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 3.10.0-327.13.1.el7 (RHEL7.2)
10 3.0.101-0.47.71 (SLES11 SP3)
11 3.0.101-71 (SLES11 SP4)
12 3.12.57-60.35 (SLES12 SP1)
13 vanilla linux 4.2.1 (ZFS only)
14 * Client known to build on unpatched kernels:
15 2.6.32-431.29.2.el6 (RHEL6.5)
16 2.6.32-504.30.3.el6 (RHEL6.6)
17 2.6.32-573.26.1.el6 (RHEL6.7)
18 3.10.0-327.13.1.el7 (RHEL7.2)
19 3.0.101-0.47.71 (SLES11 SP3)
20 3.0.101-68 (SLES11 SP4)
21 3.0.101-71 (SLES11 SP4)
22 3.12.57-60.35 (SLES12 SP1)
24 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
25 * Recommended ZFS / SPL version: 0.6.5.6
26 * Tested with ZFS / SPL version: 0.6.5.6
27 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
28 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
29 stack overflow. For more information, please refer to bugzilla 17630.
30 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
31 the re-exporting nodes to be mounted with "32bitapi" mount option
33 --------------------------------------------------------------------------------
35 02-29-2016 Intel Corporation
37 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
38 for currently supported client and server kernel versions.
39 * Server known to build on patched kernels:
40 2.6.32-431.29.2.el6 (RHEL6.5)
41 2.6.32-504.30.3.el6 (RHEL6.6)
42 2.6.32-573.12.1.el6 (RHEL6.7)
43 3.10.0-327.3.1.el7 (RHEL7.2)
44 3.0.101-0.47.71 (SLES11 SP3)
45 3.0.101-68 (SLES11 SP4)
46 vanilla linux 4.2.1 (ZFS only)
47 * Client known to build on unpatched kernels:
48 2.6.32-431.29.2.el6 (RHEL6.5)
49 2.6.32-504.30.3.el6 (RHEL6.6)
50 2.6.32-573.12.1.el6 (RHEL6.7)
51 3.10.0-327.3.1.el7 (RHEL7.2)
52 3.0.101-0.47.71 (SLES11 SP3)
53 3.0.101-68 (SLES11 SP4)
56 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
57 * Recommended ZFS / SPL version: 0.6.4.2
58 * Tested with ZFS / SPL version: 0.6.4.2
59 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
60 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
61 stack overflow. For more information, please refer to bugzilla 17630.
62 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
63 the re-exporting nodes to be mounted with "32bitapi" mount option
65 --------------------------------------------------------------------------------
67 03-10-2015 Intel Corporation
69 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
70 for currently supported client and server kernel versions.
71 * Server known to build on patched kernels:
72 2.6.32-431.29.2.el6 (RHEL6.5)
73 2.6.32-504.8.1.el6 (RHEL6.6)
74 3.0.101-0.46 (SLES11 SP3)
75 * Client known to build on unpatched kernels:
76 2.6.32-431.29.2.el6 (RHEL6.5)
77 2.6.32-504.8.1.el6 (RHEL6.6)
78 3.10.0-123.20.1.el7 (RHEL7)
79 3.0.101-0.46 (SLES11 SP3)
80 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
81 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
82 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
83 stack overflow. For more information, please refer to bugzilla 17630.
84 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
85 the re-exporting nodes to be mounted with "32bitapi" mount option
87 Severity : enhancement
89 Description: control OST-index in IDIF via ROCOMPAT flag.
90 Details : Introduce new flag OBD_ROCOMPAT_IDX_IN_IDIF that is stored in the
91 last_rcvd file. For new formatted OST device, it will be auto set;
92 for the case of upgrading from old OST device, you can enable it
93 via the lproc interface osd-ldiskfs.index_in_idif. With such flag
94 enabled, for new created OST-object, its IDIF-in-LMA will contain
95 the OST-index; for the existing OST-object, the OSD will convert
96 old format IDIF as new format IDIF with OST-index stored in the
97 LMA EA when accessing such OST-object or via OI scrub. Once such
98 flag is enabled, it cannot be reverted back, so the system cannot
99 be downgraded to the orignal incompatible version.
101 --------------------------------------------------------------------------------
103 07-30-2014 Intel Corporation
105 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
106 for currently supported client and server kernel versions.
107 * Server known to build on patched kernels:
108 2.6.32-431.20.3.el6 (RHEL6.5)
109 3.0.101-0.31 (SLES11 SP3)
110 * Client known to build on unpatched kernels:
111 2.6.32-431.20.3.el6 (RHEL6.5)
112 3.0.101-0.31 (SLES11 SP3)
115 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
116 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
117 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
118 stack overflow. For more information, please refer to bugzilla 17630.
119 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
120 the re-exporting nodes to be mounted with "32bitapi" mount option
122 --------------------------------------------------------------------------------
124 10-31-2013 Intel Corporation
126 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
127 * Server builds for kernels:
128 2.6.32-358.18.1.el6 (RHEL6.4)
129 * Client builds for unpatched kernels:
130 2.6.32-358.18.1.el6 (RHEL6.4)
131 3.0.92-0.8 (SLES11 SP3)
133 * Recommended e2fsprogs version: 1.42.7.wc2 or newer
134 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
135 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
136 stack overflow. For more information, please refer to bugzilla 17630.
137 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
138 the re-exporting nodes to be mounted with "32bitapi" mount option
140 --------------------------------------------------------------------------------
142 05-30-2013 Intel Corporation
144 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
145 * Server builds for kernels:
146 2.6.32-358.6.2.el6 (RHEL6.4)
147 * Client builds for unpatched kernels:
148 2.6.32-358.6.2.el6 (RHEL6.4)
149 3.0.74-0.6.6 (SLES11 SP2)
151 * Recommended e2fsprogs version: 1.42.7.wc1 or newer
152 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
153 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
154 stack overflow. For more information, please refer to bugzilla 17630.
155 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
156 the re-exporting nodes to be mounted with "32bitapi" mount option
158 --------------------------------------------------------------------------------
160 9-30-2012 Whamcloud, Inc.
162 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
163 * Server support for kernels:
164 2.6.32-279.5.1.el6 (RHEL6)
165 * Client support for unpatched kernels:
166 2.6.32-279.5.1.el6 (RHEL6)
168 2.6.18-238.19.1.el5 (RHEL5)
169 2.6.16.60-0.69.1 (SLES10)
170 * Recommended e2fsprogs version: 1.41.90.wc4
171 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
172 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
173 stack overflow. For more information, please refer to bugzilla 17630.
174 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
175 the re-exporting nodes to be mounted with "32bitapi" mount option
177 Severity : enhancement
179 Description: Update RHEL6.3 kernel to 2.6.32-279.5.1.el6.
181 Severity : enhancement
183 Description: Update RHEL6.3 kernel to 2.6.32-279.2.1.el6.
185 Severity : enhancement
187 Description: Update RHEL6.2 kernel to 2.6.32-220.17.1.el6
189 Severity : enhancement
191 Description: Update RHEL6.2 kernel to 2.6.32-220.13.1.el6
193 Severity : enhancement
195 Description: Update RHEL6.2 kernel to 2.6.32-220.7.1.el6
197 --------------------------------------------------------------------------------
199 04-30-2012 Whamcloud, Inc.
201 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
202 * Server support for kernels:
203 2.6.32-220.4.2.el6 (RHEL6)
204 * Client support for unpatched kernels:
205 2.6.32-220.4.2.el6 (RHEL6)
207 2.6.18-238.19.1.el5 (RHEL5)
208 2.6.16.60-0.69.1 (SLES10)
209 * Recommended e2fsprogs version: 1.41.90.wc4
210 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
211 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
212 stack overflow. For more information, please refer to bugzilla 17630.
213 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
214 the re-exporting nodes to be mounted with "32bitapi" mount option
216 Severity : enhancement
218 Description: Update RHEL6.2 kernel to 2.6.32-220.4.2.el6
220 Severity : enhancement
222 Description: Update RHEL6.2 kernel to 2.6.32-220.4.1.el6
224 Severity : enhancement
226 Description: Update RHEL6.2 kernel to 2.6.32-220.el6
228 Severity : enhancement
230 Description: Update RHEL6.1 kernel to 2.6.32-131.17.1.el6
232 --------------------------------------------------------------------------------
234 09-30-2011 Whamcloud, Inc.
236 * Server support for kernels:
237 2.6.18-238.19.1.el5 (RHEL5)
238 2.6.32-131.6.1.el6 (RHEL6)
239 * Client support for unpatched kernels:
240 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
241 2.6.18-238.19.1.el5 (RHEL5)
242 2.6.32-131.6.1.el6 (RHEL6)
243 2.6.29.4-167.fc11 (FC11)
244 2.6.16.60-0.69.1 (SLES10)
246 * Recommended e2fsprogs version: 1.41.90.wc3
247 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
248 Lustre filesystem with 4K stack may cause a stack overflow. For more
249 information, please refer to bugzilla 17630.
250 * NFSv4 reexport requires Lustre client on the re-exporting nodes to be
251 mounted with "32bitapi" mount option
253 Severity : enhancement
254 Description: Update RHEL6.1 kernel to 2.6.32-131.6.1.el6
256 Severity : enhancement
257 Description: Update RHEL5.6 kernel to 2.6.18-238.19.1.el5
259 Severity : enhancement
260 Description: Update RHEL5.6 kernel to 2.6.18-238.9.1.el5
263 Frequency : only with SLES10
265 Description: Use OFED "KMP" provided by Novell
266 Details : SLES10 SP3 ships with OFED in a separate "KMP" package.
267 Lustre is now built against this package. That means you need to
268 install the ofed-kmp package from Novell for the patchless client.
269 Note that the ofed-kmp that Novell ships may not exactly match the
270 kernel version but should still be compatible.
272 Severity : enhancement
274 Description: Update RHEL5.5 kernel to 2.6.18-194.17.1.el5
275 Update OEL5.5 kernel to 2.6.18-194.17.1.0.1.el5.
277 --------------------------------------------------------------------------------
279 2010-07-15 Oracle, Inc.
281 * Server support for kernels:
282 2.6.18-164.11.1.el5 (RHEL 5)
283 2.6.18-164.11.1.0.1.el5 (OEL 5)
284 * Client support for unpatched kernels:
285 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
286 2.6.18-164.11.1.el5 (RHEL 5),
287 2.6.18-164.11.1.0.1.el5 (OEL 5)
288 2.6.16.60-0.42.8 (SLES 10),
290 2.6.29.4-167.fc11 (FC11)
291 * Recommended e2fsprogs version: 1.41.10-sun2
293 Severity : enhancement
295 Description: Distribute lfs_migrate script with lustre
296 Details : The lfs_migrate mentioned in the manual is now included in the
297 lustre tree in lustre/scripts and will distributed in lustre RPMs.
299 Severity : enhancement
301 Description: Failover nid list added to import proc file.
303 Severity : enhancement
305 Description: "som_preview" mount option required to use SOM
306 Details : The clients will need to use som_preview mount option
310 Description: Lustre does not do 1MB IOs to HW RAID
311 Details : Bump MAX_PHYS/HW_SEGMENTS and SG_ALL to 256 in the RHEL5 kernel.
313 Severity : enhancement
315 Description: lfs quota output is not very convenient for awk/sed-parsing
316 Details : Some positions in lfs quota output table could be empty or
317 non-empty which made it hard to parse it with scripts, now
318 a dash is put instead of space where there is not supposed
323 Description: Update to OFED 1.5.1
324 Details : OFED 1.5.1 will be used in environments that don't use
325 vendor supplied OFED.
329 Description: Remove lnet. prefix from lctl params display
330 Details : Also renamed lnet.memused to lnet_memused to
335 Description: Use conf_param -d to remove permanant settings.
339 Description: Add list_param -R
340 Details : lctl list_param now supports recursive mode.
343 Bugzilla : 21829, 22312
344 Description: Fix llobdstat.
345 Details : 1. Now updating the write_bytes counter
346 2. Port script fixes from 1.8.
350 Description: decrease the usage of memory on clients.
351 Details : 1. On clients, recycle dentries and inodes unused.
352 2. Delete the code related to ll_deathrow(att 6215 in bug 1443). It
357 Description: llobdstat fix and enhancement
358 Details : add a counter to set a limit to how many samples will be returned
359 fix a wildcard in the path to limit to obdfilter stats only
363 Description: lfs check only allowed for root
364 Details : Code cleanups and added a regresion test of non-root use of
369 Description: MDS Oopses in ptlrpc_dispatch_difficult_reply()/__list_add()
370 Details : Fix hr->hr_index overload.
374 Description: Fix FIEMAP functionality.
378 Description: FC11 patchless client support added
382 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and
383 OEL5.4 kernel to 2.6.18-164.11.1.0.1.el5.
387 Description: New acc-sm script and various test-framework enhancments.
391 Description: Previously landed patch was reverted.
395 Description: Update SLES10 kernel to 2.6.16.60-0.42.8 (patchless)
399 Description: Add ldiskfs maxdirsize mount option.
403 Description: Fix mkfs.lustre support for 16TB LUNs.
407 Description: Print mmp_check_interval
408 Details : print mmp_check_interval and make it possible to abort
409 mount operation in case it takes too long.
413 Description: Add cascading_rw to lustre mpi tests.
417 Description: Store SOM data in EA.
418 Details : Also several SOM related cleanups.
422 Description: Add refcount to md_open_data
426 Description: Use refcount on lov_request_set to prevent memory corruption.
430 Description: Statahead: use parent dir's i_mutex for synchronization.
434 Description: loadgen improvements.
438 Description: Enable peer health detection in ksocklnd and ko2iblnd.
442 Description: Control DCACHE_LUSTRE_INVALID flag with MDS_INODE_LOOKCUP lock
446 Description: Fix race on obd_devcie:md_stats
450 Description: Fix SLES 11 client builds.
451 Details : Resolved uninitalized pointer gcc warnings.
455 Description: Allow non-root access for "lfs check".
456 Details : Added a check in obd_class_ioctl() for OBD_IOC_PING_TARGET.
460 Description: Clear imp_force_reconnect correctly in ptlrpc_connect_interpret()
462 Severity : enhancement
464 Description: Update kernel to RHEL5.4 2.6.18-164.6.1.el5 and
465 OEL5 2.6.18-164.6.1.0.1.el5(Both in-kernel OFED enabled).
469 Description: Sanity test_27p fails.
470 Details : OBD_FAIL_OST_ENOSPC has been moved to filter_statfs().
471 Sanity tests 27 were cleaned up and improved.
475 Description: Unify req format on client/servers
476 Details : Use new req_capsule API [almost] everywhere instead of old PTLRPC
477 buffers and swabbers approach..
482 Description: drop join file stuff.
484 Severity : enhancement
486 Description: Add 2.6.27 and SLES11 (patchless client) support
490 Description: Fix in ptlrpc_expire_one_request() to print the signed time
493 Severity : enhancement
495 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
498 Severity : enhancement
500 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
502 Severity : enhancement
504 Description: correctly handle big reply message.
505 Details : send LNet event if reply is bigger than buffer and adjust this buffer
510 Description: Drop unnecessary __GFP_NOMEMALLOC flag from filter_get_page()
512 Severity : enhancement
514 Description: Add support for OEL5.
518 Description: Conf-sanity.sh 50g test - deactivated OST should not cause a panic.
520 Severity : enhancement
522 Description: "weak-modules" support
523 Details : Implement "weak-modules" support which enables kernel modules
524 to be used with any kernel that implements the same kABI. In
525 order to achieve this modules are now installed in
526 /lib/modules/$(uname -r)/updates/kernel on all distributions.
530 Description: client could not reconnect to OST because of an active request.
531 Details : The client now retries to reconnect to the same server, if a
532 connect request failed with EBUSY or -EAGAIN. Test case 35b
533 conf-sanity.sh has been created.
537 Description: Don't allow make backward step on assignin osc next id.
538 Details : race between allocation next id and ll_sync thread can be cause
539 of set wrong osc next id and can be kill valid ost objects.
543 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
547 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
551 Description: Error handling in osc_statfs_interpret() has been improved.
552 Details : Check in osc_statfs_interpret() for EBADR.
556 Description: Do not send statfs() requests to OSTs disabled by administrator.
557 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
559 Severity : enhancement
561 Description: Added sanity-quota test to for lfs(1) to work with numeric uid/gid.
563 Severity : enhancement
565 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
567 Severity : enhancement
569 Description: Enhancement for lfs(1) command to use numeric uid/gid.
573 Description: Improved cleanup procedure for conf-sanity test_46a.
577 Description: Do not start bulk transfer, if client tried to reconnect.
578 Details : Do not start bulk transfer, if client tried to reconnect,
579 but failed due to active rpcs. Check for exp_abort_active_req
580 has been implemented.
584 Description: Prevent inconsistences between linux and lustre mount structures.
585 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
586 Make the sleep interruptible.
588 Severity : enhancement
590 Description: provide server to client comms path
591 Details : server to client communications path via new LDLM_SET_INFO rpc
595 Description: truncate starts GFP_FS allocation under transaction causing deadlock
596 Details : ldiskfs_truncate calls grab_cache_page which may start page
597 allocation under an open transaction. This may lead to
598 calling prune_icache with consequent lustre reentrance.
602 Description: send correctly lsm on open replay
603 Details : MDS is trust to LSM size on replay open, but client can set wrong size
608 Description: enable client interface failover
609 Details : When a child reconnects from another NID, properly update export
610 nid hash position and ldlm reverse import.
612 Severity : enhancement
614 Description: Communicate OST degraded/readonly state via statfs to MDS
615 Details : Flags in the statfs returned from OSTs indicate whether the
616 OST is in a degraded RAID state, or if the filesystem has
617 turned read-only after a filesystem error is detected.
622 Description: don't panic if EPROTO was hit when reading symlink
623 Details : correctly handling request reference in error cases.
625 Severity : enhancement
627 Description: Add LustreNetLink, a kernel-userspace communcation path.
629 Severity : enhancement
631 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
636 Description: access to llog context before init.
637 Details : move handling CATALOGS file at osc layer and forbid access to llog
641 Frequency : rare, only when client is writing to OST on local node
643 Description: Avoid deadlock for local client writes
644 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
645 memory freeing context. This allows OST threads to set the
646 PF_MEMALLOC flag on task structures in order to allocate memory
647 from reserved pools and complete IO.
648 Use GFP_HIGHUSER for OST allocations for non-local client writes,
649 so that the OST threads generate memory pressure and allow
650 inactive pages to be reclaimed.
652 Severity : enhancement
654 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
659 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
660 Details : move ldlm namespace creation in setup phase to avoid grab
661 _lprocfs_lock with cli_sem held.
665 Description: Temporarily disable grant shrink.
666 Details : Disable the feature for debugging.
669 Frequency : only when formatting test filesystems
671 Description: Unable to run several mkfs.lustre on loop devices at the same time
672 Details : mkfs.lustre returns error 256 on the concurrent loop devices
673 formatting. The solution is to proper handle the error.
675 Severity : enhancement
677 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
679 Severity : enhancement
681 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
686 Description: resolve race between obd_disconnect and class_disconnect_exports
687 Details : if obd_disconnect will be called to already disconnected export he
688 forget release one reference and osc module can't unloaded.
690 Severity : enhancement
692 Description: Allow tuning service thread via /proc
693 Details : For each service a new
694 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
695 created that can be used to set min/max thread counts, and get the
696 current number of running threads.
701 Descriptoin: don't return error if have partially created objects for file.
702 Details : lov_update_create_set uses set->set_success as index for created
703 objects, so if some requests will be failed, they will have hole
704 at end of array and qos_shrink_lsm for allocate correct lsm.
706 Severity : enhancement
708 Description: Update OFED support to 1.4.1
713 Description: Reduce small size read RPC
714 Details : Set read-ahead limit for every file and only do read-ahead when
715 available read-ahead pages are bigger than 1M to avoid small size
718 Severity : enhancement
720 Description: move AT tunable parameters for more consistent usage
721 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
723 Severity : enhancement
725 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
726 Details : allow skip disconnected ost for send statfs request and hide error
732 Description: fix racy locking of mballoc block bitmaps causing BUG
733 Details : The locking of the mballoc buddy bitmap and the in-memory
734 block bitmap was using two different spin locks in some
735 cases. This made it possible to incorrectly access the
736 mballoc bitmap while another process was modifying it,
737 causing a sanity assertion to fail. While no on-disk corruption
738 was reported, there was some risk of this happening.
741 Frequency : rare, on llog test 6
743 Descriptoin: don't allow connect to already connected import
744 Details : allowing connect to already connected import is hide connecting problem.
749 Descriptoin: allow kill process which wait statahead result
750 Details : for some reasons 'ls' can stick in waiting result from statahead,
751 in this case need way for kill this process.
753 Severity : enhancement
755 Description: Add state history info file, enhance import info file
756 Details : Track import connection state changes in a new osc/mdc proc file;
757 add overview-type data to the osc/mdc import proc file.
759 Severity : enhancement
761 Description: MDS create should not wait for statfs RPC while holding DLM lock.
763 Severity : enhancement
765 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
770 Descriptoin: OST grant too much space to client even there are not enough space.
771 Details : Client will shrink its grant cache to OST if there are no write
772 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
773 this grant cache if there are already not enough avaible space
774 (left_space < total_clients * 32M).
777 Frequency : start MDS on uncleanly shutdowned MDS device
779 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
780 Details : stay in waiting mds<>ost recovery finished produce random bugs
781 due race between two ll_sync thread for one lov target. send
782 ACTIVATE event only if connect realy finished and import have
786 Frequency : rare, connect and disconnect target at same time
788 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
789 Details : don't call obd_disconnect under lov_lock. this long time
790 operation and can block ptlrpcd which answer to connect request.
795 Descriptoin: don't lose wakeup for imp_recovery_waitq
796 Details : recover_import_no_retry or invalidate_import and import_close can
797 both sleep on imp_recovery_waitq, but we was send only one wakeup
801 Frequency : always with long access acl
803 Descriptoin: mds can't pack reply with long acl.
804 Details : mds don't control size of acl but they limited by reint/getattr
807 Severity : enhancement
809 Description: Update to SLES10 kernel-2.6.16.60-0.33.
811 Severity : enhancement
813 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
816 Frequency : start MDS on uncleanly shutdowned MDS device
818 Descriptoin: aborting recovery hang on MDS
819 Details : don't throttle destroy RPCs for the MDT.
822 Frequency : on remount
824 Description: external journal device not working after the remount
825 Details : clear dev_rdonly flag for external journal devices in
831 Description: shutdown vs evict race
832 Details : client_disconnect_export vs connect request race.
833 if client will evicted at this time - we start invalidate
834 thread without referece to import and import can be freed
838 Frequency : rare, need acl's on inode.
840 Description: client can't handle ost additional correctly
841 Details : if ost was added after client connected to mds client can have
842 hit lnet_try_match_md ... to big messages to wide striped files.
843 in this case need teach client to handle config events about add
844 lov target and update client max ea size at that event.
846 Severity : enhancement
848 Description: Changelogs
849 Details : Changelogs are a lightweight mechanism to track filesystem
850 metadata and namespace changes. The changelog is recorded
851 permanently on the MDTs, and is periodically "consumed" / purged
852 when records are no longer needed.
854 Severity : enhancement
856 Description: compact fld format with extents
857 Details : Store range of seq rather than every seq in FLD. Seq
858 controller update FLD rather than clients. In Case of CMD, mdt0
859 has FLD, all other metadata server act as non persistent proxy
860 for FLD queries and cache fld entries in fld cache.
865 Description: don't skip ost target if they assigned to file
866 Details : Drop slow OSCs if we can, but not for requested start idx.
867 This means "if OSC is slow and it is not the requested
868 start OST, then it can be skipped, otherwise skip it only
869 if it is inactive/recovering/out-of-space.
873 Description: more cleanup in mds_lov
874 Details : not send LOV EA under replay, we can't know about they size at this
875 time. Don't allow client connect to mds before any ost connected,
876 for avoid problems with LOV EA size and returning EIO to client.
878 Severity : enhancement
880 Description: Interoperability at server side (Disk interoperability)
882 Severity : enhancement
884 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
886 Severity : enhancement
888 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
890 Severity : enhancement
892 Description: New client IO stack (CLIO).
894 Severity : enhancement
896 Description: Commit on sharing. Eliminate inter-client dependencies between
897 uncommitted transactions by doing transaction commits.
898 Thereby clients may recovery independently.
901 Frequency : Create a symlink file with a very long name
903 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
904 Details : If there is no extra space in the request for early cancels,
905 ldlm_req_handles_avail() returns 0 instead of a negative value.
