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 2.6.32-642.15.1.el6 (RHEL6.8)
10 2.6.32-696.1.1.el6 (RHEL6.9)
11 3.10.0-514.16.1.el7 (RHEL7.3)
12 3.0.101-0.47.71 (SLES11 SP3)
13 3.0.101-100 (SLES11 SP4)
14 3.12.69-60.64.35 (SLES12 SP1)
15 4.4.59-92.17 (SLES12 SP2)
16 vanilla linux 4.6.7 (ZFS only)
17 * Client known to build on unpatched kernels:
18 2.6.32-431.29.2.el6 (RHEL6.5)
19 2.6.32-504.30.3.el6 (RHEL6.6)
20 2.6.32-573.26.1.el6 (RHEL6.7)
21 2.6.32-642.15.1.el6 (RHEL6.8)
22 2.6.32-696.1.1.el6 (RHEL6.9)
23 3.10.0-514.16.1.el7 (RHEL7.3)
24 3.0.101-0.47.71 (SLES11 SP3)
25 3.0.101-100 (SLES11 SP4)
26 3.12.69-60.64.35 (SLES12 SP1)
27 4.4.59-92.17 (SLES12 SP2)
29 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
30 * Recommended ZFS / SPL version: 0.6.5.9
31 * Tested with ZFS / SPL version: 0.6.5.9
32 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
33 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
34 stack overflow. For more information, please refer to bugzilla 17630.
35 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
36 the re-exporting nodes to be mounted with "32bitapi" mount option
38 --------------------------------------------------------------------------------
40 12-07-2016 Intel Corporation
42 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
43 for currently supported client and server kernel versions.
44 * Server known to build on patched kernels:
45 2.6.32-431.29.2.el6 (RHEL6.5)
46 2.6.32-504.30.3.el6 (RHEL6.6)
47 2.6.32-573.26.1.el6 (RHEL6.7)
48 2.6.32-642.11.1.el6 (RHEL6.8)
49 3.10.0-514.el7 (RHEL7.3)
50 3.0.101-0.47.71 (SLES11 SP3)
51 3.0.101-80 (SLES11 SP4)
52 3.12.62-60.64.8 (SLES12 SP1)
53 vanilla linux 4.5.7 (ZFS only)
54 * Client known to build on unpatched kernels:
55 2.6.32-431.29.2.el6 (RHEL6.5)
56 2.6.32-504.30.3.el6 (RHEL6.6)
57 2.6.32-573.26.1.el6 (RHEL6.7)
58 2.6.32-642.11.1.el6 (RHEL6.8)
59 3.10.0-514.el7 (RHEL7.3)
60 3.0.101-0.47.71 (SLES11 SP3)
61 3.0.101-80 (SLES11 SP4)
62 3.12.62-60.64.8 (SLES12 SP1)
64 * Recommended e2fsprogs version: 1.42.13.wc5 or newer
65 * Recommended ZFS / SPL version: 0.6.5.8
66 * Tested with ZFS / SPL version: 0.6.5.8
67 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
68 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
69 stack overflow. For more information, please refer to bugzilla 17630.
70 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
71 the re-exporting nodes to be mounted with "32bitapi" mount option
73 --------------------------------------------------------------------------------
75 02-29-2016 Intel Corporation
77 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
78 for currently supported client and server kernel versions.
79 * Server known to build on patched kernels:
80 2.6.32-431.29.2.el6 (RHEL6.5)
81 2.6.32-504.30.3.el6 (RHEL6.6)
82 2.6.32-573.12.1.el6 (RHEL6.7)
83 3.10.0-327.3.1.el7 (RHEL7.2)
84 3.0.101-0.47.71 (SLES11 SP3)
85 3.0.101-68 (SLES11 SP4)
86 vanilla linux 4.2.1 (ZFS only)
87 * Client known to build on unpatched kernels:
88 2.6.32-431.29.2.el6 (RHEL6.5)
89 2.6.32-504.30.3.el6 (RHEL6.6)
90 2.6.32-573.12.1.el6 (RHEL6.7)
91 3.10.0-327.3.1.el7 (RHEL7.2)
92 3.0.101-0.47.71 (SLES11 SP3)
93 3.0.101-68 (SLES11 SP4)
96 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
97 * Recommended ZFS / SPL version: 0.6.4.2
98 * Tested with ZFS / SPL version: 0.6.4.2
99 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
100 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
101 stack overflow. For more information, please refer to bugzilla 17630.
102 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
103 the re-exporting nodes to be mounted with "32bitapi" mount option
105 --------------------------------------------------------------------------------
107 03-10-2015 Intel Corporation
109 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
110 for currently supported client and server kernel versions.
111 * Server known to build on patched kernels:
112 2.6.32-431.29.2.el6 (RHEL6.5)
113 2.6.32-504.8.1.el6 (RHEL6.6)
114 3.0.101-0.46 (SLES11 SP3)
115 * Client known to build on unpatched kernels:
116 2.6.32-431.29.2.el6 (RHEL6.5)
117 2.6.32-504.8.1.el6 (RHEL6.6)
118 3.10.0-123.20.1.el7 (RHEL7)
119 3.0.101-0.46 (SLES11 SP3)
120 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
121 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
122 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
123 stack overflow. For more information, please refer to bugzilla 17630.
124 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
125 the re-exporting nodes to be mounted with "32bitapi" mount option
127 Severity : enhancement
129 Description: control OST-index in IDIF via ROCOMPAT flag.
130 Details : Introduce new flag OBD_ROCOMPAT_IDX_IN_IDIF that is stored in the
131 last_rcvd file. For new formatted OST device, it will be auto set;
132 for the case of upgrading from old OST device, you can enable it
133 via the lproc interface osd-ldiskfs.index_in_idif. With such flag
134 enabled, for new created OST-object, its IDIF-in-LMA will contain
135 the OST-index; for the existing OST-object, the OSD will convert
136 old format IDIF as new format IDIF with OST-index stored in the
137 LMA EA when accessing such OST-object or via OI scrub. Once such
138 flag is enabled, it cannot be reverted back, so the system cannot
139 be downgraded to the orignal incompatible version.
141 --------------------------------------------------------------------------------
143 07-30-2014 Intel Corporation
145 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
146 for currently supported client and server kernel versions.
147 * Server known to build on patched kernels:
148 2.6.32-431.20.3.el6 (RHEL6.5)
149 3.0.101-0.31 (SLES11 SP3)
150 * Client known to build on unpatched kernels:
151 2.6.32-431.20.3.el6 (RHEL6.5)
152 3.0.101-0.31 (SLES11 SP3)
155 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
156 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
157 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
158 stack overflow. For more information, please refer to bugzilla 17630.
159 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
160 the re-exporting nodes to be mounted with "32bitapi" mount option
162 --------------------------------------------------------------------------------
164 10-31-2013 Intel Corporation
166 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
167 * Server builds for kernels:
168 2.6.32-358.18.1.el6 (RHEL6.4)
169 * Client builds for unpatched kernels:
170 2.6.32-358.18.1.el6 (RHEL6.4)
171 3.0.92-0.8 (SLES11 SP3)
173 * Recommended e2fsprogs version: 1.42.7.wc2 or newer
174 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
175 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
176 stack overflow. For more information, please refer to bugzilla 17630.
177 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
178 the re-exporting nodes to be mounted with "32bitapi" mount option
180 --------------------------------------------------------------------------------
182 05-30-2013 Intel Corporation
184 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
185 * Server builds for kernels:
186 2.6.32-358.6.2.el6 (RHEL6.4)
187 * Client builds for unpatched kernels:
188 2.6.32-358.6.2.el6 (RHEL6.4)
189 3.0.74-0.6.6 (SLES11 SP2)
191 * Recommended e2fsprogs version: 1.42.7.wc1 or newer
192 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
193 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
194 stack overflow. For more information, please refer to bugzilla 17630.
195 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
196 the re-exporting nodes to be mounted with "32bitapi" mount option
198 --------------------------------------------------------------------------------
200 9-30-2012 Whamcloud, Inc.
202 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
203 * Server support for kernels:
204 2.6.32-279.5.1.el6 (RHEL6)
205 * Client support for unpatched kernels:
206 2.6.32-279.5.1.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 RHEL6.3 kernel to 2.6.32-279.5.1.el6.
221 Severity : enhancement
223 Description: Update RHEL6.3 kernel to 2.6.32-279.2.1.el6.
225 Severity : enhancement
227 Description: Update RHEL6.2 kernel to 2.6.32-220.17.1.el6
229 Severity : enhancement
231 Description: Update RHEL6.2 kernel to 2.6.32-220.13.1.el6
233 Severity : enhancement
235 Description: Update RHEL6.2 kernel to 2.6.32-220.7.1.el6
237 --------------------------------------------------------------------------------
239 04-30-2012 Whamcloud, Inc.
241 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
242 * Server support for kernels:
243 2.6.32-220.4.2.el6 (RHEL6)
244 * Client support for unpatched kernels:
245 2.6.32-220.4.2.el6 (RHEL6)
247 2.6.18-238.19.1.el5 (RHEL5)
248 2.6.16.60-0.69.1 (SLES10)
249 * Recommended e2fsprogs version: 1.41.90.wc4
250 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
251 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
252 stack overflow. For more information, please refer to bugzilla 17630.
253 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
254 the re-exporting nodes to be mounted with "32bitapi" mount option
256 Severity : enhancement
258 Description: Update RHEL6.2 kernel to 2.6.32-220.4.2.el6
260 Severity : enhancement
262 Description: Update RHEL6.2 kernel to 2.6.32-220.4.1.el6
264 Severity : enhancement
266 Description: Update RHEL6.2 kernel to 2.6.32-220.el6
268 Severity : enhancement
270 Description: Update RHEL6.1 kernel to 2.6.32-131.17.1.el6
272 --------------------------------------------------------------------------------
274 09-30-2011 Whamcloud, Inc.
276 * Server support for kernels:
277 2.6.18-238.19.1.el5 (RHEL5)
278 2.6.32-131.6.1.el6 (RHEL6)
279 * Client support for unpatched kernels:
280 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
281 2.6.18-238.19.1.el5 (RHEL5)
282 2.6.32-131.6.1.el6 (RHEL6)
283 2.6.29.4-167.fc11 (FC11)
284 2.6.16.60-0.69.1 (SLES10)
286 * Recommended e2fsprogs version: 1.41.90.wc3
287 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
288 Lustre filesystem with 4K stack may cause a stack overflow. For more
289 information, please refer to bugzilla 17630.
290 * NFSv4 reexport requires Lustre client on the re-exporting nodes to be
291 mounted with "32bitapi" mount option
293 Severity : enhancement
294 Description: Update RHEL6.1 kernel to 2.6.32-131.6.1.el6
296 Severity : enhancement
297 Description: Update RHEL5.6 kernel to 2.6.18-238.19.1.el5
299 Severity : enhancement
300 Description: Update RHEL5.6 kernel to 2.6.18-238.9.1.el5
303 Frequency : only with SLES10
305 Description: Use OFED "KMP" provided by Novell
306 Details : SLES10 SP3 ships with OFED in a separate "KMP" package.
307 Lustre is now built against this package. That means you need to
308 install the ofed-kmp package from Novell for the patchless client.
309 Note that the ofed-kmp that Novell ships may not exactly match the
310 kernel version but should still be compatible.
312 Severity : enhancement
314 Description: Update RHEL5.5 kernel to 2.6.18-194.17.1.el5
315 Update OEL5.5 kernel to 2.6.18-194.17.1.0.1.el5.
317 --------------------------------------------------------------------------------
319 2010-07-15 Oracle, Inc.
321 * Server support for kernels:
322 2.6.18-164.11.1.el5 (RHEL 5)
323 2.6.18-164.11.1.0.1.el5 (OEL 5)
324 * Client support for unpatched kernels:
325 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
326 2.6.18-164.11.1.el5 (RHEL 5),
327 2.6.18-164.11.1.0.1.el5 (OEL 5)
328 2.6.16.60-0.42.8 (SLES 10),
330 2.6.29.4-167.fc11 (FC11)
331 * Recommended e2fsprogs version: 1.41.10-sun2
333 Severity : enhancement
335 Description: Distribute lfs_migrate script with lustre
336 Details : The lfs_migrate mentioned in the manual is now included in the
337 lustre tree in lustre/scripts and will distributed in lustre RPMs.
339 Severity : enhancement
341 Description: Failover nid list added to import proc file.
343 Severity : enhancement
345 Description: "som_preview" mount option required to use SOM
346 Details : The clients will need to use som_preview mount option
350 Description: Lustre does not do 1MB IOs to HW RAID
351 Details : Bump MAX_PHYS/HW_SEGMENTS and SG_ALL to 256 in the RHEL5 kernel.
353 Severity : enhancement
355 Description: lfs quota output is not very convenient for awk/sed-parsing
356 Details : Some positions in lfs quota output table could be empty or
357 non-empty which made it hard to parse it with scripts, now
358 a dash is put instead of space where there is not supposed
363 Description: Update to OFED 1.5.1
364 Details : OFED 1.5.1 will be used in environments that don't use
365 vendor supplied OFED.
369 Description: Remove lnet. prefix from lctl params display
370 Details : Also renamed lnet.memused to lnet_memused to
375 Description: Use conf_param -d to remove permanant settings.
379 Description: Add list_param -R
380 Details : lctl list_param now supports recursive mode.
383 Bugzilla : 21829, 22312
384 Description: Fix llobdstat.
385 Details : 1. Now updating the write_bytes counter
386 2. Port script fixes from 1.8.
390 Description: decrease the usage of memory on clients.
391 Details : 1. On clients, recycle dentries and inodes unused.
392 2. Delete the code related to ll_deathrow(att 6215 in bug 1443). It
397 Description: llobdstat fix and enhancement
398 Details : add a counter to set a limit to how many samples will be returned
399 fix a wildcard in the path to limit to obdfilter stats only
403 Description: lfs check only allowed for root
404 Details : Code cleanups and added a regresion test of non-root use of
409 Description: MDS Oopses in ptlrpc_dispatch_difficult_reply()/__list_add()
410 Details : Fix hr->hr_index overload.
414 Description: Fix FIEMAP functionality.
418 Description: FC11 patchless client support added
422 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and
423 OEL5.4 kernel to 2.6.18-164.11.1.0.1.el5.
427 Description: New acc-sm script and various test-framework enhancments.
431 Description: Previously landed patch was reverted.
435 Description: Update SLES10 kernel to 2.6.16.60-0.42.8 (patchless)
439 Description: Add ldiskfs maxdirsize mount option.
443 Description: Fix mkfs.lustre support for 16TB LUNs.
447 Description: Print mmp_check_interval
448 Details : print mmp_check_interval and make it possible to abort
449 mount operation in case it takes too long.
453 Description: Add cascading_rw to lustre mpi tests.
457 Description: Store SOM data in EA.
458 Details : Also several SOM related cleanups.
462 Description: Add refcount to md_open_data
466 Description: Use refcount on lov_request_set to prevent memory corruption.
470 Description: Statahead: use parent dir's i_mutex for synchronization.
474 Description: loadgen improvements.
478 Description: Enable peer health detection in ksocklnd and ko2iblnd.
482 Description: Control DCACHE_LUSTRE_INVALID flag with MDS_INODE_LOOKCUP lock
486 Description: Fix race on obd_devcie:md_stats
490 Description: Fix SLES 11 client builds.
491 Details : Resolved uninitalized pointer gcc warnings.
495 Description: Allow non-root access for "lfs check".
496 Details : Added a check in obd_class_ioctl() for OBD_IOC_PING_TARGET.
500 Description: Clear imp_force_reconnect correctly in ptlrpc_connect_interpret()
502 Severity : enhancement
504 Description: Update kernel to RHEL5.4 2.6.18-164.6.1.el5 and
505 OEL5 2.6.18-164.6.1.0.1.el5(Both in-kernel OFED enabled).
509 Description: Sanity test_27p fails.
510 Details : OBD_FAIL_OST_ENOSPC has been moved to filter_statfs().
511 Sanity tests 27 were cleaned up and improved.
515 Description: Unify req format on client/servers
516 Details : Use new req_capsule API [almost] everywhere instead of old PTLRPC
517 buffers and swabbers approach..
522 Description: drop join file stuff.
524 Severity : enhancement
526 Description: Add 2.6.27 and SLES11 (patchless client) support
530 Description: Fix in ptlrpc_expire_one_request() to print the signed time
533 Severity : enhancement
535 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
538 Severity : enhancement
540 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
542 Severity : enhancement
544 Description: correctly handle big reply message.
545 Details : send LNet event if reply is bigger than buffer and adjust this buffer
550 Description: Drop unnecessary __GFP_NOMEMALLOC flag from filter_get_page()
552 Severity : enhancement
554 Description: Add support for OEL5.
558 Description: Conf-sanity.sh 50g test - deactivated OST should not cause a panic.
560 Severity : enhancement
562 Description: "weak-modules" support
563 Details : Implement "weak-modules" support which enables kernel modules
564 to be used with any kernel that implements the same kABI. In
565 order to achieve this modules are now installed in
566 /lib/modules/$(uname -r)/updates/kernel on all distributions.
570 Description: client could not reconnect to OST because of an active request.
571 Details : The client now retries to reconnect to the same server, if a
572 connect request failed with EBUSY or -EAGAIN. Test case 35b
573 conf-sanity.sh has been created.
577 Description: Don't allow make backward step on assignin osc next id.
578 Details : race between allocation next id and ll_sync thread can be cause
579 of set wrong osc next id and can be kill valid ost objects.
583 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
587 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
591 Description: Error handling in osc_statfs_interpret() has been improved.
592 Details : Check in osc_statfs_interpret() for EBADR.
596 Description: Do not send statfs() requests to OSTs disabled by administrator.
597 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
599 Severity : enhancement
601 Description: Added sanity-quota test to for lfs(1) to work with numeric uid/gid.
603 Severity : enhancement
605 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
607 Severity : enhancement
609 Description: Enhancement for lfs(1) command to use numeric uid/gid.
613 Description: Improved cleanup procedure for conf-sanity test_46a.
617 Description: Do not start bulk transfer, if client tried to reconnect.
618 Details : Do not start bulk transfer, if client tried to reconnect,
619 but failed due to active rpcs. Check for exp_abort_active_req
620 has been implemented.
624 Description: Prevent inconsistences between linux and lustre mount structures.
625 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
626 Make the sleep interruptible.
628 Severity : enhancement
630 Description: provide server to client comms path
631 Details : server to client communications path via new LDLM_SET_INFO rpc
635 Description: truncate starts GFP_FS allocation under transaction causing deadlock
636 Details : ldiskfs_truncate calls grab_cache_page which may start page
637 allocation under an open transaction. This may lead to
638 calling prune_icache with consequent lustre reentrance.
642 Description: send correctly lsm on open replay
643 Details : MDS is trust to LSM size on replay open, but client can set wrong size
648 Description: enable client interface failover
649 Details : When a child reconnects from another NID, properly update export
650 nid hash position and ldlm reverse import.
652 Severity : enhancement
654 Description: Communicate OST degraded/readonly state via statfs to MDS
655 Details : Flags in the statfs returned from OSTs indicate whether the
656 OST is in a degraded RAID state, or if the filesystem has
657 turned read-only after a filesystem error is detected.
662 Description: don't panic if EPROTO was hit when reading symlink
663 Details : correctly handling request reference in error cases.
665 Severity : enhancement
667 Description: Add LustreNetLink, a kernel-userspace communcation path.
669 Severity : enhancement
671 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
676 Description: access to llog context before init.
677 Details : move handling CATALOGS file at osc layer and forbid access to llog
681 Frequency : rare, only when client is writing to OST on local node
683 Description: Avoid deadlock for local client writes
684 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
685 memory freeing context. This allows OST threads to set the
686 PF_MEMALLOC flag on task structures in order to allocate memory
687 from reserved pools and complete IO.
688 Use GFP_HIGHUSER for OST allocations for non-local client writes,
689 so that the OST threads generate memory pressure and allow
690 inactive pages to be reclaimed.
692 Severity : enhancement
694 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
699 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
700 Details : move ldlm namespace creation in setup phase to avoid grab
701 _lprocfs_lock with cli_sem held.
705 Description: Temporarily disable grant shrink.
706 Details : Disable the feature for debugging.
709 Frequency : only when formatting test filesystems
711 Description: Unable to run several mkfs.lustre on loop devices at the same time
712 Details : mkfs.lustre returns error 256 on the concurrent loop devices
713 formatting. The solution is to proper handle the error.
715 Severity : enhancement
717 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
719 Severity : enhancement
721 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
726 Description: resolve race between obd_disconnect and class_disconnect_exports
727 Details : if obd_disconnect will be called to already disconnected export he
728 forget release one reference and osc module can't unloaded.
730 Severity : enhancement
732 Description: Allow tuning service thread via /proc
733 Details : For each service a new
734 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
735 created that can be used to set min/max thread counts, and get the
736 current number of running threads.
741 Descriptoin: don't return error if have partially created objects for file.
742 Details : lov_update_create_set uses set->set_success as index for created
743 objects, so if some requests will be failed, they will have hole
744 at end of array and qos_shrink_lsm for allocate correct lsm.
746 Severity : enhancement
748 Description: Update OFED support to 1.4.1
753 Description: Reduce small size read RPC
754 Details : Set read-ahead limit for every file and only do read-ahead when
755 available read-ahead pages are bigger than 1M to avoid small size
758 Severity : enhancement
760 Description: move AT tunable parameters for more consistent usage
761 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
763 Severity : enhancement
765 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
766 Details : allow skip disconnected ost for send statfs request and hide error
772 Description: fix racy locking of mballoc block bitmaps causing BUG
773 Details : The locking of the mballoc buddy bitmap and the in-memory
774 block bitmap was using two different spin locks in some
775 cases. This made it possible to incorrectly access the
776 mballoc bitmap while another process was modifying it,
777 causing a sanity assertion to fail. While no on-disk corruption
778 was reported, there was some risk of this happening.
781 Frequency : rare, on llog test 6
783 Descriptoin: don't allow connect to already connected import
784 Details : allowing connect to already connected import is hide connecting problem.
789 Descriptoin: allow kill process which wait statahead result
790 Details : for some reasons 'ls' can stick in waiting result from statahead,
791 in this case need way for kill this process.
793 Severity : enhancement
795 Description: Add state history info file, enhance import info file
796 Details : Track import connection state changes in a new osc/mdc proc file;
797 add overview-type data to the osc/mdc import proc file.
799 Severity : enhancement
801 Description: MDS create should not wait for statfs RPC while holding DLM lock.
803 Severity : enhancement
805 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
810 Descriptoin: OST grant too much space to client even there are not enough space.
811 Details : Client will shrink its grant cache to OST if there are no write
812 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
813 this grant cache if there are already not enough avaible space
814 (left_space < total_clients * 32M).
817 Frequency : start MDS on uncleanly shutdowned MDS device
819 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
820 Details : stay in waiting mds<>ost recovery finished produce random bugs
821 due race between two ll_sync thread for one lov target. send
822 ACTIVATE event only if connect realy finished and import have
826 Frequency : rare, connect and disconnect target at same time
828 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
829 Details : don't call obd_disconnect under lov_lock. this long time
830 operation and can block ptlrpcd which answer to connect request.
835 Descriptoin: don't lose wakeup for imp_recovery_waitq
836 Details : recover_import_no_retry or invalidate_import and import_close can
837 both sleep on imp_recovery_waitq, but we was send only one wakeup
841 Frequency : always with long access acl
843 Descriptoin: mds can't pack reply with long acl.
844 Details : mds don't control size of acl but they limited by reint/getattr
847 Severity : enhancement
849 Description: Update to SLES10 kernel-2.6.16.60-0.33.
851 Severity : enhancement
853 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
856 Frequency : start MDS on uncleanly shutdowned MDS device
858 Descriptoin: aborting recovery hang on MDS
859 Details : don't throttle destroy RPCs for the MDT.
862 Frequency : on remount
864 Description: external journal device not working after the remount
865 Details : clear dev_rdonly flag for external journal devices in
871 Description: shutdown vs evict race
872 Details : client_disconnect_export vs connect request race.
873 if client will evicted at this time - we start invalidate
874 thread without referece to import and import can be freed
878 Frequency : rare, need acl's on inode.
880 Description: client can't handle ost additional correctly
881 Details : if ost was added after client connected to mds client can have
882 hit lnet_try_match_md ... to big messages to wide striped files.
883 in this case need teach client to handle config events about add
884 lov target and update client max ea size at that event.
886 Severity : enhancement
888 Description: Changelogs
889 Details : Changelogs are a lightweight mechanism to track filesystem
890 metadata and namespace changes. The changelog is recorded
891 permanently on the MDTs, and is periodically "consumed" / purged
892 when records are no longer needed.
894 Severity : enhancement
896 Description: compact fld format with extents
897 Details : Store range of seq rather than every seq in FLD. Seq
898 controller update FLD rather than clients. In Case of CMD, mdt0
899 has FLD, all other metadata server act as non persistent proxy
900 for FLD queries and cache fld entries in fld cache.
