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 2.6.32-642.13.1.el6 (RHEL6.8)
10 3.10.0-514.10.2.el7 (RHEL7.3)
11 3.0.101-0.47.71 (SLES11 SP3)
12 3.0.101-94 (SLES11 SP4)
13 3.12.69-60.64.32 (SLES12 SP1)
14 4.4.38-93 (SLES12 SP2)
15 vanilla linux 4.6.7 (ZFS only)
16 * Client known to build on unpatched kernels:
17 2.6.32-431.29.2.el6 (RHEL6.5)
18 2.6.32-504.30.3.el6 (RHEL6.6)
19 2.6.32-573.26.1.el6 (RHEL6.7)
20 2.6.32-642.13.1.el6 (RHEL6.8)
21 3.10.0-514.10.2.el7 (RHEL7.3)
22 3.0.101-0.47.71 (SLES11 SP3)
23 3.0.101-94 (SLES11 SP4)
24 3.12.69-60.64.32 (SLES12 SP1)
25 4.4.38-93 (SLES12 SP2)
27 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
28 * Recommended ZFS / SPL version: 0.6.5.9
29 * Tested with ZFS / SPL version: 0.6.5.9
30 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
31 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
32 stack overflow. For more information, please refer to bugzilla 17630.
33 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
34 the re-exporting nodes to be mounted with "32bitapi" mount option
36 --------------------------------------------------------------------------------
38 12-07-2016 Intel Corporation
40 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
41 for currently supported client and server kernel versions.
42 * Server known to build on patched kernels:
43 2.6.32-431.29.2.el6 (RHEL6.5)
44 2.6.32-504.30.3.el6 (RHEL6.6)
45 2.6.32-573.26.1.el6 (RHEL6.7)
46 2.6.32-642.11.1.el6 (RHEL6.8)
47 3.10.0-514.el7 (RHEL7.3)
48 3.0.101-0.47.71 (SLES11 SP3)
49 3.0.101-80 (SLES11 SP4)
50 3.12.62-60.64.8 (SLES12 SP1)
51 vanilla linux 4.5.7 (ZFS only)
52 * Client known to build on unpatched kernels:
53 2.6.32-431.29.2.el6 (RHEL6.5)
54 2.6.32-504.30.3.el6 (RHEL6.6)
55 2.6.32-573.26.1.el6 (RHEL6.7)
56 2.6.32-642.11.1.el6 (RHEL6.8)
57 3.10.0-514.el7 (RHEL7.3)
58 3.0.101-0.47.71 (SLES11 SP3)
59 3.0.101-80 (SLES11 SP4)
60 3.12.62-60.64.8 (SLES12 SP1)
62 * Recommended e2fsprogs version: 1.42.13.wc5 or newer
63 * Recommended ZFS / SPL version: 0.6.5.8
64 * Tested with ZFS / SPL version: 0.6.5.8
65 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
66 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
67 stack overflow. For more information, please refer to bugzilla 17630.
68 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
69 the re-exporting nodes to be mounted with "32bitapi" mount option
71 --------------------------------------------------------------------------------
73 02-29-2016 Intel Corporation
75 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
76 for currently supported client and server kernel versions.
77 * Server known to build on patched kernels:
78 2.6.32-431.29.2.el6 (RHEL6.5)
79 2.6.32-504.30.3.el6 (RHEL6.6)
80 2.6.32-573.12.1.el6 (RHEL6.7)
81 3.10.0-327.3.1.el7 (RHEL7.2)
82 3.0.101-0.47.71 (SLES11 SP3)
83 3.0.101-68 (SLES11 SP4)
84 vanilla linux 4.2.1 (ZFS only)
85 * Client known to build on unpatched kernels:
86 2.6.32-431.29.2.el6 (RHEL6.5)
87 2.6.32-504.30.3.el6 (RHEL6.6)
88 2.6.32-573.12.1.el6 (RHEL6.7)
89 3.10.0-327.3.1.el7 (RHEL7.2)
90 3.0.101-0.47.71 (SLES11 SP3)
91 3.0.101-68 (SLES11 SP4)
94 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
95 * Recommended ZFS / SPL version: 0.6.4.2
96 * Tested with ZFS / SPL version: 0.6.4.2
97 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
98 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
99 stack overflow. For more information, please refer to bugzilla 17630.
100 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
101 the re-exporting nodes to be mounted with "32bitapi" mount option
103 --------------------------------------------------------------------------------
105 03-10-2015 Intel Corporation
107 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
108 for currently supported client and server kernel versions.
109 * Server known to build on patched kernels:
110 2.6.32-431.29.2.el6 (RHEL6.5)
111 2.6.32-504.8.1.el6 (RHEL6.6)
112 3.0.101-0.46 (SLES11 SP3)
113 * Client known to build on unpatched kernels:
114 2.6.32-431.29.2.el6 (RHEL6.5)
115 2.6.32-504.8.1.el6 (RHEL6.6)
116 3.10.0-123.20.1.el7 (RHEL7)
117 3.0.101-0.46 (SLES11 SP3)
118 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
119 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
120 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
121 stack overflow. For more information, please refer to bugzilla 17630.
122 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
123 the re-exporting nodes to be mounted with "32bitapi" mount option
125 Severity : enhancement
127 Description: control OST-index in IDIF via ROCOMPAT flag.
128 Details : Introduce new flag OBD_ROCOMPAT_IDX_IN_IDIF that is stored in the
129 last_rcvd file. For new formatted OST device, it will be auto set;
130 for the case of upgrading from old OST device, you can enable it
131 via the lproc interface osd-ldiskfs.index_in_idif. With such flag
132 enabled, for new created OST-object, its IDIF-in-LMA will contain
133 the OST-index; for the existing OST-object, the OSD will convert
134 old format IDIF as new format IDIF with OST-index stored in the
135 LMA EA when accessing such OST-object or via OI scrub. Once such
136 flag is enabled, it cannot be reverted back, so the system cannot
137 be downgraded to the orignal incompatible version.
139 --------------------------------------------------------------------------------
141 07-30-2014 Intel Corporation
143 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
144 for currently supported client and server kernel versions.
145 * Server known to build on patched kernels:
146 2.6.32-431.20.3.el6 (RHEL6.5)
147 3.0.101-0.31 (SLES11 SP3)
148 * Client known to build on unpatched kernels:
149 2.6.32-431.20.3.el6 (RHEL6.5)
150 3.0.101-0.31 (SLES11 SP3)
153 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
154 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
155 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
156 stack overflow. For more information, please refer to bugzilla 17630.
157 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
158 the re-exporting nodes to be mounted with "32bitapi" mount option
160 --------------------------------------------------------------------------------
162 10-31-2013 Intel Corporation
164 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
165 * Server builds for kernels:
166 2.6.32-358.18.1.el6 (RHEL6.4)
167 * Client builds for unpatched kernels:
168 2.6.32-358.18.1.el6 (RHEL6.4)
169 3.0.92-0.8 (SLES11 SP3)
171 * Recommended e2fsprogs version: 1.42.7.wc2 or newer
172 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
173 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
174 stack overflow. For more information, please refer to bugzilla 17630.
175 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
176 the re-exporting nodes to be mounted with "32bitapi" mount option
178 --------------------------------------------------------------------------------
180 05-30-2013 Intel Corporation
182 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
183 * Server builds for kernels:
184 2.6.32-358.6.2.el6 (RHEL6.4)
185 * Client builds for unpatched kernels:
186 2.6.32-358.6.2.el6 (RHEL6.4)
187 3.0.74-0.6.6 (SLES11 SP2)
189 * Recommended e2fsprogs version: 1.42.7.wc1 or newer
190 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
191 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
192 stack overflow. For more information, please refer to bugzilla 17630.
193 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
194 the re-exporting nodes to be mounted with "32bitapi" mount option
196 --------------------------------------------------------------------------------
198 9-30-2012 Whamcloud, Inc.
200 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
201 * Server support for kernels:
202 2.6.32-279.5.1.el6 (RHEL6)
203 * Client support for unpatched kernels:
204 2.6.32-279.5.1.el6 (RHEL6)
206 2.6.18-238.19.1.el5 (RHEL5)
207 2.6.16.60-0.69.1 (SLES10)
208 * Recommended e2fsprogs version: 1.41.90.wc4
209 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
210 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
211 stack overflow. For more information, please refer to bugzilla 17630.
212 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
213 the re-exporting nodes to be mounted with "32bitapi" mount option
215 Severity : enhancement
217 Description: Update RHEL6.3 kernel to 2.6.32-279.5.1.el6.
219 Severity : enhancement
221 Description: Update RHEL6.3 kernel to 2.6.32-279.2.1.el6.
223 Severity : enhancement
225 Description: Update RHEL6.2 kernel to 2.6.32-220.17.1.el6
227 Severity : enhancement
229 Description: Update RHEL6.2 kernel to 2.6.32-220.13.1.el6
231 Severity : enhancement
233 Description: Update RHEL6.2 kernel to 2.6.32-220.7.1.el6
235 --------------------------------------------------------------------------------
237 04-30-2012 Whamcloud, Inc.
239 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
240 * Server support for kernels:
241 2.6.32-220.4.2.el6 (RHEL6)
242 * Client support for unpatched kernels:
243 2.6.32-220.4.2.el6 (RHEL6)
245 2.6.18-238.19.1.el5 (RHEL5)
246 2.6.16.60-0.69.1 (SLES10)
247 * Recommended e2fsprogs version: 1.41.90.wc4
248 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
249 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
250 stack overflow. For more information, please refer to bugzilla 17630.
251 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
252 the re-exporting nodes to be mounted with "32bitapi" mount option
254 Severity : enhancement
256 Description: Update RHEL6.2 kernel to 2.6.32-220.4.2.el6
258 Severity : enhancement
260 Description: Update RHEL6.2 kernel to 2.6.32-220.4.1.el6
262 Severity : enhancement
264 Description: Update RHEL6.2 kernel to 2.6.32-220.el6
266 Severity : enhancement
268 Description: Update RHEL6.1 kernel to 2.6.32-131.17.1.el6
270 --------------------------------------------------------------------------------
272 09-30-2011 Whamcloud, Inc.
274 * Server support for kernels:
275 2.6.18-238.19.1.el5 (RHEL5)
276 2.6.32-131.6.1.el6 (RHEL6)
277 * Client support for unpatched kernels:
278 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
279 2.6.18-238.19.1.el5 (RHEL5)
280 2.6.32-131.6.1.el6 (RHEL6)
281 2.6.29.4-167.fc11 (FC11)
282 2.6.16.60-0.69.1 (SLES10)
284 * Recommended e2fsprogs version: 1.41.90.wc3
285 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
286 Lustre filesystem with 4K stack may cause a stack overflow. For more
287 information, please refer to bugzilla 17630.
288 * NFSv4 reexport requires Lustre client on the re-exporting nodes to be
289 mounted with "32bitapi" mount option
291 Severity : enhancement
292 Description: Update RHEL6.1 kernel to 2.6.32-131.6.1.el6
294 Severity : enhancement
295 Description: Update RHEL5.6 kernel to 2.6.18-238.19.1.el5
297 Severity : enhancement
298 Description: Update RHEL5.6 kernel to 2.6.18-238.9.1.el5
301 Frequency : only with SLES10
303 Description: Use OFED "KMP" provided by Novell
304 Details : SLES10 SP3 ships with OFED in a separate "KMP" package.
305 Lustre is now built against this package. That means you need to
306 install the ofed-kmp package from Novell for the patchless client.
307 Note that the ofed-kmp that Novell ships may not exactly match the
308 kernel version but should still be compatible.
310 Severity : enhancement
312 Description: Update RHEL5.5 kernel to 2.6.18-194.17.1.el5
313 Update OEL5.5 kernel to 2.6.18-194.17.1.0.1.el5.
315 --------------------------------------------------------------------------------
317 2010-07-15 Oracle, Inc.
319 * Server support for kernels:
320 2.6.18-164.11.1.el5 (RHEL 5)
321 2.6.18-164.11.1.0.1.el5 (OEL 5)
322 * Client support for unpatched kernels:
323 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
324 2.6.18-164.11.1.el5 (RHEL 5),
325 2.6.18-164.11.1.0.1.el5 (OEL 5)
326 2.6.16.60-0.42.8 (SLES 10),
328 2.6.29.4-167.fc11 (FC11)
329 * Recommended e2fsprogs version: 1.41.10-sun2
331 Severity : enhancement
333 Description: Distribute lfs_migrate script with lustre
334 Details : The lfs_migrate mentioned in the manual is now included in the
335 lustre tree in lustre/scripts and will distributed in lustre RPMs.
337 Severity : enhancement
339 Description: Failover nid list added to import proc file.
341 Severity : enhancement
343 Description: "som_preview" mount option required to use SOM
344 Details : The clients will need to use som_preview mount option
348 Description: Lustre does not do 1MB IOs to HW RAID
349 Details : Bump MAX_PHYS/HW_SEGMENTS and SG_ALL to 256 in the RHEL5 kernel.
351 Severity : enhancement
353 Description: lfs quota output is not very convenient for awk/sed-parsing
354 Details : Some positions in lfs quota output table could be empty or
355 non-empty which made it hard to parse it with scripts, now
356 a dash is put instead of space where there is not supposed
361 Description: Update to OFED 1.5.1
362 Details : OFED 1.5.1 will be used in environments that don't use
363 vendor supplied OFED.
367 Description: Remove lnet. prefix from lctl params display
368 Details : Also renamed lnet.memused to lnet_memused to
373 Description: Use conf_param -d to remove permanant settings.
377 Description: Add list_param -R
378 Details : lctl list_param now supports recursive mode.
381 Bugzilla : 21829, 22312
382 Description: Fix llobdstat.
383 Details : 1. Now updating the write_bytes counter
384 2. Port script fixes from 1.8.
388 Description: decrease the usage of memory on clients.
389 Details : 1. On clients, recycle dentries and inodes unused.
390 2. Delete the code related to ll_deathrow(att 6215 in bug 1443). It
395 Description: llobdstat fix and enhancement
396 Details : add a counter to set a limit to how many samples will be returned
397 fix a wildcard in the path to limit to obdfilter stats only
401 Description: lfs check only allowed for root
402 Details : Code cleanups and added a regresion test of non-root use of
407 Description: MDS Oopses in ptlrpc_dispatch_difficult_reply()/__list_add()
408 Details : Fix hr->hr_index overload.
412 Description: Fix FIEMAP functionality.
416 Description: FC11 patchless client support added
420 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and
421 OEL5.4 kernel to 2.6.18-164.11.1.0.1.el5.
425 Description: New acc-sm script and various test-framework enhancments.
429 Description: Previously landed patch was reverted.
433 Description: Update SLES10 kernel to 2.6.16.60-0.42.8 (patchless)
437 Description: Add ldiskfs maxdirsize mount option.
441 Description: Fix mkfs.lustre support for 16TB LUNs.
445 Description: Print mmp_check_interval
446 Details : print mmp_check_interval and make it possible to abort
447 mount operation in case it takes too long.
451 Description: Add cascading_rw to lustre mpi tests.
455 Description: Store SOM data in EA.
456 Details : Also several SOM related cleanups.
460 Description: Add refcount to md_open_data
464 Description: Use refcount on lov_request_set to prevent memory corruption.
468 Description: Statahead: use parent dir's i_mutex for synchronization.
472 Description: loadgen improvements.
476 Description: Enable peer health detection in ksocklnd and ko2iblnd.
480 Description: Control DCACHE_LUSTRE_INVALID flag with MDS_INODE_LOOKCUP lock
484 Description: Fix race on obd_devcie:md_stats
488 Description: Fix SLES 11 client builds.
489 Details : Resolved uninitalized pointer gcc warnings.
493 Description: Allow non-root access for "lfs check".
494 Details : Added a check in obd_class_ioctl() for OBD_IOC_PING_TARGET.
498 Description: Clear imp_force_reconnect correctly in ptlrpc_connect_interpret()
500 Severity : enhancement
502 Description: Update kernel to RHEL5.4 2.6.18-164.6.1.el5 and
503 OEL5 2.6.18-164.6.1.0.1.el5(Both in-kernel OFED enabled).
507 Description: Sanity test_27p fails.
508 Details : OBD_FAIL_OST_ENOSPC has been moved to filter_statfs().
509 Sanity tests 27 were cleaned up and improved.
513 Description: Unify req format on client/servers
514 Details : Use new req_capsule API [almost] everywhere instead of old PTLRPC
515 buffers and swabbers approach..
520 Description: drop join file stuff.
522 Severity : enhancement
524 Description: Add 2.6.27 and SLES11 (patchless client) support
528 Description: Fix in ptlrpc_expire_one_request() to print the signed time
531 Severity : enhancement
533 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
536 Severity : enhancement
538 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
540 Severity : enhancement
542 Description: correctly handle big reply message.
543 Details : send LNet event if reply is bigger than buffer and adjust this buffer
548 Description: Drop unnecessary __GFP_NOMEMALLOC flag from filter_get_page()
550 Severity : enhancement
552 Description: Add support for OEL5.
556 Description: Conf-sanity.sh 50g test - deactivated OST should not cause a panic.
558 Severity : enhancement
560 Description: "weak-modules" support
561 Details : Implement "weak-modules" support which enables kernel modules
562 to be used with any kernel that implements the same kABI. In
563 order to achieve this modules are now installed in
564 /lib/modules/$(uname -r)/updates/kernel on all distributions.
568 Description: client could not reconnect to OST because of an active request.
569 Details : The client now retries to reconnect to the same server, if a
570 connect request failed with EBUSY or -EAGAIN. Test case 35b
571 conf-sanity.sh has been created.
575 Description: Don't allow make backward step on assignin osc next id.
576 Details : race between allocation next id and ll_sync thread can be cause
577 of set wrong osc next id and can be kill valid ost objects.
581 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
585 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
589 Description: Error handling in osc_statfs_interpret() has been improved.
590 Details : Check in osc_statfs_interpret() for EBADR.
594 Description: Do not send statfs() requests to OSTs disabled by administrator.
595 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
597 Severity : enhancement
599 Description: Added sanity-quota test to for lfs(1) to work with numeric uid/gid.
601 Severity : enhancement
603 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
605 Severity : enhancement
607 Description: Enhancement for lfs(1) command to use numeric uid/gid.
611 Description: Improved cleanup procedure for conf-sanity test_46a.
615 Description: Do not start bulk transfer, if client tried to reconnect.
616 Details : Do not start bulk transfer, if client tried to reconnect,
617 but failed due to active rpcs. Check for exp_abort_active_req
618 has been implemented.
622 Description: Prevent inconsistences between linux and lustre mount structures.
623 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
624 Make the sleep interruptible.
626 Severity : enhancement
628 Description: provide server to client comms path
629 Details : server to client communications path via new LDLM_SET_INFO rpc
633 Description: truncate starts GFP_FS allocation under transaction causing deadlock
634 Details : ldiskfs_truncate calls grab_cache_page which may start page
635 allocation under an open transaction. This may lead to
636 calling prune_icache with consequent lustre reentrance.
640 Description: send correctly lsm on open replay
641 Details : MDS is trust to LSM size on replay open, but client can set wrong size
646 Description: enable client interface failover
647 Details : When a child reconnects from another NID, properly update export
648 nid hash position and ldlm reverse import.
650 Severity : enhancement
652 Description: Communicate OST degraded/readonly state via statfs to MDS
653 Details : Flags in the statfs returned from OSTs indicate whether the
654 OST is in a degraded RAID state, or if the filesystem has
655 turned read-only after a filesystem error is detected.
660 Description: don't panic if EPROTO was hit when reading symlink
661 Details : correctly handling request reference in error cases.
663 Severity : enhancement
665 Description: Add LustreNetLink, a kernel-userspace communcation path.
667 Severity : enhancement
669 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
674 Description: access to llog context before init.
675 Details : move handling CATALOGS file at osc layer and forbid access to llog
679 Frequency : rare, only when client is writing to OST on local node
681 Description: Avoid deadlock for local client writes
682 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
683 memory freeing context. This allows OST threads to set the
684 PF_MEMALLOC flag on task structures in order to allocate memory
685 from reserved pools and complete IO.
686 Use GFP_HIGHUSER for OST allocations for non-local client writes,
687 so that the OST threads generate memory pressure and allow
688 inactive pages to be reclaimed.
690 Severity : enhancement
692 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
697 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
698 Details : move ldlm namespace creation in setup phase to avoid grab
699 _lprocfs_lock with cli_sem held.
703 Description: Temporarily disable grant shrink.
704 Details : Disable the feature for debugging.
707 Frequency : only when formatting test filesystems
709 Description: Unable to run several mkfs.lustre on loop devices at the same time
710 Details : mkfs.lustre returns error 256 on the concurrent loop devices
711 formatting. The solution is to proper handle the error.
713 Severity : enhancement
715 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
717 Severity : enhancement
719 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
724 Description: resolve race between obd_disconnect and class_disconnect_exports
725 Details : if obd_disconnect will be called to already disconnected export he
726 forget release one reference and osc module can't unloaded.
728 Severity : enhancement
730 Description: Allow tuning service thread via /proc
731 Details : For each service a new
732 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
733 created that can be used to set min/max thread counts, and get the
734 current number of running threads.
739 Descriptoin: don't return error if have partially created objects for file.
740 Details : lov_update_create_set uses set->set_success as index for created
741 objects, so if some requests will be failed, they will have hole
742 at end of array and qos_shrink_lsm for allocate correct lsm.
744 Severity : enhancement
746 Description: Update OFED support to 1.4.1
751 Description: Reduce small size read RPC
752 Details : Set read-ahead limit for every file and only do read-ahead when
753 available read-ahead pages are bigger than 1M to avoid small size
756 Severity : enhancement
758 Description: move AT tunable parameters for more consistent usage
759 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
761 Severity : enhancement
763 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
764 Details : allow skip disconnected ost for send statfs request and hide error
770 Description: fix racy locking of mballoc block bitmaps causing BUG
771 Details : The locking of the mballoc buddy bitmap and the in-memory
772 block bitmap was using two different spin locks in some
773 cases. This made it possible to incorrectly access the
774 mballoc bitmap while another process was modifying it,
775 causing a sanity assertion to fail. While no on-disk corruption
776 was reported, there was some risk of this happening.
779 Frequency : rare, on llog test 6
781 Descriptoin: don't allow connect to already connected import
782 Details : allowing connect to already connected import is hide connecting problem.
787 Descriptoin: allow kill process which wait statahead result
788 Details : for some reasons 'ls' can stick in waiting result from statahead,
789 in this case need way for kill this process.
791 Severity : enhancement
793 Description: Add state history info file, enhance import info file
794 Details : Track import connection state changes in a new osc/mdc proc file;
795 add overview-type data to the osc/mdc import proc file.
797 Severity : enhancement
799 Description: MDS create should not wait for statfs RPC while holding DLM lock.
801 Severity : enhancement
803 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
808 Descriptoin: OST grant too much space to client even there are not enough space.
809 Details : Client will shrink its grant cache to OST if there are no write
810 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
811 this grant cache if there are already not enough avaible space
812 (left_space < total_clients * 32M).
815 Frequency : start MDS on uncleanly shutdowned MDS device
817 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
818 Details : stay in waiting mds<>ost recovery finished produce random bugs
819 due race between two ll_sync thread for one lov target. send
820 ACTIVATE event only if connect realy finished and import have
824 Frequency : rare, connect and disconnect target at same time
826 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
827 Details : don't call obd_disconnect under lov_lock. this long time
828 operation and can block ptlrpcd which answer to connect request.
833 Descriptoin: don't lose wakeup for imp_recovery_waitq
834 Details : recover_import_no_retry or invalidate_import and import_close can
835 both sleep on imp_recovery_waitq, but we was send only one wakeup
839 Frequency : always with long access acl
841 Descriptoin: mds can't pack reply with long acl.
842 Details : mds don't control size of acl but they limited by reint/getattr
845 Severity : enhancement
847 Description: Update to SLES10 kernel-2.6.16.60-0.33.
849 Severity : enhancement
851 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
854 Frequency : start MDS on uncleanly shutdowned MDS device
856 Descriptoin: aborting recovery hang on MDS
857 Details : don't throttle destroy RPCs for the MDT.
860 Frequency : on remount
862 Description: external journal device not working after the remount
863 Details : clear dev_rdonly flag for external journal devices in
869 Description: shutdown vs evict race
870 Details : client_disconnect_export vs connect request race.
871 if client will evicted at this time - we start invalidate
872 thread without referece to import and import can be freed
876 Frequency : rare, need acl's on inode.
878 Description: client can't handle ost additional correctly
879 Details : if ost was added after client connected to mds client can have
880 hit lnet_try_match_md ... to big messages to wide striped files.
881 in this case need teach client to handle config events about add
882 lov target and update client max ea size at that event.
884 Severity : enhancement
886 Description: Changelogs
887 Details : Changelogs are a lightweight mechanism to track filesystem
888 metadata and namespace changes. The changelog is recorded
889 permanently on the MDTs, and is periodically "consumed" / purged
890 when records are no longer needed.
