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 2.6.32-573.18.1.el6 (RHEL6.7)
9 3.10.0-327.10.1.el7 (RHEL7.2)
10 3.0.101-0.47.71 (SLES11 SP3)
11 3.0.101-68 (SLES11 SP4)
12 3.12.53-60.30 (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 2.6.32-573.18.1.el6 (RHEL6.7)
18 3.10.0-327.10.1.el7 (RHEL7.2)
19 3.0.101-0.47.71 (SLES11 SP3)
20 3.0.101-68 (SLES11 SP4)
21 3.12.53-60.30 (SLES12 SP1)
23 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
24 * Recommended ZFS / SPL version: 0.6.4.2
25 * Tested with ZFS / SPL version: 0.6.4.2
26 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
27 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
28 stack overflow. For more information, please refer to bugzilla 17630.
29 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
30 the re-exporting nodes to be mounted with "32bitapi" mount option
32 --------------------------------------------------------------------------------
34 02-29-2016 Intel Corporation
36 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
37 for currently supported client and server kernel versions.
38 * Server known to build on patched kernels:
39 2.6.32-431.29.2.el6 (RHEL6.5)
40 2.6.32-504.30.3.el6 (RHEL6.6)
41 2.6.32-573.12.1.el6 (RHEL6.7)
42 3.10.0-327.3.1.el7 (RHEL7.2)
43 3.0.101-0.47.71 (SLES11 SP3)
44 3.0.101-68 (SLES11 SP4)
45 vanilla linux 4.2.1 (ZFS only)
46 * Client known to build on unpatched kernels:
47 2.6.32-431.29.2.el6 (RHEL6.5)
48 2.6.32-504.30.3.el6 (RHEL6.6)
49 2.6.32-573.12.1.el6 (RHEL6.7)
50 3.10.0-327.3.1.el7 (RHEL7.2)
51 3.0.101-0.47.71 (SLES11 SP3)
52 3.0.101-68 (SLES11 SP4)
55 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
56 * Recommended ZFS / SPL version: 0.6.4.2
57 * Tested with ZFS / SPL version: 0.6.4.2
58 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
59 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
60 stack overflow. For more information, please refer to bugzilla 17630.
61 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
62 the re-exporting nodes to be mounted with "32bitapi" mount option
64 --------------------------------------------------------------------------------
66 03-10-2015 Intel Corporation
68 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
69 for currently supported client and server kernel versions.
70 * Server known to build on patched kernels:
71 2.6.32-431.29.2.el6 (RHEL6.5)
72 2.6.32-504.8.1.el6 (RHEL6.6)
73 3.0.101-0.46 (SLES11 SP3)
74 * Client known to build on unpatched kernels:
75 2.6.32-431.29.2.el6 (RHEL6.5)
76 2.6.32-504.8.1.el6 (RHEL6.6)
77 3.10.0-123.20.1.el7 (RHEL7)
78 3.0.101-0.46 (SLES11 SP3)
79 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
80 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
81 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
82 stack overflow. For more information, please refer to bugzilla 17630.
83 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
84 the re-exporting nodes to be mounted with "32bitapi" mount option
86 Severity : enhancement
88 Description: control OST-index in IDIF via ROCOMPAT flag.
89 Details : Introduce new flag OBD_ROCOMPAT_IDX_IN_IDIF that is stored in the
90 last_rcvd file. For new formatted OST device, it will be auto set;
91 for the case of upgrading from old OST device, you can enable it
92 via the lproc interface osd-ldiskfs.index_in_idif. With such flag
93 enabled, for new created OST-object, its IDIF-in-LMA will contain
94 the OST-index; for the existing OST-object, the OSD will convert
95 old format IDIF as new format IDIF with OST-index stored in the
96 LMA EA when accessing such OST-object or via OI scrub. Once such
97 flag is enabled, it cannot be reverted back, so the system cannot
98 be downgraded to the orignal incompatible version.
100 --------------------------------------------------------------------------------
102 07-30-2014 Intel Corporation
104 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
105 for currently supported client and server kernel versions.
106 * Server known to build on patched kernels:
107 2.6.32-431.20.3.el6 (RHEL6.5)
108 3.0.101-0.31 (SLES11 SP3)
109 * Client known to build on unpatched kernels:
110 2.6.32-431.20.3.el6 (RHEL6.5)
111 3.0.101-0.31 (SLES11 SP3)
114 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
115 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
116 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
117 stack overflow. For more information, please refer to bugzilla 17630.
118 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
119 the re-exporting nodes to be mounted with "32bitapi" mount option
121 --------------------------------------------------------------------------------
123 10-31-2013 Intel Corporation
125 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
126 * Server builds for kernels:
127 2.6.32-358.18.1.el6 (RHEL6.4)
128 * Client builds for unpatched kernels:
129 2.6.32-358.18.1.el6 (RHEL6.4)
130 3.0.92-0.8 (SLES11 SP3)
132 * Recommended e2fsprogs version: 1.42.7.wc2 or newer
133 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
134 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
135 stack overflow. For more information, please refer to bugzilla 17630.
136 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
137 the re-exporting nodes to be mounted with "32bitapi" mount option
139 --------------------------------------------------------------------------------
141 05-30-2013 Intel Corporation
143 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
144 * Server builds for kernels:
145 2.6.32-358.6.2.el6 (RHEL6.4)
146 * Client builds for unpatched kernels:
147 2.6.32-358.6.2.el6 (RHEL6.4)
148 3.0.74-0.6.6 (SLES11 SP2)
150 * Recommended e2fsprogs version: 1.42.7.wc1 or newer
151 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
152 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
153 stack overflow. For more information, please refer to bugzilla 17630.
154 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
155 the re-exporting nodes to be mounted with "32bitapi" mount option
157 --------------------------------------------------------------------------------
159 9-30-2012 Whamcloud, Inc.
161 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
162 * Server support for kernels:
163 2.6.32-279.5.1.el6 (RHEL6)
164 * Client support for unpatched kernels:
165 2.6.32-279.5.1.el6 (RHEL6)
167 2.6.18-238.19.1.el5 (RHEL5)
168 2.6.16.60-0.69.1 (SLES10)
169 * Recommended e2fsprogs version: 1.41.90.wc4
170 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
171 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
172 stack overflow. For more information, please refer to bugzilla 17630.
173 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
174 the re-exporting nodes to be mounted with "32bitapi" mount option
176 Severity : enhancement
178 Description: Update RHEL6.3 kernel to 2.6.32-279.5.1.el6.
180 Severity : enhancement
182 Description: Update RHEL6.3 kernel to 2.6.32-279.2.1.el6.
184 Severity : enhancement
186 Description: Update RHEL6.2 kernel to 2.6.32-220.17.1.el6
188 Severity : enhancement
190 Description: Update RHEL6.2 kernel to 2.6.32-220.13.1.el6
192 Severity : enhancement
194 Description: Update RHEL6.2 kernel to 2.6.32-220.7.1.el6
196 --------------------------------------------------------------------------------
198 04-30-2012 Whamcloud, Inc.
200 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
201 * Server support for kernels:
202 2.6.32-220.4.2.el6 (RHEL6)
203 * Client support for unpatched kernels:
204 2.6.32-220.4.2.el6 (RHEL6)
206 2.6.18-238.19.1.el5 (RHEL5)
207 2.6.16.60-0.69.1 (SLES10)
208 * Recommended e2fsprogs version: 1.41.90.wc4
209 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
210 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
211 stack overflow. For more information, please refer to bugzilla 17630.
212 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
213 the re-exporting nodes to be mounted with "32bitapi" mount option
215 Severity : enhancement
217 Description: Update RHEL6.2 kernel to 2.6.32-220.4.2.el6
219 Severity : enhancement
221 Description: Update RHEL6.2 kernel to 2.6.32-220.4.1.el6
223 Severity : enhancement
225 Description: Update RHEL6.2 kernel to 2.6.32-220.el6
227 Severity : enhancement
229 Description: Update RHEL6.1 kernel to 2.6.32-131.17.1.el6
231 --------------------------------------------------------------------------------
233 09-30-2011 Whamcloud, Inc.
235 * Server support for kernels:
236 2.6.18-238.19.1.el5 (RHEL5)
237 2.6.32-131.6.1.el6 (RHEL6)
238 * Client support for unpatched kernels:
239 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
240 2.6.18-238.19.1.el5 (RHEL5)
241 2.6.32-131.6.1.el6 (RHEL6)
242 2.6.29.4-167.fc11 (FC11)
243 2.6.16.60-0.69.1 (SLES10)
245 * Recommended e2fsprogs version: 1.41.90.wc3
246 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
247 Lustre filesystem with 4K stack may cause a stack overflow. For more
248 information, please refer to bugzilla 17630.
249 * NFSv4 reexport requires Lustre client on the re-exporting nodes to be
250 mounted with "32bitapi" mount option
252 Severity : enhancement
253 Description: Update RHEL6.1 kernel to 2.6.32-131.6.1.el6
255 Severity : enhancement
256 Description: Update RHEL5.6 kernel to 2.6.18-238.19.1.el5
258 Severity : enhancement
259 Description: Update RHEL5.6 kernel to 2.6.18-238.9.1.el5
262 Frequency : only with SLES10
264 Description: Use OFED "KMP" provided by Novell
265 Details : SLES10 SP3 ships with OFED in a separate "KMP" package.
266 Lustre is now built against this package. That means you need to
267 install the ofed-kmp package from Novell for the patchless client.
268 Note that the ofed-kmp that Novell ships may not exactly match the
269 kernel version but should still be compatible.
271 Severity : enhancement
273 Description: Update RHEL5.5 kernel to 2.6.18-194.17.1.el5
274 Update OEL5.5 kernel to 2.6.18-194.17.1.0.1.el5.
276 --------------------------------------------------------------------------------
278 2010-07-15 Oracle, Inc.
280 * Server support for kernels:
281 2.6.18-164.11.1.el5 (RHEL 5)
282 2.6.18-164.11.1.0.1.el5 (OEL 5)
283 * Client support for unpatched kernels:
284 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
285 2.6.18-164.11.1.el5 (RHEL 5),
286 2.6.18-164.11.1.0.1.el5 (OEL 5)
287 2.6.16.60-0.42.8 (SLES 10),
289 2.6.29.4-167.fc11 (FC11)
290 * Recommended e2fsprogs version: 1.41.10-sun2
292 Severity : enhancement
294 Description: Distribute lfs_migrate script with lustre
295 Details : The lfs_migrate mentioned in the manual is now included in the
296 lustre tree in lustre/scripts and will distributed in lustre RPMs.
298 Severity : enhancement
300 Description: Failover nid list added to import proc file.
302 Severity : enhancement
304 Description: "som_preview" mount option required to use SOM
305 Details : The clients will need to use som_preview mount option
309 Description: Lustre does not do 1MB IOs to HW RAID
310 Details : Bump MAX_PHYS/HW_SEGMENTS and SG_ALL to 256 in the RHEL5 kernel.
312 Severity : enhancement
314 Description: lfs quota output is not very convenient for awk/sed-parsing
315 Details : Some positions in lfs quota output table could be empty or
316 non-empty which made it hard to parse it with scripts, now
317 a dash is put instead of space where there is not supposed
322 Description: Update to OFED 1.5.1
323 Details : OFED 1.5.1 will be used in environments that don't use
324 vendor supplied OFED.
328 Description: Remove lnet. prefix from lctl params display
329 Details : Also renamed lnet.memused to lnet_memused to
334 Description: Use conf_param -d to remove permanant settings.
338 Description: Add list_param -R
339 Details : lctl list_param now supports recursive mode.
342 Bugzilla : 21829, 22312
343 Description: Fix llobdstat.
344 Details : 1. Now updating the write_bytes counter
345 2. Port script fixes from 1.8.