905 Description: don't skip ost target if they assigned to file
906 Details : Drop slow OSCs if we can, but not for requested start idx.
907 This means "if OSC is slow and it is not the requested
908 start OST, then it can be skipped, otherwise skip it only
909 if it is inactive/recovering/out-of-space.
913 Description: more cleanup in mds_lov
914 Details : not send LOV EA under replay, we can't know about they size at this
915 time. Don't allow client connect to mds before any ost connected,
916 for avoid problems with LOV EA size and returning EIO to client.
918 Severity : enhancement
920 Description: Interoperability at server side (Disk interoperability)
922 Severity : enhancement
924 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
926 Severity : enhancement
928 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
930 Severity : enhancement
932 Description: New client IO stack (CLIO).
934 Severity : enhancement
936 Description: Commit on sharing. Eliminate inter-client dependencies between
937 uncommitted transactions by doing transaction commits.
938 Thereby clients may recovery independently.
941 Frequency : Create a symlink file with a very long name
943 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
944 Details : If there is no extra space in the request for early cancels,
945 ldlm_req_handles_avail() returns 0 instead of a negative value.
947 Severity : enhancement
949 Description: Add /proc entry for import status
950 Details : The mdc, osc, and mgc import directories now have
951 an import directory that contains useful import data for debugging
954 Severity : enhancement
956 Description: Re-disable certain /proc logging
957 Details : Enable and disable client's offset_stats, extents_stats and
958 extents_stats_per_process stats logging on the fly.
961 Frequency : Only on FC kernels 2.6.22+
963 Description: oops in statahead
964 Details : Do not drop reference count for the dentry from VFS when lookup,
965 VFS will do that by itself.
967 Severity : enhancement
969 Description: Generic /proc file permissions
970 Details : Set /Proc file permissions in a more generic way to enable non-
971 root users operate on some /proc files.
975 Description: Hitting mdc_commit_close() ASSERTION
976 Details : Properly handle request reference release in
977 ll_release_openhandle().
981 Description: quota recovery deadlock during mds failover
982 Details : This patch includes att18982, att18236, att18237 in bz14840.
984 1. fix osts hang when mds does failover with quotaon
985 2. prevent watchdog storm when osts threads wait for the
990 Frequency : only patchless client
991 Description: add workaround for race between add/remove dentry from hash
993 Severity : enhancement
995 Description: Allow OST glimpses to return PW locks
999 Description: LBUG when llog conf file is full
1000 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
1004 Description: Prevent import from entering FULL state when server in recovery
1008 Description: service mount cannot take device name with ":"
1009 Details : Only when device name contains ":/" will mount treat it as
1015 Description: replace ptlrpcd with the statahead thread to interpret the async
1016 statahead RPC callback
1020 Frequency : on recovery
1021 Description: I/O failures after umount during fail back
1022 Details : if client reconnected to restarted server we need join to recovery
1023 instead of find server handler is changed and process self
1024 eviction with cancel all locks.
1026 Severity : enhancement
1028 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
1030 Severity : enhancement
1032 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
1034 Severity : enhancement
1036 Description: Upcall on Lustre log has been dumped
1037 Details : Allow for a user mode script to be called once a Lustre log has
1038 been dumped. It passes the filename of the dumped log to the
1039 script, the location of the script can be specified via
1040 /proc/sys/lnet/debug_log_upcall.
1045 Description: avoid idr_remove called for id which is not allocated.
1046 Details : Move assigment s_dev for clustered nfs to end of initialization,
1047 to avoid problem with error handling.
1052 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
1053 Details : When connection is reused this not moved from CONN_UNUSED_HASH
1054 into CONN_USED_HASH and this prodice warning when put connection
1055 again in unused hash.
1058 Severity : enhancement
1060 Description: File striping can now be set to use an arbitrary pool of OSTs.
1062 Severity : enhancement
1064 Description: Export bytes_read/bytes_write count on OSC/OST.
1068 Description: Early reply size mismatch, MGC loses connection
1069 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
1070 the connect flags are properly negotiated.
1075 Description: quotas are not honored with O_DIRECT
1076 Details : all writes with the flag O_DIRECT will use grants which leads to
1077 this problem. Now using OBD_BRW_SYNC to guard this.
1081 Description: add quota statistics
1082 Details : 1. sort out quota proc entries and proc code.
1083 2. add quota statistics
1085 Severity : enhancement
1087 Description: enable quota support for HEAD.
1091 Description: Properly propagate oinfo flags from lov to osc for statfs
1092 Details : restore missing copy oi_flags to lov requests.
1094 Severity : enhancement
1096 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
1097 : l_getgroups(8), lst(8), routerstat(8)
1098 Details : included man pages for llobdstat(8), llstat(8),
1099 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
1101 Severity : enhancement
1103 Description: configure's --enable-quota should check the
1104 : kernel .config for CONFIG_QUOTA
1105 Details : configure is terminated if --enable-quota is passed but
1106 : no quota support is in kernel
1110 Description: Remove portals compatibility
1111 Details : Remove portals compatibility, not interoperable with releases
1116 Description: Resolve device initialization race
1117 Details : Prevent proc handler from accessing devices added to the
1118 obd_devs array but yet be intialized.
1120 Severity : enhancement
1122 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
1124 Severity : enhancement
1126 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
1131 Description: Using wrong pointer in osc_brw_prep_request
1132 Details : Access to array[-1] can produce panic if kernel compiled with
1133 CONFIG_PAGE_ALLOC enabled
1137 Description: Client runs out of low memory
1138 Details : Consider only lowmem when counting initial number of llap pages
1142 Description: *optional* service tags registration
1143 Details : if the "service tags" package is installed on a Lustre node
1144 When the filesystem is mounted, a local-node service tag will
1145 be created. See http://inventory.sun.com/ for more information
1146 about the Service Tags asset management system.
1150 Description: Kernel BUG tries to release flock
1151 Details : Lustre does not destroy flock lock before last reference goes
1152 away. So always drop flock locks when client is evicted and
1153 perform unlock regardless of successfulness of speaking to MDS.
1157 Description: add refcount protection for osc callbacks, avoid panic on shutdown
1161 Description: sanity test 65a fails if stripecount of -1 is set
1162 Details : handle -1 striping on filesystem in ll_dirstripe_verify
1167 Description: ASSERTION(CheckWriteback(page,cmd)) failed
1168 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
1171 Severity : enhancement
1173 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
1177 Description: do not process already freed flock
1178 Details : flock can possibly be freed by another thread before it reaches
1179 to ldlm_flock_completion_ast.
1183 Description: LBUG during stress test
1184 Details : Need properly lock accesses the flock deadlock detection list.
1188 Description: oops in page fault handler
1189 Details : kernel page fault handler can return two special 'pages' in error
1190 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
1194 Description: timeout with invalidate import.
1195 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
1196 should be handled by ptlrpcd. This produce long age waiting and
1197 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
1199 Severity : enhancement
1201 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
1205 Description: enable MGS and MDT services start separately
1206 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
1207 a co-located MGS without starting the MGS, which is a complement
1208 to 'nosvc' mount option.
1212 Frequency : after recovery
1213 Description: precreate to many object's after del orphan.
1214 Details : del orphan st in oscc last_id == next_id and this triger growing
1215 count of precreated objects. Set flag LOW to skip increase count
1216 of precreated objects.
1220 Frequency : rare, on clear nid stats
1221 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
1222 Details : when clean nid stats sometimes try destroy live entry,
1223 and this produce panic in free.
1227 Description: Stack overflow during MDS log replay
1228 ease stack pressure by using a thread dealing llog_process.
1232 Description: wait until IO finished before start new when do lock cancel.
1233 Details : VM protocol want old IO finished before start new, in this case
1234 need wait until PG_writeback is cleared until check dirty flag and
1235 call writepages in lock cancel callback.
1237 Severity : enhancement
1239 Description: using special macro for print time and cleanup in includes.
1243 Description: mds_mfd_close() ASSERTION(rc == 0)
1244 Details : In mds_mfd_close(), we need protect inode's writecount change
1245 within its orphan write semaphore to prevent possible races.
1249 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
1253 Frequency : rare, on shutdown ost
1254 Description: don't hit live lock with umount ost.
1255 Details : shrink_dcache_parent can be in long loop with destroy dentries,
1256 use shrink_dcache_sb instead.
1260 Description: don't panic with use echo client
1261 Details : echo client pass NULL as client nid pointer and this produce null
1262 pointer dereference.
1266 Description: fix build on ppc32
1267 Details : compile code with -m64 flag produce wrong object file for ppc32.
1271 Description: add message levels for liblustreapi
1275 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
1276 Details : Move the 'good_osts' check before the 'total_bavail' check. This
1277 will result in an -EAGAIN and in the exit call path we call
1278 alloc_rr() which will with increasing aggressiveness attempt to
1279 aquire precreated objects on the minimum number of required OSCs.
1283 Description: Use old size assignment to avoid deadlock
1284 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
1285 the scheduling while holding a spinlock. We do not need locking
1286 for size in ll_update_inode() because size is only updated from
1287 the MDS for directories or files without objects, so there is no
1288 other place to do the update, and concurrent access to such inodes
1289 are protected by the inode lock.
1293 Description: resolve "_IOWR redefined" build error on SLES10
1297 Description: dump the memory debugging after all modules are unloaded to
1298 suppress false negative in conf_sanity test 39
1300 Severity : enhancement
1302 Description: build kernel-ib packages for OFED 1.3 in our release cycle
1307 Description: fix SLES kernel versioning
1308 Details : the kernel version for our SLES 10 kernel did not include a "-"
1309 before the "smp" at the end. while this was not a problem in
1310 general, it did mean that software trying to use the kernel
1311 version to try to detect a vendor specific kernel would fail.
1312 this was most evident by the OFED build scripts.
1316 Description: Don't update lov_desc members until making sure they are valid
1317 Details : When updating lov_desc members via proc fs, need fix their
1318 validities before doing the real update.
1322 Description: don't put request into delay list while invalidate in flight.
1323 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
1324 import in flight. this produce timeout for invalidate and sometimes
1325 can cause stale data.
1329 Frequency : on ppc only
1330 Description: not convert ost objects for directory because it's not exist.
1331 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
1335 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
1337 Severity : enhancement
1339 Description: Update to support 2.6.22.14 vanilla kernel.
1343 Frequency : rare, on recovery
1344 Description: read procfs can produce deadlock in some situation
1345 Details : Holding lprocfs lock which send rpc can produce block for destroy
1346 obd objects and this also block reconnect with -EALREADY. This
1347 isn't fix all lprocfs bugs - but make it rare.
1349 Severity : enhancement
1351 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
1354 Frequency : frequent on X2 node
1356 Description: mdc_set_open_replay_data LBUG
1357 Details : Set replay data for requests that are eligible for replay.
1361 Description: lustre_mgs: operation 101 on unconnected MGS
1362 Details : When MGC is disconnected from MGS long enough, MGS will evict the
1363 MGC, and late on MGC cannot successfully connect to MGS and a lot
1364 of the error messages complaining that MGS is not connected.
1368 Frequency : on network error
1369 Description: panic with double free request if network error
1370 Details : mdc_finish_enqueue is finish request if any network error ocuring,
1371 but it's true only for synchronus enqueue, for async enqueue
1372 (via ptlrpcd) this incorrect and ptlrpcd want finish request
1375 Severity : enhancement
1377 Description: client-side metadata stat-ahead during readdir(directory readahead)
1378 Details : perform client-side metadata stat-ahead when the client detects
1379 readdir and sequential stat of dir entries therein
1382 Frequency : on start mds
1384 Description: Implement get_info(last_id) in obdfilter.
1387 Frequency : occasional
1389 Description: Correctly check stale fid, not start epoch if ost not support SOM
1390 Details : open with flag O_CREATE need set old fid in op_fid3 because
1391 op_fid2 was overwritten with new generated fid, but mds can answer
1392 with one of these two fids and both is not stale. Setattr
1393 incorrectly started an epoch and assume will be called
1394 done_writing, but without SOM done_writing ever being called.
1397 Frequency : rare, depends on device drivers and load
1399 Description: MDS or OSS nodes crash due to stack overflow
1400 Details : Code changes in 1.8.0 increased the stack usage of some functions.
1401 In some cases, in conjunction with device drivers that use a lot
1402 of stack the MDS (or possibly OSS) service threads could overflow
1403 the stack. One change which was identified to consume additional
1404 stack has been reworked to avoid the extra stack usage.
1407 Frequency : occasional
1409 Description: Do not fail import if osc_interpret_create gets -EAGAIN
1410 Details : If osc_interpret_create got -EAGAIN it immediately exits and
1411 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
1412 oscc_has_objects and see OSC has no objests and call
1413 oscc_internal_create to resend create request.
1415 Severity : enhancement
1417 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
1419 Severity : enhancement
1421 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
1424 Frequency : very rare
1426 Description: Fix replay if there is an un-replied request and open
1427 Details : In some cases, older replay request will revert the
1428 mcd->mcd_last_xid on MDS which is used to record the client's
1429 latest sent request.
1431 Severity : enhancement
1433 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
1435 Severity : enhancement
1437 Description: Add rhel5 support to HEAD.
1439 Serverity : enhancement
1441 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
1446 Description: Don't allow skipping OSTs if index has been specified.
1447 Details : Don't allow skipping OSTs if index has been specified, make
1448 locking in internal create lots better.
1452 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
1453 Details : access to bitfield in structure is always rounded to long
1454 and this produce problem with not atomic change any bit.
1458 Description: Lustre make rpms failed.
1459 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
1460 builds Lustre rpms from tarball.
1463 Frequency : rare on shutdown ost
1465 Description: If llog cancel was not send before clean_exports phase, this can
1466 produce deadlock in llog code.
1467 Details : If llog thread has last reference to obd and call class_import_put
1468 this produce deadlock because llog_cleanup_commit_master wait when
1469 last llog_commit_thread exited, but this never success because was
1470 called from llog_commit_thread.
1474 Description: allow userland application know is lost one of stripes.
1475 Details : fill lvb_blocks with error code on ost and return it to
1476 application if error flag found.
1480 Description: NULL lov_tgts causing MDS oops
1481 Details : more safe checks for NULL lov_tgts for avoid oops.
1483 Severity : enhancement
1485 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
1489 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
1490 Details : If an OST is down the MDS will hang indefinitely in
1491 obd_statfs_rqset() waiting for the statfs data. While for
1492 MDS QOS usage of statfs, it should not stuck in waiting.
1494 Severity : enhancement
1496 Description: remote_acl support
1497 Details : Support ACL-based permission check for remote user.
1498 Support setfacl/getfacl for remote user with the utils
1499 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
1500 the system "{s,g}etfacl" utils.
1502 Severity : enhancement
1504 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
1506 Severity : enhancement
1508 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
1512 Description: make mgs_setparam() handle fsname containing dash
1513 Details : fsname containing a dash does not work with lctl conf_param
1515 Severity : enhancement
1517 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
1519 Severity : enhancement
1521 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
1525 Frequency : rare, at shutdown
1526 Description: access already free / zero obd_namespace.
1527 Details : if client_disconnect_export was called without force flag set,
1528 and exist connect request in flight, this can produce access to
1529 NULL pointer (or already free pointer) when connect_interpret
1530 store ocd flags in obd_namespace.
1534 Frequency : only at startup
1535 Description: not alloc memory with spinlock held.
1536 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
1537 if any spinlock held.
1539 Severity : enhancement
1541 Description: make lustre randomly fail allocating memory
1542 Details : Make lustre randomly failed allocating memory for testing purpose.
1544 Severity : enhancement
1546 Description: lost problems with lov objid file
1547 Details : Fixes some scability and access to not inited memory problems
1548 in work with lov objdid file.
1553 Description: lfs find does not continue on file error
1554 Details : Continue other files processing when a file/dir is absent.
1558 Description: Inconsistent usage of lustre_pack_reply()
1559 Details : Standardize the usage of lustre_pack_reply() such that it
1560 always generate a CERROR on failure.
1563 Frequency : occasional
1565 Description: MDS hang or stay in waiting lock
1566 Details : If client receive lock with CBPENDING flag ldlm need send lock
1567 cancel as separate rpc, to avoid situation when cancel request
1568 can't processed due all i/o threads stay in wait lock.
1572 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1576 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1580 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1584 Description: Update to RHEL4 latest kernel.
1586 Severity : enhancement
1588 Description: Build SLES10 patchless client fails
1589 Details : The configure was broken by run ./configure with
1590 --with-linux-obj=.... argument for patchless client. When the
1591 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1592 can't search header adequately. Use absolute path such as
1593 -I($LINUX)/include instead.
1597 Description: interrupt oig_wait produce painc on resend.
1598 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1599 private set, and this produce situation when rq_ptlrpcd_data not
1600 copyed to new allocated request and triggered LBUG on assert
1601 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1602 ptlrpcd set if request is changed and can be safety changed to use
1607 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1608 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1609 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1610 these macros and replace them with their difinitions instead.
1614 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1615 Details : Update which_patch & target file for SLES10 latest kernel.
1618 Description: Remove client patches from SLES 10 kernel.
1619 Details : This causes SLES 10 clients to behave as patchless clients
1620 even on a Lustre-patched (server) kernel.
1622 Severity : enhancement
1624 Description: self-adjustable client's lru lists
1625 Details : use adaptive algorithm for managing client cached locks lru
1626 lists according to current server load, other client's work
1627 pattern, memory activities, etc. Both, server and client
1628 side namespaces provide number of proc tunables for controlling
1631 Severity : enhancement
1633 Description: light-weight GSS support
1634 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
1635 and reduce performance overhead.
1637 Severity : enhancement
1639 Description: Linux keyring support
1640 Details : Support using service of Linux keyring for Lustre GSS internal
1641 context refresh/cache mechanism.
1645 Description: Fix errors in lfs documentation
1646 Details : Fixes man pages
1650 Description: don't use GFP_* in generic Lustre code.
1651 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
1655 Description: obdclass is limited by single OBD_ALLOC(idarray)
1656 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
1660 Description: warnings with build patchless client with vanila 2.6.19 and up
1661 Details : change old ctl_table style and replace ctl_table/ctl_table_header
1662 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
1666 Frequency : Only for SLES
1667 Description: sanity_quota fail test_1
1668 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
1669 file, which makes TSTID[2] inunique.
1673 Frequency : Only for SLES9
1674 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
1675 Details : Update target/ChangeLog/which_patch .
1679 Frequency : Only for test_10 in sanity-quota.sh
1680 Description: a bug in quota test code
1681 Details : checking a test flag in a wrong place
1685 Frequency : Only for test_13 in sanity-quota.sh
1686 Description: a bug in quota test script
1687 Details : shouldn't check overfull allocated quota
1691 Description: jbd statistics
1692 Details : Port older jbd statistics patch for sles10
1694 Severity : enhancement
1696 Description: Add printing inode info into message about error in writepage.
1701 Description: reply_lock_interpret crash due to race with it and lock cancel.
1702 Details : Do not replay locks that are being cancelled. Do not reference
1703 locks by their address during replay, just by their handle.
1707 Frequency : When flocks are used.
1708 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
1709 Details : Flock locks might destroy just granted lock if it could be merged
1710 with another existing flock, this is done in completion handler,
1711 so teach ldlm_cli_enquque_fini that this is a valid case for
1717 Description: Oops in read and write path when failing to allocate lock.
1718 Details : Check if lock allocation failed and return error back.
1722 Description: lstripe command fails for valid OST index
1723 Details : The stripe offset is compared to lov->desc.ld_tgt_count
1724 instead of lov->desc.ld_active_tgt_count.
1728 Description: sanity.sh failed test 103
1729 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
1734 Description: df doesn't work properly if diskfs blocksize != 4K
1735 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
1739 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
1740 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
1742 Severity : enhancement
1744 Description: Adaptive timeouts
1745 Details : RPC timeouts adapt to changing server load and network
1746 conditions to reduce resend attempts and improve recovery time.
1750 Description: llapi_file_create() does not allow some changes
1751 Details : add llapi_file_open() that allows specifying the mode and
1752 open flags, and also returns an open file handle.
1756 Description: merge and cleanup kernel patches.
1757 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
1761 Description: Add journal checksum support.(Kernel part)
1762 Details : The journal checksum feature adds two new flags i.e
1763 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
1764 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
1765 indicates that the commit block contains the checksum for
1766 the blocks described by the descriptor blocks. Now commit
1767 record can be sent to disk without waiting for descriptor
1768 blocks to be written to disk. This behavior is controlled
1769 using JBD2_FEATURE_ASYNC_COMMIT flag.
1773 Description: OSS needs mutliple precreate threads
1774 Details : Add ability to start more than one create thread per OSS.
1778 Description: Sanity.sh test_65k failed: setstripe should have succeeded
1779 Details : execute lfs setstripe on client
1783 Description: mds_obd_create error creating tmp object
1784 Details : When the user sets quota on root, llog will be affected and can't
1785 create files and write files.
1788 Frequency : Always on ia64 patchless client, and possibly others.
1790 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
1791 Details : This allows the patchless client to be loaded on architectures
1792 without this export.
1796 Description: RedHat Update kernel for RHEL5
1797 Details : Add the kernel config file for RHEL5.
1801 Description: RedHat Update kernel for RHEL5
1802 Details : Modify the kernel config file more closer RHEL5.
1806 Description: Build failure against Centos5 (RHEL5)
1807 Details : Define PAGE_SIZE when it isn't present.
1811 Description: test_7 on sanity-quota.sh can't be used on separate machines
1812 Details : add facet to handle it
1816 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
1817 Details : with NFS, the anon dentry's parent was set to itself in
1818 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
1819 corresponding dentry other than use rec->ur_name.
1821 Severity : enhancement
1823 Description: lfs setstripe enhancement
1824 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
1828 Description: enable data checksumming by default
1829 Details : enable checksum by default, allow --disable-checksum
1830 configure option and "-o nochecksum" mount option. Checksums
1831 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
1835 Description: lustre support for RHEL5
1836 Details : Add support for RHEL5.
1840 Description: Client eviction due to failover config
1841 Details : after a connection loss, the lustre client should attempt to
1842 reconnect to the last active server first before trying the
1843 other potential connections.
1847 Description: when mds and osts use different quota unit(32bit and 64bit),
1848 quota will be released repeatly.
1849 Details : void sending multiple quota reqs to mds, which will keep the
1850 status between the reqs.
1854 Description: osts not allocated evenly to files
1855 Details : change the condition to increase offset_idx
1858 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
1860 Description: Data corruption for OSTs that are formatted larger than 2TB
1862 Details : When generating the bio request for lustre file writes the
1863 sector number would overflow a temporary variable before being
1864 used for the IO. The data reads correctly from Lustre (which
1865 will overflow in a similar manner) but other file data or
1866 filesystem metadata may be corrupted in some cases.
1870 Description: Tune the kernel for good SCSI performance.
1871 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
1872 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
1877 Description: rewrite ext2-derived code in obdclass/uuid.c
1878 Details : rewrite inherited code (uuid parsing code from ext2 utils)
1879 from scratch preserving functionality.
1883 Description: "lfs find -obd UUID" prints directories
1884 Details : "lfs find -obd UUID" will return all directory names instead
1885 of just file names. It is incorrect because the directories
1886 do not reside on the OSTs.
1890 Description: use i_size_read and i_size_write in 2.6 port
1891 Details : replace inode->i_size access with i_size_read/write()
1894 Frequency : when removing large files
1896 Description: scheduling issue during removal of large Lustre files
1897 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
1898 It causes scheduling issues when removing large files (17TB in the
1902 Frequency : only with liblustre clients on XT3
1904 Description: evictions taking too long
1905 Details : allow llrd to evict clients directly on OSTs
1908 Frequency : only on ppc
1910 Description: /proc/fs/lustre/devices broken on ppc
1911 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
1912 We should make sure the type of 'index' is loff_t and then cast
1913 explicitly as needed below. Do not assign an explicitly cast
1917 Frequency : only for rhel5
1919 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
1920 Details : Modify the target file & which_kernel.
1922 Severity : enhancement
1924 Description: omit set fsid for export NFS
1925 Details : fix set/restore device id for avoid EMFILE error and mark lustre
1926 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
1930 Frequency : Always, for kernels after 2.6.16
1931 Description: Fix warning idr_remove called for id=.. which is not allocated.
1932 Details : Last kernels save old s_dev before kill super and not allow
1933 to restore from callback - restore it before call kill_anon_super.
1937 Description: buffer overruns could theoretically occur
1938 Details : llapi_semantic_traverse() modifies the "path" argument by
1939 appending values to the end of the origin string, and a
1940 overrun may occur. Adding buffer overrun check in liblustreapi.