907 Severity : enhancement
909 Description: Add /proc entry for import status
910 Details : The mdc, osc, and mgc import directories now have
911 an import directory that contains useful import data for debugging
914 Severity : enhancement
916 Description: Re-disable certain /proc logging
917 Details : Enable and disable client's offset_stats, extents_stats and
918 extents_stats_per_process stats logging on the fly.
921 Frequency : Only on FC kernels 2.6.22+
923 Description: oops in statahead
924 Details : Do not drop reference count for the dentry from VFS when lookup,
925 VFS will do that by itself.
927 Severity : enhancement
929 Description: Generic /proc file permissions
930 Details : Set /Proc file permissions in a more generic way to enable non-
931 root users operate on some /proc files.
935 Description: Hitting mdc_commit_close() ASSERTION
936 Details : Properly handle request reference release in
937 ll_release_openhandle().
941 Description: quota recovery deadlock during mds failover
942 Details : This patch includes att18982, att18236, att18237 in bz14840.
944 1. fix osts hang when mds does failover with quotaon
945 2. prevent watchdog storm when osts threads wait for the
950 Frequency : only patchless client
951 Description: add workaround for race between add/remove dentry from hash
953 Severity : enhancement
955 Description: Allow OST glimpses to return PW locks
959 Description: LBUG when llog conf file is full
960 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
964 Description: Prevent import from entering FULL state when server in recovery
968 Description: service mount cannot take device name with ":"
969 Details : Only when device name contains ":/" will mount treat it as
975 Description: replace ptlrpcd with the statahead thread to interpret the async
976 statahead RPC callback
980 Frequency : on recovery
981 Description: I/O failures after umount during fail back
982 Details : if client reconnected to restarted server we need join to recovery
983 instead of find server handler is changed and process self
984 eviction with cancel all locks.
986 Severity : enhancement
988 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
990 Severity : enhancement
992 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
994 Severity : enhancement
996 Description: Upcall on Lustre log has been dumped
997 Details : Allow for a user mode script to be called once a Lustre log has
998 been dumped. It passes the filename of the dumped log to the
999 script, the location of the script can be specified via
1000 /proc/sys/lnet/debug_log_upcall.
1005 Description: avoid idr_remove called for id which is not allocated.
1006 Details : Move assigment s_dev for clustered nfs to end of initialization,
1007 to avoid problem with error handling.
1012 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
1013 Details : When connection is reused this not moved from CONN_UNUSED_HASH
1014 into CONN_USED_HASH and this prodice warning when put connection
1015 again in unused hash.
1018 Severity : enhancement
1020 Description: File striping can now be set to use an arbitrary pool of OSTs.
1022 Severity : enhancement
1024 Description: Export bytes_read/bytes_write count on OSC/OST.
1028 Description: Early reply size mismatch, MGC loses connection
1029 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
1030 the connect flags are properly negotiated.
1035 Description: quotas are not honored with O_DIRECT
1036 Details : all writes with the flag O_DIRECT will use grants which leads to
1037 this problem. Now using OBD_BRW_SYNC to guard this.
1041 Description: add quota statistics
1042 Details : 1. sort out quota proc entries and proc code.
1043 2. add quota statistics
1045 Severity : enhancement
1047 Description: enable quota support for HEAD.
1051 Description: Properly propagate oinfo flags from lov to osc for statfs
1052 Details : restore missing copy oi_flags to lov requests.
1054 Severity : enhancement
1056 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
1057 : l_getgroups(8), lst(8), routerstat(8)
1058 Details : included man pages for llobdstat(8), llstat(8),
1059 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
1061 Severity : enhancement
1063 Description: configure's --enable-quota should check the
1064 : kernel .config for CONFIG_QUOTA
1065 Details : configure is terminated if --enable-quota is passed but
1066 : no quota support is in kernel
1070 Description: Remove portals compatibility
1071 Details : Remove portals compatibility, not interoperable with releases
1076 Description: Resolve device initialization race
1077 Details : Prevent proc handler from accessing devices added to the
1078 obd_devs array but yet be intialized.
1080 Severity : enhancement
1082 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
1084 Severity : enhancement
1086 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
1091 Description: Using wrong pointer in osc_brw_prep_request
1092 Details : Access to array[-1] can produce panic if kernel compiled with
1093 CONFIG_PAGE_ALLOC enabled
1097 Description: Client runs out of low memory
1098 Details : Consider only lowmem when counting initial number of llap pages
1102 Description: *optional* service tags registration
1103 Details : if the "service tags" package is installed on a Lustre node
1104 When the filesystem is mounted, a local-node service tag will
1105 be created. See http://inventory.sun.com/ for more information
1106 about the Service Tags asset management system.
1110 Description: Kernel BUG tries to release flock
1111 Details : Lustre does not destroy flock lock before last reference goes
1112 away. So always drop flock locks when client is evicted and
1113 perform unlock regardless of successfulness of speaking to MDS.
1117 Description: add refcount protection for osc callbacks, avoid panic on shutdown
1121 Description: sanity test 65a fails if stripecount of -1 is set
1122 Details : handle -1 striping on filesystem in ll_dirstripe_verify
1127 Description: ASSERTION(CheckWriteback(page,cmd)) failed
1128 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
1131 Severity : enhancement
1133 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
1137 Description: do not process already freed flock
1138 Details : flock can possibly be freed by another thread before it reaches
1139 to ldlm_flock_completion_ast.
1143 Description: LBUG during stress test
1144 Details : Need properly lock accesses the flock deadlock detection list.
1148 Description: oops in page fault handler
1149 Details : kernel page fault handler can return two special 'pages' in error
1150 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
1154 Description: timeout with invalidate import.
1155 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
1156 should be handled by ptlrpcd. This produce long age waiting and
1157 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
1159 Severity : enhancement
1161 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
1165 Description: enable MGS and MDT services start separately
1166 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
1167 a co-located MGS without starting the MGS, which is a complement
1168 to 'nosvc' mount option.
1172 Frequency : after recovery
1173 Description: precreate to many object's after del orphan.
1174 Details : del orphan st in oscc last_id == next_id and this triger growing
1175 count of precreated objects. Set flag LOW to skip increase count
1176 of precreated objects.
1180 Frequency : rare, on clear nid stats
1181 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
1182 Details : when clean nid stats sometimes try destroy live entry,
1183 and this produce panic in free.
1187 Description: Stack overflow during MDS log replay
1188 ease stack pressure by using a thread dealing llog_process.
1192 Description: wait until IO finished before start new when do lock cancel.
1193 Details : VM protocol want old IO finished before start new, in this case
1194 need wait until PG_writeback is cleared until check dirty flag and
1195 call writepages in lock cancel callback.
1197 Severity : enhancement
1199 Description: using special macro for print time and cleanup in includes.
1203 Description: mds_mfd_close() ASSERTION(rc == 0)
1204 Details : In mds_mfd_close(), we need protect inode's writecount change
1205 within its orphan write semaphore to prevent possible races.
1209 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
1213 Frequency : rare, on shutdown ost
1214 Description: don't hit live lock with umount ost.
1215 Details : shrink_dcache_parent can be in long loop with destroy dentries,
1216 use shrink_dcache_sb instead.
1220 Description: don't panic with use echo client
1221 Details : echo client pass NULL as client nid pointer and this produce null
1222 pointer dereference.
1226 Description: fix build on ppc32
1227 Details : compile code with -m64 flag produce wrong object file for ppc32.
1231 Description: add message levels for liblustreapi
1235 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
1236 Details : Move the 'good_osts' check before the 'total_bavail' check. This
1237 will result in an -EAGAIN and in the exit call path we call
1238 alloc_rr() which will with increasing aggressiveness attempt to
1239 aquire precreated objects on the minimum number of required OSCs.
1243 Description: Use old size assignment to avoid deadlock
1244 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
1245 the scheduling while holding a spinlock. We do not need locking
1246 for size in ll_update_inode() because size is only updated from
1247 the MDS for directories or files without objects, so there is no
1248 other place to do the update, and concurrent access to such inodes
1249 are protected by the inode lock.
1253 Description: resolve "_IOWR redefined" build error on SLES10
1257 Description: dump the memory debugging after all modules are unloaded to
1258 suppress false negative in conf_sanity test 39
1260 Severity : enhancement
1262 Description: build kernel-ib packages for OFED 1.3 in our release cycle
1267 Description: fix SLES kernel versioning
1268 Details : the kernel version for our SLES 10 kernel did not include a "-"
1269 before the "smp" at the end. while this was not a problem in
1270 general, it did mean that software trying to use the kernel
1271 version to try to detect a vendor specific kernel would fail.
1272 this was most evident by the OFED build scripts.
1276 Description: Don't update lov_desc members until making sure they are valid
1277 Details : When updating lov_desc members via proc fs, need fix their
1278 validities before doing the real update.
1282 Description: don't put request into delay list while invalidate in flight.
1283 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
1284 import in flight. this produce timeout for invalidate and sometimes
1285 can cause stale data.
1289 Frequency : on ppc only
1290 Description: not convert ost objects for directory because it's not exist.
1291 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
1295 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
1297 Severity : enhancement
1299 Description: Update to support 2.6.22.14 vanilla kernel.
1303 Frequency : rare, on recovery
1304 Description: read procfs can produce deadlock in some situation
1305 Details : Holding lprocfs lock which send rpc can produce block for destroy
1306 obd objects and this also block reconnect with -EALREADY. This
1307 isn't fix all lprocfs bugs - but make it rare.
1309 Severity : enhancement
1311 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
1314 Frequency : frequent on X2 node
1316 Description: mdc_set_open_replay_data LBUG
1317 Details : Set replay data for requests that are eligible for replay.
1321 Description: lustre_mgs: operation 101 on unconnected MGS
1322 Details : When MGC is disconnected from MGS long enough, MGS will evict the
1323 MGC, and late on MGC cannot successfully connect to MGS and a lot
1324 of the error messages complaining that MGS is not connected.
1328 Frequency : on network error
1329 Description: panic with double free request if network error
1330 Details : mdc_finish_enqueue is finish request if any network error ocuring,
1331 but it's true only for synchronus enqueue, for async enqueue
1332 (via ptlrpcd) this incorrect and ptlrpcd want finish request
1335 Severity : enhancement
1337 Description: client-side metadata stat-ahead during readdir(directory readahead)
1338 Details : perform client-side metadata stat-ahead when the client detects
1339 readdir and sequential stat of dir entries therein
1342 Frequency : on start mds
1344 Description: Implement get_info(last_id) in obdfilter.
1347 Frequency : occasional
1349 Description: Correctly check stale fid, not start epoch if ost not support SOM
1350 Details : open with flag O_CREATE need set old fid in op_fid3 because
1351 op_fid2 was overwritten with new generated fid, but mds can answer
1352 with one of these two fids and both is not stale. Setattr
1353 incorrectly started an epoch and assume will be called
1354 done_writing, but without SOM done_writing ever being called.
1357 Frequency : rare, depends on device drivers and load
1359 Description: MDS or OSS nodes crash due to stack overflow
1360 Details : Code changes in 1.8.0 increased the stack usage of some functions.
1361 In some cases, in conjunction with device drivers that use a lot
1362 of stack the MDS (or possibly OSS) service threads could overflow
1363 the stack. One change which was identified to consume additional
1364 stack has been reworked to avoid the extra stack usage.
1367 Frequency : occasional
1369 Description: Do not fail import if osc_interpret_create gets -EAGAIN
1370 Details : If osc_interpret_create got -EAGAIN it immediately exits and
1371 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
1372 oscc_has_objects and see OSC has no objests and call
1373 oscc_internal_create to resend create request.
1375 Severity : enhancement
1377 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
1379 Severity : enhancement
1381 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
1384 Frequency : very rare
1386 Description: Fix replay if there is an un-replied request and open
1387 Details : In some cases, older replay request will revert the
1388 mcd->mcd_last_xid on MDS which is used to record the client's
1389 latest sent request.
1391 Severity : enhancement
1393 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
1395 Severity : enhancement
1397 Description: Add rhel5 support to HEAD.
1399 Serverity : enhancement
1401 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
1406 Description: Don't allow skipping OSTs if index has been specified.
1407 Details : Don't allow skipping OSTs if index has been specified, make
1408 locking in internal create lots better.
1412 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
1413 Details : access to bitfield in structure is always rounded to long
1414 and this produce problem with not atomic change any bit.
1418 Description: Lustre make rpms failed.
1419 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
1420 builds Lustre rpms from tarball.
1423 Frequency : rare on shutdown ost
1425 Description: If llog cancel was not send before clean_exports phase, this can
1426 produce deadlock in llog code.
1427 Details : If llog thread has last reference to obd and call class_import_put
1428 this produce deadlock because llog_cleanup_commit_master wait when
1429 last llog_commit_thread exited, but this never success because was
1430 called from llog_commit_thread.
1434 Description: allow userland application know is lost one of stripes.
1435 Details : fill lvb_blocks with error code on ost and return it to
1436 application if error flag found.
1440 Description: NULL lov_tgts causing MDS oops
1441 Details : more safe checks for NULL lov_tgts for avoid oops.
1443 Severity : enhancement
1445 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
1449 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
1450 Details : If an OST is down the MDS will hang indefinitely in
1451 obd_statfs_rqset() waiting for the statfs data. While for
1452 MDS QOS usage of statfs, it should not stuck in waiting.
1454 Severity : enhancement
1456 Description: remote_acl support
1457 Details : Support ACL-based permission check for remote user.
1458 Support setfacl/getfacl for remote user with the utils
1459 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
1460 the system "{s,g}etfacl" utils.
1462 Severity : enhancement
1464 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
1466 Severity : enhancement
1468 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
1472 Description: make mgs_setparam() handle fsname containing dash
1473 Details : fsname containing a dash does not work with lctl conf_param
1475 Severity : enhancement
1477 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
1479 Severity : enhancement
1481 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
1485 Frequency : rare, at shutdown
1486 Description: access already free / zero obd_namespace.
1487 Details : if client_disconnect_export was called without force flag set,
1488 and exist connect request in flight, this can produce access to
1489 NULL pointer (or already free pointer) when connect_interpret
1490 store ocd flags in obd_namespace.
1494 Frequency : only at startup
1495 Description: not alloc memory with spinlock held.
1496 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
1497 if any spinlock held.
1499 Severity : enhancement
1501 Description: make lustre randomly fail allocating memory
1502 Details : Make lustre randomly failed allocating memory for testing purpose.
1504 Severity : enhancement
1506 Description: lost problems with lov objid file
1507 Details : Fixes some scability and access to not inited memory problems
1508 in work with lov objdid file.
1513 Description: lfs find does not continue on file error
1514 Details : Continue other files processing when a file/dir is absent.
1518 Description: Inconsistent usage of lustre_pack_reply()
1519 Details : Standardize the usage of lustre_pack_reply() such that it
1520 always generate a CERROR on failure.
1523 Frequency : occasional
1525 Description: MDS hang or stay in waiting lock
1526 Details : If client receive lock with CBPENDING flag ldlm need send lock
1527 cancel as separate rpc, to avoid situation when cancel request
1528 can't processed due all i/o threads stay in wait lock.
1532 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1536 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1540 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1544 Description: Update to RHEL4 latest kernel.
1546 Severity : enhancement
1548 Description: Build SLES10 patchless client fails
1549 Details : The configure was broken by run ./configure with
1550 --with-linux-obj=.... argument for patchless client. When the
1551 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1552 can't search header adequately. Use absolute path such as
1553 -I($LINUX)/include instead.
1557 Description: interrupt oig_wait produce painc on resend.
1558 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1559 private set, and this produce situation when rq_ptlrpcd_data not
1560 copyed to new allocated request and triggered LBUG on assert
1561 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1562 ptlrpcd set if request is changed and can be safety changed to use
1567 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1568 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1569 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1570 these macros and replace them with their difinitions instead.
1574 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1575 Details : Update which_patch & target file for SLES10 latest kernel.
1578 Description: Remove client patches from SLES 10 kernel.
1579 Details : This causes SLES 10 clients to behave as patchless clients
1580 even on a Lustre-patched (server) kernel.
1582 Severity : enhancement
1584 Description: self-adjustable client's lru lists
1585 Details : use adaptive algorithm for managing client cached locks lru
1586 lists according to current server load, other client's work
1587 pattern, memory activities, etc. Both, server and client
1588 side namespaces provide number of proc tunables for controlling
1591 Severity : enhancement
1593 Description: light-weight GSS support
1594 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
1595 and reduce performance overhead.
1597 Severity : enhancement
1599 Description: Linux keyring support
1600 Details : Support using service of Linux keyring for Lustre GSS internal
1601 context refresh/cache mechanism.
1605 Description: Fix errors in lfs documentation
1606 Details : Fixes man pages
1610 Description: don't use GFP_* in generic Lustre code.
1611 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
1615 Description: obdclass is limited by single OBD_ALLOC(idarray)
1616 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
1620 Description: warnings with build patchless client with vanila 2.6.19 and up
1621 Details : change old ctl_table style and replace ctl_table/ctl_table_header
1622 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
1626 Frequency : Only for SLES
1627 Description: sanity_quota fail test_1
1628 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
1629 file, which makes TSTID[2] inunique.
1633 Frequency : Only for SLES9
1634 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
1635 Details : Update target/ChangeLog/which_patch .
1639 Frequency : Only for test_10 in sanity-quota.sh
1640 Description: a bug in quota test code
1641 Details : checking a test flag in a wrong place
1645 Frequency : Only for test_13 in sanity-quota.sh
1646 Description: a bug in quota test script
1647 Details : shouldn't check overfull allocated quota
1651 Description: jbd statistics
1652 Details : Port older jbd statistics patch for sles10
1654 Severity : enhancement
1656 Description: Add printing inode info into message about error in writepage.
1661 Description: reply_lock_interpret crash due to race with it and lock cancel.
1662 Details : Do not replay locks that are being cancelled. Do not reference
1663 locks by their address during replay, just by their handle.
1667 Frequency : When flocks are used.
1668 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
1669 Details : Flock locks might destroy just granted lock if it could be merged
1670 with another existing flock, this is done in completion handler,
1671 so teach ldlm_cli_enquque_fini that this is a valid case for
1677 Description: Oops in read and write path when failing to allocate lock.
1678 Details : Check if lock allocation failed and return error back.
1682 Description: lstripe command fails for valid OST index
1683 Details : The stripe offset is compared to lov->desc.ld_tgt_count
1684 instead of lov->desc.ld_active_tgt_count.
1688 Description: sanity.sh failed test 103
1689 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
1694 Description: df doesn't work properly if diskfs blocksize != 4K
1695 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
1699 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
1700 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
1702 Severity : enhancement
1704 Description: Adaptive timeouts
1705 Details : RPC timeouts adapt to changing server load and network
1706 conditions to reduce resend attempts and improve recovery time.
1710 Description: llapi_file_create() does not allow some changes
1711 Details : add llapi_file_open() that allows specifying the mode and
1712 open flags, and also returns an open file handle.
1716 Description: merge and cleanup kernel patches.
1717 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
1721 Description: Add journal checksum support.(Kernel part)
1722 Details : The journal checksum feature adds two new flags i.e
1723 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
1724 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
1725 indicates that the commit block contains the checksum for
1726 the blocks described by the descriptor blocks. Now commit
1727 record can be sent to disk without waiting for descriptor
1728 blocks to be written to disk. This behavior is controlled
1729 using JBD2_FEATURE_ASYNC_COMMIT flag.
1733 Description: OSS needs mutliple precreate threads
1734 Details : Add ability to start more than one create thread per OSS.
1738 Description: Sanity.sh test_65k failed: setstripe should have succeeded
1739 Details : execute lfs setstripe on client
1743 Description: mds_obd_create error creating tmp object
1744 Details : When the user sets quota on root, llog will be affected and can't
1745 create files and write files.
1748 Frequency : Always on ia64 patchless client, and possibly others.
1750 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
1751 Details : This allows the patchless client to be loaded on architectures
1752 without this export.
1756 Description: RedHat Update kernel for RHEL5
1757 Details : Add the kernel config file for RHEL5.
1761 Description: RedHat Update kernel for RHEL5
1762 Details : Modify the kernel config file more closer RHEL5.
1766 Description: Build failure against Centos5 (RHEL5)
1767 Details : Define PAGE_SIZE when it isn't present.
1771 Description: test_7 on sanity-quota.sh can't be used on separate machines
1772 Details : add facet to handle it
1776 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
1777 Details : with NFS, the anon dentry's parent was set to itself in
1778 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
1779 corresponding dentry other than use rec->ur_name.
1781 Severity : enhancement
1783 Description: lfs setstripe enhancement
1784 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
1788 Description: enable data checksumming by default
1789 Details : enable checksum by default, allow --disable-checksum
1790 configure option and "-o nochecksum" mount option. Checksums
1791 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
1795 Description: lustre support for RHEL5
1796 Details : Add support for RHEL5.
1800 Description: Client eviction due to failover config
1801 Details : after a connection loss, the lustre client should attempt to
1802 reconnect to the last active server first before trying the
1803 other potential connections.
1807 Description: when mds and osts use different quota unit(32bit and 64bit),
1808 quota will be released repeatly.
1809 Details : void sending multiple quota reqs to mds, which will keep the
1810 status between the reqs.
1814 Description: osts not allocated evenly to files
1815 Details : change the condition to increase offset_idx
1818 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
1820 Description: Data corruption for OSTs that are formatted larger than 2TB
1822 Details : When generating the bio request for lustre file writes the
1823 sector number would overflow a temporary variable before being
1824 used for the IO. The data reads correctly from Lustre (which
1825 will overflow in a similar manner) but other file data or
1826 filesystem metadata may be corrupted in some cases.
1830 Description: Tune the kernel for good SCSI performance.
1831 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
1832 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
1837 Description: rewrite ext2-derived code in obdclass/uuid.c
1838 Details : rewrite inherited code (uuid parsing code from ext2 utils)
1839 from scratch preserving functionality.
1843 Description: "lfs find -obd UUID" prints directories
1844 Details : "lfs find -obd UUID" will return all directory names instead
1845 of just file names. It is incorrect because the directories
1846 do not reside on the OSTs.
1850 Description: use i_size_read and i_size_write in 2.6 port
1851 Details : replace inode->i_size access with i_size_read/write()
1854 Frequency : when removing large files
1856 Description: scheduling issue during removal of large Lustre files
1857 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
1858 It causes scheduling issues when removing large files (17TB in the
1862 Frequency : only with liblustre clients on XT3
1864 Description: evictions taking too long
1865 Details : allow llrd to evict clients directly on OSTs
1868 Frequency : only on ppc
1870 Description: /proc/fs/lustre/devices broken on ppc
1871 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
1872 We should make sure the type of 'index' is loff_t and then cast
1873 explicitly as needed below. Do not assign an explicitly cast
1877 Frequency : only for rhel5
1879 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
1880 Details : Modify the target file & which_kernel.
1882 Severity : enhancement
1884 Description: omit set fsid for export NFS
1885 Details : fix set/restore device id for avoid EMFILE error and mark lustre
1886 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
1890 Frequency : Always, for kernels after 2.6.16
1891 Description: Fix warning idr_remove called for id=.. which is not allocated.
1892 Details : Last kernels save old s_dev before kill super and not allow
1893 to restore from callback - restore it before call kill_anon_super.
1897 Description: buffer overruns could theoretically occur
1898 Details : llapi_semantic_traverse() modifies the "path" argument by
1899 appending values to the end of the origin string, and a
1900 overrun may occur. Adding buffer overrun check in liblustreapi.
1904 Description: Fix error on 'ls .' at the top of the Lustre mount.
1905 Details : Don't revalidate dentry if it is a root dentry.
1909 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
1910 Details : Modify vm-tunables-rhel4.patch.
1914 Description: Kernel config for 2.6.18-vanilla.
1915 Details : Modify targets/2.6-vanilla.target.in.
1916 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
1917 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
1918 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
1919 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
1923 Description: improve handling recoverable errors
1924 Details : if request processig with error which can be recoverable on server
1925 request should be resend, otherwise page released from cache and
1930 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
1931 Details : Modify target file & which_patch.
1932 A flaw was found in the IA32 system call emulation provided
1933 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
1934 value could be stored in the %RAX register, which could trigger an
1935 out-of-bounds system call table access. An untrusted local user
1936 could exploit this flaw to run code in the kernel
1937 (ie a root privilege escalation). (CVE-2007-4573).
1941 Description: O_DIRECT bypasses client statistics.
1942 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
1943 read_ahead_stats were updated. copied the stats section from
1944 osc_send_oap_rpc() into async_internal().
1948 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
1952 Description: change order of libsysio includes
1953 Details : '#include sysio.h' should always come before '#include xtio.h'
1957 Description: handle "serious error: objid * already exists" more gracefully
1958 Details : If LAST_ID value on disk is smaller than the objects existing in
1959 the O/0/d* directories, it indicates disk corruption and causes an
1960 LBUG(). If the object is 0-length, then we should use the existing
1961 object. This will help to avoid a full fsck in most cases.
