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 2.6.32-642.11.1.el6 (RHEL6.8)
10 3.10.0-514.el7 (RHEL7.3)
11 3.0.101-0.47.71 (SLES11 SP3)
12 3.0.101-80 (SLES11 SP4)
13 3.12.62-60.64.8 (SLES12 SP1)
14 vanilla linux 4.5.7 (ZFS only)
15 * Client known to build on unpatched kernels:
16 2.6.32-431.29.2.el6 (RHEL6.5)
17 2.6.32-504.30.3.el6 (RHEL6.6)
18 2.6.32-573.26.1.el6 (RHEL6.7)
19 2.6.32-642.11.1.el6 (RHEL6.8)
20 3.10.0-514.el7 (RHEL7.3)
21 3.0.101-0.47.71 (SLES11 SP3)
22 3.0.101-80 (SLES11 SP4)
23 3.12.62-60.64.8 (SLES12 SP1)
25 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
26 * Recommended ZFS / SPL version: 0.6.5.7
27 * Tested with ZFS / SPL version: 0.6.5.7
28 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
29 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
30 stack overflow. For more information, please refer to bugzilla 17630.
31 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
32 the re-exporting nodes to be mounted with "32bitapi" mount option
34 --------------------------------------------------------------------------------
36 12-07-2016 Intel Corporation
38 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
39 for currently supported client and server kernel versions.
40 * Server known to build on patched kernels:
41 2.6.32-431.29.2.el6 (RHEL6.5)
42 2.6.32-504.30.3.el6 (RHEL6.6)
43 2.6.32-573.26.1.el6 (RHEL6.7)
44 2.6.32-642.11.1.el6 (RHEL6.8)
45 3.10.0-514.el7 (RHEL7.3)
46 3.0.101-0.47.71 (SLES11 SP3)
47 3.0.101-80 (SLES11 SP4)
48 3.12.62-60.64.8 (SLES12 SP1)
49 vanilla linux 4.5.7 (ZFS only)
50 * Client known to build on unpatched kernels:
51 2.6.32-431.29.2.el6 (RHEL6.5)
52 2.6.32-504.30.3.el6 (RHEL6.6)
53 2.6.32-573.26.1.el6 (RHEL6.7)
54 2.6.32-642.11.1.el6 (RHEL6.8)
55 3.10.0-514.el7 (RHEL7.3)
56 3.0.101-0.47.71 (SLES11 SP3)
57 3.0.101-80 (SLES11 SP4)
58 3.12.62-60.64.8 (SLES12 SP1)
60 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
61 * Recommended ZFS / SPL version: 0.6.5.7
62 * Tested with ZFS / SPL version: 0.6.5.7
63 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
64 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
65 stack overflow. For more information, please refer to bugzilla 17630.
66 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
67 the re-exporting nodes to be mounted with "32bitapi" mount option
69 --------------------------------------------------------------------------------
71 02-29-2016 Intel Corporation
73 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
74 for currently supported client and server kernel versions.
75 * Server known to build on patched kernels:
76 2.6.32-431.29.2.el6 (RHEL6.5)
77 2.6.32-504.30.3.el6 (RHEL6.6)
78 2.6.32-573.12.1.el6 (RHEL6.7)
79 3.10.0-327.3.1.el7 (RHEL7.2)
80 3.0.101-0.47.71 (SLES11 SP3)
81 3.0.101-68 (SLES11 SP4)
82 vanilla linux 4.2.1 (ZFS only)
83 * Client known to build on unpatched kernels:
84 2.6.32-431.29.2.el6 (RHEL6.5)
85 2.6.32-504.30.3.el6 (RHEL6.6)
86 2.6.32-573.12.1.el6 (RHEL6.7)
87 3.10.0-327.3.1.el7 (RHEL7.2)
88 3.0.101-0.47.71 (SLES11 SP3)
89 3.0.101-68 (SLES11 SP4)
92 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
93 * Recommended ZFS / SPL version: 0.6.4.2
94 * Tested with ZFS / SPL version: 0.6.4.2
95 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
96 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
97 stack overflow. For more information, please refer to bugzilla 17630.
98 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
99 the re-exporting nodes to be mounted with "32bitapi" mount option
101 --------------------------------------------------------------------------------
103 03-10-2015 Intel Corporation
105 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
106 for currently supported client and server kernel versions.
107 * Server known to build on patched kernels:
108 2.6.32-431.29.2.el6 (RHEL6.5)
109 2.6.32-504.8.1.el6 (RHEL6.6)
110 3.0.101-0.46 (SLES11 SP3)
111 * Client known to build on unpatched kernels:
112 2.6.32-431.29.2.el6 (RHEL6.5)
113 2.6.32-504.8.1.el6 (RHEL6.6)
114 3.10.0-123.20.1.el7 (RHEL7)
115 3.0.101-0.46 (SLES11 SP3)
116 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
117 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
118 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
119 stack overflow. For more information, please refer to bugzilla 17630.
120 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
121 the re-exporting nodes to be mounted with "32bitapi" mount option
123 Severity : enhancement
125 Description: control OST-index in IDIF via ROCOMPAT flag.
126 Details : Introduce new flag OBD_ROCOMPAT_IDX_IN_IDIF that is stored in the
127 last_rcvd file. For new formatted OST device, it will be auto set;
128 for the case of upgrading from old OST device, you can enable it
129 via the lproc interface osd-ldiskfs.index_in_idif. With such flag
130 enabled, for new created OST-object, its IDIF-in-LMA will contain
131 the OST-index; for the existing OST-object, the OSD will convert
132 old format IDIF as new format IDIF with OST-index stored in the
133 LMA EA when accessing such OST-object or via OI scrub. Once such
134 flag is enabled, it cannot be reverted back, so the system cannot
135 be downgraded to the orignal incompatible version.
137 --------------------------------------------------------------------------------
139 07-30-2014 Intel Corporation
141 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
142 for currently supported client and server kernel versions.
143 * Server known to build on patched kernels:
144 2.6.32-431.20.3.el6 (RHEL6.5)
145 3.0.101-0.31 (SLES11 SP3)
146 * Client known to build on unpatched kernels:
147 2.6.32-431.20.3.el6 (RHEL6.5)
148 3.0.101-0.31 (SLES11 SP3)
151 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
152 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
153 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
154 stack overflow. For more information, please refer to bugzilla 17630.
155 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
156 the re-exporting nodes to be mounted with "32bitapi" mount option
158 --------------------------------------------------------------------------------
160 10-31-2013 Intel Corporation
162 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
163 * Server builds for kernels:
164 2.6.32-358.18.1.el6 (RHEL6.4)
165 * Client builds for unpatched kernels:
166 2.6.32-358.18.1.el6 (RHEL6.4)
167 3.0.92-0.8 (SLES11 SP3)
169 * Recommended e2fsprogs version: 1.42.7.wc2 or newer
170 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
171 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
172 stack overflow. For more information, please refer to bugzilla 17630.
173 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
174 the re-exporting nodes to be mounted with "32bitapi" mount option
176 --------------------------------------------------------------------------------
178 05-30-2013 Intel Corporation
180 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
181 * Server builds for kernels:
182 2.6.32-358.6.2.el6 (RHEL6.4)
183 * Client builds for unpatched kernels:
184 2.6.32-358.6.2.el6 (RHEL6.4)
185 3.0.74-0.6.6 (SLES11 SP2)
187 * Recommended e2fsprogs version: 1.42.7.wc1 or newer
188 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
189 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
190 stack overflow. For more information, please refer to bugzilla 17630.
191 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
192 the re-exporting nodes to be mounted with "32bitapi" mount option
194 --------------------------------------------------------------------------------
196 9-30-2012 Whamcloud, Inc.
198 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
199 * Server support for kernels:
200 2.6.32-279.5.1.el6 (RHEL6)
201 * Client support for unpatched kernels:
202 2.6.32-279.5.1.el6 (RHEL6)
204 2.6.18-238.19.1.el5 (RHEL5)
205 2.6.16.60-0.69.1 (SLES10)
206 * Recommended e2fsprogs version: 1.41.90.wc4
207 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
208 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
209 stack overflow. For more information, please refer to bugzilla 17630.
210 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
211 the re-exporting nodes to be mounted with "32bitapi" mount option
213 Severity : enhancement
215 Description: Update RHEL6.3 kernel to 2.6.32-279.5.1.el6.
217 Severity : enhancement
219 Description: Update RHEL6.3 kernel to 2.6.32-279.2.1.el6.
221 Severity : enhancement
223 Description: Update RHEL6.2 kernel to 2.6.32-220.17.1.el6
225 Severity : enhancement
227 Description: Update RHEL6.2 kernel to 2.6.32-220.13.1.el6
229 Severity : enhancement
231 Description: Update RHEL6.2 kernel to 2.6.32-220.7.1.el6
233 --------------------------------------------------------------------------------
235 04-30-2012 Whamcloud, Inc.
237 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
238 * Server support for kernels:
239 2.6.32-220.4.2.el6 (RHEL6)
240 * Client support for unpatched kernels:
241 2.6.32-220.4.2.el6 (RHEL6)
243 2.6.18-238.19.1.el5 (RHEL5)
244 2.6.16.60-0.69.1 (SLES10)
245 * Recommended e2fsprogs version: 1.41.90.wc4
246 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
247 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
248 stack overflow. For more information, please refer to bugzilla 17630.
249 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
250 the re-exporting nodes to be mounted with "32bitapi" mount option
252 Severity : enhancement
254 Description: Update RHEL6.2 kernel to 2.6.32-220.4.2.el6
256 Severity : enhancement
258 Description: Update RHEL6.2 kernel to 2.6.32-220.4.1.el6
260 Severity : enhancement
262 Description: Update RHEL6.2 kernel to 2.6.32-220.el6
264 Severity : enhancement
266 Description: Update RHEL6.1 kernel to 2.6.32-131.17.1.el6
268 --------------------------------------------------------------------------------
270 09-30-2011 Whamcloud, Inc.
272 * Server support for kernels:
273 2.6.18-238.19.1.el5 (RHEL5)
274 2.6.32-131.6.1.el6 (RHEL6)
275 * Client support for unpatched kernels:
276 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
277 2.6.18-238.19.1.el5 (RHEL5)
278 2.6.32-131.6.1.el6 (RHEL6)
279 2.6.29.4-167.fc11 (FC11)
280 2.6.16.60-0.69.1 (SLES10)
282 * Recommended e2fsprogs version: 1.41.90.wc3
283 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
284 Lustre filesystem with 4K stack may cause a stack overflow. For more
285 information, please refer to bugzilla 17630.
286 * NFSv4 reexport requires Lustre client on the re-exporting nodes to be
287 mounted with "32bitapi" mount option
289 Severity : enhancement
290 Description: Update RHEL6.1 kernel to 2.6.32-131.6.1.el6
292 Severity : enhancement
293 Description: Update RHEL5.6 kernel to 2.6.18-238.19.1.el5
295 Severity : enhancement
296 Description: Update RHEL5.6 kernel to 2.6.18-238.9.1.el5
299 Frequency : only with SLES10
301 Description: Use OFED "KMP" provided by Novell
302 Details : SLES10 SP3 ships with OFED in a separate "KMP" package.
303 Lustre is now built against this package. That means you need to
304 install the ofed-kmp package from Novell for the patchless client.
305 Note that the ofed-kmp that Novell ships may not exactly match the
306 kernel version but should still be compatible.
308 Severity : enhancement
310 Description: Update RHEL5.5 kernel to 2.6.18-194.17.1.el5
311 Update OEL5.5 kernel to 2.6.18-194.17.1.0.1.el5.
313 --------------------------------------------------------------------------------
315 2010-07-15 Oracle, Inc.
317 * Server support for kernels:
318 2.6.18-164.11.1.el5 (RHEL 5)
319 2.6.18-164.11.1.0.1.el5 (OEL 5)
320 * Client support for unpatched kernels:
321 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
322 2.6.18-164.11.1.el5 (RHEL 5),
323 2.6.18-164.11.1.0.1.el5 (OEL 5)
324 2.6.16.60-0.42.8 (SLES 10),
326 2.6.29.4-167.fc11 (FC11)
327 * Recommended e2fsprogs version: 1.41.10-sun2
329 Severity : enhancement
331 Description: Distribute lfs_migrate script with lustre
332 Details : The lfs_migrate mentioned in the manual is now included in the
333 lustre tree in lustre/scripts and will distributed in lustre RPMs.
335 Severity : enhancement
337 Description: Failover nid list added to import proc file.
339 Severity : enhancement
341 Description: "som_preview" mount option required to use SOM
342 Details : The clients will need to use som_preview mount option
346 Description: Lustre does not do 1MB IOs to HW RAID
347 Details : Bump MAX_PHYS/HW_SEGMENTS and SG_ALL to 256 in the RHEL5 kernel.
349 Severity : enhancement
351 Description: lfs quota output is not very convenient for awk/sed-parsing
352 Details : Some positions in lfs quota output table could be empty or
353 non-empty which made it hard to parse it with scripts, now
354 a dash is put instead of space where there is not supposed
359 Description: Update to OFED 1.5.1
360 Details : OFED 1.5.1 will be used in environments that don't use
361 vendor supplied OFED.
365 Description: Remove lnet. prefix from lctl params display
366 Details : Also renamed lnet.memused to lnet_memused to
371 Description: Use conf_param -d to remove permanant settings.
375 Description: Add list_param -R
376 Details : lctl list_param now supports recursive mode.
379 Bugzilla : 21829, 22312
380 Description: Fix llobdstat.
381 Details : 1. Now updating the write_bytes counter
382 2. Port script fixes from 1.8.
386 Description: decrease the usage of memory on clients.
387 Details : 1. On clients, recycle dentries and inodes unused.
388 2. Delete the code related to ll_deathrow(att 6215 in bug 1443). It
393 Description: llobdstat fix and enhancement
394 Details : add a counter to set a limit to how many samples will be returned
395 fix a wildcard in the path to limit to obdfilter stats only
399 Description: lfs check only allowed for root
400 Details : Code cleanups and added a regresion test of non-root use of
405 Description: MDS Oopses in ptlrpc_dispatch_difficult_reply()/__list_add()
406 Details : Fix hr->hr_index overload.
410 Description: Fix FIEMAP functionality.
414 Description: FC11 patchless client support added
418 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and
419 OEL5.4 kernel to 2.6.18-164.11.1.0.1.el5.
423 Description: New acc-sm script and various test-framework enhancments.
427 Description: Previously landed patch was reverted.
431 Description: Update SLES10 kernel to 2.6.16.60-0.42.8 (patchless)
435 Description: Add ldiskfs maxdirsize mount option.
439 Description: Fix mkfs.lustre support for 16TB LUNs.
443 Description: Print mmp_check_interval
444 Details : print mmp_check_interval and make it possible to abort
445 mount operation in case it takes too long.
449 Description: Add cascading_rw to lustre mpi tests.
453 Description: Store SOM data in EA.
454 Details : Also several SOM related cleanups.
458 Description: Add refcount to md_open_data
462 Description: Use refcount on lov_request_set to prevent memory corruption.
466 Description: Statahead: use parent dir's i_mutex for synchronization.
470 Description: loadgen improvements.
474 Description: Enable peer health detection in ksocklnd and ko2iblnd.
478 Description: Control DCACHE_LUSTRE_INVALID flag with MDS_INODE_LOOKCUP lock
482 Description: Fix race on obd_devcie:md_stats
486 Description: Fix SLES 11 client builds.
487 Details : Resolved uninitalized pointer gcc warnings.
491 Description: Allow non-root access for "lfs check".
492 Details : Added a check in obd_class_ioctl() for OBD_IOC_PING_TARGET.
496 Description: Clear imp_force_reconnect correctly in ptlrpc_connect_interpret()
498 Severity : enhancement
500 Description: Update kernel to RHEL5.4 2.6.18-164.6.1.el5 and
501 OEL5 2.6.18-164.6.1.0.1.el5(Both in-kernel OFED enabled).
505 Description: Sanity test_27p fails.
506 Details : OBD_FAIL_OST_ENOSPC has been moved to filter_statfs().
507 Sanity tests 27 were cleaned up and improved.
511 Description: Unify req format on client/servers
512 Details : Use new req_capsule API [almost] everywhere instead of old PTLRPC
513 buffers and swabbers approach..
518 Description: drop join file stuff.
520 Severity : enhancement
522 Description: Add 2.6.27 and SLES11 (patchless client) support
526 Description: Fix in ptlrpc_expire_one_request() to print the signed time
529 Severity : enhancement
531 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
534 Severity : enhancement
536 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
538 Severity : enhancement
540 Description: correctly handle big reply message.
541 Details : send LNet event if reply is bigger than buffer and adjust this buffer
546 Description: Drop unnecessary __GFP_NOMEMALLOC flag from filter_get_page()
548 Severity : enhancement
550 Description: Add support for OEL5.
554 Description: Conf-sanity.sh 50g test - deactivated OST should not cause a panic.
556 Severity : enhancement
558 Description: "weak-modules" support
559 Details : Implement "weak-modules" support which enables kernel modules
560 to be used with any kernel that implements the same kABI. In
561 order to achieve this modules are now installed in
562 /lib/modules/$(uname -r)/updates/kernel on all distributions.
566 Description: client could not reconnect to OST because of an active request.
567 Details : The client now retries to reconnect to the same server, if a
568 connect request failed with EBUSY or -EAGAIN. Test case 35b
569 conf-sanity.sh has been created.
573 Description: Don't allow make backward step on assignin osc next id.
574 Details : race between allocation next id and ll_sync thread can be cause
575 of set wrong osc next id and can be kill valid ost objects.
579 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
583 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
587 Description: Error handling in osc_statfs_interpret() has been improved.
588 Details : Check in osc_statfs_interpret() for EBADR.
592 Description: Do not send statfs() requests to OSTs disabled by administrator.
593 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
595 Severity : enhancement
597 Description: Added sanity-quota test to for lfs(1) to work with numeric uid/gid.
599 Severity : enhancement
601 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
603 Severity : enhancement
605 Description: Enhancement for lfs(1) command to use numeric uid/gid.
609 Description: Improved cleanup procedure for conf-sanity test_46a.
613 Description: Do not start bulk transfer, if client tried to reconnect.
614 Details : Do not start bulk transfer, if client tried to reconnect,
615 but failed due to active rpcs. Check for exp_abort_active_req
616 has been implemented.
620 Description: Prevent inconsistences between linux and lustre mount structures.
621 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
622 Make the sleep interruptible.
624 Severity : enhancement
626 Description: provide server to client comms path
627 Details : server to client communications path via new LDLM_SET_INFO rpc
631 Description: truncate starts GFP_FS allocation under transaction causing deadlock
632 Details : ldiskfs_truncate calls grab_cache_page which may start page
633 allocation under an open transaction. This may lead to
634 calling prune_icache with consequent lustre reentrance.
638 Description: send correctly lsm on open replay
639 Details : MDS is trust to LSM size on replay open, but client can set wrong size
644 Description: enable client interface failover
645 Details : When a child reconnects from another NID, properly update export
646 nid hash position and ldlm reverse import.
648 Severity : enhancement
650 Description: Communicate OST degraded/readonly state via statfs to MDS
651 Details : Flags in the statfs returned from OSTs indicate whether the
652 OST is in a degraded RAID state, or if the filesystem has
653 turned read-only after a filesystem error is detected.
658 Description: don't panic if EPROTO was hit when reading symlink
659 Details : correctly handling request reference in error cases.
661 Severity : enhancement
663 Description: Add LustreNetLink, a kernel-userspace communcation path.
665 Severity : enhancement
667 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
672 Description: access to llog context before init.
673 Details : move handling CATALOGS file at osc layer and forbid access to llog
677 Frequency : rare, only when client is writing to OST on local node
679 Description: Avoid deadlock for local client writes
680 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
681 memory freeing context. This allows OST threads to set the
682 PF_MEMALLOC flag on task structures in order to allocate memory
683 from reserved pools and complete IO.
684 Use GFP_HIGHUSER for OST allocations for non-local client writes,
685 so that the OST threads generate memory pressure and allow
686 inactive pages to be reclaimed.
688 Severity : enhancement
690 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
695 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
696 Details : move ldlm namespace creation in setup phase to avoid grab
697 _lprocfs_lock with cli_sem held.
701 Description: Temporarily disable grant shrink.
702 Details : Disable the feature for debugging.
705 Frequency : only when formatting test filesystems
707 Description: Unable to run several mkfs.lustre on loop devices at the same time
708 Details : mkfs.lustre returns error 256 on the concurrent loop devices
709 formatting. The solution is to proper handle the error.
711 Severity : enhancement
713 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
715 Severity : enhancement
717 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
722 Description: resolve race between obd_disconnect and class_disconnect_exports
723 Details : if obd_disconnect will be called to already disconnected export he
724 forget release one reference and osc module can't unloaded.
726 Severity : enhancement
728 Description: Allow tuning service thread via /proc
729 Details : For each service a new
730 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
731 created that can be used to set min/max thread counts, and get the
732 current number of running threads.
737 Descriptoin: don't return error if have partially created objects for file.
738 Details : lov_update_create_set uses set->set_success as index for created
739 objects, so if some requests will be failed, they will have hole
740 at end of array and qos_shrink_lsm for allocate correct lsm.
742 Severity : enhancement
744 Description: Update OFED support to 1.4.1
749 Description: Reduce small size read RPC
750 Details : Set read-ahead limit for every file and only do read-ahead when
751 available read-ahead pages are bigger than 1M to avoid small size
754 Severity : enhancement
756 Description: move AT tunable parameters for more consistent usage
757 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
759 Severity : enhancement
761 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
762 Details : allow skip disconnected ost for send statfs request and hide error
768 Description: fix racy locking of mballoc block bitmaps causing BUG
769 Details : The locking of the mballoc buddy bitmap and the in-memory
770 block bitmap was using two different spin locks in some
771 cases. This made it possible to incorrectly access the
772 mballoc bitmap while another process was modifying it,
773 causing a sanity assertion to fail. While no on-disk corruption
774 was reported, there was some risk of this happening.
777 Frequency : rare, on llog test 6
779 Descriptoin: don't allow connect to already connected import
780 Details : allowing connect to already connected import is hide connecting problem.
785 Descriptoin: allow kill process which wait statahead result
786 Details : for some reasons 'ls' can stick in waiting result from statahead,
787 in this case need way for kill this process.
789 Severity : enhancement
791 Description: Add state history info file, enhance import info file
792 Details : Track import connection state changes in a new osc/mdc proc file;
793 add overview-type data to the osc/mdc import proc file.
795 Severity : enhancement
797 Description: MDS create should not wait for statfs RPC while holding DLM lock.
799 Severity : enhancement
801 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
806 Descriptoin: OST grant too much space to client even there are not enough space.
807 Details : Client will shrink its grant cache to OST if there are no write
808 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
809 this grant cache if there are already not enough avaible space
810 (left_space < total_clients * 32M).
813 Frequency : start MDS on uncleanly shutdowned MDS device
815 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
816 Details : stay in waiting mds<>ost recovery finished produce random bugs
817 due race between two ll_sync thread for one lov target. send
818 ACTIVATE event only if connect realy finished and import have
822 Frequency : rare, connect and disconnect target at same time
824 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
825 Details : don't call obd_disconnect under lov_lock. this long time
826 operation and can block ptlrpcd which answer to connect request.
831 Descriptoin: don't lose wakeup for imp_recovery_waitq
832 Details : recover_import_no_retry or invalidate_import and import_close can
833 both sleep on imp_recovery_waitq, but we was send only one wakeup
837 Frequency : always with long access acl
839 Descriptoin: mds can't pack reply with long acl.
840 Details : mds don't control size of acl but they limited by reint/getattr
843 Severity : enhancement
845 Description: Update to SLES10 kernel-2.6.16.60-0.33.
847 Severity : enhancement
849 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
852 Frequency : start MDS on uncleanly shutdowned MDS device
854 Descriptoin: aborting recovery hang on MDS
855 Details : don't throttle destroy RPCs for the MDT.
858 Frequency : on remount
860 Description: external journal device not working after the remount
861 Details : clear dev_rdonly flag for external journal devices in
867 Description: shutdown vs evict race
868 Details : client_disconnect_export vs connect request race.
869 if client will evicted at this time - we start invalidate
870 thread without referece to import and import can be freed
874 Frequency : rare, need acl's on inode.
876 Description: client can't handle ost additional correctly
877 Details : if ost was added after client connected to mds client can have
878 hit lnet_try_match_md ... to big messages to wide striped files.
879 in this case need teach client to handle config events about add
880 lov target and update client max ea size at that event.
882 Severity : enhancement
884 Description: Changelogs
885 Details : Changelogs are a lightweight mechanism to track filesystem
886 metadata and namespace changes. The changelog is recorded
887 permanently on the MDTs, and is periodically "consumed" / purged
888 when records are no longer needed.
