
1712 Description: llapi_file_create() does not allow some changes
1713 Details : add llapi_file_open() that allows specifying the mode and
1714 open flags, and also returns an open file handle.
1718 Description: merge and cleanup kernel patches.
1719 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
1723 Description: Add journal checksum support.(Kernel part)
1724 Details : The journal checksum feature adds two new flags i.e
1725 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
1726 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
1727 indicates that the commit block contains the checksum for
1728 the blocks described by the descriptor blocks. Now commit
1729 record can be sent to disk without waiting for descriptor
1730 blocks to be written to disk. This behavior is controlled
1731 using JBD2_FEATURE_ASYNC_COMMIT flag.
1735 Description: OSS needs mutliple precreate threads
1736 Details : Add ability to start more than one create thread per OSS.
1740 Description: Sanity.sh test_65k failed: setstripe should have succeeded
1741 Details : execute lfs setstripe on client
1745 Description: mds_obd_create error creating tmp object
1746 Details : When the user sets quota on root, llog will be affected and can't
1747 create files and write files.
1750 Frequency : Always on ia64 patchless client, and possibly others.
1752 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
1753 Details : This allows the patchless client to be loaded on architectures
1754 without this export.
1758 Description: RedHat Update kernel for RHEL5
1759 Details : Add the kernel config file for RHEL5.
1763 Description: RedHat Update kernel for RHEL5
1764 Details : Modify the kernel config file more closer RHEL5.
1768 Description: Build failure against Centos5 (RHEL5)
1769 Details : Define PAGE_SIZE when it isn't present.
1773 Description: test_7 on sanity-quota.sh can't be used on separate machines
1774 Details : add facet to handle it
1778 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
1779 Details : with NFS, the anon dentry's parent was set to itself in
1780 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
1781 corresponding dentry other than use rec->ur_name.
1783 Severity : enhancement
1785 Description: lfs setstripe enhancement
1786 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
1790 Description: enable data checksumming by default
1791 Details : enable checksum by default, allow --disable-checksum
1792 configure option and "-o nochecksum" mount option. Checksums
1793 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
1797 Description: lustre support for RHEL5
1798 Details : Add support for RHEL5.
1802 Description: Client eviction due to failover config
1803 Details : after a connection loss, the lustre client should attempt to
1804 reconnect to the last active server first before trying the
1805 other potential connections.
1809 Description: when mds and osts use different quota unit(32bit and 64bit),
1810 quota will be released repeatly.
1811 Details : void sending multiple quota reqs to mds, which will keep the
1812 status between the reqs.
1816 Description: osts not allocated evenly to files
1817 Details : change the condition to increase offset_idx
1820 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
1822 Description: Data corruption for OSTs that are formatted larger than 2TB
1824 Details : When generating the bio request for lustre file writes the
1825 sector number would overflow a temporary variable before being
1826 used for the IO. The data reads correctly from Lustre (which
1827 will overflow in a similar manner) but other file data or
1828 filesystem metadata may be corrupted in some cases.
1832 Description: Tune the kernel for good SCSI performance.
1833 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
1834 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
1839 Description: rewrite ext2-derived code in obdclass/uuid.c
1840 Details : rewrite inherited code (uuid parsing code from ext2 utils)
1841 from scratch preserving functionality.
1845 Description: "lfs find -obd UUID" prints directories
1846 Details : "lfs find -obd UUID" will return all directory names instead
1847 of just file names. It is incorrect because the directories
1848 do not reside on the OSTs.
1852 Description: use i_size_read and i_size_write in 2.6 port
1853 Details : replace inode->i_size access with i_size_read/write()
1856 Frequency : when removing large files
1858 Description: scheduling issue during removal of large Lustre files
1859 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
1860 It causes scheduling issues when removing large files (17TB in the
1864 Frequency : only with liblustre clients on XT3
1866 Description: evictions taking too long
1867 Details : allow llrd to evict clients directly on OSTs
1870 Frequency : only on ppc
1872 Description: /proc/fs/lustre/devices broken on ppc
1873 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
1874 We should make sure the type of 'index' is loff_t and then cast
1875 explicitly as needed below. Do not assign an explicitly cast
1879 Frequency : only for rhel5
1881 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
1882 Details : Modify the target file & which_kernel.
1884 Severity : enhancement
1886 Description: omit set fsid for export NFS
1887 Details : fix set/restore device id for avoid EMFILE error and mark lustre
1888 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
1892 Frequency : Always, for kernels after 2.6.16
1893 Description: Fix warning idr_remove called for id=.. which is not allocated.
1894 Details : Last kernels save old s_dev before kill super and not allow
1895 to restore from callback - restore it before call kill_anon_super.
1899 Description: buffer overruns could theoretically occur
1900 Details : llapi_semantic_traverse() modifies the "path" argument by
1901 appending values to the end of the origin string, and a
1902 overrun may occur. Adding buffer overrun check in liblustreapi.
1906 Description: Fix error on 'ls .' at the top of the Lustre mount.
1907 Details : Don't revalidate dentry if it is a root dentry.
1911 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
1912 Details : Modify vm-tunables-rhel4.patch.
1916 Description: Kernel config for 2.6.18-vanilla.
1917 Details : Modify targets/2.6-vanilla.target.in.
1918 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
1919 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
1920 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
1921 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
1925 Description: improve handling recoverable errors
1926 Details : if request processig with error which can be recoverable on server
1927 request should be resend, otherwise page released from cache and
1932 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
1933 Details : Modify target file & which_patch.
1934 A flaw was found in the IA32 system call emulation provided
1935 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
1936 value could be stored in the %RAX register, which could trigger an
1937 out-of-bounds system call table access. An untrusted local user
1938 could exploit this flaw to run code in the kernel
1939 (ie a root privilege escalation). (CVE-2007-4573).
1943 Description: O_DIRECT bypasses client statistics.
1944 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
1945 read_ahead_stats were updated. copied the stats section from
1946 osc_send_oap_rpc() into async_internal().
1950 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
1954 Description: change order of libsysio includes
1955 Details : '#include sysio.h' should always come before '#include xtio.h'
1959 Description: handle "serious error: objid * already exists" more gracefully
1960 Details : If LAST_ID value on disk is smaller than the objects existing in
1961 the O/0/d* directories, it indicates disk corruption and causes an
1962 LBUG(). If the object is 0-length, then we should use the existing
1963 object. This will help to avoid a full fsck in most cases.
1965 Severity : enhancement
1967 Description: adapt the lustre_config script to support the upgrade case
1968 Details : Add "-u" option for lustre_config script to support upgrading 1.4
1969 server targets to 1.6 in parallel.
1973 Description: To avoid grant space > avaible space when the disk is almost
1974 full. Without this patch you might see the error "grant XXXX >
1975 available" or some LBUG about grant, when the disk is almost
1977 Details : In filter_check_grant, for non_grant cache write, we should
1978 check the left space by if (*left > ungranted + bytes), instead
1979 of (*left > ungranted), because only we are sure the left space
1980 is enough for another "bytes", then the ungrant space should be
1981 increase. In client, we should update cl_avail_grant only there
1982 is OBD_MD_FLGRANT in the reply.
1986 Description: Update RHEL 4 kernel to fix local root privilege escalation.
1987 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
1988 described in CVE-2007-4573. This problem could allow untrusted
1989 local users to gain root access.
1992 Frequency : when using O_DIRECT and quotas
1994 Description: Incorrect file ownership on O_DIRECT output files
1995 Details : block usage reported by 'lfs quota' does not take into account
1996 files that have been written with O_DIRECT.
2001 Description: touch file failed when fs is not full
2002 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
2003 otherwise we can get ENOSP while fs is not full.
2007 Description: parallel lock callbacks
2008 Details : Instead of sending blocking and completion callbacks as separated
2009 requests, adding them to a set and sending in parallel.
2012 Frequency : only for Cray XT3
2013 Bugzilla : 12829/13455
2014 Description: Changing primary group doesn't change the group lustre assigns to
2016 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
2017 overridden with the primary group provided by the group upcall,
2018 whereas the supplied fsgid can be trusted if it is in the list of
2019 supplementary groups returned by the group upcall.
2021 Severity : enhancement
2023 Description: Allow masking D_WARNING, D_ERROR messages from console
2024 Details : Console messages can now be disabled via lnet.printk.
2028 Description: User code with malformed file open parameter crashes client node
2029 Details : Before packing join_file req, all the related reference should be
2030 checked carefully in case some malformed flags cause fake
2031 join_file req on client.
2035 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
2036 Details : ldlm_completion_ast() assumes that a lock is granted when the req
2037 mode is equal to the granted mode. However, it should also check
2038 that LDLM_FL_CP_REQD is not set.
2042 Description: Heavy nfs access might result in deadlocks
2043 Details : After ELC code landed, it is now improper to enqueue any mds
2044 locks under och_sem, because enqueue might want to decide to
2045 cancel open locks for same inode we are holding och_sem for.
2049 Description: Client eviction while running blogbench
2050 Details : A lot of unlink operations with concurrent I/O can lead to a
2051 deadlock causing evictions. To address the problem, the number of
2052 oustanding OST_DESTROY requests is now throttled to
2053 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
2054 callbacks are processed in priority.
2058 Description: enable ACLs on MDS by default
2059 Details : ACLs must be enabled on MDS by default.
2062 Frequency : PPC/PPC64 only
2064 Description: conflicts between asm-ppc64/types.h and lustre_types.h
2065 Details : fix duplicated definitions between asm-ppc64/types.h and
2066 lustre_types.h on PPC.
2069 Frequency : PPC/PPC64 only
2071 Description: asm-ppc/segment.h does not exist
2072 Details : fix compile issue on PPC.
2076 Description: better handle error messages in extents code
2079 Frequency : RHEL4 only
2081 Description: mkfs is very slow on IA64/RHEL4
2082 Details : A performance regression has been discovered in the MPT Fusion
2083 driver between versions 3.02.73rh and 3.02.99.00rh. As a
2084 consequence, we have downgraded the MPT Fusion driver in the RHEL4
2085 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
2087 Severity : enhancement
2089 Description: SNMP support enhancement
2090 Details : Adding total number of sampled request for an MDS node in snmp
2093 Severity : enhancement
2095 Description: Optimize ldlm waiting list processing for PR extent locks
2096 Details : When processing waiting list for read extent lock and meeting read
2097 lock that is same or wider to it that is not contended, skip
2098 processing rest of the list and immediatelly return current
2099 status of conflictness, since we are guaranteed there are no
2100 conflicting locks in the rest of the list.
2104 Description: Time out and refuse to reconnect
2105 Details : When the failover node is the primary node, it is possible
2106 to have two identical connections in imp_conn_list. We must
2107 compare not conn's pointers but NIDs, otherwise we can defeat
2108 connection throttling.
2112 Description: port llog fixes from b1_6 into HEAD
2113 Details : Port llog reference couting and some llog cleanups from b1_6
2114 (bug 10800) into HEAD, for protect from panic and access to already
2115 free llog structures.
2119 Description: Detect stride IO mode in read-ahead
2120 Details : When a client does stride read, read-ahead should detect that and
2121 read-ahead pages according to the detected stride pattern.
2125 Description: data checksumming impacts single node performance
2126 Details : add support for several checksum algorithm. Currently, only CRC32
2127 and Adler-32 are supported. The checksum type can be changed on
2128 the fly via /proc/fs/lustre/osc/*/checksum_type.
2132 Description: use adler32 for page checksums
2133 Details : when available, use the Adler-32 algorithm instead of CRC32 for
2138 Description: build for x2 fails
2139 Details : fix compile issue on Cray systems.
2143 Description: Properly match for duplicate locks
2144 Details : Due to different lock order from skiplists code, we need to
2145 traverse entire list for now
2148 Frequency : only on PPC/SLES10
2150 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
2151 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
2152 be defined. Add a hack in mkfs_lustre.c to work around this compile
2157 Description: LASSERT on MDS when client holding flock lock dies
2158 Details : ldlm pool logic depends on number of granted locks equal to
2159 number of released locks which is not true for flock locks, so
2160 just exclude such locks from consideration.
2164 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
2165 Details : Use fsfilt_sync() for both the whole filesystem sync and
2166 individual file sync to eliminate dangerous inode locking
2167 with I_LOCK that can lead to a deadlock.
2171 Description: performance in 1.6.3
2172 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
2173 sure that 1MB requests are not fragmented by the block layer.
2175 Severity : enhancement
2177 Description: organize the server-side client stats on per-nid basis
2178 Details : Change the structure of stats under obdfilter and mds to
2186 The "uuid"s file would list the uuids of _active_ exports.
2187 And the clear entry is to clear all stats and stale nids.
2189 Severity : enhancement
2191 Description: eliminate client locks in face of contention
2192 Details : file contention detection and lockless i/o implementation
2193 for contended files.
2197 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
2200 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
2202 Description: MDS may lose file striping (and hence file data) in some cases
2203 Details : If there are additional extended attributes stored on the MDS,
2204 in particular ACLs, SELinux, or user attributes (if user_xattr
2205 is specified for the client mount options) then there is a risk
2206 of attribute loss. Additionally, the Lustre file striping
2207 needs to be larger than default (e.g. striped over all OSTs),
2208 and an additional attribute must be stored initially in the
2209 inode and then increase in size enough to be moved to the
2210 external attribute block (e.g. ACL growing in size) for file
2215 Description: skiplist implementation simplification
2216 Details : skiplists are used to group compatible locks on granted list
2217 that was implemented as tracking first and last lock of each
2218 lock group the patch changes that to using doubly linked lists
2222 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
2223 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
2224 because LASSERT on both the data supplied by a client, and the
2225 data on disk is dangerous and incorrect.
2227 Severity : enhancement
2229 Description: Slow truncate/writes to huge files at high offsets.
2230 Details : Directly associate cached pages to lock that protect those pages,
2231 this allows us to quickly find what pages to write and remove
2232 once lock callback is received.
2236 Description: more ldlm soft lockups
2237 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
2238 starve other threads from the resource lock for a long time in
2239 case of long waiting queue, so change the debug level from
2240 D_OTHER to the less frequently used D_INFO.
2242 Severity : enhancement
2244 Description: add -gid, -group, -uid, -user options to lfs find
2248 Description: Hung threads in invalidate_inode_pages2_range
2249 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
2250 once one is completed. As a result, some RPCs are stuck in the
2251 queue and are never sent.
2255 Description: filter threads hungs on waiting journal commit
2256 Details : Cleanup filter group llog code, then only filter group llog will
2257 be only created in the MDS/OST syncing process.
2261 Description: Procfs and llog threads access destoryed import sometimes.
2262 Details : Sync the import destoryed process with procfs and llog threads by
2263 the import refcount and semaphore.
2265 Severity : enhancement
2267 Description: openlock cache of b1_6 port to HEAD
2272 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
2273 Details : If insertion of an extent fails, then discard the inode
2274 preallocation and free data blocks else it can lead to duplicate
2279 Description: don't always update ctime in ext3_xattr_set_handle()
2280 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
2281 In some cases the ctime should not be updated, for example for
2282 2.0->1.8 compatibility it is necessary to delete an xattr and it
2283 should not update the ctime.
2287 Bugzilla : 15713/16362
2288 Description: Assertion in iopen_connect_dentry in 1.6.3
2289 Details : looking up an inode via iopen with the wrong generation number can
2290 populate the dcache with a disconneced dentry while the inode
2291 number is in the process of being reallocated. This causes an
2292 assertion failure in iopen since the inode's dentry list contains
2293 both a connected and disconnected dentry.
2297 Description: assertion failure in ldlm_handle2lock()
2298 Details : fix a race between class_handle_unhash() and class_handle2object()
2299 introduced in lustre 1.6.5 by bug 13622.
2304 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
2305 Details : remove the limit of 256 scsi disks in the sd_iostat patch
2310 Description: oops in sd_iostats_seq_show()
2311 Details : unloading/reloading the scsi low level driver triggers a kernel
2312 bug when trying to access the sd iostat file.
2317 Description: Kernel panics during QLogic driver reload
2318 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
2319 patch, causing memory corruption.
2324 Description: journal_dev option does not work in b1_6
2325 Details : pass mount option during pre-mount.
2327 Severity : enhancement
2329 Description: Add a FIEMAP(FIle Extent MAP) ioctl
2330 Details : FIEMAP ioctl will allow an application to efficiently fetch the
2331 extent information of a file. It can be used to map logical blocks
2332 in a file to physical blocks in the block device.
2336 Description: LDLM soft lockups - improvement
2337 Details : It is be possible to send the lock handle along with each read
2338 or write request because the client is already doing a lock match
2339 itself so there isn't any reason the OST should have to re-do that
2343 Frequency : only on Cray X2
2345 Description: X2 build failures
2346 Details : fix build failures on Cray X2.
2350 Description: xid & resent requests
2351 Details : Initialize RPC XID from clock at startup (randomly if clock is
2354 Severity : enhancement
2356 Description: Add lustre_start utility to start or stop multiple Lustre servers
2361 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
2362 Details : In case of memory pressure, list_del() can be called twice on
2363 req->rq_history_list, causing a kernel oops.
2367 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
2368 Details : fix stack overflow in the distributed lock manager by defering
2369 export eviction after a failed AST to the elt thread instead of
2370 handling it in the dlm interpret routine.
2374 Description: Convert some comments to new format.
2375 Details : Update documenting comments to match doxygen conventions.
2379 Description: Grammar fixes.
2380 Details : A couple of trivial spelling fixes.
2384 Description: OSD_COUNTERS-mandatory
2385 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
2386 to simplify the code.
2390 Description: simplify cmm_device freeing logic.
2391 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
2395 Description: Add lockdep support to dt_object_operations locking interface.
2396 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
2397 parameter indicating lock ordering. Update mdd code to use new
2402 Description: Introduce failloc constants for lockless IO tests.
2403 Details : Add two new failloc constants to test lockless IO. Only one of
2404 them in implemented---another is checked in yet to be landed
2409 Description: Add lockdep support for inode mutex.
2410 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
2411 in the situations where Lustre has to lock more than one inode
2416 Description: Add optional invariants checking support.
2417 Details : Add new LINVRNT() macro, optional on new --enable-invariants
2418 configure switch. This macro is to be used for consistency and
2419 sanity checks that are too expensive to be left in `production'
2424 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
2425 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
2430 Description: Add lockdep support for ldlm_lock and ldlm_resource.
2431 Details : Use spin_lock_nested() in (the only) situation where more than
2432 one ldlm_lock is locked simultaneously. Also, fix possible
2433 dead-lock in ldlm_lock_change_resource() by enforcing particular
2438 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
2439 Details : Instead of specifying each ldlm_lock call-back through separate
2440 parameter, wrap them into struct ldlm_callback_suite.
2444 Description: Kill join_lru obd method and its callers.
2445 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
2446 in memory---a requirement implemented through ->o_join_lru() obd
2447 method in HEAD. Remove this method and its users.
2451 Description: Add asynchronous ldlm ENQUEUE completion handler.
2452 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
2453 case that stock ldlm_completion_ast() cannot handle as it waits
2454 until lock is granted. Introduce new ldlm_completion_ast_async()
2455 for this. Also comment ldlm_completion_ast().
2459 Description: ldlm_error <-> errno conversion.
2460 Details : Add functions to map (rather arbitrary) between LDLM error codes
2461 and standard errno values. CLIO needs this to prevent LDLM specific
2462 constants from escaping ldlm and osc.