349 Description: decrease the usage of memory on clients.
350 Details : 1. On clients, recycle dentries and inodes unused.
351 2. Delete the code related to ll_deathrow(att 6215 in bug 1443). It
356 Description: llobdstat fix and enhancement
357 Details : add a counter to set a limit to how many samples will be returned
358 fix a wildcard in the path to limit to obdfilter stats only
362 Description: lfs check only allowed for root
363 Details : Code cleanups and added a regresion test of non-root use of
368 Description: MDS Oopses in ptlrpc_dispatch_difficult_reply()/__list_add()
369 Details : Fix hr->hr_index overload.
373 Description: Fix FIEMAP functionality.
377 Description: FC11 patchless client support added
381 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and
382 OEL5.4 kernel to 2.6.18-164.11.1.0.1.el5.
386 Description: New acc-sm script and various test-framework enhancments.
390 Description: Previously landed patch was reverted.
394 Description: Update SLES10 kernel to 2.6.16.60-0.42.8 (patchless)
398 Description: Add ldiskfs maxdirsize mount option.
402 Description: Fix mkfs.lustre support for 16TB LUNs.
406 Description: Print mmp_check_interval
407 Details : print mmp_check_interval and make it possible to abort
408 mount operation in case it takes too long.
412 Description: Add cascading_rw to lustre mpi tests.
416 Description: Store SOM data in EA.
417 Details : Also several SOM related cleanups.
421 Description: Add refcount to md_open_data
425 Description: Use refcount on lov_request_set to prevent memory corruption.
429 Description: Statahead: use parent dir's i_mutex for synchronization.
433 Description: loadgen improvements.
437 Description: Enable peer health detection in ksocklnd and ko2iblnd.
441 Description: Control DCACHE_LUSTRE_INVALID flag with MDS_INODE_LOOKCUP lock
445 Description: Fix race on obd_devcie:md_stats
449 Description: Fix SLES 11 client builds.
450 Details : Resolved uninitalized pointer gcc warnings.
454 Description: Allow non-root access for "lfs check".
455 Details : Added a check in obd_class_ioctl() for OBD_IOC_PING_TARGET.
459 Description: Clear imp_force_reconnect correctly in ptlrpc_connect_interpret()
461 Severity : enhancement
463 Description: Update kernel to RHEL5.4 2.6.18-164.6.1.el5 and
464 OEL5 2.6.18-164.6.1.0.1.el5(Both in-kernel OFED enabled).
468 Description: Sanity test_27p fails.
469 Details : OBD_FAIL_OST_ENOSPC has been moved to filter_statfs().
470 Sanity tests 27 were cleaned up and improved.
474 Description: Unify req format on client/servers
475 Details : Use new req_capsule API [almost] everywhere instead of old PTLRPC
476 buffers and swabbers approach..
481 Description: drop join file stuff.
483 Severity : enhancement
485 Description: Add 2.6.27 and SLES11 (patchless client) support
489 Description: Fix in ptlrpc_expire_one_request() to print the signed time
492 Severity : enhancement
494 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
497 Severity : enhancement
499 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
501 Severity : enhancement
503 Description: correctly handle big reply message.
504 Details : send LNet event if reply is bigger than buffer and adjust this buffer
509 Description: Drop unnecessary __GFP_NOMEMALLOC flag from filter_get_page()
511 Severity : enhancement
513 Description: Add support for OEL5.
517 Description: Conf-sanity.sh 50g test - deactivated OST should not cause a panic.
519 Severity : enhancement
521 Description: "weak-modules" support
522 Details : Implement "weak-modules" support which enables kernel modules
523 to be used with any kernel that implements the same kABI. In
524 order to achieve this modules are now installed in
525 /lib/modules/$(uname -r)/updates/kernel on all distributions.
529 Description: client could not reconnect to OST because of an active request.
530 Details : The client now retries to reconnect to the same server, if a
531 connect request failed with EBUSY or -EAGAIN. Test case 35b
532 conf-sanity.sh has been created.
536 Description: Don't allow make backward step on assignin osc next id.
537 Details : race between allocation next id and ll_sync thread can be cause
538 of set wrong osc next id and can be kill valid ost objects.
542 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
546 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
550 Description: Error handling in osc_statfs_interpret() has been improved.
551 Details : Check in osc_statfs_interpret() for EBADR.
555 Description: Do not send statfs() requests to OSTs disabled by administrator.
556 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
558 Severity : enhancement
560 Description: Added sanity-quota test to for lfs(1) to work with numeric uid/gid.
562 Severity : enhancement
564 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
566 Severity : enhancement
568 Description: Enhancement for lfs(1) command to use numeric uid/gid.
572 Description: Improved cleanup procedure for conf-sanity test_46a.
576 Description: Do not start bulk transfer, if client tried to reconnect.
577 Details : Do not start bulk transfer, if client tried to reconnect,
578 but failed due to active rpcs. Check for exp_abort_active_req
579 has been implemented.
583 Description: Prevent inconsistences between linux and lustre mount structures.
584 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
585 Make the sleep interruptible.
587 Severity : enhancement
589 Description: provide server to client comms path
590 Details : server to client communications path via new LDLM_SET_INFO rpc
594 Description: truncate starts GFP_FS allocation under transaction causing deadlock
595 Details : ldiskfs_truncate calls grab_cache_page which may start page
596 allocation under an open transaction. This may lead to
597 calling prune_icache with consequent lustre reentrance.
601 Description: send correctly lsm on open replay
602 Details : MDS is trust to LSM size on replay open, but client can set wrong size
607 Description: enable client interface failover
608 Details : When a child reconnects from another NID, properly update export
609 nid hash position and ldlm reverse import.
611 Severity : enhancement
613 Description: Communicate OST degraded/readonly state via statfs to MDS
614 Details : Flags in the statfs returned from OSTs indicate whether the
615 OST is in a degraded RAID state, or if the filesystem has
616 turned read-only after a filesystem error is detected.
621 Description: don't panic if EPROTO was hit when reading symlink
622 Details : correctly handling request reference in error cases.
624 Severity : enhancement
626 Description: Add LustreNetLink, a kernel-userspace communcation path.
628 Severity : enhancement
630 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
635 Description: access to llog context before init.
636 Details : move handling CATALOGS file at osc layer and forbid access to llog
640 Frequency : rare, only when client is writing to OST on local node
642 Description: Avoid deadlock for local client writes
643 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
644 memory freeing context. This allows OST threads to set the
645 PF_MEMALLOC flag on task structures in order to allocate memory
646 from reserved pools and complete IO.
647 Use GFP_HIGHUSER for OST allocations for non-local client writes,
648 so that the OST threads generate memory pressure and allow
649 inactive pages to be reclaimed.
651 Severity : enhancement
653 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
658 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
659 Details : move ldlm namespace creation in setup phase to avoid grab
660 _lprocfs_lock with cli_sem held.
664 Description: Temporarily disable grant shrink.
665 Details : Disable the feature for debugging.
668 Frequency : only when formatting test filesystems
670 Description: Unable to run several mkfs.lustre on loop devices at the same time
671 Details : mkfs.lustre returns error 256 on the concurrent loop devices
672 formatting. The solution is to proper handle the error.
674 Severity : enhancement
676 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
678 Severity : enhancement
680 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
685 Description: resolve race between obd_disconnect and class_disconnect_exports
686 Details : if obd_disconnect will be called to already disconnected export he
687 forget release one reference and osc module can't unloaded.
689 Severity : enhancement
691 Description: Allow tuning service thread via /proc
692 Details : For each service a new
693 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
694 created that can be used to set min/max thread counts, and get the
695 current number of running threads.
700 Descriptoin: don't return error if have partially created objects for file.
701 Details : lov_update_create_set uses set->set_success as index for created
702 objects, so if some requests will be failed, they will have hole
703 at end of array and qos_shrink_lsm for allocate correct lsm.
705 Severity : enhancement
707 Description: Update OFED support to 1.4.1
712 Description: Reduce small size read RPC
713 Details : Set read-ahead limit for every file and only do read-ahead when
714 available read-ahead pages are bigger than 1M to avoid small size
717 Severity : enhancement
719 Description: move AT tunable parameters for more consistent usage
720 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
722 Severity : enhancement
724 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
725 Details : allow skip disconnected ost for send statfs request and hide error
731 Description: fix racy locking of mballoc block bitmaps causing BUG
732 Details : The locking of the mballoc buddy bitmap and the in-memory
733 block bitmap was using two different spin locks in some
734 cases. This made it possible to incorrectly access the
735 mballoc bitmap while another process was modifying it,
736 causing a sanity assertion to fail. While no on-disk corruption
737 was reported, there was some risk of this happening.
740 Frequency : rare, on llog test 6
742 Descriptoin: don't allow connect to already connected import
743 Details : allowing connect to already connected import is hide connecting problem.
748 Descriptoin: allow kill process which wait statahead result
749 Details : for some reasons 'ls' can stick in waiting result from statahead,
750 in this case need way for kill this process.
752 Severity : enhancement
754 Description: Add state history info file, enhance import info file
755 Details : Track import connection state changes in a new osc/mdc proc file;
756 add overview-type data to the osc/mdc import proc file.
758 Severity : enhancement
760 Description: MDS create should not wait for statfs RPC while holding DLM lock.
762 Severity : enhancement
764 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
769 Descriptoin: OST grant too much space to client even there are not enough space.
770 Details : Client will shrink its grant cache to OST if there are no write
771 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
772 this grant cache if there are already not enough avaible space
773 (left_space < total_clients * 32M).
776 Frequency : start MDS on uncleanly shutdowned MDS device
778 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
779 Details : stay in waiting mds<>ost recovery finished produce random bugs
780 due race between two ll_sync thread for one lov target. send
781 ACTIVATE event only if connect realy finished and import have
785 Frequency : rare, connect and disconnect target at same time
787 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
788 Details : don't call obd_disconnect under lov_lock. this long time
789 operation and can block ptlrpcd which answer to connect request.
794 Descriptoin: don't lose wakeup for imp_recovery_waitq
795 Details : recover_import_no_retry or invalidate_import and import_close can
796 both sleep on imp_recovery_waitq, but we was send only one wakeup
800 Frequency : always with long access acl
802 Descriptoin: mds can't pack reply with long acl.
803 Details : mds don't control size of acl but they limited by reint/getattr
806 Severity : enhancement
808 Description: Update to SLES10 kernel-2.6.16.60-0.33.
810 Severity : enhancement
812 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
815 Frequency : start MDS on uncleanly shutdowned MDS device
817 Descriptoin: aborting recovery hang on MDS
818 Details : don't throttle destroy RPCs for the MDT.
821 Frequency : on remount
823 Description: external journal device not working after the remount
824 Details : clear dev_rdonly flag for external journal devices in
830 Description: shutdown vs evict race
831 Details : client_disconnect_export vs connect request race.
832 if client will evicted at this time - we start invalidate
833 thread without referece to import and import can be freed
837 Frequency : rare, need acl's on inode.
839 Description: client can't handle ost additional correctly
840 Details : if ost was added after client connected to mds client can have
841 hit lnet_try_match_md ... to big messages to wide striped files.
842 in this case need teach client to handle config events about add
843 lov target and update client max ea size at that event.
845 Severity : enhancement
847 Description: Changelogs
848 Details : Changelogs are a lightweight mechanism to track filesystem
849 metadata and namespace changes. The changelog is recorded
850 permanently on the MDTs, and is periodically "consumed" / purged
851 when records are no longer needed.
853 Severity : enhancement
855 Description: compact fld format with extents
856 Details : Store range of seq rather than every seq in FLD. Seq
857 controller update FLD rather than clients. In Case of CMD, mdt0
858 has FLD, all other metadata server act as non persistent proxy
859 for FLD queries and cache fld entries in fld cache.