890 Severity : enhancement
892 Description: compact fld format with extents
893 Details : Store range of seq rather than every seq in FLD. Seq
894 controller update FLD rather than clients. In Case of CMD, mdt0
895 has FLD, all other metadata server act as non persistent proxy
896 for FLD queries and cache fld entries in fld cache.
901 Description: don't skip ost target if they assigned to file
902 Details : Drop slow OSCs if we can, but not for requested start idx.
903 This means "if OSC is slow and it is not the requested
904 start OST, then it can be skipped, otherwise skip it only
905 if it is inactive/recovering/out-of-space.
909 Description: more cleanup in mds_lov
910 Details : not send LOV EA under replay, we can't know about they size at this
911 time. Don't allow client connect to mds before any ost connected,
912 for avoid problems with LOV EA size and returning EIO to client.
914 Severity : enhancement
916 Description: Interoperability at server side (Disk interoperability)
918 Severity : enhancement
920 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
922 Severity : enhancement
924 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
926 Severity : enhancement
928 Description: New client IO stack (CLIO).
930 Severity : enhancement
932 Description: Commit on sharing. Eliminate inter-client dependencies between
933 uncommitted transactions by doing transaction commits.
934 Thereby clients may recovery independently.
937 Frequency : Create a symlink file with a very long name
939 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
940 Details : If there is no extra space in the request for early cancels,
941 ldlm_req_handles_avail() returns 0 instead of a negative value.
943 Severity : enhancement
945 Description: Add /proc entry for import status
946 Details : The mdc, osc, and mgc import directories now have
947 an import directory that contains useful import data for debugging
950 Severity : enhancement
952 Description: Re-disable certain /proc logging
953 Details : Enable and disable client's offset_stats, extents_stats and
954 extents_stats_per_process stats logging on the fly.
957 Frequency : Only on FC kernels 2.6.22+
959 Description: oops in statahead
960 Details : Do not drop reference count for the dentry from VFS when lookup,
961 VFS will do that by itself.
963 Severity : enhancement
965 Description: Generic /proc file permissions
966 Details : Set /Proc file permissions in a more generic way to enable non-
967 root users operate on some /proc files.
971 Description: Hitting mdc_commit_close() ASSERTION
972 Details : Properly handle request reference release in
973 ll_release_openhandle().
977 Description: quota recovery deadlock during mds failover
978 Details : This patch includes att18982, att18236, att18237 in bz14840.
980 1. fix osts hang when mds does failover with quotaon
981 2. prevent watchdog storm when osts threads wait for the
986 Frequency : only patchless client
987 Description: add workaround for race between add/remove dentry from hash
989 Severity : enhancement
991 Description: Allow OST glimpses to return PW locks
995 Description: LBUG when llog conf file is full
996 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
1000 Description: Prevent import from entering FULL state when server in recovery
1004 Description: service mount cannot take device name with ":"
1005 Details : Only when device name contains ":/" will mount treat it as
1011 Description: replace ptlrpcd with the statahead thread to interpret the async
1012 statahead RPC callback
1016 Frequency : on recovery
1017 Description: I/O failures after umount during fail back
1018 Details : if client reconnected to restarted server we need join to recovery
1019 instead of find server handler is changed and process self
1020 eviction with cancel all locks.
1022 Severity : enhancement
1024 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
1026 Severity : enhancement
1028 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
1030 Severity : enhancement
1032 Description: Upcall on Lustre log has been dumped
1033 Details : Allow for a user mode script to be called once a Lustre log has
1034 been dumped. It passes the filename of the dumped log to the
1035 script, the location of the script can be specified via
1036 /proc/sys/lnet/debug_log_upcall.
1041 Description: avoid idr_remove called for id which is not allocated.
1042 Details : Move assigment s_dev for clustered nfs to end of initialization,
1043 to avoid problem with error handling.
1048 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
1049 Details : When connection is reused this not moved from CONN_UNUSED_HASH
1050 into CONN_USED_HASH and this prodice warning when put connection
1051 again in unused hash.
1054 Severity : enhancement
1056 Description: File striping can now be set to use an arbitrary pool of OSTs.
1058 Severity : enhancement
1060 Description: Export bytes_read/bytes_write count on OSC/OST.
1064 Description: Early reply size mismatch, MGC loses connection
1065 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
1066 the connect flags are properly negotiated.
1071 Description: quotas are not honored with O_DIRECT
1072 Details : all writes with the flag O_DIRECT will use grants which leads to
1073 this problem. Now using OBD_BRW_SYNC to guard this.
1077 Description: add quota statistics
1078 Details : 1. sort out quota proc entries and proc code.
1079 2. add quota statistics
1081 Severity : enhancement
1083 Description: enable quota support for HEAD.
1087 Description: Properly propagate oinfo flags from lov to osc for statfs
1088 Details : restore missing copy oi_flags to lov requests.
1090 Severity : enhancement
1092 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
1093 : l_getgroups(8), lst(8), routerstat(8)
1094 Details : included man pages for llobdstat(8), llstat(8),
1095 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
1097 Severity : enhancement
1099 Description: configure's --enable-quota should check the
1100 : kernel .config for CONFIG_QUOTA
1101 Details : configure is terminated if --enable-quota is passed but
1102 : no quota support is in kernel
1106 Description: Remove portals compatibility
1107 Details : Remove portals compatibility, not interoperable with releases
1112 Description: Resolve device initialization race
1113 Details : Prevent proc handler from accessing devices added to the
1114 obd_devs array but yet be intialized.
1116 Severity : enhancement
1118 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
1120 Severity : enhancement
1122 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
1127 Description: Using wrong pointer in osc_brw_prep_request
1128 Details : Access to array[-1] can produce panic if kernel compiled with
1129 CONFIG_PAGE_ALLOC enabled
1133 Description: Client runs out of low memory
1134 Details : Consider only lowmem when counting initial number of llap pages
1138 Description: *optional* service tags registration
1139 Details : if the "service tags" package is installed on a Lustre node
1140 When the filesystem is mounted, a local-node service tag will
1141 be created. See http://inventory.sun.com/ for more information
1142 about the Service Tags asset management system.
1146 Description: Kernel BUG tries to release flock
1147 Details : Lustre does not destroy flock lock before last reference goes
1148 away. So always drop flock locks when client is evicted and
1149 perform unlock regardless of successfulness of speaking to MDS.
1153 Description: add refcount protection for osc callbacks, avoid panic on shutdown
1157 Description: sanity test 65a fails if stripecount of -1 is set
1158 Details : handle -1 striping on filesystem in ll_dirstripe_verify
1163 Description: ASSERTION(CheckWriteback(page,cmd)) failed
1164 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
1167 Severity : enhancement
1169 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
1173 Description: do not process already freed flock
1174 Details : flock can possibly be freed by another thread before it reaches
1175 to ldlm_flock_completion_ast.
1179 Description: LBUG during stress test
1180 Details : Need properly lock accesses the flock deadlock detection list.
1184 Description: oops in page fault handler
1185 Details : kernel page fault handler can return two special 'pages' in error
1186 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
1190 Description: timeout with invalidate import.
1191 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
1192 should be handled by ptlrpcd. This produce long age waiting and
1193 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
1195 Severity : enhancement
1197 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
1201 Description: enable MGS and MDT services start separately
1202 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
1203 a co-located MGS without starting the MGS, which is a complement
1204 to 'nosvc' mount option.
1208 Frequency : after recovery
1209 Description: precreate to many object's after del orphan.
1210 Details : del orphan st in oscc last_id == next_id and this triger growing
1211 count of precreated objects. Set flag LOW to skip increase count
1212 of precreated objects.
1216 Frequency : rare, on clear nid stats
1217 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
1218 Details : when clean nid stats sometimes try destroy live entry,
1219 and this produce panic in free.
1223 Description: Stack overflow during MDS log replay
1224 ease stack pressure by using a thread dealing llog_process.
1228 Description: wait until IO finished before start new when do lock cancel.
1229 Details : VM protocol want old IO finished before start new, in this case
1230 need wait until PG_writeback is cleared until check dirty flag and
1231 call writepages in lock cancel callback.
1233 Severity : enhancement
1235 Description: using special macro for print time and cleanup in includes.
1239 Description: mds_mfd_close() ASSERTION(rc == 0)
1240 Details : In mds_mfd_close(), we need protect inode's writecount change
1241 within its orphan write semaphore to prevent possible races.
1245 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
1249 Frequency : rare, on shutdown ost
1250 Description: don't hit live lock with umount ost.
1251 Details : shrink_dcache_parent can be in long loop with destroy dentries,
1252 use shrink_dcache_sb instead.
1256 Description: don't panic with use echo client
1257 Details : echo client pass NULL as client nid pointer and this produce null
1258 pointer dereference.
1262 Description: fix build on ppc32
1263 Details : compile code with -m64 flag produce wrong object file for ppc32.
1267 Description: add message levels for liblustreapi
1271 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
1272 Details : Move the 'good_osts' check before the 'total_bavail' check. This
1273 will result in an -EAGAIN and in the exit call path we call
1274 alloc_rr() which will with increasing aggressiveness attempt to
1275 aquire precreated objects on the minimum number of required OSCs.
1279 Description: Use old size assignment to avoid deadlock
1280 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
1281 the scheduling while holding a spinlock. We do not need locking
1282 for size in ll_update_inode() because size is only updated from
1283 the MDS for directories or files without objects, so there is no
1284 other place to do the update, and concurrent access to such inodes
1285 are protected by the inode lock.
1289 Description: resolve "_IOWR redefined" build error on SLES10
1293 Description: dump the memory debugging after all modules are unloaded to
1294 suppress false negative in conf_sanity test 39
1296 Severity : enhancement
1298 Description: build kernel-ib packages for OFED 1.3 in our release cycle
1303 Description: fix SLES kernel versioning
1304 Details : the kernel version for our SLES 10 kernel did not include a "-"
1305 before the "smp" at the end. while this was not a problem in
1306 general, it did mean that software trying to use the kernel
1307 version to try to detect a vendor specific kernel would fail.
1308 this was most evident by the OFED build scripts.
1312 Description: Don't update lov_desc members until making sure they are valid
1313 Details : When updating lov_desc members via proc fs, need fix their
1314 validities before doing the real update.
1318 Description: don't put request into delay list while invalidate in flight.
1319 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
1320 import in flight. this produce timeout for invalidate and sometimes
1321 can cause stale data.
1325 Frequency : on ppc only
1326 Description: not convert ost objects for directory because it's not exist.
1327 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
1331 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
1333 Severity : enhancement
1335 Description: Update to support 2.6.22.14 vanilla kernel.
1339 Frequency : rare, on recovery
1340 Description: read procfs can produce deadlock in some situation
1341 Details : Holding lprocfs lock which send rpc can produce block for destroy
1342 obd objects and this also block reconnect with -EALREADY. This
1343 isn't fix all lprocfs bugs - but make it rare.
1345 Severity : enhancement
1347 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
1350 Frequency : frequent on X2 node
1352 Description: mdc_set_open_replay_data LBUG
1353 Details : Set replay data for requests that are eligible for replay.
1357 Description: lustre_mgs: operation 101 on unconnected MGS
1358 Details : When MGC is disconnected from MGS long enough, MGS will evict the
1359 MGC, and late on MGC cannot successfully connect to MGS and a lot
1360 of the error messages complaining that MGS is not connected.
1364 Frequency : on network error
1365 Description: panic with double free request if network error
1366 Details : mdc_finish_enqueue is finish request if any network error ocuring,
1367 but it's true only for synchronus enqueue, for async enqueue
1368 (via ptlrpcd) this incorrect and ptlrpcd want finish request
1371 Severity : enhancement
1373 Description: client-side metadata stat-ahead during readdir(directory readahead)
1374 Details : perform client-side metadata stat-ahead when the client detects
1375 readdir and sequential stat of dir entries therein
1378 Frequency : on start mds
1380 Description: Implement get_info(last_id) in obdfilter.
1383 Frequency : occasional
1385 Description: Correctly check stale fid, not start epoch if ost not support SOM
1386 Details : open with flag O_CREATE need set old fid in op_fid3 because
1387 op_fid2 was overwritten with new generated fid, but mds can answer
1388 with one of these two fids and both is not stale. Setattr
1389 incorrectly started an epoch and assume will be called
1390 done_writing, but without SOM done_writing ever being called.
1393 Frequency : rare, depends on device drivers and load
1395 Description: MDS or OSS nodes crash due to stack overflow
1396 Details : Code changes in 1.8.0 increased the stack usage of some functions.
1397 In some cases, in conjunction with device drivers that use a lot
1398 of stack the MDS (or possibly OSS) service threads could overflow
1399 the stack. One change which was identified to consume additional
1400 stack has been reworked to avoid the extra stack usage.
1403 Frequency : occasional
1405 Description: Do not fail import if osc_interpret_create gets -EAGAIN
1406 Details : If osc_interpret_create got -EAGAIN it immediately exits and
1407 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
1408 oscc_has_objects and see OSC has no objests and call
1409 oscc_internal_create to resend create request.
1411 Severity : enhancement
1413 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
1415 Severity : enhancement
1417 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
1420 Frequency : very rare
1422 Description: Fix replay if there is an un-replied request and open
1423 Details : In some cases, older replay request will revert the
1424 mcd->mcd_last_xid on MDS which is used to record the client's
1425 latest sent request.
1427 Severity : enhancement
1429 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
1431 Severity : enhancement
1433 Description: Add rhel5 support to HEAD.
1435 Serverity : enhancement
1437 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
1442 Description: Don't allow skipping OSTs if index has been specified.
1443 Details : Don't allow skipping OSTs if index has been specified, make
1444 locking in internal create lots better.
1448 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
1449 Details : access to bitfield in structure is always rounded to long
1450 and this produce problem with not atomic change any bit.
1454 Description: Lustre make rpms failed.
1455 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
1456 builds Lustre rpms from tarball.
1459 Frequency : rare on shutdown ost
1461 Description: If llog cancel was not send before clean_exports phase, this can
1462 produce deadlock in llog code.
1463 Details : If llog thread has last reference to obd and call class_import_put
1464 this produce deadlock because llog_cleanup_commit_master wait when
1465 last llog_commit_thread exited, but this never success because was
1466 called from llog_commit_thread.
1470 Description: allow userland application know is lost one of stripes.
1471 Details : fill lvb_blocks with error code on ost and return it to
1472 application if error flag found.
1476 Description: NULL lov_tgts causing MDS oops
1477 Details : more safe checks for NULL lov_tgts for avoid oops.
1479 Severity : enhancement
1481 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
1485 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
1486 Details : If an OST is down the MDS will hang indefinitely in
1487 obd_statfs_rqset() waiting for the statfs data. While for
1488 MDS QOS usage of statfs, it should not stuck in waiting.
1490 Severity : enhancement
1492 Description: remote_acl support
1493 Details : Support ACL-based permission check for remote user.
1494 Support setfacl/getfacl for remote user with the utils
1495 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
1496 the system "{s,g}etfacl" utils.
1498 Severity : enhancement
1500 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
1502 Severity : enhancement
1504 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
1508 Description: make mgs_setparam() handle fsname containing dash
1509 Details : fsname containing a dash does not work with lctl conf_param
1511 Severity : enhancement
1513 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
1515 Severity : enhancement
1517 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
1521 Frequency : rare, at shutdown
1522 Description: access already free / zero obd_namespace.
1523 Details : if client_disconnect_export was called without force flag set,
1524 and exist connect request in flight, this can produce access to
1525 NULL pointer (or already free pointer) when connect_interpret
1526 store ocd flags in obd_namespace.
1530 Frequency : only at startup
1531 Description: not alloc memory with spinlock held.
1532 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
1533 if any spinlock held.
1535 Severity : enhancement
1537 Description: make lustre randomly fail allocating memory
1538 Details : Make lustre randomly failed allocating memory for testing purpose.
1540 Severity : enhancement
1542 Description: lost problems with lov objid file
1543 Details : Fixes some scability and access to not inited memory problems
1544 in work with lov objdid file.
1549 Description: lfs find does not continue on file error
1550 Details : Continue other files processing when a file/dir is absent.
1554 Description: Inconsistent usage of lustre_pack_reply()
1555 Details : Standardize the usage of lustre_pack_reply() such that it
1556 always generate a CERROR on failure.
1559 Frequency : occasional
1561 Description: MDS hang or stay in waiting lock
1562 Details : If client receive lock with CBPENDING flag ldlm need send lock
1563 cancel as separate rpc, to avoid situation when cancel request
1564 can't processed due all i/o threads stay in wait lock.
1568 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1572 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1576 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1580 Description: Update to RHEL4 latest kernel.
1582 Severity : enhancement
1584 Description: Build SLES10 patchless client fails
1585 Details : The configure was broken by run ./configure with
1586 --with-linux-obj=.... argument for patchless client. When the
1587 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1588 can't search header adequately. Use absolute path such as
1589 -I($LINUX)/include instead.
1593 Description: interrupt oig_wait produce painc on resend.
1594 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1595 private set, and this produce situation when rq_ptlrpcd_data not
1596 copyed to new allocated request and triggered LBUG on assert
1597 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1598 ptlrpcd set if request is changed and can be safety changed to use
1603 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1604 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1605 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1606 these macros and replace them with their difinitions instead.
1610 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1611 Details : Update which_patch & target file for SLES10 latest kernel.
1614 Description: Remove client patches from SLES 10 kernel.
1615 Details : This causes SLES 10 clients to behave as patchless clients
1616 even on a Lustre-patched (server) kernel.
1618 Severity : enhancement
1620 Description: self-adjustable client's lru lists
1621 Details : use adaptive algorithm for managing client cached locks lru
1622 lists according to current server load, other client's work
1623 pattern, memory activities, etc. Both, server and client
1624 side namespaces provide number of proc tunables for controlling
1627 Severity : enhancement
1629 Description: light-weight GSS support
1630 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
1631 and reduce performance overhead.
1633 Severity : enhancement
1635 Description: Linux keyring support
1636 Details : Support using service of Linux keyring for Lustre GSS internal
1637 context refresh/cache mechanism.
1641 Description: Fix errors in lfs documentation
1642 Details : Fixes man pages
1646 Description: don't use GFP_* in generic Lustre code.
1647 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
1651 Description: obdclass is limited by single OBD_ALLOC(idarray)
1652 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
1656 Description: warnings with build patchless client with vanila 2.6.19 and up
1657 Details : change old ctl_table style and replace ctl_table/ctl_table_header
1658 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
1662 Frequency : Only for SLES
1663 Description: sanity_quota fail test_1
1664 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
1665 file, which makes TSTID[2] inunique.
1669 Frequency : Only for SLES9
1670 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
1671 Details : Update target/ChangeLog/which_patch .
1675 Frequency : Only for test_10 in sanity-quota.sh
1676 Description: a bug in quota test code
1677 Details : checking a test flag in a wrong place
1681 Frequency : Only for test_13 in sanity-quota.sh
1682 Description: a bug in quota test script
1683 Details : shouldn't check overfull allocated quota
1687 Description: jbd statistics
1688 Details : Port older jbd statistics patch for sles10
1690 Severity : enhancement
1692 Description: Add printing inode info into message about error in writepage.
1697 Description: reply_lock_interpret crash due to race with it and lock cancel.
1698 Details : Do not replay locks that are being cancelled. Do not reference
1699 locks by their address during replay, just by their handle.
1703 Frequency : When flocks are used.
1704 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
1705 Details : Flock locks might destroy just granted lock if it could be merged
1706 with another existing flock, this is done in completion handler,
1707 so teach ldlm_cli_enquque_fini that this is a valid case for
1713 Description: Oops in read and write path when failing to allocate lock.
1714 Details : Check if lock allocation failed and return error back.
1718 Description: lstripe command fails for valid OST index
1719 Details : The stripe offset is compared to lov->desc.ld_tgt_count
1720 instead of lov->desc.ld_active_tgt_count.
1724 Description: sanity.sh failed test 103
1725 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
1730 Description: df doesn't work properly if diskfs blocksize != 4K
1731 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
1735 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
1736 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
1738 Severity : enhancement
1740 Description: Adaptive timeouts
1741 Details : RPC timeouts adapt to changing server load and network
1742 conditions to reduce resend attempts and improve recovery time.
1746 Description: llapi_file_create() does not allow some changes
1747 Details : add llapi_file_open() that allows specifying the mode and
1748 open flags, and also returns an open file handle.
1752 Description: merge and cleanup kernel patches.
1753 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
1757 Description: Add journal checksum support.(Kernel part)
1758 Details : The journal checksum feature adds two new flags i.e
1759 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
1760 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
1761 indicates that the commit block contains the checksum for
1762 the blocks described by the descriptor blocks. Now commit
1763 record can be sent to disk without waiting for descriptor
1764 blocks to be written to disk. This behavior is controlled
1765 using JBD2_FEATURE_ASYNC_COMMIT flag.
1769 Description: OSS needs mutliple precreate threads
1770 Details : Add ability to start more than one create thread per OSS.
1774 Description: Sanity.sh test_65k failed: setstripe should have succeeded
1775 Details : execute lfs setstripe on client
1779 Description: mds_obd_create error creating tmp object
1780 Details : When the user sets quota on root, llog will be affected and can't
1781 create files and write files.
1784 Frequency : Always on ia64 patchless client, and possibly others.
1786 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
1787 Details : This allows the patchless client to be loaded on architectures
1788 without this export.
1792 Description: RedHat Update kernel for RHEL5
1793 Details : Add the kernel config file for RHEL5.
1797 Description: RedHat Update kernel for RHEL5
1798 Details : Modify the kernel config file more closer RHEL5.
1802 Description: Build failure against Centos5 (RHEL5)
1803 Details : Define PAGE_SIZE when it isn't present.
1807 Description: test_7 on sanity-quota.sh can't be used on separate machines
1808 Details : add facet to handle it
1812 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
1813 Details : with NFS, the anon dentry's parent was set to itself in
1814 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
1815 corresponding dentry other than use rec->ur_name.
1817 Severity : enhancement
1819 Description: lfs setstripe enhancement
1820 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
1824 Description: enable data checksumming by default
1825 Details : enable checksum by default, allow --disable-checksum
1826 configure option and "-o nochecksum" mount option. Checksums
1827 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
1831 Description: lustre support for RHEL5
1832 Details : Add support for RHEL5.
1836 Description: Client eviction due to failover config
1837 Details : after a connection loss, the lustre client should attempt to
1838 reconnect to the last active server first before trying the
1839 other potential connections.
1843 Description: when mds and osts use different quota unit(32bit and 64bit),
1844 quota will be released repeatly.
1845 Details : void sending multiple quota reqs to mds, which will keep the
1846 status between the reqs.
1850 Description: osts not allocated evenly to files
1851 Details : change the condition to increase offset_idx
1854 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
1856 Description: Data corruption for OSTs that are formatted larger than 2TB
1858 Details : When generating the bio request for lustre file writes the
1859 sector number would overflow a temporary variable before being
1860 used for the IO. The data reads correctly from Lustre (which
1861 will overflow in a similar manner) but other file data or
1862 filesystem metadata may be corrupted in some cases.
1866 Description: Tune the kernel for good SCSI performance.
1867 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
1868 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
1873 Description: rewrite ext2-derived code in obdclass/uuid.c
1874 Details : rewrite inherited code (uuid parsing code from ext2 utils)
1875 from scratch preserving functionality.
1879 Description: "lfs find -obd UUID" prints directories
1880 Details : "lfs find -obd UUID" will return all directory names instead
1881 of just file names. It is incorrect because the directories
1882 do not reside on the OSTs.
1886 Description: use i_size_read and i_size_write in 2.6 port
1887 Details : replace inode->i_size access with i_size_read/write()
1890 Frequency : when removing large files
1892 Description: scheduling issue during removal of large Lustre files
1893 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
1894 It causes scheduling issues when removing large files (17TB in the
1898 Frequency : only with liblustre clients on XT3
1900 Description: evictions taking too long
1901 Details : allow llrd to evict clients directly on OSTs
1904 Frequency : only on ppc
1906 Description: /proc/fs/lustre/devices broken on ppc
1907 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
1908 We should make sure the type of 'index' is loff_t and then cast
1909 explicitly as needed below. Do not assign an explicitly cast
1913 Frequency : only for rhel5
1915 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
1916 Details : Modify the target file & which_kernel.
1918 Severity : enhancement
1920 Description: omit set fsid for export NFS
1921 Details : fix set/restore device id for avoid EMFILE error and mark lustre
1922 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
1926 Frequency : Always, for kernels after 2.6.16
1927 Description: Fix warning idr_remove called for id=.. which is not allocated.
1928 Details : Last kernels save old s_dev before kill super and not allow
1929 to restore from callback - restore it before call kill_anon_super.
1933 Description: buffer overruns could theoretically occur
1934 Details : llapi_semantic_traverse() modifies the "path" argument by
1935 appending values to the end of the origin string, and a
1936 overrun may occur. Adding buffer overrun check in liblustreapi.