2466 Description: Kill unused ldlm_handle2lock_ns() function.
2467 Details : Kill unused ldlm_handle2lock_ns() function.
2471 Description: Add lu_ref support to ldlm_lock
2472 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
2473 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
2474 and ldlm_resource. LDLM interface has to be changed, because code that
2475 releases a reference on a lock, has to "know" what reference this is.
2476 In the most frequent case
2478 lock = ldlm_handle2lock(handle);
2480 LDLM_LOCK_PUT(lock);
2482 no changes are required. When any other reference (received _not_ from
2483 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
2484 instead of LDLM_LOCK_PUT().
2486 Arguably, changes are pervasive, and interface requires some discipline
2487 for proper use. On the other hand, it was very instrumental in finding
2488 a few leaked lock references.
2492 Description: Add ldlm_lock_addref_try().
2493 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
2494 attempts to addref a lock that might be being canceled
2499 Description: Add ldlm_weigh_callback().
2500 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
2501 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
2502 hard-coded `number of pages' logic.
2506 Description: Add lockdep annotations to llog code.
2507 Details : Use appropriately tagged _nested() locking calls in the places
2508 where llog takes more than one ->lgh_lock lock.
2512 Description: Add loi_kms_set().
2513 Details : Wrap kms updates into a helper function.
2517 Description: Constify instances of struct lsm_operations.
2518 Details : Constify instances of struct lsm_operations.
2522 Description: lu_conf support.
2523 Details : On a server, a file system object is uniquely identified
2524 by a fid, which is sufficient to locate and load all object
2525 state (inode). On a client, on the other hand, more data are
2526 necessary instantiate an object. Change lu_object_find() and
2527 friends to take additional `lu_conf' argument describing object.
2528 Typically this includes layout information.
2532 Description: lu_context fixes.
2533 Details : Introduce new lu_context functions that are needed on the client
2534 side, where some system threads (ptlrpcd) are shared by multiple
2535 modules, and so cannot be stopped during module shutdown.
2539 Description: Add start and stop methods to lu_device_type_operations.
2540 Details : Introduce two new methods in lu_device_type_operations, that are
2541 invoked when first instance of a given type is created and last one
2542 is destroyed respectively. This is need by CLIO.
2546 Description: Add lu_ref support to struct lu_device.
2547 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
2548 track leaked references.
2552 Description: Introduce lu_kmem_descr.
2553 Details : lu_kmem_descr and its companion interface allow to create
2554 and destroy a number of kmem caches at once.
2558 Description: Fix lu_object finalization race.
2559 Details : Fix a race between lu_object_find() finding an object and its
2560 concurrent finalization. This race is (most likely) not possible
2561 on the server, but might happen on the client.
2565 Description: Introduce lu_ref interface.
2566 Details : lu_ref is a debugging module allowing to track references to
2567 a given object. It is quite cpu expensive, and has to be
2568 explicitly enabled with --enable-lu_ref. See usage description
2573 Description: Factor lu_site procfs stats into a separate function.
2574 Details : Separate lu_site stats printing code into a separate function
2575 that can be reused on a client.
2579 Description: Constify instances of struct {lu,dt,md}_device_operations.
2580 Details : Constify instances of struct {lu,dt,md}_device_operations.
2584 Description: Introduce struct md_site and move meta-data specific parts of
2585 struct lu_site here.
2586 Details : Move md-specific fields out of struct lu_site into special struct
2587 md_site, so that lu_site can be used on a client.
2591 Description: Kill mdd_lov_destroy().
2592 Details : Remove unused mdd code.
2596 Description: Add st_block checking to multistat.c.
2597 Details : Add st_block checking to multistat.c.
2601 Description: Add lu_ref support to struct obd_device.
2602 Details : Add lu_ref tracking to obd_device.
2606 Description: Kill obd_set_fail_loc().
2607 Details : Remove unused code.
2611 Description: Add special type for ptlrpc_request interpret functions.
2612 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
2613 there. Actual usage will be in CLIO.
2617 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
2618 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
2619 doesn't work with lockdep.
2623 Description: Add rwv.c test program.
2624 Details : New testing program exercising readv(2) and writev(2) (Qian).
2628 Description: Add sendfile.c test program.
2629 Details : New testing program exercising sendfile(2) (Jay).
2633 Description: Ratelimit a message that can be very frequent.
2634 Details : Ratelimit a memory allocation failure message that can
2639 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2640 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2644 Description: Liblustre build fixes.
2645 Details : Liblustre build fixes.
2649 Description: libcfs: add cfs_{need,cond}_resched() interface.
2650 Details : libcfs: add cfs_{need,cond}_resched() definition and
2651 implementations for Linux, NT, and liblustre.
2653 Severity : enhancement
2655 Description: More exported tunables for mballoc
2656 Details : Add support for tunable preallocation window and new tunables for
2657 large/small requests
2661 Description: Detect corruption of block bitmap and checking for preallocations
2662 Details : Checks validity of on-disk block bitmap. Also it does better
2663 checking of number of applied preallocations. When corruption is
2664 found, it turns filesystem readonly to prevent further corruptions.
2668 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
2669 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
2670 window to 0. If it does hit cache miss, and read-pattern is still
2671 stride-io mode, does not reset the stride window, but also does not
2672 increase the stride window length in this case.
2676 Frequency : only for big-endian servers
2677 Description: Check if system is big-endian while mounting fs with extents feature
2678 Details : Mounting a filesystem with extents feature will fail on big-endian
2679 systems since ext3-based ldiskfs is not supported on big-endian
2680 systems. This can be over-riden with "bigendian_extents" mount option.
2682 Severity : enhancement
2684 Description: The root squash functionality
2685 Details : A security feature, which is to prevent users from being able
2686 to mount lustre on their desktop, run as root, and delete
2687 all of the files in the filesystem. The goal is accomplished by
2688 remapping user id (UID) and group id (GID) of the root user to
2689 a UID and GID specified by the system administartor via Lustre
2690 configuration management server (MGS). The functionality also
2691 allows to specify sets of clients for which the remapping does
2696 Description: Excessive recovery window
2697 Details : With AT enabled, the recovery window can be excessively long (6000+
2698 seconds). To address this problem, we no longer use
2699 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
2700 timeout no longer depends on the service time, it is set to
2701 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
2702 time via pb_service_time.
2706 Description: Watchdog triggered on MDS failover
2707 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
2708 the OSTs know that the MDS "UUID" can be reused for the same export
2709 from a different NID, so we do not need to wait for the export to be
2713 Frequency : rare, only if using MMP with Linux RAID
2715 Description: MMP doesn't work with Linux RAID
2716 Details : While using HA for Lustre servers with Linux RAID, it is possible
2717 that MMP will not detect multiple mounts. To make this work we
2718 need to unplug the device queue in RAID when the MMP block is being
2719 written. Also while reading the MMP block, we should read it from
2720 disk and not the cached one.
2722 Severity : enhancement
2724 Description: open file using fid
2725 Details : A file can be opened using just its fid, like
2726 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
2729 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
2732 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
2733 Details : add proper locking for creation of jbd_4k slab cache
2737 Description: MMP check in ext3_remount() fails without displaying any error
2738 Details : When multiple mount protection fails during remount, proper error
2741 Severity : enhancement
2743 Description: Allow stripe size to be up to 4G-64k
2744 Details : Fix math logic to allow large stripe sizes.
2748 Description: add check for >8TB ldiskfs filesystems
2749 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
2750 Don't allow >8TB ldiskfs filesystems to be mounted without
2751 force_over_8tb mount option
2756 Descriptoin: OSS complains "too many blocks" and crashes
2757 Details : In filter_commitrw_write():
2758 1. lquota_chkquota();
2759 2. obdo_from_inode();
2760 3. lquota_pending_commit();
2761 "2." has changed uid and gid in this case, then caused
2762 lquota_chkquota() and lquota_pending_commit() to use
2763 different uid and gid. That is the root of the bug.
2767 Descriptoin: getstripe doesn't report stripe count = -1 sanely
2768 Details : Detect and print out the 'stripe count = -1' case properly.
2772 Description: Out or order replies might be lost on replay
2773 Details : In ptlrpc_retain_replayable_request if we cannot find retained
2774 request with tid smaller than one currently being added, add it
2775 to the start, not end of the list.
2777 --------------------------------------------------------------------------------
2779 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
2781 * Support for kernels:
2782 2.6.5-7.283 (SLES 9),
2783 2.6.9-55.EL (RHEL 4),
2784 2.6.16.46-0.14 (SLES 10),
2785 2.6.18.8 vanilla (kernel.org)
2786 * Client support for unpatched kernels:
2787 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2788 2.6.16 - 2.6.22 vanilla (kernel.org)
2789 * Due to recently discovered recovery problems, we do not recommend
2790 using patchless RHEL 4 clients with this or any earlier release.
2791 * Recommended e2fsprogs version: 1.39.cfs8
2792 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2793 * Starting with this release, the ldiskfs backing filesystem required
2794 by Lustre is now in its own package, lustre-ldiskfs. This package
2795 should be installed. It is versioned separately from Lustre and
2796 may be released separately in future.
2800 Description: block reactivating mgc import until all deactivates complete
2801 Details : Fix race when failing back MDT/MGS to itself (testing)
2804 Frequency : at statup only
2806 Description: mds_lov_synchronize race leads to various problems
2807 Details : simultaneous MDT->OST connections at startup can cause the
2808 sync to abort, leaving the OSC in a bad state.
2810 Severity : enhancement
2812 Description: add optional extra BUILD_VERSION info
2813 Details : add a new environment variable (namely LUSTRE_VERS) which allows
2814 to override the lustre version.
2817 Frequency : 2.6.18 servers only
2819 Description: ll_kern_mount() doesn't release the module reference
2820 Details : The ldiskfs module reference count never drops down to 0
2821 because ll_kern_mount() doesn't release the module reference.
2826 Description: server LBUG when using old ost_num_threads parameter
2827 Details : Accept the old ost_num_threads parameter but warn that it
2828 is deprecated, and fix an off-by-one error that caused an LBUG.
2833 Description: Transient SCSI error results in persistent IO issue
2834 Details : iobuf->dr_error is not reinitialized to 0 between two
2838 Frequency : sometimes when underlying device returns I/O errors
2840 Description: OSTs not going read-only during write failures
2841 Details : OSTs are not remounted read-only when the journal commit threads
2842 get I/O errors because fsfilt_ext3 calls journal_start/stop()
2843 instead of the ext3 wrappers.
2847 Description: poor connect scaling with increasing client count
2848 Details : Don't run filter_grant_sanity_check for more than 100 exports
2849 to improve scaling for large numbers of clients.
2852 Frequency : SLES10 only
2854 Description: sanity-quota.sh quotacheck failed: rc = -22
2855 Details : Quotas cannot be enabled on SLES10.
2858 Frequency : liblustre clients only
2860 Description: getdirentries does not give error when run on compute nodes
2861 Details : getdirentries does not fail when the size specified as an argument
2862 is too small to contain at least one entry
2864 Severity : enhancement
2866 Description: Add LNET router traceability for debug purposes
2867 Details : If a checksum failure occurs with a router as part of the
2868 IO path, the NID of the last router that forwarded the bulk data
2869 is printed so it can be identified.
2874 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
2875 Details : Due to a race condition, liblustre clients were occasionally
2876 evicted incorrectly.
2878 Severity : enhancement
2880 Description: lfs setstripe use optional parameters instead of postional
2883 Severity : enhancement
2885 Description: Nanosecond timestamp support for ldiskfs
2886 Details : The on-disk ldiskfs filesystem has added support for nanosecond
2887 resolution timestamps. There is not yet support for this at
2888 the Lustre filesystem level.
2891 Frequency : during server recovery
2893 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
2894 Details : request with rq_no_resend flag not awake l_wait_event if they get
2898 Frequency : nfs export on patchless client
2900 Description: connectathon hang when test nfs export over patchless client
2901 Details : Disconnected dentry cannot be found with lookup, so we do not need
2902 to unhash it or make it invalid
2905 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
2906 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
2907 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
2912 Description: open req refcounting wrong on reconnect
2913 Details : If reconnect happened between getting open reply from server and
2914 call to mdc_set_replay_data in ll_file_open, we will schedule
2915 replay for unreferenced request that we are about to free.
2916 Subsequent close will crash in variety of ways.
2917 Check that request is still eligible for replay in
2918 mdc_set_replay_data().
2923 Description: disable writes to filesystem when reading health_check file
2924 Details : the default for reading the health_check proc file has changed
2925 to NOT do a journal transaction and write to disk, because this
2926 can cause reads of the /proc file to hang and block HA state
2927 checking on a healthy but otherwise heavily loaded system. It
2928 is possible to return to the previous behaviour during configure
2929 with --enable-health-write.
2931 Severity : enhancement
2933 Description: 64-bit inode version
2934 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
2935 to the inode. This will be required for version-based recovery.
2940 Description: MDS fails to start if a duplicate client export is detected
2941 Details : in some rare cases it was possible for a client to connect to
2942 an MDS multiple times. Upon recovery the MDS would detect this
2943 and fail during startup. Handle this more gracefully.
2945 Severity : enhancement
2947 Description: Add -o localflock option to simulate old noflock
2949 Details : This will achieve local-only flock/fcntl locks
2955 Description: log_commit_thread vs filter_destroy race leads to crash
2956 Details : Take import reference before releasing llog record semaphore
2961 Description: Wrong request locking in request set processing
2962 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
2963 imp_delayed_list, in this place should be used imp_lock.
2966 Frequency : when reconnection
2968 Description: Grant Leak when osc reconnect to OST
2969 Details : When osc reconnect ost, OST(filter) should check whether it
2970 should grant more space to client by comparing fed_grant and
2971 cl_avail_grant, and return the granted space to client instead
2972 of "new granted" space, because client will call osc_init_grant
2973 to update the client grant space info.
2976 Frequency : when client reconnect to OST
2978 Description: Grant Leak when osc do resend and replay bulk write
2979 Details : When osc reconnect to OST, OST(filter)should clear grant info of
2980 bulk write request, because the grant info while be sync between
2981 OSC and OST when reconnect, and we should ignore the grant info
2982 these of resend/replay write req.
2987 Description: Grant space more than available left space sometimes.
2988 Details : When then OST is about to be full, if two bulk writing from
2989 different clients came to OST. Accord the avaliable space of the
2990 OST, the first req should be permitted, and the second one
2991 should be denied by ENOSPC. But if the seconde arrived before
2992 the first one is committed. The OST might wrongly permit second
2993 writing, which will cause grant space > available space.
2996 Frequency : when client is evicted
2998 Description: Grant might be wrongly erased when osc is evicted by OST
2999 Details : when the import is evicted by server, it will fork another
3000 thread ptlrpc_invalidate_import_thread to invalidate the
3001 import, where the grant will be set to 0. While the original
3002 thread will update the grant it got when connecting. So if
3003 the former happened latter, the grant will be wrongly errased
3004 because of this race.
3009 Description: Checking Stale with correct fid
3010 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
3011 to check whether it is stale, because sometimes, we want the
3012 enqueue happened anyway, and op_data.fid2 will not be initialized.
3014 Severity : enhancement
3016 Description: update patchless client
3017 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
3020 Frequency : only with 2.4 kernel
3022 Description: random memory corruption
3023 Details : size of struct ll_inode_info is to big for union inode.u and this
3024 can be cause of random memory corruption.
3029 Description: Memory leak in recovery
3030 Details : Lov_mds_md was not free in an error handler in mds_create_object.
3031 It should also check obd_fail before fsfilt_start, otherwise if
3032 fsfilt_start return -EROFS,(failover mds during mds recovery).
3033 then the req will return with repmsg->transno = 0 and rc = EROFS.
3034 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
3035 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
3036 be freed no matter whether fsfilt_commit success or not.
3039 Frequency : only with huge count clients
3041 Description: Prevents from taking the superblock lock in llap_from_page for
3043 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
3044 for llap_from_page call in ll_removepage prevents from taking the
3045 superblock lock for a soon died page.
3050 Description: Not check open intent error before release open handle
3051 Details : in some rare cases, the open intent error is not checked before
3052 release open handle, which may cause
3053 ASSERTION(open_req->rq_transno != 0), because it tries to release
3054 the failed open handle.
3059 Description: Set cat log bitmap only after create log success.
3060 Details : in some rare cases, the cat log bitmap is set too early. and it
3061 should be set only after create log success.
3065 Description: Accessing a block bevice can re-enable I/O when Lustre is
3066 tearing down a device.
3067 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
3072 Description: service threads may hog cpus when there are a lot of requests
3073 Details : Insert cond_resched to give other threads a chance to use some CPU
3078 Description: the cat log was not initialized in recovery
3079 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
3080 unlink log on mds will not be initialized until mds post
3081 recovery. And also in mds post recovery, the unlink log will
3082 initialization will be done asynchronausly, so there will be race
3083 between add unlink log and unlink log initialization.
3087 Description: brw_stats were being printed incorrectly
3088 Details : brw_stats were being printed as log2 but all of them were not
3089 recorded as log2. Also remove some code duplication arising from
3090 filter_tally_{read,write}.
3094 Frequency : rare, only in recovery.
3095 Description: ASSERTION(req->rq_type != LI_POISON) failed
3096 Details : imp_lock should be held while iterating over imp_sending_list for
3097 prevent destroy request after get timeout in ptlrpc_queue_wait.
3101 Description: replay-single.sh test 52 fails
3102 Details : A lock's skiplist need to be cleanup when it being unlinked
3103 from its resource list.
3107 Description: Short directio read returns full requested size rather than
3109 Details : Direct I/O operations should return actual amount of bytes
3110 transferred rather than requested size.
3112 Severity : enhancement
3114 Description: metadata RPC reduction (e.g. for rm performance)
3115 Details : decrease the amount of synchronous RPC between clients and servers
3116 by canceling conflicing lock before the operation on the client
3117 and packing thier handles into the main operation RPC to server.
3119 Severity : enhancement
3121 Description: Async OSC create to avoid the blocking unnecessarily.
3122 Details : If an OST has no remaining object, system will block on the
3123 creation when it needs to create a new object on this OST. Now,
3124 ways use pre-created objects when available, instead of blocking on
3125 an empty osc while others are not empty. If we must block, we
3126 block for the shortest possible period of time.
3130 Description: improve handling recoverable errors
3131 Details : if request processig with error which can be recoverable on server
3132 request should be resend, otherwise page released from cache and
3135 Severity : enhancement
3137 Description: refine locking for avoid write wrong info into lov_objid file
3138 Details : fix possible races with add new target and write/update data in
3141 --------------------------------------------------------------------------------
3143 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
3148 Frequency : on some architectures
3150 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
3151 Details : Uninitialized flags sometimes cause configuration commands to
3155 Frequency : patchless clients only
3157 Description: missing __iget() symbol export
3158 Details : The __iget() symbol export is missing. To avoid the need for
3159 this on patchless clients the deathrow inode reaper is turned
3160 off, and we depend on the VM to clean up old inodes. This
3161 dependency was during via the fix for bug 12181.
3163 --------------------------------------------------------------------------------
3165 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
3167 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
3168 INTEROPERATE with older versions automatically. In many cases a
3169 special upgrade step is needed. Please read the
3170 user documentation before upgrading any part of a live system.
3171 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
3172 version will not interoperate with 1.6 betas before beta5 (1.5.95).