1963 Severity : enhancement
1965 Description: adapt the lustre_config script to support the upgrade case
1966 Details : Add "-u" option for lustre_config script to support upgrading 1.4
1967 server targets to 1.6 in parallel.
1971 Description: To avoid grant space > avaible space when the disk is almost
1972 full. Without this patch you might see the error "grant XXXX >
1973 available" or some LBUG about grant, when the disk is almost
1975 Details : In filter_check_grant, for non_grant cache write, we should
1976 check the left space by if (*left > ungranted + bytes), instead
1977 of (*left > ungranted), because only we are sure the left space
1978 is enough for another "bytes", then the ungrant space should be
1979 increase. In client, we should update cl_avail_grant only there
1980 is OBD_MD_FLGRANT in the reply.
1984 Description: Update RHEL 4 kernel to fix local root privilege escalation.
1985 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
1986 described in CVE-2007-4573. This problem could allow untrusted
1987 local users to gain root access.
1990 Frequency : when using O_DIRECT and quotas
1992 Description: Incorrect file ownership on O_DIRECT output files
1993 Details : block usage reported by 'lfs quota' does not take into account
1994 files that have been written with O_DIRECT.
1999 Description: touch file failed when fs is not full
2000 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
2001 otherwise we can get ENOSP while fs is not full.
2005 Description: parallel lock callbacks
2006 Details : Instead of sending blocking and completion callbacks as separated
2007 requests, adding them to a set and sending in parallel.
2010 Frequency : only for Cray XT3
2011 Bugzilla : 12829/13455
2012 Description: Changing primary group doesn't change the group lustre assigns to
2014 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
2015 overridden with the primary group provided by the group upcall,
2016 whereas the supplied fsgid can be trusted if it is in the list of
2017 supplementary groups returned by the group upcall.
2019 Severity : enhancement
2021 Description: Allow masking D_WARNING, D_ERROR messages from console
2022 Details : Console messages can now be disabled via lnet.printk.
2026 Description: User code with malformed file open parameter crashes client node
2027 Details : Before packing join_file req, all the related reference should be
2028 checked carefully in case some malformed flags cause fake
2029 join_file req on client.
2033 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
2034 Details : ldlm_completion_ast() assumes that a lock is granted when the req
2035 mode is equal to the granted mode. However, it should also check
2036 that LDLM_FL_CP_REQD is not set.
2040 Description: Heavy nfs access might result in deadlocks
2041 Details : After ELC code landed, it is now improper to enqueue any mds
2042 locks under och_sem, because enqueue might want to decide to
2043 cancel open locks for same inode we are holding och_sem for.
2047 Description: Client eviction while running blogbench
2048 Details : A lot of unlink operations with concurrent I/O can lead to a
2049 deadlock causing evictions. To address the problem, the number of
2050 oustanding OST_DESTROY requests is now throttled to
2051 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
2052 callbacks are processed in priority.
2056 Description: enable ACLs on MDS by default
2057 Details : ACLs must be enabled on MDS by default.
2060 Frequency : PPC/PPC64 only
2062 Description: conflicts between asm-ppc64/types.h and lustre_types.h
2063 Details : fix duplicated definitions between asm-ppc64/types.h and
2064 lustre_types.h on PPC.
2067 Frequency : PPC/PPC64 only
2069 Description: asm-ppc/segment.h does not exist
2070 Details : fix compile issue on PPC.
2074 Description: better handle error messages in extents code
2077 Frequency : RHEL4 only
2079 Description: mkfs is very slow on IA64/RHEL4
2080 Details : A performance regression has been discovered in the MPT Fusion
2081 driver between versions 3.02.73rh and 3.02.99.00rh. As a
2082 consequence, we have downgraded the MPT Fusion driver in the RHEL4
2083 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
2085 Severity : enhancement
2087 Description: SNMP support enhancement
2088 Details : Adding total number of sampled request for an MDS node in snmp
2091 Severity : enhancement
2093 Description: Optimize ldlm waiting list processing for PR extent locks
2094 Details : When processing waiting list for read extent lock and meeting read
2095 lock that is same or wider to it that is not contended, skip
2096 processing rest of the list and immediatelly return current
2097 status of conflictness, since we are guaranteed there are no
2098 conflicting locks in the rest of the list.
2102 Description: Time out and refuse to reconnect
2103 Details : When the failover node is the primary node, it is possible
2104 to have two identical connections in imp_conn_list. We must
2105 compare not conn's pointers but NIDs, otherwise we can defeat
2106 connection throttling.
2110 Description: port llog fixes from b1_6 into HEAD
2111 Details : Port llog reference couting and some llog cleanups from b1_6
2112 (bug 10800) into HEAD, for protect from panic and access to already
2113 free llog structures.
2117 Description: Detect stride IO mode in read-ahead
2118 Details : When a client does stride read, read-ahead should detect that and
2119 read-ahead pages according to the detected stride pattern.
2123 Description: data checksumming impacts single node performance
2124 Details : add support for several checksum algorithm. Currently, only CRC32
2125 and Adler-32 are supported. The checksum type can be changed on
2126 the fly via /proc/fs/lustre/osc/*/checksum_type.
2130 Description: use adler32 for page checksums
2131 Details : when available, use the Adler-32 algorithm instead of CRC32 for
2136 Description: build for x2 fails
2137 Details : fix compile issue on Cray systems.
2141 Description: Properly match for duplicate locks
2142 Details : Due to different lock order from skiplists code, we need to
2143 traverse entire list for now
2146 Frequency : only on PPC/SLES10
2148 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
2149 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
2150 be defined. Add a hack in mkfs_lustre.c to work around this compile
2155 Description: LASSERT on MDS when client holding flock lock dies
2156 Details : ldlm pool logic depends on number of granted locks equal to
2157 number of released locks which is not true for flock locks, so
2158 just exclude such locks from consideration.
2162 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
2163 Details : Use fsfilt_sync() for both the whole filesystem sync and
2164 individual file sync to eliminate dangerous inode locking
2165 with I_LOCK that can lead to a deadlock.
2169 Description: performance in 1.6.3
2170 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
2171 sure that 1MB requests are not fragmented by the block layer.
2173 Severity : enhancement
2175 Description: organize the server-side client stats on per-nid basis
2176 Details : Change the structure of stats under obdfilter and mds to
2184 The "uuid"s file would list the uuids of _active_ exports.
2185 And the clear entry is to clear all stats and stale nids.
2187 Severity : enhancement
2189 Description: eliminate client locks in face of contention
2190 Details : file contention detection and lockless i/o implementation
2191 for contended files.
2195 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
2198 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
2200 Description: MDS may lose file striping (and hence file data) in some cases
2201 Details : If there are additional extended attributes stored on the MDS,
2202 in particular ACLs, SELinux, or user attributes (if user_xattr
2203 is specified for the client mount options) then there is a risk
2204 of attribute loss. Additionally, the Lustre file striping
2205 needs to be larger than default (e.g. striped over all OSTs),
2206 and an additional attribute must be stored initially in the
2207 inode and then increase in size enough to be moved to the
2208 external attribute block (e.g. ACL growing in size) for file
2213 Description: skiplist implementation simplification
2214 Details : skiplists are used to group compatible locks on granted list
2215 that was implemented as tracking first and last lock of each
2216 lock group the patch changes that to using doubly linked lists
2220 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
2221 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
2222 because LASSERT on both the data supplied by a client, and the
2223 data on disk is dangerous and incorrect.
2225 Severity : enhancement
2227 Description: Slow truncate/writes to huge files at high offsets.
2228 Details : Directly associate cached pages to lock that protect those pages,
2229 this allows us to quickly find what pages to write and remove
2230 once lock callback is received.
2234 Description: more ldlm soft lockups
2235 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
2236 starve other threads from the resource lock for a long time in
2237 case of long waiting queue, so change the debug level from
2238 D_OTHER to the less frequently used D_INFO.
2240 Severity : enhancement
2242 Description: add -gid, -group, -uid, -user options to lfs find
2246 Description: Hung threads in invalidate_inode_pages2_range
2247 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
2248 once one is completed. As a result, some RPCs are stuck in the
2249 queue and are never sent.
2253 Description: filter threads hungs on waiting journal commit
2254 Details : Cleanup filter group llog code, then only filter group llog will
2255 be only created in the MDS/OST syncing process.
2259 Description: Procfs and llog threads access destoryed import sometimes.
2260 Details : Sync the import destoryed process with procfs and llog threads by
2261 the import refcount and semaphore.
2263 Severity : enhancement
2265 Description: openlock cache of b1_6 port to HEAD
2270 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
2271 Details : If insertion of an extent fails, then discard the inode
2272 preallocation and free data blocks else it can lead to duplicate
2277 Description: don't always update ctime in ext3_xattr_set_handle()
2278 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
2279 In some cases the ctime should not be updated, for example for
2280 2.0->1.8 compatibility it is necessary to delete an xattr and it
2281 should not update the ctime.
2285 Bugzilla : 15713/16362
2286 Description: Assertion in iopen_connect_dentry in 1.6.3
2287 Details : looking up an inode via iopen with the wrong generation number can
2288 populate the dcache with a disconneced dentry while the inode
2289 number is in the process of being reallocated. This causes an
2290 assertion failure in iopen since the inode's dentry list contains
2291 both a connected and disconnected dentry.
2295 Description: assertion failure in ldlm_handle2lock()
2296 Details : fix a race between class_handle_unhash() and class_handle2object()
2297 introduced in lustre 1.6.5 by bug 13622.
2302 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
2303 Details : remove the limit of 256 scsi disks in the sd_iostat patch
2308 Description: oops in sd_iostats_seq_show()
2309 Details : unloading/reloading the scsi low level driver triggers a kernel
2310 bug when trying to access the sd iostat file.
2315 Description: Kernel panics during QLogic driver reload
2316 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
2317 patch, causing memory corruption.
2322 Description: journal_dev option does not work in b1_6
2323 Details : pass mount option during pre-mount.
2325 Severity : enhancement
2327 Description: Add a FIEMAP(FIle Extent MAP) ioctl
2328 Details : FIEMAP ioctl will allow an application to efficiently fetch the
2329 extent information of a file. It can be used to map logical blocks
2330 in a file to physical blocks in the block device.
2334 Description: LDLM soft lockups - improvement
2335 Details : It is be possible to send the lock handle along with each read
2336 or write request because the client is already doing a lock match
2337 itself so there isn't any reason the OST should have to re-do that
2341 Frequency : only on Cray X2
2343 Description: X2 build failures
2344 Details : fix build failures on Cray X2.
2348 Description: xid & resent requests
2349 Details : Initialize RPC XID from clock at startup (randomly if clock is
2352 Severity : enhancement
2354 Description: Add lustre_start utility to start or stop multiple Lustre servers
2359 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
2360 Details : In case of memory pressure, list_del() can be called twice on
2361 req->rq_history_list, causing a kernel oops.
2365 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
2366 Details : fix stack overflow in the distributed lock manager by defering
2367 export eviction after a failed AST to the elt thread instead of
2368 handling it in the dlm interpret routine.
2372 Description: Convert some comments to new format.
2373 Details : Update documenting comments to match doxygen conventions.
2377 Description: Grammar fixes.
2378 Details : A couple of trivial spelling fixes.
2382 Description: OSD_COUNTERS-mandatory
2383 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
2384 to simplify the code.
2388 Description: simplify cmm_device freeing logic.
2389 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
2393 Description: Add lockdep support to dt_object_operations locking interface.
2394 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
2395 parameter indicating lock ordering. Update mdd code to use new
2400 Description: Introduce failloc constants for lockless IO tests.
2401 Details : Add two new failloc constants to test lockless IO. Only one of
2402 them in implemented---another is checked in yet to be landed
2407 Description: Add lockdep support for inode mutex.
2408 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
2409 in the situations where Lustre has to lock more than one inode
2414 Description: Add optional invariants checking support.
2415 Details : Add new LINVRNT() macro, optional on new --enable-invariants
2416 configure switch. This macro is to be used for consistency and
2417 sanity checks that are too expensive to be left in `production'
2422 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
2423 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
2428 Description: Add lockdep support for ldlm_lock and ldlm_resource.
2429 Details : Use spin_lock_nested() in (the only) situation where more than
2430 one ldlm_lock is locked simultaneously. Also, fix possible
2431 dead-lock in ldlm_lock_change_resource() by enforcing particular
2436 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
2437 Details : Instead of specifying each ldlm_lock call-back through separate
2438 parameter, wrap them into struct ldlm_callback_suite.
2442 Description: Kill join_lru obd method and its callers.
2443 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
2444 in memory---a requirement implemented through ->o_join_lru() obd
2445 method in HEAD. Remove this method and its users.
2449 Description: Add asynchronous ldlm ENQUEUE completion handler.
2450 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
2451 case that stock ldlm_completion_ast() cannot handle as it waits
2452 until lock is granted. Introduce new ldlm_completion_ast_async()
2453 for this. Also comment ldlm_completion_ast().
2457 Description: ldlm_error <-> errno conversion.
2458 Details : Add functions to map (rather arbitrary) between LDLM error codes
2459 and standard errno values. CLIO needs this to prevent LDLM specific
2460 constants from escaping ldlm and osc.
2464 Description: Kill unused ldlm_handle2lock_ns() function.
2465 Details : Kill unused ldlm_handle2lock_ns() function.
2469 Description: Add lu_ref support to ldlm_lock
2470 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
2471 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
2472 and ldlm_resource. LDLM interface has to be changed, because code that
2473 releases a reference on a lock, has to "know" what reference this is.
2474 In the most frequent case
2476 lock = ldlm_handle2lock(handle);
2478 LDLM_LOCK_PUT(lock);
2480 no changes are required. When any other reference (received _not_ from
2481 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
2482 instead of LDLM_LOCK_PUT().
2484 Arguably, changes are pervasive, and interface requires some discipline
2485 for proper use. On the other hand, it was very instrumental in finding
2486 a few leaked lock references.
2490 Description: Add ldlm_lock_addref_try().
2491 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
2492 attempts to addref a lock that might be being canceled
2497 Description: Add ldlm_weigh_callback().
2498 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
2499 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
2500 hard-coded `number of pages' logic.
2504 Description: Add lockdep annotations to llog code.
2505 Details : Use appropriately tagged _nested() locking calls in the places
2506 where llog takes more than one ->lgh_lock lock.
2510 Description: Add loi_kms_set().
2511 Details : Wrap kms updates into a helper function.
2515 Description: Constify instances of struct lsm_operations.
2516 Details : Constify instances of struct lsm_operations.
2520 Description: lu_conf support.
2521 Details : On a server, a file system object is uniquely identified
2522 by a fid, which is sufficient to locate and load all object
2523 state (inode). On a client, on the other hand, more data are
2524 necessary instantiate an object. Change lu_object_find() and
2525 friends to take additional `lu_conf' argument describing object.
2526 Typically this includes layout information.
2530 Description: lu_context fixes.
2531 Details : Introduce new lu_context functions that are needed on the client
2532 side, where some system threads (ptlrpcd) are shared by multiple
2533 modules, and so cannot be stopped during module shutdown.
2537 Description: Add start and stop methods to lu_device_type_operations.
2538 Details : Introduce two new methods in lu_device_type_operations, that are
2539 invoked when first instance of a given type is created and last one
2540 is destroyed respectively. This is need by CLIO.
2544 Description: Add lu_ref support to struct lu_device.
2545 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
2546 track leaked references.
2550 Description: Introduce lu_kmem_descr.
2551 Details : lu_kmem_descr and its companion interface allow to create
2552 and destroy a number of kmem caches at once.
2556 Description: Fix lu_object finalization race.
2557 Details : Fix a race between lu_object_find() finding an object and its
2558 concurrent finalization. This race is (most likely) not possible
2559 on the server, but might happen on the client.
2563 Description: Introduce lu_ref interface.
2564 Details : lu_ref is a debugging module allowing to track references to
2565 a given object. It is quite cpu expensive, and has to be
2566 explicitly enabled with --enable-lu_ref. See usage description
2571 Description: Factor lu_site procfs stats into a separate function.
2572 Details : Separate lu_site stats printing code into a separate function
2573 that can be reused on a client.
2577 Description: Constify instances of struct {lu,dt,md}_device_operations.
2578 Details : Constify instances of struct {lu,dt,md}_device_operations.
2582 Description: Introduce struct md_site and move meta-data specific parts of
2583 struct lu_site here.
2584 Details : Move md-specific fields out of struct lu_site into special struct
2585 md_site, so that lu_site can be used on a client.
2589 Description: Kill mdd_lov_destroy().
2590 Details : Remove unused mdd code.
2594 Description: Add st_block checking to multistat.c.
2595 Details : Add st_block checking to multistat.c.
2599 Description: Add lu_ref support to struct obd_device.
2600 Details : Add lu_ref tracking to obd_device.
2604 Description: Kill obd_set_fail_loc().
2605 Details : Remove unused code.
2609 Description: Add special type for ptlrpc_request interpret functions.
2610 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
2611 there. Actual usage will be in CLIO.
2615 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
2616 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
2617 doesn't work with lockdep.
2621 Description: Add rwv.c test program.
2622 Details : New testing program exercising readv(2) and writev(2) (Qian).
2626 Description: Add sendfile.c test program.
2627 Details : New testing program exercising sendfile(2) (Jay).
2631 Description: Ratelimit a message that can be very frequent.
2632 Details : Ratelimit a memory allocation failure message that can
2637 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2638 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2642 Description: Liblustre build fixes.
2643 Details : Liblustre build fixes.
2647 Description: libcfs: add cfs_{need,cond}_resched() interface.
2648 Details : libcfs: add cfs_{need,cond}_resched() definition and
2649 implementations for Linux, NT, and liblustre.
2651 Severity : enhancement
2653 Description: More exported tunables for mballoc
2654 Details : Add support for tunable preallocation window and new tunables for
2655 large/small requests
2659 Description: Detect corruption of block bitmap and checking for preallocations
2660 Details : Checks validity of on-disk block bitmap. Also it does better
2661 checking of number of applied preallocations. When corruption is
2662 found, it turns filesystem readonly to prevent further corruptions.
2666 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
2667 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
2668 window to 0. If it does hit cache miss, and read-pattern is still
2669 stride-io mode, does not reset the stride window, but also does not
2670 increase the stride window length in this case.
2674 Frequency : only for big-endian servers
2675 Description: Check if system is big-endian while mounting fs with extents feature
2676 Details : Mounting a filesystem with extents feature will fail on big-endian
2677 systems since ext3-based ldiskfs is not supported on big-endian
2678 systems. This can be over-riden with "bigendian_extents" mount option.
2680 Severity : enhancement
2682 Description: The root squash functionality
2683 Details : A security feature, which is to prevent users from being able
2684 to mount lustre on their desktop, run as root, and delete
2685 all of the files in the filesystem. The goal is accomplished by
2686 remapping user id (UID) and group id (GID) of the root user to
2687 a UID and GID specified by the system administartor via Lustre
2688 configuration management server (MGS). The functionality also
2689 allows to specify sets of clients for which the remapping does
2694 Description: Excessive recovery window
2695 Details : With AT enabled, the recovery window can be excessively long (6000+
2696 seconds). To address this problem, we no longer use
2697 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
2698 timeout no longer depends on the service time, it is set to
2699 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
2700 time via pb_service_time.
2704 Description: Watchdog triggered on MDS failover
2705 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
2706 the OSTs know that the MDS "UUID" can be reused for the same export
2707 from a different NID, so we do not need to wait for the export to be
2711 Frequency : rare, only if using MMP with Linux RAID
2713 Description: MMP doesn't work with Linux RAID
2714 Details : While using HA for Lustre servers with Linux RAID, it is possible
2715 that MMP will not detect multiple mounts. To make this work we
2716 need to unplug the device queue in RAID when the MMP block is being
2717 written. Also while reading the MMP block, we should read it from
2718 disk and not the cached one.
2720 Severity : enhancement
2722 Description: open file using fid
2723 Details : A file can be opened using just its fid, like
2724 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
2727 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
2730 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
2731 Details : add proper locking for creation of jbd_4k slab cache
2735 Description: MMP check in ext3_remount() fails without displaying any error
2736 Details : When multiple mount protection fails during remount, proper error
2739 Severity : enhancement
2741 Description: Allow stripe size to be up to 4G-64k
2742 Details : Fix math logic to allow large stripe sizes.
2746 Description: add check for >8TB ldiskfs filesystems
2747 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
2748 Don't allow >8TB ldiskfs filesystems to be mounted without
2749 force_over_8tb mount option
2754 Descriptoin: OSS complains "too many blocks" and crashes
2755 Details : In filter_commitrw_write():
2756 1. lquota_chkquota();
2757 2. obdo_from_inode();
2758 3. lquota_pending_commit();
2759 "2." has changed uid and gid in this case, then caused
2760 lquota_chkquota() and lquota_pending_commit() to use
2761 different uid and gid. That is the root of the bug.
2765 Descriptoin: getstripe doesn't report stripe count = -1 sanely
2766 Details : Detect and print out the 'stripe count = -1' case properly.
2770 Description: Out or order replies might be lost on replay
2771 Details : In ptlrpc_retain_replayable_request if we cannot find retained
2772 request with tid smaller than one currently being added, add it
2773 to the start, not end of the list.
2775 --------------------------------------------------------------------------------
2777 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
2779 * Support for kernels:
2780 2.6.5-7.283 (SLES 9),
2781 2.6.9-55.EL (RHEL 4),
2782 2.6.16.46-0.14 (SLES 10),
2783 2.6.18.8 vanilla (kernel.org)
2784 * Client support for unpatched kernels:
2785 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2786 2.6.16 - 2.6.22 vanilla (kernel.org)
2787 * Due to recently discovered recovery problems, we do not recommend
2788 using patchless RHEL 4 clients with this or any earlier release.
2789 * Recommended e2fsprogs version: 1.39.cfs8
2790 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2791 * Starting with this release, the ldiskfs backing filesystem required
2792 by Lustre is now in its own package, lustre-ldiskfs. This package
2793 should be installed. It is versioned separately from Lustre and
2794 may be released separately in future.
2798 Description: block reactivating mgc import until all deactivates complete
2799 Details : Fix race when failing back MDT/MGS to itself (testing)
2802 Frequency : at statup only
2804 Description: mds_lov_synchronize race leads to various problems
2805 Details : simultaneous MDT->OST connections at startup can cause the
2806 sync to abort, leaving the OSC in a bad state.
2808 Severity : enhancement
2810 Description: add optional extra BUILD_VERSION info
2811 Details : add a new environment variable (namely LUSTRE_VERS) which allows
2812 to override the lustre version.
2815 Frequency : 2.6.18 servers only
2817 Description: ll_kern_mount() doesn't release the module reference
2818 Details : The ldiskfs module reference count never drops down to 0
2819 because ll_kern_mount() doesn't release the module reference.
2824 Description: server LBUG when using old ost_num_threads parameter
2825 Details : Accept the old ost_num_threads parameter but warn that it
2826 is deprecated, and fix an off-by-one error that caused an LBUG.
2831 Description: Transient SCSI error results in persistent IO issue
2832 Details : iobuf->dr_error is not reinitialized to 0 between two
2836 Frequency : sometimes when underlying device returns I/O errors
2838 Description: OSTs not going read-only during write failures
2839 Details : OSTs are not remounted read-only when the journal commit threads
2840 get I/O errors because fsfilt_ext3 calls journal_start/stop()
2841 instead of the ext3 wrappers.
2845 Description: poor connect scaling with increasing client count
2846 Details : Don't run filter_grant_sanity_check for more than 100 exports
2847 to improve scaling for large numbers of clients.
2850 Frequency : SLES10 only
2852 Description: sanity-quota.sh quotacheck failed: rc = -22
2853 Details : Quotas cannot be enabled on SLES10.
2856 Frequency : liblustre clients only
2858 Description: getdirentries does not give error when run on compute nodes
2859 Details : getdirentries does not fail when the size specified as an argument
2860 is too small to contain at least one entry
2862 Severity : enhancement
2864 Description: Add LNET router traceability for debug purposes
2865 Details : If a checksum failure occurs with a router as part of the
2866 IO path, the NID of the last router that forwarded the bulk data
2867 is printed so it can be identified.
2872 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
2873 Details : Due to a race condition, liblustre clients were occasionally
2874 evicted incorrectly.
2876 Severity : enhancement
2878 Description: lfs setstripe use optional parameters instead of postional
2881 Severity : enhancement
2883 Description: Nanosecond timestamp support for ldiskfs
2884 Details : The on-disk ldiskfs filesystem has added support for nanosecond
2885 resolution timestamps. There is not yet support for this at
2886 the Lustre filesystem level.