864 Description: don't skip ost target if they assigned to file
865 Details : Drop slow OSCs if we can, but not for requested start idx.
866 This means "if OSC is slow and it is not the requested
867 start OST, then it can be skipped, otherwise skip it only
868 if it is inactive/recovering/out-of-space.
872 Description: more cleanup in mds_lov
873 Details : not send LOV EA under replay, we can't know about they size at this
874 time. Don't allow client connect to mds before any ost connected,
875 for avoid problems with LOV EA size and returning EIO to client.
877 Severity : enhancement
879 Description: Interoperability at server side (Disk interoperability)
881 Severity : enhancement
883 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
885 Severity : enhancement
887 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
889 Severity : enhancement
891 Description: New client IO stack (CLIO).
893 Severity : enhancement
895 Description: Commit on sharing. Eliminate inter-client dependencies between
896 uncommitted transactions by doing transaction commits.
897 Thereby clients may recovery independently.
900 Frequency : Create a symlink file with a very long name
902 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
903 Details : If there is no extra space in the request for early cancels,
904 ldlm_req_handles_avail() returns 0 instead of a negative value.
906 Severity : enhancement
908 Description: Add /proc entry for import status
909 Details : The mdc, osc, and mgc import directories now have
910 an import directory that contains useful import data for debugging
913 Severity : enhancement
915 Description: Re-disable certain /proc logging
916 Details : Enable and disable client's offset_stats, extents_stats and
917 extents_stats_per_process stats logging on the fly.
920 Frequency : Only on FC kernels 2.6.22+
922 Description: oops in statahead
923 Details : Do not drop reference count for the dentry from VFS when lookup,
924 VFS will do that by itself.
926 Severity : enhancement
928 Description: Generic /proc file permissions
929 Details : Set /Proc file permissions in a more generic way to enable non-
930 root users operate on some /proc files.
934 Description: Hitting mdc_commit_close() ASSERTION
935 Details : Properly handle request reference release in
936 ll_release_openhandle().
940 Description: quota recovery deadlock during mds failover
941 Details : This patch includes att18982, att18236, att18237 in bz14840.
943 1. fix osts hang when mds does failover with quotaon
944 2. prevent watchdog storm when osts threads wait for the
949 Frequency : only patchless client
950 Description: add workaround for race between add/remove dentry from hash
952 Severity : enhancement
954 Description: Allow OST glimpses to return PW locks
958 Description: LBUG when llog conf file is full
959 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
963 Description: Prevent import from entering FULL state when server in recovery
967 Description: service mount cannot take device name with ":"
968 Details : Only when device name contains ":/" will mount treat it as
974 Description: replace ptlrpcd with the statahead thread to interpret the async
975 statahead RPC callback
979 Frequency : on recovery
980 Description: I/O failures after umount during fail back
981 Details : if client reconnected to restarted server we need join to recovery
982 instead of find server handler is changed and process self
983 eviction with cancel all locks.
985 Severity : enhancement
987 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
989 Severity : enhancement
991 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
993 Severity : enhancement
995 Description: Upcall on Lustre log has been dumped
996 Details : Allow for a user mode script to be called once a Lustre log has
997 been dumped. It passes the filename of the dumped log to the
998 script, the location of the script can be specified via
999 /proc/sys/lnet/debug_log_upcall.
1004 Description: avoid idr_remove called for id which is not allocated.
1005 Details : Move assigment s_dev for clustered nfs to end of initialization,
1006 to avoid problem with error handling.
1011 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
1012 Details : When connection is reused this not moved from CONN_UNUSED_HASH
1013 into CONN_USED_HASH and this prodice warning when put connection
1014 again in unused hash.
1017 Severity : enhancement
1019 Description: File striping can now be set to use an arbitrary pool of OSTs.
1021 Severity : enhancement
1023 Description: Export bytes_read/bytes_write count on OSC/OST.
1027 Description: Early reply size mismatch, MGC loses connection
1028 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
1029 the connect flags are properly negotiated.
1034 Description: quotas are not honored with O_DIRECT
1035 Details : all writes with the flag O_DIRECT will use grants which leads to
1036 this problem. Now using OBD_BRW_SYNC to guard this.
1040 Description: add quota statistics
1041 Details : 1. sort out quota proc entries and proc code.
1042 2. add quota statistics
1044 Severity : enhancement
1046 Description: enable quota support for HEAD.
1050 Description: Properly propagate oinfo flags from lov to osc for statfs
1051 Details : restore missing copy oi_flags to lov requests.
1053 Severity : enhancement
1055 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
1056 : l_getgroups(8), lst(8), routerstat(8)
1057 Details : included man pages for llobdstat(8), llstat(8),
1058 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
1060 Severity : enhancement
1062 Description: configure's --enable-quota should check the
1063 : kernel .config for CONFIG_QUOTA
1064 Details : configure is terminated if --enable-quota is passed but
1065 : no quota support is in kernel
1069 Description: Remove portals compatibility
1070 Details : Remove portals compatibility, not interoperable with releases
1075 Description: Resolve device initialization race
1076 Details : Prevent proc handler from accessing devices added to the
1077 obd_devs array but yet be intialized.
1079 Severity : enhancement
1081 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
1083 Severity : enhancement
1085 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
1090 Description: Using wrong pointer in osc_brw_prep_request
1091 Details : Access to array[-1] can produce panic if kernel compiled with
1092 CONFIG_PAGE_ALLOC enabled
1096 Description: Client runs out of low memory
1097 Details : Consider only lowmem when counting initial number of llap pages
1101 Description: *optional* service tags registration
1102 Details : if the "service tags" package is installed on a Lustre node
1103 When the filesystem is mounted, a local-node service tag will
1104 be created. See http://inventory.sun.com/ for more information
1105 about the Service Tags asset management system.
1109 Description: Kernel BUG tries to release flock
1110 Details : Lustre does not destroy flock lock before last reference goes
1111 away. So always drop flock locks when client is evicted and
1112 perform unlock regardless of successfulness of speaking to MDS.
1116 Description: add refcount protection for osc callbacks, avoid panic on shutdown
1120 Description: sanity test 65a fails if stripecount of -1 is set
1121 Details : handle -1 striping on filesystem in ll_dirstripe_verify
1126 Description: ASSERTION(CheckWriteback(page,cmd)) failed
1127 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
1130 Severity : enhancement
1132 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
1136 Description: do not process already freed flock
1137 Details : flock can possibly be freed by another thread before it reaches
1138 to ldlm_flock_completion_ast.
1142 Description: LBUG during stress test
1143 Details : Need properly lock accesses the flock deadlock detection list.
1147 Description: oops in page fault handler
1148 Details : kernel page fault handler can return two special 'pages' in error
1149 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
1153 Description: timeout with invalidate import.
1154 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
1155 should be handled by ptlrpcd. This produce long age waiting and
1156 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
1158 Severity : enhancement
1160 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
1164 Description: enable MGS and MDT services start separately
1165 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
1166 a co-located MGS without starting the MGS, which is a complement
1167 to 'nosvc' mount option.
1171 Frequency : after recovery
1172 Description: precreate to many object's after del orphan.
1173 Details : del orphan st in oscc last_id == next_id and this triger growing
1174 count of precreated objects. Set flag LOW to skip increase count
1175 of precreated objects.
1179 Frequency : rare, on clear nid stats
1180 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
1181 Details : when clean nid stats sometimes try destroy live entry,
1182 and this produce panic in free.
1186 Description: Stack overflow during MDS log replay
1187 ease stack pressure by using a thread dealing llog_process.
1191 Description: wait until IO finished before start new when do lock cancel.
1192 Details : VM protocol want old IO finished before start new, in this case
1193 need wait until PG_writeback is cleared until check dirty flag and
1194 call writepages in lock cancel callback.
1196 Severity : enhancement
1198 Description: using special macro for print time and cleanup in includes.
1202 Description: mds_mfd_close() ASSERTION(rc == 0)
1203 Details : In mds_mfd_close(), we need protect inode's writecount change
1204 within its orphan write semaphore to prevent possible races.
1208 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
1212 Frequency : rare, on shutdown ost
1213 Description: don't hit live lock with umount ost.
1214 Details : shrink_dcache_parent can be in long loop with destroy dentries,
1215 use shrink_dcache_sb instead.
1219 Description: don't panic with use echo client
1220 Details : echo client pass NULL as client nid pointer and this produce null
1221 pointer dereference.
1225 Description: fix build on ppc32
1226 Details : compile code with -m64 flag produce wrong object file for ppc32.
1230 Description: add message levels for liblustreapi
1234 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
1235 Details : Move the 'good_osts' check before the 'total_bavail' check. This
1236 will result in an -EAGAIN and in the exit call path we call
1237 alloc_rr() which will with increasing aggressiveness attempt to
1238 aquire precreated objects on the minimum number of required OSCs.
1242 Description: Use old size assignment to avoid deadlock
1243 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
1244 the scheduling while holding a spinlock. We do not need locking
1245 for size in ll_update_inode() because size is only updated from
1246 the MDS for directories or files without objects, so there is no
1247 other place to do the update, and concurrent access to such inodes
1248 are protected by the inode lock.
1252 Description: resolve "_IOWR redefined" build error on SLES10
1256 Description: dump the memory debugging after all modules are unloaded to
1257 suppress false negative in conf_sanity test 39
1259 Severity : enhancement
1261 Description: build kernel-ib packages for OFED 1.3 in our release cycle
1266 Description: fix SLES kernel versioning
1267 Details : the kernel version for our SLES 10 kernel did not include a "-"
1268 before the "smp" at the end. while this was not a problem in
1269 general, it did mean that software trying to use the kernel
1270 version to try to detect a vendor specific kernel would fail.
1271 this was most evident by the OFED build scripts.
1275 Description: Don't update lov_desc members until making sure they are valid
1276 Details : When updating lov_desc members via proc fs, need fix their
1277 validities before doing the real update.
1281 Description: don't put request into delay list while invalidate in flight.
1282 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
1283 import in flight. this produce timeout for invalidate and sometimes
1284 can cause stale data.
1288 Frequency : on ppc only
1289 Description: not convert ost objects for directory because it's not exist.
1290 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
1294 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
1296 Severity : enhancement
1298 Description: Update to support 2.6.22.14 vanilla kernel.
1302 Frequency : rare, on recovery
1303 Description: read procfs can produce deadlock in some situation
1304 Details : Holding lprocfs lock which send rpc can produce block for destroy
1305 obd objects and this also block reconnect with -EALREADY. This
1306 isn't fix all lprocfs bugs - but make it rare.
1308 Severity : enhancement
1310 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
1313 Frequency : frequent on X2 node
1315 Description: mdc_set_open_replay_data LBUG
1316 Details : Set replay data for requests that are eligible for replay.
1320 Description: lustre_mgs: operation 101 on unconnected MGS
1321 Details : When MGC is disconnected from MGS long enough, MGS will evict the
1322 MGC, and late on MGC cannot successfully connect to MGS and a lot
1323 of the error messages complaining that MGS is not connected.
1327 Frequency : on network error
1328 Description: panic with double free request if network error
1329 Details : mdc_finish_enqueue is finish request if any network error ocuring,
1330 but it's true only for synchronus enqueue, for async enqueue
1331 (via ptlrpcd) this incorrect and ptlrpcd want finish request
1334 Severity : enhancement
1336 Description: client-side metadata stat-ahead during readdir(directory readahead)
1337 Details : perform client-side metadata stat-ahead when the client detects
1338 readdir and sequential stat of dir entries therein
1341 Frequency : on start mds
1343 Description: Implement get_info(last_id) in obdfilter.
1346 Frequency : occasional
1348 Description: Correctly check stale fid, not start epoch if ost not support SOM
1349 Details : open with flag O_CREATE need set old fid in op_fid3 because
1350 op_fid2 was overwritten with new generated fid, but mds can answer
1351 with one of these two fids and both is not stale. Setattr
1352 incorrectly started an epoch and assume will be called
1353 done_writing, but without SOM done_writing ever being called.
