3 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
4 for currently supported client and server kernel versions.
5 * Server known to build on patched kernels:
6 2.6.32-431.29.2.el6 (RHEL6.5)
7 2.6.32-504.30.3.el6 (RHEL6.6)
8 2.6.32-573.26.1.el6 (RHEL6.7)
9 2.6.32-642.6.1.el6 (RHEL6.8)
10 3.10.0-327.36.2.el7 (RHEL7.2)
11 3.0.101-0.47.71 (SLES11 SP3)
12 3.0.101-80 (SLES11 SP4)
13 3.12.62-60.62 (SLES12 SP1)
14 vanilla linux 4.5.7 (ZFS only)
15 * Client known to build on unpatched kernels:
16 2.6.32-431.29.2.el6 (RHEL6.5)
17 2.6.32-504.30.3.el6 (RHEL6.6)
18 2.6.32-573.26.1.el6 (RHEL6.7)
19 2.6.32-642.6.1.el6 (RHEL6.8)
20 3.10.0-327.36.2.el7 (RHEL7.2)
21 3.0.101-0.47.71 (SLES11 SP3)
22 3.0.101-80 (SLES11 SP4)
23 3.12.62-60.62 (SLES12 SP1)
25 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
26 * Recommended ZFS / SPL version: 0.6.5.7
27 * Tested with ZFS / SPL version: 0.6.5.7
28 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
29 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
30 stack overflow. For more information, please refer to bugzilla 17630.
31 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
32 the re-exporting nodes to be mounted with "32bitapi" mount option
34 --------------------------------------------------------------------------------
36 02-29-2016 Intel Corporation
38 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
39 for currently supported client and server kernel versions.
40 * Server known to build on patched kernels:
41 2.6.32-431.29.2.el6 (RHEL6.5)
42 2.6.32-504.30.3.el6 (RHEL6.6)
43 2.6.32-573.12.1.el6 (RHEL6.7)
44 3.10.0-327.3.1.el7 (RHEL7.2)
45 3.0.101-0.47.71 (SLES11 SP3)
46 3.0.101-68 (SLES11 SP4)
47 vanilla linux 4.2.1 (ZFS only)
48 * Client known to build on unpatched kernels:
49 2.6.32-431.29.2.el6 (RHEL6.5)
50 2.6.32-504.30.3.el6 (RHEL6.6)
51 2.6.32-573.12.1.el6 (RHEL6.7)
52 3.10.0-327.3.1.el7 (RHEL7.2)
53 3.0.101-0.47.71 (SLES11 SP3)
54 3.0.101-68 (SLES11 SP4)
57 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
58 * Recommended ZFS / SPL version: 0.6.4.2
59 * Tested with ZFS / SPL version: 0.6.4.2
60 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
61 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
62 stack overflow. For more information, please refer to bugzilla 17630.
63 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
64 the re-exporting nodes to be mounted with "32bitapi" mount option
66 --------------------------------------------------------------------------------
68 03-10-2015 Intel Corporation
70 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
71 for currently supported client and server kernel versions.
72 * Server known to build on patched kernels:
73 2.6.32-431.29.2.el6 (RHEL6.5)
74 2.6.32-504.8.1.el6 (RHEL6.6)
75 3.0.101-0.46 (SLES11 SP3)
76 * Client known to build on unpatched kernels:
77 2.6.32-431.29.2.el6 (RHEL6.5)
78 2.6.32-504.8.1.el6 (RHEL6.6)
79 3.10.0-123.20.1.el7 (RHEL7)
80 3.0.101-0.46 (SLES11 SP3)
81 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
82 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
83 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
84 stack overflow. For more information, please refer to bugzilla 17630.
85 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
86 the re-exporting nodes to be mounted with "32bitapi" mount option
88 Severity : enhancement
90 Description: control OST-index in IDIF via ROCOMPAT flag.
91 Details : Introduce new flag OBD_ROCOMPAT_IDX_IN_IDIF that is stored in the
92 last_rcvd file. For new formatted OST device, it will be auto set;
93 for the case of upgrading from old OST device, you can enable it
94 via the lproc interface osd-ldiskfs.index_in_idif. With such flag
95 enabled, for new created OST-object, its IDIF-in-LMA will contain
96 the OST-index; for the existing OST-object, the OSD will convert
97 old format IDIF as new format IDIF with OST-index stored in the
98 LMA EA when accessing such OST-object or via OI scrub. Once such
99 flag is enabled, it cannot be reverted back, so the system cannot
100 be downgraded to the orignal incompatible version.
102 --------------------------------------------------------------------------------
104 07-30-2014 Intel Corporation
106 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
107 for currently supported client and server kernel versions.
108 * Server known to build on patched kernels:
109 2.6.32-431.20.3.el6 (RHEL6.5)
110 3.0.101-0.31 (SLES11 SP3)
111 * Client known to build on unpatched kernels:
112 2.6.32-431.20.3.el6 (RHEL6.5)
113 3.0.101-0.31 (SLES11 SP3)
116 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
117 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
118 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
119 stack overflow. For more information, please refer to bugzilla 17630.
120 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
121 the re-exporting nodes to be mounted with "32bitapi" mount option
123 --------------------------------------------------------------------------------
125 10-31-2013 Intel Corporation
127 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
128 * Server builds for kernels:
129 2.6.32-358.18.1.el6 (RHEL6.4)
130 * Client builds for unpatched kernels:
131 2.6.32-358.18.1.el6 (RHEL6.4)
132 3.0.92-0.8 (SLES11 SP3)
134 * Recommended e2fsprogs version: 1.42.7.wc2 or newer
135 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
136 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
137 stack overflow. For more information, please refer to bugzilla 17630.
138 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
139 the re-exporting nodes to be mounted with "32bitapi" mount option
141 --------------------------------------------------------------------------------
143 05-30-2013 Intel Corporation
145 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
146 * Server builds for kernels:
147 2.6.32-358.6.2.el6 (RHEL6.4)
148 * Client builds for unpatched kernels:
149 2.6.32-358.6.2.el6 (RHEL6.4)
150 3.0.74-0.6.6 (SLES11 SP2)
152 * Recommended e2fsprogs version: 1.42.7.wc1 or newer
153 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
154 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
155 stack overflow. For more information, please refer to bugzilla 17630.
156 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
157 the re-exporting nodes to be mounted with "32bitapi" mount option
159 --------------------------------------------------------------------------------
161 9-30-2012 Whamcloud, Inc.
163 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
164 * Server support for kernels:
165 2.6.32-279.5.1.el6 (RHEL6)
166 * Client support for unpatched kernels:
167 2.6.32-279.5.1.el6 (RHEL6)
169 2.6.18-238.19.1.el5 (RHEL5)
170 2.6.16.60-0.69.1 (SLES10)
171 * Recommended e2fsprogs version: 1.41.90.wc4
172 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
173 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
174 stack overflow. For more information, please refer to bugzilla 17630.
175 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
176 the re-exporting nodes to be mounted with "32bitapi" mount option
178 Severity : enhancement
180 Description: Update RHEL6.3 kernel to 2.6.32-279.5.1.el6.
182 Severity : enhancement
184 Description: Update RHEL6.3 kernel to 2.6.32-279.2.1.el6.
186 Severity : enhancement
188 Description: Update RHEL6.2 kernel to 2.6.32-220.17.1.el6
190 Severity : enhancement
192 Description: Update RHEL6.2 kernel to 2.6.32-220.13.1.el6
194 Severity : enhancement
196 Description: Update RHEL6.2 kernel to 2.6.32-220.7.1.el6
198 --------------------------------------------------------------------------------
200 04-30-2012 Whamcloud, Inc.
202 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
203 * Server support for kernels:
204 2.6.32-220.4.2.el6 (RHEL6)
205 * Client support for unpatched kernels:
206 2.6.32-220.4.2.el6 (RHEL6)
208 2.6.18-238.19.1.el5 (RHEL5)
209 2.6.16.60-0.69.1 (SLES10)
210 * Recommended e2fsprogs version: 1.41.90.wc4
211 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
212 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
213 stack overflow. For more information, please refer to bugzilla 17630.
214 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
215 the re-exporting nodes to be mounted with "32bitapi" mount option
217 Severity : enhancement
219 Description: Update RHEL6.2 kernel to 2.6.32-220.4.2.el6
221 Severity : enhancement
223 Description: Update RHEL6.2 kernel to 2.6.32-220.4.1.el6
225 Severity : enhancement
227 Description: Update RHEL6.2 kernel to 2.6.32-220.el6
229 Severity : enhancement
231 Description: Update RHEL6.1 kernel to 2.6.32-131.17.1.el6
233 --------------------------------------------------------------------------------
235 09-30-2011 Whamcloud, Inc.
237 * Server support for kernels:
238 2.6.18-238.19.1.el5 (RHEL5)
239 2.6.32-131.6.1.el6 (RHEL6)
240 * Client support for unpatched kernels:
241 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
242 2.6.18-238.19.1.el5 (RHEL5)
243 2.6.32-131.6.1.el6 (RHEL6)
244 2.6.29.4-167.fc11 (FC11)
245 2.6.16.60-0.69.1 (SLES10)
247 * Recommended e2fsprogs version: 1.41.90.wc3
248 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
249 Lustre filesystem with 4K stack may cause a stack overflow. For more
250 information, please refer to bugzilla 17630.
251 * NFSv4 reexport requires Lustre client on the re-exporting nodes to be
252 mounted with "32bitapi" mount option
254 Severity : enhancement
255 Description: Update RHEL6.1 kernel to 2.6.32-131.6.1.el6
257 Severity : enhancement
258 Description: Update RHEL5.6 kernel to 2.6.18-238.19.1.el5
260 Severity : enhancement
261 Description: Update RHEL5.6 kernel to 2.6.18-238.9.1.el5
264 Frequency : only with SLES10
266 Description: Use OFED "KMP" provided by Novell
267 Details : SLES10 SP3 ships with OFED in a separate "KMP" package.
268 Lustre is now built against this package. That means you need to
269 install the ofed-kmp package from Novell for the patchless client.
270 Note that the ofed-kmp that Novell ships may not exactly match the
271 kernel version but should still be compatible.
273 Severity : enhancement
275 Description: Update RHEL5.5 kernel to 2.6.18-194.17.1.el5
276 Update OEL5.5 kernel to 2.6.18-194.17.1.0.1.el5.
278 --------------------------------------------------------------------------------
280 2010-07-15 Oracle, Inc.
282 * Server support for kernels:
283 2.6.18-164.11.1.el5 (RHEL 5)
284 2.6.18-164.11.1.0.1.el5 (OEL 5)
285 * Client support for unpatched kernels:
286 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
287 2.6.18-164.11.1.el5 (RHEL 5),
288 2.6.18-164.11.1.0.1.el5 (OEL 5)
289 2.6.16.60-0.42.8 (SLES 10),
291 2.6.29.4-167.fc11 (FC11)
292 * Recommended e2fsprogs version: 1.41.10-sun2
294 Severity : enhancement
296 Description: Distribute lfs_migrate script with lustre
297 Details : The lfs_migrate mentioned in the manual is now included in the
298 lustre tree in lustre/scripts and will distributed in lustre RPMs.
300 Severity : enhancement
302 Description: Failover nid list added to import proc file.
304 Severity : enhancement
306 Description: "som_preview" mount option required to use SOM
307 Details : The clients will need to use som_preview mount option
311 Description: Lustre does not do 1MB IOs to HW RAID
312 Details : Bump MAX_PHYS/HW_SEGMENTS and SG_ALL to 256 in the RHEL5 kernel.
314 Severity : enhancement
316 Description: lfs quota output is not very convenient for awk/sed-parsing
317 Details : Some positions in lfs quota output table could be empty or
318 non-empty which made it hard to parse it with scripts, now
319 a dash is put instead of space where there is not supposed
324 Description: Update to OFED 1.5.1
325 Details : OFED 1.5.1 will be used in environments that don't use
326 vendor supplied OFED.
330 Description: Remove lnet. prefix from lctl params display
331 Details : Also renamed lnet.memused to lnet_memused to
336 Description: Use conf_param -d to remove permanant settings.
340 Description: Add list_param -R
341 Details : lctl list_param now supports recursive mode.
344 Bugzilla : 21829, 22312
345 Description: Fix llobdstat.
346 Details : 1. Now updating the write_bytes counter
347 2. Port script fixes from 1.8.
351 Description: decrease the usage of memory on clients.
352 Details : 1. On clients, recycle dentries and inodes unused.
353 2. Delete the code related to ll_deathrow(att 6215 in bug 1443). It
358 Description: llobdstat fix and enhancement
359 Details : add a counter to set a limit to how many samples will be returned
360 fix a wildcard in the path to limit to obdfilter stats only
364 Description: lfs check only allowed for root
365 Details : Code cleanups and added a regresion test of non-root use of
370 Description: MDS Oopses in ptlrpc_dispatch_difficult_reply()/__list_add()
371 Details : Fix hr->hr_index overload.
375 Description: Fix FIEMAP functionality.
379 Description: FC11 patchless client support added
383 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and
384 OEL5.4 kernel to 2.6.18-164.11.1.0.1.el5.
388 Description: New acc-sm script and various test-framework enhancments.
392 Description: Previously landed patch was reverted.
396 Description: Update SLES10 kernel to 2.6.16.60-0.42.8 (patchless)
400 Description: Add ldiskfs maxdirsize mount option.
404 Description: Fix mkfs.lustre support for 16TB LUNs.
408 Description: Print mmp_check_interval
409 Details : print mmp_check_interval and make it possible to abort
410 mount operation in case it takes too long.
414 Description: Add cascading_rw to lustre mpi tests.
418 Description: Store SOM data in EA.
419 Details : Also several SOM related cleanups.
423 Description: Add refcount to md_open_data
427 Description: Use refcount on lov_request_set to prevent memory corruption.
431 Description: Statahead: use parent dir's i_mutex for synchronization.
435 Description: loadgen improvements.
439 Description: Enable peer health detection in ksocklnd and ko2iblnd.
443 Description: Control DCACHE_LUSTRE_INVALID flag with MDS_INODE_LOOKCUP lock
447 Description: Fix race on obd_devcie:md_stats
451 Description: Fix SLES 11 client builds.
452 Details : Resolved uninitalized pointer gcc warnings.
456 Description: Allow non-root access for "lfs check".
457 Details : Added a check in obd_class_ioctl() for OBD_IOC_PING_TARGET.
461 Description: Clear imp_force_reconnect correctly in ptlrpc_connect_interpret()
463 Severity : enhancement
465 Description: Update kernel to RHEL5.4 2.6.18-164.6.1.el5 and
466 OEL5 2.6.18-164.6.1.0.1.el5(Both in-kernel OFED enabled).
470 Description: Sanity test_27p fails.
471 Details : OBD_FAIL_OST_ENOSPC has been moved to filter_statfs().
472 Sanity tests 27 were cleaned up and improved.
476 Description: Unify req format on client/servers
477 Details : Use new req_capsule API [almost] everywhere instead of old PTLRPC
478 buffers and swabbers approach..
483 Description: drop join file stuff.
485 Severity : enhancement
487 Description: Add 2.6.27 and SLES11 (patchless client) support
491 Description: Fix in ptlrpc_expire_one_request() to print the signed time
494 Severity : enhancement
496 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
499 Severity : enhancement
501 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
503 Severity : enhancement
505 Description: correctly handle big reply message.
506 Details : send LNet event if reply is bigger than buffer and adjust this buffer
511 Description: Drop unnecessary __GFP_NOMEMALLOC flag from filter_get_page()
513 Severity : enhancement
515 Description: Add support for OEL5.
519 Description: Conf-sanity.sh 50g test - deactivated OST should not cause a panic.
521 Severity : enhancement
523 Description: "weak-modules" support
524 Details : Implement "weak-modules" support which enables kernel modules
525 to be used with any kernel that implements the same kABI. In
526 order to achieve this modules are now installed in
527 /lib/modules/$(uname -r)/updates/kernel on all distributions.
531 Description: client could not reconnect to OST because of an active request.
532 Details : The client now retries to reconnect to the same server, if a
533 connect request failed with EBUSY or -EAGAIN. Test case 35b
534 conf-sanity.sh has been created.
538 Description: Don't allow make backward step on assignin osc next id.
539 Details : race between allocation next id and ll_sync thread can be cause
540 of set wrong osc next id and can be kill valid ost objects.
544 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
548 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
552 Description: Error handling in osc_statfs_interpret() has been improved.
553 Details : Check in osc_statfs_interpret() for EBADR.
557 Description: Do not send statfs() requests to OSTs disabled by administrator.
558 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
560 Severity : enhancement
562 Description: Added sanity-quota test to for lfs(1) to work with numeric uid/gid.
564 Severity : enhancement
566 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
568 Severity : enhancement
570 Description: Enhancement for lfs(1) command to use numeric uid/gid.
574 Description: Improved cleanup procedure for conf-sanity test_46a.
578 Description: Do not start bulk transfer, if client tried to reconnect.
579 Details : Do not start bulk transfer, if client tried to reconnect,
580 but failed due to active rpcs. Check for exp_abort_active_req
581 has been implemented.
585 Description: Prevent inconsistences between linux and lustre mount structures.
586 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
587 Make the sleep interruptible.
589 Severity : enhancement
591 Description: provide server to client comms path
592 Details : server to client communications path via new LDLM_SET_INFO rpc
596 Description: truncate starts GFP_FS allocation under transaction causing deadlock
597 Details : ldiskfs_truncate calls grab_cache_page which may start page
598 allocation under an open transaction. This may lead to
599 calling prune_icache with consequent lustre reentrance.
603 Description: send correctly lsm on open replay
604 Details : MDS is trust to LSM size on replay open, but client can set wrong size
609 Description: enable client interface failover
610 Details : When a child reconnects from another NID, properly update export
611 nid hash position and ldlm reverse import.
613 Severity : enhancement
615 Description: Communicate OST degraded/readonly state via statfs to MDS
616 Details : Flags in the statfs returned from OSTs indicate whether the
617 OST is in a degraded RAID state, or if the filesystem has
618 turned read-only after a filesystem error is detected.
623 Description: don't panic if EPROTO was hit when reading symlink
624 Details : correctly handling request reference in error cases.
626 Severity : enhancement
628 Description: Add LustreNetLink, a kernel-userspace communcation path.
630 Severity : enhancement
632 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
637 Description: access to llog context before init.
638 Details : move handling CATALOGS file at osc layer and forbid access to llog
642 Frequency : rare, only when client is writing to OST on local node
644 Description: Avoid deadlock for local client writes
645 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
646 memory freeing context. This allows OST threads to set the
647 PF_MEMALLOC flag on task structures in order to allocate memory
648 from reserved pools and complete IO.
649 Use GFP_HIGHUSER for OST allocations for non-local client writes,
650 so that the OST threads generate memory pressure and allow
651 inactive pages to be reclaimed.
653 Severity : enhancement
655 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
660 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
661 Details : move ldlm namespace creation in setup phase to avoid grab
662 _lprocfs_lock with cli_sem held.
666 Description: Temporarily disable grant shrink.
667 Details : Disable the feature for debugging.
670 Frequency : only when formatting test filesystems
672 Description: Unable to run several mkfs.lustre on loop devices at the same time
673 Details : mkfs.lustre returns error 256 on the concurrent loop devices
674 formatting. The solution is to proper handle the error.
676 Severity : enhancement
678 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
680 Severity : enhancement
682 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
687 Description: resolve race between obd_disconnect and class_disconnect_exports
688 Details : if obd_disconnect will be called to already disconnected export he
689 forget release one reference and osc module can't unloaded.
691 Severity : enhancement
693 Description: Allow tuning service thread via /proc
694 Details : For each service a new
695 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
696 created that can be used to set min/max thread counts, and get the
697 current number of running threads.
702 Descriptoin: don't return error if have partially created objects for file.
703 Details : lov_update_create_set uses set->set_success as index for created
704 objects, so if some requests will be failed, they will have hole
705 at end of array and qos_shrink_lsm for allocate correct lsm.
707 Severity : enhancement
709 Description: Update OFED support to 1.4.1
714 Description: Reduce small size read RPC
715 Details : Set read-ahead limit for every file and only do read-ahead when
716 available read-ahead pages are bigger than 1M to avoid small size
719 Severity : enhancement
721 Description: move AT tunable parameters for more consistent usage
722 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
724 Severity : enhancement
726 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
727 Details : allow skip disconnected ost for send statfs request and hide error
733 Description: fix racy locking of mballoc block bitmaps causing BUG
734 Details : The locking of the mballoc buddy bitmap and the in-memory
735 block bitmap was using two different spin locks in some
736 cases. This made it possible to incorrectly access the
737 mballoc bitmap while another process was modifying it,
738 causing a sanity assertion to fail. While no on-disk corruption
739 was reported, there was some risk of this happening.
742 Frequency : rare, on llog test 6
744 Descriptoin: don't allow connect to already connected import
745 Details : allowing connect to already connected import is hide connecting problem.
750 Descriptoin: allow kill process which wait statahead result
751 Details : for some reasons 'ls' can stick in waiting result from statahead,
752 in this case need way for kill this process.
754 Severity : enhancement
756 Description: Add state history info file, enhance import info file
757 Details : Track import connection state changes in a new osc/mdc proc file;
758 add overview-type data to the osc/mdc import proc file.
760 Severity : enhancement
762 Description: MDS create should not wait for statfs RPC while holding DLM lock.
764 Severity : enhancement
766 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
771 Descriptoin: OST grant too much space to client even there are not enough space.
772 Details : Client will shrink its grant cache to OST if there are no write
773 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
774 this grant cache if there are already not enough avaible space
775 (left_space < total_clients * 32M).
778 Frequency : start MDS on uncleanly shutdowned MDS device
780 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
781 Details : stay in waiting mds<>ost recovery finished produce random bugs
782 due race between two ll_sync thread for one lov target. send
783 ACTIVATE event only if connect realy finished and import have
787 Frequency : rare, connect and disconnect target at same time
789 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
790 Details : don't call obd_disconnect under lov_lock. this long time
791 operation and can block ptlrpcd which answer to connect request.
796 Descriptoin: don't lose wakeup for imp_recovery_waitq
797 Details : recover_import_no_retry or invalidate_import and import_close can
798 both sleep on imp_recovery_waitq, but we was send only one wakeup
802 Frequency : always with long access acl
804 Descriptoin: mds can't pack reply with long acl.
805 Details : mds don't control size of acl but they limited by reint/getattr
808 Severity : enhancement
810 Description: Update to SLES10 kernel-2.6.16.60-0.33.
812 Severity : enhancement
814 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
817 Frequency : start MDS on uncleanly shutdowned MDS device
819 Descriptoin: aborting recovery hang on MDS
820 Details : don't throttle destroy RPCs for the MDT.
823 Frequency : on remount
825 Description: external journal device not working after the remount
826 Details : clear dev_rdonly flag for external journal devices in
832 Description: shutdown vs evict race
833 Details : client_disconnect_export vs connect request race.
834 if client will evicted at this time - we start invalidate
835 thread without referece to import and import can be freed
839 Frequency : rare, need acl's on inode.
841 Description: client can't handle ost additional correctly
842 Details : if ost was added after client connected to mds client can have
843 hit lnet_try_match_md ... to big messages to wide striped files.
844 in this case need teach client to handle config events about add
845 lov target and update client max ea size at that event.
847 Severity : enhancement
849 Description: Changelogs
850 Details : Changelogs are a lightweight mechanism to track filesystem
851 metadata and namespace changes. The changelog is recorded
852 permanently on the MDTs, and is periodically "consumed" / purged
853 when records are no longer needed.
855 Severity : enhancement
857 Description: compact fld format with extents
858 Details : Store range of seq rather than every seq in FLD. Seq
859 controller update FLD rather than clients. In Case of CMD, mdt0
860 has FLD, all other metadata server act as non persistent proxy
861 for FLD queries and cache fld entries in fld cache.
866 Description: don't skip ost target if they assigned to file
867 Details : Drop slow OSCs if we can, but not for requested start idx.
868 This means "if OSC is slow and it is not the requested
869 start OST, then it can be skipped, otherwise skip it only
870 if it is inactive/recovering/out-of-space.
874 Description: more cleanup in mds_lov
875 Details : not send LOV EA under replay, we can't know about they size at this
876 time. Don't allow client connect to mds before any ost connected,
877 for avoid problems with LOV EA size and returning EIO to client.
879 Severity : enhancement
881 Description: Interoperability at server side (Disk interoperability)
883 Severity : enhancement
885 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
887 Severity : enhancement
889 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
891 Severity : enhancement
893 Description: New client IO stack (CLIO).
895 Severity : enhancement
897 Description: Commit on sharing. Eliminate inter-client dependencies between
898 uncommitted transactions by doing transaction commits.