1940 Description: Fix error on 'ls .' at the top of the Lustre mount.
1941 Details : Don't revalidate dentry if it is a root dentry.
1945 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
1946 Details : Modify vm-tunables-rhel4.patch.
1950 Description: Kernel config for 2.6.18-vanilla.
1951 Details : Modify targets/2.6-vanilla.target.in.
1952 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
1953 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
1954 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
1955 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
1959 Description: improve handling recoverable errors
1960 Details : if request processig with error which can be recoverable on server
1961 request should be resend, otherwise page released from cache and
1966 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
1967 Details : Modify target file & which_patch.
1968 A flaw was found in the IA32 system call emulation provided
1969 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
1970 value could be stored in the %RAX register, which could trigger an
1971 out-of-bounds system call table access. An untrusted local user
1972 could exploit this flaw to run code in the kernel
1973 (ie a root privilege escalation). (CVE-2007-4573).
1977 Description: O_DIRECT bypasses client statistics.
1978 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
1979 read_ahead_stats were updated. copied the stats section from
1980 osc_send_oap_rpc() into async_internal().
1984 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
1988 Description: change order of libsysio includes
1989 Details : '#include sysio.h' should always come before '#include xtio.h'
1993 Description: handle "serious error: objid * already exists" more gracefully
1994 Details : If LAST_ID value on disk is smaller than the objects existing in
1995 the O/0/d* directories, it indicates disk corruption and causes an
1996 LBUG(). If the object is 0-length, then we should use the existing
1997 object. This will help to avoid a full fsck in most cases.
1999 Severity : enhancement
2001 Description: adapt the lustre_config script to support the upgrade case
2002 Details : Add "-u" option for lustre_config script to support upgrading 1.4
2003 server targets to 1.6 in parallel.
2007 Description: To avoid grant space > avaible space when the disk is almost
2008 full. Without this patch you might see the error "grant XXXX >
2009 available" or some LBUG about grant, when the disk is almost
2011 Details : In filter_check_grant, for non_grant cache write, we should
2012 check the left space by if (*left > ungranted + bytes), instead
2013 of (*left > ungranted), because only we are sure the left space
2014 is enough for another "bytes", then the ungrant space should be
2015 increase. In client, we should update cl_avail_grant only there
2016 is OBD_MD_FLGRANT in the reply.
2020 Description: Update RHEL 4 kernel to fix local root privilege escalation.
2021 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
2022 described in CVE-2007-4573. This problem could allow untrusted
2023 local users to gain root access.
2026 Frequency : when using O_DIRECT and quotas
2028 Description: Incorrect file ownership on O_DIRECT output files
2029 Details : block usage reported by 'lfs quota' does not take into account
2030 files that have been written with O_DIRECT.
2035 Description: touch file failed when fs is not full
2036 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
2037 otherwise we can get ENOSP while fs is not full.
2041 Description: parallel lock callbacks
2042 Details : Instead of sending blocking and completion callbacks as separated
2043 requests, adding them to a set and sending in parallel.
2046 Frequency : only for Cray XT3
2047 Bugzilla : 12829/13455
2048 Description: Changing primary group doesn't change the group lustre assigns to
2050 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
2051 overridden with the primary group provided by the group upcall,
2052 whereas the supplied fsgid can be trusted if it is in the list of
2053 supplementary groups returned by the group upcall.
2055 Severity : enhancement
2057 Description: Allow masking D_WARNING, D_ERROR messages from console
2058 Details : Console messages can now be disabled via lnet.printk.
2062 Description: User code with malformed file open parameter crashes client node
2063 Details : Before packing join_file req, all the related reference should be
2064 checked carefully in case some malformed flags cause fake
2065 join_file req on client.
2069 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
2070 Details : ldlm_completion_ast() assumes that a lock is granted when the req
2071 mode is equal to the granted mode. However, it should also check
2072 that LDLM_FL_CP_REQD is not set.
2076 Description: Heavy nfs access might result in deadlocks
2077 Details : After ELC code landed, it is now improper to enqueue any mds
2078 locks under och_sem, because enqueue might want to decide to
2079 cancel open locks for same inode we are holding och_sem for.
2083 Description: Client eviction while running blogbench
2084 Details : A lot of unlink operations with concurrent I/O can lead to a
2085 deadlock causing evictions. To address the problem, the number of
2086 oustanding OST_DESTROY requests is now throttled to
2087 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
2088 callbacks are processed in priority.
2092 Description: enable ACLs on MDS by default
2093 Details : ACLs must be enabled on MDS by default.
2096 Frequency : PPC/PPC64 only
2098 Description: conflicts between asm-ppc64/types.h and lustre_types.h
2099 Details : fix duplicated definitions between asm-ppc64/types.h and
2100 lustre_types.h on PPC.
2103 Frequency : PPC/PPC64 only
2105 Description: asm-ppc/segment.h does not exist
2106 Details : fix compile issue on PPC.
2110 Description: better handle error messages in extents code
2113 Frequency : RHEL4 only
2115 Description: mkfs is very slow on IA64/RHEL4
2116 Details : A performance regression has been discovered in the MPT Fusion
2117 driver between versions 3.02.73rh and 3.02.99.00rh. As a
2118 consequence, we have downgraded the MPT Fusion driver in the RHEL4
2119 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
2121 Severity : enhancement
2123 Description: SNMP support enhancement
2124 Details : Adding total number of sampled request for an MDS node in snmp
2127 Severity : enhancement
2129 Description: Optimize ldlm waiting list processing for PR extent locks
2130 Details : When processing waiting list for read extent lock and meeting read
2131 lock that is same or wider to it that is not contended, skip
2132 processing rest of the list and immediatelly return current
2133 status of conflictness, since we are guaranteed there are no
2134 conflicting locks in the rest of the list.
2138 Description: Time out and refuse to reconnect
2139 Details : When the failover node is the primary node, it is possible
2140 to have two identical connections in imp_conn_list. We must
2141 compare not conn's pointers but NIDs, otherwise we can defeat
2142 connection throttling.
2146 Description: port llog fixes from b1_6 into HEAD
2147 Details : Port llog reference couting and some llog cleanups from b1_6
2148 (bug 10800) into HEAD, for protect from panic and access to already
2149 free llog structures.
2153 Description: Detect stride IO mode in read-ahead
2154 Details : When a client does stride read, read-ahead should detect that and
2155 read-ahead pages according to the detected stride pattern.
2159 Description: data checksumming impacts single node performance
2160 Details : add support for several checksum algorithm. Currently, only CRC32
2161 and Adler-32 are supported. The checksum type can be changed on
2162 the fly via /proc/fs/lustre/osc/*/checksum_type.
2166 Description: use adler32 for page checksums
2167 Details : when available, use the Adler-32 algorithm instead of CRC32 for
2172 Description: build for x2 fails
2173 Details : fix compile issue on Cray systems.
2177 Description: Properly match for duplicate locks
2178 Details : Due to different lock order from skiplists code, we need to
2179 traverse entire list for now
2182 Frequency : only on PPC/SLES10
2184 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
2185 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
2186 be defined. Add a hack in mkfs_lustre.c to work around this compile
2191 Description: LASSERT on MDS when client holding flock lock dies
2192 Details : ldlm pool logic depends on number of granted locks equal to
2193 number of released locks which is not true for flock locks, so
2194 just exclude such locks from consideration.
2198 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
2199 Details : Use fsfilt_sync() for both the whole filesystem sync and
2200 individual file sync to eliminate dangerous inode locking
2201 with I_LOCK that can lead to a deadlock.
2205 Description: performance in 1.6.3
2206 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
2207 sure that 1MB requests are not fragmented by the block layer.
2209 Severity : enhancement
2211 Description: organize the server-side client stats on per-nid basis
2212 Details : Change the structure of stats under obdfilter and mds to
2220 The "uuid"s file would list the uuids of _active_ exports.
2221 And the clear entry is to clear all stats and stale nids.
2223 Severity : enhancement
2225 Description: eliminate client locks in face of contention
2226 Details : file contention detection and lockless i/o implementation
2227 for contended files.
2231 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
2234 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
2236 Description: MDS may lose file striping (and hence file data) in some cases
2237 Details : If there are additional extended attributes stored on the MDS,
2238 in particular ACLs, SELinux, or user attributes (if user_xattr
2239 is specified for the client mount options) then there is a risk
2240 of attribute loss. Additionally, the Lustre file striping
2241 needs to be larger than default (e.g. striped over all OSTs),
2242 and an additional attribute must be stored initially in the
2243 inode and then increase in size enough to be moved to the
2244 external attribute block (e.g. ACL growing in size) for file
2249 Description: skiplist implementation simplification
2250 Details : skiplists are used to group compatible locks on granted list
2251 that was implemented as tracking first and last lock of each
2252 lock group the patch changes that to using doubly linked lists
2256 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
2257 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
2258 because LASSERT on both the data supplied by a client, and the
2259 data on disk is dangerous and incorrect.
2261 Severity : enhancement
2263 Description: Slow truncate/writes to huge files at high offsets.
2264 Details : Directly associate cached pages to lock that protect those pages,
2265 this allows us to quickly find what pages to write and remove
2266 once lock callback is received.
2270 Description: more ldlm soft lockups
2271 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
2272 starve other threads from the resource lock for a long time in
2273 case of long waiting queue, so change the debug level from
2274 D_OTHER to the less frequently used D_INFO.
2276 Severity : enhancement
2278 Description: add -gid, -group, -uid, -user options to lfs find
2282 Description: Hung threads in invalidate_inode_pages2_range
2283 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
2284 once one is completed. As a result, some RPCs are stuck in the
2285 queue and are never sent.
2289 Description: filter threads hungs on waiting journal commit
2290 Details : Cleanup filter group llog code, then only filter group llog will
2291 be only created in the MDS/OST syncing process.
2295 Description: Procfs and llog threads access destoryed import sometimes.
2296 Details : Sync the import destoryed process with procfs and llog threads by
2297 the import refcount and semaphore.
2299 Severity : enhancement
2301 Description: openlock cache of b1_6 port to HEAD
2306 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
2307 Details : If insertion of an extent fails, then discard the inode
2308 preallocation and free data blocks else it can lead to duplicate
2313 Description: don't always update ctime in ext3_xattr_set_handle()
2314 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
2315 In some cases the ctime should not be updated, for example for
2316 2.0->1.8 compatibility it is necessary to delete an xattr and it
2317 should not update the ctime.
2321 Bugzilla : 15713/16362
2322 Description: Assertion in iopen_connect_dentry in 1.6.3
2323 Details : looking up an inode via iopen with the wrong generation number can
2324 populate the dcache with a disconneced dentry while the inode
2325 number is in the process of being reallocated. This causes an
2326 assertion failure in iopen since the inode's dentry list contains
2327 both a connected and disconnected dentry.
2331 Description: assertion failure in ldlm_handle2lock()
2332 Details : fix a race between class_handle_unhash() and class_handle2object()
2333 introduced in lustre 1.6.5 by bug 13622.
2338 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
2339 Details : remove the limit of 256 scsi disks in the sd_iostat patch
2344 Description: oops in sd_iostats_seq_show()
2345 Details : unloading/reloading the scsi low level driver triggers a kernel
2346 bug when trying to access the sd iostat file.
2351 Description: Kernel panics during QLogic driver reload
2352 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
2353 patch, causing memory corruption.
2358 Description: journal_dev option does not work in b1_6
2359 Details : pass mount option during pre-mount.
2361 Severity : enhancement
2363 Description: Add a FIEMAP(FIle Extent MAP) ioctl
2364 Details : FIEMAP ioctl will allow an application to efficiently fetch the
2365 extent information of a file. It can be used to map logical blocks
2366 in a file to physical blocks in the block device.
2370 Description: LDLM soft lockups - improvement
2371 Details : It is be possible to send the lock handle along with each read
2372 or write request because the client is already doing a lock match
2373 itself so there isn't any reason the OST should have to re-do that
2377 Frequency : only on Cray X2
2379 Description: X2 build failures
2380 Details : fix build failures on Cray X2.
2384 Description: xid & resent requests
2385 Details : Initialize RPC XID from clock at startup (randomly if clock is
2388 Severity : enhancement
2390 Description: Add lustre_start utility to start or stop multiple Lustre servers
2395 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
2396 Details : In case of memory pressure, list_del() can be called twice on
2397 req->rq_history_list, causing a kernel oops.
2401 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
2402 Details : fix stack overflow in the distributed lock manager by defering
2403 export eviction after a failed AST to the elt thread instead of
2404 handling it in the dlm interpret routine.
2408 Description: Convert some comments to new format.
2409 Details : Update documenting comments to match doxygen conventions.
2413 Description: Grammar fixes.
2414 Details : A couple of trivial spelling fixes.
2418 Description: OSD_COUNTERS-mandatory
2419 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
2420 to simplify the code.
2424 Description: simplify cmm_device freeing logic.
2425 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
2429 Description: Add lockdep support to dt_object_operations locking interface.
2430 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
2431 parameter indicating lock ordering. Update mdd code to use new
2436 Description: Introduce failloc constants for lockless IO tests.
2437 Details : Add two new failloc constants to test lockless IO. Only one of
2438 them in implemented---another is checked in yet to be landed
2443 Description: Add lockdep support for inode mutex.
2444 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
2445 in the situations where Lustre has to lock more than one inode
2450 Description: Add optional invariants checking support.
2451 Details : Add new LINVRNT() macro, optional on new --enable-invariants
2452 configure switch. This macro is to be used for consistency and
2453 sanity checks that are too expensive to be left in `production'
2458 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
2459 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
2464 Description: Add lockdep support for ldlm_lock and ldlm_resource.
2465 Details : Use spin_lock_nested() in (the only) situation where more than
2466 one ldlm_lock is locked simultaneously. Also, fix possible
2467 dead-lock in ldlm_lock_change_resource() by enforcing particular
2472 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
2473 Details : Instead of specifying each ldlm_lock call-back through separate
2474 parameter, wrap them into struct ldlm_callback_suite.
2478 Description: Kill join_lru obd method and its callers.
2479 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
2480 in memory---a requirement implemented through ->o_join_lru() obd
2481 method in HEAD. Remove this method and its users.
2485 Description: Add asynchronous ldlm ENQUEUE completion handler.
2486 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
2487 case that stock ldlm_completion_ast() cannot handle as it waits
2488 until lock is granted. Introduce new ldlm_completion_ast_async()
2489 for this. Also comment ldlm_completion_ast().
2493 Description: ldlm_error <-> errno conversion.
2494 Details : Add functions to map (rather arbitrary) between LDLM error codes
2495 and standard errno values. CLIO needs this to prevent LDLM specific
2496 constants from escaping ldlm and osc.
2500 Description: Kill unused ldlm_handle2lock_ns() function.
2501 Details : Kill unused ldlm_handle2lock_ns() function.
2505 Description: Add lu_ref support to ldlm_lock
2506 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
2507 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
2508 and ldlm_resource. LDLM interface has to be changed, because code that
2509 releases a reference on a lock, has to "know" what reference this is.
2510 In the most frequent case
2512 lock = ldlm_handle2lock(handle);
2514 LDLM_LOCK_PUT(lock);
2516 no changes are required. When any other reference (received _not_ from
2517 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
2518 instead of LDLM_LOCK_PUT().
2520 Arguably, changes are pervasive, and interface requires some discipline
2521 for proper use. On the other hand, it was very instrumental in finding
2522 a few leaked lock references.
2526 Description: Add ldlm_lock_addref_try().
2527 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
2528 attempts to addref a lock that might be being canceled
2533 Description: Add ldlm_weigh_callback().
2534 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
2535 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
2536 hard-coded `number of pages' logic.
2540 Description: Add lockdep annotations to llog code.
2541 Details : Use appropriately tagged _nested() locking calls in the places
2542 where llog takes more than one ->lgh_lock lock.
2546 Description: Add loi_kms_set().
2547 Details : Wrap kms updates into a helper function.
2551 Description: Constify instances of struct lsm_operations.
2552 Details : Constify instances of struct lsm_operations.
2556 Description: lu_conf support.
2557 Details : On a server, a file system object is uniquely identified
2558 by a fid, which is sufficient to locate and load all object
2559 state (inode). On a client, on the other hand, more data are
2560 necessary instantiate an object. Change lu_object_find() and
2561 friends to take additional `lu_conf' argument describing object.
2562 Typically this includes layout information.
2566 Description: lu_context fixes.
2567 Details : Introduce new lu_context functions that are needed on the client
2568 side, where some system threads (ptlrpcd) are shared by multiple
2569 modules, and so cannot be stopped during module shutdown.
2573 Description: Add start and stop methods to lu_device_type_operations.
2574 Details : Introduce two new methods in lu_device_type_operations, that are
2575 invoked when first instance of a given type is created and last one
2576 is destroyed respectively. This is need by CLIO.
2580 Description: Add lu_ref support to struct lu_device.
2581 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
2582 track leaked references.
2586 Description: Introduce lu_kmem_descr.
2587 Details : lu_kmem_descr and its companion interface allow to create
2588 and destroy a number of kmem caches at once.
2592 Description: Fix lu_object finalization race.
2593 Details : Fix a race between lu_object_find() finding an object and its
2594 concurrent finalization. This race is (most likely) not possible
2595 on the server, but might happen on the client.
2599 Description: Introduce lu_ref interface.
2600 Details : lu_ref is a debugging module allowing to track references to
2601 a given object. It is quite cpu expensive, and has to be
2602 explicitly enabled with --enable-lu_ref. See usage description
2607 Description: Factor lu_site procfs stats into a separate function.
2608 Details : Separate lu_site stats printing code into a separate function
2609 that can be reused on a client.
2613 Description: Constify instances of struct {lu,dt,md}_device_operations.
2614 Details : Constify instances of struct {lu,dt,md}_device_operations.
2618 Description: Introduce struct md_site and move meta-data specific parts of
2619 struct lu_site here.
2620 Details : Move md-specific fields out of struct lu_site into special struct
2621 md_site, so that lu_site can be used on a client.
2625 Description: Kill mdd_lov_destroy().
2626 Details : Remove unused mdd code.
2630 Description: Add st_block checking to multistat.c.
2631 Details : Add st_block checking to multistat.c.
2635 Description: Add lu_ref support to struct obd_device.
2636 Details : Add lu_ref tracking to obd_device.
2640 Description: Kill obd_set_fail_loc().
2641 Details : Remove unused code.
2645 Description: Add special type for ptlrpc_request interpret functions.
2646 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
2647 there. Actual usage will be in CLIO.
2651 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
2652 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
2653 doesn't work with lockdep.
2657 Description: Add rwv.c test program.
2658 Details : New testing program exercising readv(2) and writev(2) (Qian).
2662 Description: Add sendfile.c test program.
2663 Details : New testing program exercising sendfile(2) (Jay).
2667 Description: Ratelimit a message that can be very frequent.
2668 Details : Ratelimit a memory allocation failure message that can
2673 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2674 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2678 Description: Liblustre build fixes.
2679 Details : Liblustre build fixes.
2683 Description: libcfs: add cfs_{need,cond}_resched() interface.
2684 Details : libcfs: add cfs_{need,cond}_resched() definition and
2685 implementations for Linux, NT, and liblustre.
2687 Severity : enhancement
2689 Description: More exported tunables for mballoc
2690 Details : Add support for tunable preallocation window and new tunables for
2691 large/small requests
2695 Description: Detect corruption of block bitmap and checking for preallocations
2696 Details : Checks validity of on-disk block bitmap. Also it does better
2697 checking of number of applied preallocations. When corruption is
2698 found, it turns filesystem readonly to prevent further corruptions.
2702 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
2703 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
2704 window to 0. If it does hit cache miss, and read-pattern is still
2705 stride-io mode, does not reset the stride window, but also does not
2706 increase the stride window length in this case.
2710 Frequency : only for big-endian servers
2711 Description: Check if system is big-endian while mounting fs with extents feature
2712 Details : Mounting a filesystem with extents feature will fail on big-endian
2713 systems since ext3-based ldiskfs is not supported on big-endian
2714 systems. This can be over-riden with "bigendian_extents" mount option.
2716 Severity : enhancement
2718 Description: The root squash functionality
2719 Details : A security feature, which is to prevent users from being able
2720 to mount lustre on their desktop, run as root, and delete
2721 all of the files in the filesystem. The goal is accomplished by
2722 remapping user id (UID) and group id (GID) of the root user to
2723 a UID and GID specified by the system administartor via Lustre
2724 configuration management server (MGS). The functionality also
2725 allows to specify sets of clients for which the remapping does
2730 Description: Excessive recovery window
2731 Details : With AT enabled, the recovery window can be excessively long (6000+
2732 seconds). To address this problem, we no longer use
2733 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
2734 timeout no longer depends on the service time, it is set to
2735 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
2736 time via pb_service_time.
2740 Description: Watchdog triggered on MDS failover
2741 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
2742 the OSTs know that the MDS "UUID" can be reused for the same export
2743 from a different NID, so we do not need to wait for the export to be
2747 Frequency : rare, only if using MMP with Linux RAID
2749 Description: MMP doesn't work with Linux RAID
2750 Details : While using HA for Lustre servers with Linux RAID, it is possible
2751 that MMP will not detect multiple mounts. To make this work we
2752 need to unplug the device queue in RAID when the MMP block is being
2753 written. Also while reading the MMP block, we should read it from
2754 disk and not the cached one.
2756 Severity : enhancement
2758 Description: open file using fid
2759 Details : A file can be opened using just its fid, like
2760 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
2763 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
2766 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
2767 Details : add proper locking for creation of jbd_4k slab cache
2771 Description: MMP check in ext3_remount() fails without displaying any error
2772 Details : When multiple mount protection fails during remount, proper error
2775 Severity : enhancement
2777 Description: Allow stripe size to be up to 4G-64k
2778 Details : Fix math logic to allow large stripe sizes.
2782 Description: add check for >8TB ldiskfs filesystems
2783 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
2784 Don't allow >8TB ldiskfs filesystems to be mounted without
2785 force_over_8tb mount option
2790 Descriptoin: OSS complains "too many blocks" and crashes
2791 Details : In filter_commitrw_write():
2792 1. lquota_chkquota();
2793 2. obdo_from_inode();
2794 3. lquota_pending_commit();
2795 "2." has changed uid and gid in this case, then caused
2796 lquota_chkquota() and lquota_pending_commit() to use
2797 different uid and gid. That is the root of the bug.
2801 Descriptoin: getstripe doesn't report stripe count = -1 sanely
2802 Details : Detect and print out the 'stripe count = -1' case properly.
2806 Description: Out or order replies might be lost on replay
2807 Details : In ptlrpc_retain_replayable_request if we cannot find retained
2808 request with tid smaller than one currently being added, add it
2809 to the start, not end of the list.
2811 --------------------------------------------------------------------------------
2813 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
2815 * Support for kernels:
2816 2.6.5-7.283 (SLES 9),
2817 2.6.9-55.EL (RHEL 4),
2818 2.6.16.46-0.14 (SLES 10),
2819 2.6.18.8 vanilla (kernel.org)
2820 * Client support for unpatched kernels:
2821 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2822 2.6.16 - 2.6.22 vanilla (kernel.org)
2823 * Due to recently discovered recovery problems, we do not recommend
2824 using patchless RHEL 4 clients with this or any earlier release.
2825 * Recommended e2fsprogs version: 1.39.cfs8
2826 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2827 * Starting with this release, the ldiskfs backing filesystem required
2828 by Lustre is now in its own package, lustre-ldiskfs. This package
2829 should be installed. It is versioned separately from Lustre and
2830 may be released separately in future.
2834 Description: block reactivating mgc import until all deactivates complete
2835 Details : Fix race when failing back MDT/MGS to itself (testing)
2838 Frequency : at statup only
2840 Description: mds_lov_synchronize race leads to various problems
2841 Details : simultaneous MDT->OST connections at startup can cause the
2842 sync to abort, leaving the OSC in a bad state.
2844 Severity : enhancement
2846 Description: add optional extra BUILD_VERSION info
2847 Details : add a new environment variable (namely LUSTRE_VERS) which allows
2848 to override the lustre version.
2851 Frequency : 2.6.18 servers only
2853 Description: ll_kern_mount() doesn't release the module reference
2854 Details : The ldiskfs module reference count never drops down to 0
2855 because ll_kern_mount() doesn't release the module reference.
2860 Description: server LBUG when using old ost_num_threads parameter
2861 Details : Accept the old ost_num_threads parameter but warn that it
2862 is deprecated, and fix an off-by-one error that caused an LBUG.
2867 Description: Transient SCSI error results in persistent IO issue
2868 Details : iobuf->dr_error is not reinitialized to 0 between two
2872 Frequency : sometimes when underlying device returns I/O errors
2874 Description: OSTs not going read-only during write failures
2875 Details : OSTs are not remounted read-only when the journal commit threads
2876 get I/O errors because fsfilt_ext3 calls journal_start/stop()
2877 instead of the ext3 wrappers.
2881 Description: poor connect scaling with increasing client count
2882 Details : Don't run filter_grant_sanity_check for more than 100 exports
2883 to improve scaling for large numbers of clients.