3173 * WARNING: Lustre configuration and startup changes are required with
3174 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
3178 Severity : enhancement
3180 Description: MountConf
3181 Details : Lustre configuration is now managed via mkfs and mount
3182 commands instead of lmc and lconf. New obd types (MGS, MGC)
3183 are added for dynamic configuration management. See
3184 https://mail.clusterfs.com/wikis/lustre/MountConf for
3187 Severity : enhancement
3189 Description: dynamic OST addition
3190 Details : OSTs can now be added to a live filesystem
3192 Severity : enhancement
3194 Description: startup order invariance
3195 Details : MDTs and OSTs can be started in any order. Clients only
3196 require the MDT to complete startup.
3198 Severity : enhancement
3200 Description: parallel, asynchronous orphan cleanup
3201 Details : orphan cleanup is now performed in separate threads for each
3202 OST, allowing parallel non-blocking operation.
3204 Severity : enhancement
3206 Description: optimized stripe assignment
3207 Details : stripe assignments are now made based on ost space available,
3208 ost previous usage, and OSS previous usage, in order to try
3209 to optimize storage space and networking resources.
3211 Severity : enhancement
3213 Description: Permanently set tunables
3214 Details : All writable /proc/fs/lustre tunables can now be permanently
3215 set on a per-server basis, at mkfs time or on a live system.
3217 Severity : enhancement
3219 Description: Lustre message v2
3220 Details : Add lustre message format v2.
3222 Severity : enhancement
3224 Description: client OST exclusion list
3225 Details : Clients can be started with a list of OSTs that should be
3226 declared "inactive" for known non-responsive OSTs.
3230 Description: SPEC SFS validation failure on NFS v2 over lustre.
3231 Details : Changes the blocksize for regular files to be 2x RPC size,
3232 and not depend on stripe size.
3234 Severity : enhancement
3236 Description: Multiple MD RPCs in flight.
3237 Details : Further unserialise some read-only MDS RPCs - learn about intents.
3238 To avoid overly-overloading MDS, introduce a limit on number of
3239 MDS RPCs in flight for a single client and add /proc controls
3240 to adjust this limit.
3242 Severity : enhancement
3244 Description: client read/write statistics
3245 Details : Add client read/write call usage stats for performance
3246 analysis of user processes.
3247 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
3248 file access. extents_stats shows chunk size distribution.
3249 extents_stats_per_process show chunk size distribution per
3252 Severity : enhancement
3254 Description: mds statistics
3255 Details : Add detailed mds operations statistics in
3256 /proc/fs/lustre/mds/*/stats.
3260 Description: Failure of copying files with lustre special EAs.
3261 Details : Client side always return success for setxattr call for lustre
3262 special xattr (currently only "trusted.lov").
3267 Description: Refcount LNET uuids
3268 Details : The global LNET uuid list grew linearly with every startup;
3269 refcount repeated list entries instead of always adding to
3272 Severity : enhancement
3274 Description: Dynamic service threads
3275 Details : Within a small range, start extra service threads
3276 automatically when the request queue builds up.
3279 Frequency : mixed-endian client/server environments
3281 Description: mixed-endian crashes
3282 Details : The new msg_v2 system had some failures in mixed-endian
3285 Severity : enhancement
3287 Description: Easy OST removal
3288 Details : OSTs can be permanently deactivated with e.g. 'lctl
3289 conf_param lustre-OST0001.osc.active=0'
3291 Severity : enhancement
3293 Description: MGS proc entries
3294 Details : Added basic proc entries for the MGS showing what filesystems
3297 Severity : enhancement
3299 Description: provide MGS failover
3300 Details : Added config lock reacquisition after MGS server failover.
3302 Severity : enhancement
3304 Description: add Linux 2.4 support
3305 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
3309 Description: a large application tries to do I/O to the same resource and dies
3310 in the middle of it.
3311 Details : Check the req->rq_arrival time after the call to
3312 ost_brw_lock_get(), but before we do anything about
3313 processing it & sending the BULK transfer request. This
3314 should help move old stale pending locks off the queue as
3315 quickly as obd_timeout.
3318 Frequency : when an incorrect nid is specified during startup
3320 Description: ptlrpc connect to non-existant node causes kernel crash
3321 Details : LNET can't be re-entered from an event callback, which
3322 happened when we expire a message after the export has been
3323 cleaned up. Instead, hand the zombie cleanup off to another
3326 Severity : enhancement
3328 Description: plain/inodebits lock performance improvement
3329 Details : Grouping plain/inodebits in granted list by their request modes
3330 and bits policy, thus improving the performance of search through
3334 Frequency : only if OST filesystem is corrupted
3336 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3337 Details : for a short time RPCs with bulk IO are in the replay list,
3338 but replay of bulk IOs is unimplemented. If the OST filesystem
3339 is corrupted due to disk cache incoherency and then replay is
3340 started it is possible to trip an assertion. Avoid putting
3341 committed RPCs into the replay list at all to avoid this issue.
3344 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3346 Description: System hang on startup
3347 Details : This bug allowed the liblustre (e.g. catamount) client to
3348 return to the app before handling all startup RPCs. This
3349 could leave the node unresponsive to lustre network traffic
3350 and manifested as a server ptllnd timeout.
3352 Severity : enhancement
3354 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3355 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
3356 boolean to control whether to print peer debug info when a
3357 client's RPC times out.
3360 Frequency : only for kernels with patches from Lustre below 1.4.3
3362 Description: Remove old rdonly API
3363 Details : Remove old rdonly API which unused from at least lustre 1.4.3
3366 Frequency : only for devices with external journals
3368 Description: Set external device read-only also
3369 Details : During a commanded failover stop, we set the disk device
3370 read-only while the server shuts down. We now also set any
3371 external journal device read-only at the same time.
3374 Frequency : when upgrading from 1.4 while trying to change parameters
3376 Description: The wrong (new) MDC name was used when setting parameters for
3377 upgraded MDT's. Also allows changing of OSC (and MDC)
3378 parameters if --writeconf is specified at tunefs upgrade time.
3381 Frequency : when setting specific ost indicies
3383 Description: QOS code breaks on skipped indicies
3384 Details : Add checks for missing OST indicies in the QOS code, so OSTs
3385 created with --index need not be sequential.
3387 Severity : enhancement
3389 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3390 Details : The uninit_groups feature works in conjunction with the kernel
3391 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3392 up the pass1 processing of e2fsck. This is a read-only feature
3393 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3394 mount filesystems that have had this feature enabled.
3396 Severity : enhancement
3398 Description: Improve multi-block allocation algorithm to avoid fragmentation
3399 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3400 allocation locality and avoid filesystem fragmentation.
3402 ------------------------------------------------------------------------------
3404 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
3406 * Support for kernels:
3407 2.4.21-47.0.1.EL (RHEL 3)
3408 2.6.5-7.283 (SLES 9)
3409 2.6.9-42.0.10.EL (RHEL 4)
3410 2.6.12.6 vanilla (kernel.org)
3411 2.6.16.27-0.9 (SLES 10)
3412 * Recommended e2fsprogs version: 1.39.cfs5
3414 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3418 Frequency : occasional, depends on client load and configuration
3419 Bugzilla : 12181, 12203
3420 Description: data loss for recently-modified files
3422 Details : In some cases it is possible that recently written or created
3423 files may not be written to disk in a timely manner (this should
3424 normally be within 30s unless client IO load is very high).
3425 The problem appears as zero-length files or files that are a
3426 multiple of 1MB in size after a client crash or client eviction
3427 that are missing data at the end of the file.
3429 This problem is more likely to be hit on clients where files are
3430 repeatedly created and unlinked in the same directory, clients
3431 have a large amount of RAM, have many CPUs, the filesystem has
3432 many OSTs, the clients are rebooted frequently, and/or the files
3433 are not accessed by other nodes after being written.
3435 The presence of the problem can be detected by looking at
3436 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
3437 smaller than the second (nr_unused) then dirty files will not
3438 be flushed automatically to disk. "sync; sleep 10" should be
3439 run several times on the node before unmounting it to update
3440 Lustre (this is also safe to run on nodes without this problem).
3442 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
3443 that can cause this same problem, so customers using that kernel
3444 also need to update the kernel in addition to Lustre. In order
3445 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
3447 It is normal that files written just before a client crash (less
3448 than 30s) may not yet have been flushed to disk, even for local
3452 Frequency : frequent on thin XT3 nodes
3454 Description: UUID collision on thin XT3 Linux nodes
3455 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
3456 randomly, since we relied on an insufficiently-seeded PRNG.
3461 Description: OSS hangs after "All ost request buffers busy"
3462 Details : A deadlock between quota and journal operations caused OSS
3463 hangs after printing "All ost request buffers busy."
3466 Frequency : always on liblustre builds
3468 Description: Cleanup compiler warnings on liblustre
3471 Frequency : always on liblustre builds on XT3
3473 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
3477 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
3478 > 32000 subdirectories
3479 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
3480 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
3481 subdirectory count crosses 32000. This will aid e2fsck to
3482 correctly handle more than 32000 subdirectories.
3485 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3487 Description: System hang on startup
3488 Details : This bug allowed the liblustre (e.g. catamount) client to
3489 return to the app before handling all startup RPCs. This
3490 could leave the node unresponsive to lustre network traffic
3491 and manifested as a server ptllnd timeout.
3493 Severity : enhancement
3495 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3496 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
3497 boolean to control whether to print peer debug info when a
3498 client's RPC times out.
3503 Description: make O_SYNC working on 2.6 kernels
3504 Details : 2.6 kernels use different method for mark pages for write,
3505 so need add a code to lustre for O_SYNC work.
3510 Description: Failure to close file and release space on NFS
3511 Details : Put inode details into lock acquired in ll_intent_file_open.
3512 Use mdc_intent_lock in ll_intent_open to properly
3513 detect all kind of errors unhandled by mdc_enqueue.
3518 Description: proc file read during shutdown sometimes raced obd removal,
3520 Details : Add lock to prevent obd access after proc file removal.
3523 Frequency : Only for files larger than 4GB on 32-bit clients.
3525 Description: improperly doing page alignment of locks
3526 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
3527 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3532 Description: RPCs being resent when they shouldn't be
3533 Details : Some RPCs that should not be resent are being resent. This
3534 can cause inconsistencies in the RPC state machine. Do not
3535 resend such requests.
3538 Frequency : rare, only with NFS export
3540 Description: Crash on NFS re-export node
3541 Details : under very unusual load conditions an assertion is hit in
3542 ll_intent_file_open()
3545 Frequency : only if OST filesystem is corrupted
3547 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3548 Details : for a short time RPCs with bulk IO are in the replay list,
3549 but replay of bulk IOs is unimplemented. If the OST filesystem
3550 is corrupted due to disk cache incoherency and then replay is
3551 started it is possible to trip an assertion. Avoid putting
3552 committed RPCs into the replay list at all to avoid this issue.
3557 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3558 Details : Large single O_DIRECT read and write calls can fail to allocate
3559 a sufficiently large buffer to process the request. In case of
3560 allocation failure the allocation is retried with a smaller
3561 buffer and broken into smaller requests.
3563 Severity : enhancement
3565 Description: Add -o localflock option to simulate old noflock behaviour.
3566 Details : This will achieve local-only flock/fcntl locks coherentness.
3571 Description: versioning check is incomplete
3572 Details : Checking the version difference of client vs. server, report
3573 error if the gap is too big.
3578 Description: add support PG_writeback bit
3579 Details : add support for PG_writeback bit for Lustre, for more carefull
3580 work with page cache in 2.6 kernel. This also fix some deadlocks
3581 and remove hack for work O_SYNC with 2.6 kernel.
3583 Severity : enhancement
3585 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3586 Details : The uninit_groups feature works in conjunction with the kernel
3587 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3588 up the pass1 processing of e2fsck. This is a read-only feature
3589 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3590 mount filesystems that have had this feature enabled.
3592 Severity : enhancement
3594 Description: Improve multi-block allocation algorithm to avoid fragmentation
3595 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3596 allocation locality and avoid filesystem fragmentation.
3598 ------------------------------------------------------------------------------
3600 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3602 * Support for kernels:
3603 2.6.9-42.0.3.EL (RHEL 4)
3604 2.6.5-7.276 (SLES 9)
3605 2.4.21-47.0.1.EL (RHEL 3)
3606 2.6.12.6 vanilla (kernel.org)
3607 2.6.16.21-0.8 (SLES10)
3608 * Recommended e2fsprogs version: 1.39.cfs2-0
3610 * The backwards-compatible /proc/sys/portals symlink has been removed
3611 in this release. Before upgrading, please ensure that you change
3612 any configuration scripts or /etc/sysctl.conf files that access
3613 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3614 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3615 in advance of the upgrade on any system running Lustre 1.4.6 or
3616 newer, since /proc/sys/lnet was added in that version.
3617 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3621 Frequency : only when quota is used
3623 Description: avoid scanning export list for quota master
3624 Details : Change the algorithms to avoid scanning export list in order
3625 to improve the efficiency.
3628 Frequency : MDS failover only, very rarely
3630 Description: "went back in time" messages on mds failover
3631 Details : The greatest transno may be lost when the current operation
3632 finishes with an error (transno==0) and the client's last_rcvd
3633 record is over-written. Save the greatest transno in the
3634 mds_last_transno for this case.
3637 Frequency : always for specific kernels and striping counts
3639 Description: client may get "Matching packet too big" without ACL support
3640 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
3641 when trying to access files in certain configurations. The
3642 clients should in fact be denied when mounting because they do
3643 not understand ACLs.
3646 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
3648 Description: many clients connecting with IO in progress causes connect timeouts
3649 Details : Avoid synchronous journal commits to avoid delays caused by many
3650 clients connecting/disconnecting when bulk IO is in progress.
3651 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
3652 OST_IO_PORTAL to avoid delays behind potentially many pending
3656 Frequency : occasionally with multiple writers to a single file
3658 Description: shared writes to file may result in wrong size reported by stat()
3659 Details : Allow growing of kms when extent lock is cancelled
3662 Frequency : always with random mmap IO to multi-striped file
3664 Description: mmap write might be lost if we are writing to a 'hole' in stripe
3665 Details : Only if the hole is at the end of OST object so that kms is too
3666 small. Fix is to increase kms accordingly in ll_nopage.
3669 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
3671 Description: writes to a missing object would leak memory on the OST
3672 Details : If there is an inconsistency between the MDS and OST filesystems,
3673 such that the MDS references an object that doesn't exist, writes
3674 to that object will leak memory due to incorrect cleanup in the
3675 error handling path, eventually running out of memory on the OST.
3680 Description: Creating too long symlink causes lustre errors
3681 Details : Check symlink and name lengths before sending requests to MDS.
3684 Frequency : only if flock is enabled (not on by default)
3686 Description: posix locks not released on fd closure on 2.6.9+
3687 Details : We failed to add posix locks to list of inode locks on 2.6.9+
3688 kernels, this caused such locks not to be released on fd close and
3689 then assertions on fs unmount about still used locks.
3692 Frequency : MDS failover only, very rarely
3694 Description: clients may get ASSERTION(granted_lock != NULL)
3695 Details : When request was taking a long time, and a client was resending
3696 a getattr by name lock request. The were multiple lock requests
3697 with the same client lock handle and
3698 mds_getattr_name->fixup_handle_for_resent_request found one of the
3699 lock handles but later failed with ASSERTION(granted_lock != NULL).
3704 Description: handle->h_buffer_credits > 0, assertion failure
3705 Details : h_buffer_credits is zero after truncate, causing assertion
3706 failure. This patch extends the transaction or creates a new
3710 Frequency : NFS re-export or patchless client
3711 Bugzilla : 11179, 10796
3712 Description: Crash on NFS re-export node (__d_move)
3713 Details : We do not want to hash the dentry if we don't have a lock.
3714 But if this dentry is later used in d_move, we'd hit uninitialised
3715 list head d_hash, so we just do this to init d_hash field but
3716 leave dentry unhashed.
3719 Frequency : NFS re-export or patchless client
3721 Description: NFS exports has problem with symbolic link
3722 Details : lustre client didn't properly install dentry when re-exported
3723 to NFS or running patchless client.
3726 Frequency : NFS re-export or patchless client
3728 Description: Various nfs/patchless fixes.
3729 Details : fixes reuse disconected alias for lookup process - this fixes
3730 warning "find_exported_dentry: npd != pd",
3731 fix permission error with open files at nfs.
3732 fix apply umask when do revalidate.
3735 Frequency : occasional
3737 Description: Crash on NFS re-export node
3738 Details : calling clear_page() on the wrong pointer triggered oops in
3739 generic_mapping_read().
3742 Frequency : rarely, using O_DIRECT IO
3744 Description: unaligned directio crashes client with LASSERT
3745 Details : check for unaligned buffers before trying any requests.
3748 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
3750 Description: stale data returned from RAID cache
3751 Details : If only a small amount of IO is done to the RAID device before
3752 reading it again it is possible to get stale data from the RAID
3753 cache instead of reading it from disk.
3756 Frequency : always for sles10 kernel
3758 Description: sles10 support
3759 Details : ll_follow_link: compile fixes and using of nd_set_link
3760 under newer kernels.
3763 Frequency : depends on arch, kernel and compiler version, always on sles10
3766 Description: recursive or deep enough symlinks cause stack overflow
3767 Details : getting rid of large stack-allocated variable in
3771 Frequency : depends on hardware
3773 Description: lustre write performance loss in the SLES10 kernel
3774 Details : the performance loss is caused by using of write barriers in the
3775 ext3 code. The SLES10 kernel turns barrier support on by
3776 default. The fix is to undo that change for ldiskfs.
3778 ------------------------------------------------------------------------------
3780 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
3782 * Support for kernels:
3783 2.6.9-42.0.3EL (RHEL 4)
3784 2.6.5-7.276 (SLES 9)
3785 2.4.21-47.0.1.EL (RHEL 3)
3786 2.6.12.6 vanilla (kernel.org)
3790 Frequency : quota enabled and large files being deleted
3792 Description: releasing more than 4GB of quota at once hangs OST
3793 Details : If a user deletes more than 4GB of files on a single OST it
3794 will cause the OST to spin in an infinite loop. Release
3795 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3800 Description: statfs data retrieved from /proc may be stale or zero
3801 Details : When reading per-device statfs data from /proc, in the
3802 {kbytes,files}_{total,free,avail} files, it may appear
3803 as zero or be out of date.
3806 Frequency : systems with MD RAID1 external journal devices
3808 Description: lconf's call to blkid is confused by RAID1 journal devices
3809 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3810 of returning all block devices that match the journal UUID.
3813 Frequency : always, for aggregate stripe size over 4GB
3815 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
3816 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
3817 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
3820 Frequency : always if "lfs find" used on a local file/directory
3822 Description: "lfs find" segfaults if used on a local file/directory
3823 Details : The case where a directory component was not specified wasn't
3824 handled correctly. Handle this properly.
3827 Frequency : always on ppc64
3829 Description: the write to an ext3 filesystem mounted with mballoc got stuck
3830 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3831 perform endianness conversion.
3834 Frequency : rarely (truncate to non-zero file size after write under load)
3835 Bugzilla : 10730, 10687
3836 Description: Files padded with zeros to next 4K multiple
3837 Details : With filesystems mounted using the "extents" option (2.6 kernels)
3838 it is possible that files that are truncated to a non-zero size
3839 immediately after being written are filled with zero bytes beyond
3840 the truncated size. No file data is lost.