899 Thereby clients may recovery independently.
902 Frequency : Create a symlink file with a very long name
904 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
905 Details : If there is no extra space in the request for early cancels,
906 ldlm_req_handles_avail() returns 0 instead of a negative value.
908 Severity : enhancement
910 Description: Add /proc entry for import status
911 Details : The mdc, osc, and mgc import directories now have
912 an import directory that contains useful import data for debugging
915 Severity : enhancement
917 Description: Re-disable certain /proc logging
918 Details : Enable and disable client's offset_stats, extents_stats and
919 extents_stats_per_process stats logging on the fly.
922 Frequency : Only on FC kernels 2.6.22+
924 Description: oops in statahead
925 Details : Do not drop reference count for the dentry from VFS when lookup,
926 VFS will do that by itself.
928 Severity : enhancement
930 Description: Generic /proc file permissions
931 Details : Set /Proc file permissions in a more generic way to enable non-
932 root users operate on some /proc files.
936 Description: Hitting mdc_commit_close() ASSERTION
937 Details : Properly handle request reference release in
938 ll_release_openhandle().
942 Description: quota recovery deadlock during mds failover
943 Details : This patch includes att18982, att18236, att18237 in bz14840.
945 1. fix osts hang when mds does failover with quotaon
946 2. prevent watchdog storm when osts threads wait for the
951 Frequency : only patchless client
952 Description: add workaround for race between add/remove dentry from hash
954 Severity : enhancement
956 Description: Allow OST glimpses to return PW locks
960 Description: LBUG when llog conf file is full
961 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
965 Description: Prevent import from entering FULL state when server in recovery
969 Description: service mount cannot take device name with ":"
970 Details : Only when device name contains ":/" will mount treat it as
976 Description: replace ptlrpcd with the statahead thread to interpret the async
977 statahead RPC callback
981 Frequency : on recovery
982 Description: I/O failures after umount during fail back
983 Details : if client reconnected to restarted server we need join to recovery
984 instead of find server handler is changed and process self
985 eviction with cancel all locks.
987 Severity : enhancement
989 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
991 Severity : enhancement
993 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
995 Severity : enhancement
997 Description: Upcall on Lustre log has been dumped
998 Details : Allow for a user mode script to be called once a Lustre log has
999 been dumped. It passes the filename of the dumped log to the
1000 script, the location of the script can be specified via
1001 /proc/sys/lnet/debug_log_upcall.
1006 Description: avoid idr_remove called for id which is not allocated.
1007 Details : Move assigment s_dev for clustered nfs to end of initialization,
1008 to avoid problem with error handling.
1013 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
1014 Details : When connection is reused this not moved from CONN_UNUSED_HASH
1015 into CONN_USED_HASH and this prodice warning when put connection
1016 again in unused hash.
1019 Severity : enhancement
1021 Description: File striping can now be set to use an arbitrary pool of OSTs.
1023 Severity : enhancement
1025 Description: Export bytes_read/bytes_write count on OSC/OST.
1029 Description: Early reply size mismatch, MGC loses connection
1030 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
1031 the connect flags are properly negotiated.
1036 Description: quotas are not honored with O_DIRECT
1037 Details : all writes with the flag O_DIRECT will use grants which leads to
1038 this problem. Now using OBD_BRW_SYNC to guard this.
1042 Description: add quota statistics
1043 Details : 1. sort out quota proc entries and proc code.
1044 2. add quota statistics
1046 Severity : enhancement
1048 Description: enable quota support for HEAD.
1052 Description: Properly propagate oinfo flags from lov to osc for statfs
1053 Details : restore missing copy oi_flags to lov requests.
1055 Severity : enhancement
1057 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
1058 : l_getgroups(8), lst(8), routerstat(8)
1059 Details : included man pages for llobdstat(8), llstat(8),
1060 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
1062 Severity : enhancement
1064 Description: configure's --enable-quota should check the
1065 : kernel .config for CONFIG_QUOTA
1066 Details : configure is terminated if --enable-quota is passed but
1067 : no quota support is in kernel
1071 Description: Remove portals compatibility
1072 Details : Remove portals compatibility, not interoperable with releases
1077 Description: Resolve device initialization race
1078 Details : Prevent proc handler from accessing devices added to the
1079 obd_devs array but yet be intialized.
1081 Severity : enhancement
1083 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
1085 Severity : enhancement
1087 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
1092 Description: Using wrong pointer in osc_brw_prep_request
1093 Details : Access to array[-1] can produce panic if kernel compiled with
1094 CONFIG_PAGE_ALLOC enabled
1098 Description: Client runs out of low memory
1099 Details : Consider only lowmem when counting initial number of llap pages
1103 Description: *optional* service tags registration
1104 Details : if the "service tags" package is installed on a Lustre node
1105 When the filesystem is mounted, a local-node service tag will
1106 be created. See http://inventory.sun.com/ for more information
1107 about the Service Tags asset management system.
1111 Description: Kernel BUG tries to release flock
1112 Details : Lustre does not destroy flock lock before last reference goes
1113 away. So always drop flock locks when client is evicted and
1114 perform unlock regardless of successfulness of speaking to MDS.
1118 Description: add refcount protection for osc callbacks, avoid panic on shutdown
1122 Description: sanity test 65a fails if stripecount of -1 is set
1123 Details : handle -1 striping on filesystem in ll_dirstripe_verify
1128 Description: ASSERTION(CheckWriteback(page,cmd)) failed
1129 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
1132 Severity : enhancement
1134 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
1138 Description: do not process already freed flock
1139 Details : flock can possibly be freed by another thread before it reaches
1140 to ldlm_flock_completion_ast.
1144 Description: LBUG during stress test
1145 Details : Need properly lock accesses the flock deadlock detection list.
1149 Description: oops in page fault handler
1150 Details : kernel page fault handler can return two special 'pages' in error
1151 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
1155 Description: timeout with invalidate import.
1156 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
1157 should be handled by ptlrpcd. This produce long age waiting and
1158 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
1160 Severity : enhancement
1162 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
1166 Description: enable MGS and MDT services start separately
1167 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
1168 a co-located MGS without starting the MGS, which is a complement
1169 to 'nosvc' mount option.
1173 Frequency : after recovery
1174 Description: precreate to many object's after del orphan.
1175 Details : del orphan st in oscc last_id == next_id and this triger growing
1176 count of precreated objects. Set flag LOW to skip increase count
1177 of precreated objects.
1181 Frequency : rare, on clear nid stats
1182 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
1183 Details : when clean nid stats sometimes try destroy live entry,
1184 and this produce panic in free.
1188 Description: Stack overflow during MDS log replay
1189 ease stack pressure by using a thread dealing llog_process.
1193 Description: wait until IO finished before start new when do lock cancel.
1194 Details : VM protocol want old IO finished before start new, in this case
1195 need wait until PG_writeback is cleared until check dirty flag and
1196 call writepages in lock cancel callback.
1198 Severity : enhancement
1200 Description: using special macro for print time and cleanup in includes.
1204 Description: mds_mfd_close() ASSERTION(rc == 0)
1205 Details : In mds_mfd_close(), we need protect inode's writecount change
1206 within its orphan write semaphore to prevent possible races.
1210 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
1214 Frequency : rare, on shutdown ost
1215 Description: don't hit live lock with umount ost.
1216 Details : shrink_dcache_parent can be in long loop with destroy dentries,
1217 use shrink_dcache_sb instead.
1221 Description: don't panic with use echo client
1222 Details : echo client pass NULL as client nid pointer and this produce null
1223 pointer dereference.
1227 Description: fix build on ppc32
1228 Details : compile code with -m64 flag produce wrong object file for ppc32.
1232 Description: add message levels for liblustreapi
1236 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
1237 Details : Move the 'good_osts' check before the 'total_bavail' check. This
1238 will result in an -EAGAIN and in the exit call path we call
1239 alloc_rr() which will with increasing aggressiveness attempt to
1240 aquire precreated objects on the minimum number of required OSCs.
1244 Description: Use old size assignment to avoid deadlock
1245 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
1246 the scheduling while holding a spinlock. We do not need locking
1247 for size in ll_update_inode() because size is only updated from
1248 the MDS for directories or files without objects, so there is no
1249 other place to do the update, and concurrent access to such inodes
1250 are protected by the inode lock.
1254 Description: resolve "_IOWR redefined" build error on SLES10
1258 Description: dump the memory debugging after all modules are unloaded to
1259 suppress false negative in conf_sanity test 39
1261 Severity : enhancement
1263 Description: build kernel-ib packages for OFED 1.3 in our release cycle
1268 Description: fix SLES kernel versioning
1269 Details : the kernel version for our SLES 10 kernel did not include a "-"
1270 before the "smp" at the end. while this was not a problem in
1271 general, it did mean that software trying to use the kernel
1272 version to try to detect a vendor specific kernel would fail.
1273 this was most evident by the OFED build scripts.
1277 Description: Don't update lov_desc members until making sure they are valid
1278 Details : When updating lov_desc members via proc fs, need fix their
1279 validities before doing the real update.
1283 Description: don't put request into delay list while invalidate in flight.
1284 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
1285 import in flight. this produce timeout for invalidate and sometimes
1286 can cause stale data.
1290 Frequency : on ppc only
1291 Description: not convert ost objects for directory because it's not exist.
1292 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
1296 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
1298 Severity : enhancement
1300 Description: Update to support 2.6.22.14 vanilla kernel.
1304 Frequency : rare, on recovery
1305 Description: read procfs can produce deadlock in some situation
1306 Details : Holding lprocfs lock which send rpc can produce block for destroy
1307 obd objects and this also block reconnect with -EALREADY. This
1308 isn't fix all lprocfs bugs - but make it rare.
1310 Severity : enhancement
1312 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
1315 Frequency : frequent on X2 node
1317 Description: mdc_set_open_replay_data LBUG
1318 Details : Set replay data for requests that are eligible for replay.
1322 Description: lustre_mgs: operation 101 on unconnected MGS
1323 Details : When MGC is disconnected from MGS long enough, MGS will evict the
1324 MGC, and late on MGC cannot successfully connect to MGS and a lot
1325 of the error messages complaining that MGS is not connected.
1329 Frequency : on network error
1330 Description: panic with double free request if network error
1331 Details : mdc_finish_enqueue is finish request if any network error ocuring,
1332 but it's true only for synchronus enqueue, for async enqueue
1333 (via ptlrpcd) this incorrect and ptlrpcd want finish request
1336 Severity : enhancement
1338 Description: client-side metadata stat-ahead during readdir(directory readahead)
1339 Details : perform client-side metadata stat-ahead when the client detects
1340 readdir and sequential stat of dir entries therein
1343 Frequency : on start mds
1345 Description: Implement get_info(last_id) in obdfilter.
1348 Frequency : occasional
1350 Description: Correctly check stale fid, not start epoch if ost not support SOM
1351 Details : open with flag O_CREATE need set old fid in op_fid3 because
1352 op_fid2 was overwritten with new generated fid, but mds can answer
1353 with one of these two fids and both is not stale. Setattr
1354 incorrectly started an epoch and assume will be called
1355 done_writing, but without SOM done_writing ever being called.
1358 Frequency : rare, depends on device drivers and load
1360 Description: MDS or OSS nodes crash due to stack overflow
1361 Details : Code changes in 1.8.0 increased the stack usage of some functions.
1362 In some cases, in conjunction with device drivers that use a lot
1363 of stack the MDS (or possibly OSS) service threads could overflow
1364 the stack. One change which was identified to consume additional
1365 stack has been reworked to avoid the extra stack usage.
1368 Frequency : occasional
1370 Description: Do not fail import if osc_interpret_create gets -EAGAIN
1371 Details : If osc_interpret_create got -EAGAIN it immediately exits and
1372 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
1373 oscc_has_objects and see OSC has no objests and call
1374 oscc_internal_create to resend create request.
1376 Severity : enhancement
1378 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
1380 Severity : enhancement
1382 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
1385 Frequency : very rare
1387 Description: Fix replay if there is an un-replied request and open
1388 Details : In some cases, older replay request will revert the
1389 mcd->mcd_last_xid on MDS which is used to record the client's
1390 latest sent request.
1392 Severity : enhancement
1394 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
1396 Severity : enhancement
1398 Description: Add rhel5 support to HEAD.
1400 Serverity : enhancement
1402 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
1407 Description: Don't allow skipping OSTs if index has been specified.
1408 Details : Don't allow skipping OSTs if index has been specified, make
1409 locking in internal create lots better.
1413 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
1414 Details : access to bitfield in structure is always rounded to long
1415 and this produce problem with not atomic change any bit.
1419 Description: Lustre make rpms failed.
1420 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
1421 builds Lustre rpms from tarball.
1424 Frequency : rare on shutdown ost
1426 Description: If llog cancel was not send before clean_exports phase, this can
1427 produce deadlock in llog code.
1428 Details : If llog thread has last reference to obd and call class_import_put
1429 this produce deadlock because llog_cleanup_commit_master wait when
1430 last llog_commit_thread exited, but this never success because was
1431 called from llog_commit_thread.
1435 Description: allow userland application know is lost one of stripes.
1436 Details : fill lvb_blocks with error code on ost and return it to
1437 application if error flag found.
1441 Description: NULL lov_tgts causing MDS oops
1442 Details : more safe checks for NULL lov_tgts for avoid oops.
1444 Severity : enhancement
1446 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
1450 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
1451 Details : If an OST is down the MDS will hang indefinitely in
1452 obd_statfs_rqset() waiting for the statfs data. While for
1453 MDS QOS usage of statfs, it should not stuck in waiting.
1455 Severity : enhancement
1457 Description: remote_acl support
1458 Details : Support ACL-based permission check for remote user.
1459 Support setfacl/getfacl for remote user with the utils
1460 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
1461 the system "{s,g}etfacl" utils.
1463 Severity : enhancement
1465 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
1467 Severity : enhancement
1469 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
1473 Description: make mgs_setparam() handle fsname containing dash
1474 Details : fsname containing a dash does not work with lctl conf_param
1476 Severity : enhancement
1478 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
1480 Severity : enhancement
1482 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
1486 Frequency : rare, at shutdown
1487 Description: access already free / zero obd_namespace.
1488 Details : if client_disconnect_export was called without force flag set,
1489 and exist connect request in flight, this can produce access to
1490 NULL pointer (or already free pointer) when connect_interpret
1491 store ocd flags in obd_namespace.
1495 Frequency : only at startup
1496 Description: not alloc memory with spinlock held.
1497 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
1498 if any spinlock held.
1500 Severity : enhancement
1502 Description: make lustre randomly fail allocating memory
1503 Details : Make lustre randomly failed allocating memory for testing purpose.
1505 Severity : enhancement
1507 Description: lost problems with lov objid file
1508 Details : Fixes some scability and access to not inited memory problems
1509 in work with lov objdid file.
1514 Description: lfs find does not continue on file error
1515 Details : Continue other files processing when a file/dir is absent.
1519 Description: Inconsistent usage of lustre_pack_reply()
1520 Details : Standardize the usage of lustre_pack_reply() such that it
1521 always generate a CERROR on failure.
1524 Frequency : occasional
1526 Description: MDS hang or stay in waiting lock
1527 Details : If client receive lock with CBPENDING flag ldlm need send lock
1528 cancel as separate rpc, to avoid situation when cancel request
1529 can't processed due all i/o threads stay in wait lock.
1533 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1537 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1541 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1545 Description: Update to RHEL4 latest kernel.
1547 Severity : enhancement
1549 Description: Build SLES10 patchless client fails
1550 Details : The configure was broken by run ./configure with
1551 --with-linux-obj=.... argument for patchless client. When the
1552 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1553 can't search header adequately. Use absolute path such as
1554 -I($LINUX)/include instead.
1558 Description: interrupt oig_wait produce painc on resend.
1559 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1560 private set, and this produce situation when rq_ptlrpcd_data not
1561 copyed to new allocated request and triggered LBUG on assert
1562 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1563 ptlrpcd set if request is changed and can be safety changed to use
1568 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1569 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1570 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1571 these macros and replace them with their difinitions instead.
1575 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1576 Details : Update which_patch & target file for SLES10 latest kernel.
1579 Description: Remove client patches from SLES 10 kernel.
1580 Details : This causes SLES 10 clients to behave as patchless clients
1581 even on a Lustre-patched (server) kernel.
1583 Severity : enhancement
1585 Description: self-adjustable client's lru lists
1586 Details : use adaptive algorithm for managing client cached locks lru
1587 lists according to current server load, other client's work
1588 pattern, memory activities, etc. Both, server and client
1589 side namespaces provide number of proc tunables for controlling
1592 Severity : enhancement
1594 Description: light-weight GSS support
1595 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
1596 and reduce performance overhead.
1598 Severity : enhancement
1600 Description: Linux keyring support
1601 Details : Support using service of Linux keyring for Lustre GSS internal
1602 context refresh/cache mechanism.
1606 Description: Fix errors in lfs documentation
1607 Details : Fixes man pages
1611 Description: don't use GFP_* in generic Lustre code.
1612 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
1616 Description: obdclass is limited by single OBD_ALLOC(idarray)
1617 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
1621 Description: warnings with build patchless client with vanila 2.6.19 and up
1622 Details : change old ctl_table style and replace ctl_table/ctl_table_header
1623 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
1627 Frequency : Only for SLES
1628 Description: sanity_quota fail test_1
1629 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
1630 file, which makes TSTID[2] inunique.
1634 Frequency : Only for SLES9
1635 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
1636 Details : Update target/ChangeLog/which_patch .
1640 Frequency : Only for test_10 in sanity-quota.sh
1641 Description: a bug in quota test code
1642 Details : checking a test flag in a wrong place
1646 Frequency : Only for test_13 in sanity-quota.sh
1647 Description: a bug in quota test script
1648 Details : shouldn't check overfull allocated quota
1652 Description: jbd statistics
1653 Details : Port older jbd statistics patch for sles10
1655 Severity : enhancement
1657 Description: Add printing inode info into message about error in writepage.
1662 Description: reply_lock_interpret crash due to race with it and lock cancel.
1663 Details : Do not replay locks that are being cancelled. Do not reference
1664 locks by their address during replay, just by their handle.
1668 Frequency : When flocks are used.
1669 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
1670 Details : Flock locks might destroy just granted lock if it could be merged
1671 with another existing flock, this is done in completion handler,
1672 so teach ldlm_cli_enquque_fini that this is a valid case for
1678 Description: Oops in read and write path when failing to allocate lock.
1679 Details : Check if lock allocation failed and return error back.
1683 Description: lstripe command fails for valid OST index
1684 Details : The stripe offset is compared to lov->desc.ld_tgt_count
1685 instead of lov->desc.ld_active_tgt_count.
1689 Description: sanity.sh failed test 103
1690 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
1695 Description: df doesn't work properly if diskfs blocksize != 4K
1696 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
1700 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
1701 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
1703 Severity : enhancement
1705 Description: Adaptive timeouts
1706 Details : RPC timeouts adapt to changing server load and network
1707 conditions to reduce resend attempts and improve recovery time.
1711 Description: llapi_file_create() does not allow some changes
1712 Details : add llapi_file_open() that allows specifying the mode and
1713 open flags, and also returns an open file handle.
1717 Description: merge and cleanup kernel patches.
1718 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
1722 Description: Add journal checksum support.(Kernel part)
1723 Details : The journal checksum feature adds two new flags i.e
1724 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
1725 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
1726 indicates that the commit block contains the checksum for
1727 the blocks described by the descriptor blocks. Now commit
1728 record can be sent to disk without waiting for descriptor
1729 blocks to be written to disk. This behavior is controlled
1730 using JBD2_FEATURE_ASYNC_COMMIT flag.
1734 Description: OSS needs mutliple precreate threads
1735 Details : Add ability to start more than one create thread per OSS.
1739 Description: Sanity.sh test_65k failed: setstripe should have succeeded
1740 Details : execute lfs setstripe on client
1744 Description: mds_obd_create error creating tmp object
1745 Details : When the user sets quota on root, llog will be affected and can't
1746 create files and write files.
1749 Frequency : Always on ia64 patchless client, and possibly others.
1751 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
1752 Details : This allows the patchless client to be loaded on architectures
1753 without this export.
1757 Description: RedHat Update kernel for RHEL5
1758 Details : Add the kernel config file for RHEL5.
1762 Description: RedHat Update kernel for RHEL5
1763 Details : Modify the kernel config file more closer RHEL5.
1767 Description: Build failure against Centos5 (RHEL5)
1768 Details : Define PAGE_SIZE when it isn't present.
1772 Description: test_7 on sanity-quota.sh can't be used on separate machines
1773 Details : add facet to handle it
1777 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
1778 Details : with NFS, the anon dentry's parent was set to itself in
1779 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
1780 corresponding dentry other than use rec->ur_name.
1782 Severity : enhancement
1784 Description: lfs setstripe enhancement
1785 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
1789 Description: enable data checksumming by default
1790 Details : enable checksum by default, allow --disable-checksum
1791 configure option and "-o nochecksum" mount option. Checksums
1792 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
1796 Description: lustre support for RHEL5
1797 Details : Add support for RHEL5.
1801 Description: Client eviction due to failover config
1802 Details : after a connection loss, the lustre client should attempt to
1803 reconnect to the last active server first before trying the
1804 other potential connections.
1808 Description: when mds and osts use different quota unit(32bit and 64bit),
1809 quota will be released repeatly.
1810 Details : void sending multiple quota reqs to mds, which will keep the
1811 status between the reqs.
1815 Description: osts not allocated evenly to files
1816 Details : change the condition to increase offset_idx
1819 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
1821 Description: Data corruption for OSTs that are formatted larger than 2TB
1823 Details : When generating the bio request for lustre file writes the
1824 sector number would overflow a temporary variable before being
1825 used for the IO. The data reads correctly from Lustre (which
1826 will overflow in a similar manner) but other file data or
1827 filesystem metadata may be corrupted in some cases.
1831 Description: Tune the kernel for good SCSI performance.
1832 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
1833 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
1838 Description: rewrite ext2-derived code in obdclass/uuid.c
1839 Details : rewrite inherited code (uuid parsing code from ext2 utils)
1840 from scratch preserving functionality.
1844 Description: "lfs find -obd UUID" prints directories
1845 Details : "lfs find -obd UUID" will return all directory names instead
1846 of just file names. It is incorrect because the directories
1847 do not reside on the OSTs.
1851 Description: use i_size_read and i_size_write in 2.6 port
1852 Details : replace inode->i_size access with i_size_read/write()
1855 Frequency : when removing large files
1857 Description: scheduling issue during removal of large Lustre files
1858 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
1859 It causes scheduling issues when removing large files (17TB in the
1863 Frequency : only with liblustre clients on XT3
1865 Description: evictions taking too long
1866 Details : allow llrd to evict clients directly on OSTs
1869 Frequency : only on ppc
1871 Description: /proc/fs/lustre/devices broken on ppc
1872 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
1873 We should make sure the type of 'index' is loff_t and then cast
1874 explicitly as needed below. Do not assign an explicitly cast
1878 Frequency : only for rhel5
1880 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
1881 Details : Modify the target file & which_kernel.
1883 Severity : enhancement
1885 Description: omit set fsid for export NFS
1886 Details : fix set/restore device id for avoid EMFILE error and mark lustre
1887 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
1891 Frequency : Always, for kernels after 2.6.16
1892 Description: Fix warning idr_remove called for id=.. which is not allocated.
1893 Details : Last kernels save old s_dev before kill super and not allow
1894 to restore from callback - restore it before call kill_anon_super.
1898 Description: buffer overruns could theoretically occur
1899 Details : llapi_semantic_traverse() modifies the "path" argument by
1900 appending values to the end of the origin string, and a
1901 overrun may occur. Adding buffer overrun check in liblustreapi.
1905 Description: Fix error on 'ls .' at the top of the Lustre mount.
1906 Details : Don't revalidate dentry if it is a root dentry.
1910 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
1911 Details : Modify vm-tunables-rhel4.patch.
1915 Description: Kernel config for 2.6.18-vanilla.
1916 Details : Modify targets/2.6-vanilla.target.in.
1917 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
1918 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
1919 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
1920 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
1924 Description: improve handling recoverable errors
1925 Details : if request processig with error which can be recoverable on server
1926 request should be resend, otherwise page released from cache and
1931 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
1932 Details : Modify target file & which_patch.
1933 A flaw was found in the IA32 system call emulation provided
1934 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
1935 value could be stored in the %RAX register, which could trigger an
1936 out-of-bounds system call table access. An untrusted local user
1937 could exploit this flaw to run code in the kernel
1938 (ie a root privilege escalation). (CVE-2007-4573).
1942 Description: O_DIRECT bypasses client statistics.
1943 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
1944 read_ahead_stats were updated. copied the stats section from
1945 osc_send_oap_rpc() into async_internal().