892 Severity : enhancement
894 Description: compact fld format with extents
895 Details : Store range of seq rather than every seq in FLD. Seq
896 controller update FLD rather than clients. In Case of CMD, mdt0
897 has FLD, all other metadata server act as non persistent proxy
898 for FLD queries and cache fld entries in fld cache.
903 Description: don't skip ost target if they assigned to file
904 Details : Drop slow OSCs if we can, but not for requested start idx.
905 This means "if OSC is slow and it is not the requested
906 start OST, then it can be skipped, otherwise skip it only
907 if it is inactive/recovering/out-of-space.
911 Description: more cleanup in mds_lov
912 Details : not send LOV EA under replay, we can't know about they size at this
913 time. Don't allow client connect to mds before any ost connected,
914 for avoid problems with LOV EA size and returning EIO to client.
916 Severity : enhancement
918 Description: Interoperability at server side (Disk interoperability)
920 Severity : enhancement
922 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
924 Severity : enhancement
926 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
928 Severity : enhancement
930 Description: New client IO stack (CLIO).
932 Severity : enhancement
934 Description: Commit on sharing. Eliminate inter-client dependencies between
935 uncommitted transactions by doing transaction commits.
936 Thereby clients may recovery independently.
939 Frequency : Create a symlink file with a very long name
941 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
942 Details : If there is no extra space in the request for early cancels,
943 ldlm_req_handles_avail() returns 0 instead of a negative value.
945 Severity : enhancement
947 Description: Add /proc entry for import status
948 Details : The mdc, osc, and mgc import directories now have
949 an import directory that contains useful import data for debugging
952 Severity : enhancement
954 Description: Re-disable certain /proc logging
955 Details : Enable and disable client's offset_stats, extents_stats and
956 extents_stats_per_process stats logging on the fly.
959 Frequency : Only on FC kernels 2.6.22+
961 Description: oops in statahead
962 Details : Do not drop reference count for the dentry from VFS when lookup,
963 VFS will do that by itself.
965 Severity : enhancement
967 Description: Generic /proc file permissions
968 Details : Set /Proc file permissions in a more generic way to enable non-
969 root users operate on some /proc files.
973 Description: Hitting mdc_commit_close() ASSERTION
974 Details : Properly handle request reference release in
975 ll_release_openhandle().
979 Description: quota recovery deadlock during mds failover
980 Details : This patch includes att18982, att18236, att18237 in bz14840.
982 1. fix osts hang when mds does failover with quotaon
983 2. prevent watchdog storm when osts threads wait for the
988 Frequency : only patchless client
989 Description: add workaround for race between add/remove dentry from hash
991 Severity : enhancement
993 Description: Allow OST glimpses to return PW locks
997 Description: LBUG when llog conf file is full
998 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
1002 Description: Prevent import from entering FULL state when server in recovery
1006 Description: service mount cannot take device name with ":"
1007 Details : Only when device name contains ":/" will mount treat it as
1013 Description: replace ptlrpcd with the statahead thread to interpret the async
1014 statahead RPC callback
1018 Frequency : on recovery
1019 Description: I/O failures after umount during fail back
1020 Details : if client reconnected to restarted server we need join to recovery
1021 instead of find server handler is changed and process self
1022 eviction with cancel all locks.
1024 Severity : enhancement
1026 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
1028 Severity : enhancement
1030 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
1032 Severity : enhancement
1034 Description: Upcall on Lustre log has been dumped
1035 Details : Allow for a user mode script to be called once a Lustre log has
1036 been dumped. It passes the filename of the dumped log to the
1037 script, the location of the script can be specified via
1038 /proc/sys/lnet/debug_log_upcall.
1043 Description: avoid idr_remove called for id which is not allocated.
1044 Details : Move assigment s_dev for clustered nfs to end of initialization,
1045 to avoid problem with error handling.
1050 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
1051 Details : When connection is reused this not moved from CONN_UNUSED_HASH
1052 into CONN_USED_HASH and this prodice warning when put connection
1053 again in unused hash.
1056 Severity : enhancement
1058 Description: File striping can now be set to use an arbitrary pool of OSTs.
1060 Severity : enhancement
1062 Description: Export bytes_read/bytes_write count on OSC/OST.
1066 Description: Early reply size mismatch, MGC loses connection
1067 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
1068 the connect flags are properly negotiated.
1073 Description: quotas are not honored with O_DIRECT
1074 Details : all writes with the flag O_DIRECT will use grants which leads to
1075 this problem. Now using OBD_BRW_SYNC to guard this.
1079 Description: add quota statistics
1080 Details : 1. sort out quota proc entries and proc code.
1081 2. add quota statistics
1083 Severity : enhancement
1085 Description: enable quota support for HEAD.
1089 Description: Properly propagate oinfo flags from lov to osc for statfs
1090 Details : restore missing copy oi_flags to lov requests.
1092 Severity : enhancement
1094 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
1095 : l_getgroups(8), lst(8), routerstat(8)
1096 Details : included man pages for llobdstat(8), llstat(8),
1097 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
1099 Severity : enhancement
1101 Description: configure's --enable-quota should check the
1102 : kernel .config for CONFIG_QUOTA
1103 Details : configure is terminated if --enable-quota is passed but
1104 : no quota support is in kernel
1108 Description: Remove portals compatibility
1109 Details : Remove portals compatibility, not interoperable with releases
1114 Description: Resolve device initialization race
1115 Details : Prevent proc handler from accessing devices added to the
1116 obd_devs array but yet be intialized.
1118 Severity : enhancement
1120 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
1122 Severity : enhancement
1124 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
1129 Description: Using wrong pointer in osc_brw_prep_request
1130 Details : Access to array[-1] can produce panic if kernel compiled with
1131 CONFIG_PAGE_ALLOC enabled
1135 Description: Client runs out of low memory
1136 Details : Consider only lowmem when counting initial number of llap pages
1140 Description: *optional* service tags registration
1141 Details : if the "service tags" package is installed on a Lustre node
1142 When the filesystem is mounted, a local-node service tag will
1143 be created. See http://inventory.sun.com/ for more information
1144 about the Service Tags asset management system.
1148 Description: Kernel BUG tries to release flock
1149 Details : Lustre does not destroy flock lock before last reference goes
1150 away. So always drop flock locks when client is evicted and
1151 perform unlock regardless of successfulness of speaking to MDS.
1155 Description: add refcount protection for osc callbacks, avoid panic on shutdown
1159 Description: sanity test 65a fails if stripecount of -1 is set
1160 Details : handle -1 striping on filesystem in ll_dirstripe_verify
1165 Description: ASSERTION(CheckWriteback(page,cmd)) failed
1166 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
1169 Severity : enhancement
1171 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
1175 Description: do not process already freed flock
1176 Details : flock can possibly be freed by another thread before it reaches
1177 to ldlm_flock_completion_ast.
1181 Description: LBUG during stress test
1182 Details : Need properly lock accesses the flock deadlock detection list.
1186 Description: oops in page fault handler
1187 Details : kernel page fault handler can return two special 'pages' in error
1188 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
1192 Description: timeout with invalidate import.
1193 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
1194 should be handled by ptlrpcd. This produce long age waiting and
1195 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
1197 Severity : enhancement
1199 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
1203 Description: enable MGS and MDT services start separately
1204 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
1205 a co-located MGS without starting the MGS, which is a complement
1206 to 'nosvc' mount option.
1210 Frequency : after recovery
1211 Description: precreate to many object's after del orphan.
1212 Details : del orphan st in oscc last_id == next_id and this triger growing
1213 count of precreated objects. Set flag LOW to skip increase count
1214 of precreated objects.
1218 Frequency : rare, on clear nid stats
1219 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
1220 Details : when clean nid stats sometimes try destroy live entry,
1221 and this produce panic in free.
1225 Description: Stack overflow during MDS log replay
1226 ease stack pressure by using a thread dealing llog_process.
1230 Description: wait until IO finished before start new when do lock cancel.
1231 Details : VM protocol want old IO finished before start new, in this case
1232 need wait until PG_writeback is cleared until check dirty flag and
1233 call writepages in lock cancel callback.
1235 Severity : enhancement
1237 Description: using special macro for print time and cleanup in includes.
1241 Description: mds_mfd_close() ASSERTION(rc == 0)
1242 Details : In mds_mfd_close(), we need protect inode's writecount change
1243 within its orphan write semaphore to prevent possible races.
1247 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
1251 Frequency : rare, on shutdown ost
1252 Description: don't hit live lock with umount ost.
1253 Details : shrink_dcache_parent can be in long loop with destroy dentries,
1254 use shrink_dcache_sb instead.
1258 Description: don't panic with use echo client
1259 Details : echo client pass NULL as client nid pointer and this produce null
1260 pointer dereference.
1264 Description: fix build on ppc32
1265 Details : compile code with -m64 flag produce wrong object file for ppc32.
1269 Description: add message levels for liblustreapi
1273 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
1274 Details : Move the 'good_osts' check before the 'total_bavail' check. This
1275 will result in an -EAGAIN and in the exit call path we call
1276 alloc_rr() which will with increasing aggressiveness attempt to
1277 aquire precreated objects on the minimum number of required OSCs.
1281 Description: Use old size assignment to avoid deadlock
1282 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
1283 the scheduling while holding a spinlock. We do not need locking
1284 for size in ll_update_inode() because size is only updated from
1285 the MDS for directories or files without objects, so there is no
1286 other place to do the update, and concurrent access to such inodes
1287 are protected by the inode lock.
1291 Description: resolve "_IOWR redefined" build error on SLES10
1295 Description: dump the memory debugging after all modules are unloaded to
1296 suppress false negative in conf_sanity test 39
1298 Severity : enhancement
1300 Description: build kernel-ib packages for OFED 1.3 in our release cycle
1305 Description: fix SLES kernel versioning
1306 Details : the kernel version for our SLES 10 kernel did not include a "-"
1307 before the "smp" at the end. while this was not a problem in
1308 general, it did mean that software trying to use the kernel
1309 version to try to detect a vendor specific kernel would fail.
1310 this was most evident by the OFED build scripts.
1314 Description: Don't update lov_desc members until making sure they are valid
1315 Details : When updating lov_desc members via proc fs, need fix their
1316 validities before doing the real update.
1320 Description: don't put request into delay list while invalidate in flight.
1321 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
1322 import in flight. this produce timeout for invalidate and sometimes
1323 can cause stale data.
1327 Frequency : on ppc only
1328 Description: not convert ost objects for directory because it's not exist.
1329 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
1333 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
1335 Severity : enhancement
1337 Description: Update to support 2.6.22.14 vanilla kernel.
1341 Frequency : rare, on recovery
1342 Description: read procfs can produce deadlock in some situation
1343 Details : Holding lprocfs lock which send rpc can produce block for destroy
1344 obd objects and this also block reconnect with -EALREADY. This
1345 isn't fix all lprocfs bugs - but make it rare.
1347 Severity : enhancement
1349 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
1352 Frequency : frequent on X2 node
1354 Description: mdc_set_open_replay_data LBUG
1355 Details : Set replay data for requests that are eligible for replay.
1359 Description: lustre_mgs: operation 101 on unconnected MGS
1360 Details : When MGC is disconnected from MGS long enough, MGS will evict the
1361 MGC, and late on MGC cannot successfully connect to MGS and a lot
1362 of the error messages complaining that MGS is not connected.
1366 Frequency : on network error
1367 Description: panic with double free request if network error
1368 Details : mdc_finish_enqueue is finish request if any network error ocuring,
1369 but it's true only for synchronus enqueue, for async enqueue
1370 (via ptlrpcd) this incorrect and ptlrpcd want finish request
1373 Severity : enhancement
1375 Description: client-side metadata stat-ahead during readdir(directory readahead)
1376 Details : perform client-side metadata stat-ahead when the client detects
1377 readdir and sequential stat of dir entries therein
1380 Frequency : on start mds
1382 Description: Implement get_info(last_id) in obdfilter.
1385 Frequency : occasional
1387 Description: Correctly check stale fid, not start epoch if ost not support SOM
1388 Details : open with flag O_CREATE need set old fid in op_fid3 because
1389 op_fid2 was overwritten with new generated fid, but mds can answer
1390 with one of these two fids and both is not stale. Setattr
1391 incorrectly started an epoch and assume will be called
1392 done_writing, but without SOM done_writing ever being called.
1395 Frequency : rare, depends on device drivers and load
1397 Description: MDS or OSS nodes crash due to stack overflow
1398 Details : Code changes in 1.8.0 increased the stack usage of some functions.
1399 In some cases, in conjunction with device drivers that use a lot
1400 of stack the MDS (or possibly OSS) service threads could overflow
1401 the stack. One change which was identified to consume additional
1402 stack has been reworked to avoid the extra stack usage.
1405 Frequency : occasional
1407 Description: Do not fail import if osc_interpret_create gets -EAGAIN
1408 Details : If osc_interpret_create got -EAGAIN it immediately exits and
1409 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
1410 oscc_has_objects and see OSC has no objests and call
1411 oscc_internal_create to resend create request.
1413 Severity : enhancement
1415 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
1417 Severity : enhancement
1419 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
1422 Frequency : very rare
1424 Description: Fix replay if there is an un-replied request and open
1425 Details : In some cases, older replay request will revert the
1426 mcd->mcd_last_xid on MDS which is used to record the client's
1427 latest sent request.
1429 Severity : enhancement
1431 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
1433 Severity : enhancement
1435 Description: Add rhel5 support to HEAD.
1437 Serverity : enhancement
1439 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
1444 Description: Don't allow skipping OSTs if index has been specified.
1445 Details : Don't allow skipping OSTs if index has been specified, make
1446 locking in internal create lots better.
1450 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
1451 Details : access to bitfield in structure is always rounded to long
1452 and this produce problem with not atomic change any bit.
1456 Description: Lustre make rpms failed.
1457 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
1458 builds Lustre rpms from tarball.
1461 Frequency : rare on shutdown ost
1463 Description: If llog cancel was not send before clean_exports phase, this can
1464 produce deadlock in llog code.
1465 Details : If llog thread has last reference to obd and call class_import_put
1466 this produce deadlock because llog_cleanup_commit_master wait when
1467 last llog_commit_thread exited, but this never success because was
1468 called from llog_commit_thread.
1472 Description: allow userland application know is lost one of stripes.
1473 Details : fill lvb_blocks with error code on ost and return it to
1474 application if error flag found.
1478 Description: NULL lov_tgts causing MDS oops
1479 Details : more safe checks for NULL lov_tgts for avoid oops.
1481 Severity : enhancement
1483 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
1487 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
1488 Details : If an OST is down the MDS will hang indefinitely in
1489 obd_statfs_rqset() waiting for the statfs data. While for
1490 MDS QOS usage of statfs, it should not stuck in waiting.
1492 Severity : enhancement
1494 Description: remote_acl support
1495 Details : Support ACL-based permission check for remote user.
1496 Support setfacl/getfacl for remote user with the utils
1497 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
1498 the system "{s,g}etfacl" utils.
1500 Severity : enhancement
1502 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
1504 Severity : enhancement
1506 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
1510 Description: make mgs_setparam() handle fsname containing dash
1511 Details : fsname containing a dash does not work with lctl conf_param
1513 Severity : enhancement
1515 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
1517 Severity : enhancement
1519 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
1523 Frequency : rare, at shutdown
1524 Description: access already free / zero obd_namespace.
1525 Details : if client_disconnect_export was called without force flag set,
1526 and exist connect request in flight, this can produce access to
1527 NULL pointer (or already free pointer) when connect_interpret
1528 store ocd flags in obd_namespace.
1532 Frequency : only at startup
1533 Description: not alloc memory with spinlock held.
1534 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
1535 if any spinlock held.
1537 Severity : enhancement
1539 Description: make lustre randomly fail allocating memory
1540 Details : Make lustre randomly failed allocating memory for testing purpose.
1542 Severity : enhancement
1544 Description: lost problems with lov objid file
1545 Details : Fixes some scability and access to not inited memory problems
1546 in work with lov objdid file.
1551 Description: lfs find does not continue on file error
1552 Details : Continue other files processing when a file/dir is absent.
1556 Description: Inconsistent usage of lustre_pack_reply()
1557 Details : Standardize the usage of lustre_pack_reply() such that it
1558 always generate a CERROR on failure.
1561 Frequency : occasional
1563 Description: MDS hang or stay in waiting lock
1564 Details : If client receive lock with CBPENDING flag ldlm need send lock
1565 cancel as separate rpc, to avoid situation when cancel request
1566 can't processed due all i/o threads stay in wait lock.
1570 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1574 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1578 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1582 Description: Update to RHEL4 latest kernel.
1584 Severity : enhancement
1586 Description: Build SLES10 patchless client fails
1587 Details : The configure was broken by run ./configure with
1588 --with-linux-obj=.... argument for patchless client. When the
1589 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1590 can't search header adequately. Use absolute path such as
1591 -I($LINUX)/include instead.
1595 Description: interrupt oig_wait produce painc on resend.
1596 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1597 private set, and this produce situation when rq_ptlrpcd_data not
1598 copyed to new allocated request and triggered LBUG on assert
1599 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1600 ptlrpcd set if request is changed and can be safety changed to use
1605 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1606 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1607 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1608 these macros and replace them with their difinitions instead.
1612 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1613 Details : Update which_patch & target file for SLES10 latest kernel.
1616 Description: Remove client patches from SLES 10 kernel.
1617 Details : This causes SLES 10 clients to behave as patchless clients
1618 even on a Lustre-patched (server) kernel.
1620 Severity : enhancement
1622 Description: self-adjustable client's lru lists
1623 Details : use adaptive algorithm for managing client cached locks lru
1624 lists according to current server load, other client's work
1625 pattern, memory activities, etc. Both, server and client
1626 side namespaces provide number of proc tunables for controlling
1629 Severity : enhancement
1631 Description: light-weight GSS support
1632 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
1633 and reduce performance overhead.
1635 Severity : enhancement
1637 Description: Linux keyring support
1638 Details : Support using service of Linux keyring for Lustre GSS internal
1639 context refresh/cache mechanism.
1643 Description: Fix errors in lfs documentation
1644 Details : Fixes man pages
1648 Description: don't use GFP_* in generic Lustre code.
1649 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
1653 Description: obdclass is limited by single OBD_ALLOC(idarray)
1654 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
1658 Description: warnings with build patchless client with vanila 2.6.19 and up
1659 Details : change old ctl_table style and replace ctl_table/ctl_table_header
1660 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
1664 Frequency : Only for SLES
1665 Description: sanity_quota fail test_1
1666 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
1667 file, which makes TSTID[2] inunique.
1671 Frequency : Only for SLES9
1672 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
1673 Details : Update target/ChangeLog/which_patch .
1677 Frequency : Only for test_10 in sanity-quota.sh
1678 Description: a bug in quota test code
1679 Details : checking a test flag in a wrong place
1683 Frequency : Only for test_13 in sanity-quota.sh
1684 Description: a bug in quota test script
1685 Details : shouldn't check overfull allocated quota
1689 Description: jbd statistics
1690 Details : Port older jbd statistics patch for sles10
1692 Severity : enhancement
1694 Description: Add printing inode info into message about error in writepage.
1699 Description: reply_lock_interpret crash due to race with it and lock cancel.
1700 Details : Do not replay locks that are being cancelled. Do not reference
1701 locks by their address during replay, just by their handle.
1705 Frequency : When flocks are used.
1706 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
1707 Details : Flock locks might destroy just granted lock if it could be merged
1708 with another existing flock, this is done in completion handler,
1709 so teach ldlm_cli_enquque_fini that this is a valid case for
1715 Description: Oops in read and write path when failing to allocate lock.
1716 Details : Check if lock allocation failed and return error back.
1720 Description: lstripe command fails for valid OST index
1721 Details : The stripe offset is compared to lov->desc.ld_tgt_count
1722 instead of lov->desc.ld_active_tgt_count.
1726 Description: sanity.sh failed test 103
1727 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
1732 Description: df doesn't work properly if diskfs blocksize != 4K
1733 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
1737 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
1738 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
1740 Severity : enhancement
1742 Description: Adaptive timeouts
1743 Details : RPC timeouts adapt to changing server load and network
1744 conditions to reduce resend attempts and improve recovery time.
1748 Description: llapi_file_create() does not allow some changes
1749 Details : add llapi_file_open() that allows specifying the mode and
1750 open flags, and also returns an open file handle.
1754 Description: merge and cleanup kernel patches.
1755 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
1759 Description: Add journal checksum support.(Kernel part)
1760 Details : The journal checksum feature adds two new flags i.e
1761 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
1762 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
1763 indicates that the commit block contains the checksum for
1764 the blocks described by the descriptor blocks. Now commit
1765 record can be sent to disk without waiting for descriptor
1766 blocks to be written to disk. This behavior is controlled
1767 using JBD2_FEATURE_ASYNC_COMMIT flag.
1771 Description: OSS needs mutliple precreate threads
1772 Details : Add ability to start more than one create thread per OSS.
1776 Description: Sanity.sh test_65k failed: setstripe should have succeeded
1777 Details : execute lfs setstripe on client
1781 Description: mds_obd_create error creating tmp object
1782 Details : When the user sets quota on root, llog will be affected and can't
1783 create files and write files.
1786 Frequency : Always on ia64 patchless client, and possibly others.
1788 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
1789 Details : This allows the patchless client to be loaded on architectures
1790 without this export.
1794 Description: RedHat Update kernel for RHEL5
1795 Details : Add the kernel config file for RHEL5.
1799 Description: RedHat Update kernel for RHEL5
1800 Details : Modify the kernel config file more closer RHEL5.
1804 Description: Build failure against Centos5 (RHEL5)
1805 Details : Define PAGE_SIZE when it isn't present.
1809 Description: test_7 on sanity-quota.sh can't be used on separate machines
1810 Details : add facet to handle it
1814 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
1815 Details : with NFS, the anon dentry's parent was set to itself in
1816 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
1817 corresponding dentry other than use rec->ur_name.
1819 Severity : enhancement
1821 Description: lfs setstripe enhancement
1822 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
1826 Description: enable data checksumming by default
1827 Details : enable checksum by default, allow --disable-checksum
1828 configure option and "-o nochecksum" mount option. Checksums
1829 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
1833 Description: lustre support for RHEL5
1834 Details : Add support for RHEL5.
1838 Description: Client eviction due to failover config
1839 Details : after a connection loss, the lustre client should attempt to
1840 reconnect to the last active server first before trying the
1841 other potential connections.
1845 Description: when mds and osts use different quota unit(32bit and 64bit),
1846 quota will be released repeatly.
1847 Details : void sending multiple quota reqs to mds, which will keep the
1848 status between the reqs.
1852 Description: osts not allocated evenly to files
1853 Details : change the condition to increase offset_idx
1856 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
1858 Description: Data corruption for OSTs that are formatted larger than 2TB
1860 Details : When generating the bio request for lustre file writes the
1861 sector number would overflow a temporary variable before being
1862 used for the IO. The data reads correctly from Lustre (which
1863 will overflow in a similar manner) but other file data or
1864 filesystem metadata may be corrupted in some cases.
1868 Description: Tune the kernel for good SCSI performance.
1869 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
1870 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
1875 Description: rewrite ext2-derived code in obdclass/uuid.c
1876 Details : rewrite inherited code (uuid parsing code from ext2 utils)
1877 from scratch preserving functionality.
1881 Description: "lfs find -obd UUID" prints directories
1882 Details : "lfs find -obd UUID" will return all directory names instead
1883 of just file names. It is incorrect because the directories
1884 do not reside on the OSTs.
1888 Description: use i_size_read and i_size_write in 2.6 port
1889 Details : replace inode->i_size access with i_size_read/write()
1892 Frequency : when removing large files
1894 Description: scheduling issue during removal of large Lustre files
1895 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
1896 It causes scheduling issues when removing large files (17TB in the
1900 Frequency : only with liblustre clients on XT3
1902 Description: evictions taking too long
1903 Details : allow llrd to evict clients directly on OSTs
1906 Frequency : only on ppc
1908 Description: /proc/fs/lustre/devices broken on ppc
1909 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
1910 We should make sure the type of 'index' is loff_t and then cast
1911 explicitly as needed below. Do not assign an explicitly cast
1915 Frequency : only for rhel5
1917 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
1918 Details : Modify the target file & which_kernel.
1920 Severity : enhancement
1922 Description: omit set fsid for export NFS
1923 Details : fix set/restore device id for avoid EMFILE error and mark lustre
1924 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
1928 Frequency : Always, for kernels after 2.6.16
1929 Description: Fix warning idr_remove called for id=.. which is not allocated.
1930 Details : Last kernels save old s_dev before kill super and not allow
1931 to restore from callback - restore it before call kill_anon_super.
1935 Description: buffer overruns could theoretically occur
1936 Details : llapi_semantic_traverse() modifies the "path" argument by
1937 appending values to the end of the origin string, and a
1938 overrun may occur. Adding buffer overrun check in liblustreapi.