2889 Frequency : during server recovery
2891 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
2892 Details : request with rq_no_resend flag not awake l_wait_event if they get
2896 Frequency : nfs export on patchless client
2898 Description: connectathon hang when test nfs export over patchless client
2899 Details : Disconnected dentry cannot be found with lookup, so we do not need
2900 to unhash it or make it invalid
2903 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
2904 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
2905 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
2910 Description: open req refcounting wrong on reconnect
2911 Details : If reconnect happened between getting open reply from server and
2912 call to mdc_set_replay_data in ll_file_open, we will schedule
2913 replay for unreferenced request that we are about to free.
2914 Subsequent close will crash in variety of ways.
2915 Check that request is still eligible for replay in
2916 mdc_set_replay_data().
2921 Description: disable writes to filesystem when reading health_check file
2922 Details : the default for reading the health_check proc file has changed
2923 to NOT do a journal transaction and write to disk, because this
2924 can cause reads of the /proc file to hang and block HA state
2925 checking on a healthy but otherwise heavily loaded system. It
2926 is possible to return to the previous behaviour during configure
2927 with --enable-health-write.
2929 Severity : enhancement
2931 Description: 64-bit inode version
2932 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
2933 to the inode. This will be required for version-based recovery.
2938 Description: MDS fails to start if a duplicate client export is detected
2939 Details : in some rare cases it was possible for a client to connect to
2940 an MDS multiple times. Upon recovery the MDS would detect this
2941 and fail during startup. Handle this more gracefully.
2943 Severity : enhancement
2945 Description: Add -o localflock option to simulate old noflock
2947 Details : This will achieve local-only flock/fcntl locks
2953 Description: log_commit_thread vs filter_destroy race leads to crash
2954 Details : Take import reference before releasing llog record semaphore
2959 Description: Wrong request locking in request set processing
2960 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
2961 imp_delayed_list, in this place should be used imp_lock.
2964 Frequency : when reconnection
2966 Description: Grant Leak when osc reconnect to OST
2967 Details : When osc reconnect ost, OST(filter) should check whether it
2968 should grant more space to client by comparing fed_grant and
2969 cl_avail_grant, and return the granted space to client instead
2970 of "new granted" space, because client will call osc_init_grant
2971 to update the client grant space info.
2974 Frequency : when client reconnect to OST
2976 Description: Grant Leak when osc do resend and replay bulk write
2977 Details : When osc reconnect to OST, OST(filter)should clear grant info of
2978 bulk write request, because the grant info while be sync between
2979 OSC and OST when reconnect, and we should ignore the grant info
2980 these of resend/replay write req.
2985 Description: Grant space more than available left space sometimes.
2986 Details : When then OST is about to be full, if two bulk writing from
2987 different clients came to OST. Accord the avaliable space of the
2988 OST, the first req should be permitted, and the second one
2989 should be denied by ENOSPC. But if the seconde arrived before
2990 the first one is committed. The OST might wrongly permit second
2991 writing, which will cause grant space > available space.
2994 Frequency : when client is evicted
2996 Description: Grant might be wrongly erased when osc is evicted by OST
2997 Details : when the import is evicted by server, it will fork another
2998 thread ptlrpc_invalidate_import_thread to invalidate the
2999 import, where the grant will be set to 0. While the original
3000 thread will update the grant it got when connecting. So if
3001 the former happened latter, the grant will be wrongly errased
3002 because of this race.
3007 Description: Checking Stale with correct fid
3008 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
3009 to check whether it is stale, because sometimes, we want the
3010 enqueue happened anyway, and op_data.fid2 will not be initialized.
3012 Severity : enhancement
3014 Description: update patchless client
3015 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
3018 Frequency : only with 2.4 kernel
3020 Description: random memory corruption
3021 Details : size of struct ll_inode_info is to big for union inode.u and this
3022 can be cause of random memory corruption.
3027 Description: Memory leak in recovery
3028 Details : Lov_mds_md was not free in an error handler in mds_create_object.
3029 It should also check obd_fail before fsfilt_start, otherwise if
3030 fsfilt_start return -EROFS,(failover mds during mds recovery).
3031 then the req will return with repmsg->transno = 0 and rc = EROFS.
3032 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
3033 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
3034 be freed no matter whether fsfilt_commit success or not.
3037 Frequency : only with huge count clients
3039 Description: Prevents from taking the superblock lock in llap_from_page for
3041 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
3042 for llap_from_page call in ll_removepage prevents from taking the
3043 superblock lock for a soon died page.
3048 Description: Not check open intent error before release open handle
3049 Details : in some rare cases, the open intent error is not checked before
3050 release open handle, which may cause
3051 ASSERTION(open_req->rq_transno != 0), because it tries to release
3052 the failed open handle.
3057 Description: Set cat log bitmap only after create log success.
3058 Details : in some rare cases, the cat log bitmap is set too early. and it
3059 should be set only after create log success.
3063 Description: Accessing a block bevice can re-enable I/O when Lustre is
3064 tearing down a device.
3065 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
3070 Description: service threads may hog cpus when there are a lot of requests
3071 Details : Insert cond_resched to give other threads a chance to use some CPU
3076 Description: the cat log was not initialized in recovery
3077 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
3078 unlink log on mds will not be initialized until mds post
3079 recovery. And also in mds post recovery, the unlink log will
3080 initialization will be done asynchronausly, so there will be race
3081 between add unlink log and unlink log initialization.
3085 Description: brw_stats were being printed incorrectly
3086 Details : brw_stats were being printed as log2 but all of them were not
3087 recorded as log2. Also remove some code duplication arising from
3088 filter_tally_{read,write}.
3092 Frequency : rare, only in recovery.
3093 Description: ASSERTION(req->rq_type != LI_POISON) failed
3094 Details : imp_lock should be held while iterating over imp_sending_list for
3095 prevent destroy request after get timeout in ptlrpc_queue_wait.
3099 Description: replay-single.sh test 52 fails
3100 Details : A lock's skiplist need to be cleanup when it being unlinked
3101 from its resource list.
3105 Description: Short directio read returns full requested size rather than
3107 Details : Direct I/O operations should return actual amount of bytes
3108 transferred rather than requested size.
3110 Severity : enhancement
3112 Description: metadata RPC reduction (e.g. for rm performance)
3113 Details : decrease the amount of synchronous RPC between clients and servers
3114 by canceling conflicing lock before the operation on the client
3115 and packing thier handles into the main operation RPC to server.
3117 Severity : enhancement
3119 Description: Async OSC create to avoid the blocking unnecessarily.
3120 Details : If an OST has no remaining object, system will block on the
3121 creation when it needs to create a new object on this OST. Now,
3122 ways use pre-created objects when available, instead of blocking on
3123 an empty osc while others are not empty. If we must block, we
3124 block for the shortest possible period of time.
3128 Description: improve handling recoverable errors
3129 Details : if request processig with error which can be recoverable on server
3130 request should be resend, otherwise page released from cache and
3133 Severity : enhancement
3135 Description: refine locking for avoid write wrong info into lov_objid file
3136 Details : fix possible races with add new target and write/update data in
3139 --------------------------------------------------------------------------------
3141 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
3146 Frequency : on some architectures
3148 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
3149 Details : Uninitialized flags sometimes cause configuration commands to
3153 Frequency : patchless clients only
3155 Description: missing __iget() symbol export
3156 Details : The __iget() symbol export is missing. To avoid the need for
3157 this on patchless clients the deathrow inode reaper is turned
3158 off, and we depend on the VM to clean up old inodes. This
3159 dependency was during via the fix for bug 12181.
3161 --------------------------------------------------------------------------------
3163 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
3165 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
3166 INTEROPERATE with older versions automatically. In many cases a
3167 special upgrade step is needed. Please read the
3168 user documentation before upgrading any part of a live system.
3169 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
3170 version will not interoperate with 1.6 betas before beta5 (1.5.95).
3171 * WARNING: Lustre configuration and startup changes are required with
3172 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
3176 Severity : enhancement
3178 Description: MountConf
3179 Details : Lustre configuration is now managed via mkfs and mount
3180 commands instead of lmc and lconf. New obd types (MGS, MGC)
3181 are added for dynamic configuration management. See
3182 https://mail.clusterfs.com/wikis/lustre/MountConf for
3185 Severity : enhancement
3187 Description: dynamic OST addition
3188 Details : OSTs can now be added to a live filesystem
3190 Severity : enhancement
3192 Description: startup order invariance
3193 Details : MDTs and OSTs can be started in any order. Clients only
3194 require the MDT to complete startup.
3196 Severity : enhancement
3198 Description: parallel, asynchronous orphan cleanup
3199 Details : orphan cleanup is now performed in separate threads for each
3200 OST, allowing parallel non-blocking operation.
3202 Severity : enhancement
3204 Description: optimized stripe assignment
3205 Details : stripe assignments are now made based on ost space available,
3206 ost previous usage, and OSS previous usage, in order to try
3207 to optimize storage space and networking resources.
3209 Severity : enhancement
3211 Description: Permanently set tunables
3212 Details : All writable /proc/fs/lustre tunables can now be permanently
3213 set on a per-server basis, at mkfs time or on a live system.
3215 Severity : enhancement
3217 Description: Lustre message v2
3218 Details : Add lustre message format v2.
3220 Severity : enhancement
3222 Description: client OST exclusion list
3223 Details : Clients can be started with a list of OSTs that should be
3224 declared "inactive" for known non-responsive OSTs.
3228 Description: SPEC SFS validation failure on NFS v2 over lustre.
3229 Details : Changes the blocksize for regular files to be 2x RPC size,
3230 and not depend on stripe size.
3232 Severity : enhancement
3234 Description: Multiple MD RPCs in flight.
3235 Details : Further unserialise some read-only MDS RPCs - learn about intents.
3236 To avoid overly-overloading MDS, introduce a limit on number of
3237 MDS RPCs in flight for a single client and add /proc controls
3238 to adjust this limit.
3240 Severity : enhancement
3242 Description: client read/write statistics
3243 Details : Add client read/write call usage stats for performance
3244 analysis of user processes.
3245 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
3246 file access. extents_stats shows chunk size distribution.
3247 extents_stats_per_process show chunk size distribution per
3250 Severity : enhancement
3252 Description: mds statistics
3253 Details : Add detailed mds operations statistics in
3254 /proc/fs/lustre/mds/*/stats.
3258 Description: Failure of copying files with lustre special EAs.
3259 Details : Client side always return success for setxattr call for lustre
3260 special xattr (currently only "trusted.lov").
3265 Description: Refcount LNET uuids
3266 Details : The global LNET uuid list grew linearly with every startup;
3267 refcount repeated list entries instead of always adding to
3270 Severity : enhancement
3272 Description: Dynamic service threads
3273 Details : Within a small range, start extra service threads
3274 automatically when the request queue builds up.
3277 Frequency : mixed-endian client/server environments
3279 Description: mixed-endian crashes
3280 Details : The new msg_v2 system had some failures in mixed-endian
3283 Severity : enhancement
3285 Description: Easy OST removal
3286 Details : OSTs can be permanently deactivated with e.g. 'lctl
3287 conf_param lustre-OST0001.osc.active=0'
3289 Severity : enhancement
3291 Description: MGS proc entries
3292 Details : Added basic proc entries for the MGS showing what filesystems
3295 Severity : enhancement
3297 Description: provide MGS failover
3298 Details : Added config lock reacquisition after MGS server failover.
3300 Severity : enhancement
3302 Description: add Linux 2.4 support
3303 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
3307 Description: a large application tries to do I/O to the same resource and dies
3308 in the middle of it.
3309 Details : Check the req->rq_arrival time after the call to
3310 ost_brw_lock_get(), but before we do anything about
3311 processing it & sending the BULK transfer request. This
3312 should help move old stale pending locks off the queue as
3313 quickly as obd_timeout.
3316 Frequency : when an incorrect nid is specified during startup
3318 Description: ptlrpc connect to non-existant node causes kernel crash
3319 Details : LNET can't be re-entered from an event callback, which
3320 happened when we expire a message after the export has been
3321 cleaned up. Instead, hand the zombie cleanup off to another
3324 Severity : enhancement
3326 Description: plain/inodebits lock performance improvement
3327 Details : Grouping plain/inodebits in granted list by their request modes
3328 and bits policy, thus improving the performance of search through
3332 Frequency : only if OST filesystem is corrupted
3334 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3335 Details : for a short time RPCs with bulk IO are in the replay list,
3336 but replay of bulk IOs is unimplemented. If the OST filesystem
3337 is corrupted due to disk cache incoherency and then replay is
3338 started it is possible to trip an assertion. Avoid putting
3339 committed RPCs into the replay list at all to avoid this issue.
3342 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3344 Description: System hang on startup
3345 Details : This bug allowed the liblustre (e.g. catamount) client to
3346 return to the app before handling all startup RPCs. This
3347 could leave the node unresponsive to lustre network traffic
3348 and manifested as a server ptllnd timeout.
3350 Severity : enhancement
3352 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3353 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
3354 boolean to control whether to print peer debug info when a
3355 client's RPC times out.
3358 Frequency : only for kernels with patches from Lustre below 1.4.3
3360 Description: Remove old rdonly API
3361 Details : Remove old rdonly API which unused from at least lustre 1.4.3
3364 Frequency : only for devices with external journals
3366 Description: Set external device read-only also
3367 Details : During a commanded failover stop, we set the disk device
3368 read-only while the server shuts down. We now also set any
3369 external journal device read-only at the same time.
3372 Frequency : when upgrading from 1.4 while trying to change parameters
3374 Description: The wrong (new) MDC name was used when setting parameters for
3375 upgraded MDT's. Also allows changing of OSC (and MDC)
3376 parameters if --writeconf is specified at tunefs upgrade time.
3379 Frequency : when setting specific ost indicies
3381 Description: QOS code breaks on skipped indicies
3382 Details : Add checks for missing OST indicies in the QOS code, so OSTs
3383 created with --index need not be sequential.
3385 Severity : enhancement
3387 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3388 Details : The uninit_groups feature works in conjunction with the kernel
3389 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3390 up the pass1 processing of e2fsck. This is a read-only feature
3391 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3392 mount filesystems that have had this feature enabled.
3394 Severity : enhancement
3396 Description: Improve multi-block allocation algorithm to avoid fragmentation
3397 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3398 allocation locality and avoid filesystem fragmentation.
3400 ------------------------------------------------------------------------------
3402 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
3404 * Support for kernels:
3405 2.4.21-47.0.1.EL (RHEL 3)
3406 2.6.5-7.283 (SLES 9)
3407 2.6.9-42.0.10.EL (RHEL 4)
3408 2.6.12.6 vanilla (kernel.org)
3409 2.6.16.27-0.9 (SLES 10)
3410 * Recommended e2fsprogs version: 1.39.cfs5
3412 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3416 Frequency : occasional, depends on client load and configuration
3417 Bugzilla : 12181, 12203
3418 Description: data loss for recently-modified files
3420 Details : In some cases it is possible that recently written or created
3421 files may not be written to disk in a timely manner (this should
3422 normally be within 30s unless client IO load is very high).
3423 The problem appears as zero-length files or files that are a
3424 multiple of 1MB in size after a client crash or client eviction
3425 that are missing data at the end of the file.
3427 This problem is more likely to be hit on clients where files are
3428 repeatedly created and unlinked in the same directory, clients
3429 have a large amount of RAM, have many CPUs, the filesystem has
3430 many OSTs, the clients are rebooted frequently, and/or the files
3431 are not accessed by other nodes after being written.
3433 The presence of the problem can be detected by looking at
3434 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
3435 smaller than the second (nr_unused) then dirty files will not
3436 be flushed automatically to disk. "sync; sleep 10" should be
3437 run several times on the node before unmounting it to update
3438 Lustre (this is also safe to run on nodes without this problem).
3440 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
3441 that can cause this same problem, so customers using that kernel
3442 also need to update the kernel in addition to Lustre. In order
3443 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
3445 It is normal that files written just before a client crash (less
3446 than 30s) may not yet have been flushed to disk, even for local
3450 Frequency : frequent on thin XT3 nodes
3452 Description: UUID collision on thin XT3 Linux nodes
3453 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
3454 randomly, since we relied on an insufficiently-seeded PRNG.
3459 Description: OSS hangs after "All ost request buffers busy"
3460 Details : A deadlock between quota and journal operations caused OSS
3461 hangs after printing "All ost request buffers busy."
3464 Frequency : always on liblustre builds
3466 Description: Cleanup compiler warnings on liblustre
3469 Frequency : always on liblustre builds on XT3
3471 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
3475 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
3476 > 32000 subdirectories
3477 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
3478 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
3479 subdirectory count crosses 32000. This will aid e2fsck to
3480 correctly handle more than 32000 subdirectories.
3483 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3485 Description: System hang on startup
3486 Details : This bug allowed the liblustre (e.g. catamount) client to
3487 return to the app before handling all startup RPCs. This
3488 could leave the node unresponsive to lustre network traffic
3489 and manifested as a server ptllnd timeout.
3491 Severity : enhancement
3493 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3494 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
3495 boolean to control whether to print peer debug info when a
3496 client's RPC times out.
3501 Description: make O_SYNC working on 2.6 kernels
3502 Details : 2.6 kernels use different method for mark pages for write,
3503 so need add a code to lustre for O_SYNC work.
3508 Description: Failure to close file and release space on NFS
3509 Details : Put inode details into lock acquired in ll_intent_file_open.
3510 Use mdc_intent_lock in ll_intent_open to properly
3511 detect all kind of errors unhandled by mdc_enqueue.
3516 Description: proc file read during shutdown sometimes raced obd removal,
3518 Details : Add lock to prevent obd access after proc file removal.
3521 Frequency : Only for files larger than 4GB on 32-bit clients.
3523 Description: improperly doing page alignment of locks
3524 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
3525 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3530 Description: RPCs being resent when they shouldn't be
3531 Details : Some RPCs that should not be resent are being resent. This
3532 can cause inconsistencies in the RPC state machine. Do not
3533 resend such requests.
3536 Frequency : rare, only with NFS export
3538 Description: Crash on NFS re-export node
3539 Details : under very unusual load conditions an assertion is hit in
3540 ll_intent_file_open()
3543 Frequency : only if OST filesystem is corrupted
3545 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3546 Details : for a short time RPCs with bulk IO are in the replay list,
3547 but replay of bulk IOs is unimplemented. If the OST filesystem
3548 is corrupted due to disk cache incoherency and then replay is
3549 started it is possible to trip an assertion. Avoid putting
3550 committed RPCs into the replay list at all to avoid this issue.
3555 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3556 Details : Large single O_DIRECT read and write calls can fail to allocate
3557 a sufficiently large buffer to process the request. In case of
3558 allocation failure the allocation is retried with a smaller
3559 buffer and broken into smaller requests.
3561 Severity : enhancement
3563 Description: Add -o localflock option to simulate old noflock behaviour.
3564 Details : This will achieve local-only flock/fcntl locks coherentness.
3569 Description: versioning check is incomplete
3570 Details : Checking the version difference of client vs. server, report
3571 error if the gap is too big.
3576 Description: add support PG_writeback bit
3577 Details : add support for PG_writeback bit for Lustre, for more carefull
3578 work with page cache in 2.6 kernel. This also fix some deadlocks
3579 and remove hack for work O_SYNC with 2.6 kernel.
3581 Severity : enhancement
3583 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3584 Details : The uninit_groups feature works in conjunction with the kernel
3585 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3586 up the pass1 processing of e2fsck. This is a read-only feature
3587 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3588 mount filesystems that have had this feature enabled.
3590 Severity : enhancement
3592 Description: Improve multi-block allocation algorithm to avoid fragmentation
3593 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3594 allocation locality and avoid filesystem fragmentation.
3596 ------------------------------------------------------------------------------
3598 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3600 * Support for kernels:
3601 2.6.9-42.0.3.EL (RHEL 4)
3602 2.6.5-7.276 (SLES 9)
3603 2.4.21-47.0.1.EL (RHEL 3)
3604 2.6.12.6 vanilla (kernel.org)
3605 2.6.16.21-0.8 (SLES10)
3606 * Recommended e2fsprogs version: 1.39.cfs2-0
3608 * The backwards-compatible /proc/sys/portals symlink has been removed
3609 in this release. Before upgrading, please ensure that you change
3610 any configuration scripts or /etc/sysctl.conf files that access
3611 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3612 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3613 in advance of the upgrade on any system running Lustre 1.4.6 or
3614 newer, since /proc/sys/lnet was added in that version.
3615 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3619 Frequency : only when quota is used
3621 Description: avoid scanning export list for quota master
3622 Details : Change the algorithms to avoid scanning export list in order
3623 to improve the efficiency.
3626 Frequency : MDS failover only, very rarely
3628 Description: "went back in time" messages on mds failover
3629 Details : The greatest transno may be lost when the current operation
3630 finishes with an error (transno==0) and the client's last_rcvd
3631 record is over-written. Save the greatest transno in the
3632 mds_last_transno for this case.
3635 Frequency : always for specific kernels and striping counts
3637 Description: client may get "Matching packet too big" without ACL support
3638 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
3639 when trying to access files in certain configurations. The
3640 clients should in fact be denied when mounting because they do
3641 not understand ACLs.
3644 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
3646 Description: many clients connecting with IO in progress causes connect timeouts
3647 Details : Avoid synchronous journal commits to avoid delays caused by many
3648 clients connecting/disconnecting when bulk IO is in progress.
3649 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
3650 OST_IO_PORTAL to avoid delays behind potentially many pending
3654 Frequency : occasionally with multiple writers to a single file
3656 Description: shared writes to file may result in wrong size reported by stat()
3657 Details : Allow growing of kms when extent lock is cancelled
3660 Frequency : always with random mmap IO to multi-striped file
3662 Description: mmap write might be lost if we are writing to a 'hole' in stripe
3663 Details : Only if the hole is at the end of OST object so that kms is too
3664 small. Fix is to increase kms accordingly in ll_nopage.
3667 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
3669 Description: writes to a missing object would leak memory on the OST
3670 Details : If there is an inconsistency between the MDS and OST filesystems,
3671 such that the MDS references an object that doesn't exist, writes
3672 to that object will leak memory due to incorrect cleanup in the
3673 error handling path, eventually running out of memory on the OST.
3678 Description: Creating too long symlink causes lustre errors
3679 Details : Check symlink and name lengths before sending requests to MDS.
3682 Frequency : only if flock is enabled (not on by default)
3684 Description: posix locks not released on fd closure on 2.6.9+
3685 Details : We failed to add posix locks to list of inode locks on 2.6.9+
3686 kernels, this caused such locks not to be released on fd close and
3687 then assertions on fs unmount about still used locks.
3690 Frequency : MDS failover only, very rarely
3692 Description: clients may get ASSERTION(granted_lock != NULL)
3693 Details : When request was taking a long time, and a client was resending
3694 a getattr by name lock request. The were multiple lock requests
3695 with the same client lock handle and
3696 mds_getattr_name->fixup_handle_for_resent_request found one of the
3697 lock handles but later failed with ASSERTION(granted_lock != NULL).
3702 Description: handle->h_buffer_credits > 0, assertion failure
3703 Details : h_buffer_credits is zero after truncate, causing assertion
3704 failure. This patch extends the transaction or creates a new
3708 Frequency : NFS re-export or patchless client
3709 Bugzilla : 11179, 10796
3710 Description: Crash on NFS re-export node (__d_move)
3711 Details : We do not want to hash the dentry if we don't have a lock.
3712 But if this dentry is later used in d_move, we'd hit uninitialised
3713 list head d_hash, so we just do this to init d_hash field but
3714 leave dentry unhashed.
3717 Frequency : NFS re-export or patchless client
3719 Description: NFS exports has problem with symbolic link
3720 Details : lustre client didn't properly install dentry when re-exported
3721 to NFS or running patchless client.
3724 Frequency : NFS re-export or patchless client
3726 Description: Various nfs/patchless fixes.
3727 Details : fixes reuse disconected alias for lookup process - this fixes
3728 warning "find_exported_dentry: npd != pd",
3729 fix permission error with open files at nfs.
3730 fix apply umask when do revalidate.
3733 Frequency : occasional
3735 Description: Crash on NFS re-export node
3736 Details : calling clear_page() on the wrong pointer triggered oops in
3737 generic_mapping_read().
3740 Frequency : rarely, using O_DIRECT IO
3742 Description: unaligned directio crashes client with LASSERT
3743 Details : check for unaligned buffers before trying any requests.
3746 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
3748 Description: stale data returned from RAID cache
3749 Details : If only a small amount of IO is done to the RAID device before
3750 reading it again it is possible to get stale data from the RAID
3751 cache instead of reading it from disk.
3754 Frequency : always for sles10 kernel
3756 Description: sles10 support
3757 Details : ll_follow_link: compile fixes and using of nd_set_link
3758 under newer kernels.