1949 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
1953 Description: change order of libsysio includes
1954 Details : '#include sysio.h' should always come before '#include xtio.h'
1958 Description: handle "serious error: objid * already exists" more gracefully
1959 Details : If LAST_ID value on disk is smaller than the objects existing in
1960 the O/0/d* directories, it indicates disk corruption and causes an
1961 LBUG(). If the object is 0-length, then we should use the existing
1962 object. This will help to avoid a full fsck in most cases.
1964 Severity : enhancement
1966 Description: adapt the lustre_config script to support the upgrade case
1967 Details : Add "-u" option for lustre_config script to support upgrading 1.4
1968 server targets to 1.6 in parallel.
1972 Description: To avoid grant space > avaible space when the disk is almost
1973 full. Without this patch you might see the error "grant XXXX >
1974 available" or some LBUG about grant, when the disk is almost
1976 Details : In filter_check_grant, for non_grant cache write, we should
1977 check the left space by if (*left > ungranted + bytes), instead
1978 of (*left > ungranted), because only we are sure the left space
1979 is enough for another "bytes", then the ungrant space should be
1980 increase. In client, we should update cl_avail_grant only there
1981 is OBD_MD_FLGRANT in the reply.
1985 Description: Update RHEL 4 kernel to fix local root privilege escalation.
1986 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
1987 described in CVE-2007-4573. This problem could allow untrusted
1988 local users to gain root access.
1991 Frequency : when using O_DIRECT and quotas
1993 Description: Incorrect file ownership on O_DIRECT output files
1994 Details : block usage reported by 'lfs quota' does not take into account
1995 files that have been written with O_DIRECT.
2000 Description: touch file failed when fs is not full
2001 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
2002 otherwise we can get ENOSP while fs is not full.
2006 Description: parallel lock callbacks
2007 Details : Instead of sending blocking and completion callbacks as separated
2008 requests, adding them to a set and sending in parallel.
2011 Frequency : only for Cray XT3
2012 Bugzilla : 12829/13455
2013 Description: Changing primary group doesn't change the group lustre assigns to
2015 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
2016 overridden with the primary group provided by the group upcall,
2017 whereas the supplied fsgid can be trusted if it is in the list of
2018 supplementary groups returned by the group upcall.
2020 Severity : enhancement
2022 Description: Allow masking D_WARNING, D_ERROR messages from console
2023 Details : Console messages can now be disabled via lnet.printk.
2027 Description: User code with malformed file open parameter crashes client node
2028 Details : Before packing join_file req, all the related reference should be
2029 checked carefully in case some malformed flags cause fake
2030 join_file req on client.
2034 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
2035 Details : ldlm_completion_ast() assumes that a lock is granted when the req
2036 mode is equal to the granted mode. However, it should also check
2037 that LDLM_FL_CP_REQD is not set.
2041 Description: Heavy nfs access might result in deadlocks
2042 Details : After ELC code landed, it is now improper to enqueue any mds
2043 locks under och_sem, because enqueue might want to decide to
2044 cancel open locks for same inode we are holding och_sem for.
2048 Description: Client eviction while running blogbench
2049 Details : A lot of unlink operations with concurrent I/O can lead to a
2050 deadlock causing evictions. To address the problem, the number of
2051 oustanding OST_DESTROY requests is now throttled to
2052 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
2053 callbacks are processed in priority.
2057 Description: enable ACLs on MDS by default
2058 Details : ACLs must be enabled on MDS by default.
2061 Frequency : PPC/PPC64 only
2063 Description: conflicts between asm-ppc64/types.h and lustre_types.h
2064 Details : fix duplicated definitions between asm-ppc64/types.h and
2065 lustre_types.h on PPC.
2068 Frequency : PPC/PPC64 only
2070 Description: asm-ppc/segment.h does not exist
2071 Details : fix compile issue on PPC.
2075 Description: better handle error messages in extents code
2078 Frequency : RHEL4 only
2080 Description: mkfs is very slow on IA64/RHEL4
2081 Details : A performance regression has been discovered in the MPT Fusion
2082 driver between versions 3.02.73rh and 3.02.99.00rh. As a
2083 consequence, we have downgraded the MPT Fusion driver in the RHEL4
2084 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
2086 Severity : enhancement
2088 Description: SNMP support enhancement
2089 Details : Adding total number of sampled request for an MDS node in snmp
2092 Severity : enhancement
2094 Description: Optimize ldlm waiting list processing for PR extent locks
2095 Details : When processing waiting list for read extent lock and meeting read
2096 lock that is same or wider to it that is not contended, skip
2097 processing rest of the list and immediatelly return current
2098 status of conflictness, since we are guaranteed there are no
2099 conflicting locks in the rest of the list.
2103 Description: Time out and refuse to reconnect
2104 Details : When the failover node is the primary node, it is possible
2105 to have two identical connections in imp_conn_list. We must
2106 compare not conn's pointers but NIDs, otherwise we can defeat
2107 connection throttling.
2111 Description: port llog fixes from b1_6 into HEAD
2112 Details : Port llog reference couting and some llog cleanups from b1_6
2113 (bug 10800) into HEAD, for protect from panic and access to already
2114 free llog structures.
2118 Description: Detect stride IO mode in read-ahead
2119 Details : When a client does stride read, read-ahead should detect that and
2120 read-ahead pages according to the detected stride pattern.
2124 Description: data checksumming impacts single node performance
2125 Details : add support for several checksum algorithm. Currently, only CRC32
2126 and Adler-32 are supported. The checksum type can be changed on
2127 the fly via /proc/fs/lustre/osc/*/checksum_type.
2131 Description: use adler32 for page checksums
2132 Details : when available, use the Adler-32 algorithm instead of CRC32 for
2137 Description: build for x2 fails
2138 Details : fix compile issue on Cray systems.
2142 Description: Properly match for duplicate locks
2143 Details : Due to different lock order from skiplists code, we need to
2144 traverse entire list for now
2147 Frequency : only on PPC/SLES10
2149 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
2150 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
2151 be defined. Add a hack in mkfs_lustre.c to work around this compile
2156 Description: LASSERT on MDS when client holding flock lock dies
2157 Details : ldlm pool logic depends on number of granted locks equal to
2158 number of released locks which is not true for flock locks, so
2159 just exclude such locks from consideration.
2163 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
2164 Details : Use fsfilt_sync() for both the whole filesystem sync and
2165 individual file sync to eliminate dangerous inode locking
2166 with I_LOCK that can lead to a deadlock.
2170 Description: performance in 1.6.3
2171 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
2172 sure that 1MB requests are not fragmented by the block layer.
2174 Severity : enhancement
2176 Description: organize the server-side client stats on per-nid basis
2177 Details : Change the structure of stats under obdfilter and mds to
2185 The "uuid"s file would list the uuids of _active_ exports.
2186 And the clear entry is to clear all stats and stale nids.
2188 Severity : enhancement
2190 Description: eliminate client locks in face of contention
2191 Details : file contention detection and lockless i/o implementation
2192 for contended files.
2196 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
2199 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
2201 Description: MDS may lose file striping (and hence file data) in some cases
2202 Details : If there are additional extended attributes stored on the MDS,
2203 in particular ACLs, SELinux, or user attributes (if user_xattr
2204 is specified for the client mount options) then there is a risk
2205 of attribute loss. Additionally, the Lustre file striping
2206 needs to be larger than default (e.g. striped over all OSTs),
2207 and an additional attribute must be stored initially in the
2208 inode and then increase in size enough to be moved to the
2209 external attribute block (e.g. ACL growing in size) for file
2214 Description: skiplist implementation simplification
2215 Details : skiplists are used to group compatible locks on granted list
2216 that was implemented as tracking first and last lock of each
2217 lock group the patch changes that to using doubly linked lists
2221 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
2222 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
2223 because LASSERT on both the data supplied by a client, and the
2224 data on disk is dangerous and incorrect.
2226 Severity : enhancement
2228 Description: Slow truncate/writes to huge files at high offsets.
2229 Details : Directly associate cached pages to lock that protect those pages,
2230 this allows us to quickly find what pages to write and remove
2231 once lock callback is received.
2235 Description: more ldlm soft lockups
2236 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
2237 starve other threads from the resource lock for a long time in
2238 case of long waiting queue, so change the debug level from
2239 D_OTHER to the less frequently used D_INFO.
2241 Severity : enhancement
2243 Description: add -gid, -group, -uid, -user options to lfs find
2247 Description: Hung threads in invalidate_inode_pages2_range
2248 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
2249 once one is completed. As a result, some RPCs are stuck in the
2250 queue and are never sent.
2254 Description: filter threads hungs on waiting journal commit
2255 Details : Cleanup filter group llog code, then only filter group llog will
2256 be only created in the MDS/OST syncing process.
2260 Description: Procfs and llog threads access destoryed import sometimes.
2261 Details : Sync the import destoryed process with procfs and llog threads by
2262 the import refcount and semaphore.
2264 Severity : enhancement
2266 Description: openlock cache of b1_6 port to HEAD
2271 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
2272 Details : If insertion of an extent fails, then discard the inode
2273 preallocation and free data blocks else it can lead to duplicate
2278 Description: don't always update ctime in ext3_xattr_set_handle()
2279 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
2280 In some cases the ctime should not be updated, for example for
2281 2.0->1.8 compatibility it is necessary to delete an xattr and it
2282 should not update the ctime.
2286 Bugzilla : 15713/16362
2287 Description: Assertion in iopen_connect_dentry in 1.6.3
2288 Details : looking up an inode via iopen with the wrong generation number can
2289 populate the dcache with a disconneced dentry while the inode
2290 number is in the process of being reallocated. This causes an
2291 assertion failure in iopen since the inode's dentry list contains
2292 both a connected and disconnected dentry.
2296 Description: assertion failure in ldlm_handle2lock()
2297 Details : fix a race between class_handle_unhash() and class_handle2object()
2298 introduced in lustre 1.6.5 by bug 13622.
2303 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
2304 Details : remove the limit of 256 scsi disks in the sd_iostat patch
2309 Description: oops in sd_iostats_seq_show()
2310 Details : unloading/reloading the scsi low level driver triggers a kernel
2311 bug when trying to access the sd iostat file.
2316 Description: Kernel panics during QLogic driver reload
2317 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
2318 patch, causing memory corruption.
2323 Description: journal_dev option does not work in b1_6
2324 Details : pass mount option during pre-mount.
2326 Severity : enhancement
2328 Description: Add a FIEMAP(FIle Extent MAP) ioctl
2329 Details : FIEMAP ioctl will allow an application to efficiently fetch the
2330 extent information of a file. It can be used to map logical blocks
2331 in a file to physical blocks in the block device.
2335 Description: LDLM soft lockups - improvement
2336 Details : It is be possible to send the lock handle along with each read
2337 or write request because the client is already doing a lock match
2338 itself so there isn't any reason the OST should have to re-do that
2342 Frequency : only on Cray X2
2344 Description: X2 build failures
2345 Details : fix build failures on Cray X2.
2349 Description: xid & resent requests
2350 Details : Initialize RPC XID from clock at startup (randomly if clock is
2353 Severity : enhancement
2355 Description: Add lustre_start utility to start or stop multiple Lustre servers
2360 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
2361 Details : In case of memory pressure, list_del() can be called twice on
2362 req->rq_history_list, causing a kernel oops.
2366 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
2367 Details : fix stack overflow in the distributed lock manager by defering
2368 export eviction after a failed AST to the elt thread instead of
2369 handling it in the dlm interpret routine.
2373 Description: Convert some comments to new format.
2374 Details : Update documenting comments to match doxygen conventions.
2378 Description: Grammar fixes.
2379 Details : A couple of trivial spelling fixes.
2383 Description: OSD_COUNTERS-mandatory
2384 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
2385 to simplify the code.
2389 Description: simplify cmm_device freeing logic.
2390 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
2394 Description: Add lockdep support to dt_object_operations locking interface.
2395 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
2396 parameter indicating lock ordering. Update mdd code to use new
2401 Description: Introduce failloc constants for lockless IO tests.
2402 Details : Add two new failloc constants to test lockless IO. Only one of
2403 them in implemented---another is checked in yet to be landed
2408 Description: Add lockdep support for inode mutex.
2409 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
2410 in the situations where Lustre has to lock more than one inode
2415 Description: Add optional invariants checking support.
2416 Details : Add new LINVRNT() macro, optional on new --enable-invariants
2417 configure switch. This macro is to be used for consistency and
2418 sanity checks that are too expensive to be left in `production'
2423 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
2424 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
2429 Description: Add lockdep support for ldlm_lock and ldlm_resource.
2430 Details : Use spin_lock_nested() in (the only) situation where more than
2431 one ldlm_lock is locked simultaneously. Also, fix possible
2432 dead-lock in ldlm_lock_change_resource() by enforcing particular
2437 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
2438 Details : Instead of specifying each ldlm_lock call-back through separate
2439 parameter, wrap them into struct ldlm_callback_suite.
2443 Description: Kill join_lru obd method and its callers.
2444 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
2445 in memory---a requirement implemented through ->o_join_lru() obd
2446 method in HEAD. Remove this method and its users.
2450 Description: Add asynchronous ldlm ENQUEUE completion handler.
2451 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
2452 case that stock ldlm_completion_ast() cannot handle as it waits
2453 until lock is granted. Introduce new ldlm_completion_ast_async()
2454 for this. Also comment ldlm_completion_ast().
2458 Description: ldlm_error <-> errno conversion.
2459 Details : Add functions to map (rather arbitrary) between LDLM error codes
2460 and standard errno values. CLIO needs this to prevent LDLM specific
2461 constants from escaping ldlm and osc.
2465 Description: Kill unused ldlm_handle2lock_ns() function.
2466 Details : Kill unused ldlm_handle2lock_ns() function.
2470 Description: Add lu_ref support to ldlm_lock
2471 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
2472 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
2473 and ldlm_resource. LDLM interface has to be changed, because code that
2474 releases a reference on a lock, has to "know" what reference this is.
2475 In the most frequent case
2477 lock = ldlm_handle2lock(handle);
2479 LDLM_LOCK_PUT(lock);
2481 no changes are required. When any other reference (received _not_ from
2482 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
2483 instead of LDLM_LOCK_PUT().
2485 Arguably, changes are pervasive, and interface requires some discipline
2486 for proper use. On the other hand, it was very instrumental in finding
2487 a few leaked lock references.
2491 Description: Add ldlm_lock_addref_try().
2492 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
2493 attempts to addref a lock that might be being canceled
2498 Description: Add ldlm_weigh_callback().
2499 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
2500 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
2501 hard-coded `number of pages' logic.
2505 Description: Add lockdep annotations to llog code.
2506 Details : Use appropriately tagged _nested() locking calls in the places
2507 where llog takes more than one ->lgh_lock lock.
2511 Description: Add loi_kms_set().
2512 Details : Wrap kms updates into a helper function.
2516 Description: Constify instances of struct lsm_operations.
2517 Details : Constify instances of struct lsm_operations.
2521 Description: lu_conf support.
2522 Details : On a server, a file system object is uniquely identified
2523 by a fid, which is sufficient to locate and load all object
2524 state (inode). On a client, on the other hand, more data are
2525 necessary instantiate an object. Change lu_object_find() and
2526 friends to take additional `lu_conf' argument describing object.
2527 Typically this includes layout information.
2531 Description: lu_context fixes.
2532 Details : Introduce new lu_context functions that are needed on the client
2533 side, where some system threads (ptlrpcd) are shared by multiple
2534 modules, and so cannot be stopped during module shutdown.
2538 Description: Add start and stop methods to lu_device_type_operations.
2539 Details : Introduce two new methods in lu_device_type_operations, that are
2540 invoked when first instance of a given type is created and last one
2541 is destroyed respectively. This is need by CLIO.
2545 Description: Add lu_ref support to struct lu_device.
2546 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
2547 track leaked references.
2551 Description: Introduce lu_kmem_descr.
2552 Details : lu_kmem_descr and its companion interface allow to create
2553 and destroy a number of kmem caches at once.
2557 Description: Fix lu_object finalization race.
2558 Details : Fix a race between lu_object_find() finding an object and its
2559 concurrent finalization. This race is (most likely) not possible
2560 on the server, but might happen on the client.
2564 Description: Introduce lu_ref interface.
2565 Details : lu_ref is a debugging module allowing to track references to
2566 a given object. It is quite cpu expensive, and has to be
2567 explicitly enabled with --enable-lu_ref. See usage description
2572 Description: Factor lu_site procfs stats into a separate function.
2573 Details : Separate lu_site stats printing code into a separate function
2574 that can be reused on a client.
2578 Description: Constify instances of struct {lu,dt,md}_device_operations.
2579 Details : Constify instances of struct {lu,dt,md}_device_operations.
2583 Description: Introduce struct md_site and move meta-data specific parts of
2584 struct lu_site here.
2585 Details : Move md-specific fields out of struct lu_site into special struct
2586 md_site, so that lu_site can be used on a client.
2590 Description: Kill mdd_lov_destroy().
2591 Details : Remove unused mdd code.
2595 Description: Add st_block checking to multistat.c.
2596 Details : Add st_block checking to multistat.c.
2600 Description: Add lu_ref support to struct obd_device.
2601 Details : Add lu_ref tracking to obd_device.
2605 Description: Kill obd_set_fail_loc().
2606 Details : Remove unused code.
2610 Description: Add special type for ptlrpc_request interpret functions.
2611 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
2612 there. Actual usage will be in CLIO.
2616 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
2617 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
2618 doesn't work with lockdep.
2622 Description: Add rwv.c test program.
2623 Details : New testing program exercising readv(2) and writev(2) (Qian).
2627 Description: Add sendfile.c test program.
2628 Details : New testing program exercising sendfile(2) (Jay).
2632 Description: Ratelimit a message that can be very frequent.
2633 Details : Ratelimit a memory allocation failure message that can
2638 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2639 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2643 Description: Liblustre build fixes.
2644 Details : Liblustre build fixes.
2648 Description: libcfs: add cfs_{need,cond}_resched() interface.
2649 Details : libcfs: add cfs_{need,cond}_resched() definition and
2650 implementations for Linux, NT, and liblustre.
2652 Severity : enhancement
2654 Description: More exported tunables for mballoc
2655 Details : Add support for tunable preallocation window and new tunables for
2656 large/small requests
2660 Description: Detect corruption of block bitmap and checking for preallocations
2661 Details : Checks validity of on-disk block bitmap. Also it does better
2662 checking of number of applied preallocations. When corruption is
2663 found, it turns filesystem readonly to prevent further corruptions.
2667 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
2668 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
2669 window to 0. If it does hit cache miss, and read-pattern is still
2670 stride-io mode, does not reset the stride window, but also does not
2671 increase the stride window length in this case.
2675 Frequency : only for big-endian servers
2676 Description: Check if system is big-endian while mounting fs with extents feature
2677 Details : Mounting a filesystem with extents feature will fail on big-endian
2678 systems since ext3-based ldiskfs is not supported on big-endian
2679 systems. This can be over-riden with "bigendian_extents" mount option.
2681 Severity : enhancement
2683 Description: The root squash functionality
2684 Details : A security feature, which is to prevent users from being able
2685 to mount lustre on their desktop, run as root, and delete
2686 all of the files in the filesystem. The goal is accomplished by
2687 remapping user id (UID) and group id (GID) of the root user to
2688 a UID and GID specified by the system administartor via Lustre
2689 configuration management server (MGS). The functionality also
2690 allows to specify sets of clients for which the remapping does
2695 Description: Excessive recovery window
2696 Details : With AT enabled, the recovery window can be excessively long (6000+
2697 seconds). To address this problem, we no longer use
2698 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
2699 timeout no longer depends on the service time, it is set to
2700 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
2701 time via pb_service_time.
2705 Description: Watchdog triggered on MDS failover
2706 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
2707 the OSTs know that the MDS "UUID" can be reused for the same export
2708 from a different NID, so we do not need to wait for the export to be
2712 Frequency : rare, only if using MMP with Linux RAID
2714 Description: MMP doesn't work with Linux RAID
2715 Details : While using HA for Lustre servers with Linux RAID, it is possible
2716 that MMP will not detect multiple mounts. To make this work we
2717 need to unplug the device queue in RAID when the MMP block is being
2718 written. Also while reading the MMP block, we should read it from
2719 disk and not the cached one.
2721 Severity : enhancement
2723 Description: open file using fid
2724 Details : A file can be opened using just its fid, like
2725 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
2728 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
2731 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
2732 Details : add proper locking for creation of jbd_4k slab cache
2736 Description: MMP check in ext3_remount() fails without displaying any error
2737 Details : When multiple mount protection fails during remount, proper error
2740 Severity : enhancement
2742 Description: Allow stripe size to be up to 4G-64k
2743 Details : Fix math logic to allow large stripe sizes.
2747 Description: add check for >8TB ldiskfs filesystems
2748 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
2749 Don't allow >8TB ldiskfs filesystems to be mounted without
2750 force_over_8tb mount option
2755 Descriptoin: OSS complains "too many blocks" and crashes
2756 Details : In filter_commitrw_write():
2757 1. lquota_chkquota();
2758 2. obdo_from_inode();
2759 3. lquota_pending_commit();
2760 "2." has changed uid and gid in this case, then caused
2761 lquota_chkquota() and lquota_pending_commit() to use
2762 different uid and gid. That is the root of the bug.
2766 Descriptoin: getstripe doesn't report stripe count = -1 sanely
2767 Details : Detect and print out the 'stripe count = -1' case properly.
2771 Description: Out or order replies might be lost on replay
2772 Details : In ptlrpc_retain_replayable_request if we cannot find retained
2773 request with tid smaller than one currently being added, add it
2774 to the start, not end of the list.
2776 --------------------------------------------------------------------------------
2778 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
2780 * Support for kernels:
2781 2.6.5-7.283 (SLES 9),
2782 2.6.9-55.EL (RHEL 4),
2783 2.6.16.46-0.14 (SLES 10),
2784 2.6.18.8 vanilla (kernel.org)
2785 * Client support for unpatched kernels:
2786 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2787 2.6.16 - 2.6.22 vanilla (kernel.org)
2788 * Due to recently discovered recovery problems, we do not recommend
2789 using patchless RHEL 4 clients with this or any earlier release.
2790 * Recommended e2fsprogs version: 1.39.cfs8
2791 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2792 * Starting with this release, the ldiskfs backing filesystem required
2793 by Lustre is now in its own package, lustre-ldiskfs. This package
2794 should be installed. It is versioned separately from Lustre and
2795 may be released separately in future.
2799 Description: block reactivating mgc import until all deactivates complete
2800 Details : Fix race when failing back MDT/MGS to itself (testing)
2803 Frequency : at statup only
2805 Description: mds_lov_synchronize race leads to various problems
2806 Details : simultaneous MDT->OST connections at startup can cause the
2807 sync to abort, leaving the OSC in a bad state.
2809 Severity : enhancement
2811 Description: add optional extra BUILD_VERSION info
2812 Details : add a new environment variable (namely LUSTRE_VERS) which allows
2813 to override the lustre version.
2816 Frequency : 2.6.18 servers only
2818 Description: ll_kern_mount() doesn't release the module reference
2819 Details : The ldiskfs module reference count never drops down to 0
2820 because ll_kern_mount() doesn't release the module reference.
2825 Description: server LBUG when using old ost_num_threads parameter
2826 Details : Accept the old ost_num_threads parameter but warn that it
2827 is deprecated, and fix an off-by-one error that caused an LBUG.
2832 Description: Transient SCSI error results in persistent IO issue
2833 Details : iobuf->dr_error is not reinitialized to 0 between two
2837 Frequency : sometimes when underlying device returns I/O errors
2839 Description: OSTs not going read-only during write failures
2840 Details : OSTs are not remounted read-only when the journal commit threads
2841 get I/O errors because fsfilt_ext3 calls journal_start/stop()
2842 instead of the ext3 wrappers.
2846 Description: poor connect scaling with increasing client count
2847 Details : Don't run filter_grant_sanity_check for more than 100 exports
2848 to improve scaling for large numbers of clients.
2851 Frequency : SLES10 only
2853 Description: sanity-quota.sh quotacheck failed: rc = -22
2854 Details : Quotas cannot be enabled on SLES10.
2857 Frequency : liblustre clients only
2859 Description: getdirentries does not give error when run on compute nodes
2860 Details : getdirentries does not fail when the size specified as an argument
2861 is too small to contain at least one entry
2863 Severity : enhancement
2865 Description: Add LNET router traceability for debug purposes
2866 Details : If a checksum failure occurs with a router as part of the
2867 IO path, the NID of the last router that forwarded the bulk data
2868 is printed so it can be identified.
2873 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
2874 Details : Due to a race condition, liblustre clients were occasionally
2875 evicted incorrectly.