2886 Frequency : SLES10 only
2888 Description: sanity-quota.sh quotacheck failed: rc = -22
2889 Details : Quotas cannot be enabled on SLES10.
2892 Frequency : liblustre clients only
2894 Description: getdirentries does not give error when run on compute nodes
2895 Details : getdirentries does not fail when the size specified as an argument
2896 is too small to contain at least one entry
2898 Severity : enhancement
2900 Description: Add LNET router traceability for debug purposes
2901 Details : If a checksum failure occurs with a router as part of the
2902 IO path, the NID of the last router that forwarded the bulk data
2903 is printed so it can be identified.
2908 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
2909 Details : Due to a race condition, liblustre clients were occasionally
2910 evicted incorrectly.
2912 Severity : enhancement
2914 Description: lfs setstripe use optional parameters instead of postional
2917 Severity : enhancement
2919 Description: Nanosecond timestamp support for ldiskfs
2920 Details : The on-disk ldiskfs filesystem has added support for nanosecond
2921 resolution timestamps. There is not yet support for this at
2922 the Lustre filesystem level.
2925 Frequency : during server recovery
2927 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
2928 Details : request with rq_no_resend flag not awake l_wait_event if they get
2932 Frequency : nfs export on patchless client
2934 Description: connectathon hang when test nfs export over patchless client
2935 Details : Disconnected dentry cannot be found with lookup, so we do not need
2936 to unhash it or make it invalid
2939 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
2940 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
2941 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
2946 Description: open req refcounting wrong on reconnect
2947 Details : If reconnect happened between getting open reply from server and
2948 call to mdc_set_replay_data in ll_file_open, we will schedule
2949 replay for unreferenced request that we are about to free.
2950 Subsequent close will crash in variety of ways.
2951 Check that request is still eligible for replay in
2952 mdc_set_replay_data().
2957 Description: disable writes to filesystem when reading health_check file
2958 Details : the default for reading the health_check proc file has changed
2959 to NOT do a journal transaction and write to disk, because this
2960 can cause reads of the /proc file to hang and block HA state
2961 checking on a healthy but otherwise heavily loaded system. It
2962 is possible to return to the previous behaviour during configure
2963 with --enable-health-write.
2965 Severity : enhancement
2967 Description: 64-bit inode version
2968 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
2969 to the inode. This will be required for version-based recovery.
2974 Description: MDS fails to start if a duplicate client export is detected
2975 Details : in some rare cases it was possible for a client to connect to
2976 an MDS multiple times. Upon recovery the MDS would detect this
2977 and fail during startup. Handle this more gracefully.
2979 Severity : enhancement
2981 Description: Add -o localflock option to simulate old noflock
2983 Details : This will achieve local-only flock/fcntl locks
2989 Description: log_commit_thread vs filter_destroy race leads to crash
2990 Details : Take import reference before releasing llog record semaphore
2995 Description: Wrong request locking in request set processing
2996 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
2997 imp_delayed_list, in this place should be used imp_lock.
3000 Frequency : when reconnection
3002 Description: Grant Leak when osc reconnect to OST
3003 Details : When osc reconnect ost, OST(filter) should check whether it
3004 should grant more space to client by comparing fed_grant and
3005 cl_avail_grant, and return the granted space to client instead
3006 of "new granted" space, because client will call osc_init_grant
3007 to update the client grant space info.
3010 Frequency : when client reconnect to OST
3012 Description: Grant Leak when osc do resend and replay bulk write
3013 Details : When osc reconnect to OST, OST(filter)should clear grant info of
3014 bulk write request, because the grant info while be sync between
3015 OSC and OST when reconnect, and we should ignore the grant info
3016 these of resend/replay write req.
3021 Description: Grant space more than available left space sometimes.
3022 Details : When then OST is about to be full, if two bulk writing from
3023 different clients came to OST. Accord the avaliable space of the
3024 OST, the first req should be permitted, and the second one
3025 should be denied by ENOSPC. But if the seconde arrived before
3026 the first one is committed. The OST might wrongly permit second
3027 writing, which will cause grant space > available space.
3030 Frequency : when client is evicted
3032 Description: Grant might be wrongly erased when osc is evicted by OST
3033 Details : when the import is evicted by server, it will fork another
3034 thread ptlrpc_invalidate_import_thread to invalidate the
3035 import, where the grant will be set to 0. While the original
3036 thread will update the grant it got when connecting. So if
3037 the former happened latter, the grant will be wrongly errased
3038 because of this race.
3043 Description: Checking Stale with correct fid
3044 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
3045 to check whether it is stale, because sometimes, we want the
3046 enqueue happened anyway, and op_data.fid2 will not be initialized.
3048 Severity : enhancement
3050 Description: update patchless client
3051 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
3054 Frequency : only with 2.4 kernel
3056 Description: random memory corruption
3057 Details : size of struct ll_inode_info is to big for union inode.u and this
3058 can be cause of random memory corruption.
3063 Description: Memory leak in recovery
3064 Details : Lov_mds_md was not free in an error handler in mds_create_object.
3065 It should also check obd_fail before fsfilt_start, otherwise if
3066 fsfilt_start return -EROFS,(failover mds during mds recovery).
3067 then the req will return with repmsg->transno = 0 and rc = EROFS.
3068 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
3069 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
3070 be freed no matter whether fsfilt_commit success or not.
3073 Frequency : only with huge count clients
3075 Description: Prevents from taking the superblock lock in llap_from_page for
3077 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
3078 for llap_from_page call in ll_removepage prevents from taking the
3079 superblock lock for a soon died page.
3084 Description: Not check open intent error before release open handle
3085 Details : in some rare cases, the open intent error is not checked before
3086 release open handle, which may cause
3087 ASSERTION(open_req->rq_transno != 0), because it tries to release
3088 the failed open handle.
3093 Description: Set cat log bitmap only after create log success.
3094 Details : in some rare cases, the cat log bitmap is set too early. and it
3095 should be set only after create log success.
3099 Description: Accessing a block bevice can re-enable I/O when Lustre is
3100 tearing down a device.
3101 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
3106 Description: service threads may hog cpus when there are a lot of requests
3107 Details : Insert cond_resched to give other threads a chance to use some CPU
3112 Description: the cat log was not initialized in recovery
3113 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
3114 unlink log on mds will not be initialized until mds post
3115 recovery. And also in mds post recovery, the unlink log will
3116 initialization will be done asynchronausly, so there will be race
3117 between add unlink log and unlink log initialization.
3121 Description: brw_stats were being printed incorrectly
3122 Details : brw_stats were being printed as log2 but all of them were not
3123 recorded as log2. Also remove some code duplication arising from
3124 filter_tally_{read,write}.
3128 Frequency : rare, only in recovery.
3129 Description: ASSERTION(req->rq_type != LI_POISON) failed
3130 Details : imp_lock should be held while iterating over imp_sending_list for
3131 prevent destroy request after get timeout in ptlrpc_queue_wait.
3135 Description: replay-single.sh test 52 fails
3136 Details : A lock's skiplist need to be cleanup when it being unlinked
3137 from its resource list.
3141 Description: Short directio read returns full requested size rather than
3143 Details : Direct I/O operations should return actual amount of bytes
3144 transferred rather than requested size.
3146 Severity : enhancement
3148 Description: metadata RPC reduction (e.g. for rm performance)
3149 Details : decrease the amount of synchronous RPC between clients and servers
3150 by canceling conflicing lock before the operation on the client
3151 and packing thier handles into the main operation RPC to server.
3153 Severity : enhancement
3155 Description: Async OSC create to avoid the blocking unnecessarily.
3156 Details : If an OST has no remaining object, system will block on the
3157 creation when it needs to create a new object on this OST. Now,
3158 ways use pre-created objects when available, instead of blocking on
3159 an empty osc while others are not empty. If we must block, we
3160 block for the shortest possible period of time.
3164 Description: improve handling recoverable errors
3165 Details : if request processig with error which can be recoverable on server
3166 request should be resend, otherwise page released from cache and
3169 Severity : enhancement
3171 Description: refine locking for avoid write wrong info into lov_objid file
3172 Details : fix possible races with add new target and write/update data in
3175 --------------------------------------------------------------------------------
3177 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
3182 Frequency : on some architectures
3184 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
3185 Details : Uninitialized flags sometimes cause configuration commands to
3189 Frequency : patchless clients only
3191 Description: missing __iget() symbol export
3192 Details : The __iget() symbol export is missing. To avoid the need for
3193 this on patchless clients the deathrow inode reaper is turned
3194 off, and we depend on the VM to clean up old inodes. This
3195 dependency was during via the fix for bug 12181.
3197 --------------------------------------------------------------------------------
3199 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
3201 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
3202 INTEROPERATE with older versions automatically. In many cases a
3203 special upgrade step is needed. Please read the
3204 user documentation before upgrading any part of a live system.
3205 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
3206 version will not interoperate with 1.6 betas before beta5 (1.5.95).
3207 * WARNING: Lustre configuration and startup changes are required with
3208 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
3212 Severity : enhancement
3214 Description: MountConf
3215 Details : Lustre configuration is now managed via mkfs and mount
3216 commands instead of lmc and lconf. New obd types (MGS, MGC)
3217 are added for dynamic configuration management. See
3218 https://mail.clusterfs.com/wikis/lustre/MountConf for
3221 Severity : enhancement
3223 Description: dynamic OST addition
3224 Details : OSTs can now be added to a live filesystem
3226 Severity : enhancement
3228 Description: startup order invariance
3229 Details : MDTs and OSTs can be started in any order. Clients only
3230 require the MDT to complete startup.
3232 Severity : enhancement
3234 Description: parallel, asynchronous orphan cleanup
3235 Details : orphan cleanup is now performed in separate threads for each
3236 OST, allowing parallel non-blocking operation.
3238 Severity : enhancement
3240 Description: optimized stripe assignment
3241 Details : stripe assignments are now made based on ost space available,
3242 ost previous usage, and OSS previous usage, in order to try
3243 to optimize storage space and networking resources.
3245 Severity : enhancement
3247 Description: Permanently set tunables
3248 Details : All writable /proc/fs/lustre tunables can now be permanently
3249 set on a per-server basis, at mkfs time or on a live system.
3251 Severity : enhancement
3253 Description: Lustre message v2
3254 Details : Add lustre message format v2.
3256 Severity : enhancement
3258 Description: client OST exclusion list
3259 Details : Clients can be started with a list of OSTs that should be
3260 declared "inactive" for known non-responsive OSTs.
3264 Description: SPEC SFS validation failure on NFS v2 over lustre.
3265 Details : Changes the blocksize for regular files to be 2x RPC size,
3266 and not depend on stripe size.
3268 Severity : enhancement
3270 Description: Multiple MD RPCs in flight.
3271 Details : Further unserialise some read-only MDS RPCs - learn about intents.
3272 To avoid overly-overloading MDS, introduce a limit on number of
3273 MDS RPCs in flight for a single client and add /proc controls
3274 to adjust this limit.
3276 Severity : enhancement
3278 Description: client read/write statistics
3279 Details : Add client read/write call usage stats for performance
3280 analysis of user processes.
3281 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
3282 file access. extents_stats shows chunk size distribution.
3283 extents_stats_per_process show chunk size distribution per
3286 Severity : enhancement
3288 Description: mds statistics
3289 Details : Add detailed mds operations statistics in
3290 /proc/fs/lustre/mds/*/stats.
3294 Description: Failure of copying files with lustre special EAs.
3295 Details : Client side always return success for setxattr call for lustre
3296 special xattr (currently only "trusted.lov").
3301 Description: Refcount LNET uuids
3302 Details : The global LNET uuid list grew linearly with every startup;
3303 refcount repeated list entries instead of always adding to
3306 Severity : enhancement
3308 Description: Dynamic service threads
3309 Details : Within a small range, start extra service threads
3310 automatically when the request queue builds up.
3313 Frequency : mixed-endian client/server environments
3315 Description: mixed-endian crashes
3316 Details : The new msg_v2 system had some failures in mixed-endian
3319 Severity : enhancement
3321 Description: Easy OST removal
3322 Details : OSTs can be permanently deactivated with e.g. 'lctl
3323 conf_param lustre-OST0001.osc.active=0'
3325 Severity : enhancement
3327 Description: MGS proc entries
3328 Details : Added basic proc entries for the MGS showing what filesystems
3331 Severity : enhancement
3333 Description: provide MGS failover
3334 Details : Added config lock reacquisition after MGS server failover.
3336 Severity : enhancement
3338 Description: add Linux 2.4 support
3339 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
3343 Description: a large application tries to do I/O to the same resource and dies
3344 in the middle of it.
3345 Details : Check the req->rq_arrival time after the call to
3346 ost_brw_lock_get(), but before we do anything about
3347 processing it & sending the BULK transfer request. This
3348 should help move old stale pending locks off the queue as
3349 quickly as obd_timeout.
3352 Frequency : when an incorrect nid is specified during startup
3354 Description: ptlrpc connect to non-existant node causes kernel crash
3355 Details : LNET can't be re-entered from an event callback, which
3356 happened when we expire a message after the export has been
3357 cleaned up. Instead, hand the zombie cleanup off to another
3360 Severity : enhancement
3362 Description: plain/inodebits lock performance improvement
3363 Details : Grouping plain/inodebits in granted list by their request modes
3364 and bits policy, thus improving the performance of search through
3368 Frequency : only if OST filesystem is corrupted
3370 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3371 Details : for a short time RPCs with bulk IO are in the replay list,
3372 but replay of bulk IOs is unimplemented. If the OST filesystem
3373 is corrupted due to disk cache incoherency and then replay is
3374 started it is possible to trip an assertion. Avoid putting
3375 committed RPCs into the replay list at all to avoid this issue.
3378 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3380 Description: System hang on startup
3381 Details : This bug allowed the liblustre (e.g. catamount) client to
3382 return to the app before handling all startup RPCs. This
3383 could leave the node unresponsive to lustre network traffic
3384 and manifested as a server ptllnd timeout.
3386 Severity : enhancement
3388 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3389 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
3390 boolean to control whether to print peer debug info when a
3391 client's RPC times out.
3394 Frequency : only for kernels with patches from Lustre below 1.4.3
3396 Description: Remove old rdonly API
3397 Details : Remove old rdonly API which unused from at least lustre 1.4.3
3400 Frequency : only for devices with external journals
3402 Description: Set external device read-only also
3403 Details : During a commanded failover stop, we set the disk device
3404 read-only while the server shuts down. We now also set any
3405 external journal device read-only at the same time.
3408 Frequency : when upgrading from 1.4 while trying to change parameters
3410 Description: The wrong (new) MDC name was used when setting parameters for
3411 upgraded MDT's. Also allows changing of OSC (and MDC)
3412 parameters if --writeconf is specified at tunefs upgrade time.
3415 Frequency : when setting specific ost indicies
3417 Description: QOS code breaks on skipped indicies
3418 Details : Add checks for missing OST indicies in the QOS code, so OSTs
3419 created with --index need not be sequential.
3421 Severity : enhancement
3423 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3424 Details : The uninit_groups feature works in conjunction with the kernel
3425 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3426 up the pass1 processing of e2fsck. This is a read-only feature
3427 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3428 mount filesystems that have had this feature enabled.
3430 Severity : enhancement
3432 Description: Improve multi-block allocation algorithm to avoid fragmentation
3433 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3434 allocation locality and avoid filesystem fragmentation.
3436 ------------------------------------------------------------------------------
3438 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
3440 * Support for kernels:
3441 2.4.21-47.0.1.EL (RHEL 3)
3442 2.6.5-7.283 (SLES 9)
3443 2.6.9-42.0.10.EL (RHEL 4)
3444 2.6.12.6 vanilla (kernel.org)
3445 2.6.16.27-0.9 (SLES 10)
3446 * Recommended e2fsprogs version: 1.39.cfs5
3448 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3452 Frequency : occasional, depends on client load and configuration
3453 Bugzilla : 12181, 12203
3454 Description: data loss for recently-modified files
3456 Details : In some cases it is possible that recently written or created
3457 files may not be written to disk in a timely manner (this should
3458 normally be within 30s unless client IO load is very high).
3459 The problem appears as zero-length files or files that are a
3460 multiple of 1MB in size after a client crash or client eviction
3461 that are missing data at the end of the file.
3463 This problem is more likely to be hit on clients where files are
3464 repeatedly created and unlinked in the same directory, clients
3465 have a large amount of RAM, have many CPUs, the filesystem has
3466 many OSTs, the clients are rebooted frequently, and/or the files
3467 are not accessed by other nodes after being written.
3469 The presence of the problem can be detected by looking at
3470 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
3471 smaller than the second (nr_unused) then dirty files will not
3472 be flushed automatically to disk. "sync; sleep 10" should be
3473 run several times on the node before unmounting it to update
3474 Lustre (this is also safe to run on nodes without this problem).
3476 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
3477 that can cause this same problem, so customers using that kernel
3478 also need to update the kernel in addition to Lustre. In order
3479 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
3481 It is normal that files written just before a client crash (less
3482 than 30s) may not yet have been flushed to disk, even for local
3486 Frequency : frequent on thin XT3 nodes
3488 Description: UUID collision on thin XT3 Linux nodes
3489 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
3490 randomly, since we relied on an insufficiently-seeded PRNG.
3495 Description: OSS hangs after "All ost request buffers busy"
3496 Details : A deadlock between quota and journal operations caused OSS
3497 hangs after printing "All ost request buffers busy."
3500 Frequency : always on liblustre builds
3502 Description: Cleanup compiler warnings on liblustre
3505 Frequency : always on liblustre builds on XT3
3507 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
3511 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
3512 > 32000 subdirectories
3513 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
3514 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
3515 subdirectory count crosses 32000. This will aid e2fsck to
3516 correctly handle more than 32000 subdirectories.
3519 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3521 Description: System hang on startup
3522 Details : This bug allowed the liblustre (e.g. catamount) client to
3523 return to the app before handling all startup RPCs. This
3524 could leave the node unresponsive to lustre network traffic
3525 and manifested as a server ptllnd timeout.
3527 Severity : enhancement
3529 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3530 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
3531 boolean to control whether to print peer debug info when a
3532 client's RPC times out.
3537 Description: make O_SYNC working on 2.6 kernels
3538 Details : 2.6 kernels use different method for mark pages for write,
3539 so need add a code to lustre for O_SYNC work.
3544 Description: Failure to close file and release space on NFS
3545 Details : Put inode details into lock acquired in ll_intent_file_open.
3546 Use mdc_intent_lock in ll_intent_open to properly
3547 detect all kind of errors unhandled by mdc_enqueue.
3552 Description: proc file read during shutdown sometimes raced obd removal,
3554 Details : Add lock to prevent obd access after proc file removal.
3557 Frequency : Only for files larger than 4GB on 32-bit clients.
3559 Description: improperly doing page alignment of locks
3560 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
3561 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3566 Description: RPCs being resent when they shouldn't be
3567 Details : Some RPCs that should not be resent are being resent. This
3568 can cause inconsistencies in the RPC state machine. Do not
3569 resend such requests.
3572 Frequency : rare, only with NFS export
3574 Description: Crash on NFS re-export node
3575 Details : under very unusual load conditions an assertion is hit in
3576 ll_intent_file_open()
3579 Frequency : only if OST filesystem is corrupted
3581 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3582 Details : for a short time RPCs with bulk IO are in the replay list,
3583 but replay of bulk IOs is unimplemented. If the OST filesystem
3584 is corrupted due to disk cache incoherency and then replay is
3585 started it is possible to trip an assertion. Avoid putting
3586 committed RPCs into the replay list at all to avoid this issue.
3591 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3592 Details : Large single O_DIRECT read and write calls can fail to allocate
3593 a sufficiently large buffer to process the request. In case of
3594 allocation failure the allocation is retried with a smaller
3595 buffer and broken into smaller requests.
3597 Severity : enhancement
3599 Description: Add -o localflock option to simulate old noflock behaviour.
3600 Details : This will achieve local-only flock/fcntl locks coherentness.
3605 Description: versioning check is incomplete
3606 Details : Checking the version difference of client vs. server, report
3607 error if the gap is too big.
3612 Description: add support PG_writeback bit
3613 Details : add support for PG_writeback bit for Lustre, for more carefull
3614 work with page cache in 2.6 kernel. This also fix some deadlocks
3615 and remove hack for work O_SYNC with 2.6 kernel.
3617 Severity : enhancement
3619 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3620 Details : The uninit_groups feature works in conjunction with the kernel
3621 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3622 up the pass1 processing of e2fsck. This is a read-only feature
3623 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3624 mount filesystems that have had this feature enabled.
3626 Severity : enhancement
3628 Description: Improve multi-block allocation algorithm to avoid fragmentation
3629 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3630 allocation locality and avoid filesystem fragmentation.
3632 ------------------------------------------------------------------------------
3634 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3636 * Support for kernels:
3637 2.6.9-42.0.3.EL (RHEL 4)
3638 2.6.5-7.276 (SLES 9)
3639 2.4.21-47.0.1.EL (RHEL 3)
3640 2.6.12.6 vanilla (kernel.org)
3641 2.6.16.21-0.8 (SLES10)
3642 * Recommended e2fsprogs version: 1.39.cfs2-0
3644 * The backwards-compatible /proc/sys/portals symlink has been removed
3645 in this release. Before upgrading, please ensure that you change
3646 any configuration scripts or /etc/sysctl.conf files that access
3647 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3648 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3649 in advance of the upgrade on any system running Lustre 1.4.6 or
3650 newer, since /proc/sys/lnet was added in that version.
3651 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3655 Frequency : only when quota is used
3657 Description: avoid scanning export list for quota master
3658 Details : Change the algorithms to avoid scanning export list in order
3659 to improve the efficiency.
3662 Frequency : MDS failover only, very rarely
3664 Description: "went back in time" messages on mds failover
3665 Details : The greatest transno may be lost when the current operation
3666 finishes with an error (transno==0) and the client's last_rcvd
3667 record is over-written. Save the greatest transno in the
3668 mds_last_transno for this case.
3671 Frequency : always for specific kernels and striping counts
3673 Description: client may get "Matching packet too big" without ACL support
3674 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
3675 when trying to access files in certain configurations. The
3676 clients should in fact be denied when mounting because they do
3677 not understand ACLs.
3680 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
3682 Description: many clients connecting with IO in progress causes connect timeouts
3683 Details : Avoid synchronous journal commits to avoid delays caused by many
3684 clients connecting/disconnecting when bulk IO is in progress.
3685 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
3686 OST_IO_PORTAL to avoid delays behind potentially many pending
3690 Frequency : occasionally with multiple writers to a single file
3692 Description: shared writes to file may result in wrong size reported by stat()
3693 Details : Allow growing of kms when extent lock is cancelled
3696 Frequency : always with random mmap IO to multi-striped file
3698 Description: mmap write might be lost if we are writing to a 'hole' in stripe
3699 Details : Only if the hole is at the end of OST object so that kms is too
3700 small. Fix is to increase kms accordingly in ll_nopage.
3703 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
3705 Description: writes to a missing object would leak memory on the OST
3706 Details : If there is an inconsistency between the MDS and OST filesystems,
3707 such that the MDS references an object that doesn't exist, writes
3708 to that object will leak memory due to incorrect cleanup in the
3709 error handling path, eventually running out of memory on the OST.
3714 Description: Creating too long symlink causes lustre errors
3715 Details : Check symlink and name lengths before sending requests to MDS.
3718 Frequency : only if flock is enabled (not on by default)
3720 Description: posix locks not released on fd closure on 2.6.9+
3721 Details : We failed to add posix locks to list of inode locks on 2.6.9+
3722 kernels, this caused such locks not to be released on fd close and
3723 then assertions on fs unmount about still used locks.
3726 Frequency : MDS failover only, very rarely
3728 Description: clients may get ASSERTION(granted_lock != NULL)
3729 Details : When request was taking a long time, and a client was resending
3730 a getattr by name lock request. The were multiple lock requests
3731 with the same client lock handle and
3732 mds_getattr_name->fixup_handle_for_resent_request found one of the
3733 lock handles but later failed with ASSERTION(granted_lock != NULL).
3738 Description: handle->h_buffer_credits > 0, assertion failure
3739 Details : h_buffer_credits is zero after truncate, causing assertion
3740 failure. This patch extends the transaction or creates a new
3744 Frequency : NFS re-export or patchless client
3745 Bugzilla : 11179, 10796
3746 Description: Crash on NFS re-export node (__d_move)
3747 Details : We do not want to hash the dentry if we don't have a lock.
3748 But if this dentry is later used in d_move, we'd hit uninitialised
3749 list head d_hash, so we just do this to init d_hash field but
3750 leave dentry unhashed.
3753 Frequency : NFS re-export or patchless client
3755 Description: NFS exports has problem with symbolic link
3756 Details : lustre client didn't properly install dentry when re-exported
3757 to NFS or running patchless client.
3760 Frequency : NFS re-export or patchless client
3762 Description: Various nfs/patchless fixes.
3763 Details : fixes reuse disconected alias for lookup process - this fixes
3764 warning "find_exported_dentry: npd != pd",
3765 fix permission error with open files at nfs.