1944 Description: Fix error on 'ls .' at the top of the Lustre mount.
1945 Details : Don't revalidate dentry if it is a root dentry.
1949 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
1950 Details : Modify vm-tunables-rhel4.patch.
1954 Description: Kernel config for 2.6.18-vanilla.
1955 Details : Modify targets/2.6-vanilla.target.in.
1956 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
1957 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
1958 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
1959 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
1963 Description: improve handling recoverable errors
1964 Details : if request processig with error which can be recoverable on server
1965 request should be resend, otherwise page released from cache and
1970 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
1971 Details : Modify target file & which_patch.
1972 A flaw was found in the IA32 system call emulation provided
1973 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
1974 value could be stored in the %RAX register, which could trigger an
1975 out-of-bounds system call table access. An untrusted local user
1976 could exploit this flaw to run code in the kernel
1977 (ie a root privilege escalation). (CVE-2007-4573).
1981 Description: O_DIRECT bypasses client statistics.
1982 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
1983 read_ahead_stats were updated. copied the stats section from
1984 osc_send_oap_rpc() into async_internal().
1988 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
1992 Description: change order of libsysio includes
1993 Details : '#include sysio.h' should always come before '#include xtio.h'
1997 Description: handle "serious error: objid * already exists" more gracefully
1998 Details : If LAST_ID value on disk is smaller than the objects existing in
1999 the O/0/d* directories, it indicates disk corruption and causes an
2000 LBUG(). If the object is 0-length, then we should use the existing
2001 object. This will help to avoid a full fsck in most cases.
2003 Severity : enhancement
2005 Description: adapt the lustre_config script to support the upgrade case
2006 Details : Add "-u" option for lustre_config script to support upgrading 1.4
2007 server targets to 1.6 in parallel.
2011 Description: To avoid grant space > avaible space when the disk is almost
2012 full. Without this patch you might see the error "grant XXXX >
2013 available" or some LBUG about grant, when the disk is almost
2015 Details : In filter_check_grant, for non_grant cache write, we should
2016 check the left space by if (*left > ungranted + bytes), instead
2017 of (*left > ungranted), because only we are sure the left space
2018 is enough for another "bytes", then the ungrant space should be
2019 increase. In client, we should update cl_avail_grant only there
2020 is OBD_MD_FLGRANT in the reply.
2024 Description: Update RHEL 4 kernel to fix local root privilege escalation.
2025 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
2026 described in CVE-2007-4573. This problem could allow untrusted
2027 local users to gain root access.
2030 Frequency : when using O_DIRECT and quotas
2032 Description: Incorrect file ownership on O_DIRECT output files
2033 Details : block usage reported by 'lfs quota' does not take into account
2034 files that have been written with O_DIRECT.
2039 Description: touch file failed when fs is not full
2040 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
2041 otherwise we can get ENOSP while fs is not full.
2045 Description: parallel lock callbacks
2046 Details : Instead of sending blocking and completion callbacks as separated
2047 requests, adding them to a set and sending in parallel.
2050 Frequency : only for Cray XT3
2051 Bugzilla : 12829/13455
2052 Description: Changing primary group doesn't change the group lustre assigns to
2054 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
2055 overridden with the primary group provided by the group upcall,
2056 whereas the supplied fsgid can be trusted if it is in the list of
2057 supplementary groups returned by the group upcall.
2059 Severity : enhancement
2061 Description: Allow masking D_WARNING, D_ERROR messages from console
2062 Details : Console messages can now be disabled via lnet.printk.
2066 Description: User code with malformed file open parameter crashes client node
2067 Details : Before packing join_file req, all the related reference should be
2068 checked carefully in case some malformed flags cause fake
2069 join_file req on client.
2073 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
2074 Details : ldlm_completion_ast() assumes that a lock is granted when the req
2075 mode is equal to the granted mode. However, it should also check
2076 that LDLM_FL_CP_REQD is not set.
2080 Description: Heavy nfs access might result in deadlocks
2081 Details : After ELC code landed, it is now improper to enqueue any mds
2082 locks under och_sem, because enqueue might want to decide to
2083 cancel open locks for same inode we are holding och_sem for.
2087 Description: Client eviction while running blogbench
2088 Details : A lot of unlink operations with concurrent I/O can lead to a
2089 deadlock causing evictions. To address the problem, the number of
2090 oustanding OST_DESTROY requests is now throttled to
2091 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
2092 callbacks are processed in priority.
2096 Description: enable ACLs on MDS by default
2097 Details : ACLs must be enabled on MDS by default.
2100 Frequency : PPC/PPC64 only
2102 Description: conflicts between asm-ppc64/types.h and lustre_types.h
2103 Details : fix duplicated definitions between asm-ppc64/types.h and
2104 lustre_types.h on PPC.
2107 Frequency : PPC/PPC64 only
2109 Description: asm-ppc/segment.h does not exist
2110 Details : fix compile issue on PPC.
2114 Description: better handle error messages in extents code
2117 Frequency : RHEL4 only
2119 Description: mkfs is very slow on IA64/RHEL4
2120 Details : A performance regression has been discovered in the MPT Fusion
2121 driver between versions 3.02.73rh and 3.02.99.00rh. As a
2122 consequence, we have downgraded the MPT Fusion driver in the RHEL4
2123 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
2125 Severity : enhancement
2127 Description: SNMP support enhancement
2128 Details : Adding total number of sampled request for an MDS node in snmp
2131 Severity : enhancement
2133 Description: Optimize ldlm waiting list processing for PR extent locks
2134 Details : When processing waiting list for read extent lock and meeting read
2135 lock that is same or wider to it that is not contended, skip
2136 processing rest of the list and immediatelly return current
2137 status of conflictness, since we are guaranteed there are no
2138 conflicting locks in the rest of the list.
2142 Description: Time out and refuse to reconnect
2143 Details : When the failover node is the primary node, it is possible
2144 to have two identical connections in imp_conn_list. We must
2145 compare not conn's pointers but NIDs, otherwise we can defeat
2146 connection throttling.
2150 Description: port llog fixes from b1_6 into HEAD
2151 Details : Port llog reference couting and some llog cleanups from b1_6
2152 (bug 10800) into HEAD, for protect from panic and access to already
2153 free llog structures.
2157 Description: Detect stride IO mode in read-ahead
2158 Details : When a client does stride read, read-ahead should detect that and
2159 read-ahead pages according to the detected stride pattern.
2163 Description: data checksumming impacts single node performance
2164 Details : add support for several checksum algorithm. Currently, only CRC32
2165 and Adler-32 are supported. The checksum type can be changed on
2166 the fly via /proc/fs/lustre/osc/*/checksum_type.
2170 Description: use adler32 for page checksums
2171 Details : when available, use the Adler-32 algorithm instead of CRC32 for
2176 Description: build for x2 fails
2177 Details : fix compile issue on Cray systems.
2181 Description: Properly match for duplicate locks
2182 Details : Due to different lock order from skiplists code, we need to
2183 traverse entire list for now
2186 Frequency : only on PPC/SLES10
2188 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
2189 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
2190 be defined. Add a hack in mkfs_lustre.c to work around this compile
2195 Description: LASSERT on MDS when client holding flock lock dies
2196 Details : ldlm pool logic depends on number of granted locks equal to
2197 number of released locks which is not true for flock locks, so
2198 just exclude such locks from consideration.
2202 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
2203 Details : Use fsfilt_sync() for both the whole filesystem sync and
2204 individual file sync to eliminate dangerous inode locking
2205 with I_LOCK that can lead to a deadlock.
2209 Description: performance in 1.6.3
2210 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
2211 sure that 1MB requests are not fragmented by the block layer.
2213 Severity : enhancement
2215 Description: organize the server-side client stats on per-nid basis
2216 Details : Change the structure of stats under obdfilter and mds to
2224 The "uuid"s file would list the uuids of _active_ exports.
2225 And the clear entry is to clear all stats and stale nids.
2227 Severity : enhancement
2229 Description: eliminate client locks in face of contention
2230 Details : file contention detection and lockless i/o implementation
2231 for contended files.
2235 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
2238 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
2240 Description: MDS may lose file striping (and hence file data) in some cases
2241 Details : If there are additional extended attributes stored on the MDS,
2242 in particular ACLs, SELinux, or user attributes (if user_xattr
2243 is specified for the client mount options) then there is a risk
2244 of attribute loss. Additionally, the Lustre file striping
2245 needs to be larger than default (e.g. striped over all OSTs),
2246 and an additional attribute must be stored initially in the
2247 inode and then increase in size enough to be moved to the
2248 external attribute block (e.g. ACL growing in size) for file
2253 Description: skiplist implementation simplification
2254 Details : skiplists are used to group compatible locks on granted list
2255 that was implemented as tracking first and last lock of each
2256 lock group the patch changes that to using doubly linked lists
2260 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
2261 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
2262 because LASSERT on both the data supplied by a client, and the
2263 data on disk is dangerous and incorrect.
2265 Severity : enhancement
2267 Description: Slow truncate/writes to huge files at high offsets.
2268 Details : Directly associate cached pages to lock that protect those pages,
2269 this allows us to quickly find what pages to write and remove
2270 once lock callback is received.
2274 Description: more ldlm soft lockups
2275 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
2276 starve other threads from the resource lock for a long time in
2277 case of long waiting queue, so change the debug level from
2278 D_OTHER to the less frequently used D_INFO.
2280 Severity : enhancement
2282 Description: add -gid, -group, -uid, -user options to lfs find
2286 Description: Hung threads in invalidate_inode_pages2_range
2287 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
2288 once one is completed. As a result, some RPCs are stuck in the
2289 queue and are never sent.
2293 Description: filter threads hungs on waiting journal commit
2294 Details : Cleanup filter group llog code, then only filter group llog will
2295 be only created in the MDS/OST syncing process.
2299 Description: Procfs and llog threads access destoryed import sometimes.
2300 Details : Sync the import destoryed process with procfs and llog threads by
2301 the import refcount and semaphore.
2303 Severity : enhancement
2305 Description: openlock cache of b1_6 port to HEAD
2310 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
2311 Details : If insertion of an extent fails, then discard the inode
2312 preallocation and free data blocks else it can lead to duplicate
2317 Description: don't always update ctime in ext3_xattr_set_handle()
2318 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
2319 In some cases the ctime should not be updated, for example for
2320 2.0->1.8 compatibility it is necessary to delete an xattr and it
2321 should not update the ctime.
2325 Bugzilla : 15713/16362
2326 Description: Assertion in iopen_connect_dentry in 1.6.3
2327 Details : looking up an inode via iopen with the wrong generation number can
2328 populate the dcache with a disconneced dentry while the inode
2329 number is in the process of being reallocated. This causes an
2330 assertion failure in iopen since the inode's dentry list contains
2331 both a connected and disconnected dentry.
2335 Description: assertion failure in ldlm_handle2lock()
2336 Details : fix a race between class_handle_unhash() and class_handle2object()
2337 introduced in lustre 1.6.5 by bug 13622.
2342 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
2343 Details : remove the limit of 256 scsi disks in the sd_iostat patch
2348 Description: oops in sd_iostats_seq_show()
2349 Details : unloading/reloading the scsi low level driver triggers a kernel
2350 bug when trying to access the sd iostat file.
2355 Description: Kernel panics during QLogic driver reload
2356 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
2357 patch, causing memory corruption.
2362 Description: journal_dev option does not work in b1_6
2363 Details : pass mount option during pre-mount.
2365 Severity : enhancement
2367 Description: Add a FIEMAP(FIle Extent MAP) ioctl
2368 Details : FIEMAP ioctl will allow an application to efficiently fetch the
2369 extent information of a file. It can be used to map logical blocks
2370 in a file to physical blocks in the block device.
2374 Description: LDLM soft lockups - improvement
2375 Details : It is be possible to send the lock handle along with each read
2376 or write request because the client is already doing a lock match
2377 itself so there isn't any reason the OST should have to re-do that
2381 Frequency : only on Cray X2
2383 Description: X2 build failures
2384 Details : fix build failures on Cray X2.
2388 Description: xid & resent requests
2389 Details : Initialize RPC XID from clock at startup (randomly if clock is
2392 Severity : enhancement
2394 Description: Add lustre_start utility to start or stop multiple Lustre servers
2399 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
2400 Details : In case of memory pressure, list_del() can be called twice on
2401 req->rq_history_list, causing a kernel oops.
2405 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
2406 Details : fix stack overflow in the distributed lock manager by defering
2407 export eviction after a failed AST to the elt thread instead of
2408 handling it in the dlm interpret routine.
2412 Description: Convert some comments to new format.
2413 Details : Update documenting comments to match doxygen conventions.
2417 Description: Grammar fixes.
2418 Details : A couple of trivial spelling fixes.
2422 Description: OSD_COUNTERS-mandatory
2423 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
2424 to simplify the code.
2428 Description: simplify cmm_device freeing logic.
2429 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
2433 Description: Add lockdep support to dt_object_operations locking interface.
2434 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
2435 parameter indicating lock ordering. Update mdd code to use new
2440 Description: Introduce failloc constants for lockless IO tests.
2441 Details : Add two new failloc constants to test lockless IO. Only one of
2442 them in implemented---another is checked in yet to be landed
2447 Description: Add lockdep support for inode mutex.
2448 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
2449 in the situations where Lustre has to lock more than one inode
2454 Description: Add optional invariants checking support.
2455 Details : Add new LINVRNT() macro, optional on new --enable-invariants
2456 configure switch. This macro is to be used for consistency and
2457 sanity checks that are too expensive to be left in `production'
2462 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
2463 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
2468 Description: Add lockdep support for ldlm_lock and ldlm_resource.
2469 Details : Use spin_lock_nested() in (the only) situation where more than
2470 one ldlm_lock is locked simultaneously. Also, fix possible
2471 dead-lock in ldlm_lock_change_resource() by enforcing particular
2476 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
2477 Details : Instead of specifying each ldlm_lock call-back through separate
2478 parameter, wrap them into struct ldlm_callback_suite.
2482 Description: Kill join_lru obd method and its callers.
2483 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
2484 in memory---a requirement implemented through ->o_join_lru() obd
2485 method in HEAD. Remove this method and its users.
2489 Description: Add asynchronous ldlm ENQUEUE completion handler.
2490 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
2491 case that stock ldlm_completion_ast() cannot handle as it waits
2492 until lock is granted. Introduce new ldlm_completion_ast_async()
2493 for this. Also comment ldlm_completion_ast().
2497 Description: ldlm_error <-> errno conversion.
2498 Details : Add functions to map (rather arbitrary) between LDLM error codes
2499 and standard errno values. CLIO needs this to prevent LDLM specific
2500 constants from escaping ldlm and osc.
2504 Description: Kill unused ldlm_handle2lock_ns() function.
2505 Details : Kill unused ldlm_handle2lock_ns() function.
2509 Description: Add lu_ref support to ldlm_lock
2510 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
2511 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
2512 and ldlm_resource. LDLM interface has to be changed, because code that
2513 releases a reference on a lock, has to "know" what reference this is.
2514 In the most frequent case
2516 lock = ldlm_handle2lock(handle);
2518 LDLM_LOCK_PUT(lock);
2520 no changes are required. When any other reference (received _not_ from
2521 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
2522 instead of LDLM_LOCK_PUT().
2524 Arguably, changes are pervasive, and interface requires some discipline
2525 for proper use. On the other hand, it was very instrumental in finding
2526 a few leaked lock references.
2530 Description: Add ldlm_lock_addref_try().
2531 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
2532 attempts to addref a lock that might be being canceled
2537 Description: Add ldlm_weigh_callback().
2538 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
2539 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
2540 hard-coded `number of pages' logic.
2544 Description: Add lockdep annotations to llog code.
2545 Details : Use appropriately tagged _nested() locking calls in the places
2546 where llog takes more than one ->lgh_lock lock.
2550 Description: Add loi_kms_set().
2551 Details : Wrap kms updates into a helper function.
2555 Description: Constify instances of struct lsm_operations.
2556 Details : Constify instances of struct lsm_operations.
2560 Description: lu_conf support.
2561 Details : On a server, a file system object is uniquely identified
2562 by a fid, which is sufficient to locate and load all object
2563 state (inode). On a client, on the other hand, more data are
2564 necessary instantiate an object. Change lu_object_find() and
2565 friends to take additional `lu_conf' argument describing object.
2566 Typically this includes layout information.
2570 Description: lu_context fixes.
2571 Details : Introduce new lu_context functions that are needed on the client
2572 side, where some system threads (ptlrpcd) are shared by multiple
2573 modules, and so cannot be stopped during module shutdown.
2577 Description: Add start and stop methods to lu_device_type_operations.
2578 Details : Introduce two new methods in lu_device_type_operations, that are
2579 invoked when first instance of a given type is created and last one
2580 is destroyed respectively. This is need by CLIO.
2584 Description: Add lu_ref support to struct lu_device.
2585 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
2586 track leaked references.
2590 Description: Introduce lu_kmem_descr.
2591 Details : lu_kmem_descr and its companion interface allow to create
2592 and destroy a number of kmem caches at once.
2596 Description: Fix lu_object finalization race.
2597 Details : Fix a race between lu_object_find() finding an object and its
2598 concurrent finalization. This race is (most likely) not possible
2599 on the server, but might happen on the client.
2603 Description: Introduce lu_ref interface.
2604 Details : lu_ref is a debugging module allowing to track references to
2605 a given object. It is quite cpu expensive, and has to be
2606 explicitly enabled with --enable-lu_ref. See usage description
2611 Description: Factor lu_site procfs stats into a separate function.
2612 Details : Separate lu_site stats printing code into a separate function
2613 that can be reused on a client.
2617 Description: Constify instances of struct {lu,dt,md}_device_operations.
2618 Details : Constify instances of struct {lu,dt,md}_device_operations.
2622 Description: Introduce struct md_site and move meta-data specific parts of
2623 struct lu_site here.
2624 Details : Move md-specific fields out of struct lu_site into special struct
2625 md_site, so that lu_site can be used on a client.
2629 Description: Kill mdd_lov_destroy().
2630 Details : Remove unused mdd code.
2634 Description: Add st_block checking to multistat.c.
2635 Details : Add st_block checking to multistat.c.
2639 Description: Add lu_ref support to struct obd_device.
2640 Details : Add lu_ref tracking to obd_device.
2644 Description: Kill obd_set_fail_loc().
2645 Details : Remove unused code.
2649 Description: Add special type for ptlrpc_request interpret functions.
2650 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
2651 there. Actual usage will be in CLIO.
2655 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
2656 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
2657 doesn't work with lockdep.
2661 Description: Add rwv.c test program.
2662 Details : New testing program exercising readv(2) and writev(2) (Qian).
2666 Description: Add sendfile.c test program.
2667 Details : New testing program exercising sendfile(2) (Jay).
2671 Description: Ratelimit a message that can be very frequent.
2672 Details : Ratelimit a memory allocation failure message that can
2677 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2678 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2682 Description: Liblustre build fixes.
2683 Details : Liblustre build fixes.
2687 Description: libcfs: add cfs_{need,cond}_resched() interface.
2688 Details : libcfs: add cfs_{need,cond}_resched() definition and
2689 implementations for Linux, NT, and liblustre.
2691 Severity : enhancement
2693 Description: More exported tunables for mballoc
2694 Details : Add support for tunable preallocation window and new tunables for
2695 large/small requests
2699 Description: Detect corruption of block bitmap and checking for preallocations
2700 Details : Checks validity of on-disk block bitmap. Also it does better
2701 checking of number of applied preallocations. When corruption is
2702 found, it turns filesystem readonly to prevent further corruptions.
2706 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
2707 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
2708 window to 0. If it does hit cache miss, and read-pattern is still
2709 stride-io mode, does not reset the stride window, but also does not
2710 increase the stride window length in this case.
2714 Frequency : only for big-endian servers
2715 Description: Check if system is big-endian while mounting fs with extents feature
2716 Details : Mounting a filesystem with extents feature will fail on big-endian
2717 systems since ext3-based ldiskfs is not supported on big-endian
2718 systems. This can be over-riden with "bigendian_extents" mount option.
2720 Severity : enhancement
2722 Description: The root squash functionality
2723 Details : A security feature, which is to prevent users from being able
2724 to mount lustre on their desktop, run as root, and delete
2725 all of the files in the filesystem. The goal is accomplished by
2726 remapping user id (UID) and group id (GID) of the root user to
2727 a UID and GID specified by the system administartor via Lustre
2728 configuration management server (MGS). The functionality also
2729 allows to specify sets of clients for which the remapping does
2734 Description: Excessive recovery window
2735 Details : With AT enabled, the recovery window can be excessively long (6000+
2736 seconds). To address this problem, we no longer use
2737 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
2738 timeout no longer depends on the service time, it is set to
2739 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
2740 time via pb_service_time.
2744 Description: Watchdog triggered on MDS failover
2745 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
2746 the OSTs know that the MDS "UUID" can be reused for the same export
2747 from a different NID, so we do not need to wait for the export to be
2751 Frequency : rare, only if using MMP with Linux RAID
2753 Description: MMP doesn't work with Linux RAID
2754 Details : While using HA for Lustre servers with Linux RAID, it is possible
2755 that MMP will not detect multiple mounts. To make this work we
2756 need to unplug the device queue in RAID when the MMP block is being
2757 written. Also while reading the MMP block, we should read it from
2758 disk and not the cached one.
2760 Severity : enhancement
2762 Description: open file using fid
2763 Details : A file can be opened using just its fid, like
2764 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
2767 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
2770 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
2771 Details : add proper locking for creation of jbd_4k slab cache
2775 Description: MMP check in ext3_remount() fails without displaying any error
2776 Details : When multiple mount protection fails during remount, proper error
2779 Severity : enhancement
2781 Description: Allow stripe size to be up to 4G-64k
2782 Details : Fix math logic to allow large stripe sizes.
2786 Description: add check for >8TB ldiskfs filesystems
2787 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
2788 Don't allow >8TB ldiskfs filesystems to be mounted without
2789 force_over_8tb mount option
2794 Descriptoin: OSS complains "too many blocks" and crashes
2795 Details : In filter_commitrw_write():
2796 1. lquota_chkquota();
2797 2. obdo_from_inode();
2798 3. lquota_pending_commit();
2799 "2." has changed uid and gid in this case, then caused
2800 lquota_chkquota() and lquota_pending_commit() to use
2801 different uid and gid. That is the root of the bug.
2805 Descriptoin: getstripe doesn't report stripe count = -1 sanely
2806 Details : Detect and print out the 'stripe count = -1' case properly.
2810 Description: Out or order replies might be lost on replay
2811 Details : In ptlrpc_retain_replayable_request if we cannot find retained
2812 request with tid smaller than one currently being added, add it
2813 to the start, not end of the list.
2815 --------------------------------------------------------------------------------
2817 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
2819 * Support for kernels:
2820 2.6.5-7.283 (SLES 9),
2821 2.6.9-55.EL (RHEL 4),
2822 2.6.16.46-0.14 (SLES 10),
2823 2.6.18.8 vanilla (kernel.org)
2824 * Client support for unpatched kernels:
2825 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2826 2.6.16 - 2.6.22 vanilla (kernel.org)
2827 * Due to recently discovered recovery problems, we do not recommend
2828 using patchless RHEL 4 clients with this or any earlier release.
2829 * Recommended e2fsprogs version: 1.39.cfs8
2830 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2831 * Starting with this release, the ldiskfs backing filesystem required
2832 by Lustre is now in its own package, lustre-ldiskfs. This package
2833 should be installed. It is versioned separately from Lustre and
2834 may be released separately in future.
2838 Description: block reactivating mgc import until all deactivates complete
2839 Details : Fix race when failing back MDT/MGS to itself (testing)
2842 Frequency : at statup only
2844 Description: mds_lov_synchronize race leads to various problems
2845 Details : simultaneous MDT->OST connections at startup can cause the
2846 sync to abort, leaving the OSC in a bad state.
2848 Severity : enhancement
2850 Description: add optional extra BUILD_VERSION info
2851 Details : add a new environment variable (namely LUSTRE_VERS) which allows
2852 to override the lustre version.
2855 Frequency : 2.6.18 servers only
2857 Description: ll_kern_mount() doesn't release the module reference
2858 Details : The ldiskfs module reference count never drops down to 0
2859 because ll_kern_mount() doesn't release the module reference.
2864 Description: server LBUG when using old ost_num_threads parameter
2865 Details : Accept the old ost_num_threads parameter but warn that it
2866 is deprecated, and fix an off-by-one error that caused an LBUG.
2871 Description: Transient SCSI error results in persistent IO issue
2872 Details : iobuf->dr_error is not reinitialized to 0 between two
2876 Frequency : sometimes when underlying device returns I/O errors
2878 Description: OSTs not going read-only during write failures
2879 Details : OSTs are not remounted read-only when the journal commit threads
2880 get I/O errors because fsfilt_ext3 calls journal_start/stop()
2881 instead of the ext3 wrappers.
2885 Description: poor connect scaling with increasing client count
2886 Details : Don't run filter_grant_sanity_check for more than 100 exports
2887 to improve scaling for large numbers of clients.