3842 Severity : enhancement
3844 Description: Allow recovery/failover for liblustre clients.
3845 Details : liblustre clients were unaware of failover configurations until
3848 Severity : enhancement
3850 Description: user file locks should fail when not mounting with flock option
3851 Details : Set up an error-returning stub in ll_file_operations.lock field
3852 to prevent incorrect behaviour when client is mounted without
3853 flock option. Also, set up properly f_op->flock field for
3857 Frequency : always on ia64
3859 Description: "lfs df" loops on printing out MDS statfs information
3860 Details : The obd_ioctl_data was not initialized and in some systems
3861 this caused a failure during the ioctl that did not return
3862 an error. Initialize the struct and return an error on failure.
3865 Frequency : SLES 9 only
3867 Description: Error of copying files with lustre special EAs as root
3868 Details : Client side always return success for setxattr call for lustre
3869 special xattr (currently only "trusted.lov").
3872 Frequency : rarely on clusters with both ia64+i386 clients
3874 Description: ia64+i686 clients doing shared IO on the same file may LBUG
3875 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
3876 clients are doing concurrent writes to the same file it is
3877 possible that the ia64 clients may LASSERT because the OST
3878 extent locks are not PAGE_SIZE aligned. Ensure that grown
3879 locks are always aligned on the request boundary.
3882 Frequency : specific use, occasional
3884 Description: Overwriting in use executable truncates on-disk binary image
3885 Details : If one node attempts to overwrite an executable in use by
3886 another node, we now correctly return ETXTBSY instead of
3887 truncating the file.
3889 Severity : enhancement
3891 Description: Async OSC create to avoid the blocking unnecessarily.
3892 Details : If an OST has no remaining object, system will block on the creating
3893 when need to create a new object on this OST. Now, ways use
3894 pre-created objects when available, instead of blocking on an
3895 empty osc while others are not empty. If we must block, we block
3896 for the shortest possible period of time.
3901 Description: chmod on Lustre root is propagated to other clients
3902 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
3903 invalid by the follow_mount time.
3906 Frequency : liblustre clients only
3908 Description: Race in 'instant cancel' lock handling could lead to such locks
3909 never to be granted in case of SMP MDS
3910 Details : Do not destroy not yet granted but cbpending locks in
3914 Frequency : replay/resend of open
3916 Description: non null lock assetion failure in mds_intent_policy
3917 Details : Trying to replay/resend lockless open requests resulted in
3918 mds_open() returning 0 with no lock. Now it sets a flag if
3919 a lock is going to be returned.
3921 Severity : enhancement
3923 Description: Checksum enhancements
3924 Details : New checksum enhancements allow for resending RPCs that failed
3927 Severity : enhancement
3929 Description: Tunables on number of dirty pages in cacche
3930 Details : Allow to set limit on number of dirty pages cached.
3935 Description: client crash on unmount - lock still has references
3936 Details : In some error handling cases it was possible to leak a lock
3937 reference on a client while accessing a file. This was not
3938 harmful to the client during operation, but would cause the
3939 client to crash when the filesystem is unmounted.
3942 Frequency : specific case, rare
3944 Description: ETXTBSY on mds though file not in use by client
3945 Details : ETXTBSY is no longer incorrectly returned when attempting to
3946 chmod or chown a directory that the user previously tried to
3947 execute or a currently-executing binary.
3950 Frequency : extremely rare except on liblustre-based clients
3952 Description: Lustre space not freed when files are deleted
3953 Details : Clean up open-unlinked files after client eviction. Previously
3954 the unlink was skipped and the files remained as orphans.
3959 Description: OST failure "would be an LBUG" in waiting_locks_callback()
3960 Details : In some cases it was possible to send a blocking callback to
3961 a client doing a glimpse, even though that client didn't get
3962 a lock granted. When the glimpse lock is cancelled on the OST
3963 the freed lock is left on the waiting list and corrupted the list.
3966 Frequency : all core dumps
3968 Description: Broke core dumps to lustre
3969 Details : Negative dentry may be unhashed if parent does not have UPDATE
3970 lock, but some callers, e.g. do_coredump, expect dentry to be
3971 hashed after successful create, hash it in ll_create_it.
3973 ------------------------------------------------------------------------------
3975 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
3977 * Support for kernels:
3978 2.6.9-42.0.2.EL (RHEL 4)
3979 2.6.5-7.276 (SLES 9)
3980 2.4.21-40.EL (RHEL 3)
3981 2.6.12.6 vanilla (kernel.org)
3985 Frequency : always on RHEL 3
3987 Description: Number of open files grows over time
3988 Details : The number of open files grows over time, whether or not
3989 Lustre is started. This was due to a filp leak introduced
3990 by one of our kernel patches.
3992 ------------------------------------------------------------------------------
3994 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
3996 * Support for kernels:
3997 2.6.9-42.EL (RHEL 4)
3998 2.6.5-7.276 (SLES 9)
3999 2.4.21-40.EL (RHEL 3)
4000 2.6.12.6 vanilla (kernel.org)
4005 Bugzilla : 5719, 9635, 9792, 9684
4006 Description: OST (or MDS) trips assertions in (re)connection under heavy load
4007 Details : If a server is under heavy load and cannot reply to new
4008 connection requests before the client resends the (re)connect,
4009 the connection handling code can behave badly if two service
4010 threads are concurrently handing separate (re)connections from
4011 the same client. Add better locking to the connection handling
4012 code, and ensure that only a single connection will be processed
4013 for a given client UUID, even if the lock is dropped.
4015 Severity : enhancement
4017 Description: add TCP zero-copy support to kernel
4018 Details : Add support to the kernel TCP stack to allow zero-copy bulk
4019 sends if the hardware supports scatter-gather and checksumming.
4020 This allows socklnd to do client-write and server-read more
4021 efficiently and reduce CPU utilization from skbuf copying.
4024 Frequency : only if NFS exporting from client
4026 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
4027 Details : If setting attributes on a file created under NFS that had
4028 never been opened it would be possible to oops the client
4029 if the file had no objects.
4032 Frequency : always for liblustre
4034 Description: liblustre client does MDS+OSTs setattr RPC for each write
4035 Details : When doing a write from a liblustre client, the client
4036 incorrectly issued an RPC to the MDS and each OST the file was
4037 striped over in order to update the timestamps. When writing
4038 with small chunks and many clients this could overwhelm the MDS
4039 with RPCs. In all cases it would slow down the write because
4040 these RPCs are unnecessary.
4042 Severity : enhancement
4044 Description: allow number of MDS service threads to be changed at module load
4045 Details : It is now possible to change the number of MDS service threads
4046 running. Adding "options mds mds_num_threads={N}" to the MDS's
4047 /etc/modprobe.conf will set the number of threads for the next
4048 time Lustre is restarted (assuming the "mds" module is also
4049 reloaded at that time). The default number of threads will
4050 stay the same, 32 for most systems.
4055 Description: OST crash if filesystem is unformatted or corrupt
4056 Details : If an OST is started on a device that has never been formatted
4057 or if the filesystem is corrupt and cannot even mount then the
4058 error handling cleanup routines would dereference a NULL pointer.
4063 Description: NULL pointer deref in llap_from_page.
4064 Details : get_cache_page_nowait can return a page with NULL (or otherwise
4065 incorrect) mapping if the page was truncated/reclaimed while it was
4066 searched for. Check for this condition and skip such pages when
4067 doing readahead. Introduce extra check to llap_from_page() to
4068 verify page->mapping->host is non-NULL (so page is not anonymous).
4071 Frequency : Sometimes when using sys_sendfile
4073 Description: "page not covered by a lock" warnings from ll_readpage
4074 Details : sendfile called ll_readpage without right page locks present.
4075 Now we introduced ll_file_sendfile that does necessary locking
4076 around call to generic_file_sendfile() much like we do in
4080 Frequency : with certain MDS communication failures at client mount time
4082 Description: NULL pointer deref after failed client mount
4083 Details : a client connection request may delayed by the network layer
4084 and not be sent until after the PTLRPC layer has timed out the
4085 request. If the client fails the mount immediately it will try
4086 to clean up before the network times out the request. Add a
4087 reference from the request import to the obd device and delay
4088 the cleanup until the network drops the request.
4091 Frequency : occasionally during client (re)connect
4093 Description: assertion failure during client (re)connect
4094 Details : processing a client connection request may be delayed by the
4095 client or server longer than the client connect timeout. This
4096 causes the client to resend the connection request. If the
4097 original connection request is replied in this interval, the
4098 client may trip an assertion failure in ptlrpc_connect_interpret()
4099 which thought it would be the only running connect process.
4102 Frequency : only with obd_echo servers and clients that are rebooted
4104 Description: kernel BUG accessing uninitialized data structure
4105 Details : When running an obd_echo server it did not start the ping_evictor
4106 thread, and when a client was evicted an uninitialized data
4107 structure was accessed. Start the ping_evictor in the RPC
4108 service startup instead of the OBD startup.
4110 Severity : enhancement
4111 Bugzilla : 10193 (patchless)
4112 Description: Remove dependency on various unexported kernel interfaces.
4113 Details : No longer need reparent_to_init, exit_mm, exit_files,
4114 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
4117 Frequency : rare (only users of deprecated and unsupported LDAP config)
4119 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
4120 Details : LDAP apparently contains 'lustreName' attributes instead of
4121 'name'. A simple remapping of the name is sufficient.
4124 Frequency : rare (only with non-default dump_on_timeout debug enabled)
4126 Description: waiting_locks_callback trips kernel BUG if client is evicted
4127 Details : Running with the dump_on_timeout debug flag turned on makes
4128 it possible that the waiting_locks_callback() can try to dump
4129 the Lustre kernel debug logs from an interrupt handler. Defer
4130 this log dumping to the expired_lock_main() thread.
4132 Severity : enhancement
4134 Description: Support NFS exporting on 2.6 kernels.
4135 Details : Implement non-rawops metadata methods for NFS server to use without
4136 changing NFS server code.
4139 Frequency : very rare (synthetic metadata workload only)
4141 Description: two racing renames might cause an MDS thread to deadlock
4142 Details : Running the "racer" program may cause one MDS thread to rename
4143 a file from being the source of a rename to being the target of
4144 a rename at exactly the same time that another thread is doing
4145 so, and the second thread has already enqueued these locks after
4146 doing a lookup of the target and is trying to relock them in
4147 order. Ensure that we don't try to re-lock the same resource.
4150 Frequency : only very large systems with liblustre clients
4152 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
4153 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
4154 all OSTs in parallel. This allows the eviction of stale liblustre
4155 clients to proceed much faster than if they were done in series,
4156 and also offers similar improvements for other set_info RPCs.
4161 Description: excessive CPU usage during initial read phase on client
4162 Details : During the initial read phase on a client, it would agressively
4163 retry readahead on the file, consuming too much CPU and impacting
4164 performance (since 1.4.5.8). Improve the readahead algorithm
4165 to avoid this, and also improve some other common cases (read
4166 of small files in particular, where "small" is files smaller than
4167 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
4172 Description: MDS crash when receiving packet with unknown intent.
4173 Details : Do not LBUG in unknown intent case, just return -EFAULT
4175 Severity : enhancement
4176 Bugzilla : 9293, 9385
4177 Description: MDS RPCs are serialised on client. This is unnecessary for some.
4178 Details : Do not serialize getattr (non-intent version) and statfs.
4181 Frequency : occasional, when OST network is overloaded/intermittent
4183 Description: client evicted by OST after bulk IO timeout
4184 Details : If a client sends a bulk IO request (read or write) the OST
4185 may evict the client if it is unresposive to its data GET/PUT
4186 request. This is incorrect if the network is overloaded (takes
4187 too long to transfer the RPC data) or dropped the OST GET/PUT
4188 request. There is no need to evict the client at all, since
4189 the pinger and/or lock callbacks will handle this, and the
4190 client can restart the bulk request.
4193 Frequency : Always when mmapping file with no objects
4195 Description: client crashes when mmapping file with no objects
4196 Details : Check that we actually have objects in a file before doing any
4197 operations on objects in ll_vm_open, ll_vm_close and
4203 Description: Request leak when working with deleted CWD
4204 Details : Introduce advanced request refcount tracking for requests
4205 referenced from lustre intent.
4207 Severity : Enhancement
4209 Description: Cache open file handles on client.
4210 Details : MDS now will return special lock along with openhandle, if
4211 requested and client is allowed to hold openhandle, even if unused,
4212 until such a lock is revoked. Helps NFS a lot, since NFS is opening
4213 closing files for every read/write openration.
4215 Severity : Enhancement
4217 Description: Cache open negative dentries on client when possible.
4218 Details : Guard negative dentries with UPDATE lock on parent dir, drop
4219 negative dentries on lock revocation.
4224 Description: Remounting a client read-only wasn't possible with a zconf mount
4225 Details : It wasn't possible to remount a client read-only with llmount.
4227 Severity : enhancement
4228 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
4229 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
4230 MPICH is included to add Lustre-specific ADIO interfaces.
4231 This is based closely on the UFS ADIO layer and only differs
4232 in file creation, in order to allow the OST striping to be set.
4233 This is user-contributed code and not supported by CFS.
4238 Description: extended inode attributes (immutable, append-only) work improperly
4239 when 2.4 and 2.6 kernels are used on client/server or vice versa
4240 Details : Introduce kernel-independent values for these flags.
4242 Severity : enhancement
4245 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
4246 Details : Many of the /proc/ tunables can only be tuned at a megabyte
4247 granularity. Now, Fractional MB granularity is be supported,
4248 this is very useful for low memory system.
4250 Severity : enhancement
4252 Description: Getattr by fid
4253 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
4254 locks, avoids extra getattr rpc requests to MDS, allows '/' to
4255 have locks and avoids getattr rpc requests for it on every stat.
4258 Frequency : Always, for filesystems larger than 2TB
4260 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4261 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
4262 filesystems larger than 2TB to be mounted with mballoc enabled.
4265 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
4267 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4268 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
4269 filesystems larger than 2TB is mounted then the kernel will
4270 silently corrupt the start of the filesystem. CONFIG_LBD is
4271 enabled for all CFS-supported kernels, but the possibility of
4272 this happening with a modified kernel config exists.
4274 Severity : enhancement
4276 Description: add client O_DIRECT support for 2.6 kernels
4277 Details : It is now possible to do O_DIRECT reads and writes to files
4278 in the Lustre client mountpoint on 2.6 kernel clients.
4280 Severity : enhancement
4282 Description: parallel glimpse, setattr, statfs, punch, destroy requests
4283 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
4284 parallel, not waiting for response from every OST before sending
4285 a rpc to the next OST.
4290 Description: setattr vs write race when updating file timestamps
4291 Details : Client processes that update a file timestamp into the past
4292 right after writing to the file (e.g. tar) it is possible that
4293 the updated file modification time can be reset to the current
4294 time due to a race between processing the setattr and write RPC.
4296 Severity : enhancement
4298 Description: Bring 'lfs find' closer in line with regular Linux find.
4299 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
4300 -print0 options and obtains all the needed info through the lustre
4303 Severity : enhancement
4305 Description: support up to 1024 configured devices on one node
4306 Details : change obd_dev array from statically allocated to dynamically
4307 allocated structs as they are first used to reduce memory usage
4312 Description: Flush dirty partially truncated pages during truncate
4313 Details : Immediatelly flush partially truncated pages in filter_setattr,
4314 this way we completely avoid having any pages in page cache on OST
4315 and can retire ugly workarounds during writes to flush such pages.
4320 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
4321 Details : protect inode from truncation within vfs_unlink() context
4322 just take a reference before calling vfs_unlink() and release it
4323 when parent's i_sem is free.
4328 Description: last_id value checked outside lock on OST caused LASSERT failure
4329 Details : If there were multiple MDS->OST object precreate requests in
4330 flight, it was possible that the OST's last object id was checked
4331 outside a lock and incorrectly tripped an assertion. Move checks
4332 inside locks, and discard old precreate requests.
4335 Frequency : always, if extents are used on OSTs
4337 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
4338 Details : OSTs using the extents format would not zero the high 16 bits of
4339 the index physical block number. This is not a problem for any
4340 OST filesystems smaller than 16TB, and no kernels support ext3
4341 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
4342 new/modified files) and can be fixed for existing filesystems
4343 with e2fsprogs-1.39-cfs1.
4348 Description: import connection selection may be incorrect if timer wraps
4349 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
4350 import connections to be ignored if the 32-bit jiffies counter
4351 wraps. Use a 64-bit jiffies counter.
4354 Frequency : very large clusters immediately after boot
4356 Description: LNET request buffers exhausted under heavy short-term load
4357 Details : If a large number of client requests are generated on a service
4358 that has previously never seen so many requests it is possible
4359 that the request buffer growth cannot keep up with the spike in
4360 demand. Instead of dropping incoming requests, they are held in
4361 the LND until the RPC service can accept more requests.
4364 Frequency : Sometimes during replay
4366 Description: Assertion failure in ll_local_open after replay.
4367 Details : If replay happened on an open request reply before we were able
4368 to set replay handler, reply will become not swabbed tripping the
4369 assertion in ll_local_open. Now we set the handler right after
4370 recognising of open request
4373 Frequency : very rare
4375 Description: kernel reports "badness in vsnprintf"
4376 Details : Reading from the "recovery_status" /proc file in small chunks
4377 may cause a negative length in lprocfs_obd_rd_recovery_status()
4378 call to vsnprintf() (which is otherwise harmless). Exit early
4379 if there is no more space in the output buffer.
4381 Severity : enhancement
4383 Description: clear OBD RPC statistics by writing to them
4384 Details : It is now possible to clear the OBD RPC statistics by writing
4385 to the "stats" file.
4390 Description: Client mtime is not the same on different clients after utimes
4391 Details : In some cases, the client was using the utimes() syscall on
4392 a file cached on another node. The clients now validate the
4393 ctime from the MDS + OSTs to determine which one is right.
4398 Description: Inability to activate failout mode
4399 Details : lconf script incorrectly assumed that in pythong string's numeric
4400 value is used in comparisons.
4403 Frequency : always with multiple stripes per file
4405 Description: Inefficient object allocation for mutli-stripe files
4406 Details : When selecting which OSTs to stripe files over, for files with
4407 a stripe count that divides evenly into the number of OSTs,
4408 the MDS is always picking the same starting OST for each file.
4409 Return the OST selection heuristic to the original design.
4414 Description: mount failures may take full timeout to return an error
4415 Details : Under some heavy load conditions it is possible that a
4416 failed mount can wait for the full obd_timeout interval,
4417 possibly several minutes, before reporting an error.
4418 Instead return an error as soon as the status is known.
4420 Frequency : quota enabled and large files being deleted
4422 Description: releasing more than 4GB of quota at once hangs OST
4423 Details : If a user deletes more than 4GB of files on a single OST it
4424 will cause the OST to spin in an infinite loop. Release
4425 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4430 Description: statfs data retrieved from /proc may be stale or zero
4431 Details : When reading per-device statfs data from /proc, in the
4432 {kbytes,files}_{total,free,avail} files, it may appear
4433 as zero or be out of date.
4436 Frequency : systems with MD RAID1 external journal devices
4438 Description: lconf's call to blkid is confused by RAID1 journal devices
4439 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4440 of returning all block devices that match the journal UUID.