1356 Frequency : rare, depends on device drivers and load
1358 Description: MDS or OSS nodes crash due to stack overflow
1359 Details : Code changes in 1.8.0 increased the stack usage of some functions.
1360 In some cases, in conjunction with device drivers that use a lot
1361 of stack the MDS (or possibly OSS) service threads could overflow
1362 the stack. One change which was identified to consume additional
1363 stack has been reworked to avoid the extra stack usage.
1366 Frequency : occasional
1368 Description: Do not fail import if osc_interpret_create gets -EAGAIN
1369 Details : If osc_interpret_create got -EAGAIN it immediately exits and
1370 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
1371 oscc_has_objects and see OSC has no objests and call
1372 oscc_internal_create to resend create request.
1374 Severity : enhancement
1376 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
1378 Severity : enhancement
1380 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
1383 Frequency : very rare
1385 Description: Fix replay if there is an un-replied request and open
1386 Details : In some cases, older replay request will revert the
1387 mcd->mcd_last_xid on MDS which is used to record the client's
1388 latest sent request.
1390 Severity : enhancement
1392 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
1394 Severity : enhancement
1396 Description: Add rhel5 support to HEAD.
1398 Serverity : enhancement
1400 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
1405 Description: Don't allow skipping OSTs if index has been specified.
1406 Details : Don't allow skipping OSTs if index has been specified, make
1407 locking in internal create lots better.
1411 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
1412 Details : access to bitfield in structure is always rounded to long
1413 and this produce problem with not atomic change any bit.
1417 Description: Lustre make rpms failed.
1418 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
1419 builds Lustre rpms from tarball.
1422 Frequency : rare on shutdown ost
1424 Description: If llog cancel was not send before clean_exports phase, this can
1425 produce deadlock in llog code.
1426 Details : If llog thread has last reference to obd and call class_import_put
1427 this produce deadlock because llog_cleanup_commit_master wait when
1428 last llog_commit_thread exited, but this never success because was
1429 called from llog_commit_thread.
1433 Description: allow userland application know is lost one of stripes.
1434 Details : fill lvb_blocks with error code on ost and return it to
1435 application if error flag found.
1439 Description: NULL lov_tgts causing MDS oops
1440 Details : more safe checks for NULL lov_tgts for avoid oops.
1442 Severity : enhancement
1444 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
1448 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
1449 Details : If an OST is down the MDS will hang indefinitely in
1450 obd_statfs_rqset() waiting for the statfs data. While for
1451 MDS QOS usage of statfs, it should not stuck in waiting.
1453 Severity : enhancement
1455 Description: remote_acl support
1456 Details : Support ACL-based permission check for remote user.
1457 Support setfacl/getfacl for remote user with the utils
1458 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
1459 the system "{s,g}etfacl" utils.
1461 Severity : enhancement
1463 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
1465 Severity : enhancement
1467 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
1471 Description: make mgs_setparam() handle fsname containing dash
1472 Details : fsname containing a dash does not work with lctl conf_param
1474 Severity : enhancement
1476 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
1478 Severity : enhancement
1480 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
1484 Frequency : rare, at shutdown
1485 Description: access already free / zero obd_namespace.
1486 Details : if client_disconnect_export was called without force flag set,
1487 and exist connect request in flight, this can produce access to
1488 NULL pointer (or already free pointer) when connect_interpret
1489 store ocd flags in obd_namespace.
1493 Frequency : only at startup
1494 Description: not alloc memory with spinlock held.
1495 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
1496 if any spinlock held.
1498 Severity : enhancement
1500 Description: make lustre randomly fail allocating memory
1501 Details : Make lustre randomly failed allocating memory for testing purpose.
1503 Severity : enhancement
1505 Description: lost problems with lov objid file
1506 Details : Fixes some scability and access to not inited memory problems
1507 in work with lov objdid file.
1512 Description: lfs find does not continue on file error
1513 Details : Continue other files processing when a file/dir is absent.
1517 Description: Inconsistent usage of lustre_pack_reply()
1518 Details : Standardize the usage of lustre_pack_reply() such that it
1519 always generate a CERROR on failure.
1522 Frequency : occasional
1524 Description: MDS hang or stay in waiting lock
1525 Details : If client receive lock with CBPENDING flag ldlm need send lock
1526 cancel as separate rpc, to avoid situation when cancel request
1527 can't processed due all i/o threads stay in wait lock.
1531 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1535 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1539 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1543 Description: Update to RHEL4 latest kernel.
1545 Severity : enhancement
1547 Description: Build SLES10 patchless client fails
1548 Details : The configure was broken by run ./configure with
1549 --with-linux-obj=.... argument for patchless client. When the
1550 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1551 can't search header adequately. Use absolute path such as
1552 -I($LINUX)/include instead.
1556 Description: interrupt oig_wait produce painc on resend.
1557 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1558 private set, and this produce situation when rq_ptlrpcd_data not
1559 copyed to new allocated request and triggered LBUG on assert
1560 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1561 ptlrpcd set if request is changed and can be safety changed to use
1566 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1567 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1568 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1569 these macros and replace them with their difinitions instead.
1573 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1574 Details : Update which_patch & target file for SLES10 latest kernel.
1577 Description: Remove client patches from SLES 10 kernel.
1578 Details : This causes SLES 10 clients to behave as patchless clients
1579 even on a Lustre-patched (server) kernel.
1581 Severity : enhancement
1583 Description: self-adjustable client's lru lists
1584 Details : use adaptive algorithm for managing client cached locks lru
1585 lists according to current server load, other client's work
1586 pattern, memory activities, etc. Both, server and client
1587 side namespaces provide number of proc tunables for controlling
1590 Severity : enhancement
1592 Description: light-weight GSS support
1593 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
1594 and reduce performance overhead.
1596 Severity : enhancement
1598 Description: Linux keyring support
1599 Details : Support using service of Linux keyring for Lustre GSS internal
1600 context refresh/cache mechanism.
1604 Description: Fix errors in lfs documentation
1605 Details : Fixes man pages
1609 Description: don't use GFP_* in generic Lustre code.
1610 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
1614 Description: obdclass is limited by single OBD_ALLOC(idarray)
1615 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
1619 Description: warnings with build patchless client with vanila 2.6.19 and up
1620 Details : change old ctl_table style and replace ctl_table/ctl_table_header
1621 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
1625 Frequency : Only for SLES
1626 Description: sanity_quota fail test_1
1627 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
1628 file, which makes TSTID[2] inunique.
1632 Frequency : Only for SLES9
1633 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
1634 Details : Update target/ChangeLog/which_patch .
1638 Frequency : Only for test_10 in sanity-quota.sh
1639 Description: a bug in quota test code
1640 Details : checking a test flag in a wrong place
1644 Frequency : Only for test_13 in sanity-quota.sh
1645 Description: a bug in quota test script
1646 Details : shouldn't check overfull allocated quota
1650 Description: jbd statistics
1651 Details : Port older jbd statistics patch for sles10
1653 Severity : enhancement
1655 Description: Add printing inode info into message about error in writepage.
1660 Description: reply_lock_interpret crash due to race with it and lock cancel.
1661 Details : Do not replay locks that are being cancelled. Do not reference
1662 locks by their address during replay, just by their handle.
1666 Frequency : When flocks are used.
1667 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
1668 Details : Flock locks might destroy just granted lock if it could be merged
1669 with another existing flock, this is done in completion handler,
1670 so teach ldlm_cli_enquque_fini that this is a valid case for
1676 Description: Oops in read and write path when failing to allocate lock.
1677 Details : Check if lock allocation failed and return error back.
1681 Description: lstripe command fails for valid OST index
1682 Details : The stripe offset is compared to lov->desc.ld_tgt_count
1683 instead of lov->desc.ld_active_tgt_count.
1687 Description: sanity.sh failed test 103
1688 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
1693 Description: df doesn't work properly if diskfs blocksize != 4K
1694 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
1698 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
1699 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
1701 Severity : enhancement
1703 Description: Adaptive timeouts
1704 Details : RPC timeouts adapt to changing server load and network
1705 conditions to reduce resend attempts and improve recovery time.
1709 Description: llapi_file_create() does not allow some changes
1710 Details : add llapi_file_open() that allows specifying the mode and
1711 open flags, and also returns an open file handle.
1715 Description: merge and cleanup kernel patches.
1716 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
1720 Description: Add journal checksum support.(Kernel part)
1721 Details : The journal checksum feature adds two new flags i.e
1722 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
1723 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
1724 indicates that the commit block contains the checksum for
1725 the blocks described by the descriptor blocks. Now commit
1726 record can be sent to disk without waiting for descriptor
1727 blocks to be written to disk. This behavior is controlled
1728 using JBD2_FEATURE_ASYNC_COMMIT flag.
1732 Description: OSS needs mutliple precreate threads
1733 Details : Add ability to start more than one create thread per OSS.
1737 Description: Sanity.sh test_65k failed: setstripe should have succeeded
1738 Details : execute lfs setstripe on client
1742 Description: mds_obd_create error creating tmp object
1743 Details : When the user sets quota on root, llog will be affected and can't
1744 create files and write files.
1747 Frequency : Always on ia64 patchless client, and possibly others.
1749 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
1750 Details : This allows the patchless client to be loaded on architectures
1751 without this export.
1755 Description: RedHat Update kernel for RHEL5
1756 Details : Add the kernel config file for RHEL5.
1760 Description: RedHat Update kernel for RHEL5
1761 Details : Modify the kernel config file more closer RHEL5.
1765 Description: Build failure against Centos5 (RHEL5)
1766 Details : Define PAGE_SIZE when it isn't present.
1770 Description: test_7 on sanity-quota.sh can't be used on separate machines
1771 Details : add facet to handle it
1775 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
1776 Details : with NFS, the anon dentry's parent was set to itself in
1777 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
1778 corresponding dentry other than use rec->ur_name.
1780 Severity : enhancement
1782 Description: lfs setstripe enhancement
1783 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
1787 Description: enable data checksumming by default
1788 Details : enable checksum by default, allow --disable-checksum
1789 configure option and "-o nochecksum" mount option. Checksums
1790 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
1794 Description: lustre support for RHEL5
1795 Details : Add support for RHEL5.
1799 Description: Client eviction due to failover config
1800 Details : after a connection loss, the lustre client should attempt to
1801 reconnect to the last active server first before trying the
1802 other potential connections.
1806 Description: when mds and osts use different quota unit(32bit and 64bit),
1807 quota will be released repeatly.
1808 Details : void sending multiple quota reqs to mds, which will keep the
1809 status between the reqs.
1813 Description: osts not allocated evenly to files
1814 Details : change the condition to increase offset_idx
1817 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
1819 Description: Data corruption for OSTs that are formatted larger than 2TB
1821 Details : When generating the bio request for lustre file writes the
1822 sector number would overflow a temporary variable before being
1823 used for the IO. The data reads correctly from Lustre (which
1824 will overflow in a similar manner) but other file data or
1825 filesystem metadata may be corrupted in some cases.
1829 Description: Tune the kernel for good SCSI performance.
1830 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
1831 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
1836 Description: rewrite ext2-derived code in obdclass/uuid.c
1837 Details : rewrite inherited code (uuid parsing code from ext2 utils)
1838 from scratch preserving functionality.
