3 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
4 for currently supported client and server kernel versions.
5 * Server known to build on patched kernels:
6 2.6.32-431.29.2.el6 (RHEL6.5)
7 2.6.32-504.30.3.el6 (RHEL6.6)
8 2.6.32-573.26.1.el6 (RHEL6.7)
9 2.6.32-642.15.1.el6 (RHEL6.8)
10 2.6.32-696.10.2.el6 (RHEL6.9)
11 3.10.0-693.2.2.el7 (RHEL7.4)
12 3.0.101-0.47.71 (SLES11 SP3)
13 3.0.101-107 (SLES11 SP4)
14 3.12.74-60.64.40 (SLES12 SP1)
15 4.4.74-92.35 (SLES12 SP2)
16 4.4.82-6.9 (SLES12 SP3)
17 vanilla linux 4.6.7 (ZFS only)
18 * Client known to build on unpatched kernels:
19 2.6.32-431.29.2.el6 (RHEL6.5)
20 2.6.32-504.30.3.el6 (RHEL6.6)
21 2.6.32-573.26.1.el6 (RHEL6.7)
22 2.6.32-642.15.1.el6 (RHEL6.8)
23 2.6.32-696.10.2.el6 (RHEL6.9)
24 3.10.0-693.2.2.el7 (RHEL7.4)
25 3.0.101-0.47.71 (SLES11 SP3)
26 3.0.101-107 (SLES11 SP4)
27 3.12.74-60.64.40 (SLES12 SP1)
28 4.4.74-92.35 (SLES12 SP2)
29 4.4.82-6.9 (SLES12 SP3)
31 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
32 * Recommended ZFS / SPL version: 0.7.2
33 * Tested with ZFS / SPL version: 0.7.2
34 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
35 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
36 stack overflow. For more information, please refer to bugzilla 17630.
37 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
38 the re-exporting nodes to be mounted with "32bitapi" mount option
40 --------------------------------------------------------------------------------
42 2017-07-05 Intel Corporation
44 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
45 for currently supported client and server kernel versions.
46 * Server known to build on patched kernels:
47 2.6.32-431.29.2.el6 (RHEL6.5)
48 2.6.32-504.30.3.el6 (RHEL6.6)
49 2.6.32-573.26.1.el6 (RHEL6.7)
50 2.6.32-642.15.1.el6 (RHEL6.8)
51 2.6.32-696.3.1.el6 (RHEL6.9)
52 3.10.0-514.21.1.el7 (RHEL7.3)
53 3.0.101-0.47.71 (SLES11 SP3)
54 3.0.101-100 (SLES11 SP4)
55 3.12.74-60.64.40 (SLES12 SP1)
56 4.4.59-92.17 (SLES12 SP2)
57 vanilla linux 4.6.7 (ZFS only)
58 * Client known to build on unpatched kernels:
59 2.6.32-431.29.2.el6 (RHEL6.5)
60 2.6.32-504.30.3.el6 (RHEL6.6)
61 2.6.32-573.26.1.el6 (RHEL6.7)
62 2.6.32-642.15.1.el6 (RHEL6.8)
63 2.6.32-696.3.1.el6 (RHEL6.9)
64 3.10.0-514.21.1.el7 (RHEL7.3)
65 3.0.101-0.47.71 (SLES11 SP3)
66 3.0.101-100 (SLES11 SP4)
67 3.12.74-60.64.40 (SLES12 SP1)
68 4.4.59-92.17 (SLES12 SP2)
70 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
71 * Recommended ZFS / SPL version: 0.6.5.9
72 * Tested with ZFS / SPL version: 0.6.5.9
73 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
74 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
75 stack overflow. For more information, please refer to bugzilla 17630.
76 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
77 the re-exporting nodes to be mounted with "32bitapi" mount option
79 --------------------------------------------------------------------------------
81 12-07-2016 Intel Corporation
83 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
84 for currently supported client and server kernel versions.