4443 Frequency : always, for aggregate stripe size over 4GB
4445 Description: assertion fails when trying to use 4GB stripe size
4446 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
4447 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
4450 Frequency : always on ppc64
4452 Description: the first write on an ext3 filesystem with mballoc got stuck
4453 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4454 perform endianness conversion.
4456 ------------------------------------------------------------------------------
4458 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4460 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
4461 INTEROPERATE with older versions automatically. Please read the
4462 user documentation before upgrading any part of a live system.
4463 * WARNING: Lustre networking configuration changes are required with
4464 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
4467 * Support for kernels:
4468 2.6.9-22.0.2.EL (RHEL 4)
4469 2.6.5-7.244 (SLES 9)
4470 2.6.12.6 vanilla (kernel.org)
4473 Severity : enhancement
4474 Bugzilla : 7981/8208
4475 Description: Introduced Lustre Networking (LNET)
4476 Details : LNET is new networking infrastructure for Lustre, it includes
4477 a reorganized network configuration mode (see the user
4478 documentation for full details) as well as support for routing
4479 between different network fabrics. Lustre Networking Devices
4480 (LNDs) for the supported network fabrics have also been
4481 created for this new infrastructure.
4483 Severity : enhancement
4484 Description: Introduced Access control lists
4485 Details : clients can set ACLs on files and directories in order to have
4486 more fine-grained permissions than the standard Unix UGO+RWX.
4487 The MDS must be started with the "-o acl" mount option.
4489 Severity : enhancement
4490 Description: Introduced filesystem quotas
4491 Details : Administrators may now establish per-user quotas on the
4494 Severity : enhancement
4496 Description: Configuration change for the XT3
4497 The PTLLND is now used to run Lustre over Portals on the XT3
4498 The configure option(s) --with-cray-portals are no longer used.
4499 Rather --with-portals=<path-to-portals-includes> is used to
4500 enable building on the XT3. In addition to enable XT3 specific
4501 features the option --enable-cray-xt3 must be used.
4506 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
4507 Details : A race between allocating a new llcd and re-getting the llcd_lock
4508 allowed another thread to grab newly-allocated llcd.
4510 Severity : enhancement
4512 Description: 2.6 OST async journal commit and locking fix to improve performance
4513 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
4514 2.6 kernels are now async as they already were in 2.4 kernels so
4515 that they can commit concurrently with the network bulk transfer.
4516 For block-allocated files the filter allocation semaphore is held
4517 to avoid filesystem fragmentation during allocation. BKL lock
4518 removed for 2.6 xattr operations where it is no longer needed.
4523 Description: lconf incorrectly determined whether two IP networks could talk
4524 Details : In some more complicated routing and multiple-network
4525 configurations, lconf will avoid trying to make a network
4526 connection to a disjoint part of the IP space. It was doing the
4527 math incorrectly for one set of cases.
4532 Description: Fix for potential infinite loop processing records in an llog.
4533 Details : If an llog record is corrupted/zeroed, it is possible to loop
4534 forever in llog_process(). Validate the llog record length
4535 and skip the remainder of the block on error.
4538 Frequency : occasional (liblustre only)
4540 Description: liblustre could not open files whose last component is a symlink
4541 Details : sysio_path_walk() would incorrectly pass the open intent to
4542 intermediate path components.
4545 Frequency : rare (liblustre only with non-standard tuning)
4546 Bugzilla : 7201 (7350)
4547 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4548 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4549 after it has been released, only do this for referenced locks
4551 Severity : enhancement
4553 Description: specifying an (invalid) directory default stripe_size of -1
4554 would reset the directory default striping
4555 Details : stripe_size -1 was used internally to signal directory stripe
4556 removal, now use "all default" to signal dir stripe removal
4557 as a directory striping of "all default" is not useful
4560 Frequency : common for large clusters running liblustre clients
4562 Description: doing an ls when liblustre clients are running is slow
4563 Details : sending a glimpse AST to a liblustre client waits for every AST
4564 to time out, as liblustre clients will not respond. Since they
4565 cannot cache data we refresh the OST lock LVB from disk instead.
4567 Severity : enhancement
4569 Description: doing an ls at the same time as file IO can be slow
4570 Details : enqueue and other "small" requests can be blocked behind many
4571 large IO requests. Create a new OST IO portal for non-IO
4572 requests so they can be processed faster.
4575 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4577 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4578 created file with mode 000
4579 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4580 allow an ftruncate() as a non-root user to a file with mode 000.
4581 Lustre now respects this flag to disable mode checks when
4582 truncating a file owned by the user
4585 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4588 Description: Revoking locks from clients that went dead or catatonic might take
4590 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4591 cancellation of such locks instant on servers without waiting for
4592 any reply from clients. Clients drops these locks when cancel
4593 notification from server is received without replying.
4596 Frequency : liblustre-only, when liblustre client dies or becomes busy
4598 Description: Doing ls on Linux clients can take a long time with active
4600 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4601 granting such locks to it in the first place if possible. Locks
4602 are taken by proxy on the OST during the read or write and
4603 dropped immediately afterward. Add connect flags handling, do
4604 not grant locks to liblustre clients for glimpse ASTs.
4606 Severity : enhancement
4608 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4609 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4610 sequential reads and behaves badly with applications doing random
4611 reads. Improve it by reading ahead at least read region, and
4612 avoiding excessive large RPC for small reads.
4614 Severity : enhancement
4616 Description: Creating more than 1000 files for a single job may cause a load
4617 imbalance on the OSTs if there are also a large number of OSTs.
4618 Details : qos_prep_create() uses an OST index reseed value that is an
4619 even multiple of the number of available OSTs so that if the
4620 reseed happens in the middle of the object allocation it will
4621 still utilize the OSTs as uniformly as possible.
4626 Description: OST or MDS may oops in ping_evictor_main()
4627 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
4628 but doesn't restart at beginning of obd_exports_timed list
4631 Severity : enhancement
4633 Description: improve by-nid export eviction on the MDS and OST
4634 Details : allow multiple exports with the same NID to be evicted at one
4635 time without re-searching the exports list.
4638 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
4640 Description: MDS may oops in groups_free()
4641 Details : in rare race conditions a newly allocated group_info struct is
4642 freed again, and this can be NULL. The 2.4 compatibility code
4643 for groups_free() checked for a NULL pointer, but 2.6 did not.
4646 Frequency : common for liblustre clients doing little filesystem IO
4647 Bugzilla : 9352, 7313
4648 Description: server may evict liblustre clients accessing contended locks
4649 Details : if a client is granted a lock or receives a completion AST
4650 with a blocking AST already set it would not reply to the AST
4651 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
4654 Frequency : lfs setstripe, only systems with more than 160 OSTs
4656 Description: unable to set striping with a starting offset beyond OST 160
4657 Details : llapi_create_file() incorrectly limited the starting stripe
4658 index to the maximum single-file stripe count.
4661 Frequency : LDAP users only
4663 Description: lconf did not handle in-kernel recovery with LDAP properly
4664 Details : lconf/LustreDB get_refs() is searching the wrong namespace
4666 Severity : enhancement
4668 Description: bind OST threads to NUMA nodes to improve performance
4669 Details : all OST threads are uniformly bound to CPUs on a single NUMA
4670 node and do their allocations there to localize memory access
4672 Severity : enhancement
4674 Description: llmount can determine client NID directly from Myrinet (GM)
4675 Details : the client NID code from gmnalnid was moved directly into
4676 llmount, removing the need to use this or specifying the
4677 client NID explicitly when mounting GM clients with zeroconf
4680 Frequency : if client is started with down MDS
4682 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
4683 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
4684 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
4685 ptlrpc_import_delayed_req() because we didn't check if the
4686 request was interrupted, and we also didn't break out of the
4687 event loop if there was no timeout
4692 Description: data loss during non-page-aligned writes to a single file from
4693 both multiple nodes and multiple threads on one node at same time
4694 Details : updates to KMS and lsm weren't protected by common lock. Resulting
4695 inconsistency led to false short-reads, that were cached and later
4696 used by ->prepare_write() to fill in partially written page,
4697 leading to data loss.
4700 Frequency : always, if lconf --abort_recovery used
4702 Description: lconf --abort_recovery fails with 'Operation not supported'
4703 Details : lconf was attempting to abort recovery on the MDT device and not
4706 Severity : enhancement
4708 Description: remove cleanup logs
4709 Details : replace lconf-generated cleanup logs with lustre internal
4710 cleanup routines. Eliminates the need for client-cleanup and
4713 Severity : enhancement
4715 Description: add support for EAs (user and system) on lustre filesystems
4716 Details : it is now possible to store extended attributes in the Lustre
4717 client filesystem, and with the user_xattr mount option it
4718 is possible to allow users to store EAs on their files also
4720 Severity : enhancement
4722 Description: Add possibility (config option) to show minimal available OST free
4724 Details : When compiled with --enable-mindf configure option, statfs(2)
4725 (and so, df) will return least minimal free space available from
4726 all OSTs as amount of free space on FS, instead of summary of
4727 free spaces of all OSTs.
4729 Severity : enhancement
4731 Description: do not expand extent locks acquired on OST-side
4732 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
4733 by server: they are not cached anyway.
4736 Frequency : when mmap is used/binaries executed from Lustre
4738 Description: Unmmap pages before throwing them away from read cache.
4739 Details : llap_shrink cache now attempts to unmap pages before discarding
4740 them (if unmapping failed - do not discard). SLES9 kernel has
4741 extra checks that trigger if this unmapping is not done first.
4746 Description: lconf didn't resolve symlinks before checking to see whether a
4747 given mountpoint was already in use
4750 Frequency : when migrating failover services
4751 Bugzilla : 6395, 9514
4752 Description: When migrating a subset of services from a node (e.g. failback
4753 from a failover service node) the remaining services would
4754 time out and evict clients.
4755 Details : lconf --force (implied by --failover) sets the global obd_timeout
4756 to 5 seconds in order to quickly disconnect, but this caused
4757 other RPCs to time out too quickly. Do not change the global
4758 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
4760 Severity : enhancement
4761 Frequency : if MDS is started with down OST
4762 Bugzilla : 9439,5706
4763 Description: Allow startup/shutdown of an MDS without depending on the
4764 availability of the OSTs.
4765 Details : Asynchronously call mds_lov_synchronize during MDS startup.
4766 Add appropriate locking and lov-osc refcounts for safe
4767 cleaning. Add osc abort_inflight calls in case the
4768 synchronize never started.
4771 Frequency : occasional (Cray XT3 only)
4773 Description: root not authorized to access files in CRAY_PORTALS environment
4774 Details : The client process capabilities were not honoured on the MDS in
4775 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
4776 been accessed by an authorized user then root was able to access
4777 the file on the local client also. The root user capabilities
4778 are now allowed on the MDS, as this environment has secure UID.
4781 Frequency : occasional
4783 Description: ldiskfs "too long searching" message happens too often
4784 Details : A debugging message (otherwise harmless) prints too often on
4785 the OST console. This has been reduced to only happen when
4786 there are fragmentation problems on the filesystem.
4791 Description: Division by zero in statfs when all OSCs are inactive
4792 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
4793 lov_statfs divides by value returned by lov_get_stripecnt().
4797 Bugzilla : 9489, 3273
4798 Description: First write from each client to each OST was only 4kB in size,
4799 to initialize client writeback cache, which caused sub-optimal
4800 RPCs and poor layout on disk for the first written file.
4801 Details : Clients now request an initial cache grant at (re)connect time
4802 and so that they can start streaming writes to the cache right
4803 away and always do full-sized RPCs if there is enough data.
4804 If the OST is rebooted the client also re-establishes its grant
4805 so that client cached writes will be honoured under the grant.
4810 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
4811 Details : Now I/O RPCs go to different portal number and (presumably) fast
4812 lock requests (and glimses) and other RPCs get their own service
4813 threads pool that should be able to service those RPCs
4816 Severity : enhancement
4818 Description: Ability to exchange lustre version between client and servers and
4819 issue warnings at client side if client is too old. Also for
4820 liblustre clients there is ability to refuse connection of too old
4822 Details : New 'version' field is added to connect data structure that is
4823 filled with version info. That info is later checked by server and
4827 Frequency : rare, liblustre only.
4828 Bugzilla : 9296, 9581
4829 Description: Two simultaneous writes from liblustre at offset within same page
4830 might proceed at the same time overwriting eachother with stale
4832 Details : I/O lock withing llu_file_prwv was released too early, before data
4833 actually was hitting the wire. Extended lock-holding time until
4834 server acknowledges receiving data.
4837 Frequency : extremely rare. Never observed in practice.
4839 Description: avoid generating lustre_handle cookie of 0.
4840 Details : class_handle_hash() generates handle cookies by incrementing
4841 global counter, and can hit 0 occasionaly (this is unlikely, but
4842 not impossible, because initial value of cookie counter is
4843 selected randonly). Value of 0 is used as a sentinel meaning
4844 "unassigned handle" --- avoid it. Also coalesce two critical
4845 sections in this function into one.
4847 Severity : enhancement
4849 Description: allow liblustre clients to delegate truncate locking to OST
4850 Details : To avoid overhead of locking, liblustre client instructs OST to
4851 take extent lock in ost_punch() on client's behalf. New connection
4852 flag is added to handle backward compatibility.
4854 Severity : enhancement
4855 Bugzilla : 4928, 7341, 9758
4856 Description: allow number of OST service threads to be specified
4857 Details : a module parameter allows the number of OST service threads
4858 to be specified via "options ost ost_num_threads={N}" in the
4859 OSS's /etc/modules.conf or /etc/modprobe.conf.
4863 Bugzilla : 6146, 9635, 9895
4864 Description: servers crash with bad pointer in target_handle_connect()
4865 Details : In rare cases when a client is reconnecting it was possible that
4866 the connection request was the last reference for that export.
4867 We would temporarily drop the export reference and get a new
4868 one, but this may have been the last reference and the export
4869 was just destroyed. Get new reference before dropping old one.
4871 Severity : enhancement
4872 Frequency : if client is started with failover MDS
4874 Description: Allow multiple MDS hostnames in the mount command
4875 Details : Try to read the configuration from all specified MDS
4876 hostnames during a client mount in case the "primary"
4879 Severity : enhancement
4881 Description: Stop sending data to evicted clients as soon as possible.
4882 Details : Check if the client we are about to send or are sending data to
4883 was evicted already. (Check is done every second of waiting,
4884 for which l_wait_event interface was extended to allow checking
4885 of exit condition at specified intervals).
4888 Frequency : rare, normally only when NFS exporting is done from client
4890 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
4892 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
4895 Frequency : always, if srand() is called before liblustre initialization
4897 Description: Liblustre uses system PRNG disturbing its usage by user application
4898 Details : Introduce internal to lustre fast and high-quality PRNG for
4899 lustre usage and make liblustre and some other places in generic
4900 lustre code to use it.
4902 Severity : enhancement
4903 Bugzilla : 9477, 9557, 9870
4904 Description: Verify that the MDS configuration logs are updated when xml is
4905 Details : Check if the .xml configuration logs are newer than the config
4906 logs stored on the MDS and report an error if this is the case.
4907 Request --write-conf, or allow starting with --old_conf.
4909 Severity : enhancement
4911 Description: Handle symlinks in the path when checking if Lustre is mounted.
4912 Details : Resolve intermediate symlinks when checking if a client has
4913 mounted a filesystem to avoid duplicate client mounts.
4918 Description: lconf can hit an error exception but still return success.
4919 Details : The lconf command catches the Command error exception at the top
4920 level script context and will exit with the associated exit
4921 status, but doesn't ensure that this exit status is non-zero.
4926 Description: failure of ptlrpc thread startup can cause oops
4927 Details : Starting a ptlrpc service thread can fail if there are a large
4928 number of threads or the server memory is very fragmented.
4929 Handle this without oopsing.
4932 Frequency : always, only if liblustre and non-default acceptor port was used
4934 Description: liblustre cannot connect to servers with non-default acceptor port
4935 Details : tcpnal_set_default_params() was not called and was therefore
4936 ignoring the environment varaible TCPNAL_PORT, as well as other
4937 TCPNAL_ environment variables
4942 Description: two objects could be created on the same OST for a single file
4943 Details : If an OST is down, in some cases it was possible to create two
4944 objects on a single OST for a single file. No problems other
4945 than potential performance impact and spurious error messages.
4949 Bugzilla : 5681, 9562
4950 Description: Client may oops in ll_unhash_aliases
4951 Details : Client dcache may become inconsistent in race condition.
4952 In some cases "getcwd" can fail if the current directory is
4958 Description: Inode refcounting problems in NFS export code
4959 Details : link_raw functions used to call d_instantiate without obtaining
4960 extra inode reference first.
4964 Bugzilla : 9942, 9903
4965 Description: Referencing freed requests leading to crash, memleaks with NFS.
4966 Details : We used to require that call to ll_revalidate_it was always
4967 followed by ll_lookup_it. Also with revalidate_special() it is
4968 possible to call ll_revalidate_it() twice for the same dentry
4969 even if first occurence returned success. This fix changes semantic
4970 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
4971 reference on a request referred from the intent.
4972 ll_intent_release() then releases such a request.
4975 Frequency : rare, normally benchmark loads only
4977 Description: unlinked inodes were kept in memory on the client
4978 Details : If a client is repeatedly creating and unlinking files it
4979 can accumulate a lot of stale inodes in the inode slab cache.
4980 If there is no other client load running this can cause the
4981 client node to run out of memory. Instead flush old inodes
4982 from client cache that have the same inode number as a new inode.
4985 Frequency : SLES9 2.6.5 kernel and long filenames only
4986 Bugzilla : 9969, 10379
4987 Description: utime reports stale NFS file handle
4988 Details : SLES9 uses out-of-dentry names in some cases, which confused
4989 the lustre dentry revalidation. Change it to always use the
4993 Frequency : rare, unless heavy write-truncate concurrency is continuous
4994 Bugzilla : 4180, 6984, 7171, 9963, 9331
4995 Description: OST becomes very slow and/or deadlocked during object unlink
4996 Details : filter_destroy() was holding onto the parent directory lock
4997 while truncating+unlinking objects. For very large objects this
4998 may block other threads for a long time and slow overall OST
4999 responsiveness. It may also be possible to get a lock ordering
5000 deadlock in this case, or run out of journal credits because of
5001 the combined truncate+unlink. Solution is to do object truncate
5002 first in one transaction without parent lock, and then do the
5003 final unlink in a new transaction with the parent lock. This
5004 reduces the lock hold time dramatically.
5007 Frequency : rare, 2.4 kernels only
5009 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
5010 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
5011 before unmounting their filesystem in order to avoid deadlock.
5012 The kernel_locked() function in 2.4 kernels only checks whether
5013 the kernel lock is held, not whether it is this process that is
5014 holding it as 2.6 kernels do.
5019 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
5020 Details : The client ptlrpc code may be trying to reconnect to a down
5021 server before a previous connection attempt has timed out.
5022 Increase the reconnect interval to be longer than the connection
5023 timeout interval to avoid sending duplicate connections to
5027 Frequency : echo_client brw_test command
5029 Description: fix echo_client to work with OST preallocated code
5030 Details : OST preallocation code (5137) didn't take echo_client IO path
5031 into account: echo_client calls filter methods outside of any
5032 OST thread and, hence, there is no per-thread preallocated
5033 pages and buffers to use. Solution: hijack pga pages for IO. As
5034 a byproduct, this avoids unnecessary data copying.