3761 Frequency : depends on arch, kernel and compiler version, always on sles10
3764 Description: recursive or deep enough symlinks cause stack overflow
3765 Details : getting rid of large stack-allocated variable in
3769 Frequency : depends on hardware
3771 Description: lustre write performance loss in the SLES10 kernel
3772 Details : the performance loss is caused by using of write barriers in the
3773 ext3 code. The SLES10 kernel turns barrier support on by
3774 default. The fix is to undo that change for ldiskfs.
3776 ------------------------------------------------------------------------------
3778 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
3780 * Support for kernels:
3781 2.6.9-42.0.3EL (RHEL 4)
3782 2.6.5-7.276 (SLES 9)
3783 2.4.21-47.0.1.EL (RHEL 3)
3784 2.6.12.6 vanilla (kernel.org)
3788 Frequency : quota enabled and large files being deleted
3790 Description: releasing more than 4GB of quota at once hangs OST
3791 Details : If a user deletes more than 4GB of files on a single OST it
3792 will cause the OST to spin in an infinite loop. Release
3793 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3798 Description: statfs data retrieved from /proc may be stale or zero
3799 Details : When reading per-device statfs data from /proc, in the
3800 {kbytes,files}_{total,free,avail} files, it may appear
3801 as zero or be out of date.
3804 Frequency : systems with MD RAID1 external journal devices
3806 Description: lconf's call to blkid is confused by RAID1 journal devices
3807 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3808 of returning all block devices that match the journal UUID.
3811 Frequency : always, for aggregate stripe size over 4GB
3813 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
3814 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
3815 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
3818 Frequency : always if "lfs find" used on a local file/directory
3820 Description: "lfs find" segfaults if used on a local file/directory
3821 Details : The case where a directory component was not specified wasn't
3822 handled correctly. Handle this properly.
3825 Frequency : always on ppc64
3827 Description: the write to an ext3 filesystem mounted with mballoc got stuck
3828 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3829 perform endianness conversion.
3832 Frequency : rarely (truncate to non-zero file size after write under load)
3833 Bugzilla : 10730, 10687
3834 Description: Files padded with zeros to next 4K multiple
3835 Details : With filesystems mounted using the "extents" option (2.6 kernels)
3836 it is possible that files that are truncated to a non-zero size
3837 immediately after being written are filled with zero bytes beyond
3838 the truncated size. No file data is lost.
3840 Severity : enhancement
3842 Description: Allow recovery/failover for liblustre clients.
3843 Details : liblustre clients were unaware of failover configurations until
3846 Severity : enhancement
3848 Description: user file locks should fail when not mounting with flock option
3849 Details : Set up an error-returning stub in ll_file_operations.lock field
3850 to prevent incorrect behaviour when client is mounted without
3851 flock option. Also, set up properly f_op->flock field for
3855 Frequency : always on ia64
3857 Description: "lfs df" loops on printing out MDS statfs information
3858 Details : The obd_ioctl_data was not initialized and in some systems
3859 this caused a failure during the ioctl that did not return
3860 an error. Initialize the struct and return an error on failure.
3863 Frequency : SLES 9 only
3865 Description: Error of copying files with lustre special EAs as root
3866 Details : Client side always return success for setxattr call for lustre
3867 special xattr (currently only "trusted.lov").
3870 Frequency : rarely on clusters with both ia64+i386 clients
3872 Description: ia64+i686 clients doing shared IO on the same file may LBUG
3873 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
3874 clients are doing concurrent writes to the same file it is
3875 possible that the ia64 clients may LASSERT because the OST
3876 extent locks are not PAGE_SIZE aligned. Ensure that grown
3877 locks are always aligned on the request boundary.
3880 Frequency : specific use, occasional
3882 Description: Overwriting in use executable truncates on-disk binary image
3883 Details : If one node attempts to overwrite an executable in use by
3884 another node, we now correctly return ETXTBSY instead of
3885 truncating the file.
3887 Severity : enhancement
3889 Description: Async OSC create to avoid the blocking unnecessarily.
3890 Details : If an OST has no remaining object, system will block on the creating
3891 when need to create a new object on this OST. Now, ways use
3892 pre-created objects when available, instead of blocking on an
3893 empty osc while others are not empty. If we must block, we block
3894 for the shortest possible period of time.
3899 Description: chmod on Lustre root is propagated to other clients
3900 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
3901 invalid by the follow_mount time.
3904 Frequency : liblustre clients only
3906 Description: Race in 'instant cancel' lock handling could lead to such locks
3907 never to be granted in case of SMP MDS
3908 Details : Do not destroy not yet granted but cbpending locks in
3912 Frequency : replay/resend of open
3914 Description: non null lock assetion failure in mds_intent_policy
3915 Details : Trying to replay/resend lockless open requests resulted in
3916 mds_open() returning 0 with no lock. Now it sets a flag if
3917 a lock is going to be returned.
3919 Severity : enhancement
3921 Description: Checksum enhancements
3922 Details : New checksum enhancements allow for resending RPCs that failed
3925 Severity : enhancement
3927 Description: Tunables on number of dirty pages in cacche
3928 Details : Allow to set limit on number of dirty pages cached.
3933 Description: client crash on unmount - lock still has references
3934 Details : In some error handling cases it was possible to leak a lock
3935 reference on a client while accessing a file. This was not
3936 harmful to the client during operation, but would cause the
3937 client to crash when the filesystem is unmounted.
3940 Frequency : specific case, rare
3942 Description: ETXTBSY on mds though file not in use by client
3943 Details : ETXTBSY is no longer incorrectly returned when attempting to
3944 chmod or chown a directory that the user previously tried to
3945 execute or a currently-executing binary.
3948 Frequency : extremely rare except on liblustre-based clients
3950 Description: Lustre space not freed when files are deleted
3951 Details : Clean up open-unlinked files after client eviction. Previously
3952 the unlink was skipped and the files remained as orphans.
3957 Description: OST failure "would be an LBUG" in waiting_locks_callback()
3958 Details : In some cases it was possible to send a blocking callback to
3959 a client doing a glimpse, even though that client didn't get
3960 a lock granted. When the glimpse lock is cancelled on the OST
3961 the freed lock is left on the waiting list and corrupted the list.
3964 Frequency : all core dumps
3966 Description: Broke core dumps to lustre
3967 Details : Negative dentry may be unhashed if parent does not have UPDATE
3968 lock, but some callers, e.g. do_coredump, expect dentry to be
3969 hashed after successful create, hash it in ll_create_it.
3971 ------------------------------------------------------------------------------
3973 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
3975 * Support for kernels:
3976 2.6.9-42.0.2.EL (RHEL 4)
3977 2.6.5-7.276 (SLES 9)
3978 2.4.21-40.EL (RHEL 3)
3979 2.6.12.6 vanilla (kernel.org)
3983 Frequency : always on RHEL 3
3985 Description: Number of open files grows over time
3986 Details : The number of open files grows over time, whether or not
3987 Lustre is started. This was due to a filp leak introduced
3988 by one of our kernel patches.
3990 ------------------------------------------------------------------------------
3992 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
3994 * Support for kernels:
3995 2.6.9-42.EL (RHEL 4)
3996 2.6.5-7.276 (SLES 9)
3997 2.4.21-40.EL (RHEL 3)
3998 2.6.12.6 vanilla (kernel.org)
4003 Bugzilla : 5719, 9635, 9792, 9684
4004 Description: OST (or MDS) trips assertions in (re)connection under heavy load
4005 Details : If a server is under heavy load and cannot reply to new
4006 connection requests before the client resends the (re)connect,
4007 the connection handling code can behave badly if two service
4008 threads are concurrently handing separate (re)connections from
4009 the same client. Add better locking to the connection handling
4010 code, and ensure that only a single connection will be processed
4011 for a given client UUID, even if the lock is dropped.
4013 Severity : enhancement
4015 Description: add TCP zero-copy support to kernel
4016 Details : Add support to the kernel TCP stack to allow zero-copy bulk
4017 sends if the hardware supports scatter-gather and checksumming.
4018 This allows socklnd to do client-write and server-read more
4019 efficiently and reduce CPU utilization from skbuf copying.
4022 Frequency : only if NFS exporting from client
4024 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
4025 Details : If setting attributes on a file created under NFS that had
4026 never been opened it would be possible to oops the client
4027 if the file had no objects.
4030 Frequency : always for liblustre
4032 Description: liblustre client does MDS+OSTs setattr RPC for each write
4033 Details : When doing a write from a liblustre client, the client
4034 incorrectly issued an RPC to the MDS and each OST the file was
4035 striped over in order to update the timestamps. When writing
4036 with small chunks and many clients this could overwhelm the MDS
4037 with RPCs. In all cases it would slow down the write because
4038 these RPCs are unnecessary.
4040 Severity : enhancement
4042 Description: allow number of MDS service threads to be changed at module load
4043 Details : It is now possible to change the number of MDS service threads
4044 running. Adding "options mds mds_num_threads={N}" to the MDS's
4045 /etc/modprobe.conf will set the number of threads for the next
4046 time Lustre is restarted (assuming the "mds" module is also
4047 reloaded at that time). The default number of threads will
4048 stay the same, 32 for most systems.
4053 Description: OST crash if filesystem is unformatted or corrupt
4054 Details : If an OST is started on a device that has never been formatted
4055 or if the filesystem is corrupt and cannot even mount then the
4056 error handling cleanup routines would dereference a NULL pointer.
4061 Description: NULL pointer deref in llap_from_page.
4062 Details : get_cache_page_nowait can return a page with NULL (or otherwise
4063 incorrect) mapping if the page was truncated/reclaimed while it was
4064 searched for. Check for this condition and skip such pages when
4065 doing readahead. Introduce extra check to llap_from_page() to
4066 verify page->mapping->host is non-NULL (so page is not anonymous).
4069 Frequency : Sometimes when using sys_sendfile
4071 Description: "page not covered by a lock" warnings from ll_readpage
4072 Details : sendfile called ll_readpage without right page locks present.
4073 Now we introduced ll_file_sendfile that does necessary locking
4074 around call to generic_file_sendfile() much like we do in
4078 Frequency : with certain MDS communication failures at client mount time
4080 Description: NULL pointer deref after failed client mount
4081 Details : a client connection request may delayed by the network layer
4082 and not be sent until after the PTLRPC layer has timed out the
4083 request. If the client fails the mount immediately it will try
4084 to clean up before the network times out the request. Add a
4085 reference from the request import to the obd device and delay
4086 the cleanup until the network drops the request.
4089 Frequency : occasionally during client (re)connect
4091 Description: assertion failure during client (re)connect
4092 Details : processing a client connection request may be delayed by the
4093 client or server longer than the client connect timeout. This
4094 causes the client to resend the connection request. If the
4095 original connection request is replied in this interval, the
4096 client may trip an assertion failure in ptlrpc_connect_interpret()
4097 which thought it would be the only running connect process.
4100 Frequency : only with obd_echo servers and clients that are rebooted
4102 Description: kernel BUG accessing uninitialized data structure
4103 Details : When running an obd_echo server it did not start the ping_evictor
4104 thread, and when a client was evicted an uninitialized data
4105 structure was accessed. Start the ping_evictor in the RPC
4106 service startup instead of the OBD startup.
4108 Severity : enhancement
4109 Bugzilla : 10193 (patchless)
4110 Description: Remove dependency on various unexported kernel interfaces.
4111 Details : No longer need reparent_to_init, exit_mm, exit_files,
4112 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
4115 Frequency : rare (only users of deprecated and unsupported LDAP config)
4117 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
4118 Details : LDAP apparently contains 'lustreName' attributes instead of
4119 'name'. A simple remapping of the name is sufficient.
4122 Frequency : rare (only with non-default dump_on_timeout debug enabled)
4124 Description: waiting_locks_callback trips kernel BUG if client is evicted
4125 Details : Running with the dump_on_timeout debug flag turned on makes
4126 it possible that the waiting_locks_callback() can try to dump
4127 the Lustre kernel debug logs from an interrupt handler. Defer
4128 this log dumping to the expired_lock_main() thread.
4130 Severity : enhancement
4132 Description: Support NFS exporting on 2.6 kernels.
4133 Details : Implement non-rawops metadata methods for NFS server to use without
4134 changing NFS server code.
4137 Frequency : very rare (synthetic metadata workload only)
4139 Description: two racing renames might cause an MDS thread to deadlock
4140 Details : Running the "racer" program may cause one MDS thread to rename
4141 a file from being the source of a rename to being the target of
4142 a rename at exactly the same time that another thread is doing
4143 so, and the second thread has already enqueued these locks after
4144 doing a lookup of the target and is trying to relock them in
4145 order. Ensure that we don't try to re-lock the same resource.
4148 Frequency : only very large systems with liblustre clients
4150 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
4151 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
4152 all OSTs in parallel. This allows the eviction of stale liblustre
4153 clients to proceed much faster than if they were done in series,
4154 and also offers similar improvements for other set_info RPCs.
4159 Description: excessive CPU usage during initial read phase on client
4160 Details : During the initial read phase on a client, it would agressively
4161 retry readahead on the file, consuming too much CPU and impacting
4162 performance (since 1.4.5.8). Improve the readahead algorithm
4163 to avoid this, and also improve some other common cases (read
4164 of small files in particular, where "small" is files smaller than
4165 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
4170 Description: MDS crash when receiving packet with unknown intent.
4171 Details : Do not LBUG in unknown intent case, just return -EFAULT
4173 Severity : enhancement
4174 Bugzilla : 9293, 9385
4175 Description: MDS RPCs are serialised on client. This is unnecessary for some.
4176 Details : Do not serialize getattr (non-intent version) and statfs.
4179 Frequency : occasional, when OST network is overloaded/intermittent
4181 Description: client evicted by OST after bulk IO timeout
4182 Details : If a client sends a bulk IO request (read or write) the OST
4183 may evict the client if it is unresposive to its data GET/PUT
4184 request. This is incorrect if the network is overloaded (takes
4185 too long to transfer the RPC data) or dropped the OST GET/PUT
4186 request. There is no need to evict the client at all, since
4187 the pinger and/or lock callbacks will handle this, and the
4188 client can restart the bulk request.
4191 Frequency : Always when mmapping file with no objects
4193 Description: client crashes when mmapping file with no objects
4194 Details : Check that we actually have objects in a file before doing any
4195 operations on objects in ll_vm_open, ll_vm_close and
4201 Description: Request leak when working with deleted CWD
4202 Details : Introduce advanced request refcount tracking for requests
4203 referenced from lustre intent.
4205 Severity : Enhancement
4207 Description: Cache open file handles on client.
4208 Details : MDS now will return special lock along with openhandle, if
4209 requested and client is allowed to hold openhandle, even if unused,
4210 until such a lock is revoked. Helps NFS a lot, since NFS is opening
4211 closing files for every read/write openration.
4213 Severity : Enhancement
4215 Description: Cache open negative dentries on client when possible.
4216 Details : Guard negative dentries with UPDATE lock on parent dir, drop
4217 negative dentries on lock revocation.
4222 Description: Remounting a client read-only wasn't possible with a zconf mount
4223 Details : It wasn't possible to remount a client read-only with llmount.
4225 Severity : enhancement
4226 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
4227 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
4228 MPICH is included to add Lustre-specific ADIO interfaces.
4229 This is based closely on the UFS ADIO layer and only differs
4230 in file creation, in order to allow the OST striping to be set.
4231 This is user-contributed code and not supported by CFS.
4236 Description: extended inode attributes (immutable, append-only) work improperly
4237 when 2.4 and 2.6 kernels are used on client/server or vice versa
4238 Details : Introduce kernel-independent values for these flags.
4240 Severity : enhancement
4243 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
4244 Details : Many of the /proc/ tunables can only be tuned at a megabyte
4245 granularity. Now, Fractional MB granularity is be supported,
4246 this is very useful for low memory system.
4248 Severity : enhancement
4250 Description: Getattr by fid
4251 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
4252 locks, avoids extra getattr rpc requests to MDS, allows '/' to
4253 have locks and avoids getattr rpc requests for it on every stat.
4256 Frequency : Always, for filesystems larger than 2TB
4258 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4259 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
4260 filesystems larger than 2TB to be mounted with mballoc enabled.
4263 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
4265 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4266 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
4267 filesystems larger than 2TB is mounted then the kernel will
4268 silently corrupt the start of the filesystem. CONFIG_LBD is
4269 enabled for all CFS-supported kernels, but the possibility of
4270 this happening with a modified kernel config exists.
4272 Severity : enhancement
4274 Description: add client O_DIRECT support for 2.6 kernels
4275 Details : It is now possible to do O_DIRECT reads and writes to files
4276 in the Lustre client mountpoint on 2.6 kernel clients.
4278 Severity : enhancement
4280 Description: parallel glimpse, setattr, statfs, punch, destroy requests
4281 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
4282 parallel, not waiting for response from every OST before sending
4283 a rpc to the next OST.
4288 Description: setattr vs write race when updating file timestamps
4289 Details : Client processes that update a file timestamp into the past
4290 right after writing to the file (e.g. tar) it is possible that
4291 the updated file modification time can be reset to the current
4292 time due to a race between processing the setattr and write RPC.
4294 Severity : enhancement
4296 Description: Bring 'lfs find' closer in line with regular Linux find.
4297 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
4298 -print0 options and obtains all the needed info through the lustre
4301 Severity : enhancement
4303 Description: support up to 1024 configured devices on one node
4304 Details : change obd_dev array from statically allocated to dynamically
4305 allocated structs as they are first used to reduce memory usage
4310 Description: Flush dirty partially truncated pages during truncate
4311 Details : Immediatelly flush partially truncated pages in filter_setattr,
4312 this way we completely avoid having any pages in page cache on OST
4313 and can retire ugly workarounds during writes to flush such pages.
4318 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
4319 Details : protect inode from truncation within vfs_unlink() context
4320 just take a reference before calling vfs_unlink() and release it
4321 when parent's i_sem is free.
4326 Description: last_id value checked outside lock on OST caused LASSERT failure
4327 Details : If there were multiple MDS->OST object precreate requests in
4328 flight, it was possible that the OST's last object id was checked
4329 outside a lock and incorrectly tripped an assertion. Move checks
4330 inside locks, and discard old precreate requests.
4333 Frequency : always, if extents are used on OSTs
4335 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
4336 Details : OSTs using the extents format would not zero the high 16 bits of
4337 the index physical block number. This is not a problem for any
4338 OST filesystems smaller than 16TB, and no kernels support ext3
4339 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
4340 new/modified files) and can be fixed for existing filesystems
4341 with e2fsprogs-1.39-cfs1.
4346 Description: import connection selection may be incorrect if timer wraps
4347 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
4348 import connections to be ignored if the 32-bit jiffies counter
4349 wraps. Use a 64-bit jiffies counter.
4352 Frequency : very large clusters immediately after boot
4354 Description: LNET request buffers exhausted under heavy short-term load
4355 Details : If a large number of client requests are generated on a service
4356 that has previously never seen so many requests it is possible
4357 that the request buffer growth cannot keep up with the spike in
4358 demand. Instead of dropping incoming requests, they are held in
4359 the LND until the RPC service can accept more requests.
4362 Frequency : Sometimes during replay
4364 Description: Assertion failure in ll_local_open after replay.
4365 Details : If replay happened on an open request reply before we were able
4366 to set replay handler, reply will become not swabbed tripping the
4367 assertion in ll_local_open. Now we set the handler right after
4368 recognising of open request
4371 Frequency : very rare
4373 Description: kernel reports "badness in vsnprintf"
4374 Details : Reading from the "recovery_status" /proc file in small chunks
4375 may cause a negative length in lprocfs_obd_rd_recovery_status()
4376 call to vsnprintf() (which is otherwise harmless). Exit early
4377 if there is no more space in the output buffer.
4379 Severity : enhancement
4381 Description: clear OBD RPC statistics by writing to them
4382 Details : It is now possible to clear the OBD RPC statistics by writing
4383 to the "stats" file.
4388 Description: Client mtime is not the same on different clients after utimes
4389 Details : In some cases, the client was using the utimes() syscall on
4390 a file cached on another node. The clients now validate the
4391 ctime from the MDS + OSTs to determine which one is right.
4396 Description: Inability to activate failout mode
4397 Details : lconf script incorrectly assumed that in pythong string's numeric
4398 value is used in comparisons.
4401 Frequency : always with multiple stripes per file
4403 Description: Inefficient object allocation for mutli-stripe files
4404 Details : When selecting which OSTs to stripe files over, for files with
4405 a stripe count that divides evenly into the number of OSTs,
4406 the MDS is always picking the same starting OST for each file.
4407 Return the OST selection heuristic to the original design.
4412 Description: mount failures may take full timeout to return an error
4413 Details : Under some heavy load conditions it is possible that a
4414 failed mount can wait for the full obd_timeout interval,
4415 possibly several minutes, before reporting an error.
4416 Instead return an error as soon as the status is known.
4418 Frequency : quota enabled and large files being deleted
4420 Description: releasing more than 4GB of quota at once hangs OST
4421 Details : If a user deletes more than 4GB of files on a single OST it
4422 will cause the OST to spin in an infinite loop. Release
4423 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4428 Description: statfs data retrieved from /proc may be stale or zero
4429 Details : When reading per-device statfs data from /proc, in the
4430 {kbytes,files}_{total,free,avail} files, it may appear
4431 as zero or be out of date.
4434 Frequency : systems with MD RAID1 external journal devices
4436 Description: lconf's call to blkid is confused by RAID1 journal devices
4437 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4438 of returning all block devices that match the journal UUID.
4441 Frequency : always, for aggregate stripe size over 4GB
4443 Description: assertion fails when trying to use 4GB stripe size
4444 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
4445 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
4448 Frequency : always on ppc64
4450 Description: the first write on an ext3 filesystem with mballoc got stuck
4451 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4452 perform endianness conversion.
4454 ------------------------------------------------------------------------------
4456 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4458 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
4459 INTEROPERATE with older versions automatically. Please read the
4460 user documentation before upgrading any part of a live system.
4461 * WARNING: Lustre networking configuration changes are required with
4462 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
4465 * Support for kernels:
4466 2.6.9-22.0.2.EL (RHEL 4)
4467 2.6.5-7.244 (SLES 9)
4468 2.6.12.6 vanilla (kernel.org)
4471 Severity : enhancement
4472 Bugzilla : 7981/8208
4473 Description: Introduced Lustre Networking (LNET)
4474 Details : LNET is new networking infrastructure for Lustre, it includes
4475 a reorganized network configuration mode (see the user
4476 documentation for full details) as well as support for routing
4477 between different network fabrics. Lustre Networking Devices
4478 (LNDs) for the supported network fabrics have also been
4479 created for this new infrastructure.
4481 Severity : enhancement
4482 Description: Introduced Access control lists
4483 Details : clients can set ACLs on files and directories in order to have
4484 more fine-grained permissions than the standard Unix UGO+RWX.
4485 The MDS must be started with the "-o acl" mount option.
4487 Severity : enhancement
4488 Description: Introduced filesystem quotas
4489 Details : Administrators may now establish per-user quotas on the
4492 Severity : enhancement
4494 Description: Configuration change for the XT3
4495 The PTLLND is now used to run Lustre over Portals on the XT3
4496 The configure option(s) --with-cray-portals are no longer used.
4497 Rather --with-portals=<path-to-portals-includes> is used to
4498 enable building on the XT3. In addition to enable XT3 specific
4499 features the option --enable-cray-xt3 must be used.
4504 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
4505 Details : A race between allocating a new llcd and re-getting the llcd_lock
4506 allowed another thread to grab newly-allocated llcd.
4508 Severity : enhancement
4510 Description: 2.6 OST async journal commit and locking fix to improve performance
4511 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
4512 2.6 kernels are now async as they already were in 2.4 kernels so
4513 that they can commit concurrently with the network bulk transfer.
4514 For block-allocated files the filter allocation semaphore is held
4515 to avoid filesystem fragmentation during allocation. BKL lock
4516 removed for 2.6 xattr operations where it is no longer needed.
4521 Description: lconf incorrectly determined whether two IP networks could talk
4522 Details : In some more complicated routing and multiple-network
4523 configurations, lconf will avoid trying to make a network
4524 connection to a disjoint part of the IP space. It was doing the
4525 math incorrectly for one set of cases.
4530 Description: Fix for potential infinite loop processing records in an llog.
4531 Details : If an llog record is corrupted/zeroed, it is possible to loop
4532 forever in llog_process(). Validate the llog record length
4533 and skip the remainder of the block on error.
4536 Frequency : occasional (liblustre only)
4538 Description: liblustre could not open files whose last component is a symlink
4539 Details : sysio_path_walk() would incorrectly pass the open intent to
4540 intermediate path components.