85 * Server known to build on patched kernels:
86 2.6.32-431.29.2.el6 (RHEL6.5)
87 2.6.32-504.30.3.el6 (RHEL6.6)
88 2.6.32-573.26.1.el6 (RHEL6.7)
89 2.6.32-642.11.1.el6 (RHEL6.8)
90 3.10.0-514.el7 (RHEL7.3)
91 3.0.101-0.47.71 (SLES11 SP3)
92 3.0.101-80 (SLES11 SP4)
93 3.12.62-60.64.8 (SLES12 SP1)
94 vanilla linux 4.5.7 (ZFS only)
95 * Client known to build on unpatched kernels:
96 2.6.32-431.29.2.el6 (RHEL6.5)
97 2.6.32-504.30.3.el6 (RHEL6.6)
98 2.6.32-573.26.1.el6 (RHEL6.7)
99 2.6.32-642.11.1.el6 (RHEL6.8)
100 3.10.0-514.el7 (RHEL7.3)
101 3.0.101-0.47.71 (SLES11 SP3)
102 3.0.101-80 (SLES11 SP4)
103 3.12.62-60.64.8 (SLES12 SP1)
105 * Recommended e2fsprogs version: 1.42.13.wc5 or newer
106 * Recommended ZFS / SPL version: 0.6.5.8
107 * Tested with ZFS / SPL version: 0.6.5.8
108 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
109 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
110 stack overflow. For more information, please refer to bugzilla 17630.
111 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
112 the re-exporting nodes to be mounted with "32bitapi" mount option
114 --------------------------------------------------------------------------------
116 02-29-2016 Intel Corporation
118 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
119 for currently supported client and server kernel versions.
120 * Server known to build on patched kernels:
121 2.6.32-431.29.2.el6 (RHEL6.5)
122 2.6.32-504.30.3.el6 (RHEL6.6)
123 2.6.32-573.12.1.el6 (RHEL6.7)
124 3.10.0-327.3.1.el7 (RHEL7.2)
125 3.0.101-0.47.71 (SLES11 SP3)
126 3.0.101-68 (SLES11 SP4)
127 vanilla linux 4.2.1 (ZFS only)
128 * Client known to build on unpatched kernels:
129 2.6.32-431.29.2.el6 (RHEL6.5)
130 2.6.32-504.30.3.el6 (RHEL6.6)
131 2.6.32-573.12.1.el6 (RHEL6.7)
132 3.10.0-327.3.1.el7 (RHEL7.2)
133 3.0.101-0.47.71 (SLES11 SP3)
134 3.0.101-68 (SLES11 SP4)
137 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
138 * Recommended ZFS / SPL version: 0.6.4.2
139 * Tested with ZFS / SPL version: 0.6.4.2
140 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
141 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
142 stack overflow. For more information, please refer to bugzilla 17630.
143 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
144 the re-exporting nodes to be mounted with "32bitapi" mount option
146 --------------------------------------------------------------------------------
148 03-10-2015 Intel Corporation
150 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
151 for currently supported client and server kernel versions.
152 * Server known to build on patched kernels:
153 2.6.32-431.29.2.el6 (RHEL6.5)
154 2.6.32-504.8.1.el6 (RHEL6.6)
155 3.0.101-0.46 (SLES11 SP3)
156 * Client known to build on unpatched kernels:
157 2.6.32-431.29.2.el6 (RHEL6.5)
158 2.6.32-504.8.1.el6 (RHEL6.6)
159 3.10.0-123.20.1.el7 (RHEL7)
160 3.0.101-0.46 (SLES11 SP3)
161 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
162 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
163 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
164 stack overflow. For more information, please refer to bugzilla 17630.
165 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
166 the re-exporting nodes to be mounted with "32bitapi" mount option
168 Severity : enhancement
170 Description: control OST-index in IDIF via ROCOMPAT flag.
171 Details : Introduce new flag OBD_ROCOMPAT_IDX_IN_IDIF that is stored in the
172 last_rcvd file. For new formatted OST device, it will be auto set;
173 for the case of upgrading from old OST device, you can enable it
174 via the lproc interface osd-ldiskfs.index_in_idif. With such flag
175 enabled, for new created OST-object, its IDIF-in-LMA will contain
176 the OST-index; for the existing OST-object, the OSD will convert
177 old format IDIF as new format IDIF with OST-index stored in the
178 LMA EA when accessing such OST-object or via OI scrub. Once such
179 flag is enabled, it cannot be reverted back, so the system cannot
180 be downgraded to the orignal incompatible version.
182 --------------------------------------------------------------------------------
184 07-30-2014 Intel Corporation
186 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
187 for currently supported client and server kernel versions.