2877 Severity : enhancement
2879 Description: lfs setstripe use optional parameters instead of postional
2882 Severity : enhancement
2884 Description: Nanosecond timestamp support for ldiskfs
2885 Details : The on-disk ldiskfs filesystem has added support for nanosecond
2886 resolution timestamps. There is not yet support for this at
2887 the Lustre filesystem level.
2890 Frequency : during server recovery
2892 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
2893 Details : request with rq_no_resend flag not awake l_wait_event if they get
2897 Frequency : nfs export on patchless client
2899 Description: connectathon hang when test nfs export over patchless client
2900 Details : Disconnected dentry cannot be found with lookup, so we do not need
2901 to unhash it or make it invalid
2904 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
2905 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
2906 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
2911 Description: open req refcounting wrong on reconnect
2912 Details : If reconnect happened between getting open reply from server and
2913 call to mdc_set_replay_data in ll_file_open, we will schedule
2914 replay for unreferenced request that we are about to free.
2915 Subsequent close will crash in variety of ways.
2916 Check that request is still eligible for replay in
2917 mdc_set_replay_data().
2922 Description: disable writes to filesystem when reading health_check file
2923 Details : the default for reading the health_check proc file has changed
2924 to NOT do a journal transaction and write to disk, because this
2925 can cause reads of the /proc file to hang and block HA state
2926 checking on a healthy but otherwise heavily loaded system. It
2927 is possible to return to the previous behaviour during configure
2928 with --enable-health-write.
2930 Severity : enhancement
2932 Description: 64-bit inode version
2933 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
2934 to the inode. This will be required for version-based recovery.
2939 Description: MDS fails to start if a duplicate client export is detected
2940 Details : in some rare cases it was possible for a client to connect to
2941 an MDS multiple times. Upon recovery the MDS would detect this
2942 and fail during startup. Handle this more gracefully.
2944 Severity : enhancement
2946 Description: Add -o localflock option to simulate old noflock
2948 Details : This will achieve local-only flock/fcntl locks
2954 Description: log_commit_thread vs filter_destroy race leads to crash
2955 Details : Take import reference before releasing llog record semaphore
2960 Description: Wrong request locking in request set processing
2961 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
2962 imp_delayed_list, in this place should be used imp_lock.
2965 Frequency : when reconnection
2967 Description: Grant Leak when osc reconnect to OST
2968 Details : When osc reconnect ost, OST(filter) should check whether it
2969 should grant more space to client by comparing fed_grant and
2970 cl_avail_grant, and return the granted space to client instead
2971 of "new granted" space, because client will call osc_init_grant
2972 to update the client grant space info.
2975 Frequency : when client reconnect to OST
2977 Description: Grant Leak when osc do resend and replay bulk write
2978 Details : When osc reconnect to OST, OST(filter)should clear grant info of
2979 bulk write request, because the grant info while be sync between
2980 OSC and OST when reconnect, and we should ignore the grant info
2981 these of resend/replay write req.
2986 Description: Grant space more than available left space sometimes.
2987 Details : When then OST is about to be full, if two bulk writing from
2988 different clients came to OST. Accord the avaliable space of the
2989 OST, the first req should be permitted, and the second one
2990 should be denied by ENOSPC. But if the seconde arrived before
2991 the first one is committed. The OST might wrongly permit second
2992 writing, which will cause grant space > available space.
2995 Frequency : when client is evicted
2997 Description: Grant might be wrongly erased when osc is evicted by OST
2998 Details : when the import is evicted by server, it will fork another
2999 thread ptlrpc_invalidate_import_thread to invalidate the
3000 import, where the grant will be set to 0. While the original
3001 thread will update the grant it got when connecting. So if
3002 the former happened latter, the grant will be wrongly errased
3003 because of this race.
3008 Description: Checking Stale with correct fid
3009 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
3010 to check whether it is stale, because sometimes, we want the
3011 enqueue happened anyway, and op_data.fid2 will not be initialized.
3013 Severity : enhancement
3015 Description: update patchless client
3016 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
3019 Frequency : only with 2.4 kernel
3021 Description: random memory corruption
3022 Details : size of struct ll_inode_info is to big for union inode.u and this
3023 can be cause of random memory corruption.
3028 Description: Memory leak in recovery
3029 Details : Lov_mds_md was not free in an error handler in mds_create_object.
3030 It should also check obd_fail before fsfilt_start, otherwise if
3031 fsfilt_start return -EROFS,(failover mds during mds recovery).
3032 then the req will return with repmsg->transno = 0 and rc = EROFS.
3033 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
3034 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
3035 be freed no matter whether fsfilt_commit success or not.
3038 Frequency : only with huge count clients
3040 Description: Prevents from taking the superblock lock in llap_from_page for
3042 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
3043 for llap_from_page call in ll_removepage prevents from taking the
3044 superblock lock for a soon died page.
3049 Description: Not check open intent error before release open handle
3050 Details : in some rare cases, the open intent error is not checked before
3051 release open handle, which may cause
3052 ASSERTION(open_req->rq_transno != 0), because it tries to release
3053 the failed open handle.
3058 Description: Set cat log bitmap only after create log success.
3059 Details : in some rare cases, the cat log bitmap is set too early. and it
3060 should be set only after create log success.
3064 Description: Accessing a block bevice can re-enable I/O when Lustre is
3065 tearing down a device.
3066 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
3071 Description: service threads may hog cpus when there are a lot of requests
3072 Details : Insert cond_resched to give other threads a chance to use some CPU
3077 Description: the cat log was not initialized in recovery
3078 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
3079 unlink log on mds will not be initialized until mds post
3080 recovery. And also in mds post recovery, the unlink log will
3081 initialization will be done asynchronausly, so there will be race
3082 between add unlink log and unlink log initialization.
3086 Description: brw_stats were being printed incorrectly
3087 Details : brw_stats were being printed as log2 but all of them were not
3088 recorded as log2. Also remove some code duplication arising from
3089 filter_tally_{read,write}.
3093 Frequency : rare, only in recovery.
3094 Description: ASSERTION(req->rq_type != LI_POISON) failed
3095 Details : imp_lock should be held while iterating over imp_sending_list for
3096 prevent destroy request after get timeout in ptlrpc_queue_wait.
3100 Description: replay-single.sh test 52 fails
3101 Details : A lock's skiplist need to be cleanup when it being unlinked
3102 from its resource list.
3106 Description: Short directio read returns full requested size rather than
3108 Details : Direct I/O operations should return actual amount of bytes
3109 transferred rather than requested size.
3111 Severity : enhancement
3113 Description: metadata RPC reduction (e.g. for rm performance)
3114 Details : decrease the amount of synchronous RPC between clients and servers
3115 by canceling conflicing lock before the operation on the client
3116 and packing thier handles into the main operation RPC to server.
3118 Severity : enhancement
3120 Description: Async OSC create to avoid the blocking unnecessarily.
3121 Details : If an OST has no remaining object, system will block on the
3122 creation when it needs to create a new object on this OST. Now,
3123 ways use pre-created objects when available, instead of blocking on
3124 an empty osc while others are not empty. If we must block, we
3125 block for the shortest possible period of time.
3129 Description: improve handling recoverable errors
3130 Details : if request processig with error which can be recoverable on server
3131 request should be resend, otherwise page released from cache and
3134 Severity : enhancement
3136 Description: refine locking for avoid write wrong info into lov_objid file
3137 Details : fix possible races with add new target and write/update data in
3140 --------------------------------------------------------------------------------
3142 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
3147 Frequency : on some architectures
3149 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
3150 Details : Uninitialized flags sometimes cause configuration commands to
3154 Frequency : patchless clients only
3156 Description: missing __iget() symbol export
3157 Details : The __iget() symbol export is missing. To avoid the need for
3158 this on patchless clients the deathrow inode reaper is turned
3159 off, and we depend on the VM to clean up old inodes. This
3160 dependency was during via the fix for bug 12181.
3162 --------------------------------------------------------------------------------
3164 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
3166 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
3167 INTEROPERATE with older versions automatically. In many cases a
3168 special upgrade step is needed. Please read the
3169 user documentation before upgrading any part of a live system.
3170 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
3171 version will not interoperate with 1.6 betas before beta5 (1.5.95).
3172 * WARNING: Lustre configuration and startup changes are required with
3173 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
3177 Severity : enhancement
3179 Description: MountConf
3180 Details : Lustre configuration is now managed via mkfs and mount
3181 commands instead of lmc and lconf. New obd types (MGS, MGC)
3182 are added for dynamic configuration management. See
3183 https://mail.clusterfs.com/wikis/lustre/MountConf for
3186 Severity : enhancement
3188 Description: dynamic OST addition
3189 Details : OSTs can now be added to a live filesystem
3191 Severity : enhancement
3193 Description: startup order invariance
3194 Details : MDTs and OSTs can be started in any order. Clients only
3195 require the MDT to complete startup.
3197 Severity : enhancement
3199 Description: parallel, asynchronous orphan cleanup
3200 Details : orphan cleanup is now performed in separate threads for each
3201 OST, allowing parallel non-blocking operation.
3203 Severity : enhancement
3205 Description: optimized stripe assignment
3206 Details : stripe assignments are now made based on ost space available,
3207 ost previous usage, and OSS previous usage, in order to try
3208 to optimize storage space and networking resources.
3210 Severity : enhancement
3212 Description: Permanently set tunables
3213 Details : All writable /proc/fs/lustre tunables can now be permanently
3214 set on a per-server basis, at mkfs time or on a live system.
3216 Severity : enhancement
3218 Description: Lustre message v2
3219 Details : Add lustre message format v2.
3221 Severity : enhancement
3223 Description: client OST exclusion list
3224 Details : Clients can be started with a list of OSTs that should be
3225 declared "inactive" for known non-responsive OSTs.
3229 Description: SPEC SFS validation failure on NFS v2 over lustre.
3230 Details : Changes the blocksize for regular files to be 2x RPC size,
3231 and not depend on stripe size.
3233 Severity : enhancement
3235 Description: Multiple MD RPCs in flight.
3236 Details : Further unserialise some read-only MDS RPCs - learn about intents.
3237 To avoid overly-overloading MDS, introduce a limit on number of
3238 MDS RPCs in flight for a single client and add /proc controls
3239 to adjust this limit.
3241 Severity : enhancement
3243 Description: client read/write statistics
3244 Details : Add client read/write call usage stats for performance
3245 analysis of user processes.
3246 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
3247 file access. extents_stats shows chunk size distribution.
3248 extents_stats_per_process show chunk size distribution per
3251 Severity : enhancement
3253 Description: mds statistics
3254 Details : Add detailed mds operations statistics in
3255 /proc/fs/lustre/mds/*/stats.
3259 Description: Failure of copying files with lustre special EAs.
3260 Details : Client side always return success for setxattr call for lustre
3261 special xattr (currently only "trusted.lov").
3266 Description: Refcount LNET uuids
3267 Details : The global LNET uuid list grew linearly with every startup;
3268 refcount repeated list entries instead of always adding to
3271 Severity : enhancement
3273 Description: Dynamic service threads
3274 Details : Within a small range, start extra service threads
3275 automatically when the request queue builds up.
3278 Frequency : mixed-endian client/server environments
3280 Description: mixed-endian crashes
3281 Details : The new msg_v2 system had some failures in mixed-endian
3284 Severity : enhancement
3286 Description: Easy OST removal
3287 Details : OSTs can be permanently deactivated with e.g. 'lctl
3288 conf_param lustre-OST0001.osc.active=0'
3290 Severity : enhancement
3292 Description: MGS proc entries
3293 Details : Added basic proc entries for the MGS showing what filesystems
3296 Severity : enhancement
3298 Description: provide MGS failover
3299 Details : Added config lock reacquisition after MGS server failover.
3301 Severity : enhancement
3303 Description: add Linux 2.4 support
3304 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
3308 Description: a large application tries to do I/O to the same resource and dies
3309 in the middle of it.
3310 Details : Check the req->rq_arrival time after the call to
3311 ost_brw_lock_get(), but before we do anything about
3312 processing it & sending the BULK transfer request. This
3313 should help move old stale pending locks off the queue as
3314 quickly as obd_timeout.
3317 Frequency : when an incorrect nid is specified during startup
3319 Description: ptlrpc connect to non-existant node causes kernel crash
3320 Details : LNET can't be re-entered from an event callback, which
3321 happened when we expire a message after the export has been
3322 cleaned up. Instead, hand the zombie cleanup off to another
3325 Severity : enhancement
3327 Description: plain/inodebits lock performance improvement
3328 Details : Grouping plain/inodebits in granted list by their request modes
3329 and bits policy, thus improving the performance of search through
3333 Frequency : only if OST filesystem is corrupted
3335 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3336 Details : for a short time RPCs with bulk IO are in the replay list,
3337 but replay of bulk IOs is unimplemented. If the OST filesystem
3338 is corrupted due to disk cache incoherency and then replay is
3339 started it is possible to trip an assertion. Avoid putting
3340 committed RPCs into the replay list at all to avoid this issue.
3343 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3345 Description: System hang on startup
3346 Details : This bug allowed the liblustre (e.g. catamount) client to
3347 return to the app before handling all startup RPCs. This
3348 could leave the node unresponsive to lustre network traffic
3349 and manifested as a server ptllnd timeout.
3351 Severity : enhancement
3353 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3354 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
3355 boolean to control whether to print peer debug info when a
3356 client's RPC times out.
3359 Frequency : only for kernels with patches from Lustre below 1.4.3
3361 Description: Remove old rdonly API
3362 Details : Remove old rdonly API which unused from at least lustre 1.4.3
3365 Frequency : only for devices with external journals
3367 Description: Set external device read-only also
3368 Details : During a commanded failover stop, we set the disk device
3369 read-only while the server shuts down. We now also set any
3370 external journal device read-only at the same time.
3373 Frequency : when upgrading from 1.4 while trying to change parameters
3375 Description: The wrong (new) MDC name was used when setting parameters for
3376 upgraded MDT's. Also allows changing of OSC (and MDC)
3377 parameters if --writeconf is specified at tunefs upgrade time.
3380 Frequency : when setting specific ost indicies
3382 Description: QOS code breaks on skipped indicies
3383 Details : Add checks for missing OST indicies in the QOS code, so OSTs
3384 created with --index need not be sequential.
3386 Severity : enhancement
3388 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3389 Details : The uninit_groups feature works in conjunction with the kernel
3390 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3391 up the pass1 processing of e2fsck. This is a read-only feature
3392 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3393 mount filesystems that have had this feature enabled.
3395 Severity : enhancement
3397 Description: Improve multi-block allocation algorithm to avoid fragmentation
3398 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3399 allocation locality and avoid filesystem fragmentation.
3401 ------------------------------------------------------------------------------
3403 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
3405 * Support for kernels:
3406 2.4.21-47.0.1.EL (RHEL 3)
3407 2.6.5-7.283 (SLES 9)
3408 2.6.9-42.0.10.EL (RHEL 4)
3409 2.6.12.6 vanilla (kernel.org)
3410 2.6.16.27-0.9 (SLES 10)
3411 * Recommended e2fsprogs version: 1.39.cfs5
3413 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3417 Frequency : occasional, depends on client load and configuration
3418 Bugzilla : 12181, 12203
3419 Description: data loss for recently-modified files
3421 Details : In some cases it is possible that recently written or created
3422 files may not be written to disk in a timely manner (this should
3423 normally be within 30s unless client IO load is very high).
3424 The problem appears as zero-length files or files that are a
3425 multiple of 1MB in size after a client crash or client eviction
3426 that are missing data at the end of the file.
3428 This problem is more likely to be hit on clients where files are
3429 repeatedly created and unlinked in the same directory, clients
3430 have a large amount of RAM, have many CPUs, the filesystem has
3431 many OSTs, the clients are rebooted frequently, and/or the files
3432 are not accessed by other nodes after being written.
3434 The presence of the problem can be detected by looking at
3435 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
3436 smaller than the second (nr_unused) then dirty files will not
3437 be flushed automatically to disk. "sync; sleep 10" should be
3438 run several times on the node before unmounting it to update
3439 Lustre (this is also safe to run on nodes without this problem).
3441 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
3442 that can cause this same problem, so customers using that kernel
3443 also need to update the kernel in addition to Lustre. In order
3444 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
3446 It is normal that files written just before a client crash (less
3447 than 30s) may not yet have been flushed to disk, even for local
3451 Frequency : frequent on thin XT3 nodes
3453 Description: UUID collision on thin XT3 Linux nodes
3454 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
3455 randomly, since we relied on an insufficiently-seeded PRNG.
3460 Description: OSS hangs after "All ost request buffers busy"
3461 Details : A deadlock between quota and journal operations caused OSS
3462 hangs after printing "All ost request buffers busy."
3465 Frequency : always on liblustre builds
3467 Description: Cleanup compiler warnings on liblustre
3470 Frequency : always on liblustre builds on XT3
3472 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
3476 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
3477 > 32000 subdirectories
3478 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
3479 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
3480 subdirectory count crosses 32000. This will aid e2fsck to
3481 correctly handle more than 32000 subdirectories.
3484 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3486 Description: System hang on startup
3487 Details : This bug allowed the liblustre (e.g. catamount) client to
3488 return to the app before handling all startup RPCs. This
3489 could leave the node unresponsive to lustre network traffic
3490 and manifested as a server ptllnd timeout.
3492 Severity : enhancement
3494 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3495 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
3496 boolean to control whether to print peer debug info when a
3497 client's RPC times out.
3502 Description: make O_SYNC working on 2.6 kernels
3503 Details : 2.6 kernels use different method for mark pages for write,
3504 so need add a code to lustre for O_SYNC work.
3509 Description: Failure to close file and release space on NFS
3510 Details : Put inode details into lock acquired in ll_intent_file_open.
3511 Use mdc_intent_lock in ll_intent_open to properly
3512 detect all kind of errors unhandled by mdc_enqueue.
3517 Description: proc file read during shutdown sometimes raced obd removal,
3519 Details : Add lock to prevent obd access after proc file removal.
3522 Frequency : Only for files larger than 4GB on 32-bit clients.
3524 Description: improperly doing page alignment of locks
3525 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
3526 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3531 Description: RPCs being resent when they shouldn't be
3532 Details : Some RPCs that should not be resent are being resent. This
3533 can cause inconsistencies in the RPC state machine. Do not
3534 resend such requests.
3537 Frequency : rare, only with NFS export
3539 Description: Crash on NFS re-export node
3540 Details : under very unusual load conditions an assertion is hit in
3541 ll_intent_file_open()
3544 Frequency : only if OST filesystem is corrupted
3546 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3547 Details : for a short time RPCs with bulk IO are in the replay list,
3548 but replay of bulk IOs is unimplemented. If the OST filesystem
3549 is corrupted due to disk cache incoherency and then replay is
3550 started it is possible to trip an assertion. Avoid putting
3551 committed RPCs into the replay list at all to avoid this issue.
3556 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3557 Details : Large single O_DIRECT read and write calls can fail to allocate
3558 a sufficiently large buffer to process the request. In case of
3559 allocation failure the allocation is retried with a smaller
3560 buffer and broken into smaller requests.
3562 Severity : enhancement
3564 Description: Add -o localflock option to simulate old noflock behaviour.
3565 Details : This will achieve local-only flock/fcntl locks coherentness.
3570 Description: versioning check is incomplete
3571 Details : Checking the version difference of client vs. server, report
3572 error if the gap is too big.
3577 Description: add support PG_writeback bit
3578 Details : add support for PG_writeback bit for Lustre, for more carefull
3579 work with page cache in 2.6 kernel. This also fix some deadlocks
3580 and remove hack for work O_SYNC with 2.6 kernel.
3582 Severity : enhancement
3584 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3585 Details : The uninit_groups feature works in conjunction with the kernel
3586 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3587 up the pass1 processing of e2fsck. This is a read-only feature
3588 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3589 mount filesystems that have had this feature enabled.
3591 Severity : enhancement
3593 Description: Improve multi-block allocation algorithm to avoid fragmentation
3594 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3595 allocation locality and avoid filesystem fragmentation.
3597 ------------------------------------------------------------------------------
3599 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3601 * Support for kernels:
3602 2.6.9-42.0.3.EL (RHEL 4)
3603 2.6.5-7.276 (SLES 9)
3604 2.4.21-47.0.1.EL (RHEL 3)
3605 2.6.12.6 vanilla (kernel.org)
3606 2.6.16.21-0.8 (SLES10)
3607 * Recommended e2fsprogs version: 1.39.cfs2-0
3609 * The backwards-compatible /proc/sys/portals symlink has been removed
3610 in this release. Before upgrading, please ensure that you change
3611 any configuration scripts or /etc/sysctl.conf files that access
3612 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3613 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3614 in advance of the upgrade on any system running Lustre 1.4.6 or
3615 newer, since /proc/sys/lnet was added in that version.
3616 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3620 Frequency : only when quota is used
3622 Description: avoid scanning export list for quota master
3623 Details : Change the algorithms to avoid scanning export list in order
3624 to improve the efficiency.
3627 Frequency : MDS failover only, very rarely
3629 Description: "went back in time" messages on mds failover
3630 Details : The greatest transno may be lost when the current operation
3631 finishes with an error (transno==0) and the client's last_rcvd
3632 record is over-written. Save the greatest transno in the
3633 mds_last_transno for this case.
3636 Frequency : always for specific kernels and striping counts
3638 Description: client may get "Matching packet too big" without ACL support
3639 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
3640 when trying to access files in certain configurations. The
3641 clients should in fact be denied when mounting because they do
3642 not understand ACLs.
3645 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
3647 Description: many clients connecting with IO in progress causes connect timeouts
3648 Details : Avoid synchronous journal commits to avoid delays caused by many
3649 clients connecting/disconnecting when bulk IO is in progress.
3650 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
3651 OST_IO_PORTAL to avoid delays behind potentially many pending
3655 Frequency : occasionally with multiple writers to a single file
3657 Description: shared writes to file may result in wrong size reported by stat()
3658 Details : Allow growing of kms when extent lock is cancelled
3661 Frequency : always with random mmap IO to multi-striped file
3663 Description: mmap write might be lost if we are writing to a 'hole' in stripe
3664 Details : Only if the hole is at the end of OST object so that kms is too
3665 small. Fix is to increase kms accordingly in ll_nopage.
3668 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
3670 Description: writes to a missing object would leak memory on the OST
3671 Details : If there is an inconsistency between the MDS and OST filesystems,
3672 such that the MDS references an object that doesn't exist, writes
3673 to that object will leak memory due to incorrect cleanup in the
3674 error handling path, eventually running out of memory on the OST.
3679 Description: Creating too long symlink causes lustre errors
3680 Details : Check symlink and name lengths before sending requests to MDS.
3683 Frequency : only if flock is enabled (not on by default)
3685 Description: posix locks not released on fd closure on 2.6.9+
3686 Details : We failed to add posix locks to list of inode locks on 2.6.9+
3687 kernels, this caused such locks not to be released on fd close and
3688 then assertions on fs unmount about still used locks.
3691 Frequency : MDS failover only, very rarely
3693 Description: clients may get ASSERTION(granted_lock != NULL)
3694 Details : When request was taking a long time, and a client was resending
3695 a getattr by name lock request. The were multiple lock requests
3696 with the same client lock handle and
3697 mds_getattr_name->fixup_handle_for_resent_request found one of the
3698 lock handles but later failed with ASSERTION(granted_lock != NULL).
3703 Description: handle->h_buffer_credits > 0, assertion failure
3704 Details : h_buffer_credits is zero after truncate, causing assertion
3705 failure. This patch extends the transaction or creates a new
3709 Frequency : NFS re-export or patchless client
3710 Bugzilla : 11179, 10796
3711 Description: Crash on NFS re-export node (__d_move)
3712 Details : We do not want to hash the dentry if we don't have a lock.
3713 But if this dentry is later used in d_move, we'd hit uninitialised
3714 list head d_hash, so we just do this to init d_hash field but
3715 leave dentry unhashed.
3718 Frequency : NFS re-export or patchless client
3720 Description: NFS exports has problem with symbolic link
3721 Details : lustre client didn't properly install dentry when re-exported
3722 to NFS or running patchless client.
3725 Frequency : NFS re-export or patchless client
3727 Description: Various nfs/patchless fixes.
3728 Details : fixes reuse disconected alias for lookup process - this fixes
3729 warning "find_exported_dentry: npd != pd",
3730 fix permission error with open files at nfs.
3731 fix apply umask when do revalidate.
3734 Frequency : occasional
3736 Description: Crash on NFS re-export node
3737 Details : calling clear_page() on the wrong pointer triggered oops in
3738 generic_mapping_read().
3741 Frequency : rarely, using O_DIRECT IO
3743 Description: unaligned directio crashes client with LASSERT
3744 Details : check for unaligned buffers before trying any requests.
3747 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
3749 Description: stale data returned from RAID cache
3750 Details : If only a small amount of IO is done to the RAID device before
3751 reading it again it is possible to get stale data from the RAID
3752 cache instead of reading it from disk.