4543 Frequency : rare (liblustre only with non-standard tuning)
4544 Bugzilla : 7201 (7350)
4545 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4546 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4547 after it has been released, only do this for referenced locks
4549 Severity : enhancement
4551 Description: specifying an (invalid) directory default stripe_size of -1
4552 would reset the directory default striping
4553 Details : stripe_size -1 was used internally to signal directory stripe
4554 removal, now use "all default" to signal dir stripe removal
4555 as a directory striping of "all default" is not useful
4558 Frequency : common for large clusters running liblustre clients
4560 Description: doing an ls when liblustre clients are running is slow
4561 Details : sending a glimpse AST to a liblustre client waits for every AST
4562 to time out, as liblustre clients will not respond. Since they
4563 cannot cache data we refresh the OST lock LVB from disk instead.
4565 Severity : enhancement
4567 Description: doing an ls at the same time as file IO can be slow
4568 Details : enqueue and other "small" requests can be blocked behind many
4569 large IO requests. Create a new OST IO portal for non-IO
4570 requests so they can be processed faster.
4573 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4575 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4576 created file with mode 000
4577 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4578 allow an ftruncate() as a non-root user to a file with mode 000.
4579 Lustre now respects this flag to disable mode checks when
4580 truncating a file owned by the user
4583 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4586 Description: Revoking locks from clients that went dead or catatonic might take
4588 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4589 cancellation of such locks instant on servers without waiting for
4590 any reply from clients. Clients drops these locks when cancel
4591 notification from server is received without replying.
4594 Frequency : liblustre-only, when liblustre client dies or becomes busy
4596 Description: Doing ls on Linux clients can take a long time with active
4598 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4599 granting such locks to it in the first place if possible. Locks
4600 are taken by proxy on the OST during the read or write and
4601 dropped immediately afterward. Add connect flags handling, do
4602 not grant locks to liblustre clients for glimpse ASTs.
4604 Severity : enhancement
4606 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4607 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4608 sequential reads and behaves badly with applications doing random
4609 reads. Improve it by reading ahead at least read region, and
4610 avoiding excessive large RPC for small reads.
4612 Severity : enhancement
4614 Description: Creating more than 1000 files for a single job may cause a load
4615 imbalance on the OSTs if there are also a large number of OSTs.
4616 Details : qos_prep_create() uses an OST index reseed value that is an
4617 even multiple of the number of available OSTs so that if the
4618 reseed happens in the middle of the object allocation it will
4619 still utilize the OSTs as uniformly as possible.
4624 Description: OST or MDS may oops in ping_evictor_main()
4625 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
4626 but doesn't restart at beginning of obd_exports_timed list
4629 Severity : enhancement
4631 Description: improve by-nid export eviction on the MDS and OST
4632 Details : allow multiple exports with the same NID to be evicted at one
4633 time without re-searching the exports list.
4636 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
4638 Description: MDS may oops in groups_free()
4639 Details : in rare race conditions a newly allocated group_info struct is
4640 freed again, and this can be NULL. The 2.4 compatibility code
4641 for groups_free() checked for a NULL pointer, but 2.6 did not.
4644 Frequency : common for liblustre clients doing little filesystem IO
4645 Bugzilla : 9352, 7313
4646 Description: server may evict liblustre clients accessing contended locks
4647 Details : if a client is granted a lock or receives a completion AST
4648 with a blocking AST already set it would not reply to the AST
4649 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
4652 Frequency : lfs setstripe, only systems with more than 160 OSTs
4654 Description: unable to set striping with a starting offset beyond OST 160
4655 Details : llapi_create_file() incorrectly limited the starting stripe
4656 index to the maximum single-file stripe count.
4659 Frequency : LDAP users only
4661 Description: lconf did not handle in-kernel recovery with LDAP properly
4662 Details : lconf/LustreDB get_refs() is searching the wrong namespace
4664 Severity : enhancement
4666 Description: bind OST threads to NUMA nodes to improve performance
4667 Details : all OST threads are uniformly bound to CPUs on a single NUMA
4668 node and do their allocations there to localize memory access
4670 Severity : enhancement
4672 Description: llmount can determine client NID directly from Myrinet (GM)
4673 Details : the client NID code from gmnalnid was moved directly into
4674 llmount, removing the need to use this or specifying the
4675 client NID explicitly when mounting GM clients with zeroconf
4678 Frequency : if client is started with down MDS
4680 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
4681 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
4682 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
4683 ptlrpc_import_delayed_req() because we didn't check if the
4684 request was interrupted, and we also didn't break out of the
4685 event loop if there was no timeout
4690 Description: data loss during non-page-aligned writes to a single file from
4691 both multiple nodes and multiple threads on one node at same time
4692 Details : updates to KMS and lsm weren't protected by common lock. Resulting
4693 inconsistency led to false short-reads, that were cached and later
4694 used by ->prepare_write() to fill in partially written page,
4695 leading to data loss.
4698 Frequency : always, if lconf --abort_recovery used
4700 Description: lconf --abort_recovery fails with 'Operation not supported'
4701 Details : lconf was attempting to abort recovery on the MDT device and not
4704 Severity : enhancement
4706 Description: remove cleanup logs
4707 Details : replace lconf-generated cleanup logs with lustre internal
4708 cleanup routines. Eliminates the need for client-cleanup and
4711 Severity : enhancement
4713 Description: add support for EAs (user and system) on lustre filesystems
4714 Details : it is now possible to store extended attributes in the Lustre
4715 client filesystem, and with the user_xattr mount option it
4716 is possible to allow users to store EAs on their files also
4718 Severity : enhancement
4720 Description: Add possibility (config option) to show minimal available OST free
4722 Details : When compiled with --enable-mindf configure option, statfs(2)
4723 (and so, df) will return least minimal free space available from
4724 all OSTs as amount of free space on FS, instead of summary of
4725 free spaces of all OSTs.
4727 Severity : enhancement
4729 Description: do not expand extent locks acquired on OST-side
4730 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
4731 by server: they are not cached anyway.
4734 Frequency : when mmap is used/binaries executed from Lustre
4736 Description: Unmmap pages before throwing them away from read cache.
4737 Details : llap_shrink cache now attempts to unmap pages before discarding
4738 them (if unmapping failed - do not discard). SLES9 kernel has
4739 extra checks that trigger if this unmapping is not done first.
4744 Description: lconf didn't resolve symlinks before checking to see whether a
4745 given mountpoint was already in use
4748 Frequency : when migrating failover services
4749 Bugzilla : 6395, 9514
4750 Description: When migrating a subset of services from a node (e.g. failback
4751 from a failover service node) the remaining services would
4752 time out and evict clients.
4753 Details : lconf --force (implied by --failover) sets the global obd_timeout
4754 to 5 seconds in order to quickly disconnect, but this caused
4755 other RPCs to time out too quickly. Do not change the global
4756 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
4758 Severity : enhancement
4759 Frequency : if MDS is started with down OST
4760 Bugzilla : 9439,5706
4761 Description: Allow startup/shutdown of an MDS without depending on the
4762 availability of the OSTs.
4763 Details : Asynchronously call mds_lov_synchronize during MDS startup.
4764 Add appropriate locking and lov-osc refcounts for safe
4765 cleaning. Add osc abort_inflight calls in case the
4766 synchronize never started.
4769 Frequency : occasional (Cray XT3 only)
4771 Description: root not authorized to access files in CRAY_PORTALS environment
4772 Details : The client process capabilities were not honoured on the MDS in
4773 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
4774 been accessed by an authorized user then root was able to access
4775 the file on the local client also. The root user capabilities
4776 are now allowed on the MDS, as this environment has secure UID.
4779 Frequency : occasional
4781 Description: ldiskfs "too long searching" message happens too often
4782 Details : A debugging message (otherwise harmless) prints too often on
4783 the OST console. This has been reduced to only happen when
4784 there are fragmentation problems on the filesystem.
4789 Description: Division by zero in statfs when all OSCs are inactive
4790 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
4791 lov_statfs divides by value returned by lov_get_stripecnt().
4795 Bugzilla : 9489, 3273
4796 Description: First write from each client to each OST was only 4kB in size,
4797 to initialize client writeback cache, which caused sub-optimal
4798 RPCs and poor layout on disk for the first written file.
4799 Details : Clients now request an initial cache grant at (re)connect time
4800 and so that they can start streaming writes to the cache right
4801 away and always do full-sized RPCs if there is enough data.
4802 If the OST is rebooted the client also re-establishes its grant
4803 so that client cached writes will be honoured under the grant.
4808 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
4809 Details : Now I/O RPCs go to different portal number and (presumably) fast
4810 lock requests (and glimses) and other RPCs get their own service
4811 threads pool that should be able to service those RPCs
4814 Severity : enhancement
4816 Description: Ability to exchange lustre version between client and servers and
4817 issue warnings at client side if client is too old. Also for
4818 liblustre clients there is ability to refuse connection of too old
4820 Details : New 'version' field is added to connect data structure that is
4821 filled with version info. That info is later checked by server and
4825 Frequency : rare, liblustre only.
4826 Bugzilla : 9296, 9581
4827 Description: Two simultaneous writes from liblustre at offset within same page
4828 might proceed at the same time overwriting eachother with stale
4830 Details : I/O lock withing llu_file_prwv was released too early, before data
4831 actually was hitting the wire. Extended lock-holding time until
4832 server acknowledges receiving data.
4835 Frequency : extremely rare. Never observed in practice.
4837 Description: avoid generating lustre_handle cookie of 0.
4838 Details : class_handle_hash() generates handle cookies by incrementing
4839 global counter, and can hit 0 occasionaly (this is unlikely, but
4840 not impossible, because initial value of cookie counter is
4841 selected randonly). Value of 0 is used as a sentinel meaning
4842 "unassigned handle" --- avoid it. Also coalesce two critical
4843 sections in this function into one.
4845 Severity : enhancement
4847 Description: allow liblustre clients to delegate truncate locking to OST
4848 Details : To avoid overhead of locking, liblustre client instructs OST to
4849 take extent lock in ost_punch() on client's behalf. New connection
4850 flag is added to handle backward compatibility.
4852 Severity : enhancement
4853 Bugzilla : 4928, 7341, 9758
4854 Description: allow number of OST service threads to be specified
4855 Details : a module parameter allows the number of OST service threads
4856 to be specified via "options ost ost_num_threads={N}" in the
4857 OSS's /etc/modules.conf or /etc/modprobe.conf.
4861 Bugzilla : 6146, 9635, 9895
4862 Description: servers crash with bad pointer in target_handle_connect()
4863 Details : In rare cases when a client is reconnecting it was possible that
4864 the connection request was the last reference for that export.
4865 We would temporarily drop the export reference and get a new
4866 one, but this may have been the last reference and the export
4867 was just destroyed. Get new reference before dropping old one.
4869 Severity : enhancement
4870 Frequency : if client is started with failover MDS
4872 Description: Allow multiple MDS hostnames in the mount command
4873 Details : Try to read the configuration from all specified MDS
4874 hostnames during a client mount in case the "primary"
4877 Severity : enhancement
4879 Description: Stop sending data to evicted clients as soon as possible.
4880 Details : Check if the client we are about to send or are sending data to
4881 was evicted already. (Check is done every second of waiting,
4882 for which l_wait_event interface was extended to allow checking
4883 of exit condition at specified intervals).
4886 Frequency : rare, normally only when NFS exporting is done from client
4888 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
4890 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
4893 Frequency : always, if srand() is called before liblustre initialization
4895 Description: Liblustre uses system PRNG disturbing its usage by user application
4896 Details : Introduce internal to lustre fast and high-quality PRNG for
4897 lustre usage and make liblustre and some other places in generic
4898 lustre code to use it.
4900 Severity : enhancement
4901 Bugzilla : 9477, 9557, 9870
4902 Description: Verify that the MDS configuration logs are updated when xml is
4903 Details : Check if the .xml configuration logs are newer than the config
4904 logs stored on the MDS and report an error if this is the case.
4905 Request --write-conf, or allow starting with --old_conf.
4907 Severity : enhancement
4909 Description: Handle symlinks in the path when checking if Lustre is mounted.
4910 Details : Resolve intermediate symlinks when checking if a client has
4911 mounted a filesystem to avoid duplicate client mounts.
4916 Description: lconf can hit an error exception but still return success.
4917 Details : The lconf command catches the Command error exception at the top
4918 level script context and will exit with the associated exit
4919 status, but doesn't ensure that this exit status is non-zero.
4924 Description: failure of ptlrpc thread startup can cause oops
4925 Details : Starting a ptlrpc service thread can fail if there are a large
4926 number of threads or the server memory is very fragmented.
4927 Handle this without oopsing.
4930 Frequency : always, only if liblustre and non-default acceptor port was used
4932 Description: liblustre cannot connect to servers with non-default acceptor port
4933 Details : tcpnal_set_default_params() was not called and was therefore
4934 ignoring the environment varaible TCPNAL_PORT, as well as other
4935 TCPNAL_ environment variables
4940 Description: two objects could be created on the same OST for a single file
4941 Details : If an OST is down, in some cases it was possible to create two
4942 objects on a single OST for a single file. No problems other
4943 than potential performance impact and spurious error messages.
4947 Bugzilla : 5681, 9562
4948 Description: Client may oops in ll_unhash_aliases
4949 Details : Client dcache may become inconsistent in race condition.
4950 In some cases "getcwd" can fail if the current directory is
4956 Description: Inode refcounting problems in NFS export code
4957 Details : link_raw functions used to call d_instantiate without obtaining
4958 extra inode reference first.
4962 Bugzilla : 9942, 9903
4963 Description: Referencing freed requests leading to crash, memleaks with NFS.
4964 Details : We used to require that call to ll_revalidate_it was always
4965 followed by ll_lookup_it. Also with revalidate_special() it is
4966 possible to call ll_revalidate_it() twice for the same dentry
4967 even if first occurence returned success. This fix changes semantic
4968 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
4969 reference on a request referred from the intent.
4970 ll_intent_release() then releases such a request.
4973 Frequency : rare, normally benchmark loads only
4975 Description: unlinked inodes were kept in memory on the client
4976 Details : If a client is repeatedly creating and unlinking files it
4977 can accumulate a lot of stale inodes in the inode slab cache.
4978 If there is no other client load running this can cause the
4979 client node to run out of memory. Instead flush old inodes
4980 from client cache that have the same inode number as a new inode.
4983 Frequency : SLES9 2.6.5 kernel and long filenames only
4984 Bugzilla : 9969, 10379
4985 Description: utime reports stale NFS file handle
4986 Details : SLES9 uses out-of-dentry names in some cases, which confused
4987 the lustre dentry revalidation. Change it to always use the
4991 Frequency : rare, unless heavy write-truncate concurrency is continuous
4992 Bugzilla : 4180, 6984, 7171, 9963, 9331
4993 Description: OST becomes very slow and/or deadlocked during object unlink
4994 Details : filter_destroy() was holding onto the parent directory lock
4995 while truncating+unlinking objects. For very large objects this
4996 may block other threads for a long time and slow overall OST
4997 responsiveness. It may also be possible to get a lock ordering
4998 deadlock in this case, or run out of journal credits because of
4999 the combined truncate+unlink. Solution is to do object truncate
5000 first in one transaction without parent lock, and then do the
5001 final unlink in a new transaction with the parent lock. This
5002 reduces the lock hold time dramatically.
5005 Frequency : rare, 2.4 kernels only
5007 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
5008 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
5009 before unmounting their filesystem in order to avoid deadlock.
5010 The kernel_locked() function in 2.4 kernels only checks whether
5011 the kernel lock is held, not whether it is this process that is
5012 holding it as 2.6 kernels do.
5017 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
5018 Details : The client ptlrpc code may be trying to reconnect to a down
5019 server before a previous connection attempt has timed out.
5020 Increase the reconnect interval to be longer than the connection
5021 timeout interval to avoid sending duplicate connections to
5025 Frequency : echo_client brw_test command
5027 Description: fix echo_client to work with OST preallocated code
5028 Details : OST preallocation code (5137) didn't take echo_client IO path
5029 into account: echo_client calls filter methods outside of any
5030 OST thread and, hence, there is no per-thread preallocated
5031 pages and buffers to use. Solution: hijack pga pages for IO. As
5032 a byproduct, this avoids unnecessary data copying.
5036 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
5037 Description: Client can oops in mdc_commit_close() after open replay
5038 Details : It was possible for the MDS to return an open request with no
5039 transaction number in mds_finish_transno() if the client was
5040 evicted, but without actually returning an error. Clients
5041 would later try to replay that open and may trip an assertion
5042 Simplify the client close codepath, and always return an error
5043 from the MDS in case the open is not successful.
5046 Frequency : rare, 2.6 OSTs only
5048 Description: OST may deadlock under high load on fragmented files
5049 Details : If there was a heavy load and highly-fragmented OST filesystems
5050 it was possible to have all the OST threads deadlock waiting on
5051 allocation of biovecs, because the biovecs were not released
5052 until the entire RPC IO was completed. Instead, release biovecs
5053 as soon as they are complete to ensure forward IO progress.
5055 Severity : enhancement
5057 Description: Support for specifying external journal device at mount
5058 Details : If an OST or MDS device is formatted with an external journal
5059 device, this device major/minor is stored in the ext3 superblock
5060 and may not be valid for failover. Allow detecting and
5061 specifying the external journal at mount time.
5066 Description: Mounting an MDS with pending unlinked files may cause oops
5067 Details : target_finish_recovery() calls mds_postrecov() which returned
5068 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
5069 considers this an error and immediately begins cleaning up the
5070 lov, just after starting the mds_lov process
5072 Severity : enhancement
5074 Description: Implement 'lfs df' to report actual free space on per-OST basis
5075 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
5076 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
5077 usage of objects; '-h' to report in human readable format.
5079 ------------------------------------------------------------------------------
5081 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
5088 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
5089 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
5090 at startup during orphan unlinking. Instead, skip buddy update
5091 before setup, as it will be regenerated after recovery is complete.
5096 Description: If an OST is inactive, its locks might reference stale inodes.
5097 Details : lov_change_cbdata() must iterate over all namespaces, even if
5098 they are inactive to clear inode references from the lock.
5100 Severity : enhancement
5101 Frequency : occasional, if non-standard max_dirty_mb used
5103 Description: Client will block write RPCs if not enough grant
5104 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
5105 then the client will block writes while waiting for another RPC
5106 to complete instead of consuming its dirty limit. With change
5107 we get improved performance when max_dirty_mb is small.
5109 Severity : enhancement
5110 Bugzilla : 3389, 6253
5111 Description: Add support for supplementary groups on the MDS.
5112 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
5113 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
5114 lookups for user supplementary groups into a cache.
5118 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
5119 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
5120 for liblustre O_EXCL is enforced by MDS.
5124 Description: readdir never returns NULL in liblustre.
5125 Details : Corrected llu_iop_getdirentries logic, to return offset of next
5126 dentry in struct dirent.
5130 Frequency : liblustre only, depends on application IO pattern
5131 Description: liblustre clients evicted if not contacting servers
5132 Details : Don't put liblustre clients into the ping_evictor list, so
5133 they will not be evicted by the pinger ever.
5135 Severity : enhancement
5137 Description: Add ability to evict clients by NID from MDS.
5138 Details : By echoing "nid:$NID" string into
5139 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
5140 $NID would be instantly evicted from this MDS and from all active
5141 OSTs connected to it.
5145 Description: Do not query file size twice, somewhat slowing stat(2) calls.
5146 Details : lookup_it_finish() used to query file size from OSTs that was not
5151 Description: service threads change working directory to that of init
5152 Details : Starting lustre service threads may pin the working directory
5153 of the parent thread, making that filesystem busy. Threads
5154 now change to the working directory of init to avoid this.
5158 Frequency : during shutdown only
5159 Description: shutdown with a failed MDS or OST can cause unmount to hang
5160 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
5165 Frequency : 2.6 only
5166 Description: chmod/chown may include an extra supplementary group
5167 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
5168 supplementary group and if none is specified this is used.
5171 Bugzilla : 5479 (6816)
5173 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
5174 Details : If lookup is in progress on a file that is unlinked we might try
5175 to revalidate the inode and fail in revalidate after lookup is
5176 complete and ll_file_open() enqueues the open again but
5177 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
5180 Frequency : always, if lconf --abort_recovery used
5182 Description: lconf --abort_recovery fails with 'Operation not supported'
5183 Details : lconf was attempting to abort recovery on the MDT device and not
5186 ------------------------------------------------------------------------------
5188 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
5193 Frequency : rare (only unsupported configurations with a node running as an
5195 Bugzilla : 6514, 5137
5196 Description: Mounting a Lustre file system on a node running as an OST could
5198 Details : OSTs now preallocates memory needed to write out data at
5199 startup, instead of when needed, to avoid having to
5200 allocate memory in possibly low memory situations.
5201 Specifically, if the file system is mounted on on OST,
5202 memory pressure could force it to try to write out data,
5203 which it needed to allocate memory to do. Due to the low
5204 memory, it would be unable to do so and the node would
5205 become unresponsive.
5207 Severity : enhancement
5209 Description: Addition of lconf --service command line option
5210 Details : lconf now accepts a '--service <arg>' option, which is
5211 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
5213 Severity : enhancement
5215 Description: Failover mode is now the default for OSTs.
5216 Details : By default, OSTs will now run in failover mode. To return to
5217 the old behaviour, add '--failout' to the lmc line for OSTs.
5219 Severity : enhancement
5221 Description: Health checks are now provided for MDS and OSTs
5222 Details : Additional detailed health check information on MSD and OSTs
5223 is now provided through the procfs health_check value.
5226 Frequency : occasional, depends on IO load
5228 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
5229 after numerous create/delete cycles
5230 Details : The ext3 inode allocation policy would not allocate new inodes
5231 very well on the OSTs because there are no new directories
5232 being created. Instead we look for groups with free space if
5233 the parent directories are nearly full.
5238 Description: Network or server problems during mount may cause partially
5239 mounted clients instead of returning an error.
5240 Details : The config llog parsing code may overwrite the error return
5241 code during mount error handling, returning success instead
5247 Description: MDS can fail to allocate large reply buffers
5248 Details : After long uptimes the MDS can fail to allocate large reply
5249 buffers (e.g. zconf client mount config records) due to memory
5250 fragmentation or consumption by the buffer cache. Preallocate
5251 some large reply buffers so that these replies can be sent even
5252 under memory pressure.
5256 Frequency : rare (liblustre)
5257 Description: fsx running with liblustre complained that using truncate() to
5258 extend the file doesn't work. This patch corrects that issue.
5259 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
5260 ATTR_SIZE and lsm use in llu_setattr_raw.
5264 Frequency : rare, only 2.6 kernels
5265 Description: Unusual file access patterns on the MDS may result in inode
5266 data being lost in very rare circumstances.
5267 Details : Bad interaction between the ea-in-inode patch and the "no-read"
5268 code in the 2.6 kernel caused the inode and/or EA data not to
5269 be read from disk, causing single-file corruption.
5273 Frequency : rare, only 2.6 filesystems using extents
5274 Description: Heavy concurrent write and delete load may cause data corruption.
5275 Details : It was possible under high-load situations to have an extent
5276 metadata block in the block device cache from a just-unlinked
5277 file overwrite a newly-allocated data block. We now unmap any
5278 metadata buffers that alias just-allocated data blocks.
5282 Frequency : filesystems with default stripe_count larger than 77
5283 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
5284 Details : lconf specifies an inode size of 4096 bytes when the default
5285 stripe_count is larger than 77. This conflicts with the default
5286 inode density of 1 per 4096 bytes. Allocate smaller inodes in
5287 this case to avoid pinning too much memory for large EAs.
5289 ------------------------------------------------------------------------------
5291 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
5296 Frequency : rare (extremely heavy IO load with hundreds of clients)
5298 Description: Client is evicted, gets IO error writing to file
5299 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
5300 caused clients to be evicted for AST timeouts. The fixes in
5301 bug 5192 mean we no longer need to have such short AST timeouts
5302 so ldlm_timeout has been increased.
5305 Frequency : occasional during --force or --failover shutdown under load
5306 Bugzilla : 5949, 4834
5307 Description: Server oops/LBUG if stopped with --force or --failover under load
5308 Details : a collection of import/export refcount and cleanup ordering
5309 issues fixed for safer force cleanup
5312 Frequency : only filesystems larger than 120 OSTs
5313 Bugzilla : 5990, 6223
5314 Description: lfs getstripe would oops on a very large filesystem
5315 Details : lov_getconfig used kfree on vmalloc'd memory
5318 Frequency : only filesystems exporting via NFS to Solaris 10 clients
5319 Bugzilla : 6242, 6243
5320 Description: reading from files that had been truncated to a non-zero size
5321 but never opened returned no data
5322 Details : ll_file_read() reads zeros from no-object files to EOF
5327 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
5328 Details : There's little harm in aborting MDS/OSS recovery and letting it
5329 try again, so I removed the LASSERT and return an error instead.