1842 Description: "lfs find -obd UUID" prints directories
1843 Details : "lfs find -obd UUID" will return all directory names instead
1844 of just file names. It is incorrect because the directories
1845 do not reside on the OSTs.
1849 Description: use i_size_read and i_size_write in 2.6 port
1850 Details : replace inode->i_size access with i_size_read/write()
1853 Frequency : when removing large files
1855 Description: scheduling issue during removal of large Lustre files
1856 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
1857 It causes scheduling issues when removing large files (17TB in the
1861 Frequency : only with liblustre clients on XT3
1863 Description: evictions taking too long
1864 Details : allow llrd to evict clients directly on OSTs
1867 Frequency : only on ppc
1869 Description: /proc/fs/lustre/devices broken on ppc
1870 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
1871 We should make sure the type of 'index' is loff_t and then cast
1872 explicitly as needed below. Do not assign an explicitly cast
1876 Frequency : only for rhel5
1878 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
1879 Details : Modify the target file & which_kernel.
1881 Severity : enhancement
1883 Description: omit set fsid for export NFS
1884 Details : fix set/restore device id for avoid EMFILE error and mark lustre
1885 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
1889 Frequency : Always, for kernels after 2.6.16
1890 Description: Fix warning idr_remove called for id=.. which is not allocated.
1891 Details : Last kernels save old s_dev before kill super and not allow
1892 to restore from callback - restore it before call kill_anon_super.
1896 Description: buffer overruns could theoretically occur
1897 Details : llapi_semantic_traverse() modifies the "path" argument by
1898 appending values to the end of the origin string, and a
1899 overrun may occur. Adding buffer overrun check in liblustreapi.
1903 Description: Fix error on 'ls .' at the top of the Lustre mount.
1904 Details : Don't revalidate dentry if it is a root dentry.
1908 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
1909 Details : Modify vm-tunables-rhel4.patch.
1913 Description: Kernel config for 2.6.18-vanilla.
1914 Details : Modify targets/2.6-vanilla.target.in.
1915 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
1916 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
1917 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
1918 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
1922 Description: improve handling recoverable errors
1923 Details : if request processig with error which can be recoverable on server
1924 request should be resend, otherwise page released from cache and
1929 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
1930 Details : Modify target file & which_patch.
1931 A flaw was found in the IA32 system call emulation provided
1932 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
1933 value could be stored in the %RAX register, which could trigger an
1934 out-of-bounds system call table access. An untrusted local user
1935 could exploit this flaw to run code in the kernel
1936 (ie a root privilege escalation). (CVE-2007-4573).
1940 Description: O_DIRECT bypasses client statistics.
1941 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
1942 read_ahead_stats were updated. copied the stats section from
1943 osc_send_oap_rpc() into async_internal().
1947 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
1951 Description: change order of libsysio includes
1952 Details : '#include sysio.h' should always come before '#include xtio.h'
1956 Description: handle "serious error: objid * already exists" more gracefully
1957 Details : If LAST_ID value on disk is smaller than the objects existing in
1958 the O/0/d* directories, it indicates disk corruption and causes an
1959 LBUG(). If the object is 0-length, then we should use the existing
1960 object. This will help to avoid a full fsck in most cases.
1962 Severity : enhancement
1964 Description: adapt the lustre_config script to support the upgrade case
1965 Details : Add "-u" option for lustre_config script to support upgrading 1.4
1966 server targets to 1.6 in parallel.
1970 Description: To avoid grant space > avaible space when the disk is almost
1971 full. Without this patch you might see the error "grant XXXX >
1972 available" or some LBUG about grant, when the disk is almost
1974 Details : In filter_check_grant, for non_grant cache write, we should
1975 check the left space by if (*left > ungranted + bytes), instead
1976 of (*left > ungranted), because only we are sure the left space
1977 is enough for another "bytes", then the ungrant space should be
1978 increase. In client, we should update cl_avail_grant only there
1979 is OBD_MD_FLGRANT in the reply.
1983 Description: Update RHEL 4 kernel to fix local root privilege escalation.
1984 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
1985 described in CVE-2007-4573. This problem could allow untrusted
1986 local users to gain root access.
1989 Frequency : when using O_DIRECT and quotas
1991 Description: Incorrect file ownership on O_DIRECT output files
1992 Details : block usage reported by 'lfs quota' does not take into account
1993 files that have been written with O_DIRECT.
1998 Description: touch file failed when fs is not full
1999 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
2000 otherwise we can get ENOSP while fs is not full.
2004 Description: parallel lock callbacks
2005 Details : Instead of sending blocking and completion callbacks as separated
2006 requests, adding them to a set and sending in parallel.
2009 Frequency : only for Cray XT3
2010 Bugzilla : 12829/13455
2011 Description: Changing primary group doesn't change the group lustre assigns to
2013 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
2014 overridden with the primary group provided by the group upcall,
2015 whereas the supplied fsgid can be trusted if it is in the list of
2016 supplementary groups returned by the group upcall.
2018 Severity : enhancement
2020 Description: Allow masking D_WARNING, D_ERROR messages from console
2021 Details : Console messages can now be disabled via lnet.printk.
2025 Description: User code with malformed file open parameter crashes client node
2026 Details : Before packing join_file req, all the related reference should be
2027 checked carefully in case some malformed flags cause fake
2028 join_file req on client.
2032 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
2033 Details : ldlm_completion_ast() assumes that a lock is granted when the req
2034 mode is equal to the granted mode. However, it should also check
2035 that LDLM_FL_CP_REQD is not set.
2039 Description: Heavy nfs access might result in deadlocks
2040 Details : After ELC code landed, it is now improper to enqueue any mds
2041 locks under och_sem, because enqueue might want to decide to
2042 cancel open locks for same inode we are holding och_sem for.
2046 Description: Client eviction while running blogbench
2047 Details : A lot of unlink operations with concurrent I/O can lead to a
2048 deadlock causing evictions. To address the problem, the number of
2049 oustanding OST_DESTROY requests is now throttled to
2050 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
2051 callbacks are processed in priority.
2055 Description: enable ACLs on MDS by default
2056 Details : ACLs must be enabled on MDS by default.
2059 Frequency : PPC/PPC64 only
2061 Description: conflicts between asm-ppc64/types.h and lustre_types.h
2062 Details : fix duplicated definitions between asm-ppc64/types.h and
2063 lustre_types.h on PPC.
2066 Frequency : PPC/PPC64 only
2068 Description: asm-ppc/segment.h does not exist
2069 Details : fix compile issue on PPC.
2073 Description: better handle error messages in extents code
2076 Frequency : RHEL4 only
2078 Description: mkfs is very slow on IA64/RHEL4
2079 Details : A performance regression has been discovered in the MPT Fusion
2080 driver between versions 3.02.73rh and 3.02.99.00rh. As a
2081 consequence, we have downgraded the MPT Fusion driver in the RHEL4
2082 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
2084 Severity : enhancement
2086 Description: SNMP support enhancement
2087 Details : Adding total number of sampled request for an MDS node in snmp
2090 Severity : enhancement
2092 Description: Optimize ldlm waiting list processing for PR extent locks
2093 Details : When processing waiting list for read extent lock and meeting read
2094 lock that is same or wider to it that is not contended, skip
2095 processing rest of the list and immediatelly return current
2096 status of conflictness, since we are guaranteed there are no
2097 conflicting locks in the rest of the list.
2101 Description: Time out and refuse to reconnect
2102 Details : When the failover node is the primary node, it is possible
2103 to have two identical connections in imp_conn_list. We must
2104 compare not conn's pointers but NIDs, otherwise we can defeat
2105 connection throttling.
2109 Description: port llog fixes from b1_6 into HEAD
2110 Details : Port llog reference couting and some llog cleanups from b1_6
2111 (bug 10800) into HEAD, for protect from panic and access to already
2112 free llog structures.
2116 Description: Detect stride IO mode in read-ahead
2117 Details : When a client does stride read, read-ahead should detect that and
2118 read-ahead pages according to the detected stride pattern.
2122 Description: data checksumming impacts single node performance
2123 Details : add support for several checksum algorithm. Currently, only CRC32
2124 and Adler-32 are supported. The checksum type can be changed on
2125 the fly via /proc/fs/lustre/osc/*/checksum_type.
2129 Description: use adler32 for page checksums
2130 Details : when available, use the Adler-32 algorithm instead of CRC32 for
2135 Description: build for x2 fails
2136 Details : fix compile issue on Cray systems.
2140 Description: Properly match for duplicate locks
2141 Details : Due to different lock order from skiplists code, we need to
2142 traverse entire list for now
2145 Frequency : only on PPC/SLES10
2147 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
2148 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
2149 be defined. Add a hack in mkfs_lustre.c to work around this compile
2154 Description: LASSERT on MDS when client holding flock lock dies
2155 Details : ldlm pool logic depends on number of granted locks equal to
2156 number of released locks which is not true for flock locks, so
2157 just exclude such locks from consideration.
2161 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
2162 Details : Use fsfilt_sync() for both the whole filesystem sync and
2163 individual file sync to eliminate dangerous inode locking
2164 with I_LOCK that can lead to a deadlock.
2168 Description: performance in 1.6.3
2169 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
2170 sure that 1MB requests are not fragmented by the block layer.
2172 Severity : enhancement
2174 Description: organize the server-side client stats on per-nid basis
2175 Details : Change the structure of stats under obdfilter and mds to
2183 The "uuid"s file would list the uuids of _active_ exports.
2184 And the clear entry is to clear all stats and stale nids.
2186 Severity : enhancement
2188 Description: eliminate client locks in face of contention
2189 Details : file contention detection and lockless i/o implementation
2190 for contended files.
2194 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
2197 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
2199 Description: MDS may lose file striping (and hence file data) in some cases
2200 Details : If there are additional extended attributes stored on the MDS,
2201 in particular ACLs, SELinux, or user attributes (if user_xattr
2202 is specified for the client mount options) then there is a risk
2203 of attribute loss. Additionally, the Lustre file striping
2204 needs to be larger than default (e.g. striped over all OSTs),
2205 and an additional attribute must be stored initially in the
2206 inode and then increase in size enough to be moved to the
2207 external attribute block (e.g. ACL growing in size) for file
2212 Description: skiplist implementation simplification
2213 Details : skiplists are used to group compatible locks on granted list
2214 that was implemented as tracking first and last lock of each
2215 lock group the patch changes that to using doubly linked lists
2219 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
2220 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
2221 because LASSERT on both the data supplied by a client, and the
2222 data on disk is dangerous and incorrect.
2224 Severity : enhancement
2226 Description: Slow truncate/writes to huge files at high offsets.
2227 Details : Directly associate cached pages to lock that protect those pages,
2228 this allows us to quickly find what pages to write and remove
2229 once lock callback is received.
2233 Description: more ldlm soft lockups
2234 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
2235 starve other threads from the resource lock for a long time in
2236 case of long waiting queue, so change the debug level from
2237 D_OTHER to the less frequently used D_INFO.
2239 Severity : enhancement
2241 Description: add -gid, -group, -uid, -user options to lfs find
2245 Description: Hung threads in invalidate_inode_pages2_range
2246 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
2247 once one is completed. As a result, some RPCs are stuck in the
2248 queue and are never sent.
2252 Description: filter threads hungs on waiting journal commit
2253 Details : Cleanup filter group llog code, then only filter group llog will
2254 be only created in the MDS/OST syncing process.
2258 Description: Procfs and llog threads access destoryed import sometimes.
2259 Details : Sync the import destoryed process with procfs and llog threads by
2260 the import refcount and semaphore.