1942 Description: Fix error on 'ls .' at the top of the Lustre mount.
1943 Details : Don't revalidate dentry if it is a root dentry.
1947 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
1948 Details : Modify vm-tunables-rhel4.patch.
1952 Description: Kernel config for 2.6.18-vanilla.
1953 Details : Modify targets/2.6-vanilla.target.in.
1954 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
1955 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
1956 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
1957 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
1961 Description: improve handling recoverable errors
1962 Details : if request processig with error which can be recoverable on server
1963 request should be resend, otherwise page released from cache and
1968 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
1969 Details : Modify target file & which_patch.
1970 A flaw was found in the IA32 system call emulation provided
1971 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
1972 value could be stored in the %RAX register, which could trigger an
1973 out-of-bounds system call table access. An untrusted local user
1974 could exploit this flaw to run code in the kernel
1975 (ie a root privilege escalation). (CVE-2007-4573).
1979 Description: O_DIRECT bypasses client statistics.
1980 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
1981 read_ahead_stats were updated. copied the stats section from
1982 osc_send_oap_rpc() into async_internal().
1986 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
1990 Description: change order of libsysio includes
1991 Details : '#include sysio.h' should always come before '#include xtio.h'
1995 Description: handle "serious error: objid * already exists" more gracefully
1996 Details : If LAST_ID value on disk is smaller than the objects existing in
1997 the O/0/d* directories, it indicates disk corruption and causes an
1998 LBUG(). If the object is 0-length, then we should use the existing
1999 object. This will help to avoid a full fsck in most cases.
2001 Severity : enhancement
2003 Description: adapt the lustre_config script to support the upgrade case
2004 Details : Add "-u" option for lustre_config script to support upgrading 1.4
2005 server targets to 1.6 in parallel.
2009 Description: To avoid grant space > avaible space when the disk is almost
2010 full. Without this patch you might see the error "grant XXXX >
2011 available" or some LBUG about grant, when the disk is almost
2013 Details : In filter_check_grant, for non_grant cache write, we should
2014 check the left space by if (*left > ungranted + bytes), instead
2015 of (*left > ungranted), because only we are sure the left space
2016 is enough for another "bytes", then the ungrant space should be
2017 increase. In client, we should update cl_avail_grant only there
2018 is OBD_MD_FLGRANT in the reply.
2022 Description: Update RHEL 4 kernel to fix local root privilege escalation.
2023 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
2024 described in CVE-2007-4573. This problem could allow untrusted
2025 local users to gain root access.
2028 Frequency : when using O_DIRECT and quotas
2030 Description: Incorrect file ownership on O_DIRECT output files
2031 Details : block usage reported by 'lfs quota' does not take into account
2032 files that have been written with O_DIRECT.
2037 Description: touch file failed when fs is not full
2038 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
2039 otherwise we can get ENOSP while fs is not full.
2043 Description: parallel lock callbacks
2044 Details : Instead of sending blocking and completion callbacks as separated
2045 requests, adding them to a set and sending in parallel.
2048 Frequency : only for Cray XT3
2049 Bugzilla : 12829/13455
2050 Description: Changing primary group doesn't change the group lustre assigns to
2052 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
2053 overridden with the primary group provided by the group upcall,
2054 whereas the supplied fsgid can be trusted if it is in the list of
2055 supplementary groups returned by the group upcall.
2057 Severity : enhancement
2059 Description: Allow masking D_WARNING, D_ERROR messages from console
2060 Details : Console messages can now be disabled via lnet.printk.
2064 Description: User code with malformed file open parameter crashes client node
2065 Details : Before packing join_file req, all the related reference should be
2066 checked carefully in case some malformed flags cause fake
2067 join_file req on client.
2071 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
2072 Details : ldlm_completion_ast() assumes that a lock is granted when the req
2073 mode is equal to the granted mode. However, it should also check
2074 that LDLM_FL_CP_REQD is not set.
2078 Description: Heavy nfs access might result in deadlocks
2079 Details : After ELC code landed, it is now improper to enqueue any mds
2080 locks under och_sem, because enqueue might want to decide to
2081 cancel open locks for same inode we are holding och_sem for.
2085 Description: Client eviction while running blogbench
2086 Details : A lot of unlink operations with concurrent I/O can lead to a
2087 deadlock causing evictions. To address the problem, the number of
2088 oustanding OST_DESTROY requests is now throttled to
2089 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
2090 callbacks are processed in priority.
2094 Description: enable ACLs on MDS by default
2095 Details : ACLs must be enabled on MDS by default.
2098 Frequency : PPC/PPC64 only
2100 Description: conflicts between asm-ppc64/types.h and lustre_types.h
2101 Details : fix duplicated definitions between asm-ppc64/types.h and
2102 lustre_types.h on PPC.
2105 Frequency : PPC/PPC64 only
2107 Description: asm-ppc/segment.h does not exist
2108 Details : fix compile issue on PPC.
2112 Description: better handle error messages in extents code
2115 Frequency : RHEL4 only
2117 Description: mkfs is very slow on IA64/RHEL4
2118 Details : A performance regression has been discovered in the MPT Fusion
2119 driver between versions 3.02.73rh and 3.02.99.00rh. As a
2120 consequence, we have downgraded the MPT Fusion driver in the RHEL4
2121 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
2123 Severity : enhancement
2125 Description: SNMP support enhancement
2126 Details : Adding total number of sampled request for an MDS node in snmp
2129 Severity : enhancement
2131 Description: Optimize ldlm waiting list processing for PR extent locks
2132 Details : When processing waiting list for read extent lock and meeting read
2133 lock that is same or wider to it that is not contended, skip
2134 processing rest of the list and immediatelly return current
2135 status of conflictness, since we are guaranteed there are no
2136 conflicting locks in the rest of the list.
2140 Description: Time out and refuse to reconnect
2141 Details : When the failover node is the primary node, it is possible
2142 to have two identical connections in imp_conn_list. We must
2143 compare not conn's pointers but NIDs, otherwise we can defeat
2144 connection throttling.
2148 Description: port llog fixes from b1_6 into HEAD
2149 Details : Port llog reference couting and some llog cleanups from b1_6
2150 (bug 10800) into HEAD, for protect from panic and access to already
2151 free llog structures.
2155 Description: Detect stride IO mode in read-ahead
2156 Details : When a client does stride read, read-ahead should detect that and
2157 read-ahead pages according to the detected stride pattern.
2161 Description: data checksumming impacts single node performance
2162 Details : add support for several checksum algorithm. Currently, only CRC32
2163 and Adler-32 are supported. The checksum type can be changed on
2164 the fly via /proc/fs/lustre/osc/*/checksum_type.
2168 Description: use adler32 for page checksums
2169 Details : when available, use the Adler-32 algorithm instead of CRC32 for
2174 Description: build for x2 fails
2175 Details : fix compile issue on Cray systems.
2179 Description: Properly match for duplicate locks
2180 Details : Due to different lock order from skiplists code, we need to
2181 traverse entire list for now
2184 Frequency : only on PPC/SLES10
2186 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
2187 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
2188 be defined. Add a hack in mkfs_lustre.c to work around this compile
2193 Description: LASSERT on MDS when client holding flock lock dies
2194 Details : ldlm pool logic depends on number of granted locks equal to
2195 number of released locks which is not true for flock locks, so
2196 just exclude such locks from consideration.
2200 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
2201 Details : Use fsfilt_sync() for both the whole filesystem sync and
2202 individual file sync to eliminate dangerous inode locking
2203 with I_LOCK that can lead to a deadlock.
2207 Description: performance in 1.6.3
2208 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
2209 sure that 1MB requests are not fragmented by the block layer.
2211 Severity : enhancement
2213 Description: organize the server-side client stats on per-nid basis
2214 Details : Change the structure of stats under obdfilter and mds to
2222 The "uuid"s file would list the uuids of _active_ exports.
2223 And the clear entry is to clear all stats and stale nids.
2225 Severity : enhancement
2227 Description: eliminate client locks in face of contention
2228 Details : file contention detection and lockless i/o implementation
2229 for contended files.
2233 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
2236 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
2238 Description: MDS may lose file striping (and hence file data) in some cases
2239 Details : If there are additional extended attributes stored on the MDS,
2240 in particular ACLs, SELinux, or user attributes (if user_xattr
2241 is specified for the client mount options) then there is a risk
2242 of attribute loss. Additionally, the Lustre file striping
2243 needs to be larger than default (e.g. striped over all OSTs),
2244 and an additional attribute must be stored initially in the
2245 inode and then increase in size enough to be moved to the
2246 external attribute block (e.g. ACL growing in size) for file
2251 Description: skiplist implementation simplification
2252 Details : skiplists are used to group compatible locks on granted list
2253 that was implemented as tracking first and last lock of each
2254 lock group the patch changes that to using doubly linked lists
2258 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
2259 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
2260 because LASSERT on both the data supplied by a client, and the
2261 data on disk is dangerous and incorrect.
2263 Severity : enhancement
2265 Description: Slow truncate/writes to huge files at high offsets.
2266 Details : Directly associate cached pages to lock that protect those pages,
2267 this allows us to quickly find what pages to write and remove
2268 once lock callback is received.
2272 Description: more ldlm soft lockups
2273 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
2274 starve other threads from the resource lock for a long time in
2275 case of long waiting queue, so change the debug level from
2276 D_OTHER to the less frequently used D_INFO.
2278 Severity : enhancement
2280 Description: add -gid, -group, -uid, -user options to lfs find
2284 Description: Hung threads in invalidate_inode_pages2_range
2285 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
2286 once one is completed. As a result, some RPCs are stuck in the
2287 queue and are never sent.
2291 Description: filter threads hungs on waiting journal commit
2292 Details : Cleanup filter group llog code, then only filter group llog will
2293 be only created in the MDS/OST syncing process.
2297 Description: Procfs and llog threads access destoryed import sometimes.
2298 Details : Sync the import destoryed process with procfs and llog threads by
2299 the import refcount and semaphore.
2301 Severity : enhancement
2303 Description: openlock cache of b1_6 port to HEAD
2308 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
2309 Details : If insertion of an extent fails, then discard the inode
2310 preallocation and free data blocks else it can lead to duplicate
2315 Description: don't always update ctime in ext3_xattr_set_handle()
2316 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
2317 In some cases the ctime should not be updated, for example for
2318 2.0->1.8 compatibility it is necessary to delete an xattr and it
2319 should not update the ctime.
2323 Bugzilla : 15713/16362
2324 Description: Assertion in iopen_connect_dentry in 1.6.3
2325 Details : looking up an inode via iopen with the wrong generation number can
2326 populate the dcache with a disconneced dentry while the inode
2327 number is in the process of being reallocated. This causes an
2328 assertion failure in iopen since the inode's dentry list contains
2329 both a connected and disconnected dentry.
2333 Description: assertion failure in ldlm_handle2lock()
2334 Details : fix a race between class_handle_unhash() and class_handle2object()
2335 introduced in lustre 1.6.5 by bug 13622.
2340 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
2341 Details : remove the limit of 256 scsi disks in the sd_iostat patch
2346 Description: oops in sd_iostats_seq_show()
2347 Details : unloading/reloading the scsi low level driver triggers a kernel
2348 bug when trying to access the sd iostat file.
2353 Description: Kernel panics during QLogic driver reload
2354 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
2355 patch, causing memory corruption.
2360 Description: journal_dev option does not work in b1_6
2361 Details : pass mount option during pre-mount.
2363 Severity : enhancement
2365 Description: Add a FIEMAP(FIle Extent MAP) ioctl
2366 Details : FIEMAP ioctl will allow an application to efficiently fetch the
2367 extent information of a file. It can be used to map logical blocks
2368 in a file to physical blocks in the block device.
2372 Description: LDLM soft lockups - improvement
2373 Details : It is be possible to send the lock handle along with each read
2374 or write request because the client is already doing a lock match
2375 itself so there isn't any reason the OST should have to re-do that
2379 Frequency : only on Cray X2
2381 Description: X2 build failures
2382 Details : fix build failures on Cray X2.
2386 Description: xid & resent requests
2387 Details : Initialize RPC XID from clock at startup (randomly if clock is
2390 Severity : enhancement
2392 Description: Add lustre_start utility to start or stop multiple Lustre servers
2397 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
2398 Details : In case of memory pressure, list_del() can be called twice on
2399 req->rq_history_list, causing a kernel oops.
2403 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
2404 Details : fix stack overflow in the distributed lock manager by defering
2405 export eviction after a failed AST to the elt thread instead of
2406 handling it in the dlm interpret routine.
2410 Description: Convert some comments to new format.
2411 Details : Update documenting comments to match doxygen conventions.
2415 Description: Grammar fixes.
2416 Details : A couple of trivial spelling fixes.
2420 Description: OSD_COUNTERS-mandatory
2421 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
2422 to simplify the code.
2426 Description: simplify cmm_device freeing logic.
2427 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
2431 Description: Add lockdep support to dt_object_operations locking interface.
2432 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
2433 parameter indicating lock ordering. Update mdd code to use new
2438 Description: Introduce failloc constants for lockless IO tests.
2439 Details : Add two new failloc constants to test lockless IO. Only one of
2440 them in implemented---another is checked in yet to be landed
2445 Description: Add lockdep support for inode mutex.
2446 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
2447 in the situations where Lustre has to lock more than one inode
2452 Description: Add optional invariants checking support.
2453 Details : Add new LINVRNT() macro, optional on new --enable-invariants
2454 configure switch. This macro is to be used for consistency and
2455 sanity checks that are too expensive to be left in `production'
2460 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
2461 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
2466 Description: Add lockdep support for ldlm_lock and ldlm_resource.
2467 Details : Use spin_lock_nested() in (the only) situation where more than
2468 one ldlm_lock is locked simultaneously. Also, fix possible
2469 dead-lock in ldlm_lock_change_resource() by enforcing particular
2474 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
2475 Details : Instead of specifying each ldlm_lock call-back through separate
2476 parameter, wrap them into struct ldlm_callback_suite.
2480 Description: Kill join_lru obd method and its callers.
2481 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
2482 in memory---a requirement implemented through ->o_join_lru() obd
2483 method in HEAD. Remove this method and its users.
2487 Description: Add asynchronous ldlm ENQUEUE completion handler.
2488 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
2489 case that stock ldlm_completion_ast() cannot handle as it waits
2490 until lock is granted. Introduce new ldlm_completion_ast_async()
2491 for this. Also comment ldlm_completion_ast().
2495 Description: ldlm_error <-> errno conversion.
2496 Details : Add functions to map (rather arbitrary) between LDLM error codes
2497 and standard errno values. CLIO needs this to prevent LDLM specific
2498 constants from escaping ldlm and osc.
2502 Description: Kill unused ldlm_handle2lock_ns() function.
2503 Details : Kill unused ldlm_handle2lock_ns() function.
2507 Description: Add lu_ref support to ldlm_lock
2508 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
2509 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
2510 and ldlm_resource. LDLM interface has to be changed, because code that
2511 releases a reference on a lock, has to "know" what reference this is.
2512 In the most frequent case
2514 lock = ldlm_handle2lock(handle);
2516 LDLM_LOCK_PUT(lock);
2518 no changes are required. When any other reference (received _not_ from
2519 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
2520 instead of LDLM_LOCK_PUT().
2522 Arguably, changes are pervasive, and interface requires some discipline
2523 for proper use. On the other hand, it was very instrumental in finding
2524 a few leaked lock references.
2528 Description: Add ldlm_lock_addref_try().
2529 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
2530 attempts to addref a lock that might be being canceled
2535 Description: Add ldlm_weigh_callback().
2536 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
2537 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
2538 hard-coded `number of pages' logic.
2542 Description: Add lockdep annotations to llog code.
2543 Details : Use appropriately tagged _nested() locking calls in the places
2544 where llog takes more than one ->lgh_lock lock.
2548 Description: Add loi_kms_set().
2549 Details : Wrap kms updates into a helper function.
2553 Description: Constify instances of struct lsm_operations.
2554 Details : Constify instances of struct lsm_operations.
2558 Description: lu_conf support.
2559 Details : On a server, a file system object is uniquely identified
2560 by a fid, which is sufficient to locate and load all object
2561 state (inode). On a client, on the other hand, more data are
2562 necessary instantiate an object. Change lu_object_find() and
2563 friends to take additional `lu_conf' argument describing object.
2564 Typically this includes layout information.
2568 Description: lu_context fixes.
2569 Details : Introduce new lu_context functions that are needed on the client
2570 side, where some system threads (ptlrpcd) are shared by multiple
2571 modules, and so cannot be stopped during module shutdown.
2575 Description: Add start and stop methods to lu_device_type_operations.
2576 Details : Introduce two new methods in lu_device_type_operations, that are
2577 invoked when first instance of a given type is created and last one
2578 is destroyed respectively. This is need by CLIO.
2582 Description: Add lu_ref support to struct lu_device.
2583 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
2584 track leaked references.
2588 Description: Introduce lu_kmem_descr.
2589 Details : lu_kmem_descr and its companion interface allow to create
2590 and destroy a number of kmem caches at once.
2594 Description: Fix lu_object finalization race.
2595 Details : Fix a race between lu_object_find() finding an object and its
2596 concurrent finalization. This race is (most likely) not possible
2597 on the server, but might happen on the client.
2601 Description: Introduce lu_ref interface.
2602 Details : lu_ref is a debugging module allowing to track references to
2603 a given object. It is quite cpu expensive, and has to be
2604 explicitly enabled with --enable-lu_ref. See usage description
2609 Description: Factor lu_site procfs stats into a separate function.
2610 Details : Separate lu_site stats printing code into a separate function
2611 that can be reused on a client.
2615 Description: Constify instances of struct {lu,dt,md}_device_operations.
2616 Details : Constify instances of struct {lu,dt,md}_device_operations.
2620 Description: Introduce struct md_site and move meta-data specific parts of
2621 struct lu_site here.
2622 Details : Move md-specific fields out of struct lu_site into special struct
2623 md_site, so that lu_site can be used on a client.
2627 Description: Kill mdd_lov_destroy().
2628 Details : Remove unused mdd code.
2632 Description: Add st_block checking to multistat.c.
2633 Details : Add st_block checking to multistat.c.
2637 Description: Add lu_ref support to struct obd_device.
2638 Details : Add lu_ref tracking to obd_device.
2642 Description: Kill obd_set_fail_loc().
2643 Details : Remove unused code.
2647 Description: Add special type for ptlrpc_request interpret functions.
2648 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
2649 there. Actual usage will be in CLIO.
2653 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
2654 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
2655 doesn't work with lockdep.
2659 Description: Add rwv.c test program.
2660 Details : New testing program exercising readv(2) and writev(2) (Qian).
2664 Description: Add sendfile.c test program.
2665 Details : New testing program exercising sendfile(2) (Jay).
2669 Description: Ratelimit a message that can be very frequent.
2670 Details : Ratelimit a memory allocation failure message that can
2675 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2676 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2680 Description: Liblustre build fixes.
2681 Details : Liblustre build fixes.
2685 Description: libcfs: add cfs_{need,cond}_resched() interface.
2686 Details : libcfs: add cfs_{need,cond}_resched() definition and
2687 implementations for Linux, NT, and liblustre.
2689 Severity : enhancement
2691 Description: More exported tunables for mballoc
2692 Details : Add support for tunable preallocation window and new tunables for
2693 large/small requests
2697 Description: Detect corruption of block bitmap and checking for preallocations
2698 Details : Checks validity of on-disk block bitmap. Also it does better
2699 checking of number of applied preallocations. When corruption is
2700 found, it turns filesystem readonly to prevent further corruptions.
2704 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
2705 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
2706 window to 0. If it does hit cache miss, and read-pattern is still
2707 stride-io mode, does not reset the stride window, but also does not
2708 increase the stride window length in this case.
2712 Frequency : only for big-endian servers
2713 Description: Check if system is big-endian while mounting fs with extents feature
2714 Details : Mounting a filesystem with extents feature will fail on big-endian
2715 systems since ext3-based ldiskfs is not supported on big-endian
2716 systems. This can be over-riden with "bigendian_extents" mount option.
2718 Severity : enhancement
2720 Description: The root squash functionality
2721 Details : A security feature, which is to prevent users from being able
2722 to mount lustre on their desktop, run as root, and delete
2723 all of the files in the filesystem. The goal is accomplished by
2724 remapping user id (UID) and group id (GID) of the root user to
2725 a UID and GID specified by the system administartor via Lustre
2726 configuration management server (MGS). The functionality also
2727 allows to specify sets of clients for which the remapping does
2732 Description: Excessive recovery window
2733 Details : With AT enabled, the recovery window can be excessively long (6000+
2734 seconds). To address this problem, we no longer use
2735 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
2736 timeout no longer depends on the service time, it is set to
2737 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
2738 time via pb_service_time.
2742 Description: Watchdog triggered on MDS failover
2743 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
2744 the OSTs know that the MDS "UUID" can be reused for the same export
2745 from a different NID, so we do not need to wait for the export to be
2749 Frequency : rare, only if using MMP with Linux RAID
2751 Description: MMP doesn't work with Linux RAID
2752 Details : While using HA for Lustre servers with Linux RAID, it is possible
2753 that MMP will not detect multiple mounts. To make this work we
2754 need to unplug the device queue in RAID when the MMP block is being
2755 written. Also while reading the MMP block, we should read it from
2756 disk and not the cached one.
2758 Severity : enhancement
2760 Description: open file using fid
2761 Details : A file can be opened using just its fid, like
2762 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
2765 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
2768 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
2769 Details : add proper locking for creation of jbd_4k slab cache
2773 Description: MMP check in ext3_remount() fails without displaying any error
2774 Details : When multiple mount protection fails during remount, proper error
2777 Severity : enhancement
2779 Description: Allow stripe size to be up to 4G-64k
2780 Details : Fix math logic to allow large stripe sizes.
2784 Description: add check for >8TB ldiskfs filesystems
2785 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
2786 Don't allow >8TB ldiskfs filesystems to be mounted without
2787 force_over_8tb mount option
2792 Descriptoin: OSS complains "too many blocks" and crashes
2793 Details : In filter_commitrw_write():
2794 1. lquota_chkquota();
2795 2. obdo_from_inode();
2796 3. lquota_pending_commit();
2797 "2." has changed uid and gid in this case, then caused
2798 lquota_chkquota() and lquota_pending_commit() to use
2799 different uid and gid. That is the root of the bug.
2803 Descriptoin: getstripe doesn't report stripe count = -1 sanely
2804 Details : Detect and print out the 'stripe count = -1' case properly.
2808 Description: Out or order replies might be lost on replay
2809 Details : In ptlrpc_retain_replayable_request if we cannot find retained
2810 request with tid smaller than one currently being added, add it
2811 to the start, not end of the list.
2813 --------------------------------------------------------------------------------
2815 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
2817 * Support for kernels:
2818 2.6.5-7.283 (SLES 9),
2819 2.6.9-55.EL (RHEL 4),
2820 2.6.16.46-0.14 (SLES 10),
2821 2.6.18.8 vanilla (kernel.org)
2822 * Client support for unpatched kernels:
2823 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2824 2.6.16 - 2.6.22 vanilla (kernel.org)
2825 * Due to recently discovered recovery problems, we do not recommend
2826 using patchless RHEL 4 clients with this or any earlier release.
2827 * Recommended e2fsprogs version: 1.39.cfs8
2828 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2829 * Starting with this release, the ldiskfs backing filesystem required
2830 by Lustre is now in its own package, lustre-ldiskfs. This package
2831 should be installed. It is versioned separately from Lustre and
2832 may be released separately in future.
2836 Description: block reactivating mgc import until all deactivates complete
2837 Details : Fix race when failing back MDT/MGS to itself (testing)
2840 Frequency : at statup only
2842 Description: mds_lov_synchronize race leads to various problems
2843 Details : simultaneous MDT->OST connections at startup can cause the
2844 sync to abort, leaving the OSC in a bad state.
2846 Severity : enhancement
2848 Description: add optional extra BUILD_VERSION info
2849 Details : add a new environment variable (namely LUSTRE_VERS) which allows
2850 to override the lustre version.
2853 Frequency : 2.6.18 servers only
2855 Description: ll_kern_mount() doesn't release the module reference
2856 Details : The ldiskfs module reference count never drops down to 0
2857 because ll_kern_mount() doesn't release the module reference.
2862 Description: server LBUG when using old ost_num_threads parameter
2863 Details : Accept the old ost_num_threads parameter but warn that it
2864 is deprecated, and fix an off-by-one error that caused an LBUG.
2869 Description: Transient SCSI error results in persistent IO issue
2870 Details : iobuf->dr_error is not reinitialized to 0 between two
2874 Frequency : sometimes when underlying device returns I/O errors
2876 Description: OSTs not going read-only during write failures
2877 Details : OSTs are not remounted read-only when the journal commit threads
2878 get I/O errors because fsfilt_ext3 calls journal_start/stop()
2879 instead of the ext3 wrappers.