5331 Severity : enhancement
5333 Description: New debugging infrastructure for tracking down data corruption
5334 Details : The I/O checksum code was replaced to: (a) control it at runtime,
5335 (b) cover more of the client-side code path, and (c) try to narrow
5336 down where problems occurred
5340 Bugzilla : 3819, 4364, 4397, 6313
5341 Description: Racing close and eviction MDS could cause assertion in mds_close
5342 Details : It was possible to get multiple mfd references during close and
5343 client eviction, leading to one thread referencing a freed mfd.
5345 Severity: : enhancement
5346 Bugzilla : 3262, 6359
5347 Description: Attempts to reconnect to servers are now more aggressive.
5348 Details : This builds on the enhanced upcall-less recovery that was added
5349 in 1.4.2. When trying to reconnect to servers, clients will
5350 now try each server in the failover group every 10 seconds. By
5351 default, clients would previously try one server every 25 seconds.
5356 Description: After recovery, certain operations trigger a failed
5357 assertion on a client.
5358 Details : Failing over an mds, using lconf -d --failover, while a
5359 client was doing a readdir() call would cause the client to
5360 LBUG after recovery completed and the readdir() was resent.
5362 Severity : enhancement
5364 Description: Default groups are now added by lconf
5365 Details : You can now run lconf --group <servicename> without having to
5366 manually add groups with lmc.
5369 Frequency : occasional
5371 Description: Nodes with an elan id of 0 trigger a failed assertion
5374 Frequency : always when accessing e.g. tty/console device nodes
5376 Description: tty and some other devices nodes cannot be used on lustre
5377 Details : file's private_data field is used by device data and lustre
5378 values in there got lost. New field was added to struct file to
5379 store fs-specific private data.
5382 Frequency : when exporting Lustre via NFS
5384 Description: NFSD failed occasionally when looking up a path component
5385 Details : NFSD is looking up ".." which was broken in ext3 directories
5386 that had grown large enough to become hashed.
5389 Frequency : Clusters with multiple interfaces not on the same subnet
5391 Description: Nodes will repeatedly try to reconnect to an interface which it
5392 cannot reach and report an error to the log.
5393 Details : Extra peer list entries will be created by lconf with some peers
5394 unreachable. lconf now validates the peer before adding it.
5397 Frequency : Only if a default stripe is set on the filesystem root.
5399 Description: Setting a default stripe on the filesystem root prevented the
5400 filesystem from being remounted.
5401 Details : The client was sending extra request flags in the root getattr
5402 request and did not allocate a reply buffer for the dir EA.
5405 Frequency : occasional, higher if lots of files are accessed by one client
5406 Bugzilla : 6159, 6097
5407 Description: Client trips assertion regarding lsm mismatch/magic
5408 Details : While revalidating inodes the VFS looks up inodes with ifind()
5409 and in rare cases can find an inode that is being freed.
5410 The ll_test_inode() code will free the lsm during ifind()
5411 when it finds an existing inode and then the VFS later attaches
5412 this free lsm to a new inode.
5416 Bugzilla : 6422, 7030
5417 Description: MDS deadlock between mkdir and client eviction
5418 Details : Creating a new file via mkdir or mknod (starting a transaction
5419 and getting the ns lock) can deadlock with client eviction
5420 (gets ns lock and trying to finish a synchronous transaction).
5423 Frequency : occasional
5424 Description: While starting a server, the fsfilt_ext3 module could not be
5426 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
5427 kernels. Previously, lconf would still use the ext3 name
5428 when trying to load modules. Now, it will correctly use
5429 ext3 on 2.4 and ldiskfs on 2.6.
5431 Severity : enhancement
5432 Description: The default stripe count has been changed to 1
5433 Details : The interpretation of the default stripe count (0, to lfs
5434 or lmc) has been changed to mean striping across a single
5435 OST, rather than all available. For general usage we have
5436 found a stripe count of 1 or 2 works best.
5438 Severity : enhancement
5439 Description: Add support for compiling against Cray portals.
5440 Details : Conditional compiling for some areas that are different
5444 Frequency : occasional
5445 Bugzilla : 6409, 6834
5446 Description: Creating files with an explicit stripe count may lead to
5447 a failed assertion on the MDS
5448 Details : If some OSTs are full or unavailable, creating files may
5449 trigger a failed assertion on the MDS. Now, Lustre will
5450 try to use other servers or return an error to the
5454 Frequency : occasional
5456 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
5457 caused lock timeout during readahead (since 1.4.2).
5458 Details : Processes doing readahead might match a lock that hasn't been
5459 granted yet if there are overlapping and conflicting lock
5460 requests. The readahead process waits on ungranted lock
5461 (original lock is CBPENDING), while OST waits for that process
5462 to cancel CBPENDING read lock and eventually evicts client.
5464 Severity : enhancement
5466 Description: Initial enabling of flock support for clients
5467 Details : Implements fcntl advisory locking and file status functions.
5468 This feature is provided as an optional mount flag (default
5469 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
5470 locking are implemented yet, and those that are are not guaranteed
5471 to be completely correct in production environments.
5472 mount -t lustre -o [flock|noflock] ...
5475 Frequency : occasional
5477 Description: OSTs running 2.4 kernels but with extents enabled might trip an
5478 assertion in the ext3 JBD (journaling) layer.
5479 Details : The b_committed_data struct is protected by the big kernel lock
5480 in 2.4 kernels, serializing journal_commit_transaction() and
5481 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
5482 access to this struct. In 2.6 kernels there is finer grained
5483 locking to improve SMP performance of the JBD layer.
5487 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
5492 Description: merge and cleanup kernel patches.
5495 -----------------------------------------------------------------------------
5497 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
5499 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
5500 versions of Lustre. Please update all servers and clients to
5501 version 1.4.2 or later at the same time. You must also run
5502 "lconf --write-conf {config}.xml" on the MDS while it is stopped
5503 to update the configuration logs.
5505 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
5506 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
5507 - swab llog records of type '0' so we get proper header size/idx (5861)
5508 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
5509 - fix rename of one directory over another leaking an inode (5953)
5510 - avoid SetPageDirty on 2.6 (5981)
5511 - don't re-add just-being-destroyed locks to the waiting list (5653)
5512 - when creating new directories, inherit the parent's custom
5513 striping settings if present parent (3048)
5514 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
5515 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
5516 - don't hold client locks on temporary worklist from l_lru (5666)
5517 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
5518 - automatically evict dead clients (5921)
5519 - Update file size properly in create+truncate+fstat case (6196)
5520 - Do not unhash mountpoint dentries, do not allow removal of
5522 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
5523 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
5524 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
5525 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
5526 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
5527 - Correctly handle reads of files with no objects (6243)
5528 - lctl recover will also mark a device active if deactivate used (5933)
5530 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
5531 - allow --write-conf on an MDS with different nettype than client (5619)
5532 - don't write config llogs to MDS for mounts not from that MDS (5617)
5533 - lconf should create multiple TCP connections from a client (5201)
5534 - init scripts are now turned off by default; run chkconfig --on
5535 lustre and chkconfig --on lustrefs to use them
5536 - upcalls are no longer needed for clients to recover to failover
5538 - add --abort-recovery option to lconf to abort recovery on device
5540 - add support for an arbitrary number of OSTs (3026)
5541 - Quota support protocol changes.
5542 - forward compatibility changes to wire structs (6007)
5543 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5544 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5545 - improved "lustre status" script
5546 - initialize blocksize for non-regular files (6062)
5547 - added --disable-server and --disable-client configure options (5782)
5548 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5549 - Vanilla 2.4.29 support
5550 - increase maximum number of obd devices to 520 (6242)
5551 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5552 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5553 SLES 9 2.6.5 kernels
5554 - limit stripes per file to 160 (the maximum EA size) (6093)
5556 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5559 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5560 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5561 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5562 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5563 - small ext3 extents cleanups and fixes (5733)
5564 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5565 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5566 - avoid needless client->OST connect, fix handle mismatch (5317)
5567 - fix DLM error path that led to out-of-sync client, long delays (5779)
5568 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5569 - fix several locking issues related to i_size (5492,5624,5654,5672)
5570 - don't move pending lock onto export if it is already evicted (5683)
5571 - fix kernel oops when creating .foo in unlinked directory (5548)
5572 - fix deadlock in obdfilter statistics vs. object create (5811)
5573 - use time_{before,after} to avoid timer jiffies wrap (5882)
5574 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5575 - Do not leak request if server was not able to process it (5154)
5576 - If mds_open unable to find parent dir, make that negative lookup(5154)
5577 - don't create new directories with extent-mapping (5909, 5936)
5579 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5580 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5581 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5582 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5583 - support for RHEL4 kernel on i686 (5773)
5584 - provide error messages when incompatible logs are encountered (5898)
5586 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5587 * version 1.4.0.10 (1.4.1 release candidate 1)
5589 - don't keep a lock reference when lock is not granted (4238)
5590 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5591 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5592 - drop import inflight refcount on signal_completed_replay error (5255)
5593 - unlock page after async write error during send (3677)
5594 - handle missing objects in filter_preprw_read properly (5265)
5595 - no transno return for symlink open, don't save no-trasno open (3440)
5596 - don't try to complete elan receive that already failed (4012)
5597 - free RPC server reply state on error (5406)
5598 - clean up thread from ptlrpc_start_thread() on error (5160)
5599 - readahead could read extra page into cache that wasn't ejected (5388)
5600 - prevent races in class_attach/setup/cleanup/detach (5260)
5601 - don't dereference de->d_inode after l_dput of de (5458)
5602 - use "int" for stripe value returned from lock_to_stripe (5544)
5603 - mballoc allocation and error-checking fixes in 2.6 (5504)
5604 - block device patches to fix I/O request sizes in 2.6 (5482)
5605 - look up hostnames for IB nals (5602)
5606 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5607 - don't start multiple acceptors for the same port (5277)
5608 - fix incorrect LASSERT in mds_getattr_name (5635)
5609 - export a proc file for general "ping" checking (5628)
5610 - fix "lfs check" to not block when the MDS is down (5628)
5612 - service request history (4965)
5613 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5614 - create an "evict_client" /proc entry on OSTs, like the MDS has
5615 - fix mount usage message, return errors per mount(8) (5168)
5616 - change grep [] to grep "[]" in tests so they work in more UMLs
5617 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5618 - remove ancient LOV_MAGIC_V0 EA support (5047)
5619 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5620 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5621 - allow readahead allocations to fail when low on memory (5383)
5622 - mmap locking landed again, after considerable improvement (2828)
5623 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
5625 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
5628 - send OST transaction number in read/write reply to free req (4966)
5629 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5630 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5631 - the watchdog thread now runs as interruptible (5246)
5632 - flock/lockf fixes (but it's still disabled, pending 5135)
5633 - don't use EXT3 constants in llite code (5094)
5634 - memory shortage at startup could cause assertion (5176)
5636 - reorganization of lov code
5637 - single portals codebase
5639 - add extents/mballoc support (5025)
5640 - direct I/O reads in the obdfilter (4048)
5641 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
5643 tbd Cluster File Systems, Inc. <info@clusterfs.com>
5646 - send OST transaction number in read/write reply to free req (4966)
5647 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5648 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5649 - the watchdog thread now runs as interruptible (5246)
5650 - handle missing objects in filter_preprw_read properly (5265)
5651 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5652 - drop import inflight refcount on signal_completed_replay error (5255)
5653 - unlock page after async write error during send (3677)
5654 - return original error code on reconstructed replies (3761)
5655 - no transno return for symlink open, don't save no-trasno open (3440)
5657 - add pid to ldlm debugging output (4922)
5658 - bump the watchdog timeouts -- we can't handle 30sec yet
5659 - extra debugging for orphan dentry/inode bug (5259)
5661 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
5664 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
5665 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
5666 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
5667 - fix NULL dereference in /proc/sys/portals/routes (4827)
5668 - allow failed mdc_close() operations to be interrupted (4561)
5669 - stop precreate on OST before MDS would time out on it (4778)
5670 - don't send partial-page writes before EOF from client (4410)
5671 - discard client grant for sub-page writes on large-page clients (4520)
5672 - don't free dentries not owned by NFS code, check generation (4806)
5673 - fix lsm leak if mds_create_objects() fails (4801)
5674 - limit debug_daemon file size, always print CERROR messages (4789)
5675 - use transno after validating reply (3892)
5676 - process timed out requests if import state changes (3754)
5677 - update mtime on OST during writes, return in glimpse (4829)
5678 - add mkfsoptions to LDAP (4679)
5679 - use ->max_readahead method instead of zapping global ra (5039)
5680 - don't interrupt __l_wait_event() during strace
5682 - add software watchdogs to catch hung threads quickly (4941)
5683 - make lustrefs init script start after nfs is mounted
5684 - fix CWARN/ERROR duplication (4930)
5685 - return async write errors to application if possible (2248)
5686 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
5687 - print NAL number in %x format (4645)
5688 - update barely-supported suse-2.4.21-171 series (4842)
5689 - support for sles 9 %post scripts
5690 - support for building 2.6 kernel-source packages
5691 - support for sles km_* packages
5693 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
5696 - ignore -ENOENT errors in osc_destroy (3639)
5697 - notify osc create thread that OSC is being cleaned up (4600)
5698 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
5699 - reconstruct ost_handle() like mds_handle() (4657)
5700 - create a new thread to do import eviction to avoid deadlock (3969)
5701 - let lconf resolve symlinked-to devices (4629)
5702 - don't unlink "objects" from directory with default EA (4554)
5703 - hold socknal file ref over connect in case target is down (4394)
5704 - allow more than 32000 subdirectories in a single directory (3244)
5705 - fix blocks count for O_DIRECT writes (3751)
5706 - OST returns ENOSPC from object create when no space left (4539)
5707 - don't send truncate RPC if file size isn't changing (4410)
5708 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
5709 - bind to privileged port in socknal and tcpnal (3689)
5711 - rate limit CERROR/CWARN console message to avoid overload (4519)
5712 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
5713 - basic mmap support (3918)
5714 - kernel patch series update from b1_4 (4711)
5716 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
5719 - avoid crash during MDS cleanup with OST shut down (2775)
5720 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
5721 - don't use bad inodes on the MDS (3744)
5722 - dynamic object preallocation to improve recovery speed (4236)
5723 - don't hold spinlock over lock dumping or change debug flags (4401)
5724 - don't zero obd_dev when it is force cleaned (3651)
5725 - print grants to console if they go negative (4431)
5726 - "lctl deactivate" will stop automatic recovery attempts (3406)
5727 - look for existing locks in ldlm_handle_enqueue() (3764)
5728 - don't resolve lock handle twice in recovery avoiding race (4401)
5729 - revalidate should check working dir is a directory (4134)
5731 - don't always mark "slow" obdfilter messages as errors (4418)
5733 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
5736 - don't close LustreDB during write_conf until it is done (3860)
5737 - fix typo in lconf for_each_profile (3821)
5738 - allow dumping logs from multiple threads at one time (3820)
5739 - don't allow multiple threads in OSC recovery (3812)
5740 - fix debug_size parameters (3864)
5741 - fix mds_postrecov to initialize import for llog ctxt (3121)
5742 - replace config semaphore with spinlock (3306)
5743 - be sure to send a reply for a CANCEL rpc with bad export (3863)
5744 - don't allow enqueue to complete on a destroyed export (3822)
5745 - down write_lock before checking llog header bitmap (3825)
5746 - recover from lock replay timeout (3764)
5747 - up llog sem before sending rpc (3652)
5748 - reduce ns lock hold times when setting kms (3267)
5749 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
5750 - fix NULL deref and obd_dev leak on setup error (3312)
5751 - replace some LBUG about llog ops with error handling (3841)
5752 - don't match INVALID dentries from d_lookup and spin (3784)
5753 - hold dcache_lock while marking dentries INVALID and hashing (4255)
5754 - fix invalid assertion in ptlrpc_set_wait (3880)
5756 - add libwrap support for the TCP acceptor (3996)
5757 - add /proc/sys/portals/routes for non-root route listing (3994)
5758 - allow setting MDS UUID in .xml (2580)
5759 - print the stack of a process that LBUGs (4228)
5761 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
5764 - don't cleanup request in ll_file_open() on failed MDS open (3430)
5765 - make sure to unset replay flag from failed open requests (3440)
5766 - if default stripe count is 0, use OST count for inode size (3636)
5767 - update parent mtime/ctime on client for create/unlink (2611)
5768 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
5769 - free recovery state on server during a forced cleanup (3571)
5770 - unregister_reply for resent reqs (3063)
5771 - loop back devices mounting and status check on 2.6 (3563)
5772 - fix resource-creation race that can provoke i_size == 0 (3513)
5773 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
5774 - more debugging for page-accounting assertion (3746)
5775 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
5776 - avoid deadlock after precreation failure (3758)
5777 - fix race and lock order deadlock in orphan handling (3450, 3750)
5778 - add validity checks when grabbing inodes from l_ast_data (3599)
5780 - add /proc/.../recovery_status to obdfilter (3428)
5781 - lightweight CDEBUG infrastructure, debug daemon (3668)
5782 - change default OSC RPC parameters to be better on small clusters
5783 - turn off OST read cache for files smaller than 32MB
5784 - install man pages and include them in rpms (3100)
5785 - add new init script for (un)mounting lustre filesystems (2593)
5786 - run chkconfig in %post for init scripts (3701)
5787 - drop scimac NAL (unmaintained)
5789 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
5792 - clean kiobufs before and after use (3485)
5793 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
5794 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
5795 - change init script to fail more clearly if not run as root (1528)
5796 - allow clients to reconnect during replay (1742)
5797 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
5798 - don't do DNS lookups on NIDs too small for IP addresses (3442)
5799 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
5800 - fix cond_resched (3554)
5801 - only evict unfinished clients after recovery (3515)
5802 - allow bulk resend, prevent data loss (3570)
5803 - dynamic ptlrpc request buffer allocation (2102)
5804 - don't allow unlinking open directory if it isn't empty (2904)
5805 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
5806 - remove extraneous obd dereference causing LASSERT failure (3334)
5807 - don't use get_cycles() when creating temp. files on the mds (3156)
5808 - hold i_sem when setting i_size in ll_extent_lock() (3564)
5809 - handle EEXIST for set-stripe, set proper directory name (3336)
5811 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
5812 - fix ksocknal_fmb_callback() error messages (2918)
5814 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
5817 - don't copy lvb into (possibly NULL) reply on error (2983)
5818 - don't deref dentry after dput, don't free lvb on error (2922)
5819 - use the kms to determine writeback rpc length (2947)
5820 - increment oti_logcookies when osc is inactive (2948)
5821 - update client's i_blocks count via lvb messages (2543)
5822 - handle intent open/close of special files properly (1557)
5823 - mount MDS with errors=remount-ro, like obdfilter (2009)
5824 - initialize lock handle to avoid ASSERT on error cleanup (3057)
5825 - don't use cancelling-locks' kms values (2947)
5826 - use highest lock extent for kms, not last one (2925)
5827 - don't dereference ERR_PTR() dentry in error handling path (3107)
5828 - fix thread race in portals_debug_dumplog() (3122)
5829 - create lprocfs device entries at setup instead of at attach (1519)
5830 - common AST error handler, don't evict client on completion race (3145)
5831 - zero nameidata in detach_mnt in 2.6 (3118)
5832 - verify d_inode after revalidate_special is valid in 2.6 (3116)
5833 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
5834 - initialize RPC timeout timer earlier for 2.6 (3219)
5835 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
5836 - print nal/nid for unknown nid (3258)
5837 - additional checks for oscc recovery before doing precreate (3284)
5838 - fix ll_extent_lock() error return code for 64-bit systems (3043)
5839 - don't crash in mdc_close for bad permissions on open (3285)
5840 - zero i_rdev for non-device files (3147)
5841 - clear page->private before handing to FS, better assertion (3119)
5842 - tune the read pipeline (3236)
5843 - fix incorrect decref of invalidated dentry (2350)
5844 - provide read-ahead stats and refine rpc in flight stats (3328)
5845 - don't hold journal transaction open across create RPC (3313)
5846 - update atime on MDS at close time (3265)
5847 - close LDAP connection when recovering to avoid server load (3315)
5848 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
5849 - don't leak open file on MDS after open resend (3325)
5850 - serialize filter_precreate and filter_destroy_precreated (3329)
5851 - loop device shouldn't call sync_dev() for nul device (3092)
5852 - clear page cache after eviction (2766)
5853 - resynchronize MDS->OST in background (2824)
5854 - refuse to mount the same filesystem twice on same mountpoint (3394)
5855 - allow llmount to create routes for mounting behind routers (3320)
5856 - push lock cancellation to blocking thread for glimpse ASTs (3409)
5857 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
5858 - fix rare problem with rename on htree directories (3417)
5860 - allow default OST striping configuration per directory (1414)
5861 - fix compilation for qswnal for 2.6 kernels (3125)
5862 - increase maximum number of MDS request buffers for large systems
5863 - change liblustreapi to be useful for external progs like lfsck (3098)
5864 - increase local configuration timeout for slow disks (3353)
5865 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
5867 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5870 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
5871 - don't overwrite extent policy data in reply if lock was blocked (2901)
5872 - drop filter export grants atomically with removal from device (2663)
5873 - del obd_self_export from work_list in class_disconnect_exports (2908)
5874 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
5875 - swab reply message in mdc_close, other PPC fixes (2464)
5876 - fix destroying of named logs (2325)
5877 - overwrite old logs when running lconf --write_conf (2264)
5878 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
5879 - fix race in target_handle_connect (2898)
5880 - mds_reint_create() should take same inode create lock (2926)
5881 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
5882 - don't close files for self_export to avoid uninitialized obd (2936)
5883 - allow MDS with the same name as client node (2939)
5884 - hold dentry reference for closed log files for unlink (2325)
5885 - reserve space for all logs during transactions (2059)
5886 - don't evict page beyond end of stripe extent (2925)
5887 - don't oops on a deleted current working directory (2399)
5888 - handle hard links to targets without a parent properly (2517)
5889 - don't dereference NULL lock when racing during eviction (2867)
5890 - don't grow lock extents when lots of conflicting locks (2919)
5892 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
5895 - account for cache space usage on clients to avoid data loss (974)
5896 - lfsck support in lustre kernel code (2349)
5897 - reduce journal credits needed for BRW writes (2370)
5898 - orphan handling to avoid losing space on client/server crashes
5899 - ptlrpcd can be blocked, stopping ALL progress (2477)
5900 - use lock value blocks to assist in proper KMS, faster stat (1021)
5901 - takes i_sem instead of DLM locks internally on obdfilter (2720)
5902 - recovery for initial connections (2355)
5903 - fixes for mds_cleanup_orphans (1934)
5904 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
5905 - block all file creations until orphan recovery completes (1901)
5906 - client remove rq_connection from request struct (2423)
5907 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
5908 - recovery timer race (2670)
5909 - mdc_close recovey bug (2532)
5910 - ptlrpc cleanup bug (2710)
5911 - mds timeout on local locks (2588)
5912 - namespace lock held during RPCs (2431)
5913 - handle interrupted sync write properly (2503)
5914 - don't try to handle a message that hasn't been replied to (2699)
5915 - client assert failure during cleanup after abort recovery (2701)
5916 - leak mdc device after failed mount (2712)
5917 - ptlrpc_check_set allows timedout requests to complete (2714)
5918 - wait for inflight reqs when ptlrpcd finishes (2710)
5919 - make sure unregistered services are removed from the srv_list
5920 - reset bulk XID's when resending them (caught by 1138 test)
5921 - unregister_bulk after timeout
5922 - fix lconf error (2694)
5923 - handle write after unfinished setstripe, stripe-only getstripe (2388)
5924 - readahead locks pages, leaves pending causing memory pressure (2673)
5925 - increase OST request buffers to 4096 on large machines (2729)
5926 - fix up permission of existing directories in simple_mkdir (2661)