2262 Severity : enhancement
2264 Description: openlock cache of b1_6 port to HEAD
2269 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
2270 Details : If insertion of an extent fails, then discard the inode
2271 preallocation and free data blocks else it can lead to duplicate
2276 Description: don't always update ctime in ext3_xattr_set_handle()
2277 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
2278 In some cases the ctime should not be updated, for example for
2279 2.0->1.8 compatibility it is necessary to delete an xattr and it
2280 should not update the ctime.
2284 Bugzilla : 15713/16362
2285 Description: Assertion in iopen_connect_dentry in 1.6.3
2286 Details : looking up an inode via iopen with the wrong generation number can
2287 populate the dcache with a disconneced dentry while the inode
2288 number is in the process of being reallocated. This causes an
2289 assertion failure in iopen since the inode's dentry list contains
2290 both a connected and disconnected dentry.
2294 Description: assertion failure in ldlm_handle2lock()
2295 Details : fix a race between class_handle_unhash() and class_handle2object()
2296 introduced in lustre 1.6.5 by bug 13622.
2301 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
2302 Details : remove the limit of 256 scsi disks in the sd_iostat patch
2307 Description: oops in sd_iostats_seq_show()
2308 Details : unloading/reloading the scsi low level driver triggers a kernel
2309 bug when trying to access the sd iostat file.
2314 Description: Kernel panics during QLogic driver reload
2315 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
2316 patch, causing memory corruption.
2321 Description: journal_dev option does not work in b1_6
2322 Details : pass mount option during pre-mount.
2324 Severity : enhancement
2326 Description: Add a FIEMAP(FIle Extent MAP) ioctl
2327 Details : FIEMAP ioctl will allow an application to efficiently fetch the
2328 extent information of a file. It can be used to map logical blocks
2329 in a file to physical blocks in the block device.
2333 Description: LDLM soft lockups - improvement
2334 Details : It is be possible to send the lock handle along with each read
2335 or write request because the client is already doing a lock match
2336 itself so there isn't any reason the OST should have to re-do that
2340 Frequency : only on Cray X2
2342 Description: X2 build failures
2343 Details : fix build failures on Cray X2.
2347 Description: xid & resent requests
2348 Details : Initialize RPC XID from clock at startup (randomly if clock is
2351 Severity : enhancement
2353 Description: Add lustre_start utility to start or stop multiple Lustre servers
2358 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
2359 Details : In case of memory pressure, list_del() can be called twice on
2360 req->rq_history_list, causing a kernel oops.
2364 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
2365 Details : fix stack overflow in the distributed lock manager by defering
2366 export eviction after a failed AST to the elt thread instead of
2367 handling it in the dlm interpret routine.
2371 Description: Convert some comments to new format.
2372 Details : Update documenting comments to match doxygen conventions.
2376 Description: Grammar fixes.
2377 Details : A couple of trivial spelling fixes.
2381 Description: OSD_COUNTERS-mandatory
2382 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
2383 to simplify the code.
2387 Description: simplify cmm_device freeing logic.
2388 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
2392 Description: Add lockdep support to dt_object_operations locking interface.
2393 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
2394 parameter indicating lock ordering. Update mdd code to use new
2399 Description: Introduce failloc constants for lockless IO tests.
2400 Details : Add two new failloc constants to test lockless IO. Only one of
2401 them in implemented---another is checked in yet to be landed
2406 Description: Add lockdep support for inode mutex.
2407 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
2408 in the situations where Lustre has to lock more than one inode
2413 Description: Add optional invariants checking support.
2414 Details : Add new LINVRNT() macro, optional on new --enable-invariants
2415 configure switch. This macro is to be used for consistency and
2416 sanity checks that are too expensive to be left in `production'
2421 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
2422 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
2427 Description: Add lockdep support for ldlm_lock and ldlm_resource.
2428 Details : Use spin_lock_nested() in (the only) situation where more than
2429 one ldlm_lock is locked simultaneously. Also, fix possible
2430 dead-lock in ldlm_lock_change_resource() by enforcing particular
2435 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
2436 Details : Instead of specifying each ldlm_lock call-back through separate
2437 parameter, wrap them into struct ldlm_callback_suite.
2441 Description: Kill join_lru obd method and its callers.
2442 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
2443 in memory---a requirement implemented through ->o_join_lru() obd
2444 method in HEAD. Remove this method and its users.
2448 Description: Add asynchronous ldlm ENQUEUE completion handler.
2449 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
2450 case that stock ldlm_completion_ast() cannot handle as it waits
2451 until lock is granted. Introduce new ldlm_completion_ast_async()
2452 for this. Also comment ldlm_completion_ast().
2456 Description: ldlm_error <-> errno conversion.
2457 Details : Add functions to map (rather arbitrary) between LDLM error codes
2458 and standard errno values. CLIO needs this to prevent LDLM specific
2459 constants from escaping ldlm and osc.
2463 Description: Kill unused ldlm_handle2lock_ns() function.
2464 Details : Kill unused ldlm_handle2lock_ns() function.
2468 Description: Add lu_ref support to ldlm_lock
2469 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
2470 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
2471 and ldlm_resource. LDLM interface has to be changed, because code that
2472 releases a reference on a lock, has to "know" what reference this is.
2473 In the most frequent case
2475 lock = ldlm_handle2lock(handle);
2477 LDLM_LOCK_PUT(lock);
2479 no changes are required. When any other reference (received _not_ from
2480 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
2481 instead of LDLM_LOCK_PUT().
2483 Arguably, changes are pervasive, and interface requires some discipline
2484 for proper use. On the other hand, it was very instrumental in finding
2485 a few leaked lock references.
2489 Description: Add ldlm_lock_addref_try().
2490 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
2491 attempts to addref a lock that might be being canceled
2496 Description: Add ldlm_weigh_callback().
2497 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
2498 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
2499 hard-coded `number of pages' logic.
2503 Description: Add lockdep annotations to llog code.
2504 Details : Use appropriately tagged _nested() locking calls in the places
2505 where llog takes more than one ->lgh_lock lock.
2509 Description: Add loi_kms_set().
2510 Details : Wrap kms updates into a helper function.
2514 Description: Constify instances of struct lsm_operations.
2515 Details : Constify instances of struct lsm_operations.
2519 Description: lu_conf support.
2520 Details : On a server, a file system object is uniquely identified
2521 by a fid, which is sufficient to locate and load all object
2522 state (inode). On a client, on the other hand, more data are
2523 necessary instantiate an object. Change lu_object_find() and
2524 friends to take additional `lu_conf' argument describing object.
2525 Typically this includes layout information.
2529 Description: lu_context fixes.
2530 Details : Introduce new lu_context functions that are needed on the client
2531 side, where some system threads (ptlrpcd) are shared by multiple
2532 modules, and so cannot be stopped during module shutdown.
2536 Description: Add start and stop methods to lu_device_type_operations.
2537 Details : Introduce two new methods in lu_device_type_operations, that are
2538 invoked when first instance of a given type is created and last one
2539 is destroyed respectively. This is need by CLIO.
2543 Description: Add lu_ref support to struct lu_device.
2544 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
2545 track leaked references.
2549 Description: Introduce lu_kmem_descr.
2550 Details : lu_kmem_descr and its companion interface allow to create
2551 and destroy a number of kmem caches at once.
2555 Description: Fix lu_object finalization race.
2556 Details : Fix a race between lu_object_find() finding an object and its
2557 concurrent finalization. This race is (most likely) not possible
2558 on the server, but might happen on the client.
2562 Description: Introduce lu_ref interface.
2563 Details : lu_ref is a debugging module allowing to track references to
2564 a given object. It is quite cpu expensive, and has to be
2565 explicitly enabled with --enable-lu_ref. See usage description
2570 Description: Factor lu_site procfs stats into a separate function.
2571 Details : Separate lu_site stats printing code into a separate function
2572 that can be reused on a client.
2576 Description: Constify instances of struct {lu,dt,md}_device_operations.
2577 Details : Constify instances of struct {lu,dt,md}_device_operations.
2581 Description: Introduce struct md_site and move meta-data specific parts of
2582 struct lu_site here.
2583 Details : Move md-specific fields out of struct lu_site into special struct
2584 md_site, so that lu_site can be used on a client.
2588 Description: Kill mdd_lov_destroy().
2589 Details : Remove unused mdd code.
2593 Description: Add st_block checking to multistat.c.
2594 Details : Add st_block checking to multistat.c.
2598 Description: Add lu_ref support to struct obd_device.
2599 Details : Add lu_ref tracking to obd_device.
2603 Description: Kill obd_set_fail_loc().
2604 Details : Remove unused code.
2608 Description: Add special type for ptlrpc_request interpret functions.
2609 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
2610 there. Actual usage will be in CLIO.
2614 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
2615 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
2616 doesn't work with lockdep.
2620 Description: Add rwv.c test program.
2621 Details : New testing program exercising readv(2) and writev(2) (Qian).
2625 Description: Add sendfile.c test program.
2626 Details : New testing program exercising sendfile(2) (Jay).
2630 Description: Ratelimit a message that can be very frequent.
2631 Details : Ratelimit a memory allocation failure message that can
2636 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2637 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2641 Description: Liblustre build fixes.
2642 Details : Liblustre build fixes.
2646 Description: libcfs: add cfs_{need,cond}_resched() interface.
2647 Details : libcfs: add cfs_{need,cond}_resched() definition and
2648 implementations for Linux, NT, and liblustre.
2650 Severity : enhancement
2652 Description: More exported tunables for mballoc
2653 Details : Add support for tunable preallocation window and new tunables for
2654 large/small requests
2658 Description: Detect corruption of block bitmap and checking for preallocations
2659 Details : Checks validity of on-disk block bitmap. Also it does better
2660 checking of number of applied preallocations. When corruption is
2661 found, it turns filesystem readonly to prevent further corruptions.
2665 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
2666 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
2667 window to 0. If it does hit cache miss, and read-pattern is still
2668 stride-io mode, does not reset the stride window, but also does not
2669 increase the stride window length in this case.
2673 Frequency : only for big-endian servers
2674 Description: Check if system is big-endian while mounting fs with extents feature
2675 Details : Mounting a filesystem with extents feature will fail on big-endian
2676 systems since ext3-based ldiskfs is not supported on big-endian
2677 systems. This can be over-riden with "bigendian_extents" mount option.
2679 Severity : enhancement
2681 Description: The root squash functionality
2682 Details : A security feature, which is to prevent users from being able
2683 to mount lustre on their desktop, run as root, and delete
2684 all of the files in the filesystem. The goal is accomplished by
2685 remapping user id (UID) and group id (GID) of the root user to
2686 a UID and GID specified by the system administartor via Lustre
2687 configuration management server (MGS). The functionality also
2688 allows to specify sets of clients for which the remapping does
2693 Description: Excessive recovery window
2694 Details : With AT enabled, the recovery window can be excessively long (6000+
2695 seconds). To address this problem, we no longer use
2696 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
2697 timeout no longer depends on the service time, it is set to
2698 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
2699 time via pb_service_time.
2703 Description: Watchdog triggered on MDS failover
2704 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
2705 the OSTs know that the MDS "UUID" can be reused for the same export
2706 from a different NID, so we do not need to wait for the export to be
2710 Frequency : rare, only if using MMP with Linux RAID
2712 Description: MMP doesn't work with Linux RAID
2713 Details : While using HA for Lustre servers with Linux RAID, it is possible
2714 that MMP will not detect multiple mounts. To make this work we
2715 need to unplug the device queue in RAID when the MMP block is being
2716 written. Also while reading the MMP block, we should read it from
2717 disk and not the cached one.