2883 Description: poor connect scaling with increasing client count
2884 Details : Don't run filter_grant_sanity_check for more than 100 exports
2885 to improve scaling for large numbers of clients.
2888 Frequency : SLES10 only
2890 Description: sanity-quota.sh quotacheck failed: rc = -22
2891 Details : Quotas cannot be enabled on SLES10.
2894 Frequency : liblustre clients only
2896 Description: getdirentries does not give error when run on compute nodes
2897 Details : getdirentries does not fail when the size specified as an argument
2898 is too small to contain at least one entry
2900 Severity : enhancement
2902 Description: Add LNET router traceability for debug purposes
2903 Details : If a checksum failure occurs with a router as part of the
2904 IO path, the NID of the last router that forwarded the bulk data
2905 is printed so it can be identified.
2910 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
2911 Details : Due to a race condition, liblustre clients were occasionally
2912 evicted incorrectly.
2914 Severity : enhancement
2916 Description: lfs setstripe use optional parameters instead of postional
2919 Severity : enhancement
2921 Description: Nanosecond timestamp support for ldiskfs
2922 Details : The on-disk ldiskfs filesystem has added support for nanosecond
2923 resolution timestamps. There is not yet support for this at
2924 the Lustre filesystem level.
2927 Frequency : during server recovery
2929 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
2930 Details : request with rq_no_resend flag not awake l_wait_event if they get
2934 Frequency : nfs export on patchless client
2936 Description: connectathon hang when test nfs export over patchless client
2937 Details : Disconnected dentry cannot be found with lookup, so we do not need
2938 to unhash it or make it invalid
2941 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
2942 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
2943 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
2948 Description: open req refcounting wrong on reconnect
2949 Details : If reconnect happened between getting open reply from server and
2950 call to mdc_set_replay_data in ll_file_open, we will schedule
2951 replay for unreferenced request that we are about to free.
2952 Subsequent close will crash in variety of ways.
2953 Check that request is still eligible for replay in
2954 mdc_set_replay_data().
2959 Description: disable writes to filesystem when reading health_check file
2960 Details : the default for reading the health_check proc file has changed
2961 to NOT do a journal transaction and write to disk, because this
2962 can cause reads of the /proc file to hang and block HA state
2963 checking on a healthy but otherwise heavily loaded system. It
2964 is possible to return to the previous behaviour during configure
2965 with --enable-health-write.
2967 Severity : enhancement
2969 Description: 64-bit inode version
2970 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
2971 to the inode. This will be required for version-based recovery.
2976 Description: MDS fails to start if a duplicate client export is detected
2977 Details : in some rare cases it was possible for a client to connect to
2978 an MDS multiple times. Upon recovery the MDS would detect this
2979 and fail during startup. Handle this more gracefully.
2981 Severity : enhancement
2983 Description: Add -o localflock option to simulate old noflock
2985 Details : This will achieve local-only flock/fcntl locks
2991 Description: log_commit_thread vs filter_destroy race leads to crash
2992 Details : Take import reference before releasing llog record semaphore
2997 Description: Wrong request locking in request set processing
2998 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
2999 imp_delayed_list, in this place should be used imp_lock.
3002 Frequency : when reconnection
3004 Description: Grant Leak when osc reconnect to OST
3005 Details : When osc reconnect ost, OST(filter) should check whether it
3006 should grant more space to client by comparing fed_grant and
3007 cl_avail_grant, and return the granted space to client instead
3008 of "new granted" space, because client will call osc_init_grant
3009 to update the client grant space info.
3012 Frequency : when client reconnect to OST
3014 Description: Grant Leak when osc do resend and replay bulk write
3015 Details : When osc reconnect to OST, OST(filter)should clear grant info of
3016 bulk write request, because the grant info while be sync between
3017 OSC and OST when reconnect, and we should ignore the grant info
3018 these of resend/replay write req.
3023 Description: Grant space more than available left space sometimes.
3024 Details : When then OST is about to be full, if two bulk writing from
3025 different clients came to OST. Accord the avaliable space of the
3026 OST, the first req should be permitted, and the second one
3027 should be denied by ENOSPC. But if the seconde arrived before
3028 the first one is committed. The OST might wrongly permit second
3029 writing, which will cause grant space > available space.
3032 Frequency : when client is evicted
3034 Description: Grant might be wrongly erased when osc is evicted by OST
3035 Details : when the import is evicted by server, it will fork another
3036 thread ptlrpc_invalidate_import_thread to invalidate the
3037 import, where the grant will be set to 0. While the original
3038 thread will update the grant it got when connecting. So if
3039 the former happened latter, the grant will be wrongly errased
3040 because of this race.
3045 Description: Checking Stale with correct fid
3046 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
3047 to check whether it is stale, because sometimes, we want the
3048 enqueue happened anyway, and op_data.fid2 will not be initialized.
3050 Severity : enhancement
3052 Description: update patchless client
3053 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
3056 Frequency : only with 2.4 kernel
3058 Description: random memory corruption
3059 Details : size of struct ll_inode_info is to big for union inode.u and this
3060 can be cause of random memory corruption.
3065 Description: Memory leak in recovery
3066 Details : Lov_mds_md was not free in an error handler in mds_create_object.
3067 It should also check obd_fail before fsfilt_start, otherwise if
3068 fsfilt_start return -EROFS,(failover mds during mds recovery).
3069 then the req will return with repmsg->transno = 0 and rc = EROFS.
3070 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
3071 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
3072 be freed no matter whether fsfilt_commit success or not.
3075 Frequency : only with huge count clients
3077 Description: Prevents from taking the superblock lock in llap_from_page for
3079 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
3080 for llap_from_page call in ll_removepage prevents from taking the
3081 superblock lock for a soon died page.
3086 Description: Not check open intent error before release open handle
3087 Details : in some rare cases, the open intent error is not checked before
3088 release open handle, which may cause
3089 ASSERTION(open_req->rq_transno != 0), because it tries to release
3090 the failed open handle.
3095 Description: Set cat log bitmap only after create log success.
3096 Details : in some rare cases, the cat log bitmap is set too early. and it
3097 should be set only after create log success.
3101 Description: Accessing a block bevice can re-enable I/O when Lustre is
3102 tearing down a device.
3103 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
3108 Description: service threads may hog cpus when there are a lot of requests
3109 Details : Insert cond_resched to give other threads a chance to use some CPU
3114 Description: the cat log was not initialized in recovery
3115 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
3116 unlink log on mds will not be initialized until mds post
3117 recovery. And also in mds post recovery, the unlink log will
3118 initialization will be done asynchronausly, so there will be race
3119 between add unlink log and unlink log initialization.
3123 Description: brw_stats were being printed incorrectly
3124 Details : brw_stats were being printed as log2 but all of them were not
3125 recorded as log2. Also remove some code duplication arising from
3126 filter_tally_{read,write}.
3130 Frequency : rare, only in recovery.
3131 Description: ASSERTION(req->rq_type != LI_POISON) failed
3132 Details : imp_lock should be held while iterating over imp_sending_list for
3133 prevent destroy request after get timeout in ptlrpc_queue_wait.
3137 Description: replay-single.sh test 52 fails
3138 Details : A lock's skiplist need to be cleanup when it being unlinked
3139 from its resource list.
3143 Description: Short directio read returns full requested size rather than
3145 Details : Direct I/O operations should return actual amount of bytes
3146 transferred rather than requested size.
3148 Severity : enhancement
3150 Description: metadata RPC reduction (e.g. for rm performance)
3151 Details : decrease the amount of synchronous RPC between clients and servers
3152 by canceling conflicing lock before the operation on the client
3153 and packing thier handles into the main operation RPC to server.
3155 Severity : enhancement
3157 Description: Async OSC create to avoid the blocking unnecessarily.
3158 Details : If an OST has no remaining object, system will block on the
3159 creation when it needs to create a new object on this OST. Now,
3160 ways use pre-created objects when available, instead of blocking on
3161 an empty osc while others are not empty. If we must block, we
3162 block for the shortest possible period of time.
3166 Description: improve handling recoverable errors
3167 Details : if request processig with error which can be recoverable on server
3168 request should be resend, otherwise page released from cache and
3171 Severity : enhancement
3173 Description: refine locking for avoid write wrong info into lov_objid file
3174 Details : fix possible races with add new target and write/update data in
3177 --------------------------------------------------------------------------------
3179 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
3184 Frequency : on some architectures
3186 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
3187 Details : Uninitialized flags sometimes cause configuration commands to
3191 Frequency : patchless clients only
3193 Description: missing __iget() symbol export
3194 Details : The __iget() symbol export is missing. To avoid the need for
3195 this on patchless clients the deathrow inode reaper is turned
3196 off, and we depend on the VM to clean up old inodes. This
3197 dependency was during via the fix for bug 12181.
3199 --------------------------------------------------------------------------------
3201 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
3203 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
3204 INTEROPERATE with older versions automatically. In many cases a
3205 special upgrade step is needed. Please read the
3206 user documentation before upgrading any part of a live system.
3207 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
3208 version will not interoperate with 1.6 betas before beta5 (1.5.95).
3209 * WARNING: Lustre configuration and startup changes are required with
3210 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
3214 Severity : enhancement
3216 Description: MountConf
3217 Details : Lustre configuration is now managed via mkfs and mount
3218 commands instead of lmc and lconf. New obd types (MGS, MGC)
3219 are added for dynamic configuration management. See
3220 https://mail.clusterfs.com/wikis/lustre/MountConf for
3223 Severity : enhancement
3225 Description: dynamic OST addition
3226 Details : OSTs can now be added to a live filesystem
3228 Severity : enhancement
3230 Description: startup order invariance
3231 Details : MDTs and OSTs can be started in any order. Clients only
3232 require the MDT to complete startup.
3234 Severity : enhancement
3236 Description: parallel, asynchronous orphan cleanup
3237 Details : orphan cleanup is now performed in separate threads for each
3238 OST, allowing parallel non-blocking operation.
3240 Severity : enhancement
3242 Description: optimized stripe assignment
3243 Details : stripe assignments are now made based on ost space available,
3244 ost previous usage, and OSS previous usage, in order to try
3245 to optimize storage space and networking resources.
3247 Severity : enhancement
3249 Description: Permanently set tunables
3250 Details : All writable /proc/fs/lustre tunables can now be permanently
3251 set on a per-server basis, at mkfs time or on a live system.
3253 Severity : enhancement
3255 Description: Lustre message v2
3256 Details : Add lustre message format v2.
3258 Severity : enhancement
3260 Description: client OST exclusion list
3261 Details : Clients can be started with a list of OSTs that should be
3262 declared "inactive" for known non-responsive OSTs.
3266 Description: SPEC SFS validation failure on NFS v2 over lustre.
3267 Details : Changes the blocksize for regular files to be 2x RPC size,
3268 and not depend on stripe size.
3270 Severity : enhancement
3272 Description: Multiple MD RPCs in flight.
3273 Details : Further unserialise some read-only MDS RPCs - learn about intents.
3274 To avoid overly-overloading MDS, introduce a limit on number of
3275 MDS RPCs in flight for a single client and add /proc controls
3276 to adjust this limit.
3278 Severity : enhancement
3280 Description: client read/write statistics
3281 Details : Add client read/write call usage stats for performance
3282 analysis of user processes.
3283 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
3284 file access. extents_stats shows chunk size distribution.
3285 extents_stats_per_process show chunk size distribution per
3288 Severity : enhancement
3290 Description: mds statistics
3291 Details : Add detailed mds operations statistics in
3292 /proc/fs/lustre/mds/*/stats.
3296 Description: Failure of copying files with lustre special EAs.
3297 Details : Client side always return success for setxattr call for lustre
3298 special xattr (currently only "trusted.lov").
3303 Description: Refcount LNET uuids
3304 Details : The global LNET uuid list grew linearly with every startup;
3305 refcount repeated list entries instead of always adding to
3308 Severity : enhancement
3310 Description: Dynamic service threads
3311 Details : Within a small range, start extra service threads
3312 automatically when the request queue builds up.
3315 Frequency : mixed-endian client/server environments
3317 Description: mixed-endian crashes
3318 Details : The new msg_v2 system had some failures in mixed-endian
3321 Severity : enhancement
3323 Description: Easy OST removal
3324 Details : OSTs can be permanently deactivated with e.g. 'lctl
3325 conf_param lustre-OST0001.osc.active=0'
3327 Severity : enhancement
3329 Description: MGS proc entries
3330 Details : Added basic proc entries for the MGS showing what filesystems
3333 Severity : enhancement
3335 Description: provide MGS failover
3336 Details : Added config lock reacquisition after MGS server failover.
3338 Severity : enhancement
3340 Description: add Linux 2.4 support
3341 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
3345 Description: a large application tries to do I/O to the same resource and dies
3346 in the middle of it.
3347 Details : Check the req->rq_arrival time after the call to
3348 ost_brw_lock_get(), but before we do anything about
3349 processing it & sending the BULK transfer request. This
3350 should help move old stale pending locks off the queue as
3351 quickly as obd_timeout.
3354 Frequency : when an incorrect nid is specified during startup
3356 Description: ptlrpc connect to non-existant node causes kernel crash
3357 Details : LNET can't be re-entered from an event callback, which
3358 happened when we expire a message after the export has been
3359 cleaned up. Instead, hand the zombie cleanup off to another
3362 Severity : enhancement
3364 Description: plain/inodebits lock performance improvement
3365 Details : Grouping plain/inodebits in granted list by their request modes
3366 and bits policy, thus improving the performance of search through
3370 Frequency : only if OST filesystem is corrupted
3372 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3373 Details : for a short time RPCs with bulk IO are in the replay list,
3374 but replay of bulk IOs is unimplemented. If the OST filesystem
3375 is corrupted due to disk cache incoherency and then replay is
3376 started it is possible to trip an assertion. Avoid putting
3377 committed RPCs into the replay list at all to avoid this issue.
3380 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3382 Description: System hang on startup
3383 Details : This bug allowed the liblustre (e.g. catamount) client to
3384 return to the app before handling all startup RPCs. This
3385 could leave the node unresponsive to lustre network traffic
3386 and manifested as a server ptllnd timeout.
3388 Severity : enhancement
3390 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3391 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
3392 boolean to control whether to print peer debug info when a
3393 client's RPC times out.
3396 Frequency : only for kernels with patches from Lustre below 1.4.3
3398 Description: Remove old rdonly API
3399 Details : Remove old rdonly API which unused from at least lustre 1.4.3
3402 Frequency : only for devices with external journals
3404 Description: Set external device read-only also
3405 Details : During a commanded failover stop, we set the disk device
3406 read-only while the server shuts down. We now also set any
3407 external journal device read-only at the same time.
3410 Frequency : when upgrading from 1.4 while trying to change parameters
3412 Description: The wrong (new) MDC name was used when setting parameters for
3413 upgraded MDT's. Also allows changing of OSC (and MDC)
3414 parameters if --writeconf is specified at tunefs upgrade time.
3417 Frequency : when setting specific ost indicies
3419 Description: QOS code breaks on skipped indicies
3420 Details : Add checks for missing OST indicies in the QOS code, so OSTs
3421 created with --index need not be sequential.
3423 Severity : enhancement
3425 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3426 Details : The uninit_groups feature works in conjunction with the kernel
3427 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3428 up the pass1 processing of e2fsck. This is a read-only feature
3429 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3430 mount filesystems that have had this feature enabled.
3432 Severity : enhancement
3434 Description: Improve multi-block allocation algorithm to avoid fragmentation
3435 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3436 allocation locality and avoid filesystem fragmentation.
3438 ------------------------------------------------------------------------------
3440 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
3442 * Support for kernels:
3443 2.4.21-47.0.1.EL (RHEL 3)
3444 2.6.5-7.283 (SLES 9)
3445 2.6.9-42.0.10.EL (RHEL 4)
3446 2.6.12.6 vanilla (kernel.org)
3447 2.6.16.27-0.9 (SLES 10)
3448 * Recommended e2fsprogs version: 1.39.cfs5
3450 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3454 Frequency : occasional, depends on client load and configuration
3455 Bugzilla : 12181, 12203
3456 Description: data loss for recently-modified files
3458 Details : In some cases it is possible that recently written or created
3459 files may not be written to disk in a timely manner (this should
3460 normally be within 30s unless client IO load is very high).
3461 The problem appears as zero-length files or files that are a
3462 multiple of 1MB in size after a client crash or client eviction
3463 that are missing data at the end of the file.
3465 This problem is more likely to be hit on clients where files are
3466 repeatedly created and unlinked in the same directory, clients
3467 have a large amount of RAM, have many CPUs, the filesystem has
3468 many OSTs, the clients are rebooted frequently, and/or the files
3469 are not accessed by other nodes after being written.
3471 The presence of the problem can be detected by looking at
3472 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
3473 smaller than the second (nr_unused) then dirty files will not
3474 be flushed automatically to disk. "sync; sleep 10" should be
3475 run several times on the node before unmounting it to update
3476 Lustre (this is also safe to run on nodes without this problem).
3478 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
3479 that can cause this same problem, so customers using that kernel
3480 also need to update the kernel in addition to Lustre. In order
3481 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
3483 It is normal that files written just before a client crash (less
3484 than 30s) may not yet have been flushed to disk, even for local
3488 Frequency : frequent on thin XT3 nodes
3490 Description: UUID collision on thin XT3 Linux nodes
3491 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
3492 randomly, since we relied on an insufficiently-seeded PRNG.
3497 Description: OSS hangs after "All ost request buffers busy"
3498 Details : A deadlock between quota and journal operations caused OSS
3499 hangs after printing "All ost request buffers busy."
3502 Frequency : always on liblustre builds
3504 Description: Cleanup compiler warnings on liblustre
3507 Frequency : always on liblustre builds on XT3
3509 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
3513 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
3514 > 32000 subdirectories
3515 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
3516 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
3517 subdirectory count crosses 32000. This will aid e2fsck to
3518 correctly handle more than 32000 subdirectories.
3521 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3523 Description: System hang on startup
3524 Details : This bug allowed the liblustre (e.g. catamount) client to
3525 return to the app before handling all startup RPCs. This
3526 could leave the node unresponsive to lustre network traffic
3527 and manifested as a server ptllnd timeout.
3529 Severity : enhancement
3531 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3532 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
3533 boolean to control whether to print peer debug info when a
3534 client's RPC times out.
3539 Description: make O_SYNC working on 2.6 kernels
3540 Details : 2.6 kernels use different method for mark pages for write,
3541 so need add a code to lustre for O_SYNC work.
3546 Description: Failure to close file and release space on NFS
3547 Details : Put inode details into lock acquired in ll_intent_file_open.
3548 Use mdc_intent_lock in ll_intent_open to properly
3549 detect all kind of errors unhandled by mdc_enqueue.
3554 Description: proc file read during shutdown sometimes raced obd removal,
3556 Details : Add lock to prevent obd access after proc file removal.
3559 Frequency : Only for files larger than 4GB on 32-bit clients.
3561 Description: improperly doing page alignment of locks
3562 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
3563 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3568 Description: RPCs being resent when they shouldn't be
3569 Details : Some RPCs that should not be resent are being resent. This
3570 can cause inconsistencies in the RPC state machine. Do not
3571 resend such requests.
3574 Frequency : rare, only with NFS export
3576 Description: Crash on NFS re-export node
3577 Details : under very unusual load conditions an assertion is hit in
3578 ll_intent_file_open()
3581 Frequency : only if OST filesystem is corrupted
3583 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3584 Details : for a short time RPCs with bulk IO are in the replay list,
3585 but replay of bulk IOs is unimplemented. If the OST filesystem
3586 is corrupted due to disk cache incoherency and then replay is
3587 started it is possible to trip an assertion. Avoid putting
3588 committed RPCs into the replay list at all to avoid this issue.
3593 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3594 Details : Large single O_DIRECT read and write calls can fail to allocate
3595 a sufficiently large buffer to process the request. In case of
3596 allocation failure the allocation is retried with a smaller
3597 buffer and broken into smaller requests.
3599 Severity : enhancement
3601 Description: Add -o localflock option to simulate old noflock behaviour.
3602 Details : This will achieve local-only flock/fcntl locks coherentness.
3607 Description: versioning check is incomplete
3608 Details : Checking the version difference of client vs. server, report
3609 error if the gap is too big.
3614 Description: add support PG_writeback bit
3615 Details : add support for PG_writeback bit for Lustre, for more carefull
3616 work with page cache in 2.6 kernel. This also fix some deadlocks
3617 and remove hack for work O_SYNC with 2.6 kernel.
3619 Severity : enhancement
3621 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3622 Details : The uninit_groups feature works in conjunction with the kernel
3623 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3624 up the pass1 processing of e2fsck. This is a read-only feature
3625 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3626 mount filesystems that have had this feature enabled.
3628 Severity : enhancement
3630 Description: Improve multi-block allocation algorithm to avoid fragmentation
3631 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3632 allocation locality and avoid filesystem fragmentation.
3634 ------------------------------------------------------------------------------
3636 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3638 * Support for kernels:
3639 2.6.9-42.0.3.EL (RHEL 4)
3640 2.6.5-7.276 (SLES 9)
3641 2.4.21-47.0.1.EL (RHEL 3)
3642 2.6.12.6 vanilla (kernel.org)
3643 2.6.16.21-0.8 (SLES10)
3644 * Recommended e2fsprogs version: 1.39.cfs2-0
3646 * The backwards-compatible /proc/sys/portals symlink has been removed
3647 in this release. Before upgrading, please ensure that you change
3648 any configuration scripts or /etc/sysctl.conf files that access
3649 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3650 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3651 in advance of the upgrade on any system running Lustre 1.4.6 or
3652 newer, since /proc/sys/lnet was added in that version.
3653 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3657 Frequency : only when quota is used
3659 Description: avoid scanning export list for quota master
3660 Details : Change the algorithms to avoid scanning export list in order
3661 to improve the efficiency.
3664 Frequency : MDS failover only, very rarely
3666 Description: "went back in time" messages on mds failover
3667 Details : The greatest transno may be lost when the current operation
3668 finishes with an error (transno==0) and the client's last_rcvd
3669 record is over-written. Save the greatest transno in the
3670 mds_last_transno for this case.
3673 Frequency : always for specific kernels and striping counts
3675 Description: client may get "Matching packet too big" without ACL support
3676 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
3677 when trying to access files in certain configurations. The
3678 clients should in fact be denied when mounting because they do
3679 not understand ACLs.
3682 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
3684 Description: many clients connecting with IO in progress causes connect timeouts
3685 Details : Avoid synchronous journal commits to avoid delays caused by many
3686 clients connecting/disconnecting when bulk IO is in progress.
3687 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
3688 OST_IO_PORTAL to avoid delays behind potentially many pending
3692 Frequency : occasionally with multiple writers to a single file
3694 Description: shared writes to file may result in wrong size reported by stat()
3695 Details : Allow growing of kms when extent lock is cancelled
3698 Frequency : always with random mmap IO to multi-striped file
3700 Description: mmap write might be lost if we are writing to a 'hole' in stripe
3701 Details : Only if the hole is at the end of OST object so that kms is too
3702 small. Fix is to increase kms accordingly in ll_nopage.
3705 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
3707 Description: writes to a missing object would leak memory on the OST
3708 Details : If there is an inconsistency between the MDS and OST filesystems,
3709 such that the MDS references an object that doesn't exist, writes
3710 to that object will leak memory due to incorrect cleanup in the
3711 error handling path, eventually running out of memory on the OST.
3716 Description: Creating too long symlink causes lustre errors
3717 Details : Check symlink and name lengths before sending requests to MDS.
3720 Frequency : only if flock is enabled (not on by default)
3722 Description: posix locks not released on fd closure on 2.6.9+
3723 Details : We failed to add posix locks to list of inode locks on 2.6.9+
3724 kernels, this caused such locks not to be released on fd close and
3725 then assertions on fs unmount about still used locks.
3728 Frequency : MDS failover only, very rarely
3730 Description: clients may get ASSERTION(granted_lock != NULL)
3731 Details : When request was taking a long time, and a client was resending
3732 a getattr by name lock request. The were multiple lock requests
3733 with the same client lock handle and
3734 mds_getattr_name->fixup_handle_for_resent_request found one of the
3735 lock handles but later failed with ASSERTION(granted_lock != NULL).
3740 Description: handle->h_buffer_credits > 0, assertion failure
3741 Details : h_buffer_credits is zero after truncate, causing assertion
3742 failure. This patch extends the transaction or creates a new
3746 Frequency : NFS re-export or patchless client
3747 Bugzilla : 11179, 10796
3748 Description: Crash on NFS re-export node (__d_move)
3749 Details : We do not want to hash the dentry if we don't have a lock.
3750 But if this dentry is later used in d_move, we'd hit uninitialised
3751 list head d_hash, so we just do this to init d_hash field but
3752 leave dentry unhashed.
3755 Frequency : NFS re-export or patchless client
3757 Description: NFS exports has problem with symbolic link
3758 Details : lustre client didn't properly install dentry when re-exported
3759 to NFS or running patchless client.