5927 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
5928 - don't assign transno to errored transactions (2742)
5929 - don't delete objects on OST if given a bogus objid from MDS (2751)
5930 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
5931 - if rq_no_resend, then timeout request after recovery (2432)
5932 - fix MDS llog_logid record size, 64-bit array alignment (2733)
5933 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
5934 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
5935 - MDS recovery shouldn't delete objects that it has given out (2730)
5936 - if enqueue arrives after completion, don't clobber LVB (2819)
5937 - don't unlock pages twice when trigger_group_io returns error (2814)
5938 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
5939 - don't write pages to disk if there was an error (1450)
5940 - don't ping imports that have recovery disabled (2676)
5941 - take buffered bytes into account when balancing socknal conn (2817)
5942 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
5943 - reconnect unlink llog connection after MDS reconnects to OST (2816)
5944 - remove little-endian swabbing of llog records (1987)
5945 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
5946 - retry reposting request buffers if they fail (1191)
5947 - grow extent at grant time to avoid granting a revoked lock (2809)
5948 - lock revoke doesn't evict page if covered by a second lock (2765)
5949 - disable VM readahead to avoid reading outside lock extents (2805)
5951 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
5952 - updated kernel patches for hp-2.4.20 kernel (2681)
5954 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
5957 - fix truncated write corruption (2366)
5958 - fix for failed assertion in iopen_connect_dentry (1792,2517)
5960 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
5961 - ensure len(uuid) < 37 in lmc (1171)
5962 - fix ia64 OOPS in llog_test (2255)
5963 - zero end of page at obdfilter for partial page writes (2648)
5964 - don't leave stale dentries around after renames (bug 2428)
5965 - fix timeouts when evicting a client with a single lock held (2642)
5966 - set deadline for the initial HELLO message to drain (2634)
5967 - print out dotted-quad IP addresses in the socknal (2302)
5969 - additional debugging for MDS client eviction problem (2443)
5970 - fix mkfsoptions support for osts (2603, 2604)
5972 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
5975 - add series for the vanilla 2.6.0 kernel
5976 - add series for the vanilla 2.4.24 kernel
5977 - add series for a cray x86/64 UL kernel drop
5978 - fix xattr patches for the vanilla 2.4.19 series
5980 - generate true UUIDs in lmc (1171)
5981 - have portals stack dumping break in UML (2466)
5982 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
5983 - allocate with _NOFS in ldlm to avoid deadlock (1933)
5984 - wake callback waiting threads on client eviction (2460)
5985 - Add --ptldebug and --subsystem to lmc (1719)
5986 - update assertion to allow safe interrupt allocation
5987 - set rq_no_resend for cancel requests (2432)
5988 - recalculate ptlrpcd timeout after resend (2494)
5989 - call vfs_rmdir when removing pending directories (2368)
5990 - fix renaming a file to itself (2429)
5991 - lmc creates a default one-stripe lov (2454)
5992 - expand procfs space to handle large clusters (2326)
5993 - increase UML stack to avoid overflow
5994 - update lconf's list of debug and subsystem masks
5995 - fix lfs find --obd (2510)
5996 - /proc tunable for disabling filter read caching (2591)
5997 - stop rpm packages from altering slapd.conf (2301)
5998 - disable nagle in the socknal under 0conf (2578)
5999 - choose mds inode size based on stripe count (2572)
6000 - fix kernel-source rpm problems (2516)
6002 - add --disable-doc to avoid pdf generation (2421)
6003 - update documentation, tests, type-os, comments
6004 - avoid format warnings on ia64
6005 - remove the TOE NAL
6006 - tiny code cleanups by removing unused fields
6008 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
6011 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
6012 - remove the most blatant lies from BUILDING file (2371)
6013 - change default debug level to reasonable production setting
6014 - reduce client side cache size to reduce cache flush time
6015 - reduce max RPCs in flight to avoid unnecessary file fragmentation
6016 - make TCP zerocopy and pinger support enabled by default (2476)
6017 - sync writes completed after process exits caused crashes (2319)
6018 - maintain correct mount count on the MDS (2356)
6019 - backout 1557, because 2316 wasn't really fixed
6020 - better file I/O statistics gathering in /proc
6021 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
6022 - another kernel patch to fix zero-copy TCP function export
6023 - don't take duplicate lock when processing re-sent getattr (2420)
6024 - lctl uses obd_self_export instead of creating new conn (2353)
6025 - MDS/OST recovery case which requires object creation asserted (2425)
6026 - move lfs from /usr/sbin to /usr/bin in packages
6027 - fix race between mds_client_add and mds_client_free (2417)
6028 - use kmalloc instead of slabs in portals (2430)
6029 - don't create duplicate records when a failover MDS is present (2442)
6030 - remove unnecessary mount age check (2332)
6031 - don't remove directory inodes from locks prematurely (2451)
6032 - don't break if MDS service name is the same as hostname (2103)
6033 - fix races in client write RPC generation when cache full (2482)
6035 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
6038 - remove now-unused request->rq_obd (278)
6039 - if an allocation fails, print out how much memory we've used (1933)
6040 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
6041 - add the "configurable stack size" patch to most series files (1256)
6042 - ability to write large log records, for 100+ OST configs (2306)
6043 - fix NULL deref when filter_prep fails (2314)
6044 - fix operator precedence error in filter_sync
6045 - dynamic allocation of socknal TX descriptors (2315)
6046 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
6047 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
6048 - ensure that CWARN messages in Portals always get to the syslog
6049 - __init/__exit are not for prototype decls (ldlm_init/exit)
6050 - x86-64 compile warning fixes
6051 - fix gateway LMC keyword conflict (2318)
6052 - fix MDS lock inversions in getattr/reint paths (1844)
6053 - fix a rare lock re-ordering bug, which caused deadlock (2322)
6054 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
6055 - DLM race condition prevented some lock evictions (2328)
6056 - ENOMEM detection and retry on socknal sends (2230)
6057 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
6058 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
6059 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
6060 - better error messages when a client is rejected during recovery (1505)
6061 - avoid cancelling locks which were never granted, after failure (2330)
6062 - fix i_sem/journal inversion in mds_client_add (2333)
6063 - fix truncate/getattr lock cycle deadlock (2334)
6064 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
6065 - fix two rare exit paths which could leak an l_lock() ref (2321)
6066 - fencepost error in MDS/OST orphan recovery (2226)
6067 - make log record alignment 8 bytes (1988)
6068 - lstripe now fails when requested offset > ost_count (2237)
6069 - ensure that all kernel series have a complete list.h (1607)
6070 - fix crashes in special-file operations (2316)
6071 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
6073 - allow configurable automake binary, for testing new versions
6074 - small update to the lfs documentation
6076 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
6078 * fix negative export reference count in fsfilt_sync (2312)
6080 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
6081 * release candidate 0.9.1
6083 - orphans are moved into the PENDING directory for possible recovery
6084 - replayed opens now open by fid for orphan/rename safety (1042)
6085 - last close of an orphan inode generates a transno (683)
6086 - chdir() and mount() now pin the directory entry (1020)
6087 - avoid CERROR in normal ll_setattr_raw() error case (1500)
6088 - discard very old requests without processing them (1502)
6089 - remove some common, well-understood CERRORs (1505)
6090 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
6091 - clear "grant" flags in OST replies until OST grant code lands (1644)
6092 - fix read performance by not clobbering i_blksize on client (1598)
6093 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
6094 - make LRU size a /proc tunable, clears locks when reduced (707)
6095 - fix some lprocfs rot that prevented ptlbd from loading (1732)
6096 - server locks take references on exports now (1558)
6097 - build fixes for 2.4.20-rh trees (1663)
6098 - return an error from lov_create if all OSCs are inactive (1751)
6099 - fix import levels when a reconnect happens without a timeout (1597)
6100 - exit early from mds_open if we get a lookup error (1749)
6101 - partial page read at EOF wouldn't wait for disk before sending (1642)
6102 - avoid NULL deref in obdfilter when reading page past EOF (1592)
6103 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
6104 - fix LBUG in ll_it_open_error with rc = -2 (1861)
6105 - write/truncate lock inversion (1639)
6106 - Don't auto-load obdclass, portals modules during cleanup (1495)
6107 - fix timestamps from jumping to "now" (1763)
6108 - extra journal assertions (1648)
6109 - add an extra multiunlink test (1771)
6110 - fix read_record/write_record API (1776)
6111 - fix leak of offset_extent, possible incorrect i_size later (1772)
6112 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
6113 - Debugging for the kqswnal_get_idle_tx problems (1820)
6114 - Allow recovery to be attempted multiple times (1536)
6115 - Write out MDS last_rcvd file after it is first created (1600)
6116 - Fix tx_descriptor leak in failed transmit situations (1827)
6117 - ext3 journaling fixes for assertion failure after IO error (1871)
6118 - class_export_put() on freed export after completion AST error (1896)
6119 - Fix revalidate looping in VFS (1322)
6120 - Don't access a freed export during MDS_REINT timeout (1521)
6121 - Add open-unlink recovery support on the MDS (1673,1764)
6122 - Return an error if no MDS data was read from last_rcvd (1946)
6123 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
6124 - Don't setup a disk device that doesn't match exported UUID (317)
6125 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
6126 - avoid committing NULL handle in force close
6127 - local.sh is now a one-stripe LOV configuration
6128 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
6129 - fix ext3 htree duplicate directory entry corruption (1516)
6130 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
6131 - update to new LOV EA format (2097)
6132 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
6134 2003-06-15 Phil Schwan <phil@clusterfs.com>
6137 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
6138 - per-import/export recovery handling (958, 931, 959)
6139 - multiple last-rcvd slots, for serving multiple FSes (949)
6140 - connections are again shared between multiple imp/exports (963, 964)
6141 - "umount -f" would hang if any requests needed to be sent (393, 978)
6142 - avoid pinning large req buffer by copying for queued messages (989)
6143 - add "uuid" to "lctl device" command to help upcalls (991)
6144 - "open" RPCs with transnos would confuse recovery counters (1037)
6145 - do proper endian conversion of all wire messages (288, 340, 891)
6146 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
6147 - call waiting locks callback from LDLM recovery thread (1127, 1151)
6148 - fix ptlrpc_connection leak in target_handle_connect (1174)
6149 - fix import refcounting bug in OST and MDS cleanup (1134)
6150 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
6151 - fix very unlikely obd_types race condition (501)
6152 - remove osc_open hack for echo_client (1187)
6153 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
6154 - a failure in read_inode2 leads to deadlock (1139)
6155 - cancel ack-locks as soon as transaction is committed (1072)
6156 - fix major leaks and crashes in the bulk I/O path (937, 1057)
6157 - make sure to commitrw after any preprw to avoid deadlock (1162)
6158 - failing to execute a file in a lustre FS would lock inode (1203)
6159 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
6160 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
6161 - obd_uuid2tgt would walk off the end of the list (1255)
6162 - on IA64 the osc would give portals incorrect bulk size (1258)
6163 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
6164 - fix lock inversion caused by new llite matching code (1282)
6165 - limit the number of dirty pages on a client to 10MB (1286)
6166 - timed out locks were not being corrected cancelled (1289)
6167 - fix O_DIRECT above 4GB on IA-32 (1292)
6168 * major user-visible changes
6169 - fail out/fail over policy now controlled by the upcall (993)
6171 - add OBD_PING to check server availability and failure (954)
6172 - lustre messages are now sent in sending host order (288, 340, 891)
6173 - add eadatalen to MDS getattr reply (340)
6174 - OST read replies may contain second buffer, with per-page status (593)
6176 2003-03-11 Phil Schwan <phil@clusterfs.com>
6179 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
6180 - failed open()s could cause deadlock; fixed (867, 869)
6181 - stop cancelling OST locks when files are closed (481)
6182 - overlapping XID spaces caused network corruption (851, 853)
6183 - fix unsafe fsfilt counter arithmetic; change to atomic_t
6184 - setattr_raw added, to do single-RPC, server-side setattrs
6185 - lmc/lconf syntax change for OST UUIDs
6186 - fix crashy race condition between ptlrpc_free_req and osc_close
6187 - don't use request in mdc_enqueue if we hit a timeout (889)
6188 - don't set the inode i_size for regular files from the MDS (896)
6189 - handle out of order completion AST (842)
6190 - don't LBUG if a lock request times out after receiving AST (913)
6191 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
6192 - if a bad lock AST arrives, send an error instead of dropping entirely
6193 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
6194 - fix leak in bulk IO when only partially completed (899, 900, 926)
6195 - fix O_DIRECT for ia64 (55)
6196 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
6197 - C-z after timeout could hang a process forever; fixed (977)
6199 - client-side I/O cache (678, 924, 929, 941, 970)
6201 - READPAGE and SETATTRs which don't take server-side locks get
6204 2003-02-11 Phil Schwan <phil@clusterfs.com>
6207 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
6208 - Fix fsfilt_extN_readpage() to read a full page of directory
6209 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
6210 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
6211 - Fix imbalanced LOV object allocation and out-of-bound access (469)
6212 - Most intent operations were removed, in favour of a new RPC mode
6213 that does a single RPC to the server and bypasses most of the VFS
6214 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
6215 - Aggressively cancel local locks on DLM servers
6216 - mds_reint_unlink sends EA to the client if it's the last nlink.
6217 client uses that EA to unlink OST objects.
6218 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
6219 - recursive symlinks were fixed (439)
6220 - fixed NULL deref in DEBUG_REQ
6221 - filter_update_lastobjid no longer calls sync, which annoyed extN
6222 - fixed multi-client small-writes to a single file problem (445)
6223 - fixed mtime updates during file writes (607)
6224 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
6225 - fixed bug in obd_brw_read/write() (under guise of testing 367)
6226 - fixed Linux OST size reporting problem (444, 656)
6227 - OST now updates object mtime with writes or setattr (607, 619)
6228 - client verifies file size before zeroing page past EOF (445)
6229 - OST now writes last allocated objid to disk with allocation (108)
6230 - LOV on echo now works (409)
6232 - mds_reint_unlink sends a new buffer, with the EA included. this
6233 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
6234 set if a regular file was being unlinked, and it was the last link
6235 - use PtlGet from the target for bulk writes (315)
6236 - OST now updates object mtime with writes or setattr (607, 619)
6237 - LDLM now has a grant-time callback to revalidate locked items, if
6239 - Many MDS operations were reorganized to combat race conditions
6241 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
6242 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
6244 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
6247 - Fully reactivate OST imports after reconnection (512, others)
6248 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
6249 - More graceful error handling for truncating on dead OST (515)
6250 - Don't error out unless we're actually accessing dead stripes (474)
6251 - Fix garbage sizes when stripes are missing (410)
6252 - LRU counters were broken, causing constant lock purge (433, 432)
6253 - garbage on read from stripes with failed OSTs (441)
6254 - mark OSCs as active before reconnecting during recovery (438)
6255 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
6256 - lfind did not preserve OST order in output (443)
6257 - symlinks cause hung clients, incorrect data (439)
6258 - stop dereferencing request after dropping refcount (457)
6259 - don't LASSERT(spin_is_locked) on non-SMP (455)
6260 - fixes for many rename() bugs
6261 - fstat didn't correctly synchronize attributes (399)
6262 - server must handle lock cancellation during blocking AST prep (487)
6263 - bulk descriptors were free()d too soon (511)
6264 - fix paths in lconf, which would load incorrect modules (451, 507)
6265 - fix confusing lconf 'host not found' error message (386)
6266 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
6267 - fix race condition in mdc_blocking_ast() for inode access (526)
6268 - fix lov_unpackmd() unpacking wrong number of stripes (537)
6269 - fix lov_set_osc_active() marking wrong OSC inactive (440)
6270 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
6271 - fix multiple writes of stripe MD to MDS (358, maybe 519)
6272 - fix lstripe in several ways (kernel side) (527)
6273 - fix request leak in ldlm_cli_enqueue (262)
6274 - incorrect OSC was marked inactive after OST failure
6275 - call mds_fs_cleanup before unmounting filesystem (524)
6276 - fix races between taking ns_lock and ldlm_lock_change_resource
6277 - fix races updating LOV export open file list
6278 - fix lov_enqueue error path, avoid decref-ing bad lock handle
6279 - fix recovery NULL deref in ldlm_cli_cancel_unused
6280 - fix some DLM races by using new hash table for lock handles (419)
6281 - permit the client to specify desired inodes, at replay
6282 - duplicate requests when we queue them for replay reintegration
6283 - fix last_rcvd offset calculation
6284 - sync after each recovered transaction, so we always make progress
6285 - never, not always, ERESTART requests without transnos
6286 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
6288 - skip replay if the MDS says that the client is already connected
6289 - don't check for a recovery-enabled export to match lctl's UUID
6290 - don't INC_USE_COUNT for phantom exports
6291 - don't crash when cleaning up phantom exports (567)
6292 - don't double-finish or set replay data for errored mdc_open requests
6293 - abort requests when they time out, so we don't get old replies
6294 - send/receive replies for AST messages again
6295 - if the client says that it doesn't have the lock, cancel it on the
6297 - if we timeout during I/O, don't try to cancel an in-use lock; instead
6298 mark it as destroyed, it will all work out when decref is called
6299 - fix module use counts (22, 581)
6301 - ASTs now expect a reply (server cancels lock on error reply)
6303 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
6306 - fix many simultaneous client startup (392)
6307 - fix dentry->d_it clobbering
6308 - credentials weren't being shipped for readdir/getattr operations
6309 - remove invalid assertions triggered during some concurrent MD
6311 - proper Lustre versions added (336, 389)
6312 - fix memory leak for create error case (398)
6313 - fix LOV locking bug that would get cli/srv out of sync
6314 - fix echo client over LOV (409)
6315 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
6316 - fix double-O_EXCL intent crash (424)
6317 - avoid sending multiple lock CANCELs (352)
6319 - MDS can do multi-client recovery (modulo bugs in new code)
6321 - many updates, edits, cleanups
6323 2002-11-18 Phil Schwan <phil@clusterfs.com>
6326 - fix null d_it dereference (346)
6327 - fix full OST/dbench hang (333)
6328 - fix permission problem with file removal (286)
6329 - fix removal of OSCs from LOV when they fail
6330 - fix NULL deref during bulk timeout (214)
6331 - fix problems related to multiple filesystems on one MDS (241)
6332 - fixed serious subtle metadata locking bugs
6333 - free locks on clients when inodes are removed due to memory
6335 - fix inode pointer in lock data (285)
6336 - partial support for multiple MDS on a single host (241)
6337 - data locks weren't cancelled at clear_inode time (290, 311)
6338 - intent locks could lead to unbounded lock growth (205)
6339 - added a maximum lock count, an LRU list, and a flusher
6340 - fix multiple rename (365)
6341 - properly abstracted the echo client
6342 - OSC locked 1 byte too many; fixed
6343 - rewrote brw callback code:
6344 - fixed recovery bugs related to LOVs (306)
6345 - fixed too-many-pages-in-one-write crash (191)
6346 - fixed (again) crash in sync_io_timeout (214)
6347 - probably fixed callback-related race (385)
6349 - Add capability to MDS protocol
6350 - LDLM cancellations and callbacks on different portals
6352 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
6355 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
6356 - defer open object destruction to close time (601981, 138)
6357 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
6358 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
6359 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
6360 - don't call obd_setattr in ll_file_release for destroyed objects
6361 * protocol change to lustre_msg: move |version| and add |flags|
6362 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
6363 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
6364 * added replay of create, unlink, link and rename operations during
6365 MDS failover; recovery should be much more robust now
6366 * remove failed OSCs from LOVs (only lov_create uses this so far)
6367 * the lustre-HOWTO was brought (more) up to date (582544)
6369 2002-10-23 Phil Schwan <phil@clusterfs.com>
6372 - in-use dentries weren't being reused properly (617851)
6373 - prevent multiple LDLM setup (599178)
6374 - fix LOV size calculations for truncate (617853)
6375 - fix client handling of MDS intent errors (POSIX)
6376 - fix permission bug in lovstripe.c test (624321)
6377 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
6378 - truncate past end of file could corrupt data
6379 - proper cleanup after timeouts, crashes, etc (592524, 550815)
6380 - a race in recovery could return ETIMEDOUT to apps (623947)
6381 - building outside the source directory was fixed
6382 * the lustre-HOWTO was brought (more) up to date (582544)
6383 * major progress was made on recovery functionality
6385 2002-10-10 Phil Schwan <phil@clusterfs.com>
6388 - recovery deadlock fix
6389 - rm -rf causes LBUG fix (617817)
6390 - file open by multiple tasks fix (618962)
6391 - directory permissions bugs (602707 and 620007)
6392 - journal_stop fixed with locking (611313)
6393 - O_APPEND failures resolved (618273, perhaps 614459)
6394 - lconf PATH fix (619770)
6395 - IA64 build fix (621450)
6396 - RPC buffer sizes scale with amount of memory
6398 2002-10-01 Phil Schwan <phil@clusterfs.com>
6401 - locks would be cancelled without throwing away data pages,
6402 resulting in inconsistent data (605627)
6403 - inode attributes were not always being refreshed (605627, 612449)
6404 - lconf now continues to cleanup after lctl reports an error
6405 - MDS now enforces user permissions (602707)
6406 - lprocfs cleanup fixed, but not yet enabled (614157)
6407 - fixed infinite server hang, should a client not respond to an AST
6408 - avoid going into recovery if user calls readlink() with a buffer
6409 that's too small (613941)
6410 - AST RPCs no longer require replies (614867) -- this may be changed
6411 - don't crash server if client sends an IOV that's too big (611336)
6412 - fixed lock conversion deadlock (611892)
6413 - fixed the following of symlinks (614622)
6414 * recovery: the server can remove locks from a client that dies, other
6415 clients can make progress
6416 * more extN patch fixes
6417 * compile-time configurable ptlrpc buffer allocations
6419 - collaborative read cache document
6420 - Lustre Lite Performance CDR document-in-progress
6422 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6425 - fix typo in patch-2.4.18
6427 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6430 - clear ptlrpc request each time in handle_incoming_request()
6431 - unlink of files now destroys the object on the OST
6433 2002-09-19 Peter Braam <braam@clusterfs.com>
6435 * add hard link support
6436 * change obdfile creation method
6437 * kernel patch changed
6439 2002-09-19 Peter Braam <braam@clusterfs.com>
6442 - stack overflow bug in extN fixed
6444 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
6446 * documentation updates
6447 - add man pages for config tools
6448 - update tests/README to describe testing with new config tools
6449 - finish metadata API descriptions
6450 * bug fixes and cleanups
6451 - statfs workaround for 16TB limit
6452 - LOV stripe allocation improved, can stripe on subset of OSTs
6453 - LOV file size/IO offset was wrong for files > 4GB in size
6454 - object EA data was being dropped, caused files to be unreadable
6455 - memory overflow with non-LOV OST caused memory corruption
6456 - fixed regression tests to work with new config tools, obdfilter
6457 - fixed bug when directory size became larger than 1 block
6458 - fixed bug (for single client case) when PWD was deleted
6459 - invalidate local directory pages when doing intent-based ops
6460 - avoid LDLM oops when lock callback contained bad data
6462 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
6464 * documentation updates
6465 * bug fixes and cleanups
6466 - configuration tools
6469 - 64-bit compile warnings
6470 - 64-bit internal statfs data
6472 * test_brw on persistent OST devices
6474 * lprocfs (disabled)
6476 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
6478 * documentation updates
6479 * bug fixes and cleanups
6480 * configuration tools
6482 2002-08-30 Peter J. Braam <braam@clusterfs.com>
6485 * many small fixes to 0_5_4
6486 * io/network handling
6487 * thinkos in MDS operations
6489 2002-08-24 Peter J. Braam <braam@clusterfs.com>
6492 * crucial basic fixes to 0.5.3
6493 * IOR, Iozone work over Elan
6496 2002-08-07 Phil Schwan <phil@clusterfs.com>
6497 * version 0_5_3, our first alpha
6498 * we use the new Portals iovs
6499 * documentation updates
6500 * bug fixes and cleanups
6501 * small changes in the DLM wire protocol
6503 2002-07-25 Peter J. Braam <braam@clusterfs.com>
6504 * version 0_5_1 with some initial stability,
6505 * locking on MD and file I/O.
6506 * documentation updates
6507 * several bug fixes since 0.5.0
6508 * small changes in wire protocol
6510 2002-07-18 Phil Schwan <phil@clusterfs.com>
6512 * delivered as Lustre Light Alpha
6513 * fixed a crash after handling invalid MDS requests
6514 * fixed directory pages for architectures with non-4k pages sizes
6516 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
6517 * release version v0_4_4
6518 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
6520 * Fixes a number of interruption problems with OST operations.
6521 * Update documentation for portals header changes
6522 * Move all wire protocol structs/defines to lustre_idl.h
6523 * Fixes symlink length bug.
6524 * Add tcpdump to repository.
6526 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
6527 * release version v0_4_3
6528 * Fixes statfs for inodes on extN.
6529 * Fixes bug in runtests which would delete /etc/hosts.
6530 * Use 64-bit object IDs wherever possible (not into VFS though)
6531 Remove ost_get_info, which is unused by lustre, and out of date.
6533 2002-07-03 Peter Braam <braam@clusterfs.com>
6534 * release version v0_4_2 Fixes a lookup error (type not passed)
6535 * move forward to head of Portals
6536 * move forward to latest Lustre kernel
6538 2002-06-25 Peter Braam <braam@clusterfs.com>
6539 * release version v0_4_1. Hopefully stable on single node use.