188 * Server known to build on patched kernels:
189 2.6.32-431.20.3.el6 (RHEL6.5)
190 3.0.101-0.31 (SLES11 SP3)
191 * Client known to build on unpatched kernels:
192 2.6.32-431.20.3.el6 (RHEL6.5)
193 3.0.101-0.31 (SLES11 SP3)
196 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
197 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
198 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
199 stack overflow. For more information, please refer to bugzilla 17630.
200 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
201 the re-exporting nodes to be mounted with "32bitapi" mount option
203 --------------------------------------------------------------------------------
205 10-31-2013 Intel Corporation
207 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
208 * Server builds for kernels:
209 2.6.32-358.18.1.el6 (RHEL6.4)
210 * Client builds for unpatched kernels:
211 2.6.32-358.18.1.el6 (RHEL6.4)
212 3.0.92-0.8 (SLES11 SP3)
214 * Recommended e2fsprogs version: 1.42.7.wc2 or newer
215 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
216 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
217 stack overflow. For more information, please refer to bugzilla 17630.
218 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
219 the re-exporting nodes to be mounted with "32bitapi" mount option
221 --------------------------------------------------------------------------------
223 05-30-2013 Intel Corporation
225 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
226 * Server builds for kernels:
227 2.6.32-358.6.2.el6 (RHEL6.4)
228 * Client builds for unpatched kernels:
229 2.6.32-358.6.2.el6 (RHEL6.4)
230 3.0.74-0.6.6 (SLES11 SP2)
232 * Recommended e2fsprogs version: 1.42.7.wc1 or newer
233 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
234 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
235 stack overflow. For more information, please refer to bugzilla 17630.
236 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
237 the re-exporting nodes to be mounted with "32bitapi" mount option
239 --------------------------------------------------------------------------------
241 9-30-2012 Whamcloud, Inc.
243 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
244 * Server support for kernels:
245 2.6.32-279.5.1.el6 (RHEL6)
246 * Client support for unpatched kernels:
247 2.6.32-279.5.1.el6 (RHEL6)
249 2.6.18-238.19.1.el5 (RHEL5)
250 2.6.16.60-0.69.1 (SLES10)
251 * Recommended e2fsprogs version: 1.41.90.wc4
252 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
253 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
254 stack overflow. For more information, please refer to bugzilla 17630.
255 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
256 the re-exporting nodes to be mounted with "32bitapi" mount option
258 Severity : enhancement
260 Description: Update RHEL6.3 kernel to 2.6.32-279.5.1.el6.
262 Severity : enhancement
264 Description: Update RHEL6.3 kernel to 2.6.32-279.2.1.el6.
266 Severity : enhancement
268 Description: Update RHEL6.2 kernel to 2.6.32-220.17.1.el6
270 Severity : enhancement
272 Description: Update RHEL6.2 kernel to 2.6.32-220.13.1.el6
274 Severity : enhancement
276 Description: Update RHEL6.2 kernel to 2.6.32-220.7.1.el6
278 --------------------------------------------------------------------------------
280 04-30-2012 Whamcloud, Inc.
282 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
283 * Server support for kernels:
284 2.6.32-220.4.2.el6 (RHEL6)
285 * Client support for unpatched kernels:
286 2.6.32-220.4.2.el6 (RHEL6)
288 2.6.18-238.19.1.el5 (RHEL5)
289 2.6.16.60-0.69.1 (SLES10)
290 * Recommended e2fsprogs version: 1.41.90.wc4
291 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
292 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
293 stack overflow. For more information, please refer to bugzilla 17630.
294 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
295 the re-exporting nodes to be mounted with "32bitapi" mount option
297 Severity : enhancement
299 Description: Update RHEL6.2 kernel to 2.6.32-220.4.2.el6
301 Severity : enhancement
303 Description: Update RHEL6.2 kernel to 2.6.32-220.4.1.el6
305 Severity : enhancement
307 Description: Update RHEL6.2 kernel to 2.6.32-220.el6
309 Severity : enhancement
311 Description: Update RHEL6.1 kernel to 2.6.32-131.17.1.el6
313 --------------------------------------------------------------------------------
315 09-30-2011 Whamcloud, Inc.