2719 Severity : enhancement
2721 Description: open file using fid
2722 Details : A file can be opened using just its fid, like
2723 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
2726 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
2729 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
2730 Details : add proper locking for creation of jbd_4k slab cache
2734 Description: MMP check in ext3_remount() fails without displaying any error
2735 Details : When multiple mount protection fails during remount, proper error
2738 Severity : enhancement
2740 Description: Allow stripe size to be up to 4G-64k
2741 Details : Fix math logic to allow large stripe sizes.
2745 Description: add check for >8TB ldiskfs filesystems
2746 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
2747 Don't allow >8TB ldiskfs filesystems to be mounted without
2748 force_over_8tb mount option
2753 Descriptoin: OSS complains "too many blocks" and crashes
2754 Details : In filter_commitrw_write():
2755 1. lquota_chkquota();
2756 2. obdo_from_inode();
2757 3. lquota_pending_commit();
2758 "2." has changed uid and gid in this case, then caused
2759 lquota_chkquota() and lquota_pending_commit() to use
2760 different uid and gid. That is the root of the bug.
2764 Descriptoin: getstripe doesn't report stripe count = -1 sanely
2765 Details : Detect and print out the 'stripe count = -1' case properly.
2769 Description: Out or order replies might be lost on replay
2770 Details : In ptlrpc_retain_replayable_request if we cannot find retained
2771 request with tid smaller than one currently being added, add it
2772 to the start, not end of the list.
2774 --------------------------------------------------------------------------------
2776 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
2778 * Support for kernels:
2779 2.6.5-7.283 (SLES 9),
2780 2.6.9-55.EL (RHEL 4),
2781 2.6.16.46-0.14 (SLES 10),
2782 2.6.18.8 vanilla (kernel.org)
2783 * Client support for unpatched kernels:
2784 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2785 2.6.16 - 2.6.22 vanilla (kernel.org)
2786 * Due to recently discovered recovery problems, we do not recommend
2787 using patchless RHEL 4 clients with this or any earlier release.
2788 * Recommended e2fsprogs version: 1.39.cfs8
2789 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2790 * Starting with this release, the ldiskfs backing filesystem required
2791 by Lustre is now in its own package, lustre-ldiskfs. This package
2792 should be installed. It is versioned separately from Lustre and
2793 may be released separately in future.
2797 Description: block reactivating mgc import until all deactivates complete
2798 Details : Fix race when failing back MDT/MGS to itself (testing)
2801 Frequency : at statup only
2803 Description: mds_lov_synchronize race leads to various problems
2804 Details : simultaneous MDT->OST connections at startup can cause the
2805 sync to abort, leaving the OSC in a bad state.
2807 Severity : enhancement
2809 Description: add optional extra BUILD_VERSION info
2810 Details : add a new environment variable (namely LUSTRE_VERS) which allows
2811 to override the lustre version.
2814 Frequency : 2.6.18 servers only
2816 Description: ll_kern_mount() doesn't release the module reference
2817 Details : The ldiskfs module reference count never drops down to 0
2818 because ll_kern_mount() doesn't release the module reference.
2823 Description: server LBUG when using old ost_num_threads parameter
2824 Details : Accept the old ost_num_threads parameter but warn that it
2825 is deprecated, and fix an off-by-one error that caused an LBUG.
2830 Description: Transient SCSI error results in persistent IO issue
2831 Details : iobuf->dr_error is not reinitialized to 0 between two
2835 Frequency : sometimes when underlying device returns I/O errors
2837 Description: OSTs not going read-only during write failures
2838 Details : OSTs are not remounted read-only when the journal commit threads
2839 get I/O errors because fsfilt_ext3 calls journal_start/stop()
2840 instead of the ext3 wrappers.
2844 Description: poor connect scaling with increasing client count
2845 Details : Don't run filter_grant_sanity_check for more than 100 exports
2846 to improve scaling for large numbers of clients.
2849 Frequency : SLES10 only
2851 Description: sanity-quota.sh quotacheck failed: rc = -22
2852 Details : Quotas cannot be enabled on SLES10.
2855 Frequency : liblustre clients only
2857 Description: getdirentries does not give error when run on compute nodes
2858 Details : getdirentries does not fail when the size specified as an argument
2859 is too small to contain at least one entry
2861 Severity : enhancement
2863 Description: Add LNET router traceability for debug purposes
2864 Details : If a checksum failure occurs with a router as part of the
2865 IO path, the NID of the last router that forwarded the bulk data
2866 is printed so it can be identified.
2871 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
2872 Details : Due to a race condition, liblustre clients were occasionally
2873 evicted incorrectly.
2875 Severity : enhancement
2877 Description: lfs setstripe use optional parameters instead of postional
2880 Severity : enhancement
2882 Description: Nanosecond timestamp support for ldiskfs
2883 Details : The on-disk ldiskfs filesystem has added support for nanosecond
2884 resolution timestamps. There is not yet support for this at
2885 the Lustre filesystem level.
2888 Frequency : during server recovery
2890 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
2891 Details : request with rq_no_resend flag not awake l_wait_event if they get
2895 Frequency : nfs export on patchless client
2897 Description: connectathon hang when test nfs export over patchless client
2898 Details : Disconnected dentry cannot be found with lookup, so we do not need
2899 to unhash it or make it invalid
2902 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
2903 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
2904 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
2909 Description: open req refcounting wrong on reconnect
2910 Details : If reconnect happened between getting open reply from server and
2911 call to mdc_set_replay_data in ll_file_open, we will schedule
2912 replay for unreferenced request that we are about to free.
2913 Subsequent close will crash in variety of ways.
2914 Check that request is still eligible for replay in
2915 mdc_set_replay_data().
2920 Description: disable writes to filesystem when reading health_check file
2921 Details : the default for reading the health_check proc file has changed
2922 to NOT do a journal transaction and write to disk, because this
2923 can cause reads of the /proc file to hang and block HA state
2924 checking on a healthy but otherwise heavily loaded system. It
2925 is possible to return to the previous behaviour during configure
2926 with --enable-health-write.
2928 Severity : enhancement
2930 Description: 64-bit inode version
2931 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
2932 to the inode. This will be required for version-based recovery.
2937 Description: MDS fails to start if a duplicate client export is detected
2938 Details : in some rare cases it was possible for a client to connect to
2939 an MDS multiple times. Upon recovery the MDS would detect this
2940 and fail during startup. Handle this more gracefully.
2942 Severity : enhancement
2944 Description: Add -o localflock option to simulate old noflock
2946 Details : This will achieve local-only flock/fcntl locks
2952 Description: log_commit_thread vs filter_destroy race leads to crash
2953 Details : Take import reference before releasing llog record semaphore
2958 Description: Wrong request locking in request set processing
2959 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
2960 imp_delayed_list, in this place should be used imp_lock.
2963 Frequency : when reconnection
2965 Description: Grant Leak when osc reconnect to OST
2966 Details : When osc reconnect ost, OST(filter) should check whether it
2967 should grant more space to client by comparing fed_grant and
2968 cl_avail_grant, and return the granted space to client instead
2969 of "new granted" space, because client will call osc_init_grant
2970 to update the client grant space info.
2973 Frequency : when client reconnect to OST
2975 Description: Grant Leak when osc do resend and replay bulk write
2976 Details : When osc reconnect to OST, OST(filter)should clear grant info of
2977 bulk write request, because the grant info while be sync between
2978 OSC and OST when reconnect, and we should ignore the grant info
2979 these of resend/replay write req.
2984 Description: Grant space more than available left space sometimes.
2985 Details : When then OST is about to be full, if two bulk writing from
2986 different clients came to OST. Accord the avaliable space of the
2987 OST, the first req should be permitted, and the second one
2988 should be denied by ENOSPC. But if the seconde arrived before
2989 the first one is committed. The OST might wrongly permit second
2990 writing, which will cause grant space > available space.
2993 Frequency : when client is evicted
2995 Description: Grant might be wrongly erased when osc is evicted by OST
2996 Details : when the import is evicted by server, it will fork another
2997 thread ptlrpc_invalidate_import_thread to invalidate the
2998 import, where the grant will be set to 0. While the original
2999 thread will update the grant it got when connecting. So if
3000 the former happened latter, the grant will be wrongly errased
3001 because of this race.
3006 Description: Checking Stale with correct fid
3007 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
3008 to check whether it is stale, because sometimes, we want the
3009 enqueue happened anyway, and op_data.fid2 will not be initialized.
3011 Severity : enhancement
3013 Description: update patchless client
3014 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
3017 Frequency : only with 2.4 kernel
3019 Description: random memory corruption
3020 Details : size of struct ll_inode_info is to big for union inode.u and this
3021 can be cause of random memory corruption.
3026 Description: Memory leak in recovery
3027 Details : Lov_mds_md was not free in an error handler in mds_create_object.
3028 It should also check obd_fail before fsfilt_start, otherwise if
3029 fsfilt_start return -EROFS,(failover mds during mds recovery).
3030 then the req will return with repmsg->transno = 0 and rc = EROFS.
3031 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
3032 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
3033 be freed no matter whether fsfilt_commit success or not.
3036 Frequency : only with huge count clients
3038 Description: Prevents from taking the superblock lock in llap_from_page for
3040 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
3041 for llap_from_page call in ll_removepage prevents from taking the
3042 superblock lock for a soon died page.
3047 Description: Not check open intent error before release open handle
3048 Details : in some rare cases, the open intent error is not checked before
3049 release open handle, which may cause
3050 ASSERTION(open_req->rq_transno != 0), because it tries to release
3051 the failed open handle.
3056 Description: Set cat log bitmap only after create log success.
3057 Details : in some rare cases, the cat log bitmap is set too early. and it
3058 should be set only after create log success.
3062 Description: Accessing a block bevice can re-enable I/O when Lustre is
3063 tearing down a device.
3064 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
3069 Description: service threads may hog cpus when there are a lot of requests
3070 Details : Insert cond_resched to give other threads a chance to use some CPU
3075 Description: the cat log was not initialized in recovery
3076 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
3077 unlink log on mds will not be initialized until mds post
3078 recovery. And also in mds post recovery, the unlink log will
3079 initialization will be done asynchronausly, so there will be race
3080 between add unlink log and unlink log initialization.
3084 Description: brw_stats were being printed incorrectly
3085 Details : brw_stats were being printed as log2 but all of them were not
3086 recorded as log2. Also remove some code duplication arising from
3087 filter_tally_{read,write}.
3091 Frequency : rare, only in recovery.
3092 Description: ASSERTION(req->rq_type != LI_POISON) failed
3093 Details : imp_lock should be held while iterating over imp_sending_list for
3094 prevent destroy request after get timeout in ptlrpc_queue_wait.
3098 Description: replay-single.sh test 52 fails
3099 Details : A lock's skiplist need to be cleanup when it being unlinked
3100 from its resource list.
3104 Description: Short directio read returns full requested size rather than
3106 Details : Direct I/O operations should return actual amount of bytes
3107 transferred rather than requested size.
3109 Severity : enhancement
3111 Description: metadata RPC reduction (e.g. for rm performance)
3112 Details : decrease the amount of synchronous RPC between clients and servers
3113 by canceling conflicing lock before the operation on the client
3114 and packing thier handles into the main operation RPC to server.
3116 Severity : enhancement
3118 Description: Async OSC create to avoid the blocking unnecessarily.
3119 Details : If an OST has no remaining object, system will block on the
3120 creation when it needs to create a new object on this OST. Now,
3121 ways use pre-created objects when available, instead of blocking on
3122 an empty osc while others are not empty. If we must block, we
3123 block for the shortest possible period of time.