3762 Frequency : NFS re-export or patchless client
3764 Description: Various nfs/patchless fixes.
3765 Details : fixes reuse disconected alias for lookup process - this fixes
3766 warning "find_exported_dentry: npd != pd",
3767 fix permission error with open files at nfs.
3768 fix apply umask when do revalidate.
3771 Frequency : occasional
3773 Description: Crash on NFS re-export node
3774 Details : calling clear_page() on the wrong pointer triggered oops in
3775 generic_mapping_read().
3778 Frequency : rarely, using O_DIRECT IO
3780 Description: unaligned directio crashes client with LASSERT
3781 Details : check for unaligned buffers before trying any requests.
3784 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
3786 Description: stale data returned from RAID cache
3787 Details : If only a small amount of IO is done to the RAID device before
3788 reading it again it is possible to get stale data from the RAID
3789 cache instead of reading it from disk.
3792 Frequency : always for sles10 kernel
3794 Description: sles10 support
3795 Details : ll_follow_link: compile fixes and using of nd_set_link
3796 under newer kernels.
3799 Frequency : depends on arch, kernel and compiler version, always on sles10
3802 Description: recursive or deep enough symlinks cause stack overflow
3803 Details : getting rid of large stack-allocated variable in
3807 Frequency : depends on hardware
3809 Description: lustre write performance loss in the SLES10 kernel
3810 Details : the performance loss is caused by using of write barriers in the
3811 ext3 code. The SLES10 kernel turns barrier support on by
3812 default. The fix is to undo that change for ldiskfs.
3814 ------------------------------------------------------------------------------
3816 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
3818 * Support for kernels:
3819 2.6.9-42.0.3EL (RHEL 4)
3820 2.6.5-7.276 (SLES 9)
3821 2.4.21-47.0.1.EL (RHEL 3)
3822 2.6.12.6 vanilla (kernel.org)
3826 Frequency : quota enabled and large files being deleted
3828 Description: releasing more than 4GB of quota at once hangs OST
3829 Details : If a user deletes more than 4GB of files on a single OST it
3830 will cause the OST to spin in an infinite loop. Release
3831 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3836 Description: statfs data retrieved from /proc may be stale or zero
3837 Details : When reading per-device statfs data from /proc, in the
3838 {kbytes,files}_{total,free,avail} files, it may appear
3839 as zero or be out of date.
3842 Frequency : systems with MD RAID1 external journal devices
3844 Description: lconf's call to blkid is confused by RAID1 journal devices
3845 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3846 of returning all block devices that match the journal UUID.
3849 Frequency : always, for aggregate stripe size over 4GB
3851 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
3852 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
3853 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
3856 Frequency : always if "lfs find" used on a local file/directory
3858 Description: "lfs find" segfaults if used on a local file/directory
3859 Details : The case where a directory component was not specified wasn't
3860 handled correctly. Handle this properly.
3863 Frequency : always on ppc64
3865 Description: the write to an ext3 filesystem mounted with mballoc got stuck
3866 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3867 perform endianness conversion.
3870 Frequency : rarely (truncate to non-zero file size after write under load)
3871 Bugzilla : 10730, 10687
3872 Description: Files padded with zeros to next 4K multiple
3873 Details : With filesystems mounted using the "extents" option (2.6 kernels)
3874 it is possible that files that are truncated to a non-zero size
3875 immediately after being written are filled with zero bytes beyond
3876 the truncated size. No file data is lost.
3878 Severity : enhancement
3880 Description: Allow recovery/failover for liblustre clients.
3881 Details : liblustre clients were unaware of failover configurations until
3884 Severity : enhancement
3886 Description: user file locks should fail when not mounting with flock option
3887 Details : Set up an error-returning stub in ll_file_operations.lock field
3888 to prevent incorrect behaviour when client is mounted without
3889 flock option. Also, set up properly f_op->flock field for
3893 Frequency : always on ia64
3895 Description: "lfs df" loops on printing out MDS statfs information
3896 Details : The obd_ioctl_data was not initialized and in some systems
3897 this caused a failure during the ioctl that did not return
3898 an error. Initialize the struct and return an error on failure.
3901 Frequency : SLES 9 only
3903 Description: Error of copying files with lustre special EAs as root
3904 Details : Client side always return success for setxattr call for lustre
3905 special xattr (currently only "trusted.lov").
3908 Frequency : rarely on clusters with both ia64+i386 clients
3910 Description: ia64+i686 clients doing shared IO on the same file may LBUG
3911 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
3912 clients are doing concurrent writes to the same file it is
3913 possible that the ia64 clients may LASSERT because the OST
3914 extent locks are not PAGE_SIZE aligned. Ensure that grown
3915 locks are always aligned on the request boundary.
3918 Frequency : specific use, occasional
3920 Description: Overwriting in use executable truncates on-disk binary image
3921 Details : If one node attempts to overwrite an executable in use by
3922 another node, we now correctly return ETXTBSY instead of
3923 truncating the file.
3925 Severity : enhancement
3927 Description: Async OSC create to avoid the blocking unnecessarily.
3928 Details : If an OST has no remaining object, system will block on the creating
3929 when need to create a new object on this OST. Now, ways use
3930 pre-created objects when available, instead of blocking on an
3931 empty osc while others are not empty. If we must block, we block
3932 for the shortest possible period of time.
3937 Description: chmod on Lustre root is propagated to other clients
3938 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
3939 invalid by the follow_mount time.
3942 Frequency : liblustre clients only
3944 Description: Race in 'instant cancel' lock handling could lead to such locks
3945 never to be granted in case of SMP MDS
3946 Details : Do not destroy not yet granted but cbpending locks in
3950 Frequency : replay/resend of open
3952 Description: non null lock assetion failure in mds_intent_policy
3953 Details : Trying to replay/resend lockless open requests resulted in
3954 mds_open() returning 0 with no lock. Now it sets a flag if
3955 a lock is going to be returned.
3957 Severity : enhancement
3959 Description: Checksum enhancements
3960 Details : New checksum enhancements allow for resending RPCs that failed
3963 Severity : enhancement
3965 Description: Tunables on number of dirty pages in cacche
3966 Details : Allow to set limit on number of dirty pages cached.
3971 Description: client crash on unmount - lock still has references
3972 Details : In some error handling cases it was possible to leak a lock
3973 reference on a client while accessing a file. This was not
3974 harmful to the client during operation, but would cause the
3975 client to crash when the filesystem is unmounted.
3978 Frequency : specific case, rare
3980 Description: ETXTBSY on mds though file not in use by client
3981 Details : ETXTBSY is no longer incorrectly returned when attempting to
3982 chmod or chown a directory that the user previously tried to
3983 execute or a currently-executing binary.
3986 Frequency : extremely rare except on liblustre-based clients
3988 Description: Lustre space not freed when files are deleted
3989 Details : Clean up open-unlinked files after client eviction. Previously
3990 the unlink was skipped and the files remained as orphans.
3995 Description: OST failure "would be an LBUG" in waiting_locks_callback()
3996 Details : In some cases it was possible to send a blocking callback to
3997 a client doing a glimpse, even though that client didn't get
3998 a lock granted. When the glimpse lock is cancelled on the OST
3999 the freed lock is left on the waiting list and corrupted the list.
4002 Frequency : all core dumps
4004 Description: Broke core dumps to lustre
4005 Details : Negative dentry may be unhashed if parent does not have UPDATE
4006 lock, but some callers, e.g. do_coredump, expect dentry to be
4007 hashed after successful create, hash it in ll_create_it.
4009 ------------------------------------------------------------------------------
4011 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
4013 * Support for kernels:
4014 2.6.9-42.0.2.EL (RHEL 4)
4015 2.6.5-7.276 (SLES 9)
4016 2.4.21-40.EL (RHEL 3)
4017 2.6.12.6 vanilla (kernel.org)
4021 Frequency : always on RHEL 3
4023 Description: Number of open files grows over time
4024 Details : The number of open files grows over time, whether or not
4025 Lustre is started. This was due to a filp leak introduced
4026 by one of our kernel patches.
4028 ------------------------------------------------------------------------------
4030 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4032 * Support for kernels:
4033 2.6.9-42.EL (RHEL 4)
4034 2.6.5-7.276 (SLES 9)
4035 2.4.21-40.EL (RHEL 3)
4036 2.6.12.6 vanilla (kernel.org)
4041 Bugzilla : 5719, 9635, 9792, 9684
4042 Description: OST (or MDS) trips assertions in (re)connection under heavy load
4043 Details : If a server is under heavy load and cannot reply to new
4044 connection requests before the client resends the (re)connect,
4045 the connection handling code can behave badly if two service
4046 threads are concurrently handing separate (re)connections from
4047 the same client. Add better locking to the connection handling
4048 code, and ensure that only a single connection will be processed
4049 for a given client UUID, even if the lock is dropped.
4051 Severity : enhancement
4053 Description: add TCP zero-copy support to kernel
4054 Details : Add support to the kernel TCP stack to allow zero-copy bulk
4055 sends if the hardware supports scatter-gather and checksumming.
4056 This allows socklnd to do client-write and server-read more
4057 efficiently and reduce CPU utilization from skbuf copying.
4060 Frequency : only if NFS exporting from client
4062 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
4063 Details : If setting attributes on a file created under NFS that had
4064 never been opened it would be possible to oops the client
4065 if the file had no objects.
4068 Frequency : always for liblustre
4070 Description: liblustre client does MDS+OSTs setattr RPC for each write
4071 Details : When doing a write from a liblustre client, the client
4072 incorrectly issued an RPC to the MDS and each OST the file was
4073 striped over in order to update the timestamps. When writing
4074 with small chunks and many clients this could overwhelm the MDS
4075 with RPCs. In all cases it would slow down the write because
4076 these RPCs are unnecessary.
4078 Severity : enhancement
4080 Description: allow number of MDS service threads to be changed at module load
4081 Details : It is now possible to change the number of MDS service threads
4082 running. Adding "options mds mds_num_threads={N}" to the MDS's
4083 /etc/modprobe.conf will set the number of threads for the next
4084 time Lustre is restarted (assuming the "mds" module is also
4085 reloaded at that time). The default number of threads will
4086 stay the same, 32 for most systems.
4091 Description: OST crash if filesystem is unformatted or corrupt
4092 Details : If an OST is started on a device that has never been formatted
4093 or if the filesystem is corrupt and cannot even mount then the
4094 error handling cleanup routines would dereference a NULL pointer.
4099 Description: NULL pointer deref in llap_from_page.
4100 Details : get_cache_page_nowait can return a page with NULL (or otherwise
4101 incorrect) mapping if the page was truncated/reclaimed while it was
4102 searched for. Check for this condition and skip such pages when
4103 doing readahead. Introduce extra check to llap_from_page() to
4104 verify page->mapping->host is non-NULL (so page is not anonymous).
4107 Frequency : Sometimes when using sys_sendfile
4109 Description: "page not covered by a lock" warnings from ll_readpage
4110 Details : sendfile called ll_readpage without right page locks present.
4111 Now we introduced ll_file_sendfile that does necessary locking
4112 around call to generic_file_sendfile() much like we do in
4116 Frequency : with certain MDS communication failures at client mount time
4118 Description: NULL pointer deref after failed client mount
4119 Details : a client connection request may delayed by the network layer
4120 and not be sent until after the PTLRPC layer has timed out the
4121 request. If the client fails the mount immediately it will try
4122 to clean up before the network times out the request. Add a
4123 reference from the request import to the obd device and delay
4124 the cleanup until the network drops the request.
4127 Frequency : occasionally during client (re)connect
4129 Description: assertion failure during client (re)connect
4130 Details : processing a client connection request may be delayed by the
4131 client or server longer than the client connect timeout. This
4132 causes the client to resend the connection request. If the
4133 original connection request is replied in this interval, the
4134 client may trip an assertion failure in ptlrpc_connect_interpret()
4135 which thought it would be the only running connect process.
4138 Frequency : only with obd_echo servers and clients that are rebooted
4140 Description: kernel BUG accessing uninitialized data structure
4141 Details : When running an obd_echo server it did not start the ping_evictor
4142 thread, and when a client was evicted an uninitialized data
4143 structure was accessed. Start the ping_evictor in the RPC
4144 service startup instead of the OBD startup.
4146 Severity : enhancement
4147 Bugzilla : 10193 (patchless)
4148 Description: Remove dependency on various unexported kernel interfaces.
4149 Details : No longer need reparent_to_init, exit_mm, exit_files,
4150 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
4153 Frequency : rare (only users of deprecated and unsupported LDAP config)
4155 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
4156 Details : LDAP apparently contains 'lustreName' attributes instead of
4157 'name'. A simple remapping of the name is sufficient.
4160 Frequency : rare (only with non-default dump_on_timeout debug enabled)
4162 Description: waiting_locks_callback trips kernel BUG if client is evicted
4163 Details : Running with the dump_on_timeout debug flag turned on makes
4164 it possible that the waiting_locks_callback() can try to dump
4165 the Lustre kernel debug logs from an interrupt handler. Defer
4166 this log dumping to the expired_lock_main() thread.
4168 Severity : enhancement
4170 Description: Support NFS exporting on 2.6 kernels.
4171 Details : Implement non-rawops metadata methods for NFS server to use without
4172 changing NFS server code.
4175 Frequency : very rare (synthetic metadata workload only)
4177 Description: two racing renames might cause an MDS thread to deadlock
4178 Details : Running the "racer" program may cause one MDS thread to rename
4179 a file from being the source of a rename to being the target of
4180 a rename at exactly the same time that another thread is doing
4181 so, and the second thread has already enqueued these locks after
4182 doing a lookup of the target and is trying to relock them in
4183 order. Ensure that we don't try to re-lock the same resource.
4186 Frequency : only very large systems with liblustre clients
4188 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
4189 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
4190 all OSTs in parallel. This allows the eviction of stale liblustre
4191 clients to proceed much faster than if they were done in series,
4192 and also offers similar improvements for other set_info RPCs.
4197 Description: excessive CPU usage during initial read phase on client
4198 Details : During the initial read phase on a client, it would agressively
4199 retry readahead on the file, consuming too much CPU and impacting
4200 performance (since 1.4.5.8). Improve the readahead algorithm
4201 to avoid this, and also improve some other common cases (read
4202 of small files in particular, where "small" is files smaller than
4203 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
4208 Description: MDS crash when receiving packet with unknown intent.
4209 Details : Do not LBUG in unknown intent case, just return -EFAULT
4211 Severity : enhancement
4212 Bugzilla : 9293, 9385
4213 Description: MDS RPCs are serialised on client. This is unnecessary for some.
4214 Details : Do not serialize getattr (non-intent version) and statfs.
4217 Frequency : occasional, when OST network is overloaded/intermittent
4219 Description: client evicted by OST after bulk IO timeout
4220 Details : If a client sends a bulk IO request (read or write) the OST
4221 may evict the client if it is unresposive to its data GET/PUT
4222 request. This is incorrect if the network is overloaded (takes
4223 too long to transfer the RPC data) or dropped the OST GET/PUT
4224 request. There is no need to evict the client at all, since
4225 the pinger and/or lock callbacks will handle this, and the
4226 client can restart the bulk request.
4229 Frequency : Always when mmapping file with no objects
4231 Description: client crashes when mmapping file with no objects
4232 Details : Check that we actually have objects in a file before doing any
4233 operations on objects in ll_vm_open, ll_vm_close and
4239 Description: Request leak when working with deleted CWD
4240 Details : Introduce advanced request refcount tracking for requests
4241 referenced from lustre intent.
4243 Severity : Enhancement
4245 Description: Cache open file handles on client.
4246 Details : MDS now will return special lock along with openhandle, if
4247 requested and client is allowed to hold openhandle, even if unused,
4248 until such a lock is revoked. Helps NFS a lot, since NFS is opening
4249 closing files for every read/write openration.
4251 Severity : Enhancement
4253 Description: Cache open negative dentries on client when possible.
4254 Details : Guard negative dentries with UPDATE lock on parent dir, drop
4255 negative dentries on lock revocation.
4260 Description: Remounting a client read-only wasn't possible with a zconf mount
4261 Details : It wasn't possible to remount a client read-only with llmount.
4263 Severity : enhancement
4264 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
4265 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
4266 MPICH is included to add Lustre-specific ADIO interfaces.
4267 This is based closely on the UFS ADIO layer and only differs
4268 in file creation, in order to allow the OST striping to be set.
4269 This is user-contributed code and not supported by CFS.
4274 Description: extended inode attributes (immutable, append-only) work improperly
4275 when 2.4 and 2.6 kernels are used on client/server or vice versa
4276 Details : Introduce kernel-independent values for these flags.
4278 Severity : enhancement
4281 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
4282 Details : Many of the /proc/ tunables can only be tuned at a megabyte
4283 granularity. Now, Fractional MB granularity is be supported,
4284 this is very useful for low memory system.
4286 Severity : enhancement
4288 Description: Getattr by fid
4289 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
4290 locks, avoids extra getattr rpc requests to MDS, allows '/' to
4291 have locks and avoids getattr rpc requests for it on every stat.
4294 Frequency : Always, for filesystems larger than 2TB
4296 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4297 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
4298 filesystems larger than 2TB to be mounted with mballoc enabled.
4301 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
4303 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4304 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
4305 filesystems larger than 2TB is mounted then the kernel will
4306 silently corrupt the start of the filesystem. CONFIG_LBD is
4307 enabled for all CFS-supported kernels, but the possibility of
4308 this happening with a modified kernel config exists.
4310 Severity : enhancement
4312 Description: add client O_DIRECT support for 2.6 kernels
4313 Details : It is now possible to do O_DIRECT reads and writes to files
4314 in the Lustre client mountpoint on 2.6 kernel clients.
4316 Severity : enhancement
4318 Description: parallel glimpse, setattr, statfs, punch, destroy requests
4319 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
4320 parallel, not waiting for response from every OST before sending
4321 a rpc to the next OST.
4326 Description: setattr vs write race when updating file timestamps
4327 Details : Client processes that update a file timestamp into the past
4328 right after writing to the file (e.g. tar) it is possible that
4329 the updated file modification time can be reset to the current
4330 time due to a race between processing the setattr and write RPC.
4332 Severity : enhancement
4334 Description: Bring 'lfs find' closer in line with regular Linux find.
4335 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
4336 -print0 options and obtains all the needed info through the lustre
4339 Severity : enhancement
4341 Description: support up to 1024 configured devices on one node
4342 Details : change obd_dev array from statically allocated to dynamically
4343 allocated structs as they are first used to reduce memory usage
4348 Description: Flush dirty partially truncated pages during truncate
4349 Details : Immediatelly flush partially truncated pages in filter_setattr,
4350 this way we completely avoid having any pages in page cache on OST
4351 and can retire ugly workarounds during writes to flush such pages.
4356 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
4357 Details : protect inode from truncation within vfs_unlink() context
4358 just take a reference before calling vfs_unlink() and release it
4359 when parent's i_sem is free.
4364 Description: last_id value checked outside lock on OST caused LASSERT failure
4365 Details : If there were multiple MDS->OST object precreate requests in
4366 flight, it was possible that the OST's last object id was checked
4367 outside a lock and incorrectly tripped an assertion. Move checks
4368 inside locks, and discard old precreate requests.
4371 Frequency : always, if extents are used on OSTs
4373 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
4374 Details : OSTs using the extents format would not zero the high 16 bits of
4375 the index physical block number. This is not a problem for any
4376 OST filesystems smaller than 16TB, and no kernels support ext3
4377 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
4378 new/modified files) and can be fixed for existing filesystems
4379 with e2fsprogs-1.39-cfs1.
4384 Description: import connection selection may be incorrect if timer wraps
4385 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
4386 import connections to be ignored if the 32-bit jiffies counter
4387 wraps. Use a 64-bit jiffies counter.
4390 Frequency : very large clusters immediately after boot
4392 Description: LNET request buffers exhausted under heavy short-term load
4393 Details : If a large number of client requests are generated on a service
4394 that has previously never seen so many requests it is possible
4395 that the request buffer growth cannot keep up with the spike in
4396 demand. Instead of dropping incoming requests, they are held in
4397 the LND until the RPC service can accept more requests.
4400 Frequency : Sometimes during replay
4402 Description: Assertion failure in ll_local_open after replay.
4403 Details : If replay happened on an open request reply before we were able
4404 to set replay handler, reply will become not swabbed tripping the
4405 assertion in ll_local_open. Now we set the handler right after
4406 recognising of open request
4409 Frequency : very rare
4411 Description: kernel reports "badness in vsnprintf"
4412 Details : Reading from the "recovery_status" /proc file in small chunks
4413 may cause a negative length in lprocfs_obd_rd_recovery_status()
4414 call to vsnprintf() (which is otherwise harmless). Exit early
4415 if there is no more space in the output buffer.
4417 Severity : enhancement
4419 Description: clear OBD RPC statistics by writing to them
4420 Details : It is now possible to clear the OBD RPC statistics by writing
4421 to the "stats" file.
4426 Description: Client mtime is not the same on different clients after utimes
4427 Details : In some cases, the client was using the utimes() syscall on
4428 a file cached on another node. The clients now validate the
4429 ctime from the MDS + OSTs to determine which one is right.
4434 Description: Inability to activate failout mode
4435 Details : lconf script incorrectly assumed that in pythong string's numeric
4436 value is used in comparisons.
4439 Frequency : always with multiple stripes per file
4441 Description: Inefficient object allocation for mutli-stripe files
4442 Details : When selecting which OSTs to stripe files over, for files with
4443 a stripe count that divides evenly into the number of OSTs,
4444 the MDS is always picking the same starting OST for each file.
4445 Return the OST selection heuristic to the original design.
4450 Description: mount failures may take full timeout to return an error
4451 Details : Under some heavy load conditions it is possible that a
4452 failed mount can wait for the full obd_timeout interval,
4453 possibly several minutes, before reporting an error.
4454 Instead return an error as soon as the status is known.
4456 Frequency : quota enabled and large files being deleted
4458 Description: releasing more than 4GB of quota at once hangs OST
4459 Details : If a user deletes more than 4GB of files on a single OST it
4460 will cause the OST to spin in an infinite loop. Release
4461 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4466 Description: statfs data retrieved from /proc may be stale or zero
4467 Details : When reading per-device statfs data from /proc, in the
4468 {kbytes,files}_{total,free,avail} files, it may appear
4469 as zero or be out of date.
4472 Frequency : systems with MD RAID1 external journal devices
4474 Description: lconf's call to blkid is confused by RAID1 journal devices
4475 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4476 of returning all block devices that match the journal UUID.
4479 Frequency : always, for aggregate stripe size over 4GB
4481 Description: assertion fails when trying to use 4GB stripe size
4482 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
4483 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
4486 Frequency : always on ppc64
4488 Description: the first write on an ext3 filesystem with mballoc got stuck
4489 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4490 perform endianness conversion.
4492 ------------------------------------------------------------------------------
4494 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4496 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
4497 INTEROPERATE with older versions automatically. Please read the
4498 user documentation before upgrading any part of a live system.
4499 * WARNING: Lustre networking configuration changes are required with
4500 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
4503 * Support for kernels:
4504 2.6.9-22.0.2.EL (RHEL 4)
4505 2.6.5-7.244 (SLES 9)
4506 2.6.12.6 vanilla (kernel.org)
4509 Severity : enhancement
4510 Bugzilla : 7981/8208
4511 Description: Introduced Lustre Networking (LNET)
4512 Details : LNET is new networking infrastructure for Lustre, it includes
4513 a reorganized network configuration mode (see the user
4514 documentation for full details) as well as support for routing
4515 between different network fabrics. Lustre Networking Devices
4516 (LNDs) for the supported network fabrics have also been
4517 created for this new infrastructure.
4519 Severity : enhancement
4520 Description: Introduced Access control lists
4521 Details : clients can set ACLs on files and directories in order to have
4522 more fine-grained permissions than the standard Unix UGO+RWX.
4523 The MDS must be started with the "-o acl" mount option.
4525 Severity : enhancement
4526 Description: Introduced filesystem quotas
4527 Details : Administrators may now establish per-user quotas on the
4530 Severity : enhancement
4532 Description: Configuration change for the XT3
4533 The PTLLND is now used to run Lustre over Portals on the XT3
4534 The configure option(s) --with-cray-portals are no longer used.
4535 Rather --with-portals=<path-to-portals-includes> is used to
4536 enable building on the XT3. In addition to enable XT3 specific
4537 features the option --enable-cray-xt3 must be used.
4542 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
4543 Details : A race between allocating a new llcd and re-getting the llcd_lock
4544 allowed another thread to grab newly-allocated llcd.
4546 Severity : enhancement
4548 Description: 2.6 OST async journal commit and locking fix to improve performance
4549 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
4550 2.6 kernels are now async as they already were in 2.4 kernels so
4551 that they can commit concurrently with the network bulk transfer.
4552 For block-allocated files the filter allocation semaphore is held
4553 to avoid filesystem fragmentation during allocation. BKL lock
4554 removed for 2.6 xattr operations where it is no longer needed.