3766 fix apply umask when do revalidate.
3769 Frequency : occasional
3771 Description: Crash on NFS re-export node
3772 Details : calling clear_page() on the wrong pointer triggered oops in
3773 generic_mapping_read().
3776 Frequency : rarely, using O_DIRECT IO
3778 Description: unaligned directio crashes client with LASSERT
3779 Details : check for unaligned buffers before trying any requests.
3782 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
3784 Description: stale data returned from RAID cache
3785 Details : If only a small amount of IO is done to the RAID device before
3786 reading it again it is possible to get stale data from the RAID
3787 cache instead of reading it from disk.
3790 Frequency : always for sles10 kernel
3792 Description: sles10 support
3793 Details : ll_follow_link: compile fixes and using of nd_set_link
3794 under newer kernels.
3797 Frequency : depends on arch, kernel and compiler version, always on sles10
3800 Description: recursive or deep enough symlinks cause stack overflow
3801 Details : getting rid of large stack-allocated variable in
3805 Frequency : depends on hardware
3807 Description: lustre write performance loss in the SLES10 kernel
3808 Details : the performance loss is caused by using of write barriers in the
3809 ext3 code. The SLES10 kernel turns barrier support on by
3810 default. The fix is to undo that change for ldiskfs.
3812 ------------------------------------------------------------------------------
3814 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
3816 * Support for kernels:
3817 2.6.9-42.0.3EL (RHEL 4)
3818 2.6.5-7.276 (SLES 9)
3819 2.4.21-47.0.1.EL (RHEL 3)
3820 2.6.12.6 vanilla (kernel.org)
3824 Frequency : quota enabled and large files being deleted
3826 Description: releasing more than 4GB of quota at once hangs OST
3827 Details : If a user deletes more than 4GB of files on a single OST it
3828 will cause the OST to spin in an infinite loop. Release
3829 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3834 Description: statfs data retrieved from /proc may be stale or zero
3835 Details : When reading per-device statfs data from /proc, in the
3836 {kbytes,files}_{total,free,avail} files, it may appear
3837 as zero or be out of date.
3840 Frequency : systems with MD RAID1 external journal devices
3842 Description: lconf's call to blkid is confused by RAID1 journal devices
3843 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3844 of returning all block devices that match the journal UUID.
3847 Frequency : always, for aggregate stripe size over 4GB
3849 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
3850 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
3851 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
3854 Frequency : always if "lfs find" used on a local file/directory
3856 Description: "lfs find" segfaults if used on a local file/directory
3857 Details : The case where a directory component was not specified wasn't
3858 handled correctly. Handle this properly.
3861 Frequency : always on ppc64
3863 Description: the write to an ext3 filesystem mounted with mballoc got stuck
3864 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3865 perform endianness conversion.
3868 Frequency : rarely (truncate to non-zero file size after write under load)
3869 Bugzilla : 10730, 10687
3870 Description: Files padded with zeros to next 4K multiple
3871 Details : With filesystems mounted using the "extents" option (2.6 kernels)
3872 it is possible that files that are truncated to a non-zero size
3873 immediately after being written are filled with zero bytes beyond
3874 the truncated size. No file data is lost.
3876 Severity : enhancement
3878 Description: Allow recovery/failover for liblustre clients.
3879 Details : liblustre clients were unaware of failover configurations until
3882 Severity : enhancement
3884 Description: user file locks should fail when not mounting with flock option
3885 Details : Set up an error-returning stub in ll_file_operations.lock field
3886 to prevent incorrect behaviour when client is mounted without
3887 flock option. Also, set up properly f_op->flock field for
3891 Frequency : always on ia64
3893 Description: "lfs df" loops on printing out MDS statfs information
3894 Details : The obd_ioctl_data was not initialized and in some systems
3895 this caused a failure during the ioctl that did not return
3896 an error. Initialize the struct and return an error on failure.
3899 Frequency : SLES 9 only
3901 Description: Error of copying files with lustre special EAs as root
3902 Details : Client side always return success for setxattr call for lustre
3903 special xattr (currently only "trusted.lov").
3906 Frequency : rarely on clusters with both ia64+i386 clients
3908 Description: ia64+i686 clients doing shared IO on the same file may LBUG
3909 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
3910 clients are doing concurrent writes to the same file it is
3911 possible that the ia64 clients may LASSERT because the OST
3912 extent locks are not PAGE_SIZE aligned. Ensure that grown
3913 locks are always aligned on the request boundary.
3916 Frequency : specific use, occasional
3918 Description: Overwriting in use executable truncates on-disk binary image
3919 Details : If one node attempts to overwrite an executable in use by
3920 another node, we now correctly return ETXTBSY instead of
3921 truncating the file.
3923 Severity : enhancement
3925 Description: Async OSC create to avoid the blocking unnecessarily.
3926 Details : If an OST has no remaining object, system will block on the creating
3927 when need to create a new object on this OST. Now, ways use
3928 pre-created objects when available, instead of blocking on an
3929 empty osc while others are not empty. If we must block, we block
3930 for the shortest possible period of time.
3935 Description: chmod on Lustre root is propagated to other clients
3936 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
3937 invalid by the follow_mount time.
3940 Frequency : liblustre clients only
3942 Description: Race in 'instant cancel' lock handling could lead to such locks
3943 never to be granted in case of SMP MDS
3944 Details : Do not destroy not yet granted but cbpending locks in
3948 Frequency : replay/resend of open
3950 Description: non null lock assetion failure in mds_intent_policy
3951 Details : Trying to replay/resend lockless open requests resulted in
3952 mds_open() returning 0 with no lock. Now it sets a flag if
3953 a lock is going to be returned.
3955 Severity : enhancement
3957 Description: Checksum enhancements
3958 Details : New checksum enhancements allow for resending RPCs that failed
3961 Severity : enhancement
3963 Description: Tunables on number of dirty pages in cacche
3964 Details : Allow to set limit on number of dirty pages cached.
3969 Description: client crash on unmount - lock still has references
3970 Details : In some error handling cases it was possible to leak a lock
3971 reference on a client while accessing a file. This was not
3972 harmful to the client during operation, but would cause the
3973 client to crash when the filesystem is unmounted.
3976 Frequency : specific case, rare
3978 Description: ETXTBSY on mds though file not in use by client
3979 Details : ETXTBSY is no longer incorrectly returned when attempting to
3980 chmod or chown a directory that the user previously tried to
3981 execute or a currently-executing binary.
3984 Frequency : extremely rare except on liblustre-based clients
3986 Description: Lustre space not freed when files are deleted
3987 Details : Clean up open-unlinked files after client eviction. Previously
3988 the unlink was skipped and the files remained as orphans.
3993 Description: OST failure "would be an LBUG" in waiting_locks_callback()
3994 Details : In some cases it was possible to send a blocking callback to
3995 a client doing a glimpse, even though that client didn't get
3996 a lock granted. When the glimpse lock is cancelled on the OST
3997 the freed lock is left on the waiting list and corrupted the list.
4000 Frequency : all core dumps
4002 Description: Broke core dumps to lustre
4003 Details : Negative dentry may be unhashed if parent does not have UPDATE
4004 lock, but some callers, e.g. do_coredump, expect dentry to be
4005 hashed after successful create, hash it in ll_create_it.
4007 ------------------------------------------------------------------------------
4009 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
4011 * Support for kernels:
4012 2.6.9-42.0.2.EL (RHEL 4)
4013 2.6.5-7.276 (SLES 9)
4014 2.4.21-40.EL (RHEL 3)
4015 2.6.12.6 vanilla (kernel.org)
4019 Frequency : always on RHEL 3
4021 Description: Number of open files grows over time
4022 Details : The number of open files grows over time, whether or not
4023 Lustre is started. This was due to a filp leak introduced
4024 by one of our kernel patches.
4026 ------------------------------------------------------------------------------
4028 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4030 * Support for kernels:
4031 2.6.9-42.EL (RHEL 4)
4032 2.6.5-7.276 (SLES 9)
4033 2.4.21-40.EL (RHEL 3)
4034 2.6.12.6 vanilla (kernel.org)
4039 Bugzilla : 5719, 9635, 9792, 9684
4040 Description: OST (or MDS) trips assertions in (re)connection under heavy load
4041 Details : If a server is under heavy load and cannot reply to new
4042 connection requests before the client resends the (re)connect,
4043 the connection handling code can behave badly if two service
4044 threads are concurrently handing separate (re)connections from
4045 the same client. Add better locking to the connection handling
4046 code, and ensure that only a single connection will be processed
4047 for a given client UUID, even if the lock is dropped.
4049 Severity : enhancement
4051 Description: add TCP zero-copy support to kernel
4052 Details : Add support to the kernel TCP stack to allow zero-copy bulk
4053 sends if the hardware supports scatter-gather and checksumming.
4054 This allows socklnd to do client-write and server-read more
4055 efficiently and reduce CPU utilization from skbuf copying.
4058 Frequency : only if NFS exporting from client
4060 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
4061 Details : If setting attributes on a file created under NFS that had
4062 never been opened it would be possible to oops the client
4063 if the file had no objects.
4066 Frequency : always for liblustre
4068 Description: liblustre client does MDS+OSTs setattr RPC for each write
4069 Details : When doing a write from a liblustre client, the client
4070 incorrectly issued an RPC to the MDS and each OST the file was
4071 striped over in order to update the timestamps. When writing
4072 with small chunks and many clients this could overwhelm the MDS
4073 with RPCs. In all cases it would slow down the write because
4074 these RPCs are unnecessary.
4076 Severity : enhancement
4078 Description: allow number of MDS service threads to be changed at module load
4079 Details : It is now possible to change the number of MDS service threads
4080 running. Adding "options mds mds_num_threads={N}" to the MDS's
4081 /etc/modprobe.conf will set the number of threads for the next
4082 time Lustre is restarted (assuming the "mds" module is also
4083 reloaded at that time). The default number of threads will
4084 stay the same, 32 for most systems.
4089 Description: OST crash if filesystem is unformatted or corrupt
4090 Details : If an OST is started on a device that has never been formatted
4091 or if the filesystem is corrupt and cannot even mount then the
4092 error handling cleanup routines would dereference a NULL pointer.
4097 Description: NULL pointer deref in llap_from_page.
4098 Details : get_cache_page_nowait can return a page with NULL (or otherwise
4099 incorrect) mapping if the page was truncated/reclaimed while it was
4100 searched for. Check for this condition and skip such pages when
4101 doing readahead. Introduce extra check to llap_from_page() to
4102 verify page->mapping->host is non-NULL (so page is not anonymous).
4105 Frequency : Sometimes when using sys_sendfile
4107 Description: "page not covered by a lock" warnings from ll_readpage
4108 Details : sendfile called ll_readpage without right page locks present.
4109 Now we introduced ll_file_sendfile that does necessary locking
4110 around call to generic_file_sendfile() much like we do in
4114 Frequency : with certain MDS communication failures at client mount time
4116 Description: NULL pointer deref after failed client mount
4117 Details : a client connection request may delayed by the network layer
4118 and not be sent until after the PTLRPC layer has timed out the
4119 request. If the client fails the mount immediately it will try
4120 to clean up before the network times out the request. Add a
4121 reference from the request import to the obd device and delay
4122 the cleanup until the network drops the request.
4125 Frequency : occasionally during client (re)connect
4127 Description: assertion failure during client (re)connect
4128 Details : processing a client connection request may be delayed by the
4129 client or server longer than the client connect timeout. This
4130 causes the client to resend the connection request. If the
4131 original connection request is replied in this interval, the
4132 client may trip an assertion failure in ptlrpc_connect_interpret()
4133 which thought it would be the only running connect process.
4136 Frequency : only with obd_echo servers and clients that are rebooted
4138 Description: kernel BUG accessing uninitialized data structure
4139 Details : When running an obd_echo server it did not start the ping_evictor
4140 thread, and when a client was evicted an uninitialized data
4141 structure was accessed. Start the ping_evictor in the RPC
4142 service startup instead of the OBD startup.
4144 Severity : enhancement
4145 Bugzilla : 10193 (patchless)
4146 Description: Remove dependency on various unexported kernel interfaces.
4147 Details : No longer need reparent_to_init, exit_mm, exit_files,
4148 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
4151 Frequency : rare (only users of deprecated and unsupported LDAP config)
4153 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
4154 Details : LDAP apparently contains 'lustreName' attributes instead of
4155 'name'. A simple remapping of the name is sufficient.
4158 Frequency : rare (only with non-default dump_on_timeout debug enabled)
4160 Description: waiting_locks_callback trips kernel BUG if client is evicted
4161 Details : Running with the dump_on_timeout debug flag turned on makes
4162 it possible that the waiting_locks_callback() can try to dump
4163 the Lustre kernel debug logs from an interrupt handler. Defer
4164 this log dumping to the expired_lock_main() thread.
4166 Severity : enhancement
4168 Description: Support NFS exporting on 2.6 kernels.
4169 Details : Implement non-rawops metadata methods for NFS server to use without
4170 changing NFS server code.
4173 Frequency : very rare (synthetic metadata workload only)
4175 Description: two racing renames might cause an MDS thread to deadlock
4176 Details : Running the "racer" program may cause one MDS thread to rename
4177 a file from being the source of a rename to being the target of
4178 a rename at exactly the same time that another thread is doing
4179 so, and the second thread has already enqueued these locks after
4180 doing a lookup of the target and is trying to relock them in
4181 order. Ensure that we don't try to re-lock the same resource.
4184 Frequency : only very large systems with liblustre clients
4186 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
4187 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
4188 all OSTs in parallel. This allows the eviction of stale liblustre
4189 clients to proceed much faster than if they were done in series,
4190 and also offers similar improvements for other set_info RPCs.
4195 Description: excessive CPU usage during initial read phase on client
4196 Details : During the initial read phase on a client, it would agressively
4197 retry readahead on the file, consuming too much CPU and impacting
4198 performance (since 1.4.5.8). Improve the readahead algorithm
4199 to avoid this, and also improve some other common cases (read
4200 of small files in particular, where "small" is files smaller than
4201 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
4206 Description: MDS crash when receiving packet with unknown intent.
4207 Details : Do not LBUG in unknown intent case, just return -EFAULT
4209 Severity : enhancement
4210 Bugzilla : 9293, 9385
4211 Description: MDS RPCs are serialised on client. This is unnecessary for some.
4212 Details : Do not serialize getattr (non-intent version) and statfs.
4215 Frequency : occasional, when OST network is overloaded/intermittent
4217 Description: client evicted by OST after bulk IO timeout
4218 Details : If a client sends a bulk IO request (read or write) the OST
4219 may evict the client if it is unresposive to its data GET/PUT
4220 request. This is incorrect if the network is overloaded (takes
4221 too long to transfer the RPC data) or dropped the OST GET/PUT
4222 request. There is no need to evict the client at all, since
4223 the pinger and/or lock callbacks will handle this, and the
4224 client can restart the bulk request.
4227 Frequency : Always when mmapping file with no objects
4229 Description: client crashes when mmapping file with no objects
4230 Details : Check that we actually have objects in a file before doing any
4231 operations on objects in ll_vm_open, ll_vm_close and
4237 Description: Request leak when working with deleted CWD
4238 Details : Introduce advanced request refcount tracking for requests
4239 referenced from lustre intent.
4241 Severity : Enhancement
4243 Description: Cache open file handles on client.
4244 Details : MDS now will return special lock along with openhandle, if
4245 requested and client is allowed to hold openhandle, even if unused,
4246 until such a lock is revoked. Helps NFS a lot, since NFS is opening
4247 closing files for every read/write openration.
4249 Severity : Enhancement
4251 Description: Cache open negative dentries on client when possible.
4252 Details : Guard negative dentries with UPDATE lock on parent dir, drop
4253 negative dentries on lock revocation.
4258 Description: Remounting a client read-only wasn't possible with a zconf mount
4259 Details : It wasn't possible to remount a client read-only with llmount.
4261 Severity : enhancement
4262 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
4263 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
4264 MPICH is included to add Lustre-specific ADIO interfaces.
4265 This is based closely on the UFS ADIO layer and only differs
4266 in file creation, in order to allow the OST striping to be set.
4267 This is user-contributed code and not supported by CFS.
4272 Description: extended inode attributes (immutable, append-only) work improperly
4273 when 2.4 and 2.6 kernels are used on client/server or vice versa
4274 Details : Introduce kernel-independent values for these flags.
4276 Severity : enhancement
4279 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
4280 Details : Many of the /proc/ tunables can only be tuned at a megabyte
4281 granularity. Now, Fractional MB granularity is be supported,
4282 this is very useful for low memory system.
4284 Severity : enhancement
4286 Description: Getattr by fid
4287 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
4288 locks, avoids extra getattr rpc requests to MDS, allows '/' to
4289 have locks and avoids getattr rpc requests for it on every stat.
4292 Frequency : Always, for filesystems larger than 2TB
4294 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4295 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
4296 filesystems larger than 2TB to be mounted with mballoc enabled.
4299 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
4301 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4302 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
4303 filesystems larger than 2TB is mounted then the kernel will
4304 silently corrupt the start of the filesystem. CONFIG_LBD is
4305 enabled for all CFS-supported kernels, but the possibility of
4306 this happening with a modified kernel config exists.
4308 Severity : enhancement
4310 Description: add client O_DIRECT support for 2.6 kernels
4311 Details : It is now possible to do O_DIRECT reads and writes to files
4312 in the Lustre client mountpoint on 2.6 kernel clients.
4314 Severity : enhancement
4316 Description: parallel glimpse, setattr, statfs, punch, destroy requests
4317 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
4318 parallel, not waiting for response from every OST before sending
4319 a rpc to the next OST.
4324 Description: setattr vs write race when updating file timestamps
4325 Details : Client processes that update a file timestamp into the past
4326 right after writing to the file (e.g. tar) it is possible that
4327 the updated file modification time can be reset to the current
4328 time due to a race between processing the setattr and write RPC.
4330 Severity : enhancement
4332 Description: Bring 'lfs find' closer in line with regular Linux find.
4333 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
4334 -print0 options and obtains all the needed info through the lustre
4337 Severity : enhancement
4339 Description: support up to 1024 configured devices on one node
4340 Details : change obd_dev array from statically allocated to dynamically
4341 allocated structs as they are first used to reduce memory usage
4346 Description: Flush dirty partially truncated pages during truncate
4347 Details : Immediatelly flush partially truncated pages in filter_setattr,
4348 this way we completely avoid having any pages in page cache on OST
4349 and can retire ugly workarounds during writes to flush such pages.
4354 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
4355 Details : protect inode from truncation within vfs_unlink() context
4356 just take a reference before calling vfs_unlink() and release it
4357 when parent's i_sem is free.
4362 Description: last_id value checked outside lock on OST caused LASSERT failure
4363 Details : If there were multiple MDS->OST object precreate requests in
4364 flight, it was possible that the OST's last object id was checked
4365 outside a lock and incorrectly tripped an assertion. Move checks
4366 inside locks, and discard old precreate requests.
4369 Frequency : always, if extents are used on OSTs
4371 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
4372 Details : OSTs using the extents format would not zero the high 16 bits of
4373 the index physical block number. This is not a problem for any
4374 OST filesystems smaller than 16TB, and no kernels support ext3
4375 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
4376 new/modified files) and can be fixed for existing filesystems
4377 with e2fsprogs-1.39-cfs1.
4382 Description: import connection selection may be incorrect if timer wraps
4383 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
4384 import connections to be ignored if the 32-bit jiffies counter
4385 wraps. Use a 64-bit jiffies counter.
4388 Frequency : very large clusters immediately after boot
4390 Description: LNET request buffers exhausted under heavy short-term load
4391 Details : If a large number of client requests are generated on a service
4392 that has previously never seen so many requests it is possible
4393 that the request buffer growth cannot keep up with the spike in
4394 demand. Instead of dropping incoming requests, they are held in
4395 the LND until the RPC service can accept more requests.
4398 Frequency : Sometimes during replay
4400 Description: Assertion failure in ll_local_open after replay.
4401 Details : If replay happened on an open request reply before we were able
4402 to set replay handler, reply will become not swabbed tripping the
4403 assertion in ll_local_open. Now we set the handler right after
4404 recognising of open request
4407 Frequency : very rare
4409 Description: kernel reports "badness in vsnprintf"
4410 Details : Reading from the "recovery_status" /proc file in small chunks
4411 may cause a negative length in lprocfs_obd_rd_recovery_status()
4412 call to vsnprintf() (which is otherwise harmless). Exit early
4413 if there is no more space in the output buffer.
4415 Severity : enhancement
4417 Description: clear OBD RPC statistics by writing to them
4418 Details : It is now possible to clear the OBD RPC statistics by writing
4419 to the "stats" file.
4424 Description: Client mtime is not the same on different clients after utimes
4425 Details : In some cases, the client was using the utimes() syscall on
4426 a file cached on another node. The clients now validate the
4427 ctime from the MDS + OSTs to determine which one is right.
4432 Description: Inability to activate failout mode
4433 Details : lconf script incorrectly assumed that in pythong string's numeric
4434 value is used in comparisons.
4437 Frequency : always with multiple stripes per file
4439 Description: Inefficient object allocation for mutli-stripe files
4440 Details : When selecting which OSTs to stripe files over, for files with
4441 a stripe count that divides evenly into the number of OSTs,
4442 the MDS is always picking the same starting OST for each file.
4443 Return the OST selection heuristic to the original design.
4448 Description: mount failures may take full timeout to return an error
4449 Details : Under some heavy load conditions it is possible that a
4450 failed mount can wait for the full obd_timeout interval,
4451 possibly several minutes, before reporting an error.
4452 Instead return an error as soon as the status is known.
4454 Frequency : quota enabled and large files being deleted
4456 Description: releasing more than 4GB of quota at once hangs OST
4457 Details : If a user deletes more than 4GB of files on a single OST it
4458 will cause the OST to spin in an infinite loop. Release
4459 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4464 Description: statfs data retrieved from /proc may be stale or zero
4465 Details : When reading per-device statfs data from /proc, in the
4466 {kbytes,files}_{total,free,avail} files, it may appear
4467 as zero or be out of date.
4470 Frequency : systems with MD RAID1 external journal devices
4472 Description: lconf's call to blkid is confused by RAID1 journal devices
4473 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4474 of returning all block devices that match the journal UUID.
4477 Frequency : always, for aggregate stripe size over 4GB
4479 Description: assertion fails when trying to use 4GB stripe size
4480 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
4481 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
4484 Frequency : always on ppc64
4486 Description: the first write on an ext3 filesystem with mballoc got stuck
4487 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4488 perform endianness conversion.
4490 ------------------------------------------------------------------------------
4492 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4494 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
4495 INTEROPERATE with older versions automatically. Please read the
4496 user documentation before upgrading any part of a live system.
4497 * WARNING: Lustre networking configuration changes are required with
4498 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
4501 * Support for kernels:
4502 2.6.9-22.0.2.EL (RHEL 4)
4503 2.6.5-7.244 (SLES 9)
4504 2.6.12.6 vanilla (kernel.org)
4507 Severity : enhancement
4508 Bugzilla : 7981/8208
4509 Description: Introduced Lustre Networking (LNET)
4510 Details : LNET is new networking infrastructure for Lustre, it includes
4511 a reorganized network configuration mode (see the user
4512 documentation for full details) as well as support for routing
4513 between different network fabrics. Lustre Networking Devices
4514 (LNDs) for the supported network fabrics have also been
4515 created for this new infrastructure.
4517 Severity : enhancement
4518 Description: Introduced Access control lists
4519 Details : clients can set ACLs on files and directories in order to have
4520 more fine-grained permissions than the standard Unix UGO+RWX.
4521 The MDS must be started with the "-o acl" mount option.
4523 Severity : enhancement
4524 Description: Introduced filesystem quotas
4525 Details : Administrators may now establish per-user quotas on the
4528 Severity : enhancement
4530 Description: Configuration change for the XT3
4531 The PTLLND is now used to run Lustre over Portals on the XT3
4532 The configure option(s) --with-cray-portals are no longer used.
4533 Rather --with-portals=<path-to-portals-includes> is used to
4534 enable building on the XT3. In addition to enable XT3 specific
4535 features the option --enable-cray-xt3 must be used.
4540 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
4541 Details : A race between allocating a new llcd and re-getting the llcd_lock
4542 allowed another thread to grab newly-allocated llcd.
4544 Severity : enhancement
4546 Description: 2.6 OST async journal commit and locking fix to improve performance
4547 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
4548 2.6 kernels are now async as they already were in 2.4 kernels so
4549 that they can commit concurrently with the network bulk transfer.
4550 For block-allocated files the filter allocation semaphore is held
4551 to avoid filesystem fragmentation during allocation. BKL lock
4552 removed for 2.6 xattr operations where it is no longer needed.