3127 Description: improve handling recoverable errors
3128 Details : if request processig with error which can be recoverable on server
3129 request should be resend, otherwise page released from cache and
3132 Severity : enhancement
3134 Description: refine locking for avoid write wrong info into lov_objid file
3135 Details : fix possible races with add new target and write/update data in
3138 --------------------------------------------------------------------------------
3140 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
3145 Frequency : on some architectures
3147 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
3148 Details : Uninitialized flags sometimes cause configuration commands to
3152 Frequency : patchless clients only
3154 Description: missing __iget() symbol export
3155 Details : The __iget() symbol export is missing. To avoid the need for
3156 this on patchless clients the deathrow inode reaper is turned
3157 off, and we depend on the VM to clean up old inodes. This
3158 dependency was during via the fix for bug 12181.
3160 --------------------------------------------------------------------------------
3162 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
3164 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
3165 INTEROPERATE with older versions automatically. In many cases a
3166 special upgrade step is needed. Please read the
3167 user documentation before upgrading any part of a live system.
3168 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
3169 version will not interoperate with 1.6 betas before beta5 (1.5.95).
3170 * WARNING: Lustre configuration and startup changes are required with
3171 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
3175 Severity : enhancement
3177 Description: MountConf
3178 Details : Lustre configuration is now managed via mkfs and mount
3179 commands instead of lmc and lconf. New obd types (MGS, MGC)
3180 are added for dynamic configuration management. See
3181 https://mail.clusterfs.com/wikis/lustre/MountConf for
3184 Severity : enhancement
3186 Description: dynamic OST addition
3187 Details : OSTs can now be added to a live filesystem
3189 Severity : enhancement
3191 Description: startup order invariance
3192 Details : MDTs and OSTs can be started in any order. Clients only
3193 require the MDT to complete startup.
3195 Severity : enhancement
3197 Description: parallel, asynchronous orphan cleanup
3198 Details : orphan cleanup is now performed in separate threads for each
3199 OST, allowing parallel non-blocking operation.
3201 Severity : enhancement
3203 Description: optimized stripe assignment
3204 Details : stripe assignments are now made based on ost space available,
3205 ost previous usage, and OSS previous usage, in order to try
3206 to optimize storage space and networking resources.
3208 Severity : enhancement
3210 Description: Permanently set tunables
3211 Details : All writable /proc/fs/lustre tunables can now be permanently
3212 set on a per-server basis, at mkfs time or on a live system.
3214 Severity : enhancement
3216 Description: Lustre message v2
3217 Details : Add lustre message format v2.
3219 Severity : enhancement
3221 Description: client OST exclusion list
3222 Details : Clients can be started with a list of OSTs that should be
3223 declared "inactive" for known non-responsive OSTs.
3227 Description: SPEC SFS validation failure on NFS v2 over lustre.
3228 Details : Changes the blocksize for regular files to be 2x RPC size,
3229 and not depend on stripe size.
3231 Severity : enhancement
3233 Description: Multiple MD RPCs in flight.
3234 Details : Further unserialise some read-only MDS RPCs - learn about intents.
3235 To avoid overly-overloading MDS, introduce a limit on number of
3236 MDS RPCs in flight for a single client and add /proc controls
3237 to adjust this limit.
3239 Severity : enhancement
3241 Description: client read/write statistics
3242 Details : Add client read/write call usage stats for performance
3243 analysis of user processes.
3244 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
3245 file access. extents_stats shows chunk size distribution.
3246 extents_stats_per_process show chunk size distribution per
3249 Severity : enhancement
3251 Description: mds statistics
3252 Details : Add detailed mds operations statistics in
3253 /proc/fs/lustre/mds/*/stats.
3257 Description: Failure of copying files with lustre special EAs.
3258 Details : Client side always return success for setxattr call for lustre
3259 special xattr (currently only "trusted.lov").
3264 Description: Refcount LNET uuids
3265 Details : The global LNET uuid list grew linearly with every startup;
3266 refcount repeated list entries instead of always adding to
3269 Severity : enhancement
3271 Description: Dynamic service threads
3272 Details : Within a small range, start extra service threads
3273 automatically when the request queue builds up.
3276 Frequency : mixed-endian client/server environments
3278 Description: mixed-endian crashes
3279 Details : The new msg_v2 system had some failures in mixed-endian
3282 Severity : enhancement
3284 Description: Easy OST removal
3285 Details : OSTs can be permanently deactivated with e.g. 'lctl
3286 conf_param lustre-OST0001.osc.active=0'
3288 Severity : enhancement
3290 Description: MGS proc entries
3291 Details : Added basic proc entries for the MGS showing what filesystems
3294 Severity : enhancement
3296 Description: provide MGS failover
3297 Details : Added config lock reacquisition after MGS server failover.
3299 Severity : enhancement
3301 Description: add Linux 2.4 support
3302 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
3306 Description: a large application tries to do I/O to the same resource and dies
3307 in the middle of it.
3308 Details : Check the req->rq_arrival time after the call to
3309 ost_brw_lock_get(), but before we do anything about
3310 processing it & sending the BULK transfer request. This
3311 should help move old stale pending locks off the queue as
3312 quickly as obd_timeout.
3315 Frequency : when an incorrect nid is specified during startup
3317 Description: ptlrpc connect to non-existant node causes kernel crash
3318 Details : LNET can't be re-entered from an event callback, which
3319 happened when we expire a message after the export has been
3320 cleaned up. Instead, hand the zombie cleanup off to another
3323 Severity : enhancement
3325 Description: plain/inodebits lock performance improvement
3326 Details : Grouping plain/inodebits in granted list by their request modes
3327 and bits policy, thus improving the performance of search through
3331 Frequency : only if OST filesystem is corrupted
3333 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3334 Details : for a short time RPCs with bulk IO are in the replay list,
3335 but replay of bulk IOs is unimplemented. If the OST filesystem
3336 is corrupted due to disk cache incoherency and then replay is
3337 started it is possible to trip an assertion. Avoid putting
3338 committed RPCs into the replay list at all to avoid this issue.
3341 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3343 Description: System hang on startup
3344 Details : This bug allowed the liblustre (e.g. catamount) client to
3345 return to the app before handling all startup RPCs. This
3346 could leave the node unresponsive to lustre network traffic
3347 and manifested as a server ptllnd timeout.
3349 Severity : enhancement
3351 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3352 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
3353 boolean to control whether to print peer debug info when a
3354 client's RPC times out.
3357 Frequency : only for kernels with patches from Lustre below 1.4.3
3359 Description: Remove old rdonly API
3360 Details : Remove old rdonly API which unused from at least lustre 1.4.3
3363 Frequency : only for devices with external journals
3365 Description: Set external device read-only also
3366 Details : During a commanded failover stop, we set the disk device
3367 read-only while the server shuts down. We now also set any
3368 external journal device read-only at the same time.
3371 Frequency : when upgrading from 1.4 while trying to change parameters
3373 Description: The wrong (new) MDC name was used when setting parameters for
3374 upgraded MDT's. Also allows changing of OSC (and MDC)
3375 parameters if --writeconf is specified at tunefs upgrade time.
3378 Frequency : when setting specific ost indicies
3380 Description: QOS code breaks on skipped indicies
3381 Details : Add checks for missing OST indicies in the QOS code, so OSTs
3382 created with --index need not be sequential.
3384 Severity : enhancement
3386 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3387 Details : The uninit_groups feature works in conjunction with the kernel
3388 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3389 up the pass1 processing of e2fsck. This is a read-only feature
3390 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3391 mount filesystems that have had this feature enabled.
3393 Severity : enhancement
3395 Description: Improve multi-block allocation algorithm to avoid fragmentation
3396 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3397 allocation locality and avoid filesystem fragmentation.
3399 ------------------------------------------------------------------------------
3401 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
3403 * Support for kernels:
3404 2.4.21-47.0.1.EL (RHEL 3)
3405 2.6.5-7.283 (SLES 9)
3406 2.6.9-42.0.10.EL (RHEL 4)
3407 2.6.12.6 vanilla (kernel.org)
3408 2.6.16.27-0.9 (SLES 10)
3409 * Recommended e2fsprogs version: 1.39.cfs5
3411 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3415 Frequency : occasional, depends on client load and configuration
3416 Bugzilla : 12181, 12203
3417 Description: data loss for recently-modified files
3419 Details : In some cases it is possible that recently written or created
3420 files may not be written to disk in a timely manner (this should
3421 normally be within 30s unless client IO load is very high).
3422 The problem appears as zero-length files or files that are a
3423 multiple of 1MB in size after a client crash or client eviction
3424 that are missing data at the end of the file.
3426 This problem is more likely to be hit on clients where files are
3427 repeatedly created and unlinked in the same directory, clients
3428 have a large amount of RAM, have many CPUs, the filesystem has
3429 many OSTs, the clients are rebooted frequently, and/or the files
3430 are not accessed by other nodes after being written.
3432 The presence of the problem can be detected by looking at
3433 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
3434 smaller than the second (nr_unused) then dirty files will not
3435 be flushed automatically to disk. "sync; sleep 10" should be
3436 run several times on the node before unmounting it to update
3437 Lustre (this is also safe to run on nodes without this problem).
3439 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
3440 that can cause this same problem, so customers using that kernel
3441 also need to update the kernel in addition to Lustre. In order
3442 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
3444 It is normal that files written just before a client crash (less
3445 than 30s) may not yet have been flushed to disk, even for local
3449 Frequency : frequent on thin XT3 nodes
3451 Description: UUID collision on thin XT3 Linux nodes
3452 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
3453 randomly, since we relied on an insufficiently-seeded PRNG.
3458 Description: OSS hangs after "All ost request buffers busy"
3459 Details : A deadlock between quota and journal operations caused OSS
3460 hangs after printing "All ost request buffers busy."
3463 Frequency : always on liblustre builds
3465 Description: Cleanup compiler warnings on liblustre
3468 Frequency : always on liblustre builds on XT3
3470 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
3474 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
3475 > 32000 subdirectories
3476 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
3477 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
3478 subdirectory count crosses 32000. This will aid e2fsck to
3479 correctly handle more than 32000 subdirectories.
3482 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3484 Description: System hang on startup
3485 Details : This bug allowed the liblustre (e.g. catamount) client to
3486 return to the app before handling all startup RPCs. This
3487 could leave the node unresponsive to lustre network traffic
3488 and manifested as a server ptllnd timeout.
3490 Severity : enhancement
3492 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3493 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
3494 boolean to control whether to print peer debug info when a
3495 client's RPC times out.
3500 Description: make O_SYNC working on 2.6 kernels
3501 Details : 2.6 kernels use different method for mark pages for write,
3502 so need add a code to lustre for O_SYNC work.
3507 Description: Failure to close file and release space on NFS
3508 Details : Put inode details into lock acquired in ll_intent_file_open.
3509 Use mdc_intent_lock in ll_intent_open to properly
3510 detect all kind of errors unhandled by mdc_enqueue.
3515 Description: proc file read during shutdown sometimes raced obd removal,
3517 Details : Add lock to prevent obd access after proc file removal.
3520 Frequency : Only for files larger than 4GB on 32-bit clients.
3522 Description: improperly doing page alignment of locks
3523 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
3524 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3529 Description: RPCs being resent when they shouldn't be
3530 Details : Some RPCs that should not be resent are being resent. This
3531 can cause inconsistencies in the RPC state machine. Do not
3532 resend such requests.
3535 Frequency : rare, only with NFS export
3537 Description: Crash on NFS re-export node
3538 Details : under very unusual load conditions an assertion is hit in
3539 ll_intent_file_open()
3542 Frequency : only if OST filesystem is corrupted
3544 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3545 Details : for a short time RPCs with bulk IO are in the replay list,
3546 but replay of bulk IOs is unimplemented. If the OST filesystem
3547 is corrupted due to disk cache incoherency and then replay is
3548 started it is possible to trip an assertion. Avoid putting
3549 committed RPCs into the replay list at all to avoid this issue.
3554 Description: large O_DIRECT requests fail under memory pressure/fragmentation