4559 Description: lconf incorrectly determined whether two IP networks could talk
4560 Details : In some more complicated routing and multiple-network
4561 configurations, lconf will avoid trying to make a network
4562 connection to a disjoint part of the IP space. It was doing the
4563 math incorrectly for one set of cases.
4568 Description: Fix for potential infinite loop processing records in an llog.
4569 Details : If an llog record is corrupted/zeroed, it is possible to loop
4570 forever in llog_process(). Validate the llog record length
4571 and skip the remainder of the block on error.
4574 Frequency : occasional (liblustre only)
4576 Description: liblustre could not open files whose last component is a symlink
4577 Details : sysio_path_walk() would incorrectly pass the open intent to
4578 intermediate path components.
4581 Frequency : rare (liblustre only with non-standard tuning)
4582 Bugzilla : 7201 (7350)
4583 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4584 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4585 after it has been released, only do this for referenced locks
4587 Severity : enhancement
4589 Description: specifying an (invalid) directory default stripe_size of -1
4590 would reset the directory default striping
4591 Details : stripe_size -1 was used internally to signal directory stripe
4592 removal, now use "all default" to signal dir stripe removal
4593 as a directory striping of "all default" is not useful
4596 Frequency : common for large clusters running liblustre clients
4598 Description: doing an ls when liblustre clients are running is slow
4599 Details : sending a glimpse AST to a liblustre client waits for every AST
4600 to time out, as liblustre clients will not respond. Since they
4601 cannot cache data we refresh the OST lock LVB from disk instead.
4603 Severity : enhancement
4605 Description: doing an ls at the same time as file IO can be slow
4606 Details : enqueue and other "small" requests can be blocked behind many
4607 large IO requests. Create a new OST IO portal for non-IO
4608 requests so they can be processed faster.
4611 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4613 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4614 created file with mode 000
4615 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4616 allow an ftruncate() as a non-root user to a file with mode 000.
4617 Lustre now respects this flag to disable mode checks when
4618 truncating a file owned by the user
4621 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4624 Description: Revoking locks from clients that went dead or catatonic might take
4626 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4627 cancellation of such locks instant on servers without waiting for
4628 any reply from clients. Clients drops these locks when cancel
4629 notification from server is received without replying.
4632 Frequency : liblustre-only, when liblustre client dies or becomes busy
4634 Description: Doing ls on Linux clients can take a long time with active
4636 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4637 granting such locks to it in the first place if possible. Locks
4638 are taken by proxy on the OST during the read or write and
4639 dropped immediately afterward. Add connect flags handling, do
4640 not grant locks to liblustre clients for glimpse ASTs.
4642 Severity : enhancement
4644 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4645 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4646 sequential reads and behaves badly with applications doing random
4647 reads. Improve it by reading ahead at least read region, and
4648 avoiding excessive large RPC for small reads.
4650 Severity : enhancement
4652 Description: Creating more than 1000 files for a single job may cause a load
4653 imbalance on the OSTs if there are also a large number of OSTs.
4654 Details : qos_prep_create() uses an OST index reseed value that is an
4655 even multiple of the number of available OSTs so that if the
4656 reseed happens in the middle of the object allocation it will
4657 still utilize the OSTs as uniformly as possible.
4662 Description: OST or MDS may oops in ping_evictor_main()
4663 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
4664 but doesn't restart at beginning of obd_exports_timed list
4667 Severity : enhancement
4669 Description: improve by-nid export eviction on the MDS and OST
4670 Details : allow multiple exports with the same NID to be evicted at one
4671 time without re-searching the exports list.
4674 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
4676 Description: MDS may oops in groups_free()
4677 Details : in rare race conditions a newly allocated group_info struct is
4678 freed again, and this can be NULL. The 2.4 compatibility code
4679 for groups_free() checked for a NULL pointer, but 2.6 did not.
4682 Frequency : common for liblustre clients doing little filesystem IO
4683 Bugzilla : 9352, 7313
4684 Description: server may evict liblustre clients accessing contended locks
4685 Details : if a client is granted a lock or receives a completion AST
4686 with a blocking AST already set it would not reply to the AST
4687 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
4690 Frequency : lfs setstripe, only systems with more than 160 OSTs
4692 Description: unable to set striping with a starting offset beyond OST 160
4693 Details : llapi_create_file() incorrectly limited the starting stripe
4694 index to the maximum single-file stripe count.
4697 Frequency : LDAP users only
4699 Description: lconf did not handle in-kernel recovery with LDAP properly
4700 Details : lconf/LustreDB get_refs() is searching the wrong namespace
4702 Severity : enhancement
4704 Description: bind OST threads to NUMA nodes to improve performance
4705 Details : all OST threads are uniformly bound to CPUs on a single NUMA
4706 node and do their allocations there to localize memory access
4708 Severity : enhancement
4710 Description: llmount can determine client NID directly from Myrinet (GM)
4711 Details : the client NID code from gmnalnid was moved directly into
4712 llmount, removing the need to use this or specifying the
4713 client NID explicitly when mounting GM clients with zeroconf
4716 Frequency : if client is started with down MDS
4718 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
4719 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
4720 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
4721 ptlrpc_import_delayed_req() because we didn't check if the
4722 request was interrupted, and we also didn't break out of the
4723 event loop if there was no timeout
4728 Description: data loss during non-page-aligned writes to a single file from
4729 both multiple nodes and multiple threads on one node at same time
4730 Details : updates to KMS and lsm weren't protected by common lock. Resulting
4731 inconsistency led to false short-reads, that were cached and later
4732 used by ->prepare_write() to fill in partially written page,
4733 leading to data loss.
4736 Frequency : always, if lconf --abort_recovery used
4738 Description: lconf --abort_recovery fails with 'Operation not supported'
4739 Details : lconf was attempting to abort recovery on the MDT device and not
4742 Severity : enhancement
4744 Description: remove cleanup logs
4745 Details : replace lconf-generated cleanup logs with lustre internal
4746 cleanup routines. Eliminates the need for client-cleanup and
4749 Severity : enhancement
4751 Description: add support for EAs (user and system) on lustre filesystems
4752 Details : it is now possible to store extended attributes in the Lustre
4753 client filesystem, and with the user_xattr mount option it
4754 is possible to allow users to store EAs on their files also
4756 Severity : enhancement
4758 Description: Add possibility (config option) to show minimal available OST free
4760 Details : When compiled with --enable-mindf configure option, statfs(2)
4761 (and so, df) will return least minimal free space available from
4762 all OSTs as amount of free space on FS, instead of summary of
4763 free spaces of all OSTs.
4765 Severity : enhancement
4767 Description: do not expand extent locks acquired on OST-side
4768 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
4769 by server: they are not cached anyway.
4772 Frequency : when mmap is used/binaries executed from Lustre
4774 Description: Unmmap pages before throwing them away from read cache.
4775 Details : llap_shrink cache now attempts to unmap pages before discarding
4776 them (if unmapping failed - do not discard). SLES9 kernel has
4777 extra checks that trigger if this unmapping is not done first.
4782 Description: lconf didn't resolve symlinks before checking to see whether a
4783 given mountpoint was already in use
4786 Frequency : when migrating failover services
4787 Bugzilla : 6395, 9514
4788 Description: When migrating a subset of services from a node (e.g. failback
4789 from a failover service node) the remaining services would
4790 time out and evict clients.
4791 Details : lconf --force (implied by --failover) sets the global obd_timeout
4792 to 5 seconds in order to quickly disconnect, but this caused
4793 other RPCs to time out too quickly. Do not change the global
4794 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
4796 Severity : enhancement
4797 Frequency : if MDS is started with down OST
4798 Bugzilla : 9439,5706
4799 Description: Allow startup/shutdown of an MDS without depending on the
4800 availability of the OSTs.
4801 Details : Asynchronously call mds_lov_synchronize during MDS startup.
4802 Add appropriate locking and lov-osc refcounts for safe
4803 cleaning. Add osc abort_inflight calls in case the
4804 synchronize never started.
4807 Frequency : occasional (Cray XT3 only)
4809 Description: root not authorized to access files in CRAY_PORTALS environment
4810 Details : The client process capabilities were not honoured on the MDS in
4811 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
4812 been accessed by an authorized user then root was able to access
4813 the file on the local client also. The root user capabilities
4814 are now allowed on the MDS, as this environment has secure UID.
4817 Frequency : occasional
4819 Description: ldiskfs "too long searching" message happens too often
4820 Details : A debugging message (otherwise harmless) prints too often on
4821 the OST console. This has been reduced to only happen when
4822 there are fragmentation problems on the filesystem.
4827 Description: Division by zero in statfs when all OSCs are inactive
4828 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
4829 lov_statfs divides by value returned by lov_get_stripecnt().
4833 Bugzilla : 9489, 3273
4834 Description: First write from each client to each OST was only 4kB in size,
4835 to initialize client writeback cache, which caused sub-optimal
4836 RPCs and poor layout on disk for the first written file.
4837 Details : Clients now request an initial cache grant at (re)connect time
4838 and so that they can start streaming writes to the cache right
4839 away and always do full-sized RPCs if there is enough data.
4840 If the OST is rebooted the client also re-establishes its grant
4841 so that client cached writes will be honoured under the grant.
4846 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
4847 Details : Now I/O RPCs go to different portal number and (presumably) fast
4848 lock requests (and glimses) and other RPCs get their own service
4849 threads pool that should be able to service those RPCs
4852 Severity : enhancement
4854 Description: Ability to exchange lustre version between client and servers and
4855 issue warnings at client side if client is too old. Also for
4856 liblustre clients there is ability to refuse connection of too old
4858 Details : New 'version' field is added to connect data structure that is
4859 filled with version info. That info is later checked by server and
4863 Frequency : rare, liblustre only.
4864 Bugzilla : 9296, 9581
4865 Description: Two simultaneous writes from liblustre at offset within same page
4866 might proceed at the same time overwriting eachother with stale
4868 Details : I/O lock withing llu_file_prwv was released too early, before data
4869 actually was hitting the wire. Extended lock-holding time until
4870 server acknowledges receiving data.
4873 Frequency : extremely rare. Never observed in practice.
4875 Description: avoid generating lustre_handle cookie of 0.
4876 Details : class_handle_hash() generates handle cookies by incrementing
4877 global counter, and can hit 0 occasionaly (this is unlikely, but
4878 not impossible, because initial value of cookie counter is
4879 selected randonly). Value of 0 is used as a sentinel meaning
4880 "unassigned handle" --- avoid it. Also coalesce two critical
4881 sections in this function into one.
4883 Severity : enhancement
4885 Description: allow liblustre clients to delegate truncate locking to OST
4886 Details : To avoid overhead of locking, liblustre client instructs OST to
4887 take extent lock in ost_punch() on client's behalf. New connection
4888 flag is added to handle backward compatibility.
4890 Severity : enhancement
4891 Bugzilla : 4928, 7341, 9758
4892 Description: allow number of OST service threads to be specified
4893 Details : a module parameter allows the number of OST service threads
4894 to be specified via "options ost ost_num_threads={N}" in the
4895 OSS's /etc/modules.conf or /etc/modprobe.conf.
4899 Bugzilla : 6146, 9635, 9895
4900 Description: servers crash with bad pointer in target_handle_connect()
4901 Details : In rare cases when a client is reconnecting it was possible that
4902 the connection request was the last reference for that export.
4903 We would temporarily drop the export reference and get a new
4904 one, but this may have been the last reference and the export
4905 was just destroyed. Get new reference before dropping old one.
4907 Severity : enhancement
4908 Frequency : if client is started with failover MDS
4910 Description: Allow multiple MDS hostnames in the mount command
4911 Details : Try to read the configuration from all specified MDS
4912 hostnames during a client mount in case the "primary"
4915 Severity : enhancement
4917 Description: Stop sending data to evicted clients as soon as possible.
4918 Details : Check if the client we are about to send or are sending data to
4919 was evicted already. (Check is done every second of waiting,
4920 for which l_wait_event interface was extended to allow checking
4921 of exit condition at specified intervals).
4924 Frequency : rare, normally only when NFS exporting is done from client
4926 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
4928 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
4931 Frequency : always, if srand() is called before liblustre initialization
4933 Description: Liblustre uses system PRNG disturbing its usage by user application
4934 Details : Introduce internal to lustre fast and high-quality PRNG for
4935 lustre usage and make liblustre and some other places in generic
4936 lustre code to use it.
4938 Severity : enhancement
4939 Bugzilla : 9477, 9557, 9870
4940 Description: Verify that the MDS configuration logs are updated when xml is
4941 Details : Check if the .xml configuration logs are newer than the config
4942 logs stored on the MDS and report an error if this is the case.
4943 Request --write-conf, or allow starting with --old_conf.
4945 Severity : enhancement
4947 Description: Handle symlinks in the path when checking if Lustre is mounted.
4948 Details : Resolve intermediate symlinks when checking if a client has
4949 mounted a filesystem to avoid duplicate client mounts.
4954 Description: lconf can hit an error exception but still return success.
4955 Details : The lconf command catches the Command error exception at the top
4956 level script context and will exit with the associated exit
4957 status, but doesn't ensure that this exit status is non-zero.
4962 Description: failure of ptlrpc thread startup can cause oops
4963 Details : Starting a ptlrpc service thread can fail if there are a large
4964 number of threads or the server memory is very fragmented.
4965 Handle this without oopsing.
4968 Frequency : always, only if liblustre and non-default acceptor port was used
4970 Description: liblustre cannot connect to servers with non-default acceptor port
4971 Details : tcpnal_set_default_params() was not called and was therefore
4972 ignoring the environment varaible TCPNAL_PORT, as well as other
4973 TCPNAL_ environment variables
4978 Description: two objects could be created on the same OST for a single file
4979 Details : If an OST is down, in some cases it was possible to create two
4980 objects on a single OST for a single file. No problems other
4981 than potential performance impact and spurious error messages.
4985 Bugzilla : 5681, 9562
4986 Description: Client may oops in ll_unhash_aliases
4987 Details : Client dcache may become inconsistent in race condition.
4988 In some cases "getcwd" can fail if the current directory is
4994 Description: Inode refcounting problems in NFS export code
4995 Details : link_raw functions used to call d_instantiate without obtaining
4996 extra inode reference first.
5000 Bugzilla : 9942, 9903
5001 Description: Referencing freed requests leading to crash, memleaks with NFS.
5002 Details : We used to require that call to ll_revalidate_it was always
5003 followed by ll_lookup_it. Also with revalidate_special() it is
5004 possible to call ll_revalidate_it() twice for the same dentry
5005 even if first occurence returned success. This fix changes semantic
5006 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
5007 reference on a request referred from the intent.
5008 ll_intent_release() then releases such a request.
5011 Frequency : rare, normally benchmark loads only
5013 Description: unlinked inodes were kept in memory on the client
5014 Details : If a client is repeatedly creating and unlinking files it
5015 can accumulate a lot of stale inodes in the inode slab cache.
5016 If there is no other client load running this can cause the
5017 client node to run out of memory. Instead flush old inodes
5018 from client cache that have the same inode number as a new inode.
5021 Frequency : SLES9 2.6.5 kernel and long filenames only
5022 Bugzilla : 9969, 10379
5023 Description: utime reports stale NFS file handle
5024 Details : SLES9 uses out-of-dentry names in some cases, which confused
5025 the lustre dentry revalidation. Change it to always use the
5029 Frequency : rare, unless heavy write-truncate concurrency is continuous
5030 Bugzilla : 4180, 6984, 7171, 9963, 9331
5031 Description: OST becomes very slow and/or deadlocked during object unlink
5032 Details : filter_destroy() was holding onto the parent directory lock
5033 while truncating+unlinking objects. For very large objects this
5034 may block other threads for a long time and slow overall OST
5035 responsiveness. It may also be possible to get a lock ordering
5036 deadlock in this case, or run out of journal credits because of
5037 the combined truncate+unlink. Solution is to do object truncate
5038 first in one transaction without parent lock, and then do the
5039 final unlink in a new transaction with the parent lock. This
5040 reduces the lock hold time dramatically.
5043 Frequency : rare, 2.4 kernels only
5045 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
5046 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
5047 before unmounting their filesystem in order to avoid deadlock.
5048 The kernel_locked() function in 2.4 kernels only checks whether
5049 the kernel lock is held, not whether it is this process that is
5050 holding it as 2.6 kernels do.
5055 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
5056 Details : The client ptlrpc code may be trying to reconnect to a down
5057 server before a previous connection attempt has timed out.
5058 Increase the reconnect interval to be longer than the connection
5059 timeout interval to avoid sending duplicate connections to
5063 Frequency : echo_client brw_test command
5065 Description: fix echo_client to work with OST preallocated code
5066 Details : OST preallocation code (5137) didn't take echo_client IO path
5067 into account: echo_client calls filter methods outside of any
5068 OST thread and, hence, there is no per-thread preallocated
5069 pages and buffers to use. Solution: hijack pga pages for IO. As
5070 a byproduct, this avoids unnecessary data copying.
5074 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
5075 Description: Client can oops in mdc_commit_close() after open replay
5076 Details : It was possible for the MDS to return an open request with no
5077 transaction number in mds_finish_transno() if the client was
5078 evicted, but without actually returning an error. Clients
5079 would later try to replay that open and may trip an assertion
5080 Simplify the client close codepath, and always return an error
5081 from the MDS in case the open is not successful.
5084 Frequency : rare, 2.6 OSTs only
5086 Description: OST may deadlock under high load on fragmented files
5087 Details : If there was a heavy load and highly-fragmented OST filesystems
5088 it was possible to have all the OST threads deadlock waiting on
5089 allocation of biovecs, because the biovecs were not released
5090 until the entire RPC IO was completed. Instead, release biovecs
5091 as soon as they are complete to ensure forward IO progress.
5093 Severity : enhancement
5095 Description: Support for specifying external journal device at mount
5096 Details : If an OST or MDS device is formatted with an external journal
5097 device, this device major/minor is stored in the ext3 superblock
5098 and may not be valid for failover. Allow detecting and
5099 specifying the external journal at mount time.
5104 Description: Mounting an MDS with pending unlinked files may cause oops
5105 Details : target_finish_recovery() calls mds_postrecov() which returned
5106 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
5107 considers this an error and immediately begins cleaning up the
5108 lov, just after starting the mds_lov process
5110 Severity : enhancement
5112 Description: Implement 'lfs df' to report actual free space on per-OST basis
5113 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
5114 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
5115 usage of objects; '-h' to report in human readable format.
5117 ------------------------------------------------------------------------------
5119 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
5126 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
5127 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
5128 at startup during orphan unlinking. Instead, skip buddy update
5129 before setup, as it will be regenerated after recovery is complete.
5134 Description: If an OST is inactive, its locks might reference stale inodes.
5135 Details : lov_change_cbdata() must iterate over all namespaces, even if
5136 they are inactive to clear inode references from the lock.
5138 Severity : enhancement
5139 Frequency : occasional, if non-standard max_dirty_mb used
5141 Description: Client will block write RPCs if not enough grant
5142 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
5143 then the client will block writes while waiting for another RPC
5144 to complete instead of consuming its dirty limit. With change
5145 we get improved performance when max_dirty_mb is small.
5147 Severity : enhancement
5148 Bugzilla : 3389, 6253
5149 Description: Add support for supplementary groups on the MDS.
5150 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
5151 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
5152 lookups for user supplementary groups into a cache.
5156 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
5157 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
5158 for liblustre O_EXCL is enforced by MDS.
5162 Description: readdir never returns NULL in liblustre.
5163 Details : Corrected llu_iop_getdirentries logic, to return offset of next
5164 dentry in struct dirent.
5168 Frequency : liblustre only, depends on application IO pattern
5169 Description: liblustre clients evicted if not contacting servers
5170 Details : Don't put liblustre clients into the ping_evictor list, so
5171 they will not be evicted by the pinger ever.
5173 Severity : enhancement
5175 Description: Add ability to evict clients by NID from MDS.
5176 Details : By echoing "nid:$NID" string into
5177 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
5178 $NID would be instantly evicted from this MDS and from all active
5179 OSTs connected to it.
5183 Description: Do not query file size twice, somewhat slowing stat(2) calls.
5184 Details : lookup_it_finish() used to query file size from OSTs that was not
5189 Description: service threads change working directory to that of init
5190 Details : Starting lustre service threads may pin the working directory
5191 of the parent thread, making that filesystem busy. Threads
5192 now change to the working directory of init to avoid this.
5196 Frequency : during shutdown only
5197 Description: shutdown with a failed MDS or OST can cause unmount to hang
5198 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
5203 Frequency : 2.6 only
5204 Description: chmod/chown may include an extra supplementary group
5205 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
5206 supplementary group and if none is specified this is used.
5209 Bugzilla : 5479 (6816)
5211 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
5212 Details : If lookup is in progress on a file that is unlinked we might try
5213 to revalidate the inode and fail in revalidate after lookup is
5214 complete and ll_file_open() enqueues the open again but
5215 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
5218 Frequency : always, if lconf --abort_recovery used
5220 Description: lconf --abort_recovery fails with 'Operation not supported'
5221 Details : lconf was attempting to abort recovery on the MDT device and not
5224 ------------------------------------------------------------------------------
5226 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
5231 Frequency : rare (only unsupported configurations with a node running as an
5233 Bugzilla : 6514, 5137
5234 Description: Mounting a Lustre file system on a node running as an OST could
5236 Details : OSTs now preallocates memory needed to write out data at
5237 startup, instead of when needed, to avoid having to
5238 allocate memory in possibly low memory situations.
5239 Specifically, if the file system is mounted on on OST,
5240 memory pressure could force it to try to write out data,
5241 which it needed to allocate memory to do. Due to the low
5242 memory, it would be unable to do so and the node would
5243 become unresponsive.
5245 Severity : enhancement
5247 Description: Addition of lconf --service command line option
5248 Details : lconf now accepts a '--service <arg>' option, which is
5249 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
5251 Severity : enhancement
5253 Description: Failover mode is now the default for OSTs.
5254 Details : By default, OSTs will now run in failover mode. To return to
5255 the old behaviour, add '--failout' to the lmc line for OSTs.
5257 Severity : enhancement
5259 Description: Health checks are now provided for MDS and OSTs
5260 Details : Additional detailed health check information on MSD and OSTs
5261 is now provided through the procfs health_check value.
5264 Frequency : occasional, depends on IO load
5266 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
5267 after numerous create/delete cycles
5268 Details : The ext3 inode allocation policy would not allocate new inodes
5269 very well on the OSTs because there are no new directories
5270 being created. Instead we look for groups with free space if
5271 the parent directories are nearly full.
5276 Description: Network or server problems during mount may cause partially
5277 mounted clients instead of returning an error.
5278 Details : The config llog parsing code may overwrite the error return
5279 code during mount error handling, returning success instead
5285 Description: MDS can fail to allocate large reply buffers
5286 Details : After long uptimes the MDS can fail to allocate large reply
5287 buffers (e.g. zconf client mount config records) due to memory
5288 fragmentation or consumption by the buffer cache. Preallocate
5289 some large reply buffers so that these replies can be sent even
5290 under memory pressure.
5294 Frequency : rare (liblustre)
5295 Description: fsx running with liblustre complained that using truncate() to
5296 extend the file doesn't work. This patch corrects that issue.
5297 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
5298 ATTR_SIZE and lsm use in llu_setattr_raw.
5302 Frequency : rare, only 2.6 kernels
5303 Description: Unusual file access patterns on the MDS may result in inode
5304 data being lost in very rare circumstances.
5305 Details : Bad interaction between the ea-in-inode patch and the "no-read"
5306 code in the 2.6 kernel caused the inode and/or EA data not to
5307 be read from disk, causing single-file corruption.
5311 Frequency : rare, only 2.6 filesystems using extents
5312 Description: Heavy concurrent write and delete load may cause data corruption.
5313 Details : It was possible under high-load situations to have an extent
5314 metadata block in the block device cache from a just-unlinked
5315 file overwrite a newly-allocated data block. We now unmap any
5316 metadata buffers that alias just-allocated data blocks.
5320 Frequency : filesystems with default stripe_count larger than 77
5321 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
5322 Details : lconf specifies an inode size of 4096 bytes when the default
5323 stripe_count is larger than 77. This conflicts with the default
5324 inode density of 1 per 4096 bytes. Allocate smaller inodes in
5325 this case to avoid pinning too much memory for large EAs.
5327 ------------------------------------------------------------------------------
5329 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
5334 Frequency : rare (extremely heavy IO load with hundreds of clients)
5336 Description: Client is evicted, gets IO error writing to file
5337 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
5338 caused clients to be evicted for AST timeouts. The fixes in
5339 bug 5192 mean we no longer need to have such short AST timeouts
5340 so ldlm_timeout has been increased.