3755 Frequency : always for sles10 kernel
3757 Description: sles10 support
3758 Details : ll_follow_link: compile fixes and using of nd_set_link
3759 under newer kernels.
3762 Frequency : depends on arch, kernel and compiler version, always on sles10
3765 Description: recursive or deep enough symlinks cause stack overflow
3766 Details : getting rid of large stack-allocated variable in
3770 Frequency : depends on hardware
3772 Description: lustre write performance loss in the SLES10 kernel
3773 Details : the performance loss is caused by using of write barriers in the
3774 ext3 code. The SLES10 kernel turns barrier support on by
3775 default. The fix is to undo that change for ldiskfs.
3777 ------------------------------------------------------------------------------
3779 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
3781 * Support for kernels:
3782 2.6.9-42.0.3EL (RHEL 4)
3783 2.6.5-7.276 (SLES 9)
3784 2.4.21-47.0.1.EL (RHEL 3)
3785 2.6.12.6 vanilla (kernel.org)
3789 Frequency : quota enabled and large files being deleted
3791 Description: releasing more than 4GB of quota at once hangs OST
3792 Details : If a user deletes more than 4GB of files on a single OST it
3793 will cause the OST to spin in an infinite loop. Release
3794 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3799 Description: statfs data retrieved from /proc may be stale or zero
3800 Details : When reading per-device statfs data from /proc, in the
3801 {kbytes,files}_{total,free,avail} files, it may appear
3802 as zero or be out of date.
3805 Frequency : systems with MD RAID1 external journal devices
3807 Description: lconf's call to blkid is confused by RAID1 journal devices
3808 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3809 of returning all block devices that match the journal UUID.
3812 Frequency : always, for aggregate stripe size over 4GB
3814 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
3815 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
3816 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
3819 Frequency : always if "lfs find" used on a local file/directory
3821 Description: "lfs find" segfaults if used on a local file/directory
3822 Details : The case where a directory component was not specified wasn't
3823 handled correctly. Handle this properly.
3826 Frequency : always on ppc64
3828 Description: the write to an ext3 filesystem mounted with mballoc got stuck
3829 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3830 perform endianness conversion.
3833 Frequency : rarely (truncate to non-zero file size after write under load)
3834 Bugzilla : 10730, 10687
3835 Description: Files padded with zeros to next 4K multiple
3836 Details : With filesystems mounted using the "extents" option (2.6 kernels)
3837 it is possible that files that are truncated to a non-zero size
3838 immediately after being written are filled with zero bytes beyond
3839 the truncated size. No file data is lost.
3841 Severity : enhancement
3843 Description: Allow recovery/failover for liblustre clients.
3844 Details : liblustre clients were unaware of failover configurations until
3847 Severity : enhancement
3849 Description: user file locks should fail when not mounting with flock option
3850 Details : Set up an error-returning stub in ll_file_operations.lock field
3851 to prevent incorrect behaviour when client is mounted without
3852 flock option. Also, set up properly f_op->flock field for
3856 Frequency : always on ia64
3858 Description: "lfs df" loops on printing out MDS statfs information
3859 Details : The obd_ioctl_data was not initialized and in some systems
3860 this caused a failure during the ioctl that did not return
3861 an error. Initialize the struct and return an error on failure.
3864 Frequency : SLES 9 only
3866 Description: Error of copying files with lustre special EAs as root
3867 Details : Client side always return success for setxattr call for lustre
3868 special xattr (currently only "trusted.lov").
3871 Frequency : rarely on clusters with both ia64+i386 clients
3873 Description: ia64+i686 clients doing shared IO on the same file may LBUG
3874 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
3875 clients are doing concurrent writes to the same file it is
3876 possible that the ia64 clients may LASSERT because the OST
3877 extent locks are not PAGE_SIZE aligned. Ensure that grown
3878 locks are always aligned on the request boundary.
3881 Frequency : specific use, occasional
3883 Description: Overwriting in use executable truncates on-disk binary image
3884 Details : If one node attempts to overwrite an executable in use by
3885 another node, we now correctly return ETXTBSY instead of
3886 truncating the file.
3888 Severity : enhancement
3890 Description: Async OSC create to avoid the blocking unnecessarily.
3891 Details : If an OST has no remaining object, system will block on the creating
3892 when need to create a new object on this OST. Now, ways use
3893 pre-created objects when available, instead of blocking on an
3894 empty osc while others are not empty. If we must block, we block
3895 for the shortest possible period of time.
3900 Description: chmod on Lustre root is propagated to other clients
3901 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
3902 invalid by the follow_mount time.
3905 Frequency : liblustre clients only
3907 Description: Race in 'instant cancel' lock handling could lead to such locks
3908 never to be granted in case of SMP MDS
3909 Details : Do not destroy not yet granted but cbpending locks in
3913 Frequency : replay/resend of open
3915 Description: non null lock assetion failure in mds_intent_policy
3916 Details : Trying to replay/resend lockless open requests resulted in
3917 mds_open() returning 0 with no lock. Now it sets a flag if
3918 a lock is going to be returned.
3920 Severity : enhancement
3922 Description: Checksum enhancements
3923 Details : New checksum enhancements allow for resending RPCs that failed
3926 Severity : enhancement
3928 Description: Tunables on number of dirty pages in cacche
3929 Details : Allow to set limit on number of dirty pages cached.
3934 Description: client crash on unmount - lock still has references
3935 Details : In some error handling cases it was possible to leak a lock
3936 reference on a client while accessing a file. This was not
3937 harmful to the client during operation, but would cause the
3938 client to crash when the filesystem is unmounted.
3941 Frequency : specific case, rare
3943 Description: ETXTBSY on mds though file not in use by client
3944 Details : ETXTBSY is no longer incorrectly returned when attempting to
3945 chmod or chown a directory that the user previously tried to
3946 execute or a currently-executing binary.
3949 Frequency : extremely rare except on liblustre-based clients
3951 Description: Lustre space not freed when files are deleted
3952 Details : Clean up open-unlinked files after client eviction. Previously
3953 the unlink was skipped and the files remained as orphans.
3958 Description: OST failure "would be an LBUG" in waiting_locks_callback()
3959 Details : In some cases it was possible to send a blocking callback to
3960 a client doing a glimpse, even though that client didn't get
3961 a lock granted. When the glimpse lock is cancelled on the OST
3962 the freed lock is left on the waiting list and corrupted the list.
3965 Frequency : all core dumps
3967 Description: Broke core dumps to lustre
3968 Details : Negative dentry may be unhashed if parent does not have UPDATE
3969 lock, but some callers, e.g. do_coredump, expect dentry to be
3970 hashed after successful create, hash it in ll_create_it.
3972 ------------------------------------------------------------------------------
3974 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
3976 * Support for kernels:
3977 2.6.9-42.0.2.EL (RHEL 4)
3978 2.6.5-7.276 (SLES 9)
3979 2.4.21-40.EL (RHEL 3)
3980 2.6.12.6 vanilla (kernel.org)
3984 Frequency : always on RHEL 3
3986 Description: Number of open files grows over time
3987 Details : The number of open files grows over time, whether or not
3988 Lustre is started. This was due to a filp leak introduced
3989 by one of our kernel patches.
3991 ------------------------------------------------------------------------------
3993 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
3995 * Support for kernels:
3996 2.6.9-42.EL (RHEL 4)
3997 2.6.5-7.276 (SLES 9)
3998 2.4.21-40.EL (RHEL 3)
3999 2.6.12.6 vanilla (kernel.org)
4004 Bugzilla : 5719, 9635, 9792, 9684
4005 Description: OST (or MDS) trips assertions in (re)connection under heavy load
4006 Details : If a server is under heavy load and cannot reply to new
4007 connection requests before the client resends the (re)connect,
4008 the connection handling code can behave badly if two service
4009 threads are concurrently handing separate (re)connections from
4010 the same client. Add better locking to the connection handling
4011 code, and ensure that only a single connection will be processed
4012 for a given client UUID, even if the lock is dropped.
4014 Severity : enhancement
4016 Description: add TCP zero-copy support to kernel
4017 Details : Add support to the kernel TCP stack to allow zero-copy bulk
4018 sends if the hardware supports scatter-gather and checksumming.
4019 This allows socklnd to do client-write and server-read more
4020 efficiently and reduce CPU utilization from skbuf copying.
4023 Frequency : only if NFS exporting from client
4025 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
4026 Details : If setting attributes on a file created under NFS that had
4027 never been opened it would be possible to oops the client
4028 if the file had no objects.
4031 Frequency : always for liblustre
4033 Description: liblustre client does MDS+OSTs setattr RPC for each write
4034 Details : When doing a write from a liblustre client, the client
4035 incorrectly issued an RPC to the MDS and each OST the file was
4036 striped over in order to update the timestamps. When writing
4037 with small chunks and many clients this could overwhelm the MDS
4038 with RPCs. In all cases it would slow down the write because
4039 these RPCs are unnecessary.
4041 Severity : enhancement
4043 Description: allow number of MDS service threads to be changed at module load
4044 Details : It is now possible to change the number of MDS service threads
4045 running. Adding "options mds mds_num_threads={N}" to the MDS's
4046 /etc/modprobe.conf will set the number of threads for the next
4047 time Lustre is restarted (assuming the "mds" module is also
4048 reloaded at that time). The default number of threads will
4049 stay the same, 32 for most systems.
4054 Description: OST crash if filesystem is unformatted or corrupt
4055 Details : If an OST is started on a device that has never been formatted
4056 or if the filesystem is corrupt and cannot even mount then the
4057 error handling cleanup routines would dereference a NULL pointer.
4062 Description: NULL pointer deref in llap_from_page.
4063 Details : get_cache_page_nowait can return a page with NULL (or otherwise
4064 incorrect) mapping if the page was truncated/reclaimed while it was
4065 searched for. Check for this condition and skip such pages when
4066 doing readahead. Introduce extra check to llap_from_page() to
4067 verify page->mapping->host is non-NULL (so page is not anonymous).
4070 Frequency : Sometimes when using sys_sendfile
4072 Description: "page not covered by a lock" warnings from ll_readpage
4073 Details : sendfile called ll_readpage without right page locks present.
4074 Now we introduced ll_file_sendfile that does necessary locking
4075 around call to generic_file_sendfile() much like we do in
4079 Frequency : with certain MDS communication failures at client mount time
4081 Description: NULL pointer deref after failed client mount
4082 Details : a client connection request may delayed by the network layer
4083 and not be sent until after the PTLRPC layer has timed out the
4084 request. If the client fails the mount immediately it will try
4085 to clean up before the network times out the request. Add a
4086 reference from the request import to the obd device and delay
4087 the cleanup until the network drops the request.
4090 Frequency : occasionally during client (re)connect
4092 Description: assertion failure during client (re)connect
4093 Details : processing a client connection request may be delayed by the
4094 client or server longer than the client connect timeout. This
4095 causes the client to resend the connection request. If the
4096 original connection request is replied in this interval, the
4097 client may trip an assertion failure in ptlrpc_connect_interpret()
4098 which thought it would be the only running connect process.
4101 Frequency : only with obd_echo servers and clients that are rebooted
4103 Description: kernel BUG accessing uninitialized data structure
4104 Details : When running an obd_echo server it did not start the ping_evictor
4105 thread, and when a client was evicted an uninitialized data
4106 structure was accessed. Start the ping_evictor in the RPC
4107 service startup instead of the OBD startup.
4109 Severity : enhancement
4110 Bugzilla : 10193 (patchless)
4111 Description: Remove dependency on various unexported kernel interfaces.
4112 Details : No longer need reparent_to_init, exit_mm, exit_files,
4113 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
4116 Frequency : rare (only users of deprecated and unsupported LDAP config)
4118 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
4119 Details : LDAP apparently contains 'lustreName' attributes instead of
4120 'name'. A simple remapping of the name is sufficient.
4123 Frequency : rare (only with non-default dump_on_timeout debug enabled)
4125 Description: waiting_locks_callback trips kernel BUG if client is evicted
4126 Details : Running with the dump_on_timeout debug flag turned on makes
4127 it possible that the waiting_locks_callback() can try to dump
4128 the Lustre kernel debug logs from an interrupt handler. Defer
4129 this log dumping to the expired_lock_main() thread.
4131 Severity : enhancement
4133 Description: Support NFS exporting on 2.6 kernels.
4134 Details : Implement non-rawops metadata methods for NFS server to use without
4135 changing NFS server code.
4138 Frequency : very rare (synthetic metadata workload only)
4140 Description: two racing renames might cause an MDS thread to deadlock
4141 Details : Running the "racer" program may cause one MDS thread to rename
4142 a file from being the source of a rename to being the target of
4143 a rename at exactly the same time that another thread is doing
4144 so, and the second thread has already enqueued these locks after
4145 doing a lookup of the target and is trying to relock them in
4146 order. Ensure that we don't try to re-lock the same resource.
4149 Frequency : only very large systems with liblustre clients
4151 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
4152 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
4153 all OSTs in parallel. This allows the eviction of stale liblustre
4154 clients to proceed much faster than if they were done in series,
4155 and also offers similar improvements for other set_info RPCs.
4160 Description: excessive CPU usage during initial read phase on client
4161 Details : During the initial read phase on a client, it would agressively
4162 retry readahead on the file, consuming too much CPU and impacting
4163 performance (since 1.4.5.8). Improve the readahead algorithm
4164 to avoid this, and also improve some other common cases (read
4165 of small files in particular, where "small" is files smaller than
4166 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
4171 Description: MDS crash when receiving packet with unknown intent.
4172 Details : Do not LBUG in unknown intent case, just return -EFAULT
4174 Severity : enhancement
4175 Bugzilla : 9293, 9385
4176 Description: MDS RPCs are serialised on client. This is unnecessary for some.
4177 Details : Do not serialize getattr (non-intent version) and statfs.
4180 Frequency : occasional, when OST network is overloaded/intermittent
4182 Description: client evicted by OST after bulk IO timeout
4183 Details : If a client sends a bulk IO request (read or write) the OST
4184 may evict the client if it is unresposive to its data GET/PUT
4185 request. This is incorrect if the network is overloaded (takes
4186 too long to transfer the RPC data) or dropped the OST GET/PUT
4187 request. There is no need to evict the client at all, since
4188 the pinger and/or lock callbacks will handle this, and the
4189 client can restart the bulk request.
4192 Frequency : Always when mmapping file with no objects
4194 Description: client crashes when mmapping file with no objects
4195 Details : Check that we actually have objects in a file before doing any
4196 operations on objects in ll_vm_open, ll_vm_close and
4202 Description: Request leak when working with deleted CWD
4203 Details : Introduce advanced request refcount tracking for requests
4204 referenced from lustre intent.
4206 Severity : Enhancement
4208 Description: Cache open file handles on client.
4209 Details : MDS now will return special lock along with openhandle, if
4210 requested and client is allowed to hold openhandle, even if unused,
4211 until such a lock is revoked. Helps NFS a lot, since NFS is opening
4212 closing files for every read/write openration.
4214 Severity : Enhancement
4216 Description: Cache open negative dentries on client when possible.
4217 Details : Guard negative dentries with UPDATE lock on parent dir, drop
4218 negative dentries on lock revocation.
4223 Description: Remounting a client read-only wasn't possible with a zconf mount
4224 Details : It wasn't possible to remount a client read-only with llmount.
4226 Severity : enhancement
4227 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
4228 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
4229 MPICH is included to add Lustre-specific ADIO interfaces.
4230 This is based closely on the UFS ADIO layer and only differs
4231 in file creation, in order to allow the OST striping to be set.
4232 This is user-contributed code and not supported by CFS.
4237 Description: extended inode attributes (immutable, append-only) work improperly
4238 when 2.4 and 2.6 kernels are used on client/server or vice versa
4239 Details : Introduce kernel-independent values for these flags.
4241 Severity : enhancement
4244 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
4245 Details : Many of the /proc/ tunables can only be tuned at a megabyte
4246 granularity. Now, Fractional MB granularity is be supported,
4247 this is very useful for low memory system.
4249 Severity : enhancement
4251 Description: Getattr by fid
4252 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
4253 locks, avoids extra getattr rpc requests to MDS, allows '/' to
4254 have locks and avoids getattr rpc requests for it on every stat.
4257 Frequency : Always, for filesystems larger than 2TB
4259 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4260 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
4261 filesystems larger than 2TB to be mounted with mballoc enabled.
4264 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
4266 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4267 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
4268 filesystems larger than 2TB is mounted then the kernel will
4269 silently corrupt the start of the filesystem. CONFIG_LBD is
4270 enabled for all CFS-supported kernels, but the possibility of
4271 this happening with a modified kernel config exists.
4273 Severity : enhancement
4275 Description: add client O_DIRECT support for 2.6 kernels
4276 Details : It is now possible to do O_DIRECT reads and writes to files
4277 in the Lustre client mountpoint on 2.6 kernel clients.
4279 Severity : enhancement
4281 Description: parallel glimpse, setattr, statfs, punch, destroy requests
4282 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
4283 parallel, not waiting for response from every OST before sending
4284 a rpc to the next OST.
4289 Description: setattr vs write race when updating file timestamps
4290 Details : Client processes that update a file timestamp into the past
4291 right after writing to the file (e.g. tar) it is possible that
4292 the updated file modification time can be reset to the current
4293 time due to a race between processing the setattr and write RPC.
4295 Severity : enhancement
4297 Description: Bring 'lfs find' closer in line with regular Linux find.
4298 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
4299 -print0 options and obtains all the needed info through the lustre
4302 Severity : enhancement
4304 Description: support up to 1024 configured devices on one node
4305 Details : change obd_dev array from statically allocated to dynamically
4306 allocated structs as they are first used to reduce memory usage
4311 Description: Flush dirty partially truncated pages during truncate
4312 Details : Immediatelly flush partially truncated pages in filter_setattr,
4313 this way we completely avoid having any pages in page cache on OST
4314 and can retire ugly workarounds during writes to flush such pages.
4319 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
4320 Details : protect inode from truncation within vfs_unlink() context
4321 just take a reference before calling vfs_unlink() and release it
4322 when parent's i_sem is free.
4327 Description: last_id value checked outside lock on OST caused LASSERT failure
4328 Details : If there were multiple MDS->OST object precreate requests in
4329 flight, it was possible that the OST's last object id was checked
4330 outside a lock and incorrectly tripped an assertion. Move checks
4331 inside locks, and discard old precreate requests.
4334 Frequency : always, if extents are used on OSTs
4336 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
4337 Details : OSTs using the extents format would not zero the high 16 bits of
4338 the index physical block number. This is not a problem for any
4339 OST filesystems smaller than 16TB, and no kernels support ext3
4340 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
4341 new/modified files) and can be fixed for existing filesystems
4342 with e2fsprogs-1.39-cfs1.
4347 Description: import connection selection may be incorrect if timer wraps
4348 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
4349 import connections to be ignored if the 32-bit jiffies counter
4350 wraps. Use a 64-bit jiffies counter.
4353 Frequency : very large clusters immediately after boot
4355 Description: LNET request buffers exhausted under heavy short-term load
4356 Details : If a large number of client requests are generated on a service
4357 that has previously never seen so many requests it is possible
4358 that the request buffer growth cannot keep up with the spike in
4359 demand. Instead of dropping incoming requests, they are held in
4360 the LND until the RPC service can accept more requests.
4363 Frequency : Sometimes during replay
4365 Description: Assertion failure in ll_local_open after replay.
4366 Details : If replay happened on an open request reply before we were able
4367 to set replay handler, reply will become not swabbed tripping the
4368 assertion in ll_local_open. Now we set the handler right after
4369 recognising of open request
4372 Frequency : very rare
4374 Description: kernel reports "badness in vsnprintf"
4375 Details : Reading from the "recovery_status" /proc file in small chunks
4376 may cause a negative length in lprocfs_obd_rd_recovery_status()
4377 call to vsnprintf() (which is otherwise harmless). Exit early
4378 if there is no more space in the output buffer.
4380 Severity : enhancement
4382 Description: clear OBD RPC statistics by writing to them
4383 Details : It is now possible to clear the OBD RPC statistics by writing
4384 to the "stats" file.
4389 Description: Client mtime is not the same on different clients after utimes
4390 Details : In some cases, the client was using the utimes() syscall on
4391 a file cached on another node. The clients now validate the
4392 ctime from the MDS + OSTs to determine which one is right.
4397 Description: Inability to activate failout mode
4398 Details : lconf script incorrectly assumed that in pythong string's numeric
4399 value is used in comparisons.
4402 Frequency : always with multiple stripes per file
4404 Description: Inefficient object allocation for mutli-stripe files
4405 Details : When selecting which OSTs to stripe files over, for files with
4406 a stripe count that divides evenly into the number of OSTs,
4407 the MDS is always picking the same starting OST for each file.
4408 Return the OST selection heuristic to the original design.
4413 Description: mount failures may take full timeout to return an error
4414 Details : Under some heavy load conditions it is possible that a
4415 failed mount can wait for the full obd_timeout interval,
4416 possibly several minutes, before reporting an error.
4417 Instead return an error as soon as the status is known.
4419 Frequency : quota enabled and large files being deleted
4421 Description: releasing more than 4GB of quota at once hangs OST
4422 Details : If a user deletes more than 4GB of files on a single OST it
4423 will cause the OST to spin in an infinite loop. Release
4424 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4429 Description: statfs data retrieved from /proc may be stale or zero
4430 Details : When reading per-device statfs data from /proc, in the
4431 {kbytes,files}_{total,free,avail} files, it may appear
4432 as zero or be out of date.
4435 Frequency : systems with MD RAID1 external journal devices
4437 Description: lconf's call to blkid is confused by RAID1 journal devices
4438 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4439 of returning all block devices that match the journal UUID.
4442 Frequency : always, for aggregate stripe size over 4GB
4444 Description: assertion fails when trying to use 4GB stripe size
4445 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
4446 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
4449 Frequency : always on ppc64
4451 Description: the first write on an ext3 filesystem with mballoc got stuck
4452 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4453 perform endianness conversion.
4455 ------------------------------------------------------------------------------
4457 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4459 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
4460 INTEROPERATE with older versions automatically. Please read the
4461 user documentation before upgrading any part of a live system.
4462 * WARNING: Lustre networking configuration changes are required with
4463 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
4466 * Support for kernels:
4467 2.6.9-22.0.2.EL (RHEL 4)
4468 2.6.5-7.244 (SLES 9)
4469 2.6.12.6 vanilla (kernel.org)
4472 Severity : enhancement
4473 Bugzilla : 7981/8208
4474 Description: Introduced Lustre Networking (LNET)
4475 Details : LNET is new networking infrastructure for Lustre, it includes
4476 a reorganized network configuration mode (see the user
4477 documentation for full details) as well as support for routing
4478 between different network fabrics. Lustre Networking Devices
4479 (LNDs) for the supported network fabrics have also been
4480 created for this new infrastructure.
4482 Severity : enhancement
4483 Description: Introduced Access control lists
4484 Details : clients can set ACLs on files and directories in order to have
4485 more fine-grained permissions than the standard Unix UGO+RWX.
4486 The MDS must be started with the "-o acl" mount option.
4488 Severity : enhancement
4489 Description: Introduced filesystem quotas
4490 Details : Administrators may now establish per-user quotas on the
4493 Severity : enhancement
4495 Description: Configuration change for the XT3
4496 The PTLLND is now used to run Lustre over Portals on the XT3
4497 The configure option(s) --with-cray-portals are no longer used.
4498 Rather --with-portals=<path-to-portals-includes> is used to
4499 enable building on the XT3. In addition to enable XT3 specific
4500 features the option --enable-cray-xt3 must be used.
4505 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
4506 Details : A race between allocating a new llcd and re-getting the llcd_lock
4507 allowed another thread to grab newly-allocated llcd.
4509 Severity : enhancement
4511 Description: 2.6 OST async journal commit and locking fix to improve performance
4512 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
4513 2.6 kernels are now async as they already were in 2.4 kernels so
4514 that they can commit concurrently with the network bulk transfer.
4515 For block-allocated files the filter allocation semaphore is held
4516 to avoid filesystem fragmentation during allocation. BKL lock
4517 removed for 2.6 xattr operations where it is no longer needed.
4522 Description: lconf incorrectly determined whether two IP networks could talk
4523 Details : In some more complicated routing and multiple-network
4524 configurations, lconf will avoid trying to make a network
4525 connection to a disjoint part of the IP space. It was doing the
4526 math incorrectly for one set of cases.
4531 Description: Fix for potential infinite loop processing records in an llog.
4532 Details : If an llog record is corrupted/zeroed, it is possible to loop
4533 forever in llog_process(). Validate the llog record length
4534 and skip the remainder of the block on error.