4557 Description: lconf incorrectly determined whether two IP networks could talk
4558 Details : In some more complicated routing and multiple-network
4559 configurations, lconf will avoid trying to make a network
4560 connection to a disjoint part of the IP space. It was doing the
4561 math incorrectly for one set of cases.
4566 Description: Fix for potential infinite loop processing records in an llog.
4567 Details : If an llog record is corrupted/zeroed, it is possible to loop
4568 forever in llog_process(). Validate the llog record length
4569 and skip the remainder of the block on error.
4572 Frequency : occasional (liblustre only)
4574 Description: liblustre could not open files whose last component is a symlink
4575 Details : sysio_path_walk() would incorrectly pass the open intent to
4576 intermediate path components.
4579 Frequency : rare (liblustre only with non-standard tuning)
4580 Bugzilla : 7201 (7350)
4581 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4582 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4583 after it has been released, only do this for referenced locks
4585 Severity : enhancement
4587 Description: specifying an (invalid) directory default stripe_size of -1
4588 would reset the directory default striping
4589 Details : stripe_size -1 was used internally to signal directory stripe
4590 removal, now use "all default" to signal dir stripe removal
4591 as a directory striping of "all default" is not useful
4594 Frequency : common for large clusters running liblustre clients
4596 Description: doing an ls when liblustre clients are running is slow
4597 Details : sending a glimpse AST to a liblustre client waits for every AST
4598 to time out, as liblustre clients will not respond. Since they
4599 cannot cache data we refresh the OST lock LVB from disk instead.
4601 Severity : enhancement
4603 Description: doing an ls at the same time as file IO can be slow
4604 Details : enqueue and other "small" requests can be blocked behind many
4605 large IO requests. Create a new OST IO portal for non-IO
4606 requests so they can be processed faster.
4609 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4611 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4612 created file with mode 000
4613 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4614 allow an ftruncate() as a non-root user to a file with mode 000.
4615 Lustre now respects this flag to disable mode checks when
4616 truncating a file owned by the user
4619 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4622 Description: Revoking locks from clients that went dead or catatonic might take
4624 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4625 cancellation of such locks instant on servers without waiting for
4626 any reply from clients. Clients drops these locks when cancel
4627 notification from server is received without replying.
4630 Frequency : liblustre-only, when liblustre client dies or becomes busy
4632 Description: Doing ls on Linux clients can take a long time with active
4634 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4635 granting such locks to it in the first place if possible. Locks
4636 are taken by proxy on the OST during the read or write and
4637 dropped immediately afterward. Add connect flags handling, do
4638 not grant locks to liblustre clients for glimpse ASTs.
4640 Severity : enhancement
4642 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4643 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4644 sequential reads and behaves badly with applications doing random
4645 reads. Improve it by reading ahead at least read region, and
4646 avoiding excessive large RPC for small reads.
4648 Severity : enhancement
4650 Description: Creating more than 1000 files for a single job may cause a load
4651 imbalance on the OSTs if there are also a large number of OSTs.
4652 Details : qos_prep_create() uses an OST index reseed value that is an
4653 even multiple of the number of available OSTs so that if the
4654 reseed happens in the middle of the object allocation it will
4655 still utilize the OSTs as uniformly as possible.
4660 Description: OST or MDS may oops in ping_evictor_main()
4661 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
4662 but doesn't restart at beginning of obd_exports_timed list
4665 Severity : enhancement
4667 Description: improve by-nid export eviction on the MDS and OST
4668 Details : allow multiple exports with the same NID to be evicted at one
4669 time without re-searching the exports list.
4672 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
4674 Description: MDS may oops in groups_free()
4675 Details : in rare race conditions a newly allocated group_info struct is
4676 freed again, and this can be NULL. The 2.4 compatibility code
4677 for groups_free() checked for a NULL pointer, but 2.6 did not.
4680 Frequency : common for liblustre clients doing little filesystem IO
4681 Bugzilla : 9352, 7313
4682 Description: server may evict liblustre clients accessing contended locks
4683 Details : if a client is granted a lock or receives a completion AST
4684 with a blocking AST already set it would not reply to the AST
4685 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
4688 Frequency : lfs setstripe, only systems with more than 160 OSTs
4690 Description: unable to set striping with a starting offset beyond OST 160
4691 Details : llapi_create_file() incorrectly limited the starting stripe
4692 index to the maximum single-file stripe count.
4695 Frequency : LDAP users only
4697 Description: lconf did not handle in-kernel recovery with LDAP properly
4698 Details : lconf/LustreDB get_refs() is searching the wrong namespace
4700 Severity : enhancement
4702 Description: bind OST threads to NUMA nodes to improve performance
4703 Details : all OST threads are uniformly bound to CPUs on a single NUMA
4704 node and do their allocations there to localize memory access
4706 Severity : enhancement
4708 Description: llmount can determine client NID directly from Myrinet (GM)
4709 Details : the client NID code from gmnalnid was moved directly into
4710 llmount, removing the need to use this or specifying the
4711 client NID explicitly when mounting GM clients with zeroconf
4714 Frequency : if client is started with down MDS
4716 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
4717 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
4718 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
4719 ptlrpc_import_delayed_req() because we didn't check if the
4720 request was interrupted, and we also didn't break out of the
4721 event loop if there was no timeout
4726 Description: data loss during non-page-aligned writes to a single file from
4727 both multiple nodes and multiple threads on one node at same time
4728 Details : updates to KMS and lsm weren't protected by common lock. Resulting
4729 inconsistency led to false short-reads, that were cached and later
4730 used by ->prepare_write() to fill in partially written page,
4731 leading to data loss.
4734 Frequency : always, if lconf --abort_recovery used
4736 Description: lconf --abort_recovery fails with 'Operation not supported'
4737 Details : lconf was attempting to abort recovery on the MDT device and not
4740 Severity : enhancement
4742 Description: remove cleanup logs
4743 Details : replace lconf-generated cleanup logs with lustre internal
4744 cleanup routines. Eliminates the need for client-cleanup and
4747 Severity : enhancement
4749 Description: add support for EAs (user and system) on lustre filesystems
4750 Details : it is now possible to store extended attributes in the Lustre
4751 client filesystem, and with the user_xattr mount option it
4752 is possible to allow users to store EAs on their files also
4754 Severity : enhancement
4756 Description: Add possibility (config option) to show minimal available OST free
4758 Details : When compiled with --enable-mindf configure option, statfs(2)
4759 (and so, df) will return least minimal free space available from
4760 all OSTs as amount of free space on FS, instead of summary of
4761 free spaces of all OSTs.
4763 Severity : enhancement
4765 Description: do not expand extent locks acquired on OST-side
4766 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
4767 by server: they are not cached anyway.
4770 Frequency : when mmap is used/binaries executed from Lustre
4772 Description: Unmmap pages before throwing them away from read cache.
4773 Details : llap_shrink cache now attempts to unmap pages before discarding
4774 them (if unmapping failed - do not discard). SLES9 kernel has
4775 extra checks that trigger if this unmapping is not done first.
4780 Description: lconf didn't resolve symlinks before checking to see whether a
4781 given mountpoint was already in use
4784 Frequency : when migrating failover services
4785 Bugzilla : 6395, 9514
4786 Description: When migrating a subset of services from a node (e.g. failback
4787 from a failover service node) the remaining services would
4788 time out and evict clients.
4789 Details : lconf --force (implied by --failover) sets the global obd_timeout
4790 to 5 seconds in order to quickly disconnect, but this caused
4791 other RPCs to time out too quickly. Do not change the global
4792 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
4794 Severity : enhancement
4795 Frequency : if MDS is started with down OST
4796 Bugzilla : 9439,5706
4797 Description: Allow startup/shutdown of an MDS without depending on the
4798 availability of the OSTs.
4799 Details : Asynchronously call mds_lov_synchronize during MDS startup.
4800 Add appropriate locking and lov-osc refcounts for safe
4801 cleaning. Add osc abort_inflight calls in case the
4802 synchronize never started.
4805 Frequency : occasional (Cray XT3 only)
4807 Description: root not authorized to access files in CRAY_PORTALS environment
4808 Details : The client process capabilities were not honoured on the MDS in
4809 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
4810 been accessed by an authorized user then root was able to access
4811 the file on the local client also. The root user capabilities
4812 are now allowed on the MDS, as this environment has secure UID.
4815 Frequency : occasional
4817 Description: ldiskfs "too long searching" message happens too often
4818 Details : A debugging message (otherwise harmless) prints too often on
4819 the OST console. This has been reduced to only happen when
4820 there are fragmentation problems on the filesystem.
4825 Description: Division by zero in statfs when all OSCs are inactive
4826 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
4827 lov_statfs divides by value returned by lov_get_stripecnt().
4831 Bugzilla : 9489, 3273
4832 Description: First write from each client to each OST was only 4kB in size,
4833 to initialize client writeback cache, which caused sub-optimal
4834 RPCs and poor layout on disk for the first written file.
4835 Details : Clients now request an initial cache grant at (re)connect time
4836 and so that they can start streaming writes to the cache right
4837 away and always do full-sized RPCs if there is enough data.
4838 If the OST is rebooted the client also re-establishes its grant
4839 so that client cached writes will be honoured under the grant.
4844 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
4845 Details : Now I/O RPCs go to different portal number and (presumably) fast
4846 lock requests (and glimses) and other RPCs get their own service
4847 threads pool that should be able to service those RPCs
4850 Severity : enhancement
4852 Description: Ability to exchange lustre version between client and servers and
4853 issue warnings at client side if client is too old. Also for
4854 liblustre clients there is ability to refuse connection of too old
4856 Details : New 'version' field is added to connect data structure that is
4857 filled with version info. That info is later checked by server and
4861 Frequency : rare, liblustre only.
4862 Bugzilla : 9296, 9581
4863 Description: Two simultaneous writes from liblustre at offset within same page
4864 might proceed at the same time overwriting eachother with stale
4866 Details : I/O lock withing llu_file_prwv was released too early, before data
4867 actually was hitting the wire. Extended lock-holding time until
4868 server acknowledges receiving data.
4871 Frequency : extremely rare. Never observed in practice.
4873 Description: avoid generating lustre_handle cookie of 0.
4874 Details : class_handle_hash() generates handle cookies by incrementing
4875 global counter, and can hit 0 occasionaly (this is unlikely, but
4876 not impossible, because initial value of cookie counter is
4877 selected randonly). Value of 0 is used as a sentinel meaning
4878 "unassigned handle" --- avoid it. Also coalesce two critical
4879 sections in this function into one.
4881 Severity : enhancement
4883 Description: allow liblustre clients to delegate truncate locking to OST
4884 Details : To avoid overhead of locking, liblustre client instructs OST to
4885 take extent lock in ost_punch() on client's behalf. New connection
4886 flag is added to handle backward compatibility.
4888 Severity : enhancement
4889 Bugzilla : 4928, 7341, 9758
4890 Description: allow number of OST service threads to be specified
4891 Details : a module parameter allows the number of OST service threads
4892 to be specified via "options ost ost_num_threads={N}" in the
4893 OSS's /etc/modules.conf or /etc/modprobe.conf.
4897 Bugzilla : 6146, 9635, 9895
4898 Description: servers crash with bad pointer in target_handle_connect()
4899 Details : In rare cases when a client is reconnecting it was possible that
4900 the connection request was the last reference for that export.
4901 We would temporarily drop the export reference and get a new
4902 one, but this may have been the last reference and the export
4903 was just destroyed. Get new reference before dropping old one.
4905 Severity : enhancement
4906 Frequency : if client is started with failover MDS
4908 Description: Allow multiple MDS hostnames in the mount command
4909 Details : Try to read the configuration from all specified MDS
4910 hostnames during a client mount in case the "primary"
4913 Severity : enhancement
4915 Description: Stop sending data to evicted clients as soon as possible.
4916 Details : Check if the client we are about to send or are sending data to
4917 was evicted already. (Check is done every second of waiting,
4918 for which l_wait_event interface was extended to allow checking
4919 of exit condition at specified intervals).
4922 Frequency : rare, normally only when NFS exporting is done from client
4924 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
4926 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
4929 Frequency : always, if srand() is called before liblustre initialization
4931 Description: Liblustre uses system PRNG disturbing its usage by user application
4932 Details : Introduce internal to lustre fast and high-quality PRNG for
4933 lustre usage and make liblustre and some other places in generic
4934 lustre code to use it.
4936 Severity : enhancement
4937 Bugzilla : 9477, 9557, 9870
4938 Description: Verify that the MDS configuration logs are updated when xml is
4939 Details : Check if the .xml configuration logs are newer than the config
4940 logs stored on the MDS and report an error if this is the case.
4941 Request --write-conf, or allow starting with --old_conf.
4943 Severity : enhancement
4945 Description: Handle symlinks in the path when checking if Lustre is mounted.
4946 Details : Resolve intermediate symlinks when checking if a client has
4947 mounted a filesystem to avoid duplicate client mounts.
4952 Description: lconf can hit an error exception but still return success.
4953 Details : The lconf command catches the Command error exception at the top
4954 level script context and will exit with the associated exit
4955 status, but doesn't ensure that this exit status is non-zero.
4960 Description: failure of ptlrpc thread startup can cause oops
4961 Details : Starting a ptlrpc service thread can fail if there are a large
4962 number of threads or the server memory is very fragmented.
4963 Handle this without oopsing.
4966 Frequency : always, only if liblustre and non-default acceptor port was used
4968 Description: liblustre cannot connect to servers with non-default acceptor port
4969 Details : tcpnal_set_default_params() was not called and was therefore
4970 ignoring the environment varaible TCPNAL_PORT, as well as other
4971 TCPNAL_ environment variables
4976 Description: two objects could be created on the same OST for a single file
4977 Details : If an OST is down, in some cases it was possible to create two
4978 objects on a single OST for a single file. No problems other
4979 than potential performance impact and spurious error messages.
4983 Bugzilla : 5681, 9562
4984 Description: Client may oops in ll_unhash_aliases
4985 Details : Client dcache may become inconsistent in race condition.
4986 In some cases "getcwd" can fail if the current directory is
4992 Description: Inode refcounting problems in NFS export code
4993 Details : link_raw functions used to call d_instantiate without obtaining
4994 extra inode reference first.
4998 Bugzilla : 9942, 9903
4999 Description: Referencing freed requests leading to crash, memleaks with NFS.
5000 Details : We used to require that call to ll_revalidate_it was always
5001 followed by ll_lookup_it. Also with revalidate_special() it is
5002 possible to call ll_revalidate_it() twice for the same dentry
5003 even if first occurence returned success. This fix changes semantic
5004 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
5005 reference on a request referred from the intent.
5006 ll_intent_release() then releases such a request.
5009 Frequency : rare, normally benchmark loads only
5011 Description: unlinked inodes were kept in memory on the client
5012 Details : If a client is repeatedly creating and unlinking files it
5013 can accumulate a lot of stale inodes in the inode slab cache.
5014 If there is no other client load running this can cause the
5015 client node to run out of memory. Instead flush old inodes
5016 from client cache that have the same inode number as a new inode.
5019 Frequency : SLES9 2.6.5 kernel and long filenames only
5020 Bugzilla : 9969, 10379
5021 Description: utime reports stale NFS file handle
5022 Details : SLES9 uses out-of-dentry names in some cases, which confused
5023 the lustre dentry revalidation. Change it to always use the
5027 Frequency : rare, unless heavy write-truncate concurrency is continuous
5028 Bugzilla : 4180, 6984, 7171, 9963, 9331
5029 Description: OST becomes very slow and/or deadlocked during object unlink
5030 Details : filter_destroy() was holding onto the parent directory lock
5031 while truncating+unlinking objects. For very large objects this
5032 may block other threads for a long time and slow overall OST
5033 responsiveness. It may also be possible to get a lock ordering
5034 deadlock in this case, or run out of journal credits because of
5035 the combined truncate+unlink. Solution is to do object truncate
5036 first in one transaction without parent lock, and then do the
5037 final unlink in a new transaction with the parent lock. This
5038 reduces the lock hold time dramatically.
5041 Frequency : rare, 2.4 kernels only
5043 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
5044 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
5045 before unmounting their filesystem in order to avoid deadlock.
5046 The kernel_locked() function in 2.4 kernels only checks whether
5047 the kernel lock is held, not whether it is this process that is
5048 holding it as 2.6 kernels do.
5053 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
5054 Details : The client ptlrpc code may be trying to reconnect to a down
5055 server before a previous connection attempt has timed out.
5056 Increase the reconnect interval to be longer than the connection
5057 timeout interval to avoid sending duplicate connections to
5061 Frequency : echo_client brw_test command
5063 Description: fix echo_client to work with OST preallocated code
5064 Details : OST preallocation code (5137) didn't take echo_client IO path
5065 into account: echo_client calls filter methods outside of any
5066 OST thread and, hence, there is no per-thread preallocated
5067 pages and buffers to use. Solution: hijack pga pages for IO. As
5068 a byproduct, this avoids unnecessary data copying.
5072 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
5073 Description: Client can oops in mdc_commit_close() after open replay
5074 Details : It was possible for the MDS to return an open request with no
5075 transaction number in mds_finish_transno() if the client was
5076 evicted, but without actually returning an error. Clients
5077 would later try to replay that open and may trip an assertion
5078 Simplify the client close codepath, and always return an error
5079 from the MDS in case the open is not successful.
5082 Frequency : rare, 2.6 OSTs only
5084 Description: OST may deadlock under high load on fragmented files
5085 Details : If there was a heavy load and highly-fragmented OST filesystems
5086 it was possible to have all the OST threads deadlock waiting on
5087 allocation of biovecs, because the biovecs were not released
5088 until the entire RPC IO was completed. Instead, release biovecs
5089 as soon as they are complete to ensure forward IO progress.
5091 Severity : enhancement
5093 Description: Support for specifying external journal device at mount
5094 Details : If an OST or MDS device is formatted with an external journal
5095 device, this device major/minor is stored in the ext3 superblock
5096 and may not be valid for failover. Allow detecting and
5097 specifying the external journal at mount time.
5102 Description: Mounting an MDS with pending unlinked files may cause oops
5103 Details : target_finish_recovery() calls mds_postrecov() which returned
5104 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
5105 considers this an error and immediately begins cleaning up the
5106 lov, just after starting the mds_lov process
5108 Severity : enhancement
5110 Description: Implement 'lfs df' to report actual free space on per-OST basis
5111 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
5112 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
5113 usage of objects; '-h' to report in human readable format.
5115 ------------------------------------------------------------------------------
5117 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
5124 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
5125 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
5126 at startup during orphan unlinking. Instead, skip buddy update
5127 before setup, as it will be regenerated after recovery is complete.
5132 Description: If an OST is inactive, its locks might reference stale inodes.
5133 Details : lov_change_cbdata() must iterate over all namespaces, even if
5134 they are inactive to clear inode references from the lock.
5136 Severity : enhancement
5137 Frequency : occasional, if non-standard max_dirty_mb used
5139 Description: Client will block write RPCs if not enough grant
5140 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
5141 then the client will block writes while waiting for another RPC
5142 to complete instead of consuming its dirty limit. With change
5143 we get improved performance when max_dirty_mb is small.
5145 Severity : enhancement
5146 Bugzilla : 3389, 6253
5147 Description: Add support for supplementary groups on the MDS.
5148 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
5149 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
5150 lookups for user supplementary groups into a cache.
5154 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
5155 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
5156 for liblustre O_EXCL is enforced by MDS.
5160 Description: readdir never returns NULL in liblustre.
5161 Details : Corrected llu_iop_getdirentries logic, to return offset of next
5162 dentry in struct dirent.
5166 Frequency : liblustre only, depends on application IO pattern
5167 Description: liblustre clients evicted if not contacting servers
5168 Details : Don't put liblustre clients into the ping_evictor list, so
5169 they will not be evicted by the pinger ever.
5171 Severity : enhancement
5173 Description: Add ability to evict clients by NID from MDS.
5174 Details : By echoing "nid:$NID" string into
5175 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
5176 $NID would be instantly evicted from this MDS and from all active
5177 OSTs connected to it.
5181 Description: Do not query file size twice, somewhat slowing stat(2) calls.
5182 Details : lookup_it_finish() used to query file size from OSTs that was not
5187 Description: service threads change working directory to that of init
5188 Details : Starting lustre service threads may pin the working directory
5189 of the parent thread, making that filesystem busy. Threads
5190 now change to the working directory of init to avoid this.
5194 Frequency : during shutdown only
5195 Description: shutdown with a failed MDS or OST can cause unmount to hang
5196 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
5201 Frequency : 2.6 only
5202 Description: chmod/chown may include an extra supplementary group
5203 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
5204 supplementary group and if none is specified this is used.
5207 Bugzilla : 5479 (6816)
5209 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
5210 Details : If lookup is in progress on a file that is unlinked we might try
5211 to revalidate the inode and fail in revalidate after lookup is
5212 complete and ll_file_open() enqueues the open again but
5213 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
5216 Frequency : always, if lconf --abort_recovery used
5218 Description: lconf --abort_recovery fails with 'Operation not supported'
5219 Details : lconf was attempting to abort recovery on the MDT device and not
5222 ------------------------------------------------------------------------------
5224 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
5229 Frequency : rare (only unsupported configurations with a node running as an
5231 Bugzilla : 6514, 5137
5232 Description: Mounting a Lustre file system on a node running as an OST could
5234 Details : OSTs now preallocates memory needed to write out data at
5235 startup, instead of when needed, to avoid having to
5236 allocate memory in possibly low memory situations.
5237 Specifically, if the file system is mounted on on OST,
5238 memory pressure could force it to try to write out data,
5239 which it needed to allocate memory to do. Due to the low
5240 memory, it would be unable to do so and the node would
5241 become unresponsive.
5243 Severity : enhancement
5245 Description: Addition of lconf --service command line option
5246 Details : lconf now accepts a '--service <arg>' option, which is
5247 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
5249 Severity : enhancement
5251 Description: Failover mode is now the default for OSTs.
5252 Details : By default, OSTs will now run in failover mode. To return to
5253 the old behaviour, add '--failout' to the lmc line for OSTs.
5255 Severity : enhancement
5257 Description: Health checks are now provided for MDS and OSTs
5258 Details : Additional detailed health check information on MSD and OSTs
5259 is now provided through the procfs health_check value.
5262 Frequency : occasional, depends on IO load
5264 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
5265 after numerous create/delete cycles
5266 Details : The ext3 inode allocation policy would not allocate new inodes
5267 very well on the OSTs because there are no new directories
5268 being created. Instead we look for groups with free space if
5269 the parent directories are nearly full.
5274 Description: Network or server problems during mount may cause partially
5275 mounted clients instead of returning an error.
5276 Details : The config llog parsing code may overwrite the error return
5277 code during mount error handling, returning success instead
5283 Description: MDS can fail to allocate large reply buffers
5284 Details : After long uptimes the MDS can fail to allocate large reply
5285 buffers (e.g. zconf client mount config records) due to memory
5286 fragmentation or consumption by the buffer cache. Preallocate
5287 some large reply buffers so that these replies can be sent even
5288 under memory pressure.
5292 Frequency : rare (liblustre)
5293 Description: fsx running with liblustre complained that using truncate() to
5294 extend the file doesn't work. This patch corrects that issue.
5295 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
5296 ATTR_SIZE and lsm use in llu_setattr_raw.
5300 Frequency : rare, only 2.6 kernels
5301 Description: Unusual file access patterns on the MDS may result in inode
5302 data being lost in very rare circumstances.
5303 Details : Bad interaction between the ea-in-inode patch and the "no-read"
5304 code in the 2.6 kernel caused the inode and/or EA data not to
5305 be read from disk, causing single-file corruption.
5309 Frequency : rare, only 2.6 filesystems using extents
5310 Description: Heavy concurrent write and delete load may cause data corruption.
5311 Details : It was possible under high-load situations to have an extent
5312 metadata block in the block device cache from a just-unlinked
5313 file overwrite a newly-allocated data block. We now unmap any
5314 metadata buffers that alias just-allocated data blocks.
5318 Frequency : filesystems with default stripe_count larger than 77
5319 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
5320 Details : lconf specifies an inode size of 4096 bytes when the default
5321 stripe_count is larger than 77. This conflicts with the default
5322 inode density of 1 per 4096 bytes. Allocate smaller inodes in
5323 this case to avoid pinning too much memory for large EAs.
5325 ------------------------------------------------------------------------------
5327 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
5332 Frequency : rare (extremely heavy IO load with hundreds of clients)
5334 Description: Client is evicted, gets IO error writing to file
5335 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
5336 caused clients to be evicted for AST timeouts. The fixes in
5337 bug 5192 mean we no longer need to have such short AST timeouts
5338 so ldlm_timeout has been increased.