5038 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
5039 Description: Client can oops in mdc_commit_close() after open replay
5040 Details : It was possible for the MDS to return an open request with no
5041 transaction number in mds_finish_transno() if the client was
5042 evicted, but without actually returning an error. Clients
5043 would later try to replay that open and may trip an assertion
5044 Simplify the client close codepath, and always return an error
5045 from the MDS in case the open is not successful.
5048 Frequency : rare, 2.6 OSTs only
5050 Description: OST may deadlock under high load on fragmented files
5051 Details : If there was a heavy load and highly-fragmented OST filesystems
5052 it was possible to have all the OST threads deadlock waiting on
5053 allocation of biovecs, because the biovecs were not released
5054 until the entire RPC IO was completed. Instead, release biovecs
5055 as soon as they are complete to ensure forward IO progress.
5057 Severity : enhancement
5059 Description: Support for specifying external journal device at mount
5060 Details : If an OST or MDS device is formatted with an external journal
5061 device, this device major/minor is stored in the ext3 superblock
5062 and may not be valid for failover. Allow detecting and
5063 specifying the external journal at mount time.
5068 Description: Mounting an MDS with pending unlinked files may cause oops
5069 Details : target_finish_recovery() calls mds_postrecov() which returned
5070 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
5071 considers this an error and immediately begins cleaning up the
5072 lov, just after starting the mds_lov process
5074 Severity : enhancement
5076 Description: Implement 'lfs df' to report actual free space on per-OST basis
5077 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
5078 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
5079 usage of objects; '-h' to report in human readable format.
5081 ------------------------------------------------------------------------------
5083 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
5090 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
5091 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
5092 at startup during orphan unlinking. Instead, skip buddy update
5093 before setup, as it will be regenerated after recovery is complete.
5098 Description: If an OST is inactive, its locks might reference stale inodes.
5099 Details : lov_change_cbdata() must iterate over all namespaces, even if
5100 they are inactive to clear inode references from the lock.
5102 Severity : enhancement
5103 Frequency : occasional, if non-standard max_dirty_mb used
5105 Description: Client will block write RPCs if not enough grant
5106 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
5107 then the client will block writes while waiting for another RPC
5108 to complete instead of consuming its dirty limit. With change
5109 we get improved performance when max_dirty_mb is small.
5111 Severity : enhancement
5112 Bugzilla : 3389, 6253
5113 Description: Add support for supplementary groups on the MDS.
5114 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
5115 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
5116 lookups for user supplementary groups into a cache.
5120 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
5121 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
5122 for liblustre O_EXCL is enforced by MDS.
5126 Description: readdir never returns NULL in liblustre.
5127 Details : Corrected llu_iop_getdirentries logic, to return offset of next
5128 dentry in struct dirent.
5132 Frequency : liblustre only, depends on application IO pattern
5133 Description: liblustre clients evicted if not contacting servers
5134 Details : Don't put liblustre clients into the ping_evictor list, so
5135 they will not be evicted by the pinger ever.
5137 Severity : enhancement
5139 Description: Add ability to evict clients by NID from MDS.
5140 Details : By echoing "nid:$NID" string into
5141 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
5142 $NID would be instantly evicted from this MDS and from all active
5143 OSTs connected to it.
5147 Description: Do not query file size twice, somewhat slowing stat(2) calls.
5148 Details : lookup_it_finish() used to query file size from OSTs that was not
5153 Description: service threads change working directory to that of init
5154 Details : Starting lustre service threads may pin the working directory
5155 of the parent thread, making that filesystem busy. Threads
5156 now change to the working directory of init to avoid this.
5160 Frequency : during shutdown only
5161 Description: shutdown with a failed MDS or OST can cause unmount to hang
5162 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
5167 Frequency : 2.6 only
5168 Description: chmod/chown may include an extra supplementary group
5169 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
5170 supplementary group and if none is specified this is used.
5173 Bugzilla : 5479 (6816)
5175 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
5176 Details : If lookup is in progress on a file that is unlinked we might try
5177 to revalidate the inode and fail in revalidate after lookup is
5178 complete and ll_file_open() enqueues the open again but
5179 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
5182 Frequency : always, if lconf --abort_recovery used
5184 Description: lconf --abort_recovery fails with 'Operation not supported'
5185 Details : lconf was attempting to abort recovery on the MDT device and not
5188 ------------------------------------------------------------------------------
5190 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
5195 Frequency : rare (only unsupported configurations with a node running as an
5197 Bugzilla : 6514, 5137
5198 Description: Mounting a Lustre file system on a node running as an OST could
5200 Details : OSTs now preallocates memory needed to write out data at
5201 startup, instead of when needed, to avoid having to
5202 allocate memory in possibly low memory situations.
5203 Specifically, if the file system is mounted on on OST,
5204 memory pressure could force it to try to write out data,
5205 which it needed to allocate memory to do. Due to the low
5206 memory, it would be unable to do so and the node would
5207 become unresponsive.
5209 Severity : enhancement
5211 Description: Addition of lconf --service command line option
5212 Details : lconf now accepts a '--service <arg>' option, which is
5213 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
5215 Severity : enhancement
5217 Description: Failover mode is now the default for OSTs.
5218 Details : By default, OSTs will now run in failover mode. To return to
5219 the old behaviour, add '--failout' to the lmc line for OSTs.
5221 Severity : enhancement
5223 Description: Health checks are now provided for MDS and OSTs
5224 Details : Additional detailed health check information on MSD and OSTs
5225 is now provided through the procfs health_check value.
5228 Frequency : occasional, depends on IO load
5230 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
5231 after numerous create/delete cycles
5232 Details : The ext3 inode allocation policy would not allocate new inodes
5233 very well on the OSTs because there are no new directories
5234 being created. Instead we look for groups with free space if
5235 the parent directories are nearly full.
5240 Description: Network or server problems during mount may cause partially
5241 mounted clients instead of returning an error.
5242 Details : The config llog parsing code may overwrite the error return
5243 code during mount error handling, returning success instead
5249 Description: MDS can fail to allocate large reply buffers
5250 Details : After long uptimes the MDS can fail to allocate large reply
5251 buffers (e.g. zconf client mount config records) due to memory
5252 fragmentation or consumption by the buffer cache. Preallocate
5253 some large reply buffers so that these replies can be sent even
5254 under memory pressure.
5258 Frequency : rare (liblustre)
5259 Description: fsx running with liblustre complained that using truncate() to
5260 extend the file doesn't work. This patch corrects that issue.
5261 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
5262 ATTR_SIZE and lsm use in llu_setattr_raw.
5266 Frequency : rare, only 2.6 kernels
5267 Description: Unusual file access patterns on the MDS may result in inode
5268 data being lost in very rare circumstances.
5269 Details : Bad interaction between the ea-in-inode patch and the "no-read"
5270 code in the 2.6 kernel caused the inode and/or EA data not to
5271 be read from disk, causing single-file corruption.
5275 Frequency : rare, only 2.6 filesystems using extents
5276 Description: Heavy concurrent write and delete load may cause data corruption.
5277 Details : It was possible under high-load situations to have an extent
5278 metadata block in the block device cache from a just-unlinked
5279 file overwrite a newly-allocated data block. We now unmap any
5280 metadata buffers that alias just-allocated data blocks.
5284 Frequency : filesystems with default stripe_count larger than 77
5285 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
5286 Details : lconf specifies an inode size of 4096 bytes when the default
5287 stripe_count is larger than 77. This conflicts with the default
5288 inode density of 1 per 4096 bytes. Allocate smaller inodes in
5289 this case to avoid pinning too much memory for large EAs.
5291 ------------------------------------------------------------------------------
5293 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
5298 Frequency : rare (extremely heavy IO load with hundreds of clients)
5300 Description: Client is evicted, gets IO error writing to file
5301 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
5302 caused clients to be evicted for AST timeouts. The fixes in
5303 bug 5192 mean we no longer need to have such short AST timeouts
5304 so ldlm_timeout has been increased.
5307 Frequency : occasional during --force or --failover shutdown under load
5308 Bugzilla : 5949, 4834
5309 Description: Server oops/LBUG if stopped with --force or --failover under load
5310 Details : a collection of import/export refcount and cleanup ordering
5311 issues fixed for safer force cleanup
5314 Frequency : only filesystems larger than 120 OSTs
5315 Bugzilla : 5990, 6223
5316 Description: lfs getstripe would oops on a very large filesystem
5317 Details : lov_getconfig used kfree on vmalloc'd memory
5320 Frequency : only filesystems exporting via NFS to Solaris 10 clients
5321 Bugzilla : 6242, 6243
5322 Description: reading from files that had been truncated to a non-zero size
5323 but never opened returned no data
5324 Details : ll_file_read() reads zeros from no-object files to EOF
5329 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
5330 Details : There's little harm in aborting MDS/OSS recovery and letting it
5331 try again, so I removed the LASSERT and return an error instead.
5333 Severity : enhancement
5335 Description: New debugging infrastructure for tracking down data corruption
5336 Details : The I/O checksum code was replaced to: (a) control it at runtime,
5337 (b) cover more of the client-side code path, and (c) try to narrow
5338 down where problems occurred
5342 Bugzilla : 3819, 4364, 4397, 6313
5343 Description: Racing close and eviction MDS could cause assertion in mds_close
5344 Details : It was possible to get multiple mfd references during close and
5345 client eviction, leading to one thread referencing a freed mfd.
5347 Severity: : enhancement
5348 Bugzilla : 3262, 6359
5349 Description: Attempts to reconnect to servers are now more aggressive.
5350 Details : This builds on the enhanced upcall-less recovery that was added
5351 in 1.4.2. When trying to reconnect to servers, clients will
5352 now try each server in the failover group every 10 seconds. By
5353 default, clients would previously try one server every 25 seconds.
5358 Description: After recovery, certain operations trigger a failed
5359 assertion on a client.
5360 Details : Failing over an mds, using lconf -d --failover, while a
5361 client was doing a readdir() call would cause the client to
5362 LBUG after recovery completed and the readdir() was resent.
5364 Severity : enhancement
5366 Description: Default groups are now added by lconf
5367 Details : You can now run lconf --group <servicename> without having to
5368 manually add groups with lmc.
5371 Frequency : occasional
5373 Description: Nodes with an elan id of 0 trigger a failed assertion
5376 Frequency : always when accessing e.g. tty/console device nodes
5378 Description: tty and some other devices nodes cannot be used on lustre
5379 Details : file's private_data field is used by device data and lustre
5380 values in there got lost. New field was added to struct file to
5381 store fs-specific private data.
5384 Frequency : when exporting Lustre via NFS
5386 Description: NFSD failed occasionally when looking up a path component
5387 Details : NFSD is looking up ".." which was broken in ext3 directories
5388 that had grown large enough to become hashed.
5391 Frequency : Clusters with multiple interfaces not on the same subnet
5393 Description: Nodes will repeatedly try to reconnect to an interface which it
5394 cannot reach and report an error to the log.
5395 Details : Extra peer list entries will be created by lconf with some peers
5396 unreachable. lconf now validates the peer before adding it.
5399 Frequency : Only if a default stripe is set on the filesystem root.
5401 Description: Setting a default stripe on the filesystem root prevented the
5402 filesystem from being remounted.
5403 Details : The client was sending extra request flags in the root getattr
5404 request and did not allocate a reply buffer for the dir EA.
5407 Frequency : occasional, higher if lots of files are accessed by one client
5408 Bugzilla : 6159, 6097
5409 Description: Client trips assertion regarding lsm mismatch/magic
5410 Details : While revalidating inodes the VFS looks up inodes with ifind()
5411 and in rare cases can find an inode that is being freed.
5412 The ll_test_inode() code will free the lsm during ifind()
5413 when it finds an existing inode and then the VFS later attaches
5414 this free lsm to a new inode.
5418 Bugzilla : 6422, 7030
5419 Description: MDS deadlock between mkdir and client eviction
5420 Details : Creating a new file via mkdir or mknod (starting a transaction
5421 and getting the ns lock) can deadlock with client eviction
5422 (gets ns lock and trying to finish a synchronous transaction).
5425 Frequency : occasional
5426 Description: While starting a server, the fsfilt_ext3 module could not be
5428 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
5429 kernels. Previously, lconf would still use the ext3 name
5430 when trying to load modules. Now, it will correctly use
5431 ext3 on 2.4 and ldiskfs on 2.6.
5433 Severity : enhancement
5434 Description: The default stripe count has been changed to 1
5435 Details : The interpretation of the default stripe count (0, to lfs
5436 or lmc) has been changed to mean striping across a single
5437 OST, rather than all available. For general usage we have
5438 found a stripe count of 1 or 2 works best.
5440 Severity : enhancement
5441 Description: Add support for compiling against Cray portals.
5442 Details : Conditional compiling for some areas that are different
5446 Frequency : occasional
5447 Bugzilla : 6409, 6834
5448 Description: Creating files with an explicit stripe count may lead to
5449 a failed assertion on the MDS
5450 Details : If some OSTs are full or unavailable, creating files may
5451 trigger a failed assertion on the MDS. Now, Lustre will
5452 try to use other servers or return an error to the
5456 Frequency : occasional
5458 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
5459 caused lock timeout during readahead (since 1.4.2).
5460 Details : Processes doing readahead might match a lock that hasn't been
5461 granted yet if there are overlapping and conflicting lock
5462 requests. The readahead process waits on ungranted lock
5463 (original lock is CBPENDING), while OST waits for that process
5464 to cancel CBPENDING read lock and eventually evicts client.
5466 Severity : enhancement
5468 Description: Initial enabling of flock support for clients
5469 Details : Implements fcntl advisory locking and file status functions.
5470 This feature is provided as an optional mount flag (default
5471 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
5472 locking are implemented yet, and those that are are not guaranteed
5473 to be completely correct in production environments.
5474 mount -t lustre -o [flock|noflock] ...
5477 Frequency : occasional
5479 Description: OSTs running 2.4 kernels but with extents enabled might trip an
5480 assertion in the ext3 JBD (journaling) layer.
5481 Details : The b_committed_data struct is protected by the big kernel lock
5482 in 2.4 kernels, serializing journal_commit_transaction() and
5483 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
5484 access to this struct. In 2.6 kernels there is finer grained
5485 locking to improve SMP performance of the JBD layer.
5489 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
5494 Description: merge and cleanup kernel patches.
5497 -----------------------------------------------------------------------------
5499 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
5501 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
5502 versions of Lustre. Please update all servers and clients to
5503 version 1.4.2 or later at the same time. You must also run
5504 "lconf --write-conf {config}.xml" on the MDS while it is stopped
5505 to update the configuration logs.
5507 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
5508 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
5509 - swab llog records of type '0' so we get proper header size/idx (5861)
5510 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
5511 - fix rename of one directory over another leaking an inode (5953)
5512 - avoid SetPageDirty on 2.6 (5981)
5513 - don't re-add just-being-destroyed locks to the waiting list (5653)
5514 - when creating new directories, inherit the parent's custom
5515 striping settings if present parent (3048)
5516 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
5517 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
5518 - don't hold client locks on temporary worklist from l_lru (5666)
5519 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
5520 - automatically evict dead clients (5921)
5521 - Update file size properly in create+truncate+fstat case (6196)
5522 - Do not unhash mountpoint dentries, do not allow removal of
5524 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
5525 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
5526 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
5527 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
5528 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
5529 - Correctly handle reads of files with no objects (6243)
5530 - lctl recover will also mark a device active if deactivate used (5933)
5532 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
5533 - allow --write-conf on an MDS with different nettype than client (5619)
5534 - don't write config llogs to MDS for mounts not from that MDS (5617)
5535 - lconf should create multiple TCP connections from a client (5201)
5536 - init scripts are now turned off by default; run chkconfig --on
5537 lustre and chkconfig --on lustrefs to use them
5538 - upcalls are no longer needed for clients to recover to failover
5540 - add --abort-recovery option to lconf to abort recovery on device
5542 - add support for an arbitrary number of OSTs (3026)
5543 - Quota support protocol changes.