317 * Server support for kernels:
318 2.6.18-238.19.1.el5 (RHEL5)
319 2.6.32-131.6.1.el6 (RHEL6)
320 * Client support for unpatched kernels:
321 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
322 2.6.18-238.19.1.el5 (RHEL5)
323 2.6.32-131.6.1.el6 (RHEL6)
324 2.6.29.4-167.fc11 (FC11)
325 2.6.16.60-0.69.1 (SLES10)
327 * Recommended e2fsprogs version: 1.41.90.wc3
328 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
329 Lustre filesystem with 4K stack may cause a stack overflow. For more
330 information, please refer to bugzilla 17630.
331 * NFSv4 reexport requires Lustre client on the re-exporting nodes to be
332 mounted with "32bitapi" mount option
334 Severity : enhancement
335 Description: Update RHEL6.1 kernel to 2.6.32-131.6.1.el6
337 Severity : enhancement
338 Description: Update RHEL5.6 kernel to 2.6.18-238.19.1.el5
340 Severity : enhancement
341 Description: Update RHEL5.6 kernel to 2.6.18-238.9.1.el5
344 Frequency : only with SLES10
346 Description: Use OFED "KMP" provided by Novell
347 Details : SLES10 SP3 ships with OFED in a separate "KMP" package.
348 Lustre is now built against this package. That means you need to
349 install the ofed-kmp package from Novell for the patchless client.
350 Note that the ofed-kmp that Novell ships may not exactly match the
351 kernel version but should still be compatible.
353 Severity : enhancement
355 Description: Update RHEL5.5 kernel to 2.6.18-194.17.1.el5
356 Update OEL5.5 kernel to 2.6.18-194.17.1.0.1.el5.
358 --------------------------------------------------------------------------------
360 2010-07-15 Oracle, Inc.
362 * Server support for kernels:
363 2.6.18-164.11.1.el5 (RHEL 5)
364 2.6.18-164.11.1.0.1.el5 (OEL 5)
365 * Client support for unpatched kernels:
366 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
367 2.6.18-164.11.1.el5 (RHEL 5),
368 2.6.18-164.11.1.0.1.el5 (OEL 5)
369 2.6.16.60-0.42.8 (SLES 10),
371 2.6.29.4-167.fc11 (FC11)
372 * Recommended e2fsprogs version: 1.41.10-sun2
374 Severity : enhancement
376 Description: Distribute lfs_migrate script with lustre
377 Details : The lfs_migrate mentioned in the manual is now included in the
378 lustre tree in lustre/scripts and will distributed in lustre RPMs.
380 Severity : enhancement
382 Description: Failover nid list added to import proc file.
384 Severity : enhancement
386 Description: "som_preview" mount option required to use SOM
387 Details : The clients will need to use som_preview mount option
391 Description: Lustre does not do 1MB IOs to HW RAID
392 Details : Bump MAX_PHYS/HW_SEGMENTS and SG_ALL to 256 in the RHEL5 kernel.
394 Severity : enhancement
396 Description: lfs quota output is not very convenient for awk/sed-parsing
397 Details : Some positions in lfs quota output table could be empty or
398 non-empty which made it hard to parse it with scripts, now
399 a dash is put instead of space where there is not supposed
404 Description: Update to OFED 1.5.1
405 Details : OFED 1.5.1 will be used in environments that don't use
406 vendor supplied OFED.
410 Description: Remove lnet. prefix from lctl params display
411 Details : Also renamed lnet.memused to lnet_memused to
416 Description: Use conf_param -d to remove permanant settings.
420 Description: Add list_param -R
421 Details : lctl list_param now supports recursive mode.
424 Bugzilla : 21829, 22312
425 Description: Fix llobdstat.
426 Details : 1. Now updating the write_bytes counter
427 2. Port script fixes from 1.8.
431 Description: decrease the usage of memory on clients.
432 Details : 1. On clients, recycle dentries and inodes unused.
433 2. Delete the code related to ll_deathrow(att 6215 in bug 1443). It
438 Description: llobdstat fix and enhancement
439 Details : add a counter to set a limit to how many samples will be returned
440 fix a wildcard in the path to limit to obdfilter stats only
444 Description: lfs check only allowed for root
445 Details : Code cleanups and added a regresion test of non-root use of
450 Description: MDS Oopses in ptlrpc_dispatch_difficult_reply()/__list_add()
451 Details : Fix hr->hr_index overload.
455 Description: Fix FIEMAP functionality.
459 Description: FC11 patchless client support added
463 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and
464 OEL5.4 kernel to 2.6.18-164.11.1.0.1.el5.
468 Description: New acc-sm script and various test-framework enhancments.
472 Description: Previously landed patch was reverted.
476 Description: Update SLES10 kernel to 2.6.16.60-0.42.8 (patchless)
480 Description: Add ldiskfs maxdirsize mount option.
484 Description: Fix mkfs.lustre support for 16TB LUNs.
488 Description: Print mmp_check_interval
489 Details : print mmp_check_interval and make it possible to abort
490 mount operation in case it takes too long.