2890 Frequency : SLES10 only
2892 Description: sanity-quota.sh quotacheck failed: rc = -22
2893 Details : Quotas cannot be enabled on SLES10.
2896 Frequency : liblustre clients only
2898 Description: getdirentries does not give error when run on compute nodes
2899 Details : getdirentries does not fail when the size specified as an argument
2900 is too small to contain at least one entry
2902 Severity : enhancement
2904 Description: Add LNET router traceability for debug purposes
2905 Details : If a checksum failure occurs with a router as part of the
2906 IO path, the NID of the last router that forwarded the bulk data
2907 is printed so it can be identified.
2912 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
2913 Details : Due to a race condition, liblustre clients were occasionally
2914 evicted incorrectly.
2916 Severity : enhancement
2918 Description: lfs setstripe use optional parameters instead of postional
2921 Severity : enhancement
2923 Description: Nanosecond timestamp support for ldiskfs
2924 Details : The on-disk ldiskfs filesystem has added support for nanosecond
2925 resolution timestamps. There is not yet support for this at
2926 the Lustre filesystem level.
2929 Frequency : during server recovery
2931 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
2932 Details : request with rq_no_resend flag not awake l_wait_event if they get
2936 Frequency : nfs export on patchless client
2938 Description: connectathon hang when test nfs export over patchless client
2939 Details : Disconnected dentry cannot be found with lookup, so we do not need
2940 to unhash it or make it invalid
2943 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
2944 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
2945 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
2950 Description: open req refcounting wrong on reconnect
2951 Details : If reconnect happened between getting open reply from server and
2952 call to mdc_set_replay_data in ll_file_open, we will schedule
2953 replay for unreferenced request that we are about to free.
2954 Subsequent close will crash in variety of ways.
2955 Check that request is still eligible for replay in
2956 mdc_set_replay_data().
2961 Description: disable writes to filesystem when reading health_check file
2962 Details : the default for reading the health_check proc file has changed
2963 to NOT do a journal transaction and write to disk, because this
2964 can cause reads of the /proc file to hang and block HA state
2965 checking on a healthy but otherwise heavily loaded system. It
2966 is possible to return to the previous behaviour during configure
2967 with --enable-health-write.
2969 Severity : enhancement
2971 Description: 64-bit inode version
2972 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
2973 to the inode. This will be required for version-based recovery.
2978 Description: MDS fails to start if a duplicate client export is detected
2979 Details : in some rare cases it was possible for a client to connect to
2980 an MDS multiple times. Upon recovery the MDS would detect this
2981 and fail during startup. Handle this more gracefully.
2983 Severity : enhancement
2985 Description: Add -o localflock option to simulate old noflock
2987 Details : This will achieve local-only flock/fcntl locks
2993 Description: log_commit_thread vs filter_destroy race leads to crash
2994 Details : Take import reference before releasing llog record semaphore
2999 Description: Wrong request locking in request set processing
3000 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
3001 imp_delayed_list, in this place should be used imp_lock.
3004 Frequency : when reconnection
3006 Description: Grant Leak when osc reconnect to OST
3007 Details : When osc reconnect ost, OST(filter) should check whether it
3008 should grant more space to client by comparing fed_grant and
3009 cl_avail_grant, and return the granted space to client instead
3010 of "new granted" space, because client will call osc_init_grant
3011 to update the client grant space info.
3014 Frequency : when client reconnect to OST
3016 Description: Grant Leak when osc do resend and replay bulk write
3017 Details : When osc reconnect to OST, OST(filter)should clear grant info of
3018 bulk write request, because the grant info while be sync between
3019 OSC and OST when reconnect, and we should ignore the grant info
3020 these of resend/replay write req.
3025 Description: Grant space more than available left space sometimes.
3026 Details : When then OST is about to be full, if two bulk writing from
3027 different clients came to OST. Accord the avaliable space of the
3028 OST, the first req should be permitted, and the second one
3029 should be denied by ENOSPC. But if the seconde arrived before
3030 the first one is committed. The OST might wrongly permit second
3031 writing, which will cause grant space > available space.
3034 Frequency : when client is evicted
3036 Description: Grant might be wrongly erased when osc is evicted by OST
3037 Details : when the import is evicted by server, it will fork another
3038 thread ptlrpc_invalidate_import_thread to invalidate the
3039 import, where the grant will be set to 0. While the original
3040 thread will update the grant it got when connecting. So if
3041 the former happened latter, the grant will be wrongly errased
3042 because of this race.
3047 Description: Checking Stale with correct fid
3048 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
3049 to check whether it is stale, because sometimes, we want the
3050 enqueue happened anyway, and op_data.fid2 will not be initialized.
3052 Severity : enhancement
3054 Description: update patchless client
3055 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
3058 Frequency : only with 2.4 kernel
3060 Description: random memory corruption
3061 Details : size of struct ll_inode_info is to big for union inode.u and this
3062 can be cause of random memory corruption.
3067 Description: Memory leak in recovery
3068 Details : Lov_mds_md was not free in an error handler in mds_create_object.
3069 It should also check obd_fail before fsfilt_start, otherwise if
3070 fsfilt_start return -EROFS,(failover mds during mds recovery).
3071 then the req will return with repmsg->transno = 0 and rc = EROFS.
3072 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
3073 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
3074 be freed no matter whether fsfilt_commit success or not.
3077 Frequency : only with huge count clients
3079 Description: Prevents from taking the superblock lock in llap_from_page for
3081 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
3082 for llap_from_page call in ll_removepage prevents from taking the
3083 superblock lock for a soon died page.
3088 Description: Not check open intent error before release open handle
3089 Details : in some rare cases, the open intent error is not checked before
3090 release open handle, which may cause
3091 ASSERTION(open_req->rq_transno != 0), because it tries to release
3092 the failed open handle.
3097 Description: Set cat log bitmap only after create log success.
3098 Details : in some rare cases, the cat log bitmap is set too early. and it
3099 should be set only after create log success.
3103 Description: Accessing a block bevice can re-enable I/O when Lustre is
3104 tearing down a device.
3105 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
3110 Description: service threads may hog cpus when there are a lot of requests
3111 Details : Insert cond_resched to give other threads a chance to use some CPU
3116 Description: the cat log was not initialized in recovery
3117 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
3118 unlink log on mds will not be initialized until mds post
3119 recovery. And also in mds post recovery, the unlink log will
3120 initialization will be done asynchronausly, so there will be race
3121 between add unlink log and unlink log initialization.
3125 Description: brw_stats were being printed incorrectly
3126 Details : brw_stats were being printed as log2 but all of them were not
3127 recorded as log2. Also remove some code duplication arising from
3128 filter_tally_{read,write}.
3132 Frequency : rare, only in recovery.
3133 Description: ASSERTION(req->rq_type != LI_POISON) failed
3134 Details : imp_lock should be held while iterating over imp_sending_list for
3135 prevent destroy request after get timeout in ptlrpc_queue_wait.
3139 Description: replay-single.sh test 52 fails
3140 Details : A lock's skiplist need to be cleanup when it being unlinked
3141 from its resource list.
3145 Description: Short directio read returns full requested size rather than
3147 Details : Direct I/O operations should return actual amount of bytes
3148 transferred rather than requested size.
3150 Severity : enhancement
3152 Description: metadata RPC reduction (e.g. for rm performance)
3153 Details : decrease the amount of synchronous RPC between clients and servers
3154 by canceling conflicing lock before the operation on the client
3155 and packing thier handles into the main operation RPC to server.
3157 Severity : enhancement
3159 Description: Async OSC create to avoid the blocking unnecessarily.
3160 Details : If an OST has no remaining object, system will block on the
3161 creation when it needs to create a new object on this OST. Now,
3162 ways use pre-created objects when available, instead of blocking on
3163 an empty osc while others are not empty. If we must block, we
3164 block for the shortest possible period of time.
3168 Description: improve handling recoverable errors
3169 Details : if request processig with error which can be recoverable on server
3170 request should be resend, otherwise page released from cache and
3173 Severity : enhancement
3175 Description: refine locking for avoid write wrong info into lov_objid file
3176 Details : fix possible races with add new target and write/update data in
3179 --------------------------------------------------------------------------------
3181 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
3186 Frequency : on some architectures
3188 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
3189 Details : Uninitialized flags sometimes cause configuration commands to
3193 Frequency : patchless clients only
3195 Description: missing __iget() symbol export
3196 Details : The __iget() symbol export is missing. To avoid the need for
3197 this on patchless clients the deathrow inode reaper is turned
3198 off, and we depend on the VM to clean up old inodes. This
3199 dependency was during via the fix for bug 12181.
3201 --------------------------------------------------------------------------------
3203 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
3205 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
3206 INTEROPERATE with older versions automatically. In many cases a
3207 special upgrade step is needed. Please read the
3208 user documentation before upgrading any part of a live system.
3209 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
3210 version will not interoperate with 1.6 betas before beta5 (1.5.95).
3211 * WARNING: Lustre configuration and startup changes are required with
3212 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
3216 Severity : enhancement
3218 Description: MountConf
3219 Details : Lustre configuration is now managed via mkfs and mount
3220 commands instead of lmc and lconf. New obd types (MGS, MGC)
3221 are added for dynamic configuration management. See
3222 https://mail.clusterfs.com/wikis/lustre/MountConf for
3225 Severity : enhancement
3227 Description: dynamic OST addition
3228 Details : OSTs can now be added to a live filesystem
3230 Severity : enhancement
3232 Description: startup order invariance
3233 Details : MDTs and OSTs can be started in any order. Clients only
3234 require the MDT to complete startup.
3236 Severity : enhancement
3238 Description: parallel, asynchronous orphan cleanup
3239 Details : orphan cleanup is now performed in separate threads for each
3240 OST, allowing parallel non-blocking operation.
3242 Severity : enhancement
3244 Description: optimized stripe assignment
3245 Details : stripe assignments are now made based on ost space available,
3246 ost previous usage, and OSS previous usage, in order to try
3247 to optimize storage space and networking resources.
3249 Severity : enhancement
3251 Description: Permanently set tunables
3252 Details : All writable /proc/fs/lustre tunables can now be permanently
3253 set on a per-server basis, at mkfs time or on a live system.
3255 Severity : enhancement
3257 Description: Lustre message v2
3258 Details : Add lustre message format v2.
3260 Severity : enhancement
3262 Description: client OST exclusion list
3263 Details : Clients can be started with a list of OSTs that should be
3264 declared "inactive" for known non-responsive OSTs.
3268 Description: SPEC SFS validation failure on NFS v2 over lustre.
3269 Details : Changes the blocksize for regular files to be 2x RPC size,
3270 and not depend on stripe size.
3272 Severity : enhancement
3274 Description: Multiple MD RPCs in flight.
3275 Details : Further unserialise some read-only MDS RPCs - learn about intents.
3276 To avoid overly-overloading MDS, introduce a limit on number of
3277 MDS RPCs in flight for a single client and add /proc controls
3278 to adjust this limit.
3280 Severity : enhancement
3282 Description: client read/write statistics
3283 Details : Add client read/write call usage stats for performance
3284 analysis of user processes.
3285 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
3286 file access. extents_stats shows chunk size distribution.
3287 extents_stats_per_process show chunk size distribution per
3290 Severity : enhancement
3292 Description: mds statistics
3293 Details : Add detailed mds operations statistics in
3294 /proc/fs/lustre/mds/*/stats.
3298 Description: Failure of copying files with lustre special EAs.
3299 Details : Client side always return success for setxattr call for lustre
3300 special xattr (currently only "trusted.lov").
3305 Description: Refcount LNET uuids
3306 Details : The global LNET uuid list grew linearly with every startup;
3307 refcount repeated list entries instead of always adding to
3310 Severity : enhancement
3312 Description: Dynamic service threads
3313 Details : Within a small range, start extra service threads
3314 automatically when the request queue builds up.
3317 Frequency : mixed-endian client/server environments
3319 Description: mixed-endian crashes
3320 Details : The new msg_v2 system had some failures in mixed-endian
3323 Severity : enhancement
3325 Description: Easy OST removal
3326 Details : OSTs can be permanently deactivated with e.g. 'lctl
3327 conf_param lustre-OST0001.osc.active=0'
3329 Severity : enhancement
3331 Description: MGS proc entries
3332 Details : Added basic proc entries for the MGS showing what filesystems
3335 Severity : enhancement
3337 Description: provide MGS failover
3338 Details : Added config lock reacquisition after MGS server failover.
3340 Severity : enhancement
3342 Description: add Linux 2.4 support
3343 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
3347 Description: a large application tries to do I/O to the same resource and dies
3348 in the middle of it.
3349 Details : Check the req->rq_arrival time after the call to
3350 ost_brw_lock_get(), but before we do anything about
3351 processing it & sending the BULK transfer request. This
3352 should help move old stale pending locks off the queue as
3353 quickly as obd_timeout.
3356 Frequency : when an incorrect nid is specified during startup
3358 Description: ptlrpc connect to non-existant node causes kernel crash
3359 Details : LNET can't be re-entered from an event callback, which
3360 happened when we expire a message after the export has been
3361 cleaned up. Instead, hand the zombie cleanup off to another
3364 Severity : enhancement
3366 Description: plain/inodebits lock performance improvement
3367 Details : Grouping plain/inodebits in granted list by their request modes
3368 and bits policy, thus improving the performance of search through
3372 Frequency : only if OST filesystem is corrupted
3374 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3375 Details : for a short time RPCs with bulk IO are in the replay list,
3376 but replay of bulk IOs is unimplemented. If the OST filesystem
3377 is corrupted due to disk cache incoherency and then replay is
3378 started it is possible to trip an assertion. Avoid putting
3379 committed RPCs into the replay list at all to avoid this issue.
3382 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3384 Description: System hang on startup
3385 Details : This bug allowed the liblustre (e.g. catamount) client to
3386 return to the app before handling all startup RPCs. This
3387 could leave the node unresponsive to lustre network traffic
3388 and manifested as a server ptllnd timeout.
3390 Severity : enhancement
3392 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3393 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
3394 boolean to control whether to print peer debug info when a
3395 client's RPC times out.
3398 Frequency : only for kernels with patches from Lustre below 1.4.3
3400 Description: Remove old rdonly API
3401 Details : Remove old rdonly API which unused from at least lustre 1.4.3
3404 Frequency : only for devices with external journals
3406 Description: Set external device read-only also
3407 Details : During a commanded failover stop, we set the disk device
3408 read-only while the server shuts down. We now also set any
3409 external journal device read-only at the same time.
3412 Frequency : when upgrading from 1.4 while trying to change parameters
3414 Description: The wrong (new) MDC name was used when setting parameters for
3415 upgraded MDT's. Also allows changing of OSC (and MDC)
3416 parameters if --writeconf is specified at tunefs upgrade time.
3419 Frequency : when setting specific ost indicies
3421 Description: QOS code breaks on skipped indicies
3422 Details : Add checks for missing OST indicies in the QOS code, so OSTs
3423 created with --index need not be sequential.
3425 Severity : enhancement
3427 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3428 Details : The uninit_groups feature works in conjunction with the kernel
3429 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3430 up the pass1 processing of e2fsck. This is a read-only feature
3431 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3432 mount filesystems that have had this feature enabled.
3434 Severity : enhancement
3436 Description: Improve multi-block allocation algorithm to avoid fragmentation
3437 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3438 allocation locality and avoid filesystem fragmentation.
3440 ------------------------------------------------------------------------------
3442 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
3444 * Support for kernels:
3445 2.4.21-47.0.1.EL (RHEL 3)
3446 2.6.5-7.283 (SLES 9)
3447 2.6.9-42.0.10.EL (RHEL 4)
3448 2.6.12.6 vanilla (kernel.org)
3449 2.6.16.27-0.9 (SLES 10)
3450 * Recommended e2fsprogs version: 1.39.cfs5
3452 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3456 Frequency : occasional, depends on client load and configuration
3457 Bugzilla : 12181, 12203
3458 Description: data loss for recently-modified files
3460 Details : In some cases it is possible that recently written or created
3461 files may not be written to disk in a timely manner (this should
3462 normally be within 30s unless client IO load is very high).
3463 The problem appears as zero-length files or files that are a
3464 multiple of 1MB in size after a client crash or client eviction
3465 that are missing data at the end of the file.
3467 This problem is more likely to be hit on clients where files are
3468 repeatedly created and unlinked in the same directory, clients
3469 have a large amount of RAM, have many CPUs, the filesystem has
3470 many OSTs, the clients are rebooted frequently, and/or the files
3471 are not accessed by other nodes after being written.
3473 The presence of the problem can be detected by looking at
3474 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
3475 smaller than the second (nr_unused) then dirty files will not
3476 be flushed automatically to disk. "sync; sleep 10" should be
3477 run several times on the node before unmounting it to update
3478 Lustre (this is also safe to run on nodes without this problem).
3480 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
3481 that can cause this same problem, so customers using that kernel
3482 also need to update the kernel in addition to Lustre. In order
3483 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
3485 It is normal that files written just before a client crash (less
3486 than 30s) may not yet have been flushed to disk, even for local
3490 Frequency : frequent on thin XT3 nodes
3492 Description: UUID collision on thin XT3 Linux nodes
3493 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
3494 randomly, since we relied on an insufficiently-seeded PRNG.
3499 Description: OSS hangs after "All ost request buffers busy"
3500 Details : A deadlock between quota and journal operations caused OSS
3501 hangs after printing "All ost request buffers busy."
3504 Frequency : always on liblustre builds
3506 Description: Cleanup compiler warnings on liblustre
3509 Frequency : always on liblustre builds on XT3
3511 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
3515 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
3516 > 32000 subdirectories
3517 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
3518 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
3519 subdirectory count crosses 32000. This will aid e2fsck to
3520 correctly handle more than 32000 subdirectories.
3523 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3525 Description: System hang on startup
3526 Details : This bug allowed the liblustre (e.g. catamount) client to
3527 return to the app before handling all startup RPCs. This
3528 could leave the node unresponsive to lustre network traffic
3529 and manifested as a server ptllnd timeout.
3531 Severity : enhancement
3533 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3534 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
3535 boolean to control whether to print peer debug info when a
3536 client's RPC times out.
3541 Description: make O_SYNC working on 2.6 kernels
3542 Details : 2.6 kernels use different method for mark pages for write,
3543 so need add a code to lustre for O_SYNC work.
3548 Description: Failure to close file and release space on NFS
3549 Details : Put inode details into lock acquired in ll_intent_file_open.
3550 Use mdc_intent_lock in ll_intent_open to properly
3551 detect all kind of errors unhandled by mdc_enqueue.
3556 Description: proc file read during shutdown sometimes raced obd removal,
3558 Details : Add lock to prevent obd access after proc file removal.
3561 Frequency : Only for files larger than 4GB on 32-bit clients.
3563 Description: improperly doing page alignment of locks
3564 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
3565 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3570 Description: RPCs being resent when they shouldn't be
3571 Details : Some RPCs that should not be resent are being resent. This
3572 can cause inconsistencies in the RPC state machine. Do not
3573 resend such requests.
3576 Frequency : rare, only with NFS export
3578 Description: Crash on NFS re-export node
3579 Details : under very unusual load conditions an assertion is hit in
3580 ll_intent_file_open()
3583 Frequency : only if OST filesystem is corrupted
3585 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3586 Details : for a short time RPCs with bulk IO are in the replay list,
3587 but replay of bulk IOs is unimplemented. If the OST filesystem
3588 is corrupted due to disk cache incoherency and then replay is
3589 started it is possible to trip an assertion. Avoid putting
3590 committed RPCs into the replay list at all to avoid this issue.
3595 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3596 Details : Large single O_DIRECT read and write calls can fail to allocate
3597 a sufficiently large buffer to process the request. In case of
3598 allocation failure the allocation is retried with a smaller
3599 buffer and broken into smaller requests.
3601 Severity : enhancement
3603 Description: Add -o localflock option to simulate old noflock behaviour.
3604 Details : This will achieve local-only flock/fcntl locks coherentness.
3609 Description: versioning check is incomplete
3610 Details : Checking the version difference of client vs. server, report
3611 error if the gap is too big.
3616 Description: add support PG_writeback bit
3617 Details : add support for PG_writeback bit for Lustre, for more carefull
3618 work with page cache in 2.6 kernel. This also fix some deadlocks
3619 and remove hack for work O_SYNC with 2.6 kernel.
3621 Severity : enhancement
3623 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3624 Details : The uninit_groups feature works in conjunction with the kernel
3625 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3626 up the pass1 processing of e2fsck. This is a read-only feature
3627 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3628 mount filesystems that have had this feature enabled.
3630 Severity : enhancement
3632 Description: Improve multi-block allocation algorithm to avoid fragmentation
3633 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3634 allocation locality and avoid filesystem fragmentation.
3636 ------------------------------------------------------------------------------
3638 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3640 * Support for kernels:
3641 2.6.9-42.0.3.EL (RHEL 4)
3642 2.6.5-7.276 (SLES 9)
3643 2.4.21-47.0.1.EL (RHEL 3)
3644 2.6.12.6 vanilla (kernel.org)
3645 2.6.16.21-0.8 (SLES10)
3646 * Recommended e2fsprogs version: 1.39.cfs2-0
3648 * The backwards-compatible /proc/sys/portals symlink has been removed
3649 in this release. Before upgrading, please ensure that you change
3650 any configuration scripts or /etc/sysctl.conf files that access
3651 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3652 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3653 in advance of the upgrade on any system running Lustre 1.4.6 or
3654 newer, since /proc/sys/lnet was added in that version.
3655 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3659 Frequency : only when quota is used
3661 Description: avoid scanning export list for quota master
3662 Details : Change the algorithms to avoid scanning export list in order
3663 to improve the efficiency.
3666 Frequency : MDS failover only, very rarely
3668 Description: "went back in time" messages on mds failover
3669 Details : The greatest transno may be lost when the current operation
3670 finishes with an error (transno==0) and the client's last_rcvd
3671 record is over-written. Save the greatest transno in the
3672 mds_last_transno for this case.
3675 Frequency : always for specific kernels and striping counts
3677 Description: client may get "Matching packet too big" without ACL support
3678 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
3679 when trying to access files in certain configurations. The
3680 clients should in fact be denied when mounting because they do
3681 not understand ACLs.
3684 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
3686 Description: many clients connecting with IO in progress causes connect timeouts
3687 Details : Avoid synchronous journal commits to avoid delays caused by many
3688 clients connecting/disconnecting when bulk IO is in progress.
3689 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
3690 OST_IO_PORTAL to avoid delays behind potentially many pending
3694 Frequency : occasionally with multiple writers to a single file
3696 Description: shared writes to file may result in wrong size reported by stat()
3697 Details : Allow growing of kms when extent lock is cancelled
3700 Frequency : always with random mmap IO to multi-striped file
3702 Description: mmap write might be lost if we are writing to a 'hole' in stripe
3703 Details : Only if the hole is at the end of OST object so that kms is too
3704 small. Fix is to increase kms accordingly in ll_nopage.
3707 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
3709 Description: writes to a missing object would leak memory on the OST
3710 Details : If there is an inconsistency between the MDS and OST filesystems,
3711 such that the MDS references an object that doesn't exist, writes
3712 to that object will leak memory due to incorrect cleanup in the
3713 error handling path, eventually running out of memory on the OST.
3718 Description: Creating too long symlink causes lustre errors
3719 Details : Check symlink and name lengths before sending requests to MDS.
3722 Frequency : only if flock is enabled (not on by default)
3724 Description: posix locks not released on fd closure on 2.6.9+
3725 Details : We failed to add posix locks to list of inode locks on 2.6.9+
3726 kernels, this caused such locks not to be released on fd close and
3727 then assertions on fs unmount about still used locks.
3730 Frequency : MDS failover only, very rarely
3732 Description: clients may get ASSERTION(granted_lock != NULL)
3733 Details : When request was taking a long time, and a client was resending
3734 a getattr by name lock request. The were multiple lock requests
3735 with the same client lock handle and
3736 mds_getattr_name->fixup_handle_for_resent_request found one of the
3737 lock handles but later failed with ASSERTION(granted_lock != NULL).
3742 Description: handle->h_buffer_credits > 0, assertion failure
3743 Details : h_buffer_credits is zero after truncate, causing assertion
3744 failure. This patch extends the transaction or creates a new
3748 Frequency : NFS re-export or patchless client
3749 Bugzilla : 11179, 10796
3750 Description: Crash on NFS re-export node (__d_move)
3751 Details : We do not want to hash the dentry if we don't have a lock.
3752 But if this dentry is later used in d_move, we'd hit uninitialised
3753 list head d_hash, so we just do this to init d_hash field but
3754 leave dentry unhashed.
3757 Frequency : NFS re-export or patchless client
3759 Description: NFS exports has problem with symbolic link
3760 Details : lustre client didn't properly install dentry when re-exported
3761 to NFS or running patchless client.
3764 Frequency : NFS re-export or patchless client
3766 Description: Various nfs/patchless fixes.
3767 Details : fixes reuse disconected alias for lookup process - this fixes
3768 warning "find_exported_dentry: npd != pd",
3769 fix permission error with open files at nfs.