5343 Frequency : occasional during --force or --failover shutdown under load
5344 Bugzilla : 5949, 4834
5345 Description: Server oops/LBUG if stopped with --force or --failover under load
5346 Details : a collection of import/export refcount and cleanup ordering
5347 issues fixed for safer force cleanup
5350 Frequency : only filesystems larger than 120 OSTs
5351 Bugzilla : 5990, 6223
5352 Description: lfs getstripe would oops on a very large filesystem
5353 Details : lov_getconfig used kfree on vmalloc'd memory
5356 Frequency : only filesystems exporting via NFS to Solaris 10 clients
5357 Bugzilla : 6242, 6243
5358 Description: reading from files that had been truncated to a non-zero size
5359 but never opened returned no data
5360 Details : ll_file_read() reads zeros from no-object files to EOF
5365 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
5366 Details : There's little harm in aborting MDS/OSS recovery and letting it
5367 try again, so I removed the LASSERT and return an error instead.
5369 Severity : enhancement
5371 Description: New debugging infrastructure for tracking down data corruption
5372 Details : The I/O checksum code was replaced to: (a) control it at runtime,
5373 (b) cover more of the client-side code path, and (c) try to narrow
5374 down where problems occurred
5378 Bugzilla : 3819, 4364, 4397, 6313
5379 Description: Racing close and eviction MDS could cause assertion in mds_close
5380 Details : It was possible to get multiple mfd references during close and
5381 client eviction, leading to one thread referencing a freed mfd.
5383 Severity: : enhancement
5384 Bugzilla : 3262, 6359
5385 Description: Attempts to reconnect to servers are now more aggressive.
5386 Details : This builds on the enhanced upcall-less recovery that was added
5387 in 1.4.2. When trying to reconnect to servers, clients will
5388 now try each server in the failover group every 10 seconds. By
5389 default, clients would previously try one server every 25 seconds.
5394 Description: After recovery, certain operations trigger a failed
5395 assertion on a client.
5396 Details : Failing over an mds, using lconf -d --failover, while a
5397 client was doing a readdir() call would cause the client to
5398 LBUG after recovery completed and the readdir() was resent.
5400 Severity : enhancement
5402 Description: Default groups are now added by lconf
5403 Details : You can now run lconf --group <servicename> without having to
5404 manually add groups with lmc.
5407 Frequency : occasional
5409 Description: Nodes with an elan id of 0 trigger a failed assertion
5412 Frequency : always when accessing e.g. tty/console device nodes
5414 Description: tty and some other devices nodes cannot be used on lustre
5415 Details : file's private_data field is used by device data and lustre
5416 values in there got lost. New field was added to struct file to
5417 store fs-specific private data.
5420 Frequency : when exporting Lustre via NFS
5422 Description: NFSD failed occasionally when looking up a path component
5423 Details : NFSD is looking up ".." which was broken in ext3 directories
5424 that had grown large enough to become hashed.
5427 Frequency : Clusters with multiple interfaces not on the same subnet
5429 Description: Nodes will repeatedly try to reconnect to an interface which it
5430 cannot reach and report an error to the log.
5431 Details : Extra peer list entries will be created by lconf with some peers
5432 unreachable. lconf now validates the peer before adding it.
5435 Frequency : Only if a default stripe is set on the filesystem root.
5437 Description: Setting a default stripe on the filesystem root prevented the
5438 filesystem from being remounted.
5439 Details : The client was sending extra request flags in the root getattr
5440 request and did not allocate a reply buffer for the dir EA.
5443 Frequency : occasional, higher if lots of files are accessed by one client
5444 Bugzilla : 6159, 6097
5445 Description: Client trips assertion regarding lsm mismatch/magic
5446 Details : While revalidating inodes the VFS looks up inodes with ifind()
5447 and in rare cases can find an inode that is being freed.
5448 The ll_test_inode() code will free the lsm during ifind()
5449 when it finds an existing inode and then the VFS later attaches
5450 this free lsm to a new inode.
5454 Bugzilla : 6422, 7030
5455 Description: MDS deadlock between mkdir and client eviction
5456 Details : Creating a new file via mkdir or mknod (starting a transaction
5457 and getting the ns lock) can deadlock with client eviction
5458 (gets ns lock and trying to finish a synchronous transaction).
5461 Frequency : occasional
5462 Description: While starting a server, the fsfilt_ext3 module could not be
5464 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
5465 kernels. Previously, lconf would still use the ext3 name
5466 when trying to load modules. Now, it will correctly use
5467 ext3 on 2.4 and ldiskfs on 2.6.
5469 Severity : enhancement
5470 Description: The default stripe count has been changed to 1
5471 Details : The interpretation of the default stripe count (0, to lfs
5472 or lmc) has been changed to mean striping across a single
5473 OST, rather than all available. For general usage we have
5474 found a stripe count of 1 or 2 works best.
5476 Severity : enhancement
5477 Description: Add support for compiling against Cray portals.
5478 Details : Conditional compiling for some areas that are different
5482 Frequency : occasional
5483 Bugzilla : 6409, 6834
5484 Description: Creating files with an explicit stripe count may lead to
5485 a failed assertion on the MDS
5486 Details : If some OSTs are full or unavailable, creating files may
5487 trigger a failed assertion on the MDS. Now, Lustre will
5488 try to use other servers or return an error to the
5492 Frequency : occasional
5494 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
5495 caused lock timeout during readahead (since 1.4.2).
5496 Details : Processes doing readahead might match a lock that hasn't been
5497 granted yet if there are overlapping and conflicting lock
5498 requests. The readahead process waits on ungranted lock
5499 (original lock is CBPENDING), while OST waits for that process
5500 to cancel CBPENDING read lock and eventually evicts client.
5502 Severity : enhancement
5504 Description: Initial enabling of flock support for clients
5505 Details : Implements fcntl advisory locking and file status functions.
5506 This feature is provided as an optional mount flag (default
5507 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
5508 locking are implemented yet, and those that are are not guaranteed
5509 to be completely correct in production environments.
5510 mount -t lustre -o [flock|noflock] ...
5513 Frequency : occasional
5515 Description: OSTs running 2.4 kernels but with extents enabled might trip an
5516 assertion in the ext3 JBD (journaling) layer.
5517 Details : The b_committed_data struct is protected by the big kernel lock
5518 in 2.4 kernels, serializing journal_commit_transaction() and
5519 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
5520 access to this struct. In 2.6 kernels there is finer grained
5521 locking to improve SMP performance of the JBD layer.
5525 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
5530 Description: merge and cleanup kernel patches.
5533 -----------------------------------------------------------------------------
5535 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
5537 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
5538 versions of Lustre. Please update all servers and clients to
5539 version 1.4.2 or later at the same time. You must also run
5540 "lconf --write-conf {config}.xml" on the MDS while it is stopped
5541 to update the configuration logs.
5543 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
5544 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
5545 - swab llog records of type '0' so we get proper header size/idx (5861)
5546 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
5547 - fix rename of one directory over another leaking an inode (5953)
5548 - avoid SetPageDirty on 2.6 (5981)
5549 - don't re-add just-being-destroyed locks to the waiting list (5653)
5550 - when creating new directories, inherit the parent's custom
5551 striping settings if present parent (3048)
5552 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
5553 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
5554 - don't hold client locks on temporary worklist from l_lru (5666)
5555 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
5556 - automatically evict dead clients (5921)
5557 - Update file size properly in create+truncate+fstat case (6196)
5558 - Do not unhash mountpoint dentries, do not allow removal of
5560 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
5561 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
5562 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
5563 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
5564 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
5565 - Correctly handle reads of files with no objects (6243)
5566 - lctl recover will also mark a device active if deactivate used (5933)
5568 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
5569 - allow --write-conf on an MDS with different nettype than client (5619)
5570 - don't write config llogs to MDS for mounts not from that MDS (5617)
5571 - lconf should create multiple TCP connections from a client (5201)
5572 - init scripts are now turned off by default; run chkconfig --on
5573 lustre and chkconfig --on lustrefs to use them
5574 - upcalls are no longer needed for clients to recover to failover
5576 - add --abort-recovery option to lconf to abort recovery on device
5578 - add support for an arbitrary number of OSTs (3026)
5579 - Quota support protocol changes.
5580 - forward compatibility changes to wire structs (6007)
5581 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5582 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5583 - improved "lustre status" script
5584 - initialize blocksize for non-regular files (6062)
5585 - added --disable-server and --disable-client configure options (5782)
5586 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5587 - Vanilla 2.4.29 support
5588 - increase maximum number of obd devices to 520 (6242)
5589 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5590 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5591 SLES 9 2.6.5 kernels
5592 - limit stripes per file to 160 (the maximum EA size) (6093)
5594 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5597 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5598 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5599 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5600 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5601 - small ext3 extents cleanups and fixes (5733)
5602 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5603 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5604 - avoid needless client->OST connect, fix handle mismatch (5317)
5605 - fix DLM error path that led to out-of-sync client, long delays (5779)
5606 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5607 - fix several locking issues related to i_size (5492,5624,5654,5672)
5608 - don't move pending lock onto export if it is already evicted (5683)
5609 - fix kernel oops when creating .foo in unlinked directory (5548)
5610 - fix deadlock in obdfilter statistics vs. object create (5811)
5611 - use time_{before,after} to avoid timer jiffies wrap (5882)
5612 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5613 - Do not leak request if server was not able to process it (5154)
5614 - If mds_open unable to find parent dir, make that negative lookup(5154)
5615 - don't create new directories with extent-mapping (5909, 5936)
5617 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5618 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5619 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5620 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5621 - support for RHEL4 kernel on i686 (5773)
5622 - provide error messages when incompatible logs are encountered (5898)
5624 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5625 * version 1.4.0.10 (1.4.1 release candidate 1)
5627 - don't keep a lock reference when lock is not granted (4238)
5628 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5629 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5630 - drop import inflight refcount on signal_completed_replay error (5255)
5631 - unlock page after async write error during send (3677)
5632 - handle missing objects in filter_preprw_read properly (5265)
5633 - no transno return for symlink open, don't save no-trasno open (3440)
5634 - don't try to complete elan receive that already failed (4012)
5635 - free RPC server reply state on error (5406)
5636 - clean up thread from ptlrpc_start_thread() on error (5160)
5637 - readahead could read extra page into cache that wasn't ejected (5388)
5638 - prevent races in class_attach/setup/cleanup/detach (5260)
5639 - don't dereference de->d_inode after l_dput of de (5458)
5640 - use "int" for stripe value returned from lock_to_stripe (5544)
5641 - mballoc allocation and error-checking fixes in 2.6 (5504)
5642 - block device patches to fix I/O request sizes in 2.6 (5482)
5643 - look up hostnames for IB nals (5602)
5644 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5645 - don't start multiple acceptors for the same port (5277)
5646 - fix incorrect LASSERT in mds_getattr_name (5635)
5647 - export a proc file for general "ping" checking (5628)
5648 - fix "lfs check" to not block when the MDS is down (5628)
5650 - service request history (4965)
5651 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5652 - create an "evict_client" /proc entry on OSTs, like the MDS has
5653 - fix mount usage message, return errors per mount(8) (5168)
5654 - change grep [] to grep "[]" in tests so they work in more UMLs
5655 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5656 - remove ancient LOV_MAGIC_V0 EA support (5047)
5657 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5658 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5659 - allow readahead allocations to fail when low on memory (5383)
5660 - mmap locking landed again, after considerable improvement (2828)
5661 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
5663 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
5666 - send OST transaction number in read/write reply to free req (4966)
5667 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5668 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5669 - the watchdog thread now runs as interruptible (5246)
5670 - flock/lockf fixes (but it's still disabled, pending 5135)
5671 - don't use EXT3 constants in llite code (5094)
5672 - memory shortage at startup could cause assertion (5176)
5674 - reorganization of lov code
5675 - single portals codebase
5677 - add extents/mballoc support (5025)
5678 - direct I/O reads in the obdfilter (4048)
5679 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
5681 tbd Cluster File Systems, Inc. <info@clusterfs.com>
5684 - send OST transaction number in read/write reply to free req (4966)
5685 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5686 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5687 - the watchdog thread now runs as interruptible (5246)
5688 - handle missing objects in filter_preprw_read properly (5265)
5689 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5690 - drop import inflight refcount on signal_completed_replay error (5255)
5691 - unlock page after async write error during send (3677)
5692 - return original error code on reconstructed replies (3761)
5693 - no transno return for symlink open, don't save no-trasno open (3440)
5695 - add pid to ldlm debugging output (4922)
5696 - bump the watchdog timeouts -- we can't handle 30sec yet
5697 - extra debugging for orphan dentry/inode bug (5259)
5699 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
5702 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
5703 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
5704 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
5705 - fix NULL dereference in /proc/sys/portals/routes (4827)
5706 - allow failed mdc_close() operations to be interrupted (4561)
5707 - stop precreate on OST before MDS would time out on it (4778)
5708 - don't send partial-page writes before EOF from client (4410)
5709 - discard client grant for sub-page writes on large-page clients (4520)
5710 - don't free dentries not owned by NFS code, check generation (4806)
5711 - fix lsm leak if mds_create_objects() fails (4801)
5712 - limit debug_daemon file size, always print CERROR messages (4789)
5713 - use transno after validating reply (3892)
5714 - process timed out requests if import state changes (3754)
5715 - update mtime on OST during writes, return in glimpse (4829)
5716 - add mkfsoptions to LDAP (4679)
5717 - use ->max_readahead method instead of zapping global ra (5039)
5718 - don't interrupt __l_wait_event() during strace
5720 - add software watchdogs to catch hung threads quickly (4941)
5721 - make lustrefs init script start after nfs is mounted
5722 - fix CWARN/ERROR duplication (4930)
5723 - return async write errors to application if possible (2248)
5724 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
5725 - print NAL number in %x format (4645)
5726 - update barely-supported suse-2.4.21-171 series (4842)
5727 - support for sles 9 %post scripts
5728 - support for building 2.6 kernel-source packages
5729 - support for sles km_* packages
5731 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
5734 - ignore -ENOENT errors in osc_destroy (3639)
5735 - notify osc create thread that OSC is being cleaned up (4600)
5736 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
5737 - reconstruct ost_handle() like mds_handle() (4657)
5738 - create a new thread to do import eviction to avoid deadlock (3969)
5739 - let lconf resolve symlinked-to devices (4629)
5740 - don't unlink "objects" from directory with default EA (4554)
5741 - hold socknal file ref over connect in case target is down (4394)
5742 - allow more than 32000 subdirectories in a single directory (3244)
5743 - fix blocks count for O_DIRECT writes (3751)
5744 - OST returns ENOSPC from object create when no space left (4539)
5745 - don't send truncate RPC if file size isn't changing (4410)
5746 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
5747 - bind to privileged port in socknal and tcpnal (3689)
5749 - rate limit CERROR/CWARN console message to avoid overload (4519)
5750 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
5751 - basic mmap support (3918)
5752 - kernel patch series update from b1_4 (4711)
5754 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
5757 - avoid crash during MDS cleanup with OST shut down (2775)
5758 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
5759 - don't use bad inodes on the MDS (3744)
5760 - dynamic object preallocation to improve recovery speed (4236)
5761 - don't hold spinlock over lock dumping or change debug flags (4401)
5762 - don't zero obd_dev when it is force cleaned (3651)
5763 - print grants to console if they go negative (4431)
5764 - "lctl deactivate" will stop automatic recovery attempts (3406)
5765 - look for existing locks in ldlm_handle_enqueue() (3764)
5766 - don't resolve lock handle twice in recovery avoiding race (4401)
5767 - revalidate should check working dir is a directory (4134)
5769 - don't always mark "slow" obdfilter messages as errors (4418)
5771 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
5774 - don't close LustreDB during write_conf until it is done (3860)
5775 - fix typo in lconf for_each_profile (3821)
5776 - allow dumping logs from multiple threads at one time (3820)
5777 - don't allow multiple threads in OSC recovery (3812)
5778 - fix debug_size parameters (3864)
5779 - fix mds_postrecov to initialize import for llog ctxt (3121)
5780 - replace config semaphore with spinlock (3306)
5781 - be sure to send a reply for a CANCEL rpc with bad export (3863)
5782 - don't allow enqueue to complete on a destroyed export (3822)
5783 - down write_lock before checking llog header bitmap (3825)
5784 - recover from lock replay timeout (3764)
5785 - up llog sem before sending rpc (3652)
5786 - reduce ns lock hold times when setting kms (3267)
5787 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
5788 - fix NULL deref and obd_dev leak on setup error (3312)
5789 - replace some LBUG about llog ops with error handling (3841)
5790 - don't match INVALID dentries from d_lookup and spin (3784)
5791 - hold dcache_lock while marking dentries INVALID and hashing (4255)
5792 - fix invalid assertion in ptlrpc_set_wait (3880)
5794 - add libwrap support for the TCP acceptor (3996)
5795 - add /proc/sys/portals/routes for non-root route listing (3994)
5796 - allow setting MDS UUID in .xml (2580)
5797 - print the stack of a process that LBUGs (4228)
5799 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
5802 - don't cleanup request in ll_file_open() on failed MDS open (3430)
5803 - make sure to unset replay flag from failed open requests (3440)
5804 - if default stripe count is 0, use OST count for inode size (3636)
5805 - update parent mtime/ctime on client for create/unlink (2611)
5806 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
5807 - free recovery state on server during a forced cleanup (3571)
5808 - unregister_reply for resent reqs (3063)
5809 - loop back devices mounting and status check on 2.6 (3563)
5810 - fix resource-creation race that can provoke i_size == 0 (3513)
5811 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
5812 - more debugging for page-accounting assertion (3746)
5813 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
5814 - avoid deadlock after precreation failure (3758)
5815 - fix race and lock order deadlock in orphan handling (3450, 3750)
5816 - add validity checks when grabbing inodes from l_ast_data (3599)
5818 - add /proc/.../recovery_status to obdfilter (3428)
5819 - lightweight CDEBUG infrastructure, debug daemon (3668)
5820 - change default OSC RPC parameters to be better on small clusters
5821 - turn off OST read cache for files smaller than 32MB
5822 - install man pages and include them in rpms (3100)
5823 - add new init script for (un)mounting lustre filesystems (2593)
5824 - run chkconfig in %post for init scripts (3701)
5825 - drop scimac NAL (unmaintained)
5827 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
5830 - clean kiobufs before and after use (3485)
5831 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
5832 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
5833 - change init script to fail more clearly if not run as root (1528)
5834 - allow clients to reconnect during replay (1742)
5835 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
5836 - don't do DNS lookups on NIDs too small for IP addresses (3442)
5837 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
5838 - fix cond_resched (3554)
5839 - only evict unfinished clients after recovery (3515)
5840 - allow bulk resend, prevent data loss (3570)
5841 - dynamic ptlrpc request buffer allocation (2102)
5842 - don't allow unlinking open directory if it isn't empty (2904)
5843 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
5844 - remove extraneous obd dereference causing LASSERT failure (3334)
5845 - don't use get_cycles() when creating temp. files on the mds (3156)
5846 - hold i_sem when setting i_size in ll_extent_lock() (3564)
5847 - handle EEXIST for set-stripe, set proper directory name (3336)
5849 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
5850 - fix ksocknal_fmb_callback() error messages (2918)
5852 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
5855 - don't copy lvb into (possibly NULL) reply on error (2983)
5856 - don't deref dentry after dput, don't free lvb on error (2922)
5857 - use the kms to determine writeback rpc length (2947)
5858 - increment oti_logcookies when osc is inactive (2948)
5859 - update client's i_blocks count via lvb messages (2543)
5860 - handle intent open/close of special files properly (1557)
5861 - mount MDS with errors=remount-ro, like obdfilter (2009)
5862 - initialize lock handle to avoid ASSERT on error cleanup (3057)
5863 - don't use cancelling-locks' kms values (2947)
5864 - use highest lock extent for kms, not last one (2925)
5865 - don't dereference ERR_PTR() dentry in error handling path (3107)
5866 - fix thread race in portals_debug_dumplog() (3122)
5867 - create lprocfs device entries at setup instead of at attach (1519)
5868 - common AST error handler, don't evict client on completion race (3145)
5869 - zero nameidata in detach_mnt in 2.6 (3118)
5870 - verify d_inode after revalidate_special is valid in 2.6 (3116)
5871 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
5872 - initialize RPC timeout timer earlier for 2.6 (3219)
5873 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
5874 - print nal/nid for unknown nid (3258)
5875 - additional checks for oscc recovery before doing precreate (3284)
5876 - fix ll_extent_lock() error return code for 64-bit systems (3043)
5877 - don't crash in mdc_close for bad permissions on open (3285)
5878 - zero i_rdev for non-device files (3147)
5879 - clear page->private before handing to FS, better assertion (3119)
5880 - tune the read pipeline (3236)
5881 - fix incorrect decref of invalidated dentry (2350)
5882 - provide read-ahead stats and refine rpc in flight stats (3328)
5883 - don't hold journal transaction open across create RPC (3313)
5884 - update atime on MDS at close time (3265)
5885 - close LDAP connection when recovering to avoid server load (3315)
5886 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
5887 - don't leak open file on MDS after open resend (3325)
5888 - serialize filter_precreate and filter_destroy_precreated (3329)
5889 - loop device shouldn't call sync_dev() for nul device (3092)
5890 - clear page cache after eviction (2766)
5891 - resynchronize MDS->OST in background (2824)
5892 - refuse to mount the same filesystem twice on same mountpoint (3394)
5893 - allow llmount to create routes for mounting behind routers (3320)
5894 - push lock cancellation to blocking thread for glimpse ASTs (3409)
5895 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
5896 - fix rare problem with rename on htree directories (3417)
5898 - allow default OST striping configuration per directory (1414)
5899 - fix compilation for qswnal for 2.6 kernels (3125)
5900 - increase maximum number of MDS request buffers for large systems
5901 - change liblustreapi to be useful for external progs like lfsck (3098)
5902 - increase local configuration timeout for slow disks (3353)
5903 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
5905 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5908 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
5909 - don't overwrite extent policy data in reply if lock was blocked (2901)
5910 - drop filter export grants atomically with removal from device (2663)
5911 - del obd_self_export from work_list in class_disconnect_exports (2908)
5912 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
5913 - swab reply message in mdc_close, other PPC fixes (2464)
5914 - fix destroying of named logs (2325)
5915 - overwrite old logs when running lconf --write_conf (2264)
5916 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
5917 - fix race in target_handle_connect (2898)
5918 - mds_reint_create() should take same inode create lock (2926)
5919 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
5920 - don't close files for self_export to avoid uninitialized obd (2936)
5921 - allow MDS with the same name as client node (2939)
5922 - hold dentry reference for closed log files for unlink (2325)
5923 - reserve space for all logs during transactions (2059)
5924 - don't evict page beyond end of stripe extent (2925)
5925 - don't oops on a deleted current working directory (2399)
5926 - handle hard links to targets without a parent properly (2517)
5927 - don't dereference NULL lock when racing during eviction (2867)
5928 - don't grow lock extents when lots of conflicting locks (2919)
5930 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
5933 - account for cache space usage on clients to avoid data loss (974)
5934 - lfsck support in lustre kernel code (2349)
5935 - reduce journal credits needed for BRW writes (2370)
5936 - orphan handling to avoid losing space on client/server crashes
5937 - ptlrpcd can be blocked, stopping ALL progress (2477)
5938 - use lock value blocks to assist in proper KMS, faster stat (1021)
5939 - takes i_sem instead of DLM locks internally on obdfilter (2720)
5940 - recovery for initial connections (2355)
5941 - fixes for mds_cleanup_orphans (1934)
5942 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
5943 - block all file creations until orphan recovery completes (1901)
5944 - client remove rq_connection from request struct (2423)
5945 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
5946 - recovery timer race (2670)
5947 - mdc_close recovey bug (2532)
5948 - ptlrpc cleanup bug (2710)
5949 - mds timeout on local locks (2588)
5950 - namespace lock held during RPCs (2431)
5951 - handle interrupted sync write properly (2503)
5952 - don't try to handle a message that hasn't been replied to (2699)
5953 - client assert failure during cleanup after abort recovery (2701)
5954 - leak mdc device after failed mount (2712)
5955 - ptlrpc_check_set allows timedout requests to complete (2714)