4537 Frequency : occasional (liblustre only)
4539 Description: liblustre could not open files whose last component is a symlink
4540 Details : sysio_path_walk() would incorrectly pass the open intent to
4541 intermediate path components.
4544 Frequency : rare (liblustre only with non-standard tuning)
4545 Bugzilla : 7201 (7350)
4546 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4547 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4548 after it has been released, only do this for referenced locks
4550 Severity : enhancement
4552 Description: specifying an (invalid) directory default stripe_size of -1
4553 would reset the directory default striping
4554 Details : stripe_size -1 was used internally to signal directory stripe
4555 removal, now use "all default" to signal dir stripe removal
4556 as a directory striping of "all default" is not useful
4559 Frequency : common for large clusters running liblustre clients
4561 Description: doing an ls when liblustre clients are running is slow
4562 Details : sending a glimpse AST to a liblustre client waits for every AST
4563 to time out, as liblustre clients will not respond. Since they
4564 cannot cache data we refresh the OST lock LVB from disk instead.
4566 Severity : enhancement
4568 Description: doing an ls at the same time as file IO can be slow
4569 Details : enqueue and other "small" requests can be blocked behind many
4570 large IO requests. Create a new OST IO portal for non-IO
4571 requests so they can be processed faster.
4574 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4576 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4577 created file with mode 000
4578 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4579 allow an ftruncate() as a non-root user to a file with mode 000.
4580 Lustre now respects this flag to disable mode checks when
4581 truncating a file owned by the user
4584 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4587 Description: Revoking locks from clients that went dead or catatonic might take
4589 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4590 cancellation of such locks instant on servers without waiting for
4591 any reply from clients. Clients drops these locks when cancel
4592 notification from server is received without replying.
4595 Frequency : liblustre-only, when liblustre client dies or becomes busy
4597 Description: Doing ls on Linux clients can take a long time with active
4599 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4600 granting such locks to it in the first place if possible. Locks
4601 are taken by proxy on the OST during the read or write and
4602 dropped immediately afterward. Add connect flags handling, do
4603 not grant locks to liblustre clients for glimpse ASTs.
4605 Severity : enhancement
4607 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4608 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4609 sequential reads and behaves badly with applications doing random
4610 reads. Improve it by reading ahead at least read region, and
4611 avoiding excessive large RPC for small reads.
4613 Severity : enhancement
4615 Description: Creating more than 1000 files for a single job may cause a load
4616 imbalance on the OSTs if there are also a large number of OSTs.
4617 Details : qos_prep_create() uses an OST index reseed value that is an
4618 even multiple of the number of available OSTs so that if the
4619 reseed happens in the middle of the object allocation it will
4620 still utilize the OSTs as uniformly as possible.
4625 Description: OST or MDS may oops in ping_evictor_main()
4626 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
4627 but doesn't restart at beginning of obd_exports_timed list
4630 Severity : enhancement
4632 Description: improve by-nid export eviction on the MDS and OST
4633 Details : allow multiple exports with the same NID to be evicted at one
4634 time without re-searching the exports list.
4637 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
4639 Description: MDS may oops in groups_free()
4640 Details : in rare race conditions a newly allocated group_info struct is
4641 freed again, and this can be NULL. The 2.4 compatibility code
4642 for groups_free() checked for a NULL pointer, but 2.6 did not.
4645 Frequency : common for liblustre clients doing little filesystem IO
4646 Bugzilla : 9352, 7313
4647 Description: server may evict liblustre clients accessing contended locks
4648 Details : if a client is granted a lock or receives a completion AST
4649 with a blocking AST already set it would not reply to the AST
4650 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
4653 Frequency : lfs setstripe, only systems with more than 160 OSTs
4655 Description: unable to set striping with a starting offset beyond OST 160
4656 Details : llapi_create_file() incorrectly limited the starting stripe
4657 index to the maximum single-file stripe count.
4660 Frequency : LDAP users only
4662 Description: lconf did not handle in-kernel recovery with LDAP properly
4663 Details : lconf/LustreDB get_refs() is searching the wrong namespace
4665 Severity : enhancement
4667 Description: bind OST threads to NUMA nodes to improve performance
4668 Details : all OST threads are uniformly bound to CPUs on a single NUMA
4669 node and do their allocations there to localize memory access
4671 Severity : enhancement
4673 Description: llmount can determine client NID directly from Myrinet (GM)
4674 Details : the client NID code from gmnalnid was moved directly into
4675 llmount, removing the need to use this or specifying the
4676 client NID explicitly when mounting GM clients with zeroconf
4679 Frequency : if client is started with down MDS
4681 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
4682 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
4683 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
4684 ptlrpc_import_delayed_req() because we didn't check if the
4685 request was interrupted, and we also didn't break out of the
4686 event loop if there was no timeout
4691 Description: data loss during non-page-aligned writes to a single file from
4692 both multiple nodes and multiple threads on one node at same time
4693 Details : updates to KMS and lsm weren't protected by common lock. Resulting
4694 inconsistency led to false short-reads, that were cached and later
4695 used by ->prepare_write() to fill in partially written page,
4696 leading to data loss.
4699 Frequency : always, if lconf --abort_recovery used
4701 Description: lconf --abort_recovery fails with 'Operation not supported'
4702 Details : lconf was attempting to abort recovery on the MDT device and not
4705 Severity : enhancement
4707 Description: remove cleanup logs
4708 Details : replace lconf-generated cleanup logs with lustre internal
4709 cleanup routines. Eliminates the need for client-cleanup and
4712 Severity : enhancement
4714 Description: add support for EAs (user and system) on lustre filesystems
4715 Details : it is now possible to store extended attributes in the Lustre
4716 client filesystem, and with the user_xattr mount option it
4717 is possible to allow users to store EAs on their files also
4719 Severity : enhancement
4721 Description: Add possibility (config option) to show minimal available OST free
4723 Details : When compiled with --enable-mindf configure option, statfs(2)
4724 (and so, df) will return least minimal free space available from
4725 all OSTs as amount of free space on FS, instead of summary of
4726 free spaces of all OSTs.
4728 Severity : enhancement
4730 Description: do not expand extent locks acquired on OST-side
4731 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
4732 by server: they are not cached anyway.
4735 Frequency : when mmap is used/binaries executed from Lustre
4737 Description: Unmmap pages before throwing them away from read cache.
4738 Details : llap_shrink cache now attempts to unmap pages before discarding
4739 them (if unmapping failed - do not discard). SLES9 kernel has
4740 extra checks that trigger if this unmapping is not done first.
4745 Description: lconf didn't resolve symlinks before checking to see whether a
4746 given mountpoint was already in use
4749 Frequency : when migrating failover services
4750 Bugzilla : 6395, 9514
4751 Description: When migrating a subset of services from a node (e.g. failback
4752 from a failover service node) the remaining services would
4753 time out and evict clients.
4754 Details : lconf --force (implied by --failover) sets the global obd_timeout
4755 to 5 seconds in order to quickly disconnect, but this caused
4756 other RPCs to time out too quickly. Do not change the global
4757 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
4759 Severity : enhancement
4760 Frequency : if MDS is started with down OST
4761 Bugzilla : 9439,5706
4762 Description: Allow startup/shutdown of an MDS without depending on the
4763 availability of the OSTs.
4764 Details : Asynchronously call mds_lov_synchronize during MDS startup.
4765 Add appropriate locking and lov-osc refcounts for safe
4766 cleaning. Add osc abort_inflight calls in case the
4767 synchronize never started.
4770 Frequency : occasional (Cray XT3 only)
4772 Description: root not authorized to access files in CRAY_PORTALS environment
4773 Details : The client process capabilities were not honoured on the MDS in
4774 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
4775 been accessed by an authorized user then root was able to access
4776 the file on the local client also. The root user capabilities
4777 are now allowed on the MDS, as this environment has secure UID.
4780 Frequency : occasional
4782 Description: ldiskfs "too long searching" message happens too often
4783 Details : A debugging message (otherwise harmless) prints too often on
4784 the OST console. This has been reduced to only happen when
4785 there are fragmentation problems on the filesystem.
4790 Description: Division by zero in statfs when all OSCs are inactive
4791 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
4792 lov_statfs divides by value returned by lov_get_stripecnt().
4796 Bugzilla : 9489, 3273
4797 Description: First write from each client to each OST was only 4kB in size,
4798 to initialize client writeback cache, which caused sub-optimal
4799 RPCs and poor layout on disk for the first written file.
4800 Details : Clients now request an initial cache grant at (re)connect time
4801 and so that they can start streaming writes to the cache right
4802 away and always do full-sized RPCs if there is enough data.
4803 If the OST is rebooted the client also re-establishes its grant
4804 so that client cached writes will be honoured under the grant.
4809 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
4810 Details : Now I/O RPCs go to different portal number and (presumably) fast
4811 lock requests (and glimses) and other RPCs get their own service
4812 threads pool that should be able to service those RPCs
4815 Severity : enhancement
4817 Description: Ability to exchange lustre version between client and servers and
4818 issue warnings at client side if client is too old. Also for
4819 liblustre clients there is ability to refuse connection of too old
4821 Details : New 'version' field is added to connect data structure that is
4822 filled with version info. That info is later checked by server and
4826 Frequency : rare, liblustre only.
4827 Bugzilla : 9296, 9581
4828 Description: Two simultaneous writes from liblustre at offset within same page
4829 might proceed at the same time overwriting eachother with stale
4831 Details : I/O lock withing llu_file_prwv was released too early, before data
4832 actually was hitting the wire. Extended lock-holding time until
4833 server acknowledges receiving data.
4836 Frequency : extremely rare. Never observed in practice.
4838 Description: avoid generating lustre_handle cookie of 0.
4839 Details : class_handle_hash() generates handle cookies by incrementing
4840 global counter, and can hit 0 occasionaly (this is unlikely, but
4841 not impossible, because initial value of cookie counter is
4842 selected randonly). Value of 0 is used as a sentinel meaning
4843 "unassigned handle" --- avoid it. Also coalesce two critical
4844 sections in this function into one.
4846 Severity : enhancement
4848 Description: allow liblustre clients to delegate truncate locking to OST
4849 Details : To avoid overhead of locking, liblustre client instructs OST to
4850 take extent lock in ost_punch() on client's behalf. New connection
4851 flag is added to handle backward compatibility.
4853 Severity : enhancement
4854 Bugzilla : 4928, 7341, 9758
4855 Description: allow number of OST service threads to be specified
4856 Details : a module parameter allows the number of OST service threads
4857 to be specified via "options ost ost_num_threads={N}" in the
4858 OSS's /etc/modules.conf or /etc/modprobe.conf.
4862 Bugzilla : 6146, 9635, 9895
4863 Description: servers crash with bad pointer in target_handle_connect()
4864 Details : In rare cases when a client is reconnecting it was possible that
4865 the connection request was the last reference for that export.
4866 We would temporarily drop the export reference and get a new
4867 one, but this may have been the last reference and the export
4868 was just destroyed. Get new reference before dropping old one.
4870 Severity : enhancement
4871 Frequency : if client is started with failover MDS
4873 Description: Allow multiple MDS hostnames in the mount command
4874 Details : Try to read the configuration from all specified MDS
4875 hostnames during a client mount in case the "primary"
4878 Severity : enhancement
4880 Description: Stop sending data to evicted clients as soon as possible.
4881 Details : Check if the client we are about to send or are sending data to
4882 was evicted already. (Check is done every second of waiting,
4883 for which l_wait_event interface was extended to allow checking
4884 of exit condition at specified intervals).
4887 Frequency : rare, normally only when NFS exporting is done from client
4889 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
4891 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
4894 Frequency : always, if srand() is called before liblustre initialization
4896 Description: Liblustre uses system PRNG disturbing its usage by user application
4897 Details : Introduce internal to lustre fast and high-quality PRNG for
4898 lustre usage and make liblustre and some other places in generic
4899 lustre code to use it.
4901 Severity : enhancement
4902 Bugzilla : 9477, 9557, 9870
4903 Description: Verify that the MDS configuration logs are updated when xml is
4904 Details : Check if the .xml configuration logs are newer than the config
4905 logs stored on the MDS and report an error if this is the case.
4906 Request --write-conf, or allow starting with --old_conf.
4908 Severity : enhancement
4910 Description: Handle symlinks in the path when checking if Lustre is mounted.
4911 Details : Resolve intermediate symlinks when checking if a client has
4912 mounted a filesystem to avoid duplicate client mounts.
4917 Description: lconf can hit an error exception but still return success.
4918 Details : The lconf command catches the Command error exception at the top
4919 level script context and will exit with the associated exit
4920 status, but doesn't ensure that this exit status is non-zero.
4925 Description: failure of ptlrpc thread startup can cause oops
4926 Details : Starting a ptlrpc service thread can fail if there are a large
4927 number of threads or the server memory is very fragmented.
4928 Handle this without oopsing.
4931 Frequency : always, only if liblustre and non-default acceptor port was used
4933 Description: liblustre cannot connect to servers with non-default acceptor port
4934 Details : tcpnal_set_default_params() was not called and was therefore
4935 ignoring the environment varaible TCPNAL_PORT, as well as other
4936 TCPNAL_ environment variables
4941 Description: two objects could be created on the same OST for a single file
4942 Details : If an OST is down, in some cases it was possible to create two
4943 objects on a single OST for a single file. No problems other
4944 than potential performance impact and spurious error messages.
4948 Bugzilla : 5681, 9562
4949 Description: Client may oops in ll_unhash_aliases
4950 Details : Client dcache may become inconsistent in race condition.
4951 In some cases "getcwd" can fail if the current directory is
4957 Description: Inode refcounting problems in NFS export code
4958 Details : link_raw functions used to call d_instantiate without obtaining
4959 extra inode reference first.
4963 Bugzilla : 9942, 9903
4964 Description: Referencing freed requests leading to crash, memleaks with NFS.
4965 Details : We used to require that call to ll_revalidate_it was always
4966 followed by ll_lookup_it. Also with revalidate_special() it is
4967 possible to call ll_revalidate_it() twice for the same dentry
4968 even if first occurence returned success. This fix changes semantic
4969 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
4970 reference on a request referred from the intent.
4971 ll_intent_release() then releases such a request.
4974 Frequency : rare, normally benchmark loads only
4976 Description: unlinked inodes were kept in memory on the client
4977 Details : If a client is repeatedly creating and unlinking files it
4978 can accumulate a lot of stale inodes in the inode slab cache.
4979 If there is no other client load running this can cause the
4980 client node to run out of memory. Instead flush old inodes
4981 from client cache that have the same inode number as a new inode.
4984 Frequency : SLES9 2.6.5 kernel and long filenames only
4985 Bugzilla : 9969, 10379
4986 Description: utime reports stale NFS file handle
4987 Details : SLES9 uses out-of-dentry names in some cases, which confused
4988 the lustre dentry revalidation. Change it to always use the
4992 Frequency : rare, unless heavy write-truncate concurrency is continuous
4993 Bugzilla : 4180, 6984, 7171, 9963, 9331
4994 Description: OST becomes very slow and/or deadlocked during object unlink
4995 Details : filter_destroy() was holding onto the parent directory lock
4996 while truncating+unlinking objects. For very large objects this
4997 may block other threads for a long time and slow overall OST
4998 responsiveness. It may also be possible to get a lock ordering
4999 deadlock in this case, or run out of journal credits because of
5000 the combined truncate+unlink. Solution is to do object truncate
5001 first in one transaction without parent lock, and then do the
5002 final unlink in a new transaction with the parent lock. This
5003 reduces the lock hold time dramatically.
5006 Frequency : rare, 2.4 kernels only
5008 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
5009 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
5010 before unmounting their filesystem in order to avoid deadlock.
5011 The kernel_locked() function in 2.4 kernels only checks whether
5012 the kernel lock is held, not whether it is this process that is
5013 holding it as 2.6 kernels do.
5018 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
5019 Details : The client ptlrpc code may be trying to reconnect to a down
5020 server before a previous connection attempt has timed out.
5021 Increase the reconnect interval to be longer than the connection
5022 timeout interval to avoid sending duplicate connections to
5026 Frequency : echo_client brw_test command
5028 Description: fix echo_client to work with OST preallocated code
5029 Details : OST preallocation code (5137) didn't take echo_client IO path
5030 into account: echo_client calls filter methods outside of any
5031 OST thread and, hence, there is no per-thread preallocated
5032 pages and buffers to use. Solution: hijack pga pages for IO. As
5033 a byproduct, this avoids unnecessary data copying.
5037 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
5038 Description: Client can oops in mdc_commit_close() after open replay
5039 Details : It was possible for the MDS to return an open request with no
5040 transaction number in mds_finish_transno() if the client was
5041 evicted, but without actually returning an error. Clients
5042 would later try to replay that open and may trip an assertion
5043 Simplify the client close codepath, and always return an error
5044 from the MDS in case the open is not successful.
5047 Frequency : rare, 2.6 OSTs only
5049 Description: OST may deadlock under high load on fragmented files
5050 Details : If there was a heavy load and highly-fragmented OST filesystems
5051 it was possible to have all the OST threads deadlock waiting on
5052 allocation of biovecs, because the biovecs were not released
5053 until the entire RPC IO was completed. Instead, release biovecs
5054 as soon as they are complete to ensure forward IO progress.
5056 Severity : enhancement
5058 Description: Support for specifying external journal device at mount
5059 Details : If an OST or MDS device is formatted with an external journal
5060 device, this device major/minor is stored in the ext3 superblock
5061 and may not be valid for failover. Allow detecting and
5062 specifying the external journal at mount time.
5067 Description: Mounting an MDS with pending unlinked files may cause oops
5068 Details : target_finish_recovery() calls mds_postrecov() which returned
5069 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
5070 considers this an error and immediately begins cleaning up the
5071 lov, just after starting the mds_lov process
5073 Severity : enhancement
5075 Description: Implement 'lfs df' to report actual free space on per-OST basis
5076 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
5077 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
5078 usage of objects; '-h' to report in human readable format.
5080 ------------------------------------------------------------------------------
5082 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
5089 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
5090 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
5091 at startup during orphan unlinking. Instead, skip buddy update
5092 before setup, as it will be regenerated after recovery is complete.
5097 Description: If an OST is inactive, its locks might reference stale inodes.
5098 Details : lov_change_cbdata() must iterate over all namespaces, even if
5099 they are inactive to clear inode references from the lock.
5101 Severity : enhancement
5102 Frequency : occasional, if non-standard max_dirty_mb used
5104 Description: Client will block write RPCs if not enough grant
5105 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
5106 then the client will block writes while waiting for another RPC
5107 to complete instead of consuming its dirty limit. With change
5108 we get improved performance when max_dirty_mb is small.
5110 Severity : enhancement
5111 Bugzilla : 3389, 6253
5112 Description: Add support for supplementary groups on the MDS.
5113 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
5114 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
5115 lookups for user supplementary groups into a cache.
5119 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
5120 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
5121 for liblustre O_EXCL is enforced by MDS.
5125 Description: readdir never returns NULL in liblustre.
5126 Details : Corrected llu_iop_getdirentries logic, to return offset of next
5127 dentry in struct dirent.
5131 Frequency : liblustre only, depends on application IO pattern
5132 Description: liblustre clients evicted if not contacting servers
5133 Details : Don't put liblustre clients into the ping_evictor list, so
5134 they will not be evicted by the pinger ever.
5136 Severity : enhancement
5138 Description: Add ability to evict clients by NID from MDS.
5139 Details : By echoing "nid:$NID" string into
5140 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
5141 $NID would be instantly evicted from this MDS and from all active
5142 OSTs connected to it.
5146 Description: Do not query file size twice, somewhat slowing stat(2) calls.
5147 Details : lookup_it_finish() used to query file size from OSTs that was not
5152 Description: service threads change working directory to that of init
5153 Details : Starting lustre service threads may pin the working directory
5154 of the parent thread, making that filesystem busy. Threads
5155 now change to the working directory of init to avoid this.
5159 Frequency : during shutdown only
5160 Description: shutdown with a failed MDS or OST can cause unmount to hang
5161 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
5166 Frequency : 2.6 only
5167 Description: chmod/chown may include an extra supplementary group
5168 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
5169 supplementary group and if none is specified this is used.
5172 Bugzilla : 5479 (6816)
5174 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
5175 Details : If lookup is in progress on a file that is unlinked we might try
5176 to revalidate the inode and fail in revalidate after lookup is
5177 complete and ll_file_open() enqueues the open again but
5178 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
5181 Frequency : always, if lconf --abort_recovery used
5183 Description: lconf --abort_recovery fails with 'Operation not supported'
5184 Details : lconf was attempting to abort recovery on the MDT device and not
5187 ------------------------------------------------------------------------------
5189 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
5194 Frequency : rare (only unsupported configurations with a node running as an
5196 Bugzilla : 6514, 5137
5197 Description: Mounting a Lustre file system on a node running as an OST could
5199 Details : OSTs now preallocates memory needed to write out data at
5200 startup, instead of when needed, to avoid having to
5201 allocate memory in possibly low memory situations.
5202 Specifically, if the file system is mounted on on OST,
5203 memory pressure could force it to try to write out data,
5204 which it needed to allocate memory to do. Due to the low
5205 memory, it would be unable to do so and the node would
5206 become unresponsive.
5208 Severity : enhancement
5210 Description: Addition of lconf --service command line option
5211 Details : lconf now accepts a '--service <arg>' option, which is
5212 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
5214 Severity : enhancement
5216 Description: Failover mode is now the default for OSTs.
5217 Details : By default, OSTs will now run in failover mode. To return to
5218 the old behaviour, add '--failout' to the lmc line for OSTs.
5220 Severity : enhancement
5222 Description: Health checks are now provided for MDS and OSTs
5223 Details : Additional detailed health check information on MSD and OSTs
5224 is now provided through the procfs health_check value.
5227 Frequency : occasional, depends on IO load
5229 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
5230 after numerous create/delete cycles
5231 Details : The ext3 inode allocation policy would not allocate new inodes
5232 very well on the OSTs because there are no new directories
5233 being created. Instead we look for groups with free space if
5234 the parent directories are nearly full.
5239 Description: Network or server problems during mount may cause partially
5240 mounted clients instead of returning an error.
5241 Details : The config llog parsing code may overwrite the error return
5242 code during mount error handling, returning success instead
5248 Description: MDS can fail to allocate large reply buffers
5249 Details : After long uptimes the MDS can fail to allocate large reply
5250 buffers (e.g. zconf client mount config records) due to memory
5251 fragmentation or consumption by the buffer cache. Preallocate
5252 some large reply buffers so that these replies can be sent even
5253 under memory pressure.
5257 Frequency : rare (liblustre)
5258 Description: fsx running with liblustre complained that using truncate() to
5259 extend the file doesn't work. This patch corrects that issue.
5260 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
5261 ATTR_SIZE and lsm use in llu_setattr_raw.
5265 Frequency : rare, only 2.6 kernels
5266 Description: Unusual file access patterns on the MDS may result in inode
5267 data being lost in very rare circumstances.
5268 Details : Bad interaction between the ea-in-inode patch and the "no-read"
5269 code in the 2.6 kernel caused the inode and/or EA data not to
5270 be read from disk, causing single-file corruption.
5274 Frequency : rare, only 2.6 filesystems using extents
5275 Description: Heavy concurrent write and delete load may cause data corruption.
5276 Details : It was possible under high-load situations to have an extent
5277 metadata block in the block device cache from a just-unlinked
5278 file overwrite a newly-allocated data block. We now unmap any
5279 metadata buffers that alias just-allocated data blocks.
5283 Frequency : filesystems with default stripe_count larger than 77
5284 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
5285 Details : lconf specifies an inode size of 4096 bytes when the default
5286 stripe_count is larger than 77. This conflicts with the default
5287 inode density of 1 per 4096 bytes. Allocate smaller inodes in
5288 this case to avoid pinning too much memory for large EAs.
5290 ------------------------------------------------------------------------------
5292 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
5297 Frequency : rare (extremely heavy IO load with hundreds of clients)
5299 Description: Client is evicted, gets IO error writing to file
5300 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
5301 caused clients to be evicted for AST timeouts. The fixes in
5302 bug 5192 mean we no longer need to have such short AST timeouts
5303 so ldlm_timeout has been increased.
5306 Frequency : occasional during --force or --failover shutdown under load
5307 Bugzilla : 5949, 4834
5308 Description: Server oops/LBUG if stopped with --force or --failover under load
5309 Details : a collection of import/export refcount and cleanup ordering
5310 issues fixed for safer force cleanup
5313 Frequency : only filesystems larger than 120 OSTs
5314 Bugzilla : 5990, 6223
5315 Description: lfs getstripe would oops on a very large filesystem
5316 Details : lov_getconfig used kfree on vmalloc'd memory
5319 Frequency : only filesystems exporting via NFS to Solaris 10 clients
5320 Bugzilla : 6242, 6243
5321 Description: reading from files that had been truncated to a non-zero size
5322 but never opened returned no data
5323 Details : ll_file_read() reads zeros from no-object files to EOF
5328 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
5329 Details : There's little harm in aborting MDS/OSS recovery and letting it
5330 try again, so I removed the LASSERT and return an error instead.