5544 - forward compatibility changes to wire structs (6007)
5545 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5546 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5547 - improved "lustre status" script
5548 - initialize blocksize for non-regular files (6062)
5549 - added --disable-server and --disable-client configure options (5782)
5550 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5551 - Vanilla 2.4.29 support
5552 - increase maximum number of obd devices to 520 (6242)
5553 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5554 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5555 SLES 9 2.6.5 kernels
5556 - limit stripes per file to 160 (the maximum EA size) (6093)
5558 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5561 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5562 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5563 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5564 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5565 - small ext3 extents cleanups and fixes (5733)
5566 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5567 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5568 - avoid needless client->OST connect, fix handle mismatch (5317)
5569 - fix DLM error path that led to out-of-sync client, long delays (5779)
5570 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5571 - fix several locking issues related to i_size (5492,5624,5654,5672)
5572 - don't move pending lock onto export if it is already evicted (5683)
5573 - fix kernel oops when creating .foo in unlinked directory (5548)
5574 - fix deadlock in obdfilter statistics vs. object create (5811)
5575 - use time_{before,after} to avoid timer jiffies wrap (5882)
5576 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5577 - Do not leak request if server was not able to process it (5154)
5578 - If mds_open unable to find parent dir, make that negative lookup(5154)
5579 - don't create new directories with extent-mapping (5909, 5936)
5581 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5582 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5583 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5584 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5585 - support for RHEL4 kernel on i686 (5773)
5586 - provide error messages when incompatible logs are encountered (5898)
5588 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5589 * version 1.4.0.10 (1.4.1 release candidate 1)
5591 - don't keep a lock reference when lock is not granted (4238)
5592 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5593 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5594 - drop import inflight refcount on signal_completed_replay error (5255)
5595 - unlock page after async write error during send (3677)
5596 - handle missing objects in filter_preprw_read properly (5265)
5597 - no transno return for symlink open, don't save no-trasno open (3440)
5598 - don't try to complete elan receive that already failed (4012)
5599 - free RPC server reply state on error (5406)
5600 - clean up thread from ptlrpc_start_thread() on error (5160)
5601 - readahead could read extra page into cache that wasn't ejected (5388)
5602 - prevent races in class_attach/setup/cleanup/detach (5260)
5603 - don't dereference de->d_inode after l_dput of de (5458)
5604 - use "int" for stripe value returned from lock_to_stripe (5544)
5605 - mballoc allocation and error-checking fixes in 2.6 (5504)
5606 - block device patches to fix I/O request sizes in 2.6 (5482)
5607 - look up hostnames for IB nals (5602)
5608 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5609 - don't start multiple acceptors for the same port (5277)
5610 - fix incorrect LASSERT in mds_getattr_name (5635)
5611 - export a proc file for general "ping" checking (5628)
5612 - fix "lfs check" to not block when the MDS is down (5628)
5614 - service request history (4965)
5615 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5616 - create an "evict_client" /proc entry on OSTs, like the MDS has
5617 - fix mount usage message, return errors per mount(8) (5168)
5618 - change grep [] to grep "[]" in tests so they work in more UMLs
5619 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5620 - remove ancient LOV_MAGIC_V0 EA support (5047)
5621 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5622 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5623 - allow readahead allocations to fail when low on memory (5383)
5624 - mmap locking landed again, after considerable improvement (2828)
5625 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
5627 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
5630 - send OST transaction number in read/write reply to free req (4966)
5631 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5632 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5633 - the watchdog thread now runs as interruptible (5246)
5634 - flock/lockf fixes (but it's still disabled, pending 5135)
5635 - don't use EXT3 constants in llite code (5094)
5636 - memory shortage at startup could cause assertion (5176)
5638 - reorganization of lov code
5639 - single portals codebase
5641 - add extents/mballoc support (5025)
5642 - direct I/O reads in the obdfilter (4048)
5643 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
5645 tbd Cluster File Systems, Inc. <info@clusterfs.com>
5648 - send OST transaction number in read/write reply to free req (4966)
5649 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5650 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5651 - the watchdog thread now runs as interruptible (5246)
5652 - handle missing objects in filter_preprw_read properly (5265)
5653 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5654 - drop import inflight refcount on signal_completed_replay error (5255)
5655 - unlock page after async write error during send (3677)
5656 - return original error code on reconstructed replies (3761)
5657 - no transno return for symlink open, don't save no-trasno open (3440)
5659 - add pid to ldlm debugging output (4922)
5660 - bump the watchdog timeouts -- we can't handle 30sec yet
5661 - extra debugging for orphan dentry/inode bug (5259)
5663 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
5666 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
5667 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
5668 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
5669 - fix NULL dereference in /proc/sys/portals/routes (4827)
5670 - allow failed mdc_close() operations to be interrupted (4561)
5671 - stop precreate on OST before MDS would time out on it (4778)
5672 - don't send partial-page writes before EOF from client (4410)
5673 - discard client grant for sub-page writes on large-page clients (4520)
5674 - don't free dentries not owned by NFS code, check generation (4806)
5675 - fix lsm leak if mds_create_objects() fails (4801)
5676 - limit debug_daemon file size, always print CERROR messages (4789)
5677 - use transno after validating reply (3892)
5678 - process timed out requests if import state changes (3754)
5679 - update mtime on OST during writes, return in glimpse (4829)
5680 - add mkfsoptions to LDAP (4679)
5681 - use ->max_readahead method instead of zapping global ra (5039)
5682 - don't interrupt __l_wait_event() during strace
5684 - add software watchdogs to catch hung threads quickly (4941)
5685 - make lustrefs init script start after nfs is mounted
5686 - fix CWARN/ERROR duplication (4930)
5687 - return async write errors to application if possible (2248)
5688 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
5689 - print NAL number in %x format (4645)
5690 - update barely-supported suse-2.4.21-171 series (4842)
5691 - support for sles 9 %post scripts
5692 - support for building 2.6 kernel-source packages
5693 - support for sles km_* packages
5695 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
5698 - ignore -ENOENT errors in osc_destroy (3639)
5699 - notify osc create thread that OSC is being cleaned up (4600)
5700 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
5701 - reconstruct ost_handle() like mds_handle() (4657)
5702 - create a new thread to do import eviction to avoid deadlock (3969)
5703 - let lconf resolve symlinked-to devices (4629)
5704 - don't unlink "objects" from directory with default EA (4554)
5705 - hold socknal file ref over connect in case target is down (4394)
5706 - allow more than 32000 subdirectories in a single directory (3244)
5707 - fix blocks count for O_DIRECT writes (3751)
5708 - OST returns ENOSPC from object create when no space left (4539)
5709 - don't send truncate RPC if file size isn't changing (4410)
5710 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
5711 - bind to privileged port in socknal and tcpnal (3689)
5713 - rate limit CERROR/CWARN console message to avoid overload (4519)
5714 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
5715 - basic mmap support (3918)
5716 - kernel patch series update from b1_4 (4711)
5718 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
5721 - avoid crash during MDS cleanup with OST shut down (2775)
5722 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
5723 - don't use bad inodes on the MDS (3744)
5724 - dynamic object preallocation to improve recovery speed (4236)
5725 - don't hold spinlock over lock dumping or change debug flags (4401)
5726 - don't zero obd_dev when it is force cleaned (3651)
5727 - print grants to console if they go negative (4431)
5728 - "lctl deactivate" will stop automatic recovery attempts (3406)
5729 - look for existing locks in ldlm_handle_enqueue() (3764)
5730 - don't resolve lock handle twice in recovery avoiding race (4401)
5731 - revalidate should check working dir is a directory (4134)
5733 - don't always mark "slow" obdfilter messages as errors (4418)
5735 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
5738 - don't close LustreDB during write_conf until it is done (3860)
5739 - fix typo in lconf for_each_profile (3821)
5740 - allow dumping logs from multiple threads at one time (3820)
5741 - don't allow multiple threads in OSC recovery (3812)
5742 - fix debug_size parameters (3864)
5743 - fix mds_postrecov to initialize import for llog ctxt (3121)
5744 - replace config semaphore with spinlock (3306)
5745 - be sure to send a reply for a CANCEL rpc with bad export (3863)
5746 - don't allow enqueue to complete on a destroyed export (3822)
5747 - down write_lock before checking llog header bitmap (3825)
5748 - recover from lock replay timeout (3764)
5749 - up llog sem before sending rpc (3652)
5750 - reduce ns lock hold times when setting kms (3267)
5751 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
5752 - fix NULL deref and obd_dev leak on setup error (3312)
5753 - replace some LBUG about llog ops with error handling (3841)
5754 - don't match INVALID dentries from d_lookup and spin (3784)
5755 - hold dcache_lock while marking dentries INVALID and hashing (4255)
5756 - fix invalid assertion in ptlrpc_set_wait (3880)
5758 - add libwrap support for the TCP acceptor (3996)
5759 - add /proc/sys/portals/routes for non-root route listing (3994)
5760 - allow setting MDS UUID in .xml (2580)
5761 - print the stack of a process that LBUGs (4228)
5763 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
5766 - don't cleanup request in ll_file_open() on failed MDS open (3430)
5767 - make sure to unset replay flag from failed open requests (3440)
5768 - if default stripe count is 0, use OST count for inode size (3636)
5769 - update parent mtime/ctime on client for create/unlink (2611)
5770 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
5771 - free recovery state on server during a forced cleanup (3571)
5772 - unregister_reply for resent reqs (3063)
5773 - loop back devices mounting and status check on 2.6 (3563)
5774 - fix resource-creation race that can provoke i_size == 0 (3513)
5775 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
5776 - more debugging for page-accounting assertion (3746)
5777 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
5778 - avoid deadlock after precreation failure (3758)
5779 - fix race and lock order deadlock in orphan handling (3450, 3750)
5780 - add validity checks when grabbing inodes from l_ast_data (3599)
5782 - add /proc/.../recovery_status to obdfilter (3428)
5783 - lightweight CDEBUG infrastructure, debug daemon (3668)
5784 - change default OSC RPC parameters to be better on small clusters
5785 - turn off OST read cache for files smaller than 32MB
5786 - install man pages and include them in rpms (3100)
5787 - add new init script for (un)mounting lustre filesystems (2593)
5788 - run chkconfig in %post for init scripts (3701)
5789 - drop scimac NAL (unmaintained)
5791 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
5794 - clean kiobufs before and after use (3485)
5795 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
5796 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
5797 - change init script to fail more clearly if not run as root (1528)
5798 - allow clients to reconnect during replay (1742)
5799 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
5800 - don't do DNS lookups on NIDs too small for IP addresses (3442)
5801 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
5802 - fix cond_resched (3554)
5803 - only evict unfinished clients after recovery (3515)
5804 - allow bulk resend, prevent data loss (3570)
5805 - dynamic ptlrpc request buffer allocation (2102)
5806 - don't allow unlinking open directory if it isn't empty (2904)
5807 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
5808 - remove extraneous obd dereference causing LASSERT failure (3334)
5809 - don't use get_cycles() when creating temp. files on the mds (3156)
5810 - hold i_sem when setting i_size in ll_extent_lock() (3564)
5811 - handle EEXIST for set-stripe, set proper directory name (3336)
5813 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
5814 - fix ksocknal_fmb_callback() error messages (2918)
5816 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
5819 - don't copy lvb into (possibly NULL) reply on error (2983)
5820 - don't deref dentry after dput, don't free lvb on error (2922)
5821 - use the kms to determine writeback rpc length (2947)
5822 - increment oti_logcookies when osc is inactive (2948)
5823 - update client's i_blocks count via lvb messages (2543)
5824 - handle intent open/close of special files properly (1557)
5825 - mount MDS with errors=remount-ro, like obdfilter (2009)
5826 - initialize lock handle to avoid ASSERT on error cleanup (3057)
5827 - don't use cancelling-locks' kms values (2947)
5828 - use highest lock extent for kms, not last one (2925)
5829 - don't dereference ERR_PTR() dentry in error handling path (3107)
5830 - fix thread race in portals_debug_dumplog() (3122)
5831 - create lprocfs device entries at setup instead of at attach (1519)
5832 - common AST error handler, don't evict client on completion race (3145)
5833 - zero nameidata in detach_mnt in 2.6 (3118)
5834 - verify d_inode after revalidate_special is valid in 2.6 (3116)
5835 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
5836 - initialize RPC timeout timer earlier for 2.6 (3219)
5837 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
5838 - print nal/nid for unknown nid (3258)
5839 - additional checks for oscc recovery before doing precreate (3284)
5840 - fix ll_extent_lock() error return code for 64-bit systems (3043)
5841 - don't crash in mdc_close for bad permissions on open (3285)
5842 - zero i_rdev for non-device files (3147)
5843 - clear page->private before handing to FS, better assertion (3119)
5844 - tune the read pipeline (3236)
5845 - fix incorrect decref of invalidated dentry (2350)
5846 - provide read-ahead stats and refine rpc in flight stats (3328)
5847 - don't hold journal transaction open across create RPC (3313)
5848 - update atime on MDS at close time (3265)
5849 - close LDAP connection when recovering to avoid server load (3315)
5850 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
5851 - don't leak open file on MDS after open resend (3325)
5852 - serialize filter_precreate and filter_destroy_precreated (3329)
5853 - loop device shouldn't call sync_dev() for nul device (3092)
5854 - clear page cache after eviction (2766)
5855 - resynchronize MDS->OST in background (2824)
5856 - refuse to mount the same filesystem twice on same mountpoint (3394)
5857 - allow llmount to create routes for mounting behind routers (3320)
5858 - push lock cancellation to blocking thread for glimpse ASTs (3409)
5859 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
5860 - fix rare problem with rename on htree directories (3417)
5862 - allow default OST striping configuration per directory (1414)
5863 - fix compilation for qswnal for 2.6 kernels (3125)
5864 - increase maximum number of MDS request buffers for large systems
5865 - change liblustreapi to be useful for external progs like lfsck (3098)
5866 - increase local configuration timeout for slow disks (3353)
5867 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
5869 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5872 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
5873 - don't overwrite extent policy data in reply if lock was blocked (2901)
5874 - drop filter export grants atomically with removal from device (2663)
5875 - del obd_self_export from work_list in class_disconnect_exports (2908)
5876 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
5877 - swab reply message in mdc_close, other PPC fixes (2464)
5878 - fix destroying of named logs (2325)
5879 - overwrite old logs when running lconf --write_conf (2264)
5880 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
5881 - fix race in target_handle_connect (2898)
5882 - mds_reint_create() should take same inode create lock (2926)
5883 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
5884 - don't close files for self_export to avoid uninitialized obd (2936)
5885 - allow MDS with the same name as client node (2939)
5886 - hold dentry reference for closed log files for unlink (2325)
5887 - reserve space for all logs during transactions (2059)
5888 - don't evict page beyond end of stripe extent (2925)
5889 - don't oops on a deleted current working directory (2399)
5890 - handle hard links to targets without a parent properly (2517)
5891 - don't dereference NULL lock when racing during eviction (2867)
5892 - don't grow lock extents when lots of conflicting locks (2919)
5894 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
5897 - account for cache space usage on clients to avoid data loss (974)
5898 - lfsck support in lustre kernel code (2349)
5899 - reduce journal credits needed for BRW writes (2370)
5900 - orphan handling to avoid losing space on client/server crashes
5901 - ptlrpcd can be blocked, stopping ALL progress (2477)
5902 - use lock value blocks to assist in proper KMS, faster stat (1021)
5903 - takes i_sem instead of DLM locks internally on obdfilter (2720)
5904 - recovery for initial connections (2355)
5905 - fixes for mds_cleanup_orphans (1934)
5906 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
5907 - block all file creations until orphan recovery completes (1901)
5908 - client remove rq_connection from request struct (2423)
5909 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
5910 - recovery timer race (2670)
5911 - mdc_close recovey bug (2532)
5912 - ptlrpc cleanup bug (2710)
5913 - mds timeout on local locks (2588)
5914 - namespace lock held during RPCs (2431)
5915 - handle interrupted sync write properly (2503)
5916 - don't try to handle a message that hasn't been replied to (2699)
5917 - client assert failure during cleanup after abort recovery (2701)
5918 - leak mdc device after failed mount (2712)
5919 - ptlrpc_check_set allows timedout requests to complete (2714)
5920 - wait for inflight reqs when ptlrpcd finishes (2710)
5921 - make sure unregistered services are removed from the srv_list
5922 - reset bulk XID's when resending them (caught by 1138 test)
5923 - unregister_bulk after timeout
5924 - fix lconf error (2694)
5925 - handle write after unfinished setstripe, stripe-only getstripe (2388)
5926 - readahead locks pages, leaves pending causing memory pressure (2673)
5927 - increase OST request buffers to 4096 on large machines (2729)
5928 - fix up permission of existing directories in simple_mkdir (2661)
5929 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
5930 - don't assign transno to errored transactions (2742)
5931 - don't delete objects on OST if given a bogus objid from MDS (2751)
5932 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
5933 - if rq_no_resend, then timeout request after recovery (2432)
5934 - fix MDS llog_logid record size, 64-bit array alignment (2733)
5935 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
5936 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
5937 - MDS recovery shouldn't delete objects that it has given out (2730)
5938 - if enqueue arrives after completion, don't clobber LVB (2819)
5939 - don't unlock pages twice when trigger_group_io returns error (2814)
5940 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
5941 - don't write pages to disk if there was an error (1450)
5942 - don't ping imports that have recovery disabled (2676)
5943 - take buffered bytes into account when balancing socknal conn (2817)
5944 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
5945 - reconnect unlink llog connection after MDS reconnects to OST (2816)
5946 - remove little-endian swabbing of llog records (1987)
5947 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
5948 - retry reposting request buffers if they fail (1191)
5949 - grow extent at grant time to avoid granting a revoked lock (2809)
5950 - lock revoke doesn't evict page if covered by a second lock (2765)
5951 - disable VM readahead to avoid reading outside lock extents (2805)
5953 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
5954 - updated kernel patches for hp-2.4.20 kernel (2681)
5956 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
5959 - fix truncated write corruption (2366)
5960 - fix for failed assertion in iopen_connect_dentry (1792,2517)
5962 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
5963 - ensure len(uuid) < 37 in lmc (1171)
5964 - fix ia64 OOPS in llog_test (2255)
5965 - zero end of page at obdfilter for partial page writes (2648)
5966 - don't leave stale dentries around after renames (bug 2428)
5967 - fix timeouts when evicting a client with a single lock held (2642)
5968 - set deadline for the initial HELLO message to drain (2634)
5969 - print out dotted-quad IP addresses in the socknal (2302)
5971 - additional debugging for MDS client eviction problem (2443)
5972 - fix mkfsoptions support for osts (2603, 2604)
5974 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
5977 - add series for the vanilla 2.6.0 kernel
5978 - add series for the vanilla 2.4.24 kernel
5979 - add series for a cray x86/64 UL kernel drop
5980 - fix xattr patches for the vanilla 2.4.19 series
5982 - generate true UUIDs in lmc (1171)
5983 - have portals stack dumping break in UML (2466)
5984 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
5985 - allocate with _NOFS in ldlm to avoid deadlock (1933)
5986 - wake callback waiting threads on client eviction (2460)
5987 - Add --ptldebug and --subsystem to lmc (1719)
5988 - update assertion to allow safe interrupt allocation
5989 - set rq_no_resend for cancel requests (2432)
5990 - recalculate ptlrpcd timeout after resend (2494)
5991 - call vfs_rmdir when removing pending directories (2368)
5992 - fix renaming a file to itself (2429)
5993 - lmc creates a default one-stripe lov (2454)
5994 - expand procfs space to handle large clusters (2326)
5995 - increase UML stack to avoid overflow
5996 - update lconf's list of debug and subsystem masks
5997 - fix lfs find --obd (2510)
5998 - /proc tunable for disabling filter read caching (2591)
5999 - stop rpm packages from altering slapd.conf (2301)
6000 - disable nagle in the socknal under 0conf (2578)
6001 - choose mds inode size based on stripe count (2572)
6002 - fix kernel-source rpm problems (2516)
6004 - add --disable-doc to avoid pdf generation (2421)
6005 - update documentation, tests, type-os, comments
6006 - avoid format warnings on ia64
6007 - remove the TOE NAL
6008 - tiny code cleanups by removing unused fields
6010 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
6013 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
6014 - remove the most blatant lies from BUILDING file (2371)
6015 - change default debug level to reasonable production setting
6016 - reduce client side cache size to reduce cache flush time
6017 - reduce max RPCs in flight to avoid unnecessary file fragmentation