494 Description: Add cascading_rw to lustre mpi tests.
498 Description: Store SOM data in EA.
499 Details : Also several SOM related cleanups.
503 Description: Add refcount to md_open_data
507 Description: Use refcount on lov_request_set to prevent memory corruption.
511 Description: Statahead: use parent dir's i_mutex for synchronization.
515 Description: loadgen improvements.
519 Description: Enable peer health detection in ksocklnd and ko2iblnd.
523 Description: Control DCACHE_LUSTRE_INVALID flag with MDS_INODE_LOOKCUP lock
527 Description: Fix race on obd_devcie:md_stats
531 Description: Fix SLES 11 client builds.
532 Details : Resolved uninitalized pointer gcc warnings.
536 Description: Allow non-root access for "lfs check".
537 Details : Added a check in obd_class_ioctl() for OBD_IOC_PING_TARGET.
541 Description: Clear imp_force_reconnect correctly in ptlrpc_connect_interpret()
543 Severity : enhancement
545 Description: Update kernel to RHEL5.4 2.6.18-164.6.1.el5 and
546 OEL5 2.6.18-164.6.1.0.1.el5(Both in-kernel OFED enabled).
550 Description: Sanity test_27p fails.
551 Details : OBD_FAIL_OST_ENOSPC has been moved to filter_statfs().
552 Sanity tests 27 were cleaned up and improved.
556 Description: Unify req format on client/servers
557 Details : Use new req_capsule API [almost] everywhere instead of old PTLRPC
558 buffers and swabbers approach..
563 Description: drop join file stuff.
565 Severity : enhancement
567 Description: Add 2.6.27 and SLES11 (patchless client) support
571 Description: Fix in ptlrpc_expire_one_request() to print the signed time
574 Severity : enhancement
576 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
579 Severity : enhancement
581 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
583 Severity : enhancement
585 Description: correctly handle big reply message.
586 Details : send LNet event if reply is bigger than buffer and adjust this buffer
591 Description: Drop unnecessary __GFP_NOMEMALLOC flag from filter_get_page()
593 Severity : enhancement
595 Description: Add support for OEL5.
599 Description: Conf-sanity.sh 50g test - deactivated OST should not cause a panic.
601 Severity : enhancement
603 Description: "weak-modules" support
604 Details : Implement "weak-modules" support which enables kernel modules
605 to be used with any kernel that implements the same kABI. In
606 order to achieve this modules are now installed in
607 /lib/modules/$(uname -r)/updates/kernel on all distributions.
611 Description: client could not reconnect to OST because of an active request.
612 Details : The client now retries to reconnect to the same server, if a
613 connect request failed with EBUSY or -EAGAIN. Test case 35b
614 conf-sanity.sh has been created.
618 Description: Don't allow make backward step on assignin osc next id.
619 Details : race between allocation next id and ll_sync thread can be cause
620 of set wrong osc next id and can be kill valid ost objects.
624 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
628 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
632 Description: Error handling in osc_statfs_interpret() has been improved.
633 Details : Check in osc_statfs_interpret() for EBADR.
637 Description: Do not send statfs() requests to OSTs disabled by administrator.
638 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
640 Severity : enhancement
642 Description: Added sanity-quota test to for lfs(1) to work with numeric uid/gid.
644 Severity : enhancement
646 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
648 Severity : enhancement
650 Description: Enhancement for lfs(1) command to use numeric uid/gid.
654 Description: Improved cleanup procedure for conf-sanity test_46a.
658 Description: Do not start bulk transfer, if client tried to reconnect.
659 Details : Do not start bulk transfer, if client tried to reconnect,
660 but failed due to active rpcs. Check for exp_abort_active_req
661 has been implemented.