5332 Severity : enhancement
5334 Description: New debugging infrastructure for tracking down data corruption
5335 Details : The I/O checksum code was replaced to: (a) control it at runtime,
5336 (b) cover more of the client-side code path, and (c) try to narrow
5337 down where problems occurred
5341 Bugzilla : 3819, 4364, 4397, 6313
5342 Description: Racing close and eviction MDS could cause assertion in mds_close
5343 Details : It was possible to get multiple mfd references during close and
5344 client eviction, leading to one thread referencing a freed mfd.
5346 Severity: : enhancement
5347 Bugzilla : 3262, 6359
5348 Description: Attempts to reconnect to servers are now more aggressive.
5349 Details : This builds on the enhanced upcall-less recovery that was added
5350 in 1.4.2. When trying to reconnect to servers, clients will
5351 now try each server in the failover group every 10 seconds. By
5352 default, clients would previously try one server every 25 seconds.
5357 Description: After recovery, certain operations trigger a failed
5358 assertion on a client.
5359 Details : Failing over an mds, using lconf -d --failover, while a
5360 client was doing a readdir() call would cause the client to
5361 LBUG after recovery completed and the readdir() was resent.
5363 Severity : enhancement
5365 Description: Default groups are now added by lconf
5366 Details : You can now run lconf --group <servicename> without having to
5367 manually add groups with lmc.
5370 Frequency : occasional
5372 Description: Nodes with an elan id of 0 trigger a failed assertion
5375 Frequency : always when accessing e.g. tty/console device nodes
5377 Description: tty and some other devices nodes cannot be used on lustre
5378 Details : file's private_data field is used by device data and lustre
5379 values in there got lost. New field was added to struct file to
5380 store fs-specific private data.
5383 Frequency : when exporting Lustre via NFS
5385 Description: NFSD failed occasionally when looking up a path component
5386 Details : NFSD is looking up ".." which was broken in ext3 directories
5387 that had grown large enough to become hashed.
5390 Frequency : Clusters with multiple interfaces not on the same subnet
5392 Description: Nodes will repeatedly try to reconnect to an interface which it
5393 cannot reach and report an error to the log.
5394 Details : Extra peer list entries will be created by lconf with some peers
5395 unreachable. lconf now validates the peer before adding it.
5398 Frequency : Only if a default stripe is set on the filesystem root.
5400 Description: Setting a default stripe on the filesystem root prevented the
5401 filesystem from being remounted.
5402 Details : The client was sending extra request flags in the root getattr
5403 request and did not allocate a reply buffer for the dir EA.
5406 Frequency : occasional, higher if lots of files are accessed by one client
5407 Bugzilla : 6159, 6097
5408 Description: Client trips assertion regarding lsm mismatch/magic
5409 Details : While revalidating inodes the VFS looks up inodes with ifind()
5410 and in rare cases can find an inode that is being freed.
5411 The ll_test_inode() code will free the lsm during ifind()
5412 when it finds an existing inode and then the VFS later attaches
5413 this free lsm to a new inode.
5417 Bugzilla : 6422, 7030
5418 Description: MDS deadlock between mkdir and client eviction
5419 Details : Creating a new file via mkdir or mknod (starting a transaction
5420 and getting the ns lock) can deadlock with client eviction
5421 (gets ns lock and trying to finish a synchronous transaction).
5424 Frequency : occasional
5425 Description: While starting a server, the fsfilt_ext3 module could not be
5427 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
5428 kernels. Previously, lconf would still use the ext3 name
5429 when trying to load modules. Now, it will correctly use
5430 ext3 on 2.4 and ldiskfs on 2.6.
5432 Severity : enhancement
5433 Description: The default stripe count has been changed to 1
5434 Details : The interpretation of the default stripe count (0, to lfs
5435 or lmc) has been changed to mean striping across a single
5436 OST, rather than all available. For general usage we have
5437 found a stripe count of 1 or 2 works best.
5439 Severity : enhancement
5440 Description: Add support for compiling against Cray portals.
5441 Details : Conditional compiling for some areas that are different
5445 Frequency : occasional
5446 Bugzilla : 6409, 6834
5447 Description: Creating files with an explicit stripe count may lead to
5448 a failed assertion on the MDS
5449 Details : If some OSTs are full or unavailable, creating files may
5450 trigger a failed assertion on the MDS. Now, Lustre will
5451 try to use other servers or return an error to the
5455 Frequency : occasional
5457 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
5458 caused lock timeout during readahead (since 1.4.2).
5459 Details : Processes doing readahead might match a lock that hasn't been
5460 granted yet if there are overlapping and conflicting lock
5461 requests. The readahead process waits on ungranted lock
5462 (original lock is CBPENDING), while OST waits for that process
5463 to cancel CBPENDING read lock and eventually evicts client.
5465 Severity : enhancement
5467 Description: Initial enabling of flock support for clients
5468 Details : Implements fcntl advisory locking and file status functions.
5469 This feature is provided as an optional mount flag (default
5470 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
5471 locking are implemented yet, and those that are are not guaranteed
5472 to be completely correct in production environments.
5473 mount -t lustre -o [flock|noflock] ...
5476 Frequency : occasional
5478 Description: OSTs running 2.4 kernels but with extents enabled might trip an
5479 assertion in the ext3 JBD (journaling) layer.
5480 Details : The b_committed_data struct is protected by the big kernel lock
5481 in 2.4 kernels, serializing journal_commit_transaction() and
5482 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
5483 access to this struct. In 2.6 kernels there is finer grained
5484 locking to improve SMP performance of the JBD layer.
5488 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
5493 Description: merge and cleanup kernel patches.
5496 -----------------------------------------------------------------------------
5498 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
5500 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
5501 versions of Lustre. Please update all servers and clients to
5502 version 1.4.2 or later at the same time. You must also run
5503 "lconf --write-conf {config}.xml" on the MDS while it is stopped
5504 to update the configuration logs.
5506 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
5507 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
5508 - swab llog records of type '0' so we get proper header size/idx (5861)
5509 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
5510 - fix rename of one directory over another leaking an inode (5953)
5511 - avoid SetPageDirty on 2.6 (5981)
5512 - don't re-add just-being-destroyed locks to the waiting list (5653)
5513 - when creating new directories, inherit the parent's custom
5514 striping settings if present parent (3048)
5515 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
5516 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
5517 - don't hold client locks on temporary worklist from l_lru (5666)
5518 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
5519 - automatically evict dead clients (5921)
5520 - Update file size properly in create+truncate+fstat case (6196)
5521 - Do not unhash mountpoint dentries, do not allow removal of
5523 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
5524 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
5525 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
5526 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
5527 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
5528 - Correctly handle reads of files with no objects (6243)
5529 - lctl recover will also mark a device active if deactivate used (5933)
5531 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
5532 - allow --write-conf on an MDS with different nettype than client (5619)
5533 - don't write config llogs to MDS for mounts not from that MDS (5617)
5534 - lconf should create multiple TCP connections from a client (5201)
5535 - init scripts are now turned off by default; run chkconfig --on
5536 lustre and chkconfig --on lustrefs to use them
5537 - upcalls are no longer needed for clients to recover to failover
5539 - add --abort-recovery option to lconf to abort recovery on device
5541 - add support for an arbitrary number of OSTs (3026)
5542 - Quota support protocol changes.
5543 - forward compatibility changes to wire structs (6007)
5544 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5545 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5546 - improved "lustre status" script
5547 - initialize blocksize for non-regular files (6062)
5548 - added --disable-server and --disable-client configure options (5782)
5549 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5550 - Vanilla 2.4.29 support
5551 - increase maximum number of obd devices to 520 (6242)
5552 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5553 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5554 SLES 9 2.6.5 kernels
5555 - limit stripes per file to 160 (the maximum EA size) (6093)
5557 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5560 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5561 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5562 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5563 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5564 - small ext3 extents cleanups and fixes (5733)
5565 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5566 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5567 - avoid needless client->OST connect, fix handle mismatch (5317)
5568 - fix DLM error path that led to out-of-sync client, long delays (5779)
5569 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5570 - fix several locking issues related to i_size (5492,5624,5654,5672)
5571 - don't move pending lock onto export if it is already evicted (5683)
5572 - fix kernel oops when creating .foo in unlinked directory (5548)
5573 - fix deadlock in obdfilter statistics vs. object create (5811)
5574 - use time_{before,after} to avoid timer jiffies wrap (5882)
5575 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5576 - Do not leak request if server was not able to process it (5154)
5577 - If mds_open unable to find parent dir, make that negative lookup(5154)
5578 - don't create new directories with extent-mapping (5909, 5936)
5580 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5581 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5582 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5583 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5584 - support for RHEL4 kernel on i686 (5773)
5585 - provide error messages when incompatible logs are encountered (5898)
5587 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5588 * version 1.4.0.10 (1.4.1 release candidate 1)
5590 - don't keep a lock reference when lock is not granted (4238)
5591 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5592 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5593 - drop import inflight refcount on signal_completed_replay error (5255)
5594 - unlock page after async write error during send (3677)
5595 - handle missing objects in filter_preprw_read properly (5265)
5596 - no transno return for symlink open, don't save no-trasno open (3440)
5597 - don't try to complete elan receive that already failed (4012)
5598 - free RPC server reply state on error (5406)
5599 - clean up thread from ptlrpc_start_thread() on error (5160)
5600 - readahead could read extra page into cache that wasn't ejected (5388)
5601 - prevent races in class_attach/setup/cleanup/detach (5260)
5602 - don't dereference de->d_inode after l_dput of de (5458)
5603 - use "int" for stripe value returned from lock_to_stripe (5544)
5604 - mballoc allocation and error-checking fixes in 2.6 (5504)
5605 - block device patches to fix I/O request sizes in 2.6 (5482)
5606 - look up hostnames for IB nals (5602)
5607 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5608 - don't start multiple acceptors for the same port (5277)
5609 - fix incorrect LASSERT in mds_getattr_name (5635)
5610 - export a proc file for general "ping" checking (5628)
5611 - fix "lfs check" to not block when the MDS is down (5628)
5613 - service request history (4965)
5614 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5615 - create an "evict_client" /proc entry on OSTs, like the MDS has
5616 - fix mount usage message, return errors per mount(8) (5168)
5617 - change grep [] to grep "[]" in tests so they work in more UMLs
5618 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5619 - remove ancient LOV_MAGIC_V0 EA support (5047)
5620 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5621 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5622 - allow readahead allocations to fail when low on memory (5383)
5623 - mmap locking landed again, after considerable improvement (2828)
5624 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
5626 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
5629 - send OST transaction number in read/write reply to free req (4966)
5630 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5631 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5632 - the watchdog thread now runs as interruptible (5246)
5633 - flock/lockf fixes (but it's still disabled, pending 5135)
5634 - don't use EXT3 constants in llite code (5094)
5635 - memory shortage at startup could cause assertion (5176)
5637 - reorganization of lov code
5638 - single portals codebase
5640 - add extents/mballoc support (5025)
5641 - direct I/O reads in the obdfilter (4048)
5642 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
5644 tbd Cluster File Systems, Inc. <info@clusterfs.com>
5647 - send OST transaction number in read/write reply to free req (4966)
5648 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5649 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5650 - the watchdog thread now runs as interruptible (5246)
5651 - handle missing objects in filter_preprw_read properly (5265)
5652 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5653 - drop import inflight refcount on signal_completed_replay error (5255)
5654 - unlock page after async write error during send (3677)
5655 - return original error code on reconstructed replies (3761)
5656 - no transno return for symlink open, don't save no-trasno open (3440)
5658 - add pid to ldlm debugging output (4922)
5659 - bump the watchdog timeouts -- we can't handle 30sec yet
5660 - extra debugging for orphan dentry/inode bug (5259)
5662 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
5665 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
5666 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
5667 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
5668 - fix NULL dereference in /proc/sys/portals/routes (4827)
5669 - allow failed mdc_close() operations to be interrupted (4561)
5670 - stop precreate on OST before MDS would time out on it (4778)
5671 - don't send partial-page writes before EOF from client (4410)
5672 - discard client grant for sub-page writes on large-page clients (4520)
5673 - don't free dentries not owned by NFS code, check generation (4806)
5674 - fix lsm leak if mds_create_objects() fails (4801)
5675 - limit debug_daemon file size, always print CERROR messages (4789)
5676 - use transno after validating reply (3892)
5677 - process timed out requests if import state changes (3754)
5678 - update mtime on OST during writes, return in glimpse (4829)
5679 - add mkfsoptions to LDAP (4679)
5680 - use ->max_readahead method instead of zapping global ra (5039)
5681 - don't interrupt __l_wait_event() during strace
5683 - add software watchdogs to catch hung threads quickly (4941)
5684 - make lustrefs init script start after nfs is mounted
5685 - fix CWARN/ERROR duplication (4930)
5686 - return async write errors to application if possible (2248)
5687 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
5688 - print NAL number in %x format (4645)
5689 - update barely-supported suse-2.4.21-171 series (4842)
5690 - support for sles 9 %post scripts
5691 - support for building 2.6 kernel-source packages
5692 - support for sles km_* packages
5694 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
5697 - ignore -ENOENT errors in osc_destroy (3639)
5698 - notify osc create thread that OSC is being cleaned up (4600)
5699 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
5700 - reconstruct ost_handle() like mds_handle() (4657)
5701 - create a new thread to do import eviction to avoid deadlock (3969)
5702 - let lconf resolve symlinked-to devices (4629)
5703 - don't unlink "objects" from directory with default EA (4554)
5704 - hold socknal file ref over connect in case target is down (4394)
5705 - allow more than 32000 subdirectories in a single directory (3244)
5706 - fix blocks count for O_DIRECT writes (3751)
5707 - OST returns ENOSPC from object create when no space left (4539)
5708 - don't send truncate RPC if file size isn't changing (4410)
5709 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
5710 - bind to privileged port in socknal and tcpnal (3689)
5712 - rate limit CERROR/CWARN console message to avoid overload (4519)
5713 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
5714 - basic mmap support (3918)
5715 - kernel patch series update from b1_4 (4711)
5717 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
5720 - avoid crash during MDS cleanup with OST shut down (2775)
5721 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
5722 - don't use bad inodes on the MDS (3744)
5723 - dynamic object preallocation to improve recovery speed (4236)
5724 - don't hold spinlock over lock dumping or change debug flags (4401)
5725 - don't zero obd_dev when it is force cleaned (3651)
5726 - print grants to console if they go negative (4431)
5727 - "lctl deactivate" will stop automatic recovery attempts (3406)
5728 - look for existing locks in ldlm_handle_enqueue() (3764)
5729 - don't resolve lock handle twice in recovery avoiding race (4401)
5730 - revalidate should check working dir is a directory (4134)
5732 - don't always mark "slow" obdfilter messages as errors (4418)
5734 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
5737 - don't close LustreDB during write_conf until it is done (3860)
5738 - fix typo in lconf for_each_profile (3821)
5739 - allow dumping logs from multiple threads at one time (3820)
5740 - don't allow multiple threads in OSC recovery (3812)
5741 - fix debug_size parameters (3864)
5742 - fix mds_postrecov to initialize import for llog ctxt (3121)
5743 - replace config semaphore with spinlock (3306)
5744 - be sure to send a reply for a CANCEL rpc with bad export (3863)
5745 - don't allow enqueue to complete on a destroyed export (3822)
5746 - down write_lock before checking llog header bitmap (3825)
5747 - recover from lock replay timeout (3764)
5748 - up llog sem before sending rpc (3652)
5749 - reduce ns lock hold times when setting kms (3267)
5750 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
5751 - fix NULL deref and obd_dev leak on setup error (3312)
5752 - replace some LBUG about llog ops with error handling (3841)
5753 - don't match INVALID dentries from d_lookup and spin (3784)
5754 - hold dcache_lock while marking dentries INVALID and hashing (4255)
5755 - fix invalid assertion in ptlrpc_set_wait (3880)
5757 - add libwrap support for the TCP acceptor (3996)
5758 - add /proc/sys/portals/routes for non-root route listing (3994)
5759 - allow setting MDS UUID in .xml (2580)
5760 - print the stack of a process that LBUGs (4228)
5762 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
5765 - don't cleanup request in ll_file_open() on failed MDS open (3430)
5766 - make sure to unset replay flag from failed open requests (3440)
5767 - if default stripe count is 0, use OST count for inode size (3636)
5768 - update parent mtime/ctime on client for create/unlink (2611)
5769 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
5770 - free recovery state on server during a forced cleanup (3571)
5771 - unregister_reply for resent reqs (3063)
5772 - loop back devices mounting and status check on 2.6 (3563)
5773 - fix resource-creation race that can provoke i_size == 0 (3513)
5774 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
5775 - more debugging for page-accounting assertion (3746)
5776 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
5777 - avoid deadlock after precreation failure (3758)
5778 - fix race and lock order deadlock in orphan handling (3450, 3750)
5779 - add validity checks when grabbing inodes from l_ast_data (3599)
5781 - add /proc/.../recovery_status to obdfilter (3428)
5782 - lightweight CDEBUG infrastructure, debug daemon (3668)
5783 - change default OSC RPC parameters to be better on small clusters
5784 - turn off OST read cache for files smaller than 32MB
5785 - install man pages and include them in rpms (3100)
5786 - add new init script for (un)mounting lustre filesystems (2593)
5787 - run chkconfig in %post for init scripts (3701)
5788 - drop scimac NAL (unmaintained)
5790 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
5793 - clean kiobufs before and after use (3485)
5794 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
5795 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
5796 - change init script to fail more clearly if not run as root (1528)
5797 - allow clients to reconnect during replay (1742)
5798 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
5799 - don't do DNS lookups on NIDs too small for IP addresses (3442)
5800 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
5801 - fix cond_resched (3554)
5802 - only evict unfinished clients after recovery (3515)
5803 - allow bulk resend, prevent data loss (3570)
5804 - dynamic ptlrpc request buffer allocation (2102)
5805 - don't allow unlinking open directory if it isn't empty (2904)
5806 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
5807 - remove extraneous obd dereference causing LASSERT failure (3334)
5808 - don't use get_cycles() when creating temp. files on the mds (3156)
5809 - hold i_sem when setting i_size in ll_extent_lock() (3564)
5810 - handle EEXIST for set-stripe, set proper directory name (3336)
5812 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
5813 - fix ksocknal_fmb_callback() error messages (2918)
5815 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
5818 - don't copy lvb into (possibly NULL) reply on error (2983)
5819 - don't deref dentry after dput, don't free lvb on error (2922)
5820 - use the kms to determine writeback rpc length (2947)
5821 - increment oti_logcookies when osc is inactive (2948)
5822 - update client's i_blocks count via lvb messages (2543)
5823 - handle intent open/close of special files properly (1557)
5824 - mount MDS with errors=remount-ro, like obdfilter (2009)
5825 - initialize lock handle to avoid ASSERT on error cleanup (3057)
5826 - don't use cancelling-locks' kms values (2947)
5827 - use highest lock extent for kms, not last one (2925)
5828 - don't dereference ERR_PTR() dentry in error handling path (3107)
5829 - fix thread race in portals_debug_dumplog() (3122)
5830 - create lprocfs device entries at setup instead of at attach (1519)
5831 - common AST error handler, don't evict client on completion race (3145)
5832 - zero nameidata in detach_mnt in 2.6 (3118)
5833 - verify d_inode after revalidate_special is valid in 2.6 (3116)
5834 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
5835 - initialize RPC timeout timer earlier for 2.6 (3219)
5836 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
5837 - print nal/nid for unknown nid (3258)
5838 - additional checks for oscc recovery before doing precreate (3284)
5839 - fix ll_extent_lock() error return code for 64-bit systems (3043)
5840 - don't crash in mdc_close for bad permissions on open (3285)
5841 - zero i_rdev for non-device files (3147)
5842 - clear page->private before handing to FS, better assertion (3119)
5843 - tune the read pipeline (3236)
5844 - fix incorrect decref of invalidated dentry (2350)
5845 - provide read-ahead stats and refine rpc in flight stats (3328)
5846 - don't hold journal transaction open across create RPC (3313)
5847 - update atime on MDS at close time (3265)
5848 - close LDAP connection when recovering to avoid server load (3315)
5849 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
5850 - don't leak open file on MDS after open resend (3325)
5851 - serialize filter_precreate and filter_destroy_precreated (3329)
5852 - loop device shouldn't call sync_dev() for nul device (3092)
5853 - clear page cache after eviction (2766)
5854 - resynchronize MDS->OST in background (2824)
5855 - refuse to mount the same filesystem twice on same mountpoint (3394)
5856 - allow llmount to create routes for mounting behind routers (3320)
5857 - push lock cancellation to blocking thread for glimpse ASTs (3409)
5858 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
5859 - fix rare problem with rename on htree directories (3417)
5861 - allow default OST striping configuration per directory (1414)
5862 - fix compilation for qswnal for 2.6 kernels (3125)
5863 - increase maximum number of MDS request buffers for large systems
5864 - change liblustreapi to be useful for external progs like lfsck (3098)
5865 - increase local configuration timeout for slow disks (3353)
5866 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
5868 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5871 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
5872 - don't overwrite extent policy data in reply if lock was blocked (2901)
5873 - drop filter export grants atomically with removal from device (2663)
5874 - del obd_self_export from work_list in class_disconnect_exports (2908)
5875 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
5876 - swab reply message in mdc_close, other PPC fixes (2464)
5877 - fix destroying of named logs (2325)
5878 - overwrite old logs when running lconf --write_conf (2264)
5879 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
5880 - fix race in target_handle_connect (2898)
5881 - mds_reint_create() should take same inode create lock (2926)
5882 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
5883 - don't close files for self_export to avoid uninitialized obd (2936)
5884 - allow MDS with the same name as client node (2939)
5885 - hold dentry reference for closed log files for unlink (2325)
5886 - reserve space for all logs during transactions (2059)
5887 - don't evict page beyond end of stripe extent (2925)
5888 - don't oops on a deleted current working directory (2399)
5889 - handle hard links to targets without a parent properly (2517)
5890 - don't dereference NULL lock when racing during eviction (2867)
5891 - don't grow lock extents when lots of conflicting locks (2919)
5893 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
5896 - account for cache space usage on clients to avoid data loss (974)
5897 - lfsck support in lustre kernel code (2349)
5898 - reduce journal credits needed for BRW writes (2370)
5899 - orphan handling to avoid losing space on client/server crashes
5900 - ptlrpcd can be blocked, stopping ALL progress (2477)
5901 - use lock value blocks to assist in proper KMS, faster stat (1021)
5902 - takes i_sem instead of DLM locks internally on obdfilter (2720)
5903 - recovery for initial connections (2355)
5904 - fixes for mds_cleanup_orphans (1934)
5905 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
5906 - block all file creations until orphan recovery completes (1901)
5907 - client remove rq_connection from request struct (2423)
5908 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
5909 - recovery timer race (2670)
5910 - mdc_close recovey bug (2532)
5911 - ptlrpc cleanup bug (2710)
5912 - mds timeout on local locks (2588)
5913 - namespace lock held during RPCs (2431)
5914 - handle interrupted sync write properly (2503)
5915 - don't try to handle a message that hasn't been replied to (2699)
5916 - client assert failure during cleanup after abort recovery (2701)
5917 - leak mdc device after failed mount (2712)
5918 - ptlrpc_check_set allows timedout requests to complete (2714)
5919 - wait for inflight reqs when ptlrpcd finishes (2710)
5920 - make sure unregistered services are removed from the srv_list
5921 - reset bulk XID's when resending them (caught by 1138 test)
5922 - unregister_bulk after timeout
5923 - fix lconf error (2694)
5924 - handle write after unfinished setstripe, stripe-only getstripe (2388)
5925 - readahead locks pages, leaves pending causing memory pressure (2673)
5926 - increase OST request buffers to 4096 on large machines (2729)
5927 - fix up permission of existing directories in simple_mkdir (2661)