6018 - make TCP zerocopy and pinger support enabled by default (2476)
6019 - sync writes completed after process exits caused crashes (2319)
6020 - maintain correct mount count on the MDS (2356)
6021 - backout 1557, because 2316 wasn't really fixed
6022 - better file I/O statistics gathering in /proc
6023 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
6024 - another kernel patch to fix zero-copy TCP function export
6025 - don't take duplicate lock when processing re-sent getattr (2420)
6026 - lctl uses obd_self_export instead of creating new conn (2353)
6027 - MDS/OST recovery case which requires object creation asserted (2425)
6028 - move lfs from /usr/sbin to /usr/bin in packages
6029 - fix race between mds_client_add and mds_client_free (2417)
6030 - use kmalloc instead of slabs in portals (2430)
6031 - don't create duplicate records when a failover MDS is present (2442)
6032 - remove unnecessary mount age check (2332)
6033 - don't remove directory inodes from locks prematurely (2451)
6034 - don't break if MDS service name is the same as hostname (2103)
6035 - fix races in client write RPC generation when cache full (2482)
6037 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
6040 - remove now-unused request->rq_obd (278)
6041 - if an allocation fails, print out how much memory we've used (1933)
6042 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
6043 - add the "configurable stack size" patch to most series files (1256)
6044 - ability to write large log records, for 100+ OST configs (2306)
6045 - fix NULL deref when filter_prep fails (2314)
6046 - fix operator precedence error in filter_sync
6047 - dynamic allocation of socknal TX descriptors (2315)
6048 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
6049 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
6050 - ensure that CWARN messages in Portals always get to the syslog
6051 - __init/__exit are not for prototype decls (ldlm_init/exit)
6052 - x86-64 compile warning fixes
6053 - fix gateway LMC keyword conflict (2318)
6054 - fix MDS lock inversions in getattr/reint paths (1844)
6055 - fix a rare lock re-ordering bug, which caused deadlock (2322)
6056 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
6057 - DLM race condition prevented some lock evictions (2328)
6058 - ENOMEM detection and retry on socknal sends (2230)
6059 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
6060 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
6061 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
6062 - better error messages when a client is rejected during recovery (1505)
6063 - avoid cancelling locks which were never granted, after failure (2330)
6064 - fix i_sem/journal inversion in mds_client_add (2333)
6065 - fix truncate/getattr lock cycle deadlock (2334)
6066 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
6067 - fix two rare exit paths which could leak an l_lock() ref (2321)
6068 - fencepost error in MDS/OST orphan recovery (2226)
6069 - make log record alignment 8 bytes (1988)
6070 - lstripe now fails when requested offset > ost_count (2237)
6071 - ensure that all kernel series have a complete list.h (1607)
6072 - fix crashes in special-file operations (2316)
6073 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
6075 - allow configurable automake binary, for testing new versions
6076 - small update to the lfs documentation
6078 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
6080 * fix negative export reference count in fsfilt_sync (2312)
6082 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
6083 * release candidate 0.9.1
6085 - orphans are moved into the PENDING directory for possible recovery
6086 - replayed opens now open by fid for orphan/rename safety (1042)
6087 - last close of an orphan inode generates a transno (683)
6088 - chdir() and mount() now pin the directory entry (1020)
6089 - avoid CERROR in normal ll_setattr_raw() error case (1500)
6090 - discard very old requests without processing them (1502)
6091 - remove some common, well-understood CERRORs (1505)
6092 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
6093 - clear "grant" flags in OST replies until OST grant code lands (1644)
6094 - fix read performance by not clobbering i_blksize on client (1598)
6095 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
6096 - make LRU size a /proc tunable, clears locks when reduced (707)
6097 - fix some lprocfs rot that prevented ptlbd from loading (1732)
6098 - server locks take references on exports now (1558)
6099 - build fixes for 2.4.20-rh trees (1663)
6100 - return an error from lov_create if all OSCs are inactive (1751)
6101 - fix import levels when a reconnect happens without a timeout (1597)
6102 - exit early from mds_open if we get a lookup error (1749)
6103 - partial page read at EOF wouldn't wait for disk before sending (1642)
6104 - avoid NULL deref in obdfilter when reading page past EOF (1592)
6105 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
6106 - fix LBUG in ll_it_open_error with rc = -2 (1861)
6107 - write/truncate lock inversion (1639)
6108 - Don't auto-load obdclass, portals modules during cleanup (1495)
6109 - fix timestamps from jumping to "now" (1763)
6110 - extra journal assertions (1648)
6111 - add an extra multiunlink test (1771)
6112 - fix read_record/write_record API (1776)
6113 - fix leak of offset_extent, possible incorrect i_size later (1772)
6114 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
6115 - Debugging for the kqswnal_get_idle_tx problems (1820)
6116 - Allow recovery to be attempted multiple times (1536)
6117 - Write out MDS last_rcvd file after it is first created (1600)
6118 - Fix tx_descriptor leak in failed transmit situations (1827)
6119 - ext3 journaling fixes for assertion failure after IO error (1871)
6120 - class_export_put() on freed export after completion AST error (1896)
6121 - Fix revalidate looping in VFS (1322)
6122 - Don't access a freed export during MDS_REINT timeout (1521)
6123 - Add open-unlink recovery support on the MDS (1673,1764)
6124 - Return an error if no MDS data was read from last_rcvd (1946)
6125 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
6126 - Don't setup a disk device that doesn't match exported UUID (317)
6127 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
6128 - avoid committing NULL handle in force close
6129 - local.sh is now a one-stripe LOV configuration
6130 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
6131 - fix ext3 htree duplicate directory entry corruption (1516)
6132 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
6133 - update to new LOV EA format (2097)
6134 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
6136 2003-06-15 Phil Schwan <phil@clusterfs.com>
6139 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
6140 - per-import/export recovery handling (958, 931, 959)
6141 - multiple last-rcvd slots, for serving multiple FSes (949)
6142 - connections are again shared between multiple imp/exports (963, 964)
6143 - "umount -f" would hang if any requests needed to be sent (393, 978)
6144 - avoid pinning large req buffer by copying for queued messages (989)
6145 - add "uuid" to "lctl device" command to help upcalls (991)
6146 - "open" RPCs with transnos would confuse recovery counters (1037)
6147 - do proper endian conversion of all wire messages (288, 340, 891)
6148 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
6149 - call waiting locks callback from LDLM recovery thread (1127, 1151)
6150 - fix ptlrpc_connection leak in target_handle_connect (1174)
6151 - fix import refcounting bug in OST and MDS cleanup (1134)
6152 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
6153 - fix very unlikely obd_types race condition (501)
6154 - remove osc_open hack for echo_client (1187)
6155 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
6156 - a failure in read_inode2 leads to deadlock (1139)
6157 - cancel ack-locks as soon as transaction is committed (1072)
6158 - fix major leaks and crashes in the bulk I/O path (937, 1057)
6159 - make sure to commitrw after any preprw to avoid deadlock (1162)
6160 - failing to execute a file in a lustre FS would lock inode (1203)
6161 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
6162 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
6163 - obd_uuid2tgt would walk off the end of the list (1255)
6164 - on IA64 the osc would give portals incorrect bulk size (1258)
6165 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
6166 - fix lock inversion caused by new llite matching code (1282)
6167 - limit the number of dirty pages on a client to 10MB (1286)
6168 - timed out locks were not being corrected cancelled (1289)
6169 - fix O_DIRECT above 4GB on IA-32 (1292)
6170 * major user-visible changes
6171 - fail out/fail over policy now controlled by the upcall (993)
6173 - add OBD_PING to check server availability and failure (954)
6174 - lustre messages are now sent in sending host order (288, 340, 891)
6175 - add eadatalen to MDS getattr reply (340)
6176 - OST read replies may contain second buffer, with per-page status (593)
6178 2003-03-11 Phil Schwan <phil@clusterfs.com>
6181 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
6182 - failed open()s could cause deadlock; fixed (867, 869)
6183 - stop cancelling OST locks when files are closed (481)
6184 - overlapping XID spaces caused network corruption (851, 853)
6185 - fix unsafe fsfilt counter arithmetic; change to atomic_t
6186 - setattr_raw added, to do single-RPC, server-side setattrs
6187 - lmc/lconf syntax change for OST UUIDs
6188 - fix crashy race condition between ptlrpc_free_req and osc_close
6189 - don't use request in mdc_enqueue if we hit a timeout (889)
6190 - don't set the inode i_size for regular files from the MDS (896)
6191 - handle out of order completion AST (842)
6192 - don't LBUG if a lock request times out after receiving AST (913)
6193 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
6194 - if a bad lock AST arrives, send an error instead of dropping entirely
6195 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
6196 - fix leak in bulk IO when only partially completed (899, 900, 926)
6197 - fix O_DIRECT for ia64 (55)
6198 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
6199 - C-z after timeout could hang a process forever; fixed (977)
6201 - client-side I/O cache (678, 924, 929, 941, 970)
6203 - READPAGE and SETATTRs which don't take server-side locks get
6206 2003-02-11 Phil Schwan <phil@clusterfs.com>
6209 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
6210 - Fix fsfilt_extN_readpage() to read a full page of directory
6211 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
6212 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
6213 - Fix imbalanced LOV object allocation and out-of-bound access (469)
6214 - Most intent operations were removed, in favour of a new RPC mode
6215 that does a single RPC to the server and bypasses most of the VFS
6216 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
6217 - Aggressively cancel local locks on DLM servers
6218 - mds_reint_unlink sends EA to the client if it's the last nlink.
6219 client uses that EA to unlink OST objects.
6220 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
6221 - recursive symlinks were fixed (439)
6222 - fixed NULL deref in DEBUG_REQ
6223 - filter_update_lastobjid no longer calls sync, which annoyed extN
6224 - fixed multi-client small-writes to a single file problem (445)
6225 - fixed mtime updates during file writes (607)
6226 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
6227 - fixed bug in obd_brw_read/write() (under guise of testing 367)
6228 - fixed Linux OST size reporting problem (444, 656)
6229 - OST now updates object mtime with writes or setattr (607, 619)
6230 - client verifies file size before zeroing page past EOF (445)
6231 - OST now writes last allocated objid to disk with allocation (108)
6232 - LOV on echo now works (409)
6234 - mds_reint_unlink sends a new buffer, with the EA included. this
6235 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
6236 set if a regular file was being unlinked, and it was the last link
6237 - use PtlGet from the target for bulk writes (315)
6238 - OST now updates object mtime with writes or setattr (607, 619)
6239 - LDLM now has a grant-time callback to revalidate locked items, if
6241 - Many MDS operations were reorganized to combat race conditions
6243 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
6244 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
6246 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
6249 - Fully reactivate OST imports after reconnection (512, others)
6250 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
6251 - More graceful error handling for truncating on dead OST (515)
6252 - Don't error out unless we're actually accessing dead stripes (474)
6253 - Fix garbage sizes when stripes are missing (410)
6254 - LRU counters were broken, causing constant lock purge (433, 432)
6255 - garbage on read from stripes with failed OSTs (441)
6256 - mark OSCs as active before reconnecting during recovery (438)
6257 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
6258 - lfind did not preserve OST order in output (443)
6259 - symlinks cause hung clients, incorrect data (439)
6260 - stop dereferencing request after dropping refcount (457)
6261 - don't LASSERT(spin_is_locked) on non-SMP (455)
6262 - fixes for many rename() bugs
6263 - fstat didn't correctly synchronize attributes (399)
6264 - server must handle lock cancellation during blocking AST prep (487)
6265 - bulk descriptors were free()d too soon (511)
6266 - fix paths in lconf, which would load incorrect modules (451, 507)
6267 - fix confusing lconf 'host not found' error message (386)
6268 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
6269 - fix race condition in mdc_blocking_ast() for inode access (526)
6270 - fix lov_unpackmd() unpacking wrong number of stripes (537)
6271 - fix lov_set_osc_active() marking wrong OSC inactive (440)
6272 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
6273 - fix multiple writes of stripe MD to MDS (358, maybe 519)
6274 - fix lstripe in several ways (kernel side) (527)
6275 - fix request leak in ldlm_cli_enqueue (262)
6276 - incorrect OSC was marked inactive after OST failure
6277 - call mds_fs_cleanup before unmounting filesystem (524)
6278 - fix races between taking ns_lock and ldlm_lock_change_resource
6279 - fix races updating LOV export open file list
6280 - fix lov_enqueue error path, avoid decref-ing bad lock handle
6281 - fix recovery NULL deref in ldlm_cli_cancel_unused
6282 - fix some DLM races by using new hash table for lock handles (419)
6283 - permit the client to specify desired inodes, at replay
6284 - duplicate requests when we queue them for replay reintegration
6285 - fix last_rcvd offset calculation
6286 - sync after each recovered transaction, so we always make progress
6287 - never, not always, ERESTART requests without transnos
6288 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
6290 - skip replay if the MDS says that the client is already connected
6291 - don't check for a recovery-enabled export to match lctl's UUID
6292 - don't INC_USE_COUNT for phantom exports
6293 - don't crash when cleaning up phantom exports (567)
6294 - don't double-finish or set replay data for errored mdc_open requests
6295 - abort requests when they time out, so we don't get old replies
6296 - send/receive replies for AST messages again
6297 - if the client says that it doesn't have the lock, cancel it on the
6299 - if we timeout during I/O, don't try to cancel an in-use lock; instead
6300 mark it as destroyed, it will all work out when decref is called
6301 - fix module use counts (22, 581)
6303 - ASTs now expect a reply (server cancels lock on error reply)
6305 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
6308 - fix many simultaneous client startup (392)
6309 - fix dentry->d_it clobbering
6310 - credentials weren't being shipped for readdir/getattr operations
6311 - remove invalid assertions triggered during some concurrent MD
6313 - proper Lustre versions added (336, 389)
6314 - fix memory leak for create error case (398)
6315 - fix LOV locking bug that would get cli/srv out of sync
6316 - fix echo client over LOV (409)
6317 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
6318 - fix double-O_EXCL intent crash (424)
6319 - avoid sending multiple lock CANCELs (352)
6321 - MDS can do multi-client recovery (modulo bugs in new code)
6323 - many updates, edits, cleanups
6325 2002-11-18 Phil Schwan <phil@clusterfs.com>
6328 - fix null d_it dereference (346)
6329 - fix full OST/dbench hang (333)
6330 - fix permission problem with file removal (286)
6331 - fix removal of OSCs from LOV when they fail
6332 - fix NULL deref during bulk timeout (214)
6333 - fix problems related to multiple filesystems on one MDS (241)
6334 - fixed serious subtle metadata locking bugs
6335 - free locks on clients when inodes are removed due to memory
6337 - fix inode pointer in lock data (285)
6338 - partial support for multiple MDS on a single host (241)
6339 - data locks weren't cancelled at clear_inode time (290, 311)
6340 - intent locks could lead to unbounded lock growth (205)
6341 - added a maximum lock count, an LRU list, and a flusher
6342 - fix multiple rename (365)
6343 - properly abstracted the echo client
6344 - OSC locked 1 byte too many; fixed
6345 - rewrote brw callback code:
6346 - fixed recovery bugs related to LOVs (306)
6347 - fixed too-many-pages-in-one-write crash (191)
6348 - fixed (again) crash in sync_io_timeout (214)
6349 - probably fixed callback-related race (385)
6351 - Add capability to MDS protocol
6352 - LDLM cancellations and callbacks on different portals
6354 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
6357 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
6358 - defer open object destruction to close time (601981, 138)
6359 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
6360 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
6361 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
6362 - don't call obd_setattr in ll_file_release for destroyed objects
6363 * protocol change to lustre_msg: move |version| and add |flags|
6364 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
6365 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
6366 * added replay of create, unlink, link and rename operations during
6367 MDS failover; recovery should be much more robust now
6368 * remove failed OSCs from LOVs (only lov_create uses this so far)
6369 * the lustre-HOWTO was brought (more) up to date (582544)
6371 2002-10-23 Phil Schwan <phil@clusterfs.com>
6374 - in-use dentries weren't being reused properly (617851)
6375 - prevent multiple LDLM setup (599178)
6376 - fix LOV size calculations for truncate (617853)
6377 - fix client handling of MDS intent errors (POSIX)
6378 - fix permission bug in lovstripe.c test (624321)
6379 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
6380 - truncate past end of file could corrupt data
6381 - proper cleanup after timeouts, crashes, etc (592524, 550815)
6382 - a race in recovery could return ETIMEDOUT to apps (623947)
6383 - building outside the source directory was fixed
6384 * the lustre-HOWTO was brought (more) up to date (582544)
6385 * major progress was made on recovery functionality
6387 2002-10-10 Phil Schwan <phil@clusterfs.com>
6390 - recovery deadlock fix
6391 - rm -rf causes LBUG fix (617817)
6392 - file open by multiple tasks fix (618962)
6393 - directory permissions bugs (602707 and 620007)
6394 - journal_stop fixed with locking (611313)
6395 - O_APPEND failures resolved (618273, perhaps 614459)
6396 - lconf PATH fix (619770)
6397 - IA64 build fix (621450)
6398 - RPC buffer sizes scale with amount of memory
6400 2002-10-01 Phil Schwan <phil@clusterfs.com>
6403 - locks would be cancelled without throwing away data pages,
6404 resulting in inconsistent data (605627)
6405 - inode attributes were not always being refreshed (605627, 612449)
6406 - lconf now continues to cleanup after lctl reports an error
6407 - MDS now enforces user permissions (602707)
6408 - lprocfs cleanup fixed, but not yet enabled (614157)
6409 - fixed infinite server hang, should a client not respond to an AST
6410 - avoid going into recovery if user calls readlink() with a buffer
6411 that's too small (613941)
6412 - AST RPCs no longer require replies (614867) -- this may be changed
6413 - don't crash server if client sends an IOV that's too big (611336)
6414 - fixed lock conversion deadlock (611892)
6415 - fixed the following of symlinks (614622)
6416 * recovery: the server can remove locks from a client that dies, other
6417 clients can make progress
6418 * more extN patch fixes
6419 * compile-time configurable ptlrpc buffer allocations
6421 - collaborative read cache document
6422 - Lustre Lite Performance CDR document-in-progress
6424 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6427 - fix typo in patch-2.4.18
6429 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6432 - clear ptlrpc request each time in handle_incoming_request()
6433 - unlink of files now destroys the object on the OST
6435 2002-09-19 Peter Braam <braam@clusterfs.com>
6437 * add hard link support
6438 * change obdfile creation method
6439 * kernel patch changed
6441 2002-09-19 Peter Braam <braam@clusterfs.com>
6444 - stack overflow bug in extN fixed
6446 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
6448 * documentation updates
6449 - add man pages for config tools
6450 - update tests/README to describe testing with new config tools
6451 - finish metadata API descriptions
6452 * bug fixes and cleanups
6453 - statfs workaround for 16TB limit
6454 - LOV stripe allocation improved, can stripe on subset of OSTs
6455 - LOV file size/IO offset was wrong for files > 4GB in size
6456 - object EA data was being dropped, caused files to be unreadable
6457 - memory overflow with non-LOV OST caused memory corruption
6458 - fixed regression tests to work with new config tools, obdfilter
6459 - fixed bug when directory size became larger than 1 block
6460 - fixed bug (for single client case) when PWD was deleted
6461 - invalidate local directory pages when doing intent-based ops
6462 - avoid LDLM oops when lock callback contained bad data
6464 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
6466 * documentation updates
6467 * bug fixes and cleanups
6468 - configuration tools
6471 - 64-bit compile warnings
6472 - 64-bit internal statfs data
6474 * test_brw on persistent OST devices
6476 * lprocfs (disabled)
6478 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
6480 * documentation updates
6481 * bug fixes and cleanups
6482 * configuration tools
6484 2002-08-30 Peter J. Braam <braam@clusterfs.com>
6487 * many small fixes to 0_5_4
6488 * io/network handling
6489 * thinkos in MDS operations
6491 2002-08-24 Peter J. Braam <braam@clusterfs.com>
6494 * crucial basic fixes to 0.5.3
6495 * IOR, Iozone work over Elan
6498 2002-08-07 Phil Schwan <phil@clusterfs.com>
6499 * version 0_5_3, our first alpha
6500 * we use the new Portals iovs
6501 * documentation updates
6502 * bug fixes and cleanups
6503 * small changes in the DLM wire protocol
6505 2002-07-25 Peter J. Braam <braam@clusterfs.com>
6506 * version 0_5_1 with some initial stability,
6507 * locking on MD and file I/O.
6508 * documentation updates
6509 * several bug fixes since 0.5.0
6510 * small changes in wire protocol
6512 2002-07-18 Phil Schwan <phil@clusterfs.com>
6514 * delivered as Lustre Light Alpha
6515 * fixed a crash after handling invalid MDS requests
6516 * fixed directory pages for architectures with non-4k pages sizes
6518 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
6519 * release version v0_4_4
6520 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
6522 * Fixes a number of interruption problems with OST operations.
6523 * Update documentation for portals header changes
6524 * Move all wire protocol structs/defines to lustre_idl.h
6525 * Fixes symlink length bug.
6526 * Add tcpdump to repository.
6528 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
6529 * release version v0_4_3
6530 * Fixes statfs for inodes on extN.
6531 * Fixes bug in runtests which would delete /etc/hosts.
6532 * Use 64-bit object IDs wherever possible (not into VFS though)
6533 Remove ost_get_info, which is unused by lustre, and out of date.
6535 2002-07-03 Peter Braam <braam@clusterfs.com>
6536 * release version v0_4_2 Fixes a lookup error (type not passed)
6537 * move forward to head of Portals
6538 * move forward to latest Lustre kernel
6540 2002-06-25 Peter Braam <braam@clusterfs.com>
6541 * release version v0_4_1. Hopefully stable on single node use.