3770 fix apply umask when do revalidate.
3773 Frequency : occasional
3775 Description: Crash on NFS re-export node
3776 Details : calling clear_page() on the wrong pointer triggered oops in
3777 generic_mapping_read().
3780 Frequency : rarely, using O_DIRECT IO
3782 Description: unaligned directio crashes client with LASSERT
3783 Details : check for unaligned buffers before trying any requests.
3786 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
3788 Description: stale data returned from RAID cache
3789 Details : If only a small amount of IO is done to the RAID device before
3790 reading it again it is possible to get stale data from the RAID
3791 cache instead of reading it from disk.
3794 Frequency : always for sles10 kernel
3796 Description: sles10 support
3797 Details : ll_follow_link: compile fixes and using of nd_set_link
3798 under newer kernels.
3801 Frequency : depends on arch, kernel and compiler version, always on sles10
3804 Description: recursive or deep enough symlinks cause stack overflow
3805 Details : getting rid of large stack-allocated variable in
3809 Frequency : depends on hardware
3811 Description: lustre write performance loss in the SLES10 kernel
3812 Details : the performance loss is caused by using of write barriers in the
3813 ext3 code. The SLES10 kernel turns barrier support on by
3814 default. The fix is to undo that change for ldiskfs.
3816 ------------------------------------------------------------------------------
3818 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
3820 * Support for kernels:
3821 2.6.9-42.0.3EL (RHEL 4)
3822 2.6.5-7.276 (SLES 9)
3823 2.4.21-47.0.1.EL (RHEL 3)
3824 2.6.12.6 vanilla (kernel.org)
3828 Frequency : quota enabled and large files being deleted
3830 Description: releasing more than 4GB of quota at once hangs OST
3831 Details : If a user deletes more than 4GB of files on a single OST it
3832 will cause the OST to spin in an infinite loop. Release
3833 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3838 Description: statfs data retrieved from /proc may be stale or zero
3839 Details : When reading per-device statfs data from /proc, in the
3840 {kbytes,files}_{total,free,avail} files, it may appear
3841 as zero or be out of date.
3844 Frequency : systems with MD RAID1 external journal devices
3846 Description: lconf's call to blkid is confused by RAID1 journal devices
3847 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3848 of returning all block devices that match the journal UUID.
3851 Frequency : always, for aggregate stripe size over 4GB
3853 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
3854 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
3855 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
3858 Frequency : always if "lfs find" used on a local file/directory
3860 Description: "lfs find" segfaults if used on a local file/directory
3861 Details : The case where a directory component was not specified wasn't
3862 handled correctly. Handle this properly.
3865 Frequency : always on ppc64
3867 Description: the write to an ext3 filesystem mounted with mballoc got stuck
3868 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3869 perform endianness conversion.
3872 Frequency : rarely (truncate to non-zero file size after write under load)
3873 Bugzilla : 10730, 10687
3874 Description: Files padded with zeros to next 4K multiple
3875 Details : With filesystems mounted using the "extents" option (2.6 kernels)
3876 it is possible that files that are truncated to a non-zero size
3877 immediately after being written are filled with zero bytes beyond
3878 the truncated size. No file data is lost.
3880 Severity : enhancement
3882 Description: Allow recovery/failover for liblustre clients.
3883 Details : liblustre clients were unaware of failover configurations until
3886 Severity : enhancement
3888 Description: user file locks should fail when not mounting with flock option
3889 Details : Set up an error-returning stub in ll_file_operations.lock field
3890 to prevent incorrect behaviour when client is mounted without
3891 flock option. Also, set up properly f_op->flock field for
3895 Frequency : always on ia64
3897 Description: "lfs df" loops on printing out MDS statfs information
3898 Details : The obd_ioctl_data was not initialized and in some systems
3899 this caused a failure during the ioctl that did not return
3900 an error. Initialize the struct and return an error on failure.
3903 Frequency : SLES 9 only
3905 Description: Error of copying files with lustre special EAs as root
3906 Details : Client side always return success for setxattr call for lustre
3907 special xattr (currently only "trusted.lov").
3910 Frequency : rarely on clusters with both ia64+i386 clients
3912 Description: ia64+i686 clients doing shared IO on the same file may LBUG
3913 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
3914 clients are doing concurrent writes to the same file it is
3915 possible that the ia64 clients may LASSERT because the OST
3916 extent locks are not PAGE_SIZE aligned. Ensure that grown
3917 locks are always aligned on the request boundary.
3920 Frequency : specific use, occasional
3922 Description: Overwriting in use executable truncates on-disk binary image
3923 Details : If one node attempts to overwrite an executable in use by
3924 another node, we now correctly return ETXTBSY instead of
3925 truncating the file.
3927 Severity : enhancement
3929 Description: Async OSC create to avoid the blocking unnecessarily.
3930 Details : If an OST has no remaining object, system will block on the creating
3931 when need to create a new object on this OST. Now, ways use
3932 pre-created objects when available, instead of blocking on an
3933 empty osc while others are not empty. If we must block, we block
3934 for the shortest possible period of time.
3939 Description: chmod on Lustre root is propagated to other clients
3940 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
3941 invalid by the follow_mount time.
3944 Frequency : liblustre clients only
3946 Description: Race in 'instant cancel' lock handling could lead to such locks
3947 never to be granted in case of SMP MDS
3948 Details : Do not destroy not yet granted but cbpending locks in
3952 Frequency : replay/resend of open
3954 Description: non null lock assetion failure in mds_intent_policy
3955 Details : Trying to replay/resend lockless open requests resulted in
3956 mds_open() returning 0 with no lock. Now it sets a flag if
3957 a lock is going to be returned.
3959 Severity : enhancement
3961 Description: Checksum enhancements
3962 Details : New checksum enhancements allow for resending RPCs that failed
3965 Severity : enhancement
3967 Description: Tunables on number of dirty pages in cacche
3968 Details : Allow to set limit on number of dirty pages cached.
3973 Description: client crash on unmount - lock still has references
3974 Details : In some error handling cases it was possible to leak a lock
3975 reference on a client while accessing a file. This was not
3976 harmful to the client during operation, but would cause the
3977 client to crash when the filesystem is unmounted.
3980 Frequency : specific case, rare
3982 Description: ETXTBSY on mds though file not in use by client
3983 Details : ETXTBSY is no longer incorrectly returned when attempting to
3984 chmod or chown a directory that the user previously tried to
3985 execute or a currently-executing binary.
3988 Frequency : extremely rare except on liblustre-based clients
3990 Description: Lustre space not freed when files are deleted
3991 Details : Clean up open-unlinked files after client eviction. Previously
3992 the unlink was skipped and the files remained as orphans.
3997 Description: OST failure "would be an LBUG" in waiting_locks_callback()
3998 Details : In some cases it was possible to send a blocking callback to
3999 a client doing a glimpse, even though that client didn't get
4000 a lock granted. When the glimpse lock is cancelled on the OST
4001 the freed lock is left on the waiting list and corrupted the list.
4004 Frequency : all core dumps
4006 Description: Broke core dumps to lustre
4007 Details : Negative dentry may be unhashed if parent does not have UPDATE
4008 lock, but some callers, e.g. do_coredump, expect dentry to be
4009 hashed after successful create, hash it in ll_create_it.
4011 ------------------------------------------------------------------------------
4013 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
4015 * Support for kernels:
4016 2.6.9-42.0.2.EL (RHEL 4)
4017 2.6.5-7.276 (SLES 9)
4018 2.4.21-40.EL (RHEL 3)
4019 2.6.12.6 vanilla (kernel.org)
4023 Frequency : always on RHEL 3
4025 Description: Number of open files grows over time
4026 Details : The number of open files grows over time, whether or not
4027 Lustre is started. This was due to a filp leak introduced
4028 by one of our kernel patches.
4030 ------------------------------------------------------------------------------
4032 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4034 * Support for kernels:
4035 2.6.9-42.EL (RHEL 4)
4036 2.6.5-7.276 (SLES 9)
4037 2.4.21-40.EL (RHEL 3)
4038 2.6.12.6 vanilla (kernel.org)
4043 Bugzilla : 5719, 9635, 9792, 9684
4044 Description: OST (or MDS) trips assertions in (re)connection under heavy load
4045 Details : If a server is under heavy load and cannot reply to new
4046 connection requests before the client resends the (re)connect,
4047 the connection handling code can behave badly if two service
4048 threads are concurrently handing separate (re)connections from
4049 the same client. Add better locking to the connection handling
4050 code, and ensure that only a single connection will be processed
4051 for a given client UUID, even if the lock is dropped.
4053 Severity : enhancement
4055 Description: add TCP zero-copy support to kernel
4056 Details : Add support to the kernel TCP stack to allow zero-copy bulk
4057 sends if the hardware supports scatter-gather and checksumming.
4058 This allows socklnd to do client-write and server-read more
4059 efficiently and reduce CPU utilization from skbuf copying.
4062 Frequency : only if NFS exporting from client
4064 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
4065 Details : If setting attributes on a file created under NFS that had
4066 never been opened it would be possible to oops the client
4067 if the file had no objects.
4070 Frequency : always for liblustre
4072 Description: liblustre client does MDS+OSTs setattr RPC for each write
4073 Details : When doing a write from a liblustre client, the client
4074 incorrectly issued an RPC to the MDS and each OST the file was
4075 striped over in order to update the timestamps. When writing
4076 with small chunks and many clients this could overwhelm the MDS
4077 with RPCs. In all cases it would slow down the write because
4078 these RPCs are unnecessary.
4080 Severity : enhancement
4082 Description: allow number of MDS service threads to be changed at module load
4083 Details : It is now possible to change the number of MDS service threads
4084 running. Adding "options mds mds_num_threads={N}" to the MDS's
4085 /etc/modprobe.conf will set the number of threads for the next
4086 time Lustre is restarted (assuming the "mds" module is also
4087 reloaded at that time). The default number of threads will
4088 stay the same, 32 for most systems.
4093 Description: OST crash if filesystem is unformatted or corrupt
4094 Details : If an OST is started on a device that has never been formatted
4095 or if the filesystem is corrupt and cannot even mount then the
4096 error handling cleanup routines would dereference a NULL pointer.
4101 Description: NULL pointer deref in llap_from_page.
4102 Details : get_cache_page_nowait can return a page with NULL (or otherwise
4103 incorrect) mapping if the page was truncated/reclaimed while it was
4104 searched for. Check for this condition and skip such pages when
4105 doing readahead. Introduce extra check to llap_from_page() to
4106 verify page->mapping->host is non-NULL (so page is not anonymous).
4109 Frequency : Sometimes when using sys_sendfile
4111 Description: "page not covered by a lock" warnings from ll_readpage
4112 Details : sendfile called ll_readpage without right page locks present.
4113 Now we introduced ll_file_sendfile that does necessary locking
4114 around call to generic_file_sendfile() much like we do in
4118 Frequency : with certain MDS communication failures at client mount time
4120 Description: NULL pointer deref after failed client mount
4121 Details : a client connection request may delayed by the network layer
4122 and not be sent until after the PTLRPC layer has timed out the
4123 request. If the client fails the mount immediately it will try
4124 to clean up before the network times out the request. Add a
4125 reference from the request import to the obd device and delay
4126 the cleanup until the network drops the request.
4129 Frequency : occasionally during client (re)connect
4131 Description: assertion failure during client (re)connect
4132 Details : processing a client connection request may be delayed by the
4133 client or server longer than the client connect timeout. This
4134 causes the client to resend the connection request. If the
4135 original connection request is replied in this interval, the
4136 client may trip an assertion failure in ptlrpc_connect_interpret()
4137 which thought it would be the only running connect process.
4140 Frequency : only with obd_echo servers and clients that are rebooted
4142 Description: kernel BUG accessing uninitialized data structure
4143 Details : When running an obd_echo server it did not start the ping_evictor
4144 thread, and when a client was evicted an uninitialized data
4145 structure was accessed. Start the ping_evictor in the RPC
4146 service startup instead of the OBD startup.
4148 Severity : enhancement
4149 Bugzilla : 10193 (patchless)
4150 Description: Remove dependency on various unexported kernel interfaces.
4151 Details : No longer need reparent_to_init, exit_mm, exit_files,
4152 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
4155 Frequency : rare (only users of deprecated and unsupported LDAP config)
4157 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
4158 Details : LDAP apparently contains 'lustreName' attributes instead of
4159 'name'. A simple remapping of the name is sufficient.
4162 Frequency : rare (only with non-default dump_on_timeout debug enabled)
4164 Description: waiting_locks_callback trips kernel BUG if client is evicted
4165 Details : Running with the dump_on_timeout debug flag turned on makes
4166 it possible that the waiting_locks_callback() can try to dump
4167 the Lustre kernel debug logs from an interrupt handler. Defer
4168 this log dumping to the expired_lock_main() thread.
4170 Severity : enhancement
4172 Description: Support NFS exporting on 2.6 kernels.
4173 Details : Implement non-rawops metadata methods for NFS server to use without
4174 changing NFS server code.
4177 Frequency : very rare (synthetic metadata workload only)
4179 Description: two racing renames might cause an MDS thread to deadlock
4180 Details : Running the "racer" program may cause one MDS thread to rename
4181 a file from being the source of a rename to being the target of
4182 a rename at exactly the same time that another thread is doing
4183 so, and the second thread has already enqueued these locks after
4184 doing a lookup of the target and is trying to relock them in
4185 order. Ensure that we don't try to re-lock the same resource.
4188 Frequency : only very large systems with liblustre clients
4190 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
4191 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
4192 all OSTs in parallel. This allows the eviction of stale liblustre
4193 clients to proceed much faster than if they were done in series,
4194 and also offers similar improvements for other set_info RPCs.
4199 Description: excessive CPU usage during initial read phase on client
4200 Details : During the initial read phase on a client, it would agressively
4201 retry readahead on the file, consuming too much CPU and impacting
4202 performance (since 1.4.5.8). Improve the readahead algorithm
4203 to avoid this, and also improve some other common cases (read
4204 of small files in particular, where "small" is files smaller than
4205 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
4210 Description: MDS crash when receiving packet with unknown intent.
4211 Details : Do not LBUG in unknown intent case, just return -EFAULT
4213 Severity : enhancement
4214 Bugzilla : 9293, 9385
4215 Description: MDS RPCs are serialised on client. This is unnecessary for some.
4216 Details : Do not serialize getattr (non-intent version) and statfs.
4219 Frequency : occasional, when OST network is overloaded/intermittent
4221 Description: client evicted by OST after bulk IO timeout
4222 Details : If a client sends a bulk IO request (read or write) the OST
4223 may evict the client if it is unresposive to its data GET/PUT
4224 request. This is incorrect if the network is overloaded (takes
4225 too long to transfer the RPC data) or dropped the OST GET/PUT
4226 request. There is no need to evict the client at all, since
4227 the pinger and/or lock callbacks will handle this, and the
4228 client can restart the bulk request.
4231 Frequency : Always when mmapping file with no objects
4233 Description: client crashes when mmapping file with no objects
4234 Details : Check that we actually have objects in a file before doing any
4235 operations on objects in ll_vm_open, ll_vm_close and
4241 Description: Request leak when working with deleted CWD
4242 Details : Introduce advanced request refcount tracking for requests
4243 referenced from lustre intent.
4245 Severity : Enhancement
4247 Description: Cache open file handles on client.
4248 Details : MDS now will return special lock along with openhandle, if
4249 requested and client is allowed to hold openhandle, even if unused,
4250 until such a lock is revoked. Helps NFS a lot, since NFS is opening
4251 closing files for every read/write openration.
4253 Severity : Enhancement
4255 Description: Cache open negative dentries on client when possible.
4256 Details : Guard negative dentries with UPDATE lock on parent dir, drop
4257 negative dentries on lock revocation.
4262 Description: Remounting a client read-only wasn't possible with a zconf mount
4263 Details : It wasn't possible to remount a client read-only with llmount.
4265 Severity : enhancement
4266 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
4267 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
4268 MPICH is included to add Lustre-specific ADIO interfaces.
4269 This is based closely on the UFS ADIO layer and only differs
4270 in file creation, in order to allow the OST striping to be set.
4271 This is user-contributed code and not supported by CFS.
4276 Description: extended inode attributes (immutable, append-only) work improperly
4277 when 2.4 and 2.6 kernels are used on client/server or vice versa
4278 Details : Introduce kernel-independent values for these flags.
4280 Severity : enhancement
4283 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
4284 Details : Many of the /proc/ tunables can only be tuned at a megabyte
4285 granularity. Now, Fractional MB granularity is be supported,
4286 this is very useful for low memory system.
4288 Severity : enhancement
4290 Description: Getattr by fid
4291 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
4292 locks, avoids extra getattr rpc requests to MDS, allows '/' to
4293 have locks and avoids getattr rpc requests for it on every stat.
4296 Frequency : Always, for filesystems larger than 2TB
4298 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4299 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
4300 filesystems larger than 2TB to be mounted with mballoc enabled.
4303 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
4305 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4306 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
4307 filesystems larger than 2TB is mounted then the kernel will
4308 silently corrupt the start of the filesystem. CONFIG_LBD is
4309 enabled for all CFS-supported kernels, but the possibility of
4310 this happening with a modified kernel config exists.
4312 Severity : enhancement
4314 Description: add client O_DIRECT support for 2.6 kernels
4315 Details : It is now possible to do O_DIRECT reads and writes to files
4316 in the Lustre client mountpoint on 2.6 kernel clients.
4318 Severity : enhancement
4320 Description: parallel glimpse, setattr, statfs, punch, destroy requests
4321 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
4322 parallel, not waiting for response from every OST before sending
4323 a rpc to the next OST.
4328 Description: setattr vs write race when updating file timestamps
4329 Details : Client processes that update a file timestamp into the past
4330 right after writing to the file (e.g. tar) it is possible that
4331 the updated file modification time can be reset to the current
4332 time due to a race between processing the setattr and write RPC.
4334 Severity : enhancement
4336 Description: Bring 'lfs find' closer in line with regular Linux find.
4337 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
4338 -print0 options and obtains all the needed info through the lustre
4341 Severity : enhancement
4343 Description: support up to 1024 configured devices on one node
4344 Details : change obd_dev array from statically allocated to dynamically
4345 allocated structs as they are first used to reduce memory usage
4350 Description: Flush dirty partially truncated pages during truncate
4351 Details : Immediatelly flush partially truncated pages in filter_setattr,
4352 this way we completely avoid having any pages in page cache on OST
4353 and can retire ugly workarounds during writes to flush such pages.
4358 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
4359 Details : protect inode from truncation within vfs_unlink() context
4360 just take a reference before calling vfs_unlink() and release it
4361 when parent's i_sem is free.
4366 Description: last_id value checked outside lock on OST caused LASSERT failure
4367 Details : If there were multiple MDS->OST object precreate requests in
4368 flight, it was possible that the OST's last object id was checked
4369 outside a lock and incorrectly tripped an assertion. Move checks
4370 inside locks, and discard old precreate requests.
4373 Frequency : always, if extents are used on OSTs
4375 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
4376 Details : OSTs using the extents format would not zero the high 16 bits of
4377 the index physical block number. This is not a problem for any
4378 OST filesystems smaller than 16TB, and no kernels support ext3
4379 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
4380 new/modified files) and can be fixed for existing filesystems
4381 with e2fsprogs-1.39-cfs1.
4386 Description: import connection selection may be incorrect if timer wraps
4387 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
4388 import connections to be ignored if the 32-bit jiffies counter
4389 wraps. Use a 64-bit jiffies counter.
4392 Frequency : very large clusters immediately after boot
4394 Description: LNET request buffers exhausted under heavy short-term load
4395 Details : If a large number of client requests are generated on a service
4396 that has previously never seen so many requests it is possible
4397 that the request buffer growth cannot keep up with the spike in
4398 demand. Instead of dropping incoming requests, they are held in
4399 the LND until the RPC service can accept more requests.
4402 Frequency : Sometimes during replay
4404 Description: Assertion failure in ll_local_open after replay.
4405 Details : If replay happened on an open request reply before we were able
4406 to set replay handler, reply will become not swabbed tripping the
4407 assertion in ll_local_open. Now we set the handler right after
4408 recognising of open request
4411 Frequency : very rare
4413 Description: kernel reports "badness in vsnprintf"
4414 Details : Reading from the "recovery_status" /proc file in small chunks
4415 may cause a negative length in lprocfs_obd_rd_recovery_status()
4416 call to vsnprintf() (which is otherwise harmless). Exit early
4417 if there is no more space in the output buffer.
4419 Severity : enhancement
4421 Description: clear OBD RPC statistics by writing to them
4422 Details : It is now possible to clear the OBD RPC statistics by writing
4423 to the "stats" file.
4428 Description: Client mtime is not the same on different clients after utimes
4429 Details : In some cases, the client was using the utimes() syscall on
4430 a file cached on another node. The clients now validate the
4431 ctime from the MDS + OSTs to determine which one is right.
4436 Description: Inability to activate failout mode
4437 Details : lconf script incorrectly assumed that in pythong string's numeric
4438 value is used in comparisons.
4441 Frequency : always with multiple stripes per file
4443 Description: Inefficient object allocation for mutli-stripe files
4444 Details : When selecting which OSTs to stripe files over, for files with
4445 a stripe count that divides evenly into the number of OSTs,
4446 the MDS is always picking the same starting OST for each file.
4447 Return the OST selection heuristic to the original design.
4452 Description: mount failures may take full timeout to return an error
4453 Details : Under some heavy load conditions it is possible that a
4454 failed mount can wait for the full obd_timeout interval,
4455 possibly several minutes, before reporting an error.
4456 Instead return an error as soon as the status is known.
4458 Frequency : quota enabled and large files being deleted
4460 Description: releasing more than 4GB of quota at once hangs OST
4461 Details : If a user deletes more than 4GB of files on a single OST it
4462 will cause the OST to spin in an infinite loop. Release
4463 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4468 Description: statfs data retrieved from /proc may be stale or zero
4469 Details : When reading per-device statfs data from /proc, in the
4470 {kbytes,files}_{total,free,avail} files, it may appear
4471 as zero or be out of date.
4474 Frequency : systems with MD RAID1 external journal devices
4476 Description: lconf's call to blkid is confused by RAID1 journal devices
4477 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4478 of returning all block devices that match the journal UUID.
4481 Frequency : always, for aggregate stripe size over 4GB
4483 Description: assertion fails when trying to use 4GB stripe size
4484 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
4485 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
4488 Frequency : always on ppc64
4490 Description: the first write on an ext3 filesystem with mballoc got stuck
4491 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4492 perform endianness conversion.
4494 ------------------------------------------------------------------------------
4496 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4498 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
4499 INTEROPERATE with older versions automatically. Please read the
4500 user documentation before upgrading any part of a live system.
4501 * WARNING: Lustre networking configuration changes are required with
4502 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
4505 * Support for kernels:
4506 2.6.9-22.0.2.EL (RHEL 4)
4507 2.6.5-7.244 (SLES 9)
4508 2.6.12.6 vanilla (kernel.org)
4511 Severity : enhancement
4512 Bugzilla : 7981/8208
4513 Description: Introduced Lustre Networking (LNET)
4514 Details : LNET is new networking infrastructure for Lustre, it includes
4515 a reorganized network configuration mode (see the user
4516 documentation for full details) as well as support for routing
4517 between different network fabrics. Lustre Networking Devices
4518 (LNDs) for the supported network fabrics have also been
4519 created for this new infrastructure.
4521 Severity : enhancement
4522 Description: Introduced Access control lists
4523 Details : clients can set ACLs on files and directories in order to have
4524 more fine-grained permissions than the standard Unix UGO+RWX.
4525 The MDS must be started with the "-o acl" mount option.
4527 Severity : enhancement
4528 Description: Introduced filesystem quotas
4529 Details : Administrators may now establish per-user quotas on the
4532 Severity : enhancement
4534 Description: Configuration change for the XT3
4535 The PTLLND is now used to run Lustre over Portals on the XT3
4536 The configure option(s) --with-cray-portals are no longer used.
4537 Rather --with-portals=<path-to-portals-includes> is used to
4538 enable building on the XT3. In addition to enable XT3 specific
4539 features the option --enable-cray-xt3 must be used.
4544 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
4545 Details : A race between allocating a new llcd and re-getting the llcd_lock
4546 allowed another thread to grab newly-allocated llcd.
4548 Severity : enhancement
4550 Description: 2.6 OST async journal commit and locking fix to improve performance
4551 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
4552 2.6 kernels are now async as they already were in 2.4 kernels so
4553 that they can commit concurrently with the network bulk transfer.
4554 For block-allocated files the filter allocation semaphore is held
4555 to avoid filesystem fragmentation during allocation. BKL lock
4556 removed for 2.6 xattr operations where it is no longer needed.