665 Description: Prevent inconsistences between linux and lustre mount structures.
666 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
667 Make the sleep interruptible.
669 Severity : enhancement
671 Description: provide server to client comms path
672 Details : server to client communications path via new LDLM_SET_INFO rpc
676 Description: truncate starts GFP_FS allocation under transaction causing deadlock
677 Details : ldiskfs_truncate calls grab_cache_page which may start page
678 allocation under an open transaction. This may lead to
679 calling prune_icache with consequent lustre reentrance.
683 Description: send correctly lsm on open replay
684 Details : MDS is trust to LSM size on replay open, but client can set wrong size
689 Description: enable client interface failover
690 Details : When a child reconnects from another NID, properly update export
691 nid hash position and ldlm reverse import.
693 Severity : enhancement
695 Description: Communicate OST degraded/readonly state via statfs to MDS
696 Details : Flags in the statfs returned from OSTs indicate whether the
697 OST is in a degraded RAID state, or if the filesystem has
698 turned read-only after a filesystem error is detected.
703 Description: don't panic if EPROTO was hit when reading symlink
704 Details : correctly handling request reference in error cases.
706 Severity : enhancement
708 Description: Add LustreNetLink, a kernel-userspace communcation path.
710 Severity : enhancement
712 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
717 Description: access to llog context before init.
718 Details : move handling CATALOGS file at osc layer and forbid access to llog
722 Frequency : rare, only when client is writing to OST on local node
724 Description: Avoid deadlock for local client writes
725 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
726 memory freeing context. This allows OST threads to set the
727 PF_MEMALLOC flag on task structures in order to allocate memory
728 from reserved pools and complete IO.
729 Use GFP_HIGHUSER for OST allocations for non-local client writes,
730 so that the OST threads generate memory pressure and allow
731 inactive pages to be reclaimed.
733 Severity : enhancement
735 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
740 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
741 Details : move ldlm namespace creation in setup phase to avoid grab
742 _lprocfs_lock with cli_sem held.
746 Description: Temporarily disable grant shrink.
747 Details : Disable the feature for debugging.
750 Frequency : only when formatting test filesystems
752 Description: Unable to run several mkfs.lustre on loop devices at the same time
753 Details : mkfs.lustre returns error 256 on the concurrent loop devices
754 formatting. The solution is to proper handle the error.
756 Severity : enhancement
758 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
760 Severity : enhancement
762 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
767 Description: resolve race between obd_disconnect and class_disconnect_exports
768 Details : if obd_disconnect will be called to already disconnected export he
769 forget release one reference and osc module can't unloaded.
771 Severity : enhancement
773 Description: Allow tuning service thread via /proc
774 Details : For each service a new
775 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
776 created that can be used to set min/max thread counts, and get the
777 current number of running threads.
782 Descriptoin: don't return error if have partially created objects for file.
783 Details : lov_update_create_set uses set->set_success as index for created
784 objects, so if some requests will be failed, they will have hole
785 at end of array and qos_shrink_lsm for allocate correct lsm.
787 Severity : enhancement
789 Description: Update OFED support to 1.4.1
794 Description: Reduce small size read RPC
795 Details : Set read-ahead limit for every file and only do read-ahead when
796 available read-ahead pages are bigger than 1M to avoid small size
799 Severity : enhancement
801 Description: move AT tunable parameters for more consistent usage
802 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
804 Severity : enhancement
806 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
807 Details : allow skip disconnected ost for send statfs request and hide error
813 Description: fix racy locking of mballoc block bitmaps causing BUG
814 Details : The locking of the mballoc buddy bitmap and the in-memory
815 block bitmap was using two different spin locks in some
816 cases. This made it possible to incorrectly access the
817 mballoc bitmap while another process was modifying it,
818 causing a sanity assertion to fail. While no on-disk corruption
819 was reported, there was some risk of this happening.
822 Frequency : rare, on llog test 6
824 Descriptoin: don't allow connect to already connected import
825 Details : allowing connect to already connected import is hide connecting problem.
830 Descriptoin: allow kill process which wait statahead result
831 Details : for some reasons 'ls' can stick in waiting result from statahead,
832 in this case need way for kill this process.
834 Severity : enhancement
836 Description: Add state history info file, enhance import info file
837 Details : Track import connection state changes in a new osc/mdc proc file;
838 add overview-type data to the osc/mdc import proc file.