5956 - wait for inflight reqs when ptlrpcd finishes (2710)
5957 - make sure unregistered services are removed from the srv_list
5958 - reset bulk XID's when resending them (caught by 1138 test)
5959 - unregister_bulk after timeout
5960 - fix lconf error (2694)
5961 - handle write after unfinished setstripe, stripe-only getstripe (2388)
5962 - readahead locks pages, leaves pending causing memory pressure (2673)
5963 - increase OST request buffers to 4096 on large machines (2729)
5964 - fix up permission of existing directories in simple_mkdir (2661)
5965 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
5966 - don't assign transno to errored transactions (2742)
5967 - don't delete objects on OST if given a bogus objid from MDS (2751)
5968 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
5969 - if rq_no_resend, then timeout request after recovery (2432)
5970 - fix MDS llog_logid record size, 64-bit array alignment (2733)
5971 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
5972 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
5973 - MDS recovery shouldn't delete objects that it has given out (2730)
5974 - if enqueue arrives after completion, don't clobber LVB (2819)
5975 - don't unlock pages twice when trigger_group_io returns error (2814)
5976 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
5977 - don't write pages to disk if there was an error (1450)
5978 - don't ping imports that have recovery disabled (2676)
5979 - take buffered bytes into account when balancing socknal conn (2817)
5980 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
5981 - reconnect unlink llog connection after MDS reconnects to OST (2816)
5982 - remove little-endian swabbing of llog records (1987)
5983 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
5984 - retry reposting request buffers if they fail (1191)
5985 - grow extent at grant time to avoid granting a revoked lock (2809)
5986 - lock revoke doesn't evict page if covered by a second lock (2765)
5987 - disable VM readahead to avoid reading outside lock extents (2805)
5989 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
5990 - updated kernel patches for hp-2.4.20 kernel (2681)
5992 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
5995 - fix truncated write corruption (2366)
5996 - fix for failed assertion in iopen_connect_dentry (1792,2517)
5998 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
5999 - ensure len(uuid) < 37 in lmc (1171)
6000 - fix ia64 OOPS in llog_test (2255)
6001 - zero end of page at obdfilter for partial page writes (2648)
6002 - don't leave stale dentries around after renames (bug 2428)
6003 - fix timeouts when evicting a client with a single lock held (2642)
6004 - set deadline for the initial HELLO message to drain (2634)
6005 - print out dotted-quad IP addresses in the socknal (2302)
6007 - additional debugging for MDS client eviction problem (2443)
6008 - fix mkfsoptions support for osts (2603, 2604)
6010 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
6013 - add series for the vanilla 2.6.0 kernel
6014 - add series for the vanilla 2.4.24 kernel
6015 - add series for a cray x86/64 UL kernel drop
6016 - fix xattr patches for the vanilla 2.4.19 series
6018 - generate true UUIDs in lmc (1171)
6019 - have portals stack dumping break in UML (2466)
6020 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
6021 - allocate with _NOFS in ldlm to avoid deadlock (1933)
6022 - wake callback waiting threads on client eviction (2460)
6023 - Add --ptldebug and --subsystem to lmc (1719)
6024 - update assertion to allow safe interrupt allocation
6025 - set rq_no_resend for cancel requests (2432)
6026 - recalculate ptlrpcd timeout after resend (2494)
6027 - call vfs_rmdir when removing pending directories (2368)
6028 - fix renaming a file to itself (2429)
6029 - lmc creates a default one-stripe lov (2454)
6030 - expand procfs space to handle large clusters (2326)
6031 - increase UML stack to avoid overflow
6032 - update lconf's list of debug and subsystem masks
6033 - fix lfs find --obd (2510)
6034 - /proc tunable for disabling filter read caching (2591)
6035 - stop rpm packages from altering slapd.conf (2301)
6036 - disable nagle in the socknal under 0conf (2578)
6037 - choose mds inode size based on stripe count (2572)
6038 - fix kernel-source rpm problems (2516)
6040 - add --disable-doc to avoid pdf generation (2421)
6041 - update documentation, tests, type-os, comments
6042 - avoid format warnings on ia64
6043 - remove the TOE NAL
6044 - tiny code cleanups by removing unused fields
6046 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
6049 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
6050 - remove the most blatant lies from BUILDING file (2371)
6051 - change default debug level to reasonable production setting
6052 - reduce client side cache size to reduce cache flush time
6053 - reduce max RPCs in flight to avoid unnecessary file fragmentation
6054 - make TCP zerocopy and pinger support enabled by default (2476)
6055 - sync writes completed after process exits caused crashes (2319)
6056 - maintain correct mount count on the MDS (2356)
6057 - backout 1557, because 2316 wasn't really fixed
6058 - better file I/O statistics gathering in /proc
6059 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
6060 - another kernel patch to fix zero-copy TCP function export
6061 - don't take duplicate lock when processing re-sent getattr (2420)
6062 - lctl uses obd_self_export instead of creating new conn (2353)
6063 - MDS/OST recovery case which requires object creation asserted (2425)
6064 - move lfs from /usr/sbin to /usr/bin in packages
6065 - fix race between mds_client_add and mds_client_free (2417)
6066 - use kmalloc instead of slabs in portals (2430)
6067 - don't create duplicate records when a failover MDS is present (2442)
6068 - remove unnecessary mount age check (2332)
6069 - don't remove directory inodes from locks prematurely (2451)
6070 - don't break if MDS service name is the same as hostname (2103)
6071 - fix races in client write RPC generation when cache full (2482)
6073 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
6076 - remove now-unused request->rq_obd (278)
6077 - if an allocation fails, print out how much memory we've used (1933)
6078 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
6079 - add the "configurable stack size" patch to most series files (1256)
6080 - ability to write large log records, for 100+ OST configs (2306)
6081 - fix NULL deref when filter_prep fails (2314)
6082 - fix operator precedence error in filter_sync
6083 - dynamic allocation of socknal TX descriptors (2315)
6084 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
6085 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
6086 - ensure that CWARN messages in Portals always get to the syslog
6087 - __init/__exit are not for prototype decls (ldlm_init/exit)
6088 - x86-64 compile warning fixes
6089 - fix gateway LMC keyword conflict (2318)
6090 - fix MDS lock inversions in getattr/reint paths (1844)
6091 - fix a rare lock re-ordering bug, which caused deadlock (2322)
6092 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
6093 - DLM race condition prevented some lock evictions (2328)
6094 - ENOMEM detection and retry on socknal sends (2230)
6095 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
6096 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
6097 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
6098 - better error messages when a client is rejected during recovery (1505)
6099 - avoid cancelling locks which were never granted, after failure (2330)
6100 - fix i_sem/journal inversion in mds_client_add (2333)
6101 - fix truncate/getattr lock cycle deadlock (2334)
6102 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
6103 - fix two rare exit paths which could leak an l_lock() ref (2321)
6104 - fencepost error in MDS/OST orphan recovery (2226)
6105 - make log record alignment 8 bytes (1988)
6106 - lstripe now fails when requested offset > ost_count (2237)
6107 - ensure that all kernel series have a complete list.h (1607)
6108 - fix crashes in special-file operations (2316)
6109 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
6111 - allow configurable automake binary, for testing new versions
6112 - small update to the lfs documentation
6114 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
6116 * fix negative export reference count in fsfilt_sync (2312)
6118 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
6119 * release candidate 0.9.1
6121 - orphans are moved into the PENDING directory for possible recovery
6122 - replayed opens now open by fid for orphan/rename safety (1042)
6123 - last close of an orphan inode generates a transno (683)
6124 - chdir() and mount() now pin the directory entry (1020)
6125 - avoid CERROR in normal ll_setattr_raw() error case (1500)
6126 - discard very old requests without processing them (1502)
6127 - remove some common, well-understood CERRORs (1505)
6128 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
6129 - clear "grant" flags in OST replies until OST grant code lands (1644)
6130 - fix read performance by not clobbering i_blksize on client (1598)
6131 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
6132 - make LRU size a /proc tunable, clears locks when reduced (707)
6133 - fix some lprocfs rot that prevented ptlbd from loading (1732)
6134 - server locks take references on exports now (1558)
6135 - build fixes for 2.4.20-rh trees (1663)
6136 - return an error from lov_create if all OSCs are inactive (1751)
6137 - fix import levels when a reconnect happens without a timeout (1597)
6138 - exit early from mds_open if we get a lookup error (1749)
6139 - partial page read at EOF wouldn't wait for disk before sending (1642)
6140 - avoid NULL deref in obdfilter when reading page past EOF (1592)
6141 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
6142 - fix LBUG in ll_it_open_error with rc = -2 (1861)
6143 - write/truncate lock inversion (1639)
6144 - Don't auto-load obdclass, portals modules during cleanup (1495)
6145 - fix timestamps from jumping to "now" (1763)
6146 - extra journal assertions (1648)
6147 - add an extra multiunlink test (1771)
6148 - fix read_record/write_record API (1776)
6149 - fix leak of offset_extent, possible incorrect i_size later (1772)
6150 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
6151 - Debugging for the kqswnal_get_idle_tx problems (1820)
6152 - Allow recovery to be attempted multiple times (1536)
6153 - Write out MDS last_rcvd file after it is first created (1600)
6154 - Fix tx_descriptor leak in failed transmit situations (1827)
6155 - ext3 journaling fixes for assertion failure after IO error (1871)
6156 - class_export_put() on freed export after completion AST error (1896)
6157 - Fix revalidate looping in VFS (1322)
6158 - Don't access a freed export during MDS_REINT timeout (1521)
6159 - Add open-unlink recovery support on the MDS (1673,1764)
6160 - Return an error if no MDS data was read from last_rcvd (1946)
6161 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
6162 - Don't setup a disk device that doesn't match exported UUID (317)
6163 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
6164 - avoid committing NULL handle in force close
6165 - local.sh is now a one-stripe LOV configuration
6166 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
6167 - fix ext3 htree duplicate directory entry corruption (1516)
6168 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
6169 - update to new LOV EA format (2097)
6170 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
6172 2003-06-15 Phil Schwan <phil@clusterfs.com>
6175 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
6176 - per-import/export recovery handling (958, 931, 959)
6177 - multiple last-rcvd slots, for serving multiple FSes (949)
6178 - connections are again shared between multiple imp/exports (963, 964)
6179 - "umount -f" would hang if any requests needed to be sent (393, 978)
6180 - avoid pinning large req buffer by copying for queued messages (989)
6181 - add "uuid" to "lctl device" command to help upcalls (991)
6182 - "open" RPCs with transnos would confuse recovery counters (1037)
6183 - do proper endian conversion of all wire messages (288, 340, 891)
6184 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
6185 - call waiting locks callback from LDLM recovery thread (1127, 1151)
6186 - fix ptlrpc_connection leak in target_handle_connect (1174)
6187 - fix import refcounting bug in OST and MDS cleanup (1134)
6188 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
6189 - fix very unlikely obd_types race condition (501)
6190 - remove osc_open hack for echo_client (1187)
6191 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
6192 - a failure in read_inode2 leads to deadlock (1139)
6193 - cancel ack-locks as soon as transaction is committed (1072)
6194 - fix major leaks and crashes in the bulk I/O path (937, 1057)
6195 - make sure to commitrw after any preprw to avoid deadlock (1162)
6196 - failing to execute a file in a lustre FS would lock inode (1203)
6197 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
6198 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
6199 - obd_uuid2tgt would walk off the end of the list (1255)
6200 - on IA64 the osc would give portals incorrect bulk size (1258)
6201 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
6202 - fix lock inversion caused by new llite matching code (1282)
6203 - limit the number of dirty pages on a client to 10MB (1286)
6204 - timed out locks were not being corrected cancelled (1289)
6205 - fix O_DIRECT above 4GB on IA-32 (1292)
6206 * major user-visible changes
6207 - fail out/fail over policy now controlled by the upcall (993)
6209 - add OBD_PING to check server availability and failure (954)
6210 - lustre messages are now sent in sending host order (288, 340, 891)
6211 - add eadatalen to MDS getattr reply (340)
6212 - OST read replies may contain second buffer, with per-page status (593)
6214 2003-03-11 Phil Schwan <phil@clusterfs.com>
6217 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
6218 - failed open()s could cause deadlock; fixed (867, 869)
6219 - stop cancelling OST locks when files are closed (481)
6220 - overlapping XID spaces caused network corruption (851, 853)
6221 - fix unsafe fsfilt counter arithmetic; change to atomic_t
6222 - setattr_raw added, to do single-RPC, server-side setattrs
6223 - lmc/lconf syntax change for OST UUIDs
6224 - fix crashy race condition between ptlrpc_free_req and osc_close
6225 - don't use request in mdc_enqueue if we hit a timeout (889)
6226 - don't set the inode i_size for regular files from the MDS (896)
6227 - handle out of order completion AST (842)
6228 - don't LBUG if a lock request times out after receiving AST (913)
6229 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
6230 - if a bad lock AST arrives, send an error instead of dropping entirely
6231 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
6232 - fix leak in bulk IO when only partially completed (899, 900, 926)
6233 - fix O_DIRECT for ia64 (55)
6234 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
6235 - C-z after timeout could hang a process forever; fixed (977)
6237 - client-side I/O cache (678, 924, 929, 941, 970)
6239 - READPAGE and SETATTRs which don't take server-side locks get
6242 2003-02-11 Phil Schwan <phil@clusterfs.com>
6245 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
6246 - Fix fsfilt_extN_readpage() to read a full page of directory
6247 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
6248 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
6249 - Fix imbalanced LOV object allocation and out-of-bound access (469)
6250 - Most intent operations were removed, in favour of a new RPC mode
6251 that does a single RPC to the server and bypasses most of the VFS
6252 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
6253 - Aggressively cancel local locks on DLM servers
6254 - mds_reint_unlink sends EA to the client if it's the last nlink.
6255 client uses that EA to unlink OST objects.
6256 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
6257 - recursive symlinks were fixed (439)
6258 - fixed NULL deref in DEBUG_REQ
6259 - filter_update_lastobjid no longer calls sync, which annoyed extN
6260 - fixed multi-client small-writes to a single file problem (445)
6261 - fixed mtime updates during file writes (607)
6262 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
6263 - fixed bug in obd_brw_read/write() (under guise of testing 367)
6264 - fixed Linux OST size reporting problem (444, 656)
6265 - OST now updates object mtime with writes or setattr (607, 619)
6266 - client verifies file size before zeroing page past EOF (445)
6267 - OST now writes last allocated objid to disk with allocation (108)
6268 - LOV on echo now works (409)
6270 - mds_reint_unlink sends a new buffer, with the EA included. this
6271 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
6272 set if a regular file was being unlinked, and it was the last link
6273 - use PtlGet from the target for bulk writes (315)
6274 - OST now updates object mtime with writes or setattr (607, 619)
6275 - LDLM now has a grant-time callback to revalidate locked items, if
6277 - Many MDS operations were reorganized to combat race conditions
6279 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
6280 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
6282 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
6285 - Fully reactivate OST imports after reconnection (512, others)
6286 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
6287 - More graceful error handling for truncating on dead OST (515)
6288 - Don't error out unless we're actually accessing dead stripes (474)
6289 - Fix garbage sizes when stripes are missing (410)
6290 - LRU counters were broken, causing constant lock purge (433, 432)
6291 - garbage on read from stripes with failed OSTs (441)
6292 - mark OSCs as active before reconnecting during recovery (438)
6293 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
6294 - lfind did not preserve OST order in output (443)
6295 - symlinks cause hung clients, incorrect data (439)
6296 - stop dereferencing request after dropping refcount (457)
6297 - don't LASSERT(spin_is_locked) on non-SMP (455)
6298 - fixes for many rename() bugs
6299 - fstat didn't correctly synchronize attributes (399)
6300 - server must handle lock cancellation during blocking AST prep (487)
6301 - bulk descriptors were free()d too soon (511)
6302 - fix paths in lconf, which would load incorrect modules (451, 507)
6303 - fix confusing lconf 'host not found' error message (386)
6304 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
6305 - fix race condition in mdc_blocking_ast() for inode access (526)
6306 - fix lov_unpackmd() unpacking wrong number of stripes (537)
6307 - fix lov_set_osc_active() marking wrong OSC inactive (440)
6308 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
6309 - fix multiple writes of stripe MD to MDS (358, maybe 519)
6310 - fix lstripe in several ways (kernel side) (527)
6311 - fix request leak in ldlm_cli_enqueue (262)
6312 - incorrect OSC was marked inactive after OST failure
6313 - call mds_fs_cleanup before unmounting filesystem (524)
6314 - fix races between taking ns_lock and ldlm_lock_change_resource
6315 - fix races updating LOV export open file list
6316 - fix lov_enqueue error path, avoid decref-ing bad lock handle
6317 - fix recovery NULL deref in ldlm_cli_cancel_unused
6318 - fix some DLM races by using new hash table for lock handles (419)
6319 - permit the client to specify desired inodes, at replay
6320 - duplicate requests when we queue them for replay reintegration
6321 - fix last_rcvd offset calculation
6322 - sync after each recovered transaction, so we always make progress
6323 - never, not always, ERESTART requests without transnos
6324 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
6326 - skip replay if the MDS says that the client is already connected
6327 - don't check for a recovery-enabled export to match lctl's UUID
6328 - don't INC_USE_COUNT for phantom exports
6329 - don't crash when cleaning up phantom exports (567)
6330 - don't double-finish or set replay data for errored mdc_open requests
6331 - abort requests when they time out, so we don't get old replies
6332 - send/receive replies for AST messages again
6333 - if the client says that it doesn't have the lock, cancel it on the
6335 - if we timeout during I/O, don't try to cancel an in-use lock; instead
6336 mark it as destroyed, it will all work out when decref is called
6337 - fix module use counts (22, 581)
6339 - ASTs now expect a reply (server cancels lock on error reply)
6341 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
6344 - fix many simultaneous client startup (392)
6345 - fix dentry->d_it clobbering
6346 - credentials weren't being shipped for readdir/getattr operations
6347 - remove invalid assertions triggered during some concurrent MD
6349 - proper Lustre versions added (336, 389)
6350 - fix memory leak for create error case (398)
6351 - fix LOV locking bug that would get cli/srv out of sync
6352 - fix echo client over LOV (409)
6353 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
6354 - fix double-O_EXCL intent crash (424)
6355 - avoid sending multiple lock CANCELs (352)
6357 - MDS can do multi-client recovery (modulo bugs in new code)
6359 - many updates, edits, cleanups
6361 2002-11-18 Phil Schwan <phil@clusterfs.com>
6364 - fix null d_it dereference (346)
6365 - fix full OST/dbench hang (333)
6366 - fix permission problem with file removal (286)
6367 - fix removal of OSCs from LOV when they fail
6368 - fix NULL deref during bulk timeout (214)
6369 - fix problems related to multiple filesystems on one MDS (241)
6370 - fixed serious subtle metadata locking bugs
6371 - free locks on clients when inodes are removed due to memory
6373 - fix inode pointer in lock data (285)
6374 - partial support for multiple MDS on a single host (241)
6375 - data locks weren't cancelled at clear_inode time (290, 311)
6376 - intent locks could lead to unbounded lock growth (205)
6377 - added a maximum lock count, an LRU list, and a flusher
6378 - fix multiple rename (365)
6379 - properly abstracted the echo client
6380 - OSC locked 1 byte too many; fixed
6381 - rewrote brw callback code:
6382 - fixed recovery bugs related to LOVs (306)
6383 - fixed too-many-pages-in-one-write crash (191)
6384 - fixed (again) crash in sync_io_timeout (214)
6385 - probably fixed callback-related race (385)
6387 - Add capability to MDS protocol
6388 - LDLM cancellations and callbacks on different portals
6390 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
6393 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
6394 - defer open object destruction to close time (601981, 138)
6395 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
6396 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
6397 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
6398 - don't call obd_setattr in ll_file_release for destroyed objects
6399 * protocol change to lustre_msg: move |version| and add |flags|
6400 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
6401 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
6402 * added replay of create, unlink, link and rename operations during
6403 MDS failover; recovery should be much more robust now
6404 * remove failed OSCs from LOVs (only lov_create uses this so far)
6405 * the lustre-HOWTO was brought (more) up to date (582544)
6407 2002-10-23 Phil Schwan <phil@clusterfs.com>
6410 - in-use dentries weren't being reused properly (617851)
6411 - prevent multiple LDLM setup (599178)
6412 - fix LOV size calculations for truncate (617853)
6413 - fix client handling of MDS intent errors (POSIX)
6414 - fix permission bug in lovstripe.c test (624321)
6415 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
6416 - truncate past end of file could corrupt data
6417 - proper cleanup after timeouts, crashes, etc (592524, 550815)
6418 - a race in recovery could return ETIMEDOUT to apps (623947)
6419 - building outside the source directory was fixed
6420 * the lustre-HOWTO was brought (more) up to date (582544)
6421 * major progress was made on recovery functionality
6423 2002-10-10 Phil Schwan <phil@clusterfs.com>
6426 - recovery deadlock fix
6427 - rm -rf causes LBUG fix (617817)
6428 - file open by multiple tasks fix (618962)
6429 - directory permissions bugs (602707 and 620007)
6430 - journal_stop fixed with locking (611313)
6431 - O_APPEND failures resolved (618273, perhaps 614459)
6432 - lconf PATH fix (619770)
6433 - IA64 build fix (621450)
6434 - RPC buffer sizes scale with amount of memory
6436 2002-10-01 Phil Schwan <phil@clusterfs.com>
6439 - locks would be cancelled without throwing away data pages,
6440 resulting in inconsistent data (605627)
6441 - inode attributes were not always being refreshed (605627, 612449)
6442 - lconf now continues to cleanup after lctl reports an error
6443 - MDS now enforces user permissions (602707)
6444 - lprocfs cleanup fixed, but not yet enabled (614157)
6445 - fixed infinite server hang, should a client not respond to an AST
6446 - avoid going into recovery if user calls readlink() with a buffer
6447 that's too small (613941)
6448 - AST RPCs no longer require replies (614867) -- this may be changed
6449 - don't crash server if client sends an IOV that's too big (611336)
6450 - fixed lock conversion deadlock (611892)
6451 - fixed the following of symlinks (614622)
6452 * recovery: the server can remove locks from a client that dies, other
6453 clients can make progress
6454 * more extN patch fixes
6455 * compile-time configurable ptlrpc buffer allocations
6457 - collaborative read cache document
6458 - Lustre Lite Performance CDR document-in-progress
6460 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6463 - fix typo in patch-2.4.18
6465 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6468 - clear ptlrpc request each time in handle_incoming_request()
6469 - unlink of files now destroys the object on the OST
6471 2002-09-19 Peter Braam <braam@clusterfs.com>
6473 * add hard link support
6474 * change obdfile creation method
6475 * kernel patch changed
6477 2002-09-19 Peter Braam <braam@clusterfs.com>
6480 - stack overflow bug in extN fixed
6482 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
6484 * documentation updates
6485 - add man pages for config tools
6486 - update tests/README to describe testing with new config tools
6487 - finish metadata API descriptions
6488 * bug fixes and cleanups
6489 - statfs workaround for 16TB limit
6490 - LOV stripe allocation improved, can stripe on subset of OSTs
6491 - LOV file size/IO offset was wrong for files > 4GB in size
6492 - object EA data was being dropped, caused files to be unreadable
6493 - memory overflow with non-LOV OST caused memory corruption
6494 - fixed regression tests to work with new config tools, obdfilter
6495 - fixed bug when directory size became larger than 1 block
6496 - fixed bug (for single client case) when PWD was deleted
6497 - invalidate local directory pages when doing intent-based ops
6498 - avoid LDLM oops when lock callback contained bad data
6500 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
6502 * documentation updates
6503 * bug fixes and cleanups
6504 - configuration tools
6507 - 64-bit compile warnings
6508 - 64-bit internal statfs data
6510 * test_brw on persistent OST devices
6512 * lprocfs (disabled)
6514 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
6516 * documentation updates
6517 * bug fixes and cleanups
6518 * configuration tools
6520 2002-08-30 Peter J. Braam <braam@clusterfs.com>
6523 * many small fixes to 0_5_4
6524 * io/network handling
6525 * thinkos in MDS operations
6527 2002-08-24 Peter J. Braam <braam@clusterfs.com>
6530 * crucial basic fixes to 0.5.3
6531 * IOR, Iozone work over Elan
6534 2002-08-07 Phil Schwan <phil@clusterfs.com>
6535 * version 0_5_3, our first alpha
6536 * we use the new Portals iovs
6537 * documentation updates
6538 * bug fixes and cleanups
6539 * small changes in the DLM wire protocol
6541 2002-07-25 Peter J. Braam <braam@clusterfs.com>
6542 * version 0_5_1 with some initial stability,
6543 * locking on MD and file I/O.
6544 * documentation updates
6545 * several bug fixes since 0.5.0
6546 * small changes in wire protocol
6548 2002-07-18 Phil Schwan <phil@clusterfs.com>
6550 * delivered as Lustre Light Alpha
6551 * fixed a crash after handling invalid MDS requests
6552 * fixed directory pages for architectures with non-4k pages sizes
6554 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
6555 * release version v0_4_4
6556 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
6558 * Fixes a number of interruption problems with OST operations.
6559 * Update documentation for portals header changes
6560 * Move all wire protocol structs/defines to lustre_idl.h
6561 * Fixes symlink length bug.
6562 * Add tcpdump to repository.
6564 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
6565 * release version v0_4_3
6566 * Fixes statfs for inodes on extN.
6567 * Fixes bug in runtests which would delete /etc/hosts.
6568 * Use 64-bit object IDs wherever possible (not into VFS though)
6569 Remove ost_get_info, which is unused by lustre, and out of date.
6571 2002-07-03 Peter Braam <braam@clusterfs.com>
6572 * release version v0_4_2 Fixes a lookup error (type not passed)
6573 * move forward to head of Portals
6574 * move forward to latest Lustre kernel
6576 2002-06-25 Peter Braam <braam@clusterfs.com>
6577 * release version v0_4_1. Hopefully stable on single node use.