4561 Description: lconf incorrectly determined whether two IP networks could talk
4562 Details : In some more complicated routing and multiple-network
4563 configurations, lconf will avoid trying to make a network
4564 connection to a disjoint part of the IP space. It was doing the
4565 math incorrectly for one set of cases.
4570 Description: Fix for potential infinite loop processing records in an llog.
4571 Details : If an llog record is corrupted/zeroed, it is possible to loop
4572 forever in llog_process(). Validate the llog record length
4573 and skip the remainder of the block on error.
4576 Frequency : occasional (liblustre only)
4578 Description: liblustre could not open files whose last component is a symlink
4579 Details : sysio_path_walk() would incorrectly pass the open intent to
4580 intermediate path components.
4583 Frequency : rare (liblustre only with non-standard tuning)
4584 Bugzilla : 7201 (7350)
4585 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4586 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4587 after it has been released, only do this for referenced locks
4589 Severity : enhancement
4591 Description: specifying an (invalid) directory default stripe_size of -1
4592 would reset the directory default striping
4593 Details : stripe_size -1 was used internally to signal directory stripe
4594 removal, now use "all default" to signal dir stripe removal
4595 as a directory striping of "all default" is not useful
4598 Frequency : common for large clusters running liblustre clients
4600 Description: doing an ls when liblustre clients are running is slow
4601 Details : sending a glimpse AST to a liblustre client waits for every AST
4602 to time out, as liblustre clients will not respond. Since they
4603 cannot cache data we refresh the OST lock LVB from disk instead.
4605 Severity : enhancement
4607 Description: doing an ls at the same time as file IO can be slow
4608 Details : enqueue and other "small" requests can be blocked behind many
4609 large IO requests. Create a new OST IO portal for non-IO
4610 requests so they can be processed faster.
4613 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4615 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4616 created file with mode 000
4617 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4618 allow an ftruncate() as a non-root user to a file with mode 000.
4619 Lustre now respects this flag to disable mode checks when
4620 truncating a file owned by the user
4623 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4626 Description: Revoking locks from clients that went dead or catatonic might take
4628 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4629 cancellation of such locks instant on servers without waiting for
4630 any reply from clients. Clients drops these locks when cancel
4631 notification from server is received without replying.
4634 Frequency : liblustre-only, when liblustre client dies or becomes busy
4636 Description: Doing ls on Linux clients can take a long time with active
4638 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4639 granting such locks to it in the first place if possible. Locks
4640 are taken by proxy on the OST during the read or write and
4641 dropped immediately afterward. Add connect flags handling, do
4642 not grant locks to liblustre clients for glimpse ASTs.
4644 Severity : enhancement
4646 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4647 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4648 sequential reads and behaves badly with applications doing random
4649 reads. Improve it by reading ahead at least read region, and
4650 avoiding excessive large RPC for small reads.
4652 Severity : enhancement
4654 Description: Creating more than 1000 files for a single job may cause a load
4655 imbalance on the OSTs if there are also a large number of OSTs.
4656 Details : qos_prep_create() uses an OST index reseed value that is an
4657 even multiple of the number of available OSTs so that if the
4658 reseed happens in the middle of the object allocation it will
4659 still utilize the OSTs as uniformly as possible.
4664 Description: OST or MDS may oops in ping_evictor_main()
4665 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
4666 but doesn't restart at beginning of obd_exports_timed list
4669 Severity : enhancement
4671 Description: improve by-nid export eviction on the MDS and OST
4672 Details : allow multiple exports with the same NID to be evicted at one
4673 time without re-searching the exports list.
4676 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
4678 Description: MDS may oops in groups_free()
4679 Details : in rare race conditions a newly allocated group_info struct is
4680 freed again, and this can be NULL. The 2.4 compatibility code
4681 for groups_free() checked for a NULL pointer, but 2.6 did not.
4684 Frequency : common for liblustre clients doing little filesystem IO
4685 Bugzilla : 9352, 7313
4686 Description: server may evict liblustre clients accessing contended locks
4687 Details : if a client is granted a lock or receives a completion AST
4688 with a blocking AST already set it would not reply to the AST
4689 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
4692 Frequency : lfs setstripe, only systems with more than 160 OSTs
4694 Description: unable to set striping with a starting offset beyond OST 160
4695 Details : llapi_create_file() incorrectly limited the starting stripe
4696 index to the maximum single-file stripe count.
4699 Frequency : LDAP users only
4701 Description: lconf did not handle in-kernel recovery with LDAP properly
4702 Details : lconf/LustreDB get_refs() is searching the wrong namespace
4704 Severity : enhancement
4706 Description: bind OST threads to NUMA nodes to improve performance
4707 Details : all OST threads are uniformly bound to CPUs on a single NUMA
4708 node and do their allocations there to localize memory access
4710 Severity : enhancement
4712 Description: llmount can determine client NID directly from Myrinet (GM)
4713 Details : the client NID code from gmnalnid was moved directly into
4714 llmount, removing the need to use this or specifying the
4715 client NID explicitly when mounting GM clients with zeroconf
4718 Frequency : if client is started with down MDS
4720 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
4721 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
4722 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
4723 ptlrpc_import_delayed_req() because we didn't check if the
4724 request was interrupted, and we also didn't break out of the
4725 event loop if there was no timeout
4730 Description: data loss during non-page-aligned writes to a single file from
4731 both multiple nodes and multiple threads on one node at same time
4732 Details : updates to KMS and lsm weren't protected by common lock. Resulting
4733 inconsistency led to false short-reads, that were cached and later
4734 used by ->prepare_write() to fill in partially written page,
4735 leading to data loss.
4738 Frequency : always, if lconf --abort_recovery used
4740 Description: lconf --abort_recovery fails with 'Operation not supported'
4741 Details : lconf was attempting to abort recovery on the MDT device and not
4744 Severity : enhancement
4746 Description: remove cleanup logs
4747 Details : replace lconf-generated cleanup logs with lustre internal
4748 cleanup routines. Eliminates the need for client-cleanup and
4751 Severity : enhancement
4753 Description: add support for EAs (user and system) on lustre filesystems
4754 Details : it is now possible to store extended attributes in the Lustre
4755 client filesystem, and with the user_xattr mount option it
4756 is possible to allow users to store EAs on their files also
4758 Severity : enhancement
4760 Description: Add possibility (config option) to show minimal available OST free
4762 Details : When compiled with --enable-mindf configure option, statfs(2)
4763 (and so, df) will return least minimal free space available from
4764 all OSTs as amount of free space on FS, instead of summary of
4765 free spaces of all OSTs.
4767 Severity : enhancement
4769 Description: do not expand extent locks acquired on OST-side
4770 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
4771 by server: they are not cached anyway.
4774 Frequency : when mmap is used/binaries executed from Lustre
4776 Description: Unmmap pages before throwing them away from read cache.
4777 Details : llap_shrink cache now attempts to unmap pages before discarding
4778 them (if unmapping failed - do not discard). SLES9 kernel has
4779 extra checks that trigger if this unmapping is not done first.
4784 Description: lconf didn't resolve symlinks before checking to see whether a
4785 given mountpoint was already in use
4788 Frequency : when migrating failover services
4789 Bugzilla : 6395, 9514
4790 Description: When migrating a subset of services from a node (e.g. failback
4791 from a failover service node) the remaining services would
4792 time out and evict clients.
4793 Details : lconf --force (implied by --failover) sets the global obd_timeout
4794 to 5 seconds in order to quickly disconnect, but this caused
4795 other RPCs to time out too quickly. Do not change the global
4796 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
4798 Severity : enhancement
4799 Frequency : if MDS is started with down OST
4800 Bugzilla : 9439,5706
4801 Description: Allow startup/shutdown of an MDS without depending on the
4802 availability of the OSTs.
4803 Details : Asynchronously call mds_lov_synchronize during MDS startup.
4804 Add appropriate locking and lov-osc refcounts for safe
4805 cleaning. Add osc abort_inflight calls in case the
4806 synchronize never started.
4809 Frequency : occasional (Cray XT3 only)
4811 Description: root not authorized to access files in CRAY_PORTALS environment
4812 Details : The client process capabilities were not honoured on the MDS in
4813 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
4814 been accessed by an authorized user then root was able to access
4815 the file on the local client also. The root user capabilities
4816 are now allowed on the MDS, as this environment has secure UID.
4819 Frequency : occasional
4821 Description: ldiskfs "too long searching" message happens too often
4822 Details : A debugging message (otherwise harmless) prints too often on
4823 the OST console. This has been reduced to only happen when
4824 there are fragmentation problems on the filesystem.
4829 Description: Division by zero in statfs when all OSCs are inactive
4830 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
4831 lov_statfs divides by value returned by lov_get_stripecnt().
4835 Bugzilla : 9489, 3273
4836 Description: First write from each client to each OST was only 4kB in size,
4837 to initialize client writeback cache, which caused sub-optimal
4838 RPCs and poor layout on disk for the first written file.
4839 Details : Clients now request an initial cache grant at (re)connect time
4840 and so that they can start streaming writes to the cache right
4841 away and always do full-sized RPCs if there is enough data.
4842 If the OST is rebooted the client also re-establishes its grant
4843 so that client cached writes will be honoured under the grant.
4848 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
4849 Details : Now I/O RPCs go to different portal number and (presumably) fast
4850 lock requests (and glimses) and other RPCs get their own service
4851 threads pool that should be able to service those RPCs
4854 Severity : enhancement
4856 Description: Ability to exchange lustre version between client and servers and
4857 issue warnings at client side if client is too old. Also for
4858 liblustre clients there is ability to refuse connection of too old
4860 Details : New 'version' field is added to connect data structure that is
4861 filled with version info. That info is later checked by server and
4865 Frequency : rare, liblustre only.
4866 Bugzilla : 9296, 9581
4867 Description: Two simultaneous writes from liblustre at offset within same page
4868 might proceed at the same time overwriting eachother with stale
4870 Details : I/O lock withing llu_file_prwv was released too early, before data
4871 actually was hitting the wire. Extended lock-holding time until
4872 server acknowledges receiving data.
4875 Frequency : extremely rare. Never observed in practice.
4877 Description: avoid generating lustre_handle cookie of 0.
4878 Details : class_handle_hash() generates handle cookies by incrementing
4879 global counter, and can hit 0 occasionaly (this is unlikely, but
4880 not impossible, because initial value of cookie counter is
4881 selected randonly). Value of 0 is used as a sentinel meaning
4882 "unassigned handle" --- avoid it. Also coalesce two critical
4883 sections in this function into one.
4885 Severity : enhancement
4887 Description: allow liblustre clients to delegate truncate locking to OST
4888 Details : To avoid overhead of locking, liblustre client instructs OST to
4889 take extent lock in ost_punch() on client's behalf. New connection
4890 flag is added to handle backward compatibility.
4892 Severity : enhancement
4893 Bugzilla : 4928, 7341, 9758
4894 Description: allow number of OST service threads to be specified
4895 Details : a module parameter allows the number of OST service threads
4896 to be specified via "options ost ost_num_threads={N}" in the
4897 OSS's /etc/modules.conf or /etc/modprobe.conf.
4901 Bugzilla : 6146, 9635, 9895
4902 Description: servers crash with bad pointer in target_handle_connect()
4903 Details : In rare cases when a client is reconnecting it was possible that
4904 the connection request was the last reference for that export.
4905 We would temporarily drop the export reference and get a new
4906 one, but this may have been the last reference and the export
4907 was just destroyed. Get new reference before dropping old one.
4909 Severity : enhancement
4910 Frequency : if client is started with failover MDS
4912 Description: Allow multiple MDS hostnames in the mount command
4913 Details : Try to read the configuration from all specified MDS
4914 hostnames during a client mount in case the "primary"
4917 Severity : enhancement
4919 Description: Stop sending data to evicted clients as soon as possible.
4920 Details : Check if the client we are about to send or are sending data to
4921 was evicted already. (Check is done every second of waiting,
4922 for which l_wait_event interface was extended to allow checking
4923 of exit condition at specified intervals).
4926 Frequency : rare, normally only when NFS exporting is done from client
4928 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
4930 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
4933 Frequency : always, if srand() is called before liblustre initialization
4935 Description: Liblustre uses system PRNG disturbing its usage by user application
4936 Details : Introduce internal to lustre fast and high-quality PRNG for
4937 lustre usage and make liblustre and some other places in generic
4938 lustre code to use it.
4940 Severity : enhancement
4941 Bugzilla : 9477, 9557, 9870
4942 Description: Verify that the MDS configuration logs are updated when xml is
4943 Details : Check if the .xml configuration logs are newer than the config
4944 logs stored on the MDS and report an error if this is the case.
4945 Request --write-conf, or allow starting with --old_conf.
4947 Severity : enhancement
4949 Description: Handle symlinks in the path when checking if Lustre is mounted.
4950 Details : Resolve intermediate symlinks when checking if a client has
4951 mounted a filesystem to avoid duplicate client mounts.
4956 Description: lconf can hit an error exception but still return success.
4957 Details : The lconf command catches the Command error exception at the top
4958 level script context and will exit with the associated exit
4959 status, but doesn't ensure that this exit status is non-zero.
4964 Description: failure of ptlrpc thread startup can cause oops
4965 Details : Starting a ptlrpc service thread can fail if there are a large
4966 number of threads or the server memory is very fragmented.
4967 Handle this without oopsing.
4970 Frequency : always, only if liblustre and non-default acceptor port was used
4972 Description: liblustre cannot connect to servers with non-default acceptor port
4973 Details : tcpnal_set_default_params() was not called and was therefore
4974 ignoring the environment varaible TCPNAL_PORT, as well as other
4975 TCPNAL_ environment variables
4980 Description: two objects could be created on the same OST for a single file
4981 Details : If an OST is down, in some cases it was possible to create two
4982 objects on a single OST for a single file. No problems other
4983 than potential performance impact and spurious error messages.
4987 Bugzilla : 5681, 9562
4988 Description: Client may oops in ll_unhash_aliases
4989 Details : Client dcache may become inconsistent in race condition.
4990 In some cases "getcwd" can fail if the current directory is
4996 Description: Inode refcounting problems in NFS export code
4997 Details : link_raw functions used to call d_instantiate without obtaining
4998 extra inode reference first.
5002 Bugzilla : 9942, 9903
5003 Description: Referencing freed requests leading to crash, memleaks with NFS.
5004 Details : We used to require that call to ll_revalidate_it was always
5005 followed by ll_lookup_it. Also with revalidate_special() it is
5006 possible to call ll_revalidate_it() twice for the same dentry
5007 even if first occurence returned success. This fix changes semantic
5008 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
5009 reference on a request referred from the intent.
5010 ll_intent_release() then releases such a request.
5013 Frequency : rare, normally benchmark loads only
5015 Description: unlinked inodes were kept in memory on the client
5016 Details : If a client is repeatedly creating and unlinking files it
5017 can accumulate a lot of stale inodes in the inode slab cache.
5018 If there is no other client load running this can cause the
5019 client node to run out of memory. Instead flush old inodes
5020 from client cache that have the same inode number as a new inode.
5023 Frequency : SLES9 2.6.5 kernel and long filenames only
5024 Bugzilla : 9969, 10379
5025 Description: utime reports stale NFS file handle
5026 Details : SLES9 uses out-of-dentry names in some cases, which confused
5027 the lustre dentry revalidation. Change it to always use the
5031 Frequency : rare, unless heavy write-truncate concurrency is continuous
5032 Bugzilla : 4180, 6984, 7171, 9963, 9331
5033 Description: OST becomes very slow and/or deadlocked during object unlink
5034 Details : filter_destroy() was holding onto the parent directory lock
5035 while truncating+unlinking objects. For very large objects this
5036 may block other threads for a long time and slow overall OST
5037 responsiveness. It may also be possible to get a lock ordering
5038 deadlock in this case, or run out of journal credits because of
5039 the combined truncate+unlink. Solution is to do object truncate
5040 first in one transaction without parent lock, and then do the
5041 final unlink in a new transaction with the parent lock. This
5042 reduces the lock hold time dramatically.
5045 Frequency : rare, 2.4 kernels only
5047 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
5048 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
5049 before unmounting their filesystem in order to avoid deadlock.
5050 The kernel_locked() function in 2.4 kernels only checks whether
5051 the kernel lock is held, not whether it is this process that is
5052 holding it as 2.6 kernels do.
5057 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
5058 Details : The client ptlrpc code may be trying to reconnect to a down
5059 server before a previous connection attempt has timed out.
5060 Increase the reconnect interval to be longer than the connection
5061 timeout interval to avoid sending duplicate connections to
5065 Frequency : echo_client brw_test command
5067 Description: fix echo_client to work with OST preallocated code
5068 Details : OST preallocation code (5137) didn't take echo_client IO path
5069 into account: echo_client calls filter methods outside of any
5070 OST thread and, hence, there is no per-thread preallocated
5071 pages and buffers to use. Solution: hijack pga pages for IO. As
5072 a byproduct, this avoids unnecessary data copying.
5076 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
5077 Description: Client can oops in mdc_commit_close() after open replay
5078 Details : It was possible for the MDS to return an open request with no
5079 transaction number in mds_finish_transno() if the client was
5080 evicted, but without actually returning an error. Clients
5081 would later try to replay that open and may trip an assertion
5082 Simplify the client close codepath, and always return an error
5083 from the MDS in case the open is not successful.
5086 Frequency : rare, 2.6 OSTs only
5088 Description: OST may deadlock under high load on fragmented files
5089 Details : If there was a heavy load and highly-fragmented OST filesystems
5090 it was possible to have all the OST threads deadlock waiting on
5091 allocation of biovecs, because the biovecs were not released
5092 until the entire RPC IO was completed. Instead, release biovecs
5093 as soon as they are complete to ensure forward IO progress.
5095 Severity : enhancement
5097 Description: Support for specifying external journal device at mount
5098 Details : If an OST or MDS device is formatted with an external journal
5099 device, this device major/minor is stored in the ext3 superblock
5100 and may not be valid for failover. Allow detecting and
5101 specifying the external journal at mount time.
5106 Description: Mounting an MDS with pending unlinked files may cause oops
5107 Details : target_finish_recovery() calls mds_postrecov() which returned
5108 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
5109 considers this an error and immediately begins cleaning up the
5110 lov, just after starting the mds_lov process
5112 Severity : enhancement
5114 Description: Implement 'lfs df' to report actual free space on per-OST basis
5115 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
5116 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
5117 usage of objects; '-h' to report in human readable format.
5119 ------------------------------------------------------------------------------
5121 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
5128 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
5129 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
5130 at startup during orphan unlinking. Instead, skip buddy update
5131 before setup, as it will be regenerated after recovery is complete.
5136 Description: If an OST is inactive, its locks might reference stale inodes.
5137 Details : lov_change_cbdata() must iterate over all namespaces, even if
5138 they are inactive to clear inode references from the lock.
5140 Severity : enhancement
5141 Frequency : occasional, if non-standard max_dirty_mb used
5143 Description: Client will block write RPCs if not enough grant
5144 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
5145 then the client will block writes while waiting for another RPC
5146 to complete instead of consuming its dirty limit. With change
5147 we get improved performance when max_dirty_mb is small.
5149 Severity : enhancement
5150 Bugzilla : 3389, 6253
5151 Description: Add support for supplementary groups on the MDS.
5152 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
5153 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
5154 lookups for user supplementary groups into a cache.
5158 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
5159 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
5160 for liblustre O_EXCL is enforced by MDS.
5164 Description: readdir never returns NULL in liblustre.
5165 Details : Corrected llu_iop_getdirentries logic, to return offset of next
5166 dentry in struct dirent.
5170 Frequency : liblustre only, depends on application IO pattern
5171 Description: liblustre clients evicted if not contacting servers
5172 Details : Don't put liblustre clients into the ping_evictor list, so
5173 they will not be evicted by the pinger ever.
5175 Severity : enhancement
5177 Description: Add ability to evict clients by NID from MDS.
5178 Details : By echoing "nid:$NID" string into
5179 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
5180 $NID would be instantly evicted from this MDS and from all active
5181 OSTs connected to it.
5185 Description: Do not query file size twice, somewhat slowing stat(2) calls.
5186 Details : lookup_it_finish() used to query file size from OSTs that was not
5191 Description: service threads change working directory to that of init
5192 Details : Starting lustre service threads may pin the working directory
5193 of the parent thread, making that filesystem busy. Threads
5194 now change to the working directory of init to avoid this.
5198 Frequency : during shutdown only
5199 Description: shutdown with a failed MDS or OST can cause unmount to hang
5200 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
5205 Frequency : 2.6 only
5206 Description: chmod/chown may include an extra supplementary group
5207 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
5208 supplementary group and if none is specified this is used.
5211 Bugzilla : 5479 (6816)
5213 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
5214 Details : If lookup is in progress on a file that is unlinked we might try
5215 to revalidate the inode and fail in revalidate after lookup is
5216 complete and ll_file_open() enqueues the open again but
5217 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
5220 Frequency : always, if lconf --abort_recovery used
5222 Description: lconf --abort_recovery fails with 'Operation not supported'
5223 Details : lconf was attempting to abort recovery on the MDT device and not
5226 ------------------------------------------------------------------------------
5228 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
5233 Frequency : rare (only unsupported configurations with a node running as an
5235 Bugzilla : 6514, 5137
5236 Description: Mounting a Lustre file system on a node running as an OST could
5238 Details : OSTs now preallocates memory needed to write out data at
5239 startup, instead of when needed, to avoid having to
5240 allocate memory in possibly low memory situations.
5241 Specifically, if the file system is mounted on on OST,
5242 memory pressure could force it to try to write out data,
5243 which it needed to allocate memory to do. Due to the low
5244 memory, it would be unable to do so and the node would
5245 become unresponsive.
5247 Severity : enhancement
5249 Description: Addition of lconf --service command line option
5250 Details : lconf now accepts a '--service <arg>' option, which is
5251 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
5253 Severity : enhancement
5255 Description: Failover mode is now the default for OSTs.
5256 Details : By default, OSTs will now run in failover mode. To return to
5257 the old behaviour, add '--failout' to the lmc line for OSTs.
5259 Severity : enhancement
5261 Description: Health checks are now provided for MDS and OSTs
5262 Details : Additional detailed health check information on MSD and OSTs
5263 is now provided through the procfs health_check value.
5266 Frequency : occasional, depends on IO load
5268 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
5269 after numerous create/delete cycles
5270 Details : The ext3 inode allocation policy would not allocate new inodes
5271 very well on the OSTs because there are no new directories
5272 being created. Instead we look for groups with free space if
5273 the parent directories are nearly full.
5278 Description: Network or server problems during mount may cause partially
5279 mounted clients instead of returning an error.
5280 Details : The config llog parsing code may overwrite the error return
5281 code during mount error handling, returning success instead
5287 Description: MDS can fail to allocate large reply buffers
5288 Details : After long uptimes the MDS can fail to allocate large reply
5289 buffers (e.g. zconf client mount config records) due to memory
5290 fragmentation or consumption by the buffer cache. Preallocate
5291 some large reply buffers so that these replies can be sent even
5292 under memory pressure.
5296 Frequency : rare (liblustre)
5297 Description: fsx running with liblustre complained that using truncate() to
5298 extend the file doesn't work. This patch corrects that issue.
5299 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
5300 ATTR_SIZE and lsm use in llu_setattr_raw.
5304 Frequency : rare, only 2.6 kernels
5305 Description: Unusual file access patterns on the MDS may result in inode
5306 data being lost in very rare circumstances.
5307 Details : Bad interaction between the ea-in-inode patch and the "no-read"
5308 code in the 2.6 kernel caused the inode and/or EA data not to
5309 be read from disk, causing single-file corruption.
5313 Frequency : rare, only 2.6 filesystems using extents
5314 Description: Heavy concurrent write and delete load may cause data corruption.
5315 Details : It was possible under high-load situations to have an extent
5316 metadata block in the block device cache from a just-unlinked
5317 file overwrite a newly-allocated data block. We now unmap any
5318 metadata buffers that alias just-allocated data blocks.
5322 Frequency : filesystems with default stripe_count larger than 77
5323 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
5324 Details : lconf specifies an inode size of 4096 bytes when the default
5325 stripe_count is larger than 77. This conflicts with the default
5326 inode density of 1 per 4096 bytes. Allocate smaller inodes in
5327 this case to avoid pinning too much memory for large EAs.
5329 ------------------------------------------------------------------------------
5331 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
5336 Frequency : rare (extremely heavy IO load with hundreds of clients)
5338 Description: Client is evicted, gets IO error writing to file
5339 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
5340 caused clients to be evicted for AST timeouts. The fixes in
5341 bug 5192 mean we no longer need to have such short AST timeouts
5342 so ldlm_timeout has been increased.