840 Severity : enhancement
842 Description: MDS create should not wait for statfs RPC while holding DLM lock.
844 Severity : enhancement
846 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
851 Descriptoin: OST grant too much space to client even there are not enough space.
852 Details : Client will shrink its grant cache to OST if there are no write
853 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
854 this grant cache if there are already not enough avaible space
855 (left_space < total_clients * 32M).
858 Frequency : start MDS on uncleanly shutdowned MDS device
860 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
861 Details : stay in waiting mds<>ost recovery finished produce random bugs
862 due race between two ll_sync thread for one lov target. send
863 ACTIVATE event only if connect realy finished and import have
867 Frequency : rare, connect and disconnect target at same time
869 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
870 Details : don't call obd_disconnect under lov_lock. this long time
871 operation and can block ptlrpcd which answer to connect request.
876 Descriptoin: don't lose wakeup for imp_recovery_waitq
877 Details : recover_import_no_retry or invalidate_import and import_close can
878 both sleep on imp_recovery_waitq, but we was send only one wakeup
882 Frequency : always with long access acl
884 Descriptoin: mds can't pack reply with long acl.
885 Details : mds don't control size of acl but they limited by reint/getattr
888 Severity : enhancement
890 Description: Update to SLES10 kernel-2.6.16.60-0.33.
892 Severity : enhancement
894 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
897 Frequency : start MDS on uncleanly shutdowned MDS device
899 Descriptoin: aborting recovery hang on MDS
900 Details : don't throttle destroy RPCs for the MDT.
903 Frequency : on remount
905 Description: external journal device not working after the remount
906 Details : clear dev_rdonly flag for external journal devices in
912 Description: shutdown vs evict race
913 Details : client_disconnect_export vs connect request race.
914 if client will evicted at this time - we start invalidate
915 thread without referece to import and import can be freed
919 Frequency : rare, need acl's on inode.
921 Description: client can't handle ost additional correctly
922 Details : if ost was added after client connected to mds client can have
923 hit lnet_try_match_md ... to big messages to wide striped files.
924 in this case need teach client to handle config events about add
925 lov target and update client max ea size at that event.
927 Severity : enhancement
929 Description: Changelogs
930 Details : Changelogs are a lightweight mechanism to track filesystem
931 metadata and namespace changes. The changelog is recorded
932 permanently on the MDTs, and is periodically "consumed" / purged
933 when records are no longer needed.
935 Severity : enhancement
937 Description: compact fld format with extents
938 Details : Store range of seq rather than every seq in FLD. Seq
939 controller update FLD rather than clients. In Case of CMD, mdt0
940 has FLD, all other metadata server act as non persistent proxy
941 for FLD queries and cache fld entries in fld cache.
946 Description: don't skip ost target if they assigned to file
947 Details : Drop slow OSCs if we can, but not for requested start idx.
948 This means "if OSC is slow and it is not the requested
949 start OST, then it can be skipped, otherwise skip it only
950 if it is inactive/recovering/out-of-space.
954 Description: more cleanup in mds_lov
955 Details : not send LOV EA under replay, we can't know about they size at this
956 time. Don't allow client connect to mds before any ost connected,
957 for avoid problems with LOV EA size and returning EIO to client.
959 Severity : enhancement
961 Description: Interoperability at server side (Disk interoperability)
963 Severity : enhancement
965 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
967 Severity : enhancement
969 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
971 Severity : enhancement
973 Description: New client IO stack (CLIO).
975 Severity : enhancement
977 Description: Commit on sharing. Eliminate inter-client dependencies between
978 uncommitted transactions by doing transaction commits.
979 Thereby clients may recovery independently.
982 Frequency : Create a symlink file with a very long name
984 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
985 Details : If there is no extra space in the request for early cancels,
986 ldlm_req_handles_avail() returns 0 instead of a negative value.
988 Severity : enhancement
990 Description: Add /proc entry for import status
991 Details : The mdc, osc, and mgc import directories now have
992 an import directory that contains useful import data for debugging
995 Severity : enhancement
997 Description: Re-disable certain /proc logging
998 Details : Enable and disable client's offset_stats, extents_stats and
999 extents_stats_per_process stats logging on the fly.
1002 Frequency : Only on FC kernels 2.6.22+
1004 Description: oops in statahead
1005 Details : Do not drop reference count for the dentry from VFS when lookup,
1006 VFS will do that by itself.
1008 Severity : enhancement
1010 Description: Generic /proc file permissions
1011 Details : Set /Proc file permissions in a more generic way to enable non-
1012 root users operate on some /proc files.