5928 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
5929 - don't assign transno to errored transactions (2742)
5930 - don't delete objects on OST if given a bogus objid from MDS (2751)
5931 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
5932 - if rq_no_resend, then timeout request after recovery (2432)
5933 - fix MDS llog_logid record size, 64-bit array alignment (2733)
5934 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
5935 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
5936 - MDS recovery shouldn't delete objects that it has given out (2730)
5937 - if enqueue arrives after completion, don't clobber LVB (2819)
5938 - don't unlock pages twice when trigger_group_io returns error (2814)
5939 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
5940 - don't write pages to disk if there was an error (1450)
5941 - don't ping imports that have recovery disabled (2676)
5942 - take buffered bytes into account when balancing socknal conn (2817)
5943 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
5944 - reconnect unlink llog connection after MDS reconnects to OST (2816)
5945 - remove little-endian swabbing of llog records (1987)
5946 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
5947 - retry reposting request buffers if they fail (1191)
5948 - grow extent at grant time to avoid granting a revoked lock (2809)
5949 - lock revoke doesn't evict page if covered by a second lock (2765)
5950 - disable VM readahead to avoid reading outside lock extents (2805)
5952 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
5953 - updated kernel patches for hp-2.4.20 kernel (2681)
5955 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
5958 - fix truncated write corruption (2366)
5959 - fix for failed assertion in iopen_connect_dentry (1792,2517)
5961 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
5962 - ensure len(uuid) < 37 in lmc (1171)
5963 - fix ia64 OOPS in llog_test (2255)
5964 - zero end of page at obdfilter for partial page writes (2648)
5965 - don't leave stale dentries around after renames (bug 2428)
5966 - fix timeouts when evicting a client with a single lock held (2642)
5967 - set deadline for the initial HELLO message to drain (2634)
5968 - print out dotted-quad IP addresses in the socknal (2302)
5970 - additional debugging for MDS client eviction problem (2443)
5971 - fix mkfsoptions support for osts (2603, 2604)
5973 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
5976 - add series for the vanilla 2.6.0 kernel
5977 - add series for the vanilla 2.4.24 kernel
5978 - add series for a cray x86/64 UL kernel drop
5979 - fix xattr patches for the vanilla 2.4.19 series
5981 - generate true UUIDs in lmc (1171)
5982 - have portals stack dumping break in UML (2466)
5983 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
5984 - allocate with _NOFS in ldlm to avoid deadlock (1933)
5985 - wake callback waiting threads on client eviction (2460)
5986 - Add --ptldebug and --subsystem to lmc (1719)
5987 - update assertion to allow safe interrupt allocation
5988 - set rq_no_resend for cancel requests (2432)
5989 - recalculate ptlrpcd timeout after resend (2494)
5990 - call vfs_rmdir when removing pending directories (2368)
5991 - fix renaming a file to itself (2429)
5992 - lmc creates a default one-stripe lov (2454)
5993 - expand procfs space to handle large clusters (2326)
5994 - increase UML stack to avoid overflow
5995 - update lconf's list of debug and subsystem masks
5996 - fix lfs find --obd (2510)
5997 - /proc tunable for disabling filter read caching (2591)
5998 - stop rpm packages from altering slapd.conf (2301)
5999 - disable nagle in the socknal under 0conf (2578)
6000 - choose mds inode size based on stripe count (2572)
6001 - fix kernel-source rpm problems (2516)
6003 - add --disable-doc to avoid pdf generation (2421)
6004 - update documentation, tests, type-os, comments
6005 - avoid format warnings on ia64
6006 - remove the TOE NAL
6007 - tiny code cleanups by removing unused fields
6009 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
6012 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
6013 - remove the most blatant lies from BUILDING file (2371)
6014 - change default debug level to reasonable production setting
6015 - reduce client side cache size to reduce cache flush time
6016 - reduce max RPCs in flight to avoid unnecessary file fragmentation
6017 - make TCP zerocopy and pinger support enabled by default (2476)
6018 - sync writes completed after process exits caused crashes (2319)
6019 - maintain correct mount count on the MDS (2356)
6020 - backout 1557, because 2316 wasn't really fixed
6021 - better file I/O statistics gathering in /proc
6022 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
6023 - another kernel patch to fix zero-copy TCP function export
6024 - don't take duplicate lock when processing re-sent getattr (2420)
6025 - lctl uses obd_self_export instead of creating new conn (2353)
6026 - MDS/OST recovery case which requires object creation asserted (2425)
6027 - move lfs from /usr/sbin to /usr/bin in packages
6028 - fix race between mds_client_add and mds_client_free (2417)
6029 - use kmalloc instead of slabs in portals (2430)
6030 - don't create duplicate records when a failover MDS is present (2442)
6031 - remove unnecessary mount age check (2332)
6032 - don't remove directory inodes from locks prematurely (2451)
6033 - don't break if MDS service name is the same as hostname (2103)
6034 - fix races in client write RPC generation when cache full (2482)
6036 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
6039 - remove now-unused request->rq_obd (278)
6040 - if an allocation fails, print out how much memory we've used (1933)
6041 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
6042 - add the "configurable stack size" patch to most series files (1256)
6043 - ability to write large log records, for 100+ OST configs (2306)
6044 - fix NULL deref when filter_prep fails (2314)
6045 - fix operator precedence error in filter_sync
6046 - dynamic allocation of socknal TX descriptors (2315)
6047 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
6048 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
6049 - ensure that CWARN messages in Portals always get to the syslog
6050 - __init/__exit are not for prototype decls (ldlm_init/exit)
6051 - x86-64 compile warning fixes
6052 - fix gateway LMC keyword conflict (2318)
6053 - fix MDS lock inversions in getattr/reint paths (1844)
6054 - fix a rare lock re-ordering bug, which caused deadlock (2322)
6055 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
6056 - DLM race condition prevented some lock evictions (2328)
6057 - ENOMEM detection and retry on socknal sends (2230)
6058 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
6059 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
6060 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
6061 - better error messages when a client is rejected during recovery (1505)
6062 - avoid cancelling locks which were never granted, after failure (2330)
6063 - fix i_sem/journal inversion in mds_client_add (2333)
6064 - fix truncate/getattr lock cycle deadlock (2334)
6065 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
6066 - fix two rare exit paths which could leak an l_lock() ref (2321)
6067 - fencepost error in MDS/OST orphan recovery (2226)
6068 - make log record alignment 8 bytes (1988)
6069 - lstripe now fails when requested offset > ost_count (2237)
6070 - ensure that all kernel series have a complete list.h (1607)
6071 - fix crashes in special-file operations (2316)
6072 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
6074 - allow configurable automake binary, for testing new versions
6075 - small update to the lfs documentation
6077 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
6079 * fix negative export reference count in fsfilt_sync (2312)
6081 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
6082 * release candidate 0.9.1
6084 - orphans are moved into the PENDING directory for possible recovery
6085 - replayed opens now open by fid for orphan/rename safety (1042)
6086 - last close of an orphan inode generates a transno (683)
6087 - chdir() and mount() now pin the directory entry (1020)
6088 - avoid CERROR in normal ll_setattr_raw() error case (1500)
6089 - discard very old requests without processing them (1502)
6090 - remove some common, well-understood CERRORs (1505)
6091 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
6092 - clear "grant" flags in OST replies until OST grant code lands (1644)
6093 - fix read performance by not clobbering i_blksize on client (1598)
6094 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
6095 - make LRU size a /proc tunable, clears locks when reduced (707)
6096 - fix some lprocfs rot that prevented ptlbd from loading (1732)
6097 - server locks take references on exports now (1558)
6098 - build fixes for 2.4.20-rh trees (1663)
6099 - return an error from lov_create if all OSCs are inactive (1751)
6100 - fix import levels when a reconnect happens without a timeout (1597)
6101 - exit early from mds_open if we get a lookup error (1749)
6102 - partial page read at EOF wouldn't wait for disk before sending (1642)
6103 - avoid NULL deref in obdfilter when reading page past EOF (1592)
6104 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
6105 - fix LBUG in ll_it_open_error with rc = -2 (1861)
6106 - write/truncate lock inversion (1639)
6107 - Don't auto-load obdclass, portals modules during cleanup (1495)
6108 - fix timestamps from jumping to "now" (1763)
6109 - extra journal assertions (1648)
6110 - add an extra multiunlink test (1771)
6111 - fix read_record/write_record API (1776)
6112 - fix leak of offset_extent, possible incorrect i_size later (1772)
6113 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
6114 - Debugging for the kqswnal_get_idle_tx problems (1820)
6115 - Allow recovery to be attempted multiple times (1536)
6116 - Write out MDS last_rcvd file after it is first created (1600)
6117 - Fix tx_descriptor leak in failed transmit situations (1827)
6118 - ext3 journaling fixes for assertion failure after IO error (1871)
6119 - class_export_put() on freed export after completion AST error (1896)
6120 - Fix revalidate looping in VFS (1322)
6121 - Don't access a freed export during MDS_REINT timeout (1521)
6122 - Add open-unlink recovery support on the MDS (1673,1764)
6123 - Return an error if no MDS data was read from last_rcvd (1946)
6124 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
6125 - Don't setup a disk device that doesn't match exported UUID (317)
6126 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
6127 - avoid committing NULL handle in force close
6128 - local.sh is now a one-stripe LOV configuration
6129 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
6130 - fix ext3 htree duplicate directory entry corruption (1516)
6131 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
6132 - update to new LOV EA format (2097)
6133 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
6135 2003-06-15 Phil Schwan <phil@clusterfs.com>
6138 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
6139 - per-import/export recovery handling (958, 931, 959)
6140 - multiple last-rcvd slots, for serving multiple FSes (949)
6141 - connections are again shared between multiple imp/exports (963, 964)
6142 - "umount -f" would hang if any requests needed to be sent (393, 978)
6143 - avoid pinning large req buffer by copying for queued messages (989)
6144 - add "uuid" to "lctl device" command to help upcalls (991)
6145 - "open" RPCs with transnos would confuse recovery counters (1037)
6146 - do proper endian conversion of all wire messages (288, 340, 891)
6147 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
6148 - call waiting locks callback from LDLM recovery thread (1127, 1151)
6149 - fix ptlrpc_connection leak in target_handle_connect (1174)
6150 - fix import refcounting bug in OST and MDS cleanup (1134)
6151 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
6152 - fix very unlikely obd_types race condition (501)
6153 - remove osc_open hack for echo_client (1187)
6154 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
6155 - a failure in read_inode2 leads to deadlock (1139)
6156 - cancel ack-locks as soon as transaction is committed (1072)
6157 - fix major leaks and crashes in the bulk I/O path (937, 1057)
6158 - make sure to commitrw after any preprw to avoid deadlock (1162)
6159 - failing to execute a file in a lustre FS would lock inode (1203)
6160 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
6161 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
6162 - obd_uuid2tgt would walk off the end of the list (1255)
6163 - on IA64 the osc would give portals incorrect bulk size (1258)
6164 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
6165 - fix lock inversion caused by new llite matching code (1282)
6166 - limit the number of dirty pages on a client to 10MB (1286)
6167 - timed out locks were not being corrected cancelled (1289)
6168 - fix O_DIRECT above 4GB on IA-32 (1292)
6169 * major user-visible changes
6170 - fail out/fail over policy now controlled by the upcall (993)
6172 - add OBD_PING to check server availability and failure (954)
6173 - lustre messages are now sent in sending host order (288, 340, 891)
6174 - add eadatalen to MDS getattr reply (340)
6175 - OST read replies may contain second buffer, with per-page status (593)
6177 2003-03-11 Phil Schwan <phil@clusterfs.com>
6180 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
6181 - failed open()s could cause deadlock; fixed (867, 869)
6182 - stop cancelling OST locks when files are closed (481)
6183 - overlapping XID spaces caused network corruption (851, 853)
6184 - fix unsafe fsfilt counter arithmetic; change to atomic_t
6185 - setattr_raw added, to do single-RPC, server-side setattrs
6186 - lmc/lconf syntax change for OST UUIDs
6187 - fix crashy race condition between ptlrpc_free_req and osc_close
6188 - don't use request in mdc_enqueue if we hit a timeout (889)
6189 - don't set the inode i_size for regular files from the MDS (896)
6190 - handle out of order completion AST (842)
6191 - don't LBUG if a lock request times out after receiving AST (913)
6192 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
6193 - if a bad lock AST arrives, send an error instead of dropping entirely
6194 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
6195 - fix leak in bulk IO when only partially completed (899, 900, 926)
6196 - fix O_DIRECT for ia64 (55)
6197 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
6198 - C-z after timeout could hang a process forever; fixed (977)
6200 - client-side I/O cache (678, 924, 929, 941, 970)
6202 - READPAGE and SETATTRs which don't take server-side locks get
6205 2003-02-11 Phil Schwan <phil@clusterfs.com>
6208 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
6209 - Fix fsfilt_extN_readpage() to read a full page of directory
6210 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
6211 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
6212 - Fix imbalanced LOV object allocation and out-of-bound access (469)
6213 - Most intent operations were removed, in favour of a new RPC mode
6214 that does a single RPC to the server and bypasses most of the VFS
6215 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
6216 - Aggressively cancel local locks on DLM servers
6217 - mds_reint_unlink sends EA to the client if it's the last nlink.
6218 client uses that EA to unlink OST objects.
6219 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
6220 - recursive symlinks were fixed (439)
6221 - fixed NULL deref in DEBUG_REQ
6222 - filter_update_lastobjid no longer calls sync, which annoyed extN
6223 - fixed multi-client small-writes to a single file problem (445)
6224 - fixed mtime updates during file writes (607)
6225 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
6226 - fixed bug in obd_brw_read/write() (under guise of testing 367)
6227 - fixed Linux OST size reporting problem (444, 656)
6228 - OST now updates object mtime with writes or setattr (607, 619)
6229 - client verifies file size before zeroing page past EOF (445)
6230 - OST now writes last allocated objid to disk with allocation (108)
6231 - LOV on echo now works (409)
6233 - mds_reint_unlink sends a new buffer, with the EA included. this
6234 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
6235 set if a regular file was being unlinked, and it was the last link
6236 - use PtlGet from the target for bulk writes (315)
6237 - OST now updates object mtime with writes or setattr (607, 619)
6238 - LDLM now has a grant-time callback to revalidate locked items, if
6240 - Many MDS operations were reorganized to combat race conditions
6242 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
6243 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
6245 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
6248 - Fully reactivate OST imports after reconnection (512, others)
6249 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
6250 - More graceful error handling for truncating on dead OST (515)
6251 - Don't error out unless we're actually accessing dead stripes (474)
6252 - Fix garbage sizes when stripes are missing (410)
6253 - LRU counters were broken, causing constant lock purge (433, 432)
6254 - garbage on read from stripes with failed OSTs (441)
6255 - mark OSCs as active before reconnecting during recovery (438)
6256 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
6257 - lfind did not preserve OST order in output (443)
6258 - symlinks cause hung clients, incorrect data (439)
6259 - stop dereferencing request after dropping refcount (457)
6260 - don't LASSERT(spin_is_locked) on non-SMP (455)
6261 - fixes for many rename() bugs
6262 - fstat didn't correctly synchronize attributes (399)
6263 - server must handle lock cancellation during blocking AST prep (487)
6264 - bulk descriptors were free()d too soon (511)
6265 - fix paths in lconf, which would load incorrect modules (451, 507)
6266 - fix confusing lconf 'host not found' error message (386)
6267 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
6268 - fix race condition in mdc_blocking_ast() for inode access (526)
6269 - fix lov_unpackmd() unpacking wrong number of stripes (537)
6270 - fix lov_set_osc_active() marking wrong OSC inactive (440)
6271 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
6272 - fix multiple writes of stripe MD to MDS (358, maybe 519)
6273 - fix lstripe in several ways (kernel side) (527)
6274 - fix request leak in ldlm_cli_enqueue (262)
6275 - incorrect OSC was marked inactive after OST failure
6276 - call mds_fs_cleanup before unmounting filesystem (524)
6277 - fix races between taking ns_lock and ldlm_lock_change_resource
6278 - fix races updating LOV export open file list
6279 - fix lov_enqueue error path, avoid decref-ing bad lock handle
6280 - fix recovery NULL deref in ldlm_cli_cancel_unused
6281 - fix some DLM races by using new hash table for lock handles (419)
6282 - permit the client to specify desired inodes, at replay
6283 - duplicate requests when we queue them for replay reintegration
6284 - fix last_rcvd offset calculation
6285 - sync after each recovered transaction, so we always make progress
6286 - never, not always, ERESTART requests without transnos
6287 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
6289 - skip replay if the MDS says that the client is already connected
6290 - don't check for a recovery-enabled export to match lctl's UUID
6291 - don't INC_USE_COUNT for phantom exports
6292 - don't crash when cleaning up phantom exports (567)
6293 - don't double-finish or set replay data for errored mdc_open requests
6294 - abort requests when they time out, so we don't get old replies
6295 - send/receive replies for AST messages again
6296 - if the client says that it doesn't have the lock, cancel it on the
6298 - if we timeout during I/O, don't try to cancel an in-use lock; instead
6299 mark it as destroyed, it will all work out when decref is called
6300 - fix module use counts (22, 581)
6302 - ASTs now expect a reply (server cancels lock on error reply)
6304 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
6307 - fix many simultaneous client startup (392)
6308 - fix dentry->d_it clobbering
6309 - credentials weren't being shipped for readdir/getattr operations
6310 - remove invalid assertions triggered during some concurrent MD
6312 - proper Lustre versions added (336, 389)
6313 - fix memory leak for create error case (398)
6314 - fix LOV locking bug that would get cli/srv out of sync
6315 - fix echo client over LOV (409)
6316 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
6317 - fix double-O_EXCL intent crash (424)
6318 - avoid sending multiple lock CANCELs (352)
6320 - MDS can do multi-client recovery (modulo bugs in new code)
6322 - many updates, edits, cleanups
6324 2002-11-18 Phil Schwan <phil@clusterfs.com>
6327 - fix null d_it dereference (346)
6328 - fix full OST/dbench hang (333)
6329 - fix permission problem with file removal (286)
6330 - fix removal of OSCs from LOV when they fail
6331 - fix NULL deref during bulk timeout (214)
6332 - fix problems related to multiple filesystems on one MDS (241)
6333 - fixed serious subtle metadata locking bugs
6334 - free locks on clients when inodes are removed due to memory
6336 - fix inode pointer in lock data (285)
6337 - partial support for multiple MDS on a single host (241)
6338 - data locks weren't cancelled at clear_inode time (290, 311)
6339 - intent locks could lead to unbounded lock growth (205)
6340 - added a maximum lock count, an LRU list, and a flusher
6341 - fix multiple rename (365)
6342 - properly abstracted the echo client
6343 - OSC locked 1 byte too many; fixed
6344 - rewrote brw callback code:
6345 - fixed recovery bugs related to LOVs (306)
6346 - fixed too-many-pages-in-one-write crash (191)
6347 - fixed (again) crash in sync_io_timeout (214)
6348 - probably fixed callback-related race (385)
6350 - Add capability to MDS protocol
6351 - LDLM cancellations and callbacks on different portals
6353 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
6356 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
6357 - defer open object destruction to close time (601981, 138)
6358 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
6359 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
6360 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
6361 - don't call obd_setattr in ll_file_release for destroyed objects
6362 * protocol change to lustre_msg: move |version| and add |flags|
6363 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
6364 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
6365 * added replay of create, unlink, link and rename operations during
6366 MDS failover; recovery should be much more robust now
6367 * remove failed OSCs from LOVs (only lov_create uses this so far)
6368 * the lustre-HOWTO was brought (more) up to date (582544)
6370 2002-10-23 Phil Schwan <phil@clusterfs.com>
6373 - in-use dentries weren't being reused properly (617851)
6374 - prevent multiple LDLM setup (599178)
6375 - fix LOV size calculations for truncate (617853)
6376 - fix client handling of MDS intent errors (POSIX)
6377 - fix permission bug in lovstripe.c test (624321)
6378 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
6379 - truncate past end of file could corrupt data
6380 - proper cleanup after timeouts, crashes, etc (592524, 550815)
6381 - a race in recovery could return ETIMEDOUT to apps (623947)
6382 - building outside the source directory was fixed
6383 * the lustre-HOWTO was brought (more) up to date (582544)
6384 * major progress was made on recovery functionality
6386 2002-10-10 Phil Schwan <phil@clusterfs.com>
6389 - recovery deadlock fix
6390 - rm -rf causes LBUG fix (617817)
6391 - file open by multiple tasks fix (618962)
6392 - directory permissions bugs (602707 and 620007)
6393 - journal_stop fixed with locking (611313)
6394 - O_APPEND failures resolved (618273, perhaps 614459)
6395 - lconf PATH fix (619770)
6396 - IA64 build fix (621450)
6397 - RPC buffer sizes scale with amount of memory
6399 2002-10-01 Phil Schwan <phil@clusterfs.com>
6402 - locks would be cancelled without throwing away data pages,
6403 resulting in inconsistent data (605627)
6404 - inode attributes were not always being refreshed (605627, 612449)
6405 - lconf now continues to cleanup after lctl reports an error
6406 - MDS now enforces user permissions (602707)
6407 - lprocfs cleanup fixed, but not yet enabled (614157)
6408 - fixed infinite server hang, should a client not respond to an AST
6409 - avoid going into recovery if user calls readlink() with a buffer
6410 that's too small (613941)
6411 - AST RPCs no longer require replies (614867) -- this may be changed
6412 - don't crash server if client sends an IOV that's too big (611336)
6413 - fixed lock conversion deadlock (611892)
6414 - fixed the following of symlinks (614622)
6415 * recovery: the server can remove locks from a client that dies, other
6416 clients can make progress
6417 * more extN patch fixes
6418 * compile-time configurable ptlrpc buffer allocations
6420 - collaborative read cache document
6421 - Lustre Lite Performance CDR document-in-progress
6423 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6426 - fix typo in patch-2.4.18
6428 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6431 - clear ptlrpc request each time in handle_incoming_request()
6432 - unlink of files now destroys the object on the OST
6434 2002-09-19 Peter Braam <braam@clusterfs.com>
6436 * add hard link support
6437 * change obdfile creation method
6438 * kernel patch changed
6440 2002-09-19 Peter Braam <braam@clusterfs.com>
6443 - stack overflow bug in extN fixed
6445 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
6447 * documentation updates
6448 - add man pages for config tools
6449 - update tests/README to describe testing with new config tools
6450 - finish metadata API descriptions
6451 * bug fixes and cleanups
6452 - statfs workaround for 16TB limit
6453 - LOV stripe allocation improved, can stripe on subset of OSTs
6454 - LOV file size/IO offset was wrong for files > 4GB in size
6455 - object EA data was being dropped, caused files to be unreadable
6456 - memory overflow with non-LOV OST caused memory corruption
6457 - fixed regression tests to work with new config tools, obdfilter
6458 - fixed bug when directory size became larger than 1 block
6459 - fixed bug (for single client case) when PWD was deleted
6460 - invalidate local directory pages when doing intent-based ops
6461 - avoid LDLM oops when lock callback contained bad data
6463 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
6465 * documentation updates
6466 * bug fixes and cleanups
6467 - configuration tools
6470 - 64-bit compile warnings
6471 - 64-bit internal statfs data
6473 * test_brw on persistent OST devices
6475 * lprocfs (disabled)
6477 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
6479 * documentation updates
6480 * bug fixes and cleanups
6481 * configuration tools
6483 2002-08-30 Peter J. Braam <braam@clusterfs.com>
6486 * many small fixes to 0_5_4
6487 * io/network handling
6488 * thinkos in MDS operations
6490 2002-08-24 Peter J. Braam <braam@clusterfs.com>
6493 * crucial basic fixes to 0.5.3
6494 * IOR, Iozone work over Elan
6497 2002-08-07 Phil Schwan <phil@clusterfs.com>
6498 * version 0_5_3, our first alpha
6499 * we use the new Portals iovs
6500 * documentation updates
6501 * bug fixes and cleanups
6502 * small changes in the DLM wire protocol
6504 2002-07-25 Peter J. Braam <braam@clusterfs.com>
6505 * version 0_5_1 with some initial stability,
6506 * locking on MD and file I/O.
6507 * documentation updates
6508 * several bug fixes since 0.5.0
6509 * small changes in wire protocol
6511 2002-07-18 Phil Schwan <phil@clusterfs.com>
6513 * delivered as Lustre Light Alpha
6514 * fixed a crash after handling invalid MDS requests
6515 * fixed directory pages for architectures with non-4k pages sizes
6517 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
6518 * release version v0_4_4
6519 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
6521 * Fixes a number of interruption problems with OST operations.
6522 * Update documentation for portals header changes
6523 * Move all wire protocol structs/defines to lustre_idl.h
6524 * Fixes symlink length bug.
6525 * Add tcpdump to repository.
6527 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
6528 * release version v0_4_3
6529 * Fixes statfs for inodes on extN.
6530 * Fixes bug in runtests which would delete /etc/hosts.
6531 * Use 64-bit object IDs wherever possible (not into VFS though)
6532 Remove ost_get_info, which is unused by lustre, and out of date.
6534 2002-07-03 Peter Braam <braam@clusterfs.com>
6535 * release version v0_4_2 Fixes a lookup error (type not passed)
6536 * move forward to head of Portals
6537 * move forward to latest Lustre kernel
6539 2002-06-25 Peter Braam <braam@clusterfs.com>
6540 * release version v0_4_1. Hopefully stable on single node use.