5341 Frequency : occasional during --force or --failover shutdown under load
5342 Bugzilla : 5949, 4834
5343 Description: Server oops/LBUG if stopped with --force or --failover under load
5344 Details : a collection of import/export refcount and cleanup ordering
5345 issues fixed for safer force cleanup
5348 Frequency : only filesystems larger than 120 OSTs
5349 Bugzilla : 5990, 6223
5350 Description: lfs getstripe would oops on a very large filesystem
5351 Details : lov_getconfig used kfree on vmalloc'd memory
5354 Frequency : only filesystems exporting via NFS to Solaris 10 clients
5355 Bugzilla : 6242, 6243
5356 Description: reading from files that had been truncated to a non-zero size
5357 but never opened returned no data
5358 Details : ll_file_read() reads zeros from no-object files to EOF
5363 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
5364 Details : There's little harm in aborting MDS/OSS recovery and letting it
5365 try again, so I removed the LASSERT and return an error instead.
5367 Severity : enhancement
5369 Description: New debugging infrastructure for tracking down data corruption
5370 Details : The I/O checksum code was replaced to: (a) control it at runtime,
5371 (b) cover more of the client-side code path, and (c) try to narrow
5372 down where problems occurred
5376 Bugzilla : 3819, 4364, 4397, 6313
5377 Description: Racing close and eviction MDS could cause assertion in mds_close
5378 Details : It was possible to get multiple mfd references during close and
5379 client eviction, leading to one thread referencing a freed mfd.
5381 Severity: : enhancement
5382 Bugzilla : 3262, 6359
5383 Description: Attempts to reconnect to servers are now more aggressive.
5384 Details : This builds on the enhanced upcall-less recovery that was added
5385 in 1.4.2. When trying to reconnect to servers, clients will
5386 now try each server in the failover group every 10 seconds. By
5387 default, clients would previously try one server every 25 seconds.
5392 Description: After recovery, certain operations trigger a failed
5393 assertion on a client.
5394 Details : Failing over an mds, using lconf -d --failover, while a
5395 client was doing a readdir() call would cause the client to
5396 LBUG after recovery completed and the readdir() was resent.
5398 Severity : enhancement
5400 Description: Default groups are now added by lconf
5401 Details : You can now run lconf --group <servicename> without having to
5402 manually add groups with lmc.
5405 Frequency : occasional
5407 Description: Nodes with an elan id of 0 trigger a failed assertion
5410 Frequency : always when accessing e.g. tty/console device nodes
5412 Description: tty and some other devices nodes cannot be used on lustre
5413 Details : file's private_data field is used by device data and lustre
5414 values in there got lost. New field was added to struct file to
5415 store fs-specific private data.
5418 Frequency : when exporting Lustre via NFS
5420 Description: NFSD failed occasionally when looking up a path component
5421 Details : NFSD is looking up ".." which was broken in ext3 directories
5422 that had grown large enough to become hashed.
5425 Frequency : Clusters with multiple interfaces not on the same subnet
5427 Description: Nodes will repeatedly try to reconnect to an interface which it
5428 cannot reach and report an error to the log.
5429 Details : Extra peer list entries will be created by lconf with some peers
5430 unreachable. lconf now validates the peer before adding it.
5433 Frequency : Only if a default stripe is set on the filesystem root.
5435 Description: Setting a default stripe on the filesystem root prevented the
5436 filesystem from being remounted.
5437 Details : The client was sending extra request flags in the root getattr
5438 request and did not allocate a reply buffer for the dir EA.
5441 Frequency : occasional, higher if lots of files are accessed by one client
5442 Bugzilla : 6159, 6097
5443 Description: Client trips assertion regarding lsm mismatch/magic
5444 Details : While revalidating inodes the VFS looks up inodes with ifind()
5445 and in rare cases can find an inode that is being freed.
5446 The ll_test_inode() code will free the lsm during ifind()
5447 when it finds an existing inode and then the VFS later attaches
5448 this free lsm to a new inode.
5452 Bugzilla : 6422, 7030
5453 Description: MDS deadlock between mkdir and client eviction
5454 Details : Creating a new file via mkdir or mknod (starting a transaction
5455 and getting the ns lock) can deadlock with client eviction
5456 (gets ns lock and trying to finish a synchronous transaction).
5459 Frequency : occasional
5460 Description: While starting a server, the fsfilt_ext3 module could not be
5462 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
5463 kernels. Previously, lconf would still use the ext3 name
5464 when trying to load modules. Now, it will correctly use
5465 ext3 on 2.4 and ldiskfs on 2.6.
5467 Severity : enhancement
5468 Description: The default stripe count has been changed to 1
5469 Details : The interpretation of the default stripe count (0, to lfs
5470 or lmc) has been changed to mean striping across a single
5471 OST, rather than all available. For general usage we have
5472 found a stripe count of 1 or 2 works best.
5474 Severity : enhancement
5475 Description: Add support for compiling against Cray portals.
5476 Details : Conditional compiling for some areas that are different
5480 Frequency : occasional
5481 Bugzilla : 6409, 6834
5482 Description: Creating files with an explicit stripe count may lead to
5483 a failed assertion on the MDS
5484 Details : If some OSTs are full or unavailable, creating files may
5485 trigger a failed assertion on the MDS. Now, Lustre will
5486 try to use other servers or return an error to the
5490 Frequency : occasional
5492 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
5493 caused lock timeout during readahead (since 1.4.2).
5494 Details : Processes doing readahead might match a lock that hasn't been
5495 granted yet if there are overlapping and conflicting lock
5496 requests. The readahead process waits on ungranted lock
5497 (original lock is CBPENDING), while OST waits for that process
5498 to cancel CBPENDING read lock and eventually evicts client.
5500 Severity : enhancement
5502 Description: Initial enabling of flock support for clients
5503 Details : Implements fcntl advisory locking and file status functions.
5504 This feature is provided as an optional mount flag (default
5505 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
5506 locking are implemented yet, and those that are are not guaranteed
5507 to be completely correct in production environments.
5508 mount -t lustre -o [flock|noflock] ...
5511 Frequency : occasional
5513 Description: OSTs running 2.4 kernels but with extents enabled might trip an
5514 assertion in the ext3 JBD (journaling) layer.
5515 Details : The b_committed_data struct is protected by the big kernel lock
5516 in 2.4 kernels, serializing journal_commit_transaction() and
5517 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
5518 access to this struct. In 2.6 kernels there is finer grained
5519 locking to improve SMP performance of the JBD layer.
5523 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
5528 Description: merge and cleanup kernel patches.
5531 -----------------------------------------------------------------------------
5533 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
5535 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
5536 versions of Lustre. Please update all servers and clients to
5537 version 1.4.2 or later at the same time. You must also run
5538 "lconf --write-conf {config}.xml" on the MDS while it is stopped
5539 to update the configuration logs.
5541 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
5542 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
5543 - swab llog records of type '0' so we get proper header size/idx (5861)
5544 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
5545 - fix rename of one directory over another leaking an inode (5953)
5546 - avoid SetPageDirty on 2.6 (5981)
5547 - don't re-add just-being-destroyed locks to the waiting list (5653)
5548 - when creating new directories, inherit the parent's custom
5549 striping settings if present parent (3048)
5550 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
5551 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
5552 - don't hold client locks on temporary worklist from l_lru (5666)
5553 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
5554 - automatically evict dead clients (5921)
5555 - Update file size properly in create+truncate+fstat case (6196)
5556 - Do not unhash mountpoint dentries, do not allow removal of
5558 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
5559 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
5560 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
5561 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
5562 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
5563 - Correctly handle reads of files with no objects (6243)
5564 - lctl recover will also mark a device active if deactivate used (5933)
5566 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
5567 - allow --write-conf on an MDS with different nettype than client (5619)
5568 - don't write config llogs to MDS for mounts not from that MDS (5617)
5569 - lconf should create multiple TCP connections from a client (5201)
5570 - init scripts are now turned off by default; run chkconfig --on
5571 lustre and chkconfig --on lustrefs to use them
5572 - upcalls are no longer needed for clients to recover to failover
5574 - add --abort-recovery option to lconf to abort recovery on device
5576 - add support for an arbitrary number of OSTs (3026)
5577 - Quota support protocol changes.
5578 - forward compatibility changes to wire structs (6007)
5579 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5580 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5581 - improved "lustre status" script
5582 - initialize blocksize for non-regular files (6062)
5583 - added --disable-server and --disable-client configure options (5782)
5584 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5585 - Vanilla 2.4.29 support
5586 - increase maximum number of obd devices to 520 (6242)
5587 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5588 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5589 SLES 9 2.6.5 kernels
5590 - limit stripes per file to 160 (the maximum EA size) (6093)
5592 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5595 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5596 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5597 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5598 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5599 - small ext3 extents cleanups and fixes (5733)
5600 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5601 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5602 - avoid needless client->OST connect, fix handle mismatch (5317)
5603 - fix DLM error path that led to out-of-sync client, long delays (5779)
5604 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5605 - fix several locking issues related to i_size (5492,5624,5654,5672)
5606 - don't move pending lock onto export if it is already evicted (5683)
5607 - fix kernel oops when creating .foo in unlinked directory (5548)
5608 - fix deadlock in obdfilter statistics vs. object create (5811)
5609 - use time_{before,after} to avoid timer jiffies wrap (5882)
5610 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5611 - Do not leak request if server was not able to process it (5154)
5612 - If mds_open unable to find parent dir, make that negative lookup(5154)
5613 - don't create new directories with extent-mapping (5909, 5936)
5615 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5616 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5617 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5618 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5619 - support for RHEL4 kernel on i686 (5773)
5620 - provide error messages when incompatible logs are encountered (5898)
5622 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5623 * version 1.4.0.10 (1.4.1 release candidate 1)
5625 - don't keep a lock reference when lock is not granted (4238)
5626 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5627 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5628 - drop import inflight refcount on signal_completed_replay error (5255)
5629 - unlock page after async write error during send (3677)
5630 - handle missing objects in filter_preprw_read properly (5265)
5631 - no transno return for symlink open, don't save no-trasno open (3440)
5632 - don't try to complete elan receive that already failed (4012)
5633 - free RPC server reply state on error (5406)
5634 - clean up thread from ptlrpc_start_thread() on error (5160)
5635 - readahead could read extra page into cache that wasn't ejected (5388)
5636 - prevent races in class_attach/setup/cleanup/detach (5260)
5637 - don't dereference de->d_inode after l_dput of de (5458)
5638 - use "int" for stripe value returned from lock_to_stripe (5544)
5639 - mballoc allocation and error-checking fixes in 2.6 (5504)
5640 - block device patches to fix I/O request sizes in 2.6 (5482)
5641 - look up hostnames for IB nals (5602)
5642 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5643 - don't start multiple acceptors for the same port (5277)
5644 - fix incorrect LASSERT in mds_getattr_name (5635)
5645 - export a proc file for general "ping" checking (5628)
5646 - fix "lfs check" to not block when the MDS is down (5628)
5648 - service request history (4965)
5649 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5650 - create an "evict_client" /proc entry on OSTs, like the MDS has
5651 - fix mount usage message, return errors per mount(8) (5168)
5652 - change grep [] to grep "[]" in tests so they work in more UMLs
5653 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5654 - remove ancient LOV_MAGIC_V0 EA support (5047)
5655 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5656 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5657 - allow readahead allocations to fail when low on memory (5383)
5658 - mmap locking landed again, after considerable improvement (2828)
5659 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
5661 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
5664 - send OST transaction number in read/write reply to free req (4966)
5665 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5666 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5667 - the watchdog thread now runs as interruptible (5246)
5668 - flock/lockf fixes (but it's still disabled, pending 5135)
5669 - don't use EXT3 constants in llite code (5094)
5670 - memory shortage at startup could cause assertion (5176)
5672 - reorganization of lov code
5673 - single portals codebase
5675 - add extents/mballoc support (5025)
5676 - direct I/O reads in the obdfilter (4048)
5677 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
5679 tbd Cluster File Systems, Inc. <info@clusterfs.com>
5682 - send OST transaction number in read/write reply to free req (4966)
5683 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5684 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5685 - the watchdog thread now runs as interruptible (5246)
5686 - handle missing objects in filter_preprw_read properly (5265)
5687 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5688 - drop import inflight refcount on signal_completed_replay error (5255)
5689 - unlock page after async write error during send (3677)
5690 - return original error code on reconstructed replies (3761)
5691 - no transno return for symlink open, don't save no-trasno open (3440)
5693 - add pid to ldlm debugging output (4922)
5694 - bump the watchdog timeouts -- we can't handle 30sec yet
5695 - extra debugging for orphan dentry/inode bug (5259)
5697 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
5700 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
5701 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
5702 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
5703 - fix NULL dereference in /proc/sys/portals/routes (4827)
5704 - allow failed mdc_close() operations to be interrupted (4561)
5705 - stop precreate on OST before MDS would time out on it (4778)
5706 - don't send partial-page writes before EOF from client (4410)
5707 - discard client grant for sub-page writes on large-page clients (4520)
5708 - don't free dentries not owned by NFS code, check generation (4806)
5709 - fix lsm leak if mds_create_objects() fails (4801)
5710 - limit debug_daemon file size, always print CERROR messages (4789)
5711 - use transno after validating reply (3892)
5712 - process timed out requests if import state changes (3754)
5713 - update mtime on OST during writes, return in glimpse (4829)
5714 - add mkfsoptions to LDAP (4679)
5715 - use ->max_readahead method instead of zapping global ra (5039)
5716 - don't interrupt __l_wait_event() during strace
5718 - add software watchdogs to catch hung threads quickly (4941)
5719 - make lustrefs init script start after nfs is mounted
5720 - fix CWARN/ERROR duplication (4930)
5721 - return async write errors to application if possible (2248)
5722 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
5723 - print NAL number in %x format (4645)
5724 - update barely-supported suse-2.4.21-171 series (4842)
5725 - support for sles 9 %post scripts
5726 - support for building 2.6 kernel-source packages
5727 - support for sles km_* packages
5729 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
5732 - ignore -ENOENT errors in osc_destroy (3639)
5733 - notify osc create thread that OSC is being cleaned up (4600)
5734 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
5735 - reconstruct ost_handle() like mds_handle() (4657)
5736 - create a new thread to do import eviction to avoid deadlock (3969)
5737 - let lconf resolve symlinked-to devices (4629)
5738 - don't unlink "objects" from directory with default EA (4554)
5739 - hold socknal file ref over connect in case target is down (4394)
5740 - allow more than 32000 subdirectories in a single directory (3244)
5741 - fix blocks count for O_DIRECT writes (3751)
5742 - OST returns ENOSPC from object create when no space left (4539)
5743 - don't send truncate RPC if file size isn't changing (4410)
5744 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
5745 - bind to privileged port in socknal and tcpnal (3689)
5747 - rate limit CERROR/CWARN console message to avoid overload (4519)
5748 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
5749 - basic mmap support (3918)
5750 - kernel patch series update from b1_4 (4711)
5752 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
5755 - avoid crash during MDS cleanup with OST shut down (2775)
5756 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
5757 - don't use bad inodes on the MDS (3744)
5758 - dynamic object preallocation to improve recovery speed (4236)
5759 - don't hold spinlock over lock dumping or change debug flags (4401)
5760 - don't zero obd_dev when it is force cleaned (3651)
5761 - print grants to console if they go negative (4431)
5762 - "lctl deactivate" will stop automatic recovery attempts (3406)
5763 - look for existing locks in ldlm_handle_enqueue() (3764)
5764 - don't resolve lock handle twice in recovery avoiding race (4401)
5765 - revalidate should check working dir is a directory (4134)
5767 - don't always mark "slow" obdfilter messages as errors (4418)
5769 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
5772 - don't close LustreDB during write_conf until it is done (3860)
5773 - fix typo in lconf for_each_profile (3821)
5774 - allow dumping logs from multiple threads at one time (3820)
5775 - don't allow multiple threads in OSC recovery (3812)
5776 - fix debug_size parameters (3864)
5777 - fix mds_postrecov to initialize import for llog ctxt (3121)
5778 - replace config semaphore with spinlock (3306)
5779 - be sure to send a reply for a CANCEL rpc with bad export (3863)
5780 - don't allow enqueue to complete on a destroyed export (3822)
5781 - down write_lock before checking llog header bitmap (3825)
5782 - recover from lock replay timeout (3764)
5783 - up llog sem before sending rpc (3652)
5784 - reduce ns lock hold times when setting kms (3267)
5785 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
5786 - fix NULL deref and obd_dev leak on setup error (3312)
5787 - replace some LBUG about llog ops with error handling (3841)
5788 - don't match INVALID dentries from d_lookup and spin (3784)
5789 - hold dcache_lock while marking dentries INVALID and hashing (4255)
5790 - fix invalid assertion in ptlrpc_set_wait (3880)
5792 - add libwrap support for the TCP acceptor (3996)
5793 - add /proc/sys/portals/routes for non-root route listing (3994)
5794 - allow setting MDS UUID in .xml (2580)
5795 - print the stack of a process that LBUGs (4228)
5797 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
5800 - don't cleanup request in ll_file_open() on failed MDS open (3430)
5801 - make sure to unset replay flag from failed open requests (3440)
5802 - if default stripe count is 0, use OST count for inode size (3636)
5803 - update parent mtime/ctime on client for create/unlink (2611)
5804 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
5805 - free recovery state on server during a forced cleanup (3571)
5806 - unregister_reply for resent reqs (3063)
5807 - loop back devices mounting and status check on 2.6 (3563)
5808 - fix resource-creation race that can provoke i_size == 0 (3513)
5809 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
5810 - more debugging for page-accounting assertion (3746)
5811 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
5812 - avoid deadlock after precreation failure (3758)
5813 - fix race and lock order deadlock in orphan handling (3450, 3750)
5814 - add validity checks when grabbing inodes from l_ast_data (3599)
5816 - add /proc/.../recovery_status to obdfilter (3428)
5817 - lightweight CDEBUG infrastructure, debug daemon (3668)
5818 - change default OSC RPC parameters to be better on small clusters
5819 - turn off OST read cache for files smaller than 32MB
5820 - install man pages and include them in rpms (3100)
5821 - add new init script for (un)mounting lustre filesystems (2593)
5822 - run chkconfig in %post for init scripts (3701)
5823 - drop scimac NAL (unmaintained)
5825 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
5828 - clean kiobufs before and after use (3485)
5829 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
5830 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
5831 - change init script to fail more clearly if not run as root (1528)
5832 - allow clients to reconnect during replay (1742)
5833 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
5834 - don't do DNS lookups on NIDs too small for IP addresses (3442)
5835 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
5836 - fix cond_resched (3554)
5837 - only evict unfinished clients after recovery (3515)
5838 - allow bulk resend, prevent data loss (3570)
5839 - dynamic ptlrpc request buffer allocation (2102)
5840 - don't allow unlinking open directory if it isn't empty (2904)
5841 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
5842 - remove extraneous obd dereference causing LASSERT failure (3334)
5843 - don't use get_cycles() when creating temp. files on the mds (3156)
5844 - hold i_sem when setting i_size in ll_extent_lock() (3564)
5845 - handle EEXIST for set-stripe, set proper directory name (3336)
5847 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
5848 - fix ksocknal_fmb_callback() error messages (2918)
5850 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
5853 - don't copy lvb into (possibly NULL) reply on error (2983)
5854 - don't deref dentry after dput, don't free lvb on error (2922)
5855 - use the kms to determine writeback rpc length (2947)
5856 - increment oti_logcookies when osc is inactive (2948)
5857 - update client's i_blocks count via lvb messages (2543)
5858 - handle intent open/close of special files properly (1557)
5859 - mount MDS with errors=remount-ro, like obdfilter (2009)
5860 - initialize lock handle to avoid ASSERT on error cleanup (3057)
5861 - don't use cancelling-locks' kms values (2947)
5862 - use highest lock extent for kms, not last one (2925)
5863 - don't dereference ERR_PTR() dentry in error handling path (3107)
5864 - fix thread race in portals_debug_dumplog() (3122)
5865 - create lprocfs device entries at setup instead of at attach (1519)
5866 - common AST error handler, don't evict client on completion race (3145)
5867 - zero nameidata in detach_mnt in 2.6 (3118)
5868 - verify d_inode after revalidate_special is valid in 2.6 (3116)
5869 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
5870 - initialize RPC timeout timer earlier for 2.6 (3219)
5871 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
5872 - print nal/nid for unknown nid (3258)
5873 - additional checks for oscc recovery before doing precreate (3284)
5874 - fix ll_extent_lock() error return code for 64-bit systems (3043)
5875 - don't crash in mdc_close for bad permissions on open (3285)
5876 - zero i_rdev for non-device files (3147)
5877 - clear page->private before handing to FS, better assertion (3119)
5878 - tune the read pipeline (3236)
5879 - fix incorrect decref of invalidated dentry (2350)
5880 - provide read-ahead stats and refine rpc in flight stats (3328)
5881 - don't hold journal transaction open across create RPC (3313)
5882 - update atime on MDS at close time (3265)
5883 - close LDAP connection when recovering to avoid server load (3315)
5884 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
5885 - don't leak open file on MDS after open resend (3325)
5886 - serialize filter_precreate and filter_destroy_precreated (3329)
5887 - loop device shouldn't call sync_dev() for nul device (3092)
5888 - clear page cache after eviction (2766)
5889 - resynchronize MDS->OST in background (2824)
5890 - refuse to mount the same filesystem twice on same mountpoint (3394)
5891 - allow llmount to create routes for mounting behind routers (3320)
5892 - push lock cancellation to blocking thread for glimpse ASTs (3409)
5893 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
5894 - fix rare problem with rename on htree directories (3417)
5896 - allow default OST striping configuration per directory (1414)
5897 - fix compilation for qswnal for 2.6 kernels (3125)
5898 - increase maximum number of MDS request buffers for large systems
5899 - change liblustreapi to be useful for external progs like lfsck (3098)
5900 - increase local configuration timeout for slow disks (3353)
5901 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
5903 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5906 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
5907 - don't overwrite extent policy data in reply if lock was blocked (2901)
5908 - drop filter export grants atomically with removal from device (2663)
5909 - del obd_self_export from work_list in class_disconnect_exports (2908)
5910 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
5911 - swab reply message in mdc_close, other PPC fixes (2464)
5912 - fix destroying of named logs (2325)
5913 - overwrite old logs when running lconf --write_conf (2264)
5914 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
5915 - fix race in target_handle_connect (2898)
5916 - mds_reint_create() should take same inode create lock (2926)
5917 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
5918 - don't close files for self_export to avoid uninitialized obd (2936)
5919 - allow MDS with the same name as client node (2939)
5920 - hold dentry reference for closed log files for unlink (2325)
5921 - reserve space for all logs during transactions (2059)
5922 - don't evict page beyond end of stripe extent (2925)
5923 - don't oops on a deleted current working directory (2399)
5924 - handle hard links to targets without a parent properly (2517)
5925 - don't dereference NULL lock when racing during eviction (2867)
5926 - don't grow lock extents when lots of conflicting locks (2919)
5928 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
5931 - account for cache space usage on clients to avoid data loss (974)
5932 - lfsck support in lustre kernel code (2349)
5933 - reduce journal credits needed for BRW writes (2370)
5934 - orphan handling to avoid losing space on client/server crashes
5935 - ptlrpcd can be blocked, stopping ALL progress (2477)
5936 - use lock value blocks to assist in proper KMS, faster stat (1021)
5937 - takes i_sem instead of DLM locks internally on obdfilter (2720)
5938 - recovery for initial connections (2355)
5939 - fixes for mds_cleanup_orphans (1934)
5940 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
5941 - block all file creations until orphan recovery completes (1901)
5942 - client remove rq_connection from request struct (2423)
5943 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
5944 - recovery timer race (2670)
5945 - mdc_close recovey bug (2532)
5946 - ptlrpc cleanup bug (2710)
5947 - mds timeout on local locks (2588)
5948 - namespace lock held during RPCs (2431)
5949 - handle interrupted sync write properly (2503)
5950 - don't try to handle a message that hasn't been replied to (2699)
5951 - client assert failure during cleanup after abort recovery (2701)
5952 - leak mdc device after failed mount (2712)
5953 - ptlrpc_check_set allows timedout requests to complete (2714)