5345 Frequency : occasional during --force or --failover shutdown under load
5346 Bugzilla : 5949, 4834
5347 Description: Server oops/LBUG if stopped with --force or --failover under load
5348 Details : a collection of import/export refcount and cleanup ordering
5349 issues fixed for safer force cleanup
5352 Frequency : only filesystems larger than 120 OSTs
5353 Bugzilla : 5990, 6223
5354 Description: lfs getstripe would oops on a very large filesystem
5355 Details : lov_getconfig used kfree on vmalloc'd memory
5358 Frequency : only filesystems exporting via NFS to Solaris 10 clients
5359 Bugzilla : 6242, 6243
5360 Description: reading from files that had been truncated to a non-zero size
5361 but never opened returned no data
5362 Details : ll_file_read() reads zeros from no-object files to EOF
5367 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
5368 Details : There's little harm in aborting MDS/OSS recovery and letting it
5369 try again, so I removed the LASSERT and return an error instead.
5371 Severity : enhancement
5373 Description: New debugging infrastructure for tracking down data corruption
5374 Details : The I/O checksum code was replaced to: (a) control it at runtime,
5375 (b) cover more of the client-side code path, and (c) try to narrow
5376 down where problems occurred
5380 Bugzilla : 3819, 4364, 4397, 6313
5381 Description: Racing close and eviction MDS could cause assertion in mds_close
5382 Details : It was possible to get multiple mfd references during close and
5383 client eviction, leading to one thread referencing a freed mfd.
5385 Severity: : enhancement
5386 Bugzilla : 3262, 6359
5387 Description: Attempts to reconnect to servers are now more aggressive.
5388 Details : This builds on the enhanced upcall-less recovery that was added
5389 in 1.4.2. When trying to reconnect to servers, clients will
5390 now try each server in the failover group every 10 seconds. By
5391 default, clients would previously try one server every 25 seconds.
5396 Description: After recovery, certain operations trigger a failed
5397 assertion on a client.
5398 Details : Failing over an mds, using lconf -d --failover, while a
5399 client was doing a readdir() call would cause the client to
5400 LBUG after recovery completed and the readdir() was resent.
5402 Severity : enhancement
5404 Description: Default groups are now added by lconf
5405 Details : You can now run lconf --group <servicename> without having to
5406 manually add groups with lmc.
5409 Frequency : occasional
5411 Description: Nodes with an elan id of 0 trigger a failed assertion
5414 Frequency : always when accessing e.g. tty/console device nodes
5416 Description: tty and some other devices nodes cannot be used on lustre
5417 Details : file's private_data field is used by device data and lustre
5418 values in there got lost. New field was added to struct file to
5419 store fs-specific private data.
5422 Frequency : when exporting Lustre via NFS
5424 Description: NFSD failed occasionally when looking up a path component
5425 Details : NFSD is looking up ".." which was broken in ext3 directories
5426 that had grown large enough to become hashed.
5429 Frequency : Clusters with multiple interfaces not on the same subnet
5431 Description: Nodes will repeatedly try to reconnect to an interface which it
5432 cannot reach and report an error to the log.
5433 Details : Extra peer list entries will be created by lconf with some peers
5434 unreachable. lconf now validates the peer before adding it.
5437 Frequency : Only if a default stripe is set on the filesystem root.
5439 Description: Setting a default stripe on the filesystem root prevented the
5440 filesystem from being remounted.
5441 Details : The client was sending extra request flags in the root getattr
5442 request and did not allocate a reply buffer for the dir EA.
5445 Frequency : occasional, higher if lots of files are accessed by one client
5446 Bugzilla : 6159, 6097
5447 Description: Client trips assertion regarding lsm mismatch/magic
5448 Details : While revalidating inodes the VFS looks up inodes with ifind()
5449 and in rare cases can find an inode that is being freed.
5450 The ll_test_inode() code will free the lsm during ifind()
5451 when it finds an existing inode and then the VFS later attaches
5452 this free lsm to a new inode.
5456 Bugzilla : 6422, 7030
5457 Description: MDS deadlock between mkdir and client eviction
5458 Details : Creating a new file via mkdir or mknod (starting a transaction
5459 and getting the ns lock) can deadlock with client eviction
5460 (gets ns lock and trying to finish a synchronous transaction).
5463 Frequency : occasional
5464 Description: While starting a server, the fsfilt_ext3 module could not be
5466 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
5467 kernels. Previously, lconf would still use the ext3 name
5468 when trying to load modules. Now, it will correctly use
5469 ext3 on 2.4 and ldiskfs on 2.6.
5471 Severity : enhancement
5472 Description: The default stripe count has been changed to 1
5473 Details : The interpretation of the default stripe count (0, to lfs
5474 or lmc) has been changed to mean striping across a single
5475 OST, rather than all available. For general usage we have
5476 found a stripe count of 1 or 2 works best.
5478 Severity : enhancement
5479 Description: Add support for compiling against Cray portals.
5480 Details : Conditional compiling for some areas that are different
5484 Frequency : occasional
5485 Bugzilla : 6409, 6834
5486 Description: Creating files with an explicit stripe count may lead to
5487 a failed assertion on the MDS
5488 Details : If some OSTs are full or unavailable, creating files may
5489 trigger a failed assertion on the MDS. Now, Lustre will
5490 try to use other servers or return an error to the
5494 Frequency : occasional
5496 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
5497 caused lock timeout during readahead (since 1.4.2).
5498 Details : Processes doing readahead might match a lock that hasn't been
5499 granted yet if there are overlapping and conflicting lock
5500 requests. The readahead process waits on ungranted lock
5501 (original lock is CBPENDING), while OST waits for that process
5502 to cancel CBPENDING read lock and eventually evicts client.
5504 Severity : enhancement
5506 Description: Initial enabling of flock support for clients
5507 Details : Implements fcntl advisory locking and file status functions.
5508 This feature is provided as an optional mount flag (default
5509 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
5510 locking are implemented yet, and those that are are not guaranteed
5511 to be completely correct in production environments.
5512 mount -t lustre -o [flock|noflock] ...
5515 Frequency : occasional
5517 Description: OSTs running 2.4 kernels but with extents enabled might trip an
5518 assertion in the ext3 JBD (journaling) layer.
5519 Details : The b_committed_data struct is protected by the big kernel lock
5520 in 2.4 kernels, serializing journal_commit_transaction() and
5521 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
5522 access to this struct. In 2.6 kernels there is finer grained
5523 locking to improve SMP performance of the JBD layer.
5527 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
5532 Description: merge and cleanup kernel patches.
5535 -----------------------------------------------------------------------------
5537 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
5539 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
5540 versions of Lustre. Please update all servers and clients to
5541 version 1.4.2 or later at the same time. You must also run
5542 "lconf --write-conf {config}.xml" on the MDS while it is stopped
5543 to update the configuration logs.
5545 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
5546 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
5547 - swab llog records of type '0' so we get proper header size/idx (5861)
5548 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
5549 - fix rename of one directory over another leaking an inode (5953)
5550 - avoid SetPageDirty on 2.6 (5981)
5551 - don't re-add just-being-destroyed locks to the waiting list (5653)
5552 - when creating new directories, inherit the parent's custom
5553 striping settings if present parent (3048)
5554 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
5555 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
5556 - don't hold client locks on temporary worklist from l_lru (5666)
5557 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
5558 - automatically evict dead clients (5921)
5559 - Update file size properly in create+truncate+fstat case (6196)
5560 - Do not unhash mountpoint dentries, do not allow removal of
5562 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
5563 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
5564 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
5565 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
5566 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
5567 - Correctly handle reads of files with no objects (6243)
5568 - lctl recover will also mark a device active if deactivate used (5933)
5570 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
5571 - allow --write-conf on an MDS with different nettype than client (5619)
5572 - don't write config llogs to MDS for mounts not from that MDS (5617)
5573 - lconf should create multiple TCP connections from a client (5201)
5574 - init scripts are now turned off by default; run chkconfig --on
5575 lustre and chkconfig --on lustrefs to use them
5576 - upcalls are no longer needed for clients to recover to failover
5578 - add --abort-recovery option to lconf to abort recovery on device
5580 - add support for an arbitrary number of OSTs (3026)
5581 - Quota support protocol changes.
5582 - forward compatibility changes to wire structs (6007)
5583 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5584 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5585 - improved "lustre status" script
5586 - initialize blocksize for non-regular files (6062)
5587 - added --disable-server and --disable-client configure options (5782)
5588 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5589 - Vanilla 2.4.29 support
5590 - increase maximum number of obd devices to 520 (6242)
5591 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5592 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5593 SLES 9 2.6.5 kernels
5594 - limit stripes per file to 160 (the maximum EA size) (6093)
5596 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5599 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5600 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5601 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5602 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5603 - small ext3 extents cleanups and fixes (5733)
5604 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5605 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5606 - avoid needless client->OST connect, fix handle mismatch (5317)
5607 - fix DLM error path that led to out-of-sync client, long delays (5779)
5608 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5609 - fix several locking issues related to i_size (5492,5624,5654,5672)
5610 - don't move pending lock onto export if it is already evicted (5683)
5611 - fix kernel oops when creating .foo in unlinked directory (5548)
5612 - fix deadlock in obdfilter statistics vs. object create (5811)
5613 - use time_{before,after} to avoid timer jiffies wrap (5882)
5614 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5615 - Do not leak request if server was not able to process it (5154)
5616 - If mds_open unable to find parent dir, make that negative lookup(5154)
5617 - don't create new directories with extent-mapping (5909, 5936)
5619 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5620 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5621 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5622 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5623 - support for RHEL4 kernel on i686 (5773)
5624 - provide error messages when incompatible logs are encountered (5898)
5626 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5627 * version 1.4.0.10 (1.4.1 release candidate 1)
5629 - don't keep a lock reference when lock is not granted (4238)
5630 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5631 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5632 - drop import inflight refcount on signal_completed_replay error (5255)
5633 - unlock page after async write error during send (3677)
5634 - handle missing objects in filter_preprw_read properly (5265)
5635 - no transno return for symlink open, don't save no-trasno open (3440)
5636 - don't try to complete elan receive that already failed (4012)
5637 - free RPC server reply state on error (5406)
5638 - clean up thread from ptlrpc_start_thread() on error (5160)
5639 - readahead could read extra page into cache that wasn't ejected (5388)
5640 - prevent races in class_attach/setup/cleanup/detach (5260)
5641 - don't dereference de->d_inode after l_dput of de (5458)
5642 - use "int" for stripe value returned from lock_to_stripe (5544)
5643 - mballoc allocation and error-checking fixes in 2.6 (5504)
5644 - block device patches to fix I/O request sizes in 2.6 (5482)
5645 - look up hostnames for IB nals (5602)
5646 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5647 - don't start multiple acceptors for the same port (5277)
5648 - fix incorrect LASSERT in mds_getattr_name (5635)
5649 - export a proc file for general "ping" checking (5628)
5650 - fix "lfs check" to not block when the MDS is down (5628)
5652 - service request history (4965)
5653 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5654 - create an "evict_client" /proc entry on OSTs, like the MDS has
5655 - fix mount usage message, return errors per mount(8) (5168)
5656 - change grep [] to grep "[]" in tests so they work in more UMLs
5657 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5658 - remove ancient LOV_MAGIC_V0 EA support (5047)
5659 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5660 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5661 - allow readahead allocations to fail when low on memory (5383)
5662 - mmap locking landed again, after considerable improvement (2828)
5663 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
5665 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
5668 - send OST transaction number in read/write reply to free req (4966)
5669 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5670 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5671 - the watchdog thread now runs as interruptible (5246)
5672 - flock/lockf fixes (but it's still disabled, pending 5135)
5673 - don't use EXT3 constants in llite code (5094)
5674 - memory shortage at startup could cause assertion (5176)
5676 - reorganization of lov code
5677 - single portals codebase
5679 - add extents/mballoc support (5025)
5680 - direct I/O reads in the obdfilter (4048)
5681 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
5683 tbd Cluster File Systems, Inc. <info@clusterfs.com>
5686 - send OST transaction number in read/write reply to free req (4966)
5687 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5688 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5689 - the watchdog thread now runs as interruptible (5246)
5690 - handle missing objects in filter_preprw_read properly (5265)
5691 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5692 - drop import inflight refcount on signal_completed_replay error (5255)
5693 - unlock page after async write error during send (3677)
5694 - return original error code on reconstructed replies (3761)
5695 - no transno return for symlink open, don't save no-trasno open (3440)
5697 - add pid to ldlm debugging output (4922)
5698 - bump the watchdog timeouts -- we can't handle 30sec yet
5699 - extra debugging for orphan dentry/inode bug (5259)
5701 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
5704 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
5705 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
5706 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
5707 - fix NULL dereference in /proc/sys/portals/routes (4827)
5708 - allow failed mdc_close() operations to be interrupted (4561)
5709 - stop precreate on OST before MDS would time out on it (4778)
5710 - don't send partial-page writes before EOF from client (4410)
5711 - discard client grant for sub-page writes on large-page clients (4520)
5712 - don't free dentries not owned by NFS code, check generation (4806)
5713 - fix lsm leak if mds_create_objects() fails (4801)
5714 - limit debug_daemon file size, always print CERROR messages (4789)
5715 - use transno after validating reply (3892)
5716 - process timed out requests if import state changes (3754)
5717 - update mtime on OST during writes, return in glimpse (4829)
5718 - add mkfsoptions to LDAP (4679)
5719 - use ->max_readahead method instead of zapping global ra (5039)
5720 - don't interrupt __l_wait_event() during strace
5722 - add software watchdogs to catch hung threads quickly (4941)
5723 - make lustrefs init script start after nfs is mounted
5724 - fix CWARN/ERROR duplication (4930)
5725 - return async write errors to application if possible (2248)
5726 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
5727 - print NAL number in %x format (4645)
5728 - update barely-supported suse-2.4.21-171 series (4842)
5729 - support for sles 9 %post scripts
5730 - support for building 2.6 kernel-source packages
5731 - support for sles km_* packages
5733 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
5736 - ignore -ENOENT errors in osc_destroy (3639)
5737 - notify osc create thread that OSC is being cleaned up (4600)
5738 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
5739 - reconstruct ost_handle() like mds_handle() (4657)
5740 - create a new thread to do import eviction to avoid deadlock (3969)
5741 - let lconf resolve symlinked-to devices (4629)
5742 - don't unlink "objects" from directory with default EA (4554)
5743 - hold socknal file ref over connect in case target is down (4394)
5744 - allow more than 32000 subdirectories in a single directory (3244)
5745 - fix blocks count for O_DIRECT writes (3751)
5746 - OST returns ENOSPC from object create when no space left (4539)
5747 - don't send truncate RPC if file size isn't changing (4410)
5748 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
5749 - bind to privileged port in socknal and tcpnal (3689)
5751 - rate limit CERROR/CWARN console message to avoid overload (4519)
5752 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
5753 - basic mmap support (3918)
5754 - kernel patch series update from b1_4 (4711)
5756 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
5759 - avoid crash during MDS cleanup with OST shut down (2775)
5760 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
5761 - don't use bad inodes on the MDS (3744)
5762 - dynamic object preallocation to improve recovery speed (4236)
5763 - don't hold spinlock over lock dumping or change debug flags (4401)
5764 - don't zero obd_dev when it is force cleaned (3651)
5765 - print grants to console if they go negative (4431)
5766 - "lctl deactivate" will stop automatic recovery attempts (3406)
5767 - look for existing locks in ldlm_handle_enqueue() (3764)
5768 - don't resolve lock handle twice in recovery avoiding race (4401)
5769 - revalidate should check working dir is a directory (4134)
5771 - don't always mark "slow" obdfilter messages as errors (4418)
5773 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
5776 - don't close LustreDB during write_conf until it is done (3860)
5777 - fix typo in lconf for_each_profile (3821)
5778 - allow dumping logs from multiple threads at one time (3820)
5779 - don't allow multiple threads in OSC recovery (3812)
5780 - fix debug_size parameters (3864)
5781 - fix mds_postrecov to initialize import for llog ctxt (3121)
5782 - replace config semaphore with spinlock (3306)
5783 - be sure to send a reply for a CANCEL rpc with bad export (3863)
5784 - don't allow enqueue to complete on a destroyed export (3822)
5785 - down write_lock before checking llog header bitmap (3825)
5786 - recover from lock replay timeout (3764)
5787 - up llog sem before sending rpc (3652)
5788 - reduce ns lock hold times when setting kms (3267)
5789 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
5790 - fix NULL deref and obd_dev leak on setup error (3312)
5791 - replace some LBUG about llog ops with error handling (3841)
5792 - don't match INVALID dentries from d_lookup and spin (3784)
5793 - hold dcache_lock while marking dentries INVALID and hashing (4255)
5794 - fix invalid assertion in ptlrpc_set_wait (3880)
5796 - add libwrap support for the TCP acceptor (3996)
5797 - add /proc/sys/portals/routes for non-root route listing (3994)
5798 - allow setting MDS UUID in .xml (2580)
5799 - print the stack of a process that LBUGs (4228)
5801 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
5804 - don't cleanup request in ll_file_open() on failed MDS open (3430)
5805 - make sure to unset replay flag from failed open requests (3440)
5806 - if default stripe count is 0, use OST count for inode size (3636)
5807 - update parent mtime/ctime on client for create/unlink (2611)
5808 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
5809 - free recovery state on server during a forced cleanup (3571)
5810 - unregister_reply for resent reqs (3063)
5811 - loop back devices mounting and status check on 2.6 (3563)
5812 - fix resource-creation race that can provoke i_size == 0 (3513)
5813 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
5814 - more debugging for page-accounting assertion (3746)
5815 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
5816 - avoid deadlock after precreation failure (3758)
5817 - fix race and lock order deadlock in orphan handling (3450, 3750)
5818 - add validity checks when grabbing inodes from l_ast_data (3599)
5820 - add /proc/.../recovery_status to obdfilter (3428)
5821 - lightweight CDEBUG infrastructure, debug daemon (3668)
5822 - change default OSC RPC parameters to be better on small clusters
5823 - turn off OST read cache for files smaller than 32MB
5824 - install man pages and include them in rpms (3100)
5825 - add new init script for (un)mounting lustre filesystems (2593)
5826 - run chkconfig in %post for init scripts (3701)
5827 - drop scimac NAL (unmaintained)
5829 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
5832 - clean kiobufs before and after use (3485)
5833 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
5834 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
5835 - change init script to fail more clearly if not run as root (1528)
5836 - allow clients to reconnect during replay (1742)
5837 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
5838 - don't do DNS lookups on NIDs too small for IP addresses (3442)
5839 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
5840 - fix cond_resched (3554)
5841 - only evict unfinished clients after recovery (3515)
5842 - allow bulk resend, prevent data loss (3570)
5843 - dynamic ptlrpc request buffer allocation (2102)
5844 - don't allow unlinking open directory if it isn't empty (2904)
5845 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
5846 - remove extraneous obd dereference causing LASSERT failure (3334)
5847 - don't use get_cycles() when creating temp. files on the mds (3156)
5848 - hold i_sem when setting i_size in ll_extent_lock() (3564)
5849 - handle EEXIST for set-stripe, set proper directory name (3336)
5851 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
5852 - fix ksocknal_fmb_callback() error messages (2918)
5854 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
5857 - don't copy lvb into (possibly NULL) reply on error (2983)
5858 - don't deref dentry after dput, don't free lvb on error (2922)
5859 - use the kms to determine writeback rpc length (2947)
5860 - increment oti_logcookies when osc is inactive (2948)
5861 - update client's i_blocks count via lvb messages (2543)
5862 - handle intent open/close of special files properly (1557)
5863 - mount MDS with errors=remount-ro, like obdfilter (2009)
5864 - initialize lock handle to avoid ASSERT on error cleanup (3057)
5865 - don't use cancelling-locks' kms values (2947)
5866 - use highest lock extent for kms, not last one (2925)
5867 - don't dereference ERR_PTR() dentry in error handling path (3107)
5868 - fix thread race in portals_debug_dumplog() (3122)
5869 - create lprocfs device entries at setup instead of at attach (1519)
5870 - common AST error handler, don't evict client on completion race (3145)
5871 - zero nameidata in detach_mnt in 2.6 (3118)
5872 - verify d_inode after revalidate_special is valid in 2.6 (3116)
5873 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
5874 - initialize RPC timeout timer earlier for 2.6 (3219)
5875 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
5876 - print nal/nid for unknown nid (3258)
5877 - additional checks for oscc recovery before doing precreate (3284)
5878 - fix ll_extent_lock() error return code for 64-bit systems (3043)
5879 - don't crash in mdc_close for bad permissions on open (3285)
5880 - zero i_rdev for non-device files (3147)
5881 - clear page->private before handing to FS, better assertion (3119)
5882 - tune the read pipeline (3236)
5883 - fix incorrect decref of invalidated dentry (2350)
5884 - provide read-ahead stats and refine rpc in flight stats (3328)
5885 - don't hold journal transaction open across create RPC (3313)
5886 - update atime on MDS at close time (3265)
5887 - close LDAP connection when recovering to avoid server load (3315)
5888 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
5889 - don't leak open file on MDS after open resend (3325)
5890 - serialize filter_precreate and filter_destroy_precreated (3329)
5891 - loop device shouldn't call sync_dev() for nul device (3092)
5892 - clear page cache after eviction (2766)
5893 - resynchronize MDS->OST in background (2824)
5894 - refuse to mount the same filesystem twice on same mountpoint (3394)
5895 - allow llmount to create routes for mounting behind routers (3320)
5896 - push lock cancellation to blocking thread for glimpse ASTs (3409)
5897 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
5898 - fix rare problem with rename on htree directories (3417)
5900 - allow default OST striping configuration per directory (1414)
5901 - fix compilation for qswnal for 2.6 kernels (3125)
5902 - increase maximum number of MDS request buffers for large systems
5903 - change liblustreapi to be useful for external progs like lfsck (3098)
5904 - increase local configuration timeout for slow disks (3353)
5905 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
5907 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5910 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
5911 - don't overwrite extent policy data in reply if lock was blocked (2901)
5912 - drop filter export grants atomically with removal from device (2663)
5913 - del obd_self_export from work_list in class_disconnect_exports (2908)
5914 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
5915 - swab reply message in mdc_close, other PPC fixes (2464)
5916 - fix destroying of named logs (2325)
5917 - overwrite old logs when running lconf --write_conf (2264)
5918 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
5919 - fix race in target_handle_connect (2898)
5920 - mds_reint_create() should take same inode create lock (2926)
5921 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
5922 - don't close files for self_export to avoid uninitialized obd (2936)
5923 - allow MDS with the same name as client node (2939)
5924 - hold dentry reference for closed log files for unlink (2325)
5925 - reserve space for all logs during transactions (2059)
5926 - don't evict page beyond end of stripe extent (2925)
5927 - don't oops on a deleted current working directory (2399)
5928 - handle hard links to targets without a parent properly (2517)
5929 - don't dereference NULL lock when racing during eviction (2867)
5930 - don't grow lock extents when lots of conflicting locks (2919)
5932 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
5935 - account for cache space usage on clients to avoid data loss (974)
5936 - lfsck support in lustre kernel code (2349)
5937 - reduce journal credits needed for BRW writes (2370)
5938 - orphan handling to avoid losing space on client/server crashes
5939 - ptlrpcd can be blocked, stopping ALL progress (2477)
5940 - use lock value blocks to assist in proper KMS, faster stat (1021)
5941 - takes i_sem instead of DLM locks internally on obdfilter (2720)
5942 - recovery for initial connections (2355)
5943 - fixes for mds_cleanup_orphans (1934)
5944 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
5945 - block all file creations until orphan recovery completes (1901)
5946 - client remove rq_connection from request struct (2423)
5947 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
5948 - recovery timer race (2670)
5949 - mdc_close recovey bug (2532)
5950 - ptlrpc cleanup bug (2710)
5951 - mds timeout on local locks (2588)
5952 - namespace lock held during RPCs (2431)
5953 - handle interrupted sync write properly (2503)
5954 - don't try to handle a message that hasn't been replied to (2699)
5955 - client assert failure during cleanup after abort recovery (2701)
5956 - leak mdc device after failed mount (2712)
5957 - ptlrpc_check_set allows timedout requests to complete (2714)
5958 - wait for inflight reqs when ptlrpcd finishes (2710)
5959 - make sure unregistered services are removed from the srv_list
5960 - reset bulk XID's when resending them (caught by 1138 test)
5961 - unregister_bulk after timeout
5962 - fix lconf error (2694)
5963 - handle write after unfinished setstripe, stripe-only getstripe (2388)
5964 - readahead locks pages, leaves pending causing memory pressure (2673)
5965 - increase OST request buffers to 4096 on large machines (2729)
5966 - fix up permission of existing directories in simple_mkdir (2661)
5967 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
5968 - don't assign transno to errored transactions (2742)
5969 - don't delete objects on OST if given a bogus objid from MDS (2751)
5970 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
5971 - if rq_no_resend, then timeout request after recovery (2432)
5972 - fix MDS llog_logid record size, 64-bit array alignment (2733)
5973 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
5974 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
5975 - MDS recovery shouldn't delete objects that it has given out (2730)
5976 - if enqueue arrives after completion, don't clobber LVB (2819)
5977 - don't unlock pages twice when trigger_group_io returns error (2814)
5978 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
5979 - don't write pages to disk if there was an error (1450)
5980 - don't ping imports that have recovery disabled (2676)
5981 - take buffered bytes into account when balancing socknal conn (2817)
5982 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
5983 - reconnect unlink llog connection after MDS reconnects to OST (2816)
5984 - remove little-endian swabbing of llog records (1987)
5985 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
5986 - retry reposting request buffers if they fail (1191)
5987 - grow extent at grant time to avoid granting a revoked lock (2809)
5988 - lock revoke doesn't evict page if covered by a second lock (2765)
5989 - disable VM readahead to avoid reading outside lock extents (2805)
5991 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
5992 - updated kernel patches for hp-2.4.20 kernel (2681)
5994 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
5997 - fix truncated write corruption (2366)
5998 - fix for failed assertion in iopen_connect_dentry (1792,2517)
6000 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
6001 - ensure len(uuid) < 37 in lmc (1171)
6002 - fix ia64 OOPS in llog_test (2255)
6003 - zero end of page at obdfilter for partial page writes (2648)
6004 - don't leave stale dentries around after renames (bug 2428)
6005 - fix timeouts when evicting a client with a single lock held (2642)
6006 - set deadline for the initial HELLO message to drain (2634)
6007 - print out dotted-quad IP addresses in the socknal (2302)
6009 - additional debugging for MDS client eviction problem (2443)
6010 - fix mkfsoptions support for osts (2603, 2604)
6012 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
6015 - add series for the vanilla 2.6.0 kernel
6016 - add series for the vanilla 2.4.24 kernel
6017 - add series for a cray x86/64 UL kernel drop
6018 - fix xattr patches for the vanilla 2.4.19 series
6020 - generate true UUIDs in lmc (1171)
6021 - have portals stack dumping break in UML (2466)
6022 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
6023 - allocate with _NOFS in ldlm to avoid deadlock (1933)
6024 - wake callback waiting threads on client eviction (2460)
6025 - Add --ptldebug and --subsystem to lmc (1719)
6026 - update assertion to allow safe interrupt allocation
6027 - set rq_no_resend for cancel requests (2432)
6028 - recalculate ptlrpcd timeout after resend (2494)
6029 - call vfs_rmdir when removing pending directories (2368)
6030 - fix renaming a file to itself (2429)
6031 - lmc creates a default one-stripe lov (2454)
6032 - expand procfs space to handle large clusters (2326)
6033 - increase UML stack to avoid overflow
6034 - update lconf's list of debug and subsystem masks
6035 - fix lfs find --obd (2510)
6036 - /proc tunable for disabling filter read caching (2591)
6037 - stop rpm packages from altering slapd.conf (2301)
6038 - disable nagle in the socknal under 0conf (2578)
6039 - choose mds inode size based on stripe count (2572)
6040 - fix kernel-source rpm problems (2516)
6042 - add --disable-doc to avoid pdf generation (2421)
6043 - update documentation, tests, type-os, comments
6044 - avoid format warnings on ia64
6045 - remove the TOE NAL
6046 - tiny code cleanups by removing unused fields
6048 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
6051 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
6052 - remove the most blatant lies from BUILDING file (2371)
6053 - change default debug level to reasonable production setting
6054 - reduce client side cache size to reduce cache flush time
6055 - reduce max RPCs in flight to avoid unnecessary file fragmentation
6056 - make TCP zerocopy and pinger support enabled by default (2476)
6057 - sync writes completed after process exits caused crashes (2319)
6058 - maintain correct mount count on the MDS (2356)
6059 - backout 1557, because 2316 wasn't really fixed
6060 - better file I/O statistics gathering in /proc
6061 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
6062 - another kernel patch to fix zero-copy TCP function export
6063 - don't take duplicate lock when processing re-sent getattr (2420)
6064 - lctl uses obd_self_export instead of creating new conn (2353)
6065 - MDS/OST recovery case which requires object creation asserted (2425)
6066 - move lfs from /usr/sbin to /usr/bin in packages
6067 - fix race between mds_client_add and mds_client_free (2417)
6068 - use kmalloc instead of slabs in portals (2430)
6069 - don't create duplicate records when a failover MDS is present (2442)
6070 - remove unnecessary mount age check (2332)
6071 - don't remove directory inodes from locks prematurely (2451)
6072 - don't break if MDS service name is the same as hostname (2103)
6073 - fix races in client write RPC generation when cache full (2482)
6075 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
6078 - remove now-unused request->rq_obd (278)
6079 - if an allocation fails, print out how much memory we've used (1933)
6080 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
6081 - add the "configurable stack size" patch to most series files (1256)
6082 - ability to write large log records, for 100+ OST configs (2306)
6083 - fix NULL deref when filter_prep fails (2314)
6084 - fix operator precedence error in filter_sync
6085 - dynamic allocation of socknal TX descriptors (2315)
6086 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
6087 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
6088 - ensure that CWARN messages in Portals always get to the syslog
6089 - __init/__exit are not for prototype decls (ldlm_init/exit)
6090 - x86-64 compile warning fixes
6091 - fix gateway LMC keyword conflict (2318)
6092 - fix MDS lock inversions in getattr/reint paths (1844)
6093 - fix a rare lock re-ordering bug, which caused deadlock (2322)
6094 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
6095 - DLM race condition prevented some lock evictions (2328)
6096 - ENOMEM detection and retry on socknal sends (2230)
6097 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
6098 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
6099 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
6100 - better error messages when a client is rejected during recovery (1505)
6101 - avoid cancelling locks which were never granted, after failure (2330)
6102 - fix i_sem/journal inversion in mds_client_add (2333)
6103 - fix truncate/getattr lock cycle deadlock (2334)
6104 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
6105 - fix two rare exit paths which could leak an l_lock() ref (2321)
6106 - fencepost error in MDS/OST orphan recovery (2226)
6107 - make log record alignment 8 bytes (1988)
6108 - lstripe now fails when requested offset > ost_count (2237)
6109 - ensure that all kernel series have a complete list.h (1607)
6110 - fix crashes in special-file operations (2316)
6111 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
6113 - allow configurable automake binary, for testing new versions
6114 - small update to the lfs documentation
6116 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
6118 * fix negative export reference count in fsfilt_sync (2312)
6120 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
6121 * release candidate 0.9.1
6123 - orphans are moved into the PENDING directory for possible recovery
6124 - replayed opens now open by fid for orphan/rename safety (1042)
6125 - last close of an orphan inode generates a transno (683)
6126 - chdir() and mount() now pin the directory entry (1020)
6127 - avoid CERROR in normal ll_setattr_raw() error case (1500)
6128 - discard very old requests without processing them (1502)
6129 - remove some common, well-understood CERRORs (1505)
6130 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
6131 - clear "grant" flags in OST replies until OST grant code lands (1644)
6132 - fix read performance by not clobbering i_blksize on client (1598)
6133 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
6134 - make LRU size a /proc tunable, clears locks when reduced (707)
6135 - fix some lprocfs rot that prevented ptlbd from loading (1732)
6136 - server locks take references on exports now (1558)
6137 - build fixes for 2.4.20-rh trees (1663)
6138 - return an error from lov_create if all OSCs are inactive (1751)
6139 - fix import levels when a reconnect happens without a timeout (1597)
6140 - exit early from mds_open if we get a lookup error (1749)
6141 - partial page read at EOF wouldn't wait for disk before sending (1642)
6142 - avoid NULL deref in obdfilter when reading page past EOF (1592)
6143 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
6144 - fix LBUG in ll_it_open_error with rc = -2 (1861)
6145 - write/truncate lock inversion (1639)
6146 - Don't auto-load obdclass, portals modules during cleanup (1495)
6147 - fix timestamps from jumping to "now" (1763)
6148 - extra journal assertions (1648)
6149 - add an extra multiunlink test (1771)
6150 - fix read_record/write_record API (1776)
6151 - fix leak of offset_extent, possible incorrect i_size later (1772)
6152 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
6153 - Debugging for the kqswnal_get_idle_tx problems (1820)
6154 - Allow recovery to be attempted multiple times (1536)
6155 - Write out MDS last_rcvd file after it is first created (1600)
6156 - Fix tx_descriptor leak in failed transmit situations (1827)
6157 - ext3 journaling fixes for assertion failure after IO error (1871)
6158 - class_export_put() on freed export after completion AST error (1896)
6159 - Fix revalidate looping in VFS (1322)
6160 - Don't access a freed export during MDS_REINT timeout (1521)
6161 - Add open-unlink recovery support on the MDS (1673,1764)
6162 - Return an error if no MDS data was read from last_rcvd (1946)
6163 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
6164 - Don't setup a disk device that doesn't match exported UUID (317)
6165 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
6166 - avoid committing NULL handle in force close
6167 - local.sh is now a one-stripe LOV configuration
6168 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
6169 - fix ext3 htree duplicate directory entry corruption (1516)
6170 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
6171 - update to new LOV EA format (2097)
6172 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
6174 2003-06-15 Phil Schwan <phil@clusterfs.com>
6177 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
6178 - per-import/export recovery handling (958, 931, 959)
6179 - multiple last-rcvd slots, for serving multiple FSes (949)
6180 - connections are again shared between multiple imp/exports (963, 964)
6181 - "umount -f" would hang if any requests needed to be sent (393, 978)
6182 - avoid pinning large req buffer by copying for queued messages (989)
6183 - add "uuid" to "lctl device" command to help upcalls (991)
6184 - "open" RPCs with transnos would confuse recovery counters (1037)
6185 - do proper endian conversion of all wire messages (288, 340, 891)
6186 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
6187 - call waiting locks callback from LDLM recovery thread (1127, 1151)
6188 - fix ptlrpc_connection leak in target_handle_connect (1174)
6189 - fix import refcounting bug in OST and MDS cleanup (1134)
6190 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
6191 - fix very unlikely obd_types race condition (501)
6192 - remove osc_open hack for echo_client (1187)
6193 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
6194 - a failure in read_inode2 leads to deadlock (1139)
6195 - cancel ack-locks as soon as transaction is committed (1072)
6196 - fix major leaks and crashes in the bulk I/O path (937, 1057)
6197 - make sure to commitrw after any preprw to avoid deadlock (1162)
6198 - failing to execute a file in a lustre FS would lock inode (1203)
6199 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
6200 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
6201 - obd_uuid2tgt would walk off the end of the list (1255)
6202 - on IA64 the osc would give portals incorrect bulk size (1258)
6203 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
6204 - fix lock inversion caused by new llite matching code (1282)
6205 - limit the number of dirty pages on a client to 10MB (1286)
6206 - timed out locks were not being corrected cancelled (1289)
6207 - fix O_DIRECT above 4GB on IA-32 (1292)
6208 * major user-visible changes
6209 - fail out/fail over policy now controlled by the upcall (993)
6211 - add OBD_PING to check server availability and failure (954)
6212 - lustre messages are now sent in sending host order (288, 340, 891)
6213 - add eadatalen to MDS getattr reply (340)
6214 - OST read replies may contain second buffer, with per-page status (593)
6216 2003-03-11 Phil Schwan <phil@clusterfs.com>
6219 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
6220 - failed open()s could cause deadlock; fixed (867, 869)
6221 - stop cancelling OST locks when files are closed (481)
6222 - overlapping XID spaces caused network corruption (851, 853)
6223 - fix unsafe fsfilt counter arithmetic; change to atomic_t
6224 - setattr_raw added, to do single-RPC, server-side setattrs
6225 - lmc/lconf syntax change for OST UUIDs
6226 - fix crashy race condition between ptlrpc_free_req and osc_close
6227 - don't use request in mdc_enqueue if we hit a timeout (889)
6228 - don't set the inode i_size for regular files from the MDS (896)
6229 - handle out of order completion AST (842)
6230 - don't LBUG if a lock request times out after receiving AST (913)
6231 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
6232 - if a bad lock AST arrives, send an error instead of dropping entirely
6233 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
6234 - fix leak in bulk IO when only partially completed (899, 900, 926)
6235 - fix O_DIRECT for ia64 (55)
6236 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
6237 - C-z after timeout could hang a process forever; fixed (977)
6239 - client-side I/O cache (678, 924, 929, 941, 970)
6241 - READPAGE and SETATTRs which don't take server-side locks get
6244 2003-02-11 Phil Schwan <phil@clusterfs.com>
6247 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
6248 - Fix fsfilt_extN_readpage() to read a full page of directory
6249 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
6250 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
6251 - Fix imbalanced LOV object allocation and out-of-bound access (469)
6252 - Most intent operations were removed, in favour of a new RPC mode
6253 that does a single RPC to the server and bypasses most of the VFS
6254 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
6255 - Aggressively cancel local locks on DLM servers
6256 - mds_reint_unlink sends EA to the client if it's the last nlink.
6257 client uses that EA to unlink OST objects.
6258 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
6259 - recursive symlinks were fixed (439)
6260 - fixed NULL deref in DEBUG_REQ
6261 - filter_update_lastobjid no longer calls sync, which annoyed extN
6262 - fixed multi-client small-writes to a single file problem (445)
6263 - fixed mtime updates during file writes (607)
6264 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
6265 - fixed bug in obd_brw_read/write() (under guise of testing 367)
6266 - fixed Linux OST size reporting problem (444, 656)
6267 - OST now updates object mtime with writes or setattr (607, 619)
6268 - client verifies file size before zeroing page past EOF (445)
6269 - OST now writes last allocated objid to disk with allocation (108)
6270 - LOV on echo now works (409)
6272 - mds_reint_unlink sends a new buffer, with the EA included. this
6273 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
6274 set if a regular file was being unlinked, and it was the last link
6275 - use PtlGet from the target for bulk writes (315)
6276 - OST now updates object mtime with writes or setattr (607, 619)
6277 - LDLM now has a grant-time callback to revalidate locked items, if
6279 - Many MDS operations were reorganized to combat race conditions
6281 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
6282 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
6284 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
6287 - Fully reactivate OST imports after reconnection (512, others)
6288 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
6289 - More graceful error handling for truncating on dead OST (515)
6290 - Don't error out unless we're actually accessing dead stripes (474)
6291 - Fix garbage sizes when stripes are missing (410)
6292 - LRU counters were broken, causing constant lock purge (433, 432)
6293 - garbage on read from stripes with failed OSTs (441)
6294 - mark OSCs as active before reconnecting during recovery (438)
6295 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
6296 - lfind did not preserve OST order in output (443)
6297 - symlinks cause hung clients, incorrect data (439)
6298 - stop dereferencing request after dropping refcount (457)
6299 - don't LASSERT(spin_is_locked) on non-SMP (455)
6300 - fixes for many rename() bugs
6301 - fstat didn't correctly synchronize attributes (399)
6302 - server must handle lock cancellation during blocking AST prep (487)
6303 - bulk descriptors were free()d too soon (511)
6304 - fix paths in lconf, which would load incorrect modules (451, 507)
6305 - fix confusing lconf 'host not found' error message (386)
6306 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
6307 - fix race condition in mdc_blocking_ast() for inode access (526)
6308 - fix lov_unpackmd() unpacking wrong number of stripes (537)
6309 - fix lov_set_osc_active() marking wrong OSC inactive (440)
6310 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
6311 - fix multiple writes of stripe MD to MDS (358, maybe 519)
6312 - fix lstripe in several ways (kernel side) (527)
6313 - fix request leak in ldlm_cli_enqueue (262)
6314 - incorrect OSC was marked inactive after OST failure
6315 - call mds_fs_cleanup before unmounting filesystem (524)
6316 - fix races between taking ns_lock and ldlm_lock_change_resource
6317 - fix races updating LOV export open file list
6318 - fix lov_enqueue error path, avoid decref-ing bad lock handle
6319 - fix recovery NULL deref in ldlm_cli_cancel_unused
6320 - fix some DLM races by using new hash table for lock handles (419)
6321 - permit the client to specify desired inodes, at replay
6322 - duplicate requests when we queue them for replay reintegration
6323 - fix last_rcvd offset calculation
6324 - sync after each recovered transaction, so we always make progress
6325 - never, not always, ERESTART requests without transnos
6326 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
6328 - skip replay if the MDS says that the client is already connected
6329 - don't check for a recovery-enabled export to match lctl's UUID
6330 - don't INC_USE_COUNT for phantom exports
6331 - don't crash when cleaning up phantom exports (567)
6332 - don't double-finish or set replay data for errored mdc_open requests
6333 - abort requests when they time out, so we don't get old replies
6334 - send/receive replies for AST messages again
6335 - if the client says that it doesn't have the lock, cancel it on the
6337 - if we timeout during I/O, don't try to cancel an in-use lock; instead
6338 mark it as destroyed, it will all work out when decref is called
6339 - fix module use counts (22, 581)
6341 - ASTs now expect a reply (server cancels lock on error reply)
6343 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
6346 - fix many simultaneous client startup (392)
6347 - fix dentry->d_it clobbering
6348 - credentials weren't being shipped for readdir/getattr operations
6349 - remove invalid assertions triggered during some concurrent MD
6351 - proper Lustre versions added (336, 389)
6352 - fix memory leak for create error case (398)
6353 - fix LOV locking bug that would get cli/srv out of sync
6354 - fix echo client over LOV (409)
6355 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
6356 - fix double-O_EXCL intent crash (424)
6357 - avoid sending multiple lock CANCELs (352)
6359 - MDS can do multi-client recovery (modulo bugs in new code)
6361 - many updates, edits, cleanups
6363 2002-11-18 Phil Schwan <phil@clusterfs.com>
6366 - fix null d_it dereference (346)
6367 - fix full OST/dbench hang (333)
6368 - fix permission problem with file removal (286)
6369 - fix removal of OSCs from LOV when they fail
6370 - fix NULL deref during bulk timeout (214)
6371 - fix problems related to multiple filesystems on one MDS (241)
6372 - fixed serious subtle metadata locking bugs
6373 - free locks on clients when inodes are removed due to memory
6375 - fix inode pointer in lock data (285)
6376 - partial support for multiple MDS on a single host (241)
6377 - data locks weren't cancelled at clear_inode time (290, 311)
6378 - intent locks could lead to unbounded lock growth (205)
6379 - added a maximum lock count, an LRU list, and a flusher
6380 - fix multiple rename (365)
6381 - properly abstracted the echo client
6382 - OSC locked 1 byte too many; fixed
6383 - rewrote brw callback code:
6384 - fixed recovery bugs related to LOVs (306)
6385 - fixed too-many-pages-in-one-write crash (191)
6386 - fixed (again) crash in sync_io_timeout (214)
6387 - probably fixed callback-related race (385)
6389 - Add capability to MDS protocol
6390 - LDLM cancellations and callbacks on different portals
6392 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
6395 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
6396 - defer open object destruction to close time (601981, 138)
6397 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
6398 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
6399 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
6400 - don't call obd_setattr in ll_file_release for destroyed objects
6401 * protocol change to lustre_msg: move |version| and add |flags|
6402 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
6403 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
6404 * added replay of create, unlink, link and rename operations during
6405 MDS failover; recovery should be much more robust now
6406 * remove failed OSCs from LOVs (only lov_create uses this so far)
6407 * the lustre-HOWTO was brought (more) up to date (582544)
6409 2002-10-23 Phil Schwan <phil@clusterfs.com>
6412 - in-use dentries weren't being reused properly (617851)
6413 - prevent multiple LDLM setup (599178)
6414 - fix LOV size calculations for truncate (617853)
6415 - fix client handling of MDS intent errors (POSIX)
6416 - fix permission bug in lovstripe.c test (624321)
6417 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
6418 - truncate past end of file could corrupt data
6419 - proper cleanup after timeouts, crashes, etc (592524, 550815)
6420 - a race in recovery could return ETIMEDOUT to apps (623947)
6421 - building outside the source directory was fixed
6422 * the lustre-HOWTO was brought (more) up to date (582544)
6423 * major progress was made on recovery functionality
6425 2002-10-10 Phil Schwan <phil@clusterfs.com>
6428 - recovery deadlock fix
6429 - rm -rf causes LBUG fix (617817)
6430 - file open by multiple tasks fix (618962)
6431 - directory permissions bugs (602707 and 620007)
6432 - journal_stop fixed with locking (611313)
6433 - O_APPEND failures resolved (618273, perhaps 614459)
6434 - lconf PATH fix (619770)
6435 - IA64 build fix (621450)
6436 - RPC buffer sizes scale with amount of memory
6438 2002-10-01 Phil Schwan <phil@clusterfs.com>
6441 - locks would be cancelled without throwing away data pages,
6442 resulting in inconsistent data (605627)
6443 - inode attributes were not always being refreshed (605627, 612449)
6444 - lconf now continues to cleanup after lctl reports an error
6445 - MDS now enforces user permissions (602707)
6446 - lprocfs cleanup fixed, but not yet enabled (614157)
6447 - fixed infinite server hang, should a client not respond to an AST
6448 - avoid going into recovery if user calls readlink() with a buffer
6449 that's too small (613941)
6450 - AST RPCs no longer require replies (614867) -- this may be changed
6451 - don't crash server if client sends an IOV that's too big (611336)
6452 - fixed lock conversion deadlock (611892)
6453 - fixed the following of symlinks (614622)
6454 * recovery: the server can remove locks from a client that dies, other
6455 clients can make progress
6456 * more extN patch fixes
6457 * compile-time configurable ptlrpc buffer allocations
6459 - collaborative read cache document
6460 - Lustre Lite Performance CDR document-in-progress
6462 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6465 - fix typo in patch-2.4.18
6467 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6470 - clear ptlrpc request each time in handle_incoming_request()
6471 - unlink of files now destroys the object on the OST
6473 2002-09-19 Peter Braam <braam@clusterfs.com>
6475 * add hard link support
6476 * change obdfile creation method
6477 * kernel patch changed
6479 2002-09-19 Peter Braam <braam@clusterfs.com>
6482 - stack overflow bug in extN fixed
6484 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
6486 * documentation updates
6487 - add man pages for config tools
6488 - update tests/README to describe testing with new config tools
6489 - finish metadata API descriptions
6490 * bug fixes and cleanups
6491 - statfs workaround for 16TB limit
6492 - LOV stripe allocation improved, can stripe on subset of OSTs
6493 - LOV file size/IO offset was wrong for files > 4GB in size
6494 - object EA data was being dropped, caused files to be unreadable
6495 - memory overflow with non-LOV OST caused memory corruption
6496 - fixed regression tests to work with new config tools, obdfilter
6497 - fixed bug when directory size became larger than 1 block
6498 - fixed bug (for single client case) when PWD was deleted
6499 - invalidate local directory pages when doing intent-based ops
6500 - avoid LDLM oops when lock callback contained bad data
6502 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
6504 * documentation updates
6505 * bug fixes and cleanups
6506 - configuration tools
6509 - 64-bit compile warnings
6510 - 64-bit internal statfs data
6512 * test_brw on persistent OST devices
6514 * lprocfs (disabled)
6516 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
6518 * documentation updates
6519 * bug fixes and cleanups
6520 * configuration tools
6522 2002-08-30 Peter J. Braam <braam@clusterfs.com>
6525 * many small fixes to 0_5_4
6526 * io/network handling
6527 * thinkos in MDS operations
6529 2002-08-24 Peter J. Braam <braam@clusterfs.com>
6532 * crucial basic fixes to 0.5.3
6533 * IOR, Iozone work over Elan
6536 2002-08-07 Phil Schwan <phil@clusterfs.com>
6537 * version 0_5_3, our first alpha
6538 * we use the new Portals iovs
6539 * documentation updates
6540 * bug fixes and cleanups
6541 * small changes in the DLM wire protocol
6543 2002-07-25 Peter J. Braam <braam@clusterfs.com>
6544 * version 0_5_1 with some initial stability,
6545 * locking on MD and file I/O.
6546 * documentation updates
6547 * several bug fixes since 0.5.0
6548 * small changes in wire protocol
6550 2002-07-18 Phil Schwan <phil@clusterfs.com>
6552 * delivered as Lustre Light Alpha
6553 * fixed a crash after handling invalid MDS requests
6554 * fixed directory pages for architectures with non-4k pages sizes
6556 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
6557 * release version v0_4_4
6558 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
6560 * Fixes a number of interruption problems with OST operations.
6561 * Update documentation for portals header changes
6562 * Move all wire protocol structs/defines to lustre_idl.h
6563 * Fixes symlink length bug.
6564 * Add tcpdump to repository.
6566 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
6567 * release version v0_4_3
6568 * Fixes statfs for inodes on extN.
6569 * Fixes bug in runtests which would delete /etc/hosts.
6570 * Use 64-bit object IDs wherever possible (not into VFS though)
6571 Remove ost_get_info, which is unused by lustre, and out of date.
6573 2002-07-03 Peter Braam <braam@clusterfs.com>
6574 * release version v0_4_2 Fixes a lookup error (type not passed)
6575 * move forward to head of Portals
6576 * move forward to latest Lustre kernel
6578 2002-06-25 Peter Braam <braam@clusterfs.com>
6579 * release version v0_4_1. Hopefully stable on single node use.