5954 - wait for inflight reqs when ptlrpcd finishes (2710)
5955 - make sure unregistered services are removed from the srv_list
5956 - reset bulk XID's when resending them (caught by 1138 test)
5957 - unregister_bulk after timeout
5958 - fix lconf error (2694)
5959 - handle write after unfinished setstripe, stripe-only getstripe (2388)
5960 - readahead locks pages, leaves pending causing memory pressure (2673)
5961 - increase OST request buffers to 4096 on large machines (2729)
5962 - fix up permission of existing directories in simple_mkdir (2661)
5963 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
5964 - don't assign transno to errored transactions (2742)
5965 - don't delete objects on OST if given a bogus objid from MDS (2751)
5966 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
5967 - if rq_no_resend, then timeout request after recovery (2432)
5968 - fix MDS llog_logid record size, 64-bit array alignment (2733)
5969 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
5970 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
5971 - MDS recovery shouldn't delete objects that it has given out (2730)
5972 - if enqueue arrives after completion, don't clobber LVB (2819)
5973 - don't unlock pages twice when trigger_group_io returns error (2814)
5974 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
5975 - don't write pages to disk if there was an error (1450)
5976 - don't ping imports that have recovery disabled (2676)
5977 - take buffered bytes into account when balancing socknal conn (2817)
5978 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
5979 - reconnect unlink llog connection after MDS reconnects to OST (2816)
5980 - remove little-endian swabbing of llog records (1987)
5981 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
5982 - retry reposting request buffers if they fail (1191)
5983 - grow extent at grant time to avoid granting a revoked lock (2809)
5984 - lock revoke doesn't evict page if covered by a second lock (2765)
5985 - disable VM readahead to avoid reading outside lock extents (2805)
5987 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
5988 - updated kernel patches for hp-2.4.20 kernel (2681)
5990 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
5993 - fix truncated write corruption (2366)
5994 - fix for failed assertion in iopen_connect_dentry (1792,2517)
5996 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
5997 - ensure len(uuid) < 37 in lmc (1171)
5998 - fix ia64 OOPS in llog_test (2255)
5999 - zero end of page at obdfilter for partial page writes (2648)
6000 - don't leave stale dentries around after renames (bug 2428)
6001 - fix timeouts when evicting a client with a single lock held (2642)
6002 - set deadline for the initial HELLO message to drain (2634)
6003 - print out dotted-quad IP addresses in the socknal (2302)
6005 - additional debugging for MDS client eviction problem (2443)
6006 - fix mkfsoptions support for osts (2603, 2604)
6008 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
6011 - add series for the vanilla 2.6.0 kernel
6012 - add series for the vanilla 2.4.24 kernel
6013 - add series for a cray x86/64 UL kernel drop
6014 - fix xattr patches for the vanilla 2.4.19 series
6016 - generate true UUIDs in lmc (1171)
6017 - have portals stack dumping break in UML (2466)
6018 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
6019 - allocate with _NOFS in ldlm to avoid deadlock (1933)
6020 - wake callback waiting threads on client eviction (2460)
6021 - Add --ptldebug and --subsystem to lmc (1719)
6022 - update assertion to allow safe interrupt allocation
6023 - set rq_no_resend for cancel requests (2432)
6024 - recalculate ptlrpcd timeout after resend (2494)
6025 - call vfs_rmdir when removing pending directories (2368)
6026 - fix renaming a file to itself (2429)
6027 - lmc creates a default one-stripe lov (2454)
6028 - expand procfs space to handle large clusters (2326)
6029 - increase UML stack to avoid overflow
6030 - update lconf's list of debug and subsystem masks
6031 - fix lfs find --obd (2510)
6032 - /proc tunable for disabling filter read caching (2591)
6033 - stop rpm packages from altering slapd.conf (2301)
6034 - disable nagle in the socknal under 0conf (2578)
6035 - choose mds inode size based on stripe count (2572)
6036 - fix kernel-source rpm problems (2516)
6038 - add --disable-doc to avoid pdf generation (2421)
6039 - update documentation, tests, type-os, comments
6040 - avoid format warnings on ia64
6041 - remove the TOE NAL
6042 - tiny code cleanups by removing unused fields
6044 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
6047 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
6048 - remove the most blatant lies from BUILDING file (2371)
6049 - change default debug level to reasonable production setting
6050 - reduce client side cache size to reduce cache flush time
6051 - reduce max RPCs in flight to avoid unnecessary file fragmentation
6052 - make TCP zerocopy and pinger support enabled by default (2476)
6053 - sync writes completed after process exits caused crashes (2319)
6054 - maintain correct mount count on the MDS (2356)
6055 - backout 1557, because 2316 wasn't really fixed
6056 - better file I/O statistics gathering in /proc
6057 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
6058 - another kernel patch to fix zero-copy TCP function export
6059 - don't take duplicate lock when processing re-sent getattr (2420)
6060 - lctl uses obd_self_export instead of creating new conn (2353)
6061 - MDS/OST recovery case which requires object creation asserted (2425)
6062 - move lfs from /usr/sbin to /usr/bin in packages
6063 - fix race between mds_client_add and mds_client_free (2417)
6064 - use kmalloc instead of slabs in portals (2430)
6065 - don't create duplicate records when a failover MDS is present (2442)
6066 - remove unnecessary mount age check (2332)
6067 - don't remove directory inodes from locks prematurely (2451)
6068 - don't break if MDS service name is the same as hostname (2103)
6069 - fix races in client write RPC generation when cache full (2482)
6071 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
6074 - remove now-unused request->rq_obd (278)
6075 - if an allocation fails, print out how much memory we've used (1933)
6076 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
6077 - add the "configurable stack size" patch to most series files (1256)
6078 - ability to write large log records, for 100+ OST configs (2306)
6079 - fix NULL deref when filter_prep fails (2314)
6080 - fix operator precedence error in filter_sync
6081 - dynamic allocation of socknal TX descriptors (2315)
6082 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
6083 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
6084 - ensure that CWARN messages in Portals always get to the syslog
6085 - __init/__exit are not for prototype decls (ldlm_init/exit)
6086 - x86-64 compile warning fixes
6087 - fix gateway LMC keyword conflict (2318)
6088 - fix MDS lock inversions in getattr/reint paths (1844)
6089 - fix a rare lock re-ordering bug, which caused deadlock (2322)
6090 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
6091 - DLM race condition prevented some lock evictions (2328)
6092 - ENOMEM detection and retry on socknal sends (2230)
6093 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
6094 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
6095 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
6096 - better error messages when a client is rejected during recovery (1505)
6097 - avoid cancelling locks which were never granted, after failure (2330)
6098 - fix i_sem/journal inversion in mds_client_add (2333)
6099 - fix truncate/getattr lock cycle deadlock (2334)
6100 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
6101 - fix two rare exit paths which could leak an l_lock() ref (2321)
6102 - fencepost error in MDS/OST orphan recovery (2226)
6103 - make log record alignment 8 bytes (1988)
6104 - lstripe now fails when requested offset > ost_count (2237)
6105 - ensure that all kernel series have a complete list.h (1607)
6106 - fix crashes in special-file operations (2316)
6107 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
6109 - allow configurable automake binary, for testing new versions
6110 - small update to the lfs documentation
6112 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
6114 * fix negative export reference count in fsfilt_sync (2312)
6116 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
6117 * release candidate 0.9.1
6119 - orphans are moved into the PENDING directory for possible recovery
6120 - replayed opens now open by fid for orphan/rename safety (1042)
6121 - last close of an orphan inode generates a transno (683)
6122 - chdir() and mount() now pin the directory entry (1020)
6123 - avoid CERROR in normal ll_setattr_raw() error case (1500)
6124 - discard very old requests without processing them (1502)
6125 - remove some common, well-understood CERRORs (1505)
6126 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
6127 - clear "grant" flags in OST replies until OST grant code lands (1644)
6128 - fix read performance by not clobbering i_blksize on client (1598)
6129 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
6130 - make LRU size a /proc tunable, clears locks when reduced (707)
6131 - fix some lprocfs rot that prevented ptlbd from loading (1732)
6132 - server locks take references on exports now (1558)
6133 - build fixes for 2.4.20-rh trees (1663)
6134 - return an error from lov_create if all OSCs are inactive (1751)
6135 - fix import levels when a reconnect happens without a timeout (1597)
6136 - exit early from mds_open if we get a lookup error (1749)
6137 - partial page read at EOF wouldn't wait for disk before sending (1642)
6138 - avoid NULL deref in obdfilter when reading page past EOF (1592)
6139 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
6140 - fix LBUG in ll_it_open_error with rc = -2 (1861)
6141 - write/truncate lock inversion (1639)
6142 - Don't auto-load obdclass, portals modules during cleanup (1495)
6143 - fix timestamps from jumping to "now" (1763)
6144 - extra journal assertions (1648)
6145 - add an extra multiunlink test (1771)
6146 - fix read_record/write_record API (1776)
6147 - fix leak of offset_extent, possible incorrect i_size later (1772)
6148 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
6149 - Debugging for the kqswnal_get_idle_tx problems (1820)
6150 - Allow recovery to be attempted multiple times (1536)
6151 - Write out MDS last_rcvd file after it is first created (1600)
6152 - Fix tx_descriptor leak in failed transmit situations (1827)
6153 - ext3 journaling fixes for assertion failure after IO error (1871)
6154 - class_export_put() on freed export after completion AST error (1896)
6155 - Fix revalidate looping in VFS (1322)
6156 - Don't access a freed export during MDS_REINT timeout (1521)
6157 - Add open-unlink recovery support on the MDS (1673,1764)
6158 - Return an error if no MDS data was read from last_rcvd (1946)
6159 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
6160 - Don't setup a disk device that doesn't match exported UUID (317)
6161 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
6162 - avoid committing NULL handle in force close
6163 - local.sh is now a one-stripe LOV configuration
6164 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
6165 - fix ext3 htree duplicate directory entry corruption (1516)
6166 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
6167 - update to new LOV EA format (2097)
6168 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
6170 2003-06-15 Phil Schwan <phil@clusterfs.com>
6173 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
6174 - per-import/export recovery handling (958, 931, 959)
6175 - multiple last-rcvd slots, for serving multiple FSes (949)
6176 - connections are again shared between multiple imp/exports (963, 964)
6177 - "umount -f" would hang if any requests needed to be sent (393, 978)
6178 - avoid pinning large req buffer by copying for queued messages (989)
6179 - add "uuid" to "lctl device" command to help upcalls (991)
6180 - "open" RPCs with transnos would confuse recovery counters (1037)
6181 - do proper endian conversion of all wire messages (288, 340, 891)
6182 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
6183 - call waiting locks callback from LDLM recovery thread (1127, 1151)
6184 - fix ptlrpc_connection leak in target_handle_connect (1174)
6185 - fix import refcounting bug in OST and MDS cleanup (1134)
6186 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
6187 - fix very unlikely obd_types race condition (501)
6188 - remove osc_open hack for echo_client (1187)
6189 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
6190 - a failure in read_inode2 leads to deadlock (1139)
6191 - cancel ack-locks as soon as transaction is committed (1072)
6192 - fix major leaks and crashes in the bulk I/O path (937, 1057)
6193 - make sure to commitrw after any preprw to avoid deadlock (1162)
6194 - failing to execute a file in a lustre FS would lock inode (1203)
6195 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
6196 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
6197 - obd_uuid2tgt would walk off the end of the list (1255)
6198 - on IA64 the osc would give portals incorrect bulk size (1258)
6199 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
6200 - fix lock inversion caused by new llite matching code (1282)
6201 - limit the number of dirty pages on a client to 10MB (1286)
6202 - timed out locks were not being corrected cancelled (1289)
6203 - fix O_DIRECT above 4GB on IA-32 (1292)
6204 * major user-visible changes
6205 - fail out/fail over policy now controlled by the upcall (993)
6207 - add OBD_PING to check server availability and failure (954)
6208 - lustre messages are now sent in sending host order (288, 340, 891)
6209 - add eadatalen to MDS getattr reply (340)
6210 - OST read replies may contain second buffer, with per-page status (593)
6212 2003-03-11 Phil Schwan <phil@clusterfs.com>
6215 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
6216 - failed open()s could cause deadlock; fixed (867, 869)
6217 - stop cancelling OST locks when files are closed (481)
6218 - overlapping XID spaces caused network corruption (851, 853)
6219 - fix unsafe fsfilt counter arithmetic; change to atomic_t
6220 - setattr_raw added, to do single-RPC, server-side setattrs
6221 - lmc/lconf syntax change for OST UUIDs
6222 - fix crashy race condition between ptlrpc_free_req and osc_close
6223 - don't use request in mdc_enqueue if we hit a timeout (889)
6224 - don't set the inode i_size for regular files from the MDS (896)
6225 - handle out of order completion AST (842)
6226 - don't LBUG if a lock request times out after receiving AST (913)
6227 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
6228 - if a bad lock AST arrives, send an error instead of dropping entirely
6229 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
6230 - fix leak in bulk IO when only partially completed (899, 900, 926)
6231 - fix O_DIRECT for ia64 (55)
6232 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
6233 - C-z after timeout could hang a process forever; fixed (977)
6235 - client-side I/O cache (678, 924, 929, 941, 970)
6237 - READPAGE and SETATTRs which don't take server-side locks get
6240 2003-02-11 Phil Schwan <phil@clusterfs.com>
6243 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
6244 - Fix fsfilt_extN_readpage() to read a full page of directory
6245 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
6246 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
6247 - Fix imbalanced LOV object allocation and out-of-bound access (469)
6248 - Most intent operations were removed, in favour of a new RPC mode
6249 that does a single RPC to the server and bypasses most of the VFS
6250 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
6251 - Aggressively cancel local locks on DLM servers
6252 - mds_reint_unlink sends EA to the client if it's the last nlink.
6253 client uses that EA to unlink OST objects.
6254 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
6255 - recursive symlinks were fixed (439)
6256 - fixed NULL deref in DEBUG_REQ
6257 - filter_update_lastobjid no longer calls sync, which annoyed extN
6258 - fixed multi-client small-writes to a single file problem (445)
6259 - fixed mtime updates during file writes (607)
6260 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
6261 - fixed bug in obd_brw_read/write() (under guise of testing 367)
6262 - fixed Linux OST size reporting problem (444, 656)
6263 - OST now updates object mtime with writes or setattr (607, 619)
6264 - client verifies file size before zeroing page past EOF (445)
6265 - OST now writes last allocated objid to disk with allocation (108)
6266 - LOV on echo now works (409)
6268 - mds_reint_unlink sends a new buffer, with the EA included. this
6269 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
6270 set if a regular file was being unlinked, and it was the last link
6271 - use PtlGet from the target for bulk writes (315)
6272 - OST now updates object mtime with writes or setattr (607, 619)
6273 - LDLM now has a grant-time callback to revalidate locked items, if
6275 - Many MDS operations were reorganized to combat race conditions
6277 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
6278 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
6280 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
6283 - Fully reactivate OST imports after reconnection (512, others)
6284 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
6285 - More graceful error handling for truncating on dead OST (515)
6286 - Don't error out unless we're actually accessing dead stripes (474)
6287 - Fix garbage sizes when stripes are missing (410)
6288 - LRU counters were broken, causing constant lock purge (433, 432)
6289 - garbage on read from stripes with failed OSTs (441)
6290 - mark OSCs as active before reconnecting during recovery (438)
6291 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
6292 - lfind did not preserve OST order in output (443)
6293 - symlinks cause hung clients, incorrect data (439)
6294 - stop dereferencing request after dropping refcount (457)
6295 - don't LASSERT(spin_is_locked) on non-SMP (455)
6296 - fixes for many rename() bugs
6297 - fstat didn't correctly synchronize attributes (399)
6298 - server must handle lock cancellation during blocking AST prep (487)
6299 - bulk descriptors were free()d too soon (511)
6300 - fix paths in lconf, which would load incorrect modules (451, 507)
6301 - fix confusing lconf 'host not found' error message (386)
6302 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
6303 - fix race condition in mdc_blocking_ast() for inode access (526)
6304 - fix lov_unpackmd() unpacking wrong number of stripes (537)
6305 - fix lov_set_osc_active() marking wrong OSC inactive (440)
6306 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
6307 - fix multiple writes of stripe MD to MDS (358, maybe 519)
6308 - fix lstripe in several ways (kernel side) (527)
6309 - fix request leak in ldlm_cli_enqueue (262)
6310 - incorrect OSC was marked inactive after OST failure
6311 - call mds_fs_cleanup before unmounting filesystem (524)
6312 - fix races between taking ns_lock and ldlm_lock_change_resource
6313 - fix races updating LOV export open file list
6314 - fix lov_enqueue error path, avoid decref-ing bad lock handle
6315 - fix recovery NULL deref in ldlm_cli_cancel_unused
6316 - fix some DLM races by using new hash table for lock handles (419)
6317 - permit the client to specify desired inodes, at replay
6318 - duplicate requests when we queue them for replay reintegration
6319 - fix last_rcvd offset calculation
6320 - sync after each recovered transaction, so we always make progress
6321 - never, not always, ERESTART requests without transnos
6322 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
6324 - skip replay if the MDS says that the client is already connected
6325 - don't check for a recovery-enabled export to match lctl's UUID
6326 - don't INC_USE_COUNT for phantom exports
6327 - don't crash when cleaning up phantom exports (567)
6328 - don't double-finish or set replay data for errored mdc_open requests
6329 - abort requests when they time out, so we don't get old replies
6330 - send/receive replies for AST messages again
6331 - if the client says that it doesn't have the lock, cancel it on the
6333 - if we timeout during I/O, don't try to cancel an in-use lock; instead
6334 mark it as destroyed, it will all work out when decref is called
6335 - fix module use counts (22, 581)
6337 - ASTs now expect a reply (server cancels lock on error reply)
6339 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
6342 - fix many simultaneous client startup (392)
6343 - fix dentry->d_it clobbering
6344 - credentials weren't being shipped for readdir/getattr operations
6345 - remove invalid assertions triggered during some concurrent MD
6347 - proper Lustre versions added (336, 389)
6348 - fix memory leak for create error case (398)
6349 - fix LOV locking bug that would get cli/srv out of sync
6350 - fix echo client over LOV (409)
6351 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
6352 - fix double-O_EXCL intent crash (424)
6353 - avoid sending multiple lock CANCELs (352)
6355 - MDS can do multi-client recovery (modulo bugs in new code)
6357 - many updates, edits, cleanups
6359 2002-11-18 Phil Schwan <phil@clusterfs.com>
6362 - fix null d_it dereference (346)
6363 - fix full OST/dbench hang (333)
6364 - fix permission problem with file removal (286)
6365 - fix removal of OSCs from LOV when they fail
6366 - fix NULL deref during bulk timeout (214)
6367 - fix problems related to multiple filesystems on one MDS (241)
6368 - fixed serious subtle metadata locking bugs
6369 - free locks on clients when inodes are removed due to memory
6371 - fix inode pointer in lock data (285)
6372 - partial support for multiple MDS on a single host (241)
6373 - data locks weren't cancelled at clear_inode time (290, 311)
6374 - intent locks could lead to unbounded lock growth (205)
6375 - added a maximum lock count, an LRU list, and a flusher
6376 - fix multiple rename (365)
6377 - properly abstracted the echo client
6378 - OSC locked 1 byte too many; fixed
6379 - rewrote brw callback code:
6380 - fixed recovery bugs related to LOVs (306)
6381 - fixed too-many-pages-in-one-write crash (191)
6382 - fixed (again) crash in sync_io_timeout (214)
6383 - probably fixed callback-related race (385)
6385 - Add capability to MDS protocol
6386 - LDLM cancellations and callbacks on different portals
6388 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
6391 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
6392 - defer open object destruction to close time (601981, 138)
6393 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
6394 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
6395 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
6396 - don't call obd_setattr in ll_file_release for destroyed objects
6397 * protocol change to lustre_msg: move |version| and add |flags|
6398 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
6399 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
6400 * added replay of create, unlink, link and rename operations during
6401 MDS failover; recovery should be much more robust now
6402 * remove failed OSCs from LOVs (only lov_create uses this so far)
6403 * the lustre-HOWTO was brought (more) up to date (582544)
6405 2002-10-23 Phil Schwan <phil@clusterfs.com>
6408 - in-use dentries weren't being reused properly (617851)
6409 - prevent multiple LDLM setup (599178)
6410 - fix LOV size calculations for truncate (617853)
6411 - fix client handling of MDS intent errors (POSIX)
6412 - fix permission bug in lovstripe.c test (624321)
6413 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
6414 - truncate past end of file could corrupt data
6415 - proper cleanup after timeouts, crashes, etc (592524, 550815)
6416 - a race in recovery could return ETIMEDOUT to apps (623947)
6417 - building outside the source directory was fixed
6418 * the lustre-HOWTO was brought (more) up to date (582544)
6419 * major progress was made on recovery functionality
6421 2002-10-10 Phil Schwan <phil@clusterfs.com>
6424 - recovery deadlock fix
6425 - rm -rf causes LBUG fix (617817)
6426 - file open by multiple tasks fix (618962)
6427 - directory permissions bugs (602707 and 620007)
6428 - journal_stop fixed with locking (611313)
6429 - O_APPEND failures resolved (618273, perhaps 614459)
6430 - lconf PATH fix (619770)
6431 - IA64 build fix (621450)
6432 - RPC buffer sizes scale with amount of memory
6434 2002-10-01 Phil Schwan <phil@clusterfs.com>
6437 - locks would be cancelled without throwing away data pages,
6438 resulting in inconsistent data (605627)
6439 - inode attributes were not always being refreshed (605627, 612449)
6440 - lconf now continues to cleanup after lctl reports an error
6441 - MDS now enforces user permissions (602707)
6442 - lprocfs cleanup fixed, but not yet enabled (614157)
6443 - fixed infinite server hang, should a client not respond to an AST
6444 - avoid going into recovery if user calls readlink() with a buffer
6445 that's too small (613941)
6446 - AST RPCs no longer require replies (614867) -- this may be changed
6447 - don't crash server if client sends an IOV that's too big (611336)
6448 - fixed lock conversion deadlock (611892)
6449 - fixed the following of symlinks (614622)
6450 * recovery: the server can remove locks from a client that dies, other
6451 clients can make progress
6452 * more extN patch fixes
6453 * compile-time configurable ptlrpc buffer allocations
6455 - collaborative read cache document
6456 - Lustre Lite Performance CDR document-in-progress
6458 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6461 - fix typo in patch-2.4.18
6463 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6466 - clear ptlrpc request each time in handle_incoming_request()
6467 - unlink of files now destroys the object on the OST
6469 2002-09-19 Peter Braam <braam@clusterfs.com>
6471 * add hard link support
6472 * change obdfile creation method
6473 * kernel patch changed
6475 2002-09-19 Peter Braam <braam@clusterfs.com>
6478 - stack overflow bug in extN fixed
6480 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
6482 * documentation updates
6483 - add man pages for config tools
6484 - update tests/README to describe testing with new config tools
6485 - finish metadata API descriptions
6486 * bug fixes and cleanups
6487 - statfs workaround for 16TB limit
6488 - LOV stripe allocation improved, can stripe on subset of OSTs
6489 - LOV file size/IO offset was wrong for files > 4GB in size
6490 - object EA data was being dropped, caused files to be unreadable
6491 - memory overflow with non-LOV OST caused memory corruption
6492 - fixed regression tests to work with new config tools, obdfilter
6493 - fixed bug when directory size became larger than 1 block
6494 - fixed bug (for single client case) when PWD was deleted
6495 - invalidate local directory pages when doing intent-based ops
6496 - avoid LDLM oops when lock callback contained bad data
6498 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
6500 * documentation updates
6501 * bug fixes and cleanups
6502 - configuration tools
6505 - 64-bit compile warnings
6506 - 64-bit internal statfs data
6508 * test_brw on persistent OST devices
6510 * lprocfs (disabled)
6512 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
6514 * documentation updates
6515 * bug fixes and cleanups
6516 * configuration tools
6518 2002-08-30 Peter J. Braam <braam@clusterfs.com>
6521 * many small fixes to 0_5_4
6522 * io/network handling
6523 * thinkos in MDS operations
6525 2002-08-24 Peter J. Braam <braam@clusterfs.com>
6528 * crucial basic fixes to 0.5.3
6529 * IOR, Iozone work over Elan
6532 2002-08-07 Phil Schwan <phil@clusterfs.com>
6533 * version 0_5_3, our first alpha
6534 * we use the new Portals iovs
6535 * documentation updates
6536 * bug fixes and cleanups
6537 * small changes in the DLM wire protocol
6539 2002-07-25 Peter J. Braam <braam@clusterfs.com>
6540 * version 0_5_1 with some initial stability,
6541 * locking on MD and file I/O.
6542 * documentation updates
6543 * several bug fixes since 0.5.0
6544 * small changes in wire protocol
6546 2002-07-18 Phil Schwan <phil@clusterfs.com>
6548 * delivered as Lustre Light Alpha
6549 * fixed a crash after handling invalid MDS requests
6550 * fixed directory pages for architectures with non-4k pages sizes
6552 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
6553 * release version v0_4_4
6554 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
6556 * Fixes a number of interruption problems with OST operations.
6557 * Update documentation for portals header changes
6558 * Move all wire protocol structs/defines to lustre_idl.h
6559 * Fixes symlink length bug.
6560 * Add tcpdump to repository.
6562 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
6563 * release version v0_4_3
6564 * Fixes statfs for inodes on extN.
6565 * Fixes bug in runtests which would delete /etc/hosts.
6566 * Use 64-bit object IDs wherever possible (not into VFS though)
6567 Remove ost_get_info, which is unused by lustre, and out of date.
6569 2002-07-03 Peter Braam <braam@clusterfs.com>
6570 * release version v0_4_2 Fixes a lookup error (type not passed)
6571 * move forward to head of Portals
6572 * move forward to latest Lustre kernel
6574 2002-06-25 Peter Braam <braam@clusterfs.com>
6575 * release version v0_4_1. Hopefully stable on single node use.