1016 Description: Hitting mdc_commit_close() ASSERTION
1017 Details : Properly handle request reference release in
1018 ll_release_openhandle().
1022 Description: quota recovery deadlock during mds failover
1023 Details : This patch includes att18982, att18236, att18237 in bz14840.
1025 1. fix osts hang when mds does failover with quotaon
1026 2. prevent watchdog storm when osts threads wait for the
1031 Frequency : only patchless client
1032 Description: add workaround for race between add/remove dentry from hash
1034 Severity : enhancement
1036 Description: Allow OST glimpses to return PW locks
1040 Description: LBUG when llog conf file is full
1041 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
1045 Description: Prevent import from entering FULL state when server in recovery
1049 Description: service mount cannot take device name with ":"
1050 Details : Only when device name contains ":/" will mount treat it as
1056 Description: replace ptlrpcd with the statahead thread to interpret the async
1057 statahead RPC callback
1061 Frequency : on recovery
1062 Description: I/O failures after umount during fail back
1063 Details : if client reconnected to restarted server we need join to recovery
1064 instead of find server handler is changed and process self
1065 eviction with cancel all locks.
1067 Severity : enhancement
1069 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
1071 Severity : enhancement
1073 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
1075 Severity : enhancement
1077 Description: Upcall on Lustre log has been dumped
1078 Details : Allow for a user mode script to be called once a Lustre log has
1079 been dumped. It passes the filename of the dumped log to the
1080 script, the location of the script can be specified via
1081 /proc/sys/lnet/debug_log_upcall.
1086 Description: avoid idr_remove called for id which is not allocated.
1087 Details : Move assigment s_dev for clustered nfs to end of initialization,
1088 to avoid problem with error handling.
1093 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
1094 Details : When connection is reused this not moved from CONN_UNUSED_HASH
1095 into CONN_USED_HASH and this prodice warning when put connection
1096 again in unused hash.
1099 Severity : enhancement
1101 Description: File striping can now be set to use an arbitrary pool of OSTs.
1103 Severity : enhancement
1105 Description: Export bytes_read/bytes_write count on OSC/OST.
1109 Description: Early reply size mismatch, MGC loses connection
1110 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
1111 the connect flags are properly negotiated.
1116 Description: quotas are not honored with O_DIRECT
1117 Details : all writes with the flag O_DIRECT will use grants which leads to
1118 this problem. Now using OBD_BRW_SYNC to guard this.
1122 Description: add quota statistics
1123 Details : 1. sort out quota proc entries and proc code.
1124 2. add quota statistics
1126 Severity : enhancement
1128 Description: enable quota support for HEAD.
1132 Description: Properly propagate oinfo flags from lov to osc for statfs
1133 Details : restore missing copy oi_flags to lov requests.
1135 Severity : enhancement
1137 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
1138 : l_getgroups(8), lst(8), routerstat(8)
1139 Details : included man pages for llobdstat(8), llstat(8),
1140 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
1142 Severity : enhancement
1144 Description: configure's --enable-quota should check the
1145 : kernel .config for CONFIG_QUOTA
1146 Details : configure is terminated if --enable-quota is passed but
1147 : no quota support is in kernel
1151 Description: Remove portals compatibility
1152 Details : Remove portals compatibility, not interoperable with releases
1157 Description: Resolve device initialization race
1158 Details : Prevent proc handler from accessing devices added to the
1159 obd_devs array but yet be intialized.
1161 Severity : enhancement
1163 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
1165 Severity : enhancement
1167 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
1172 Description: Using wrong pointer in osc_brw_prep_request
1173 Details : Access to array[-1] can produce panic if kernel compiled with
1174 CONFIG_PAGE_ALLOC enabled
1178 Description: Client runs out of low memory
1179 Details : Consider only lowmem when counting initial number of llap pages
1183 Description: *optional* service tags registration
1184 Details : if the "service tags" package is installed on a Lustre node
1185 When the filesystem is mounted, a local-node service tag will
1186 be created. See http://inventory.sun.com/ for more information
1187 about the Service Tags asset management system.
1191 Description: Kernel BUG tries to release flock
1192 Details : Lustre does not destroy flock lock before last reference goes
1193 away. So always drop flock locks when client is evicted and
1194 perform unlock regardless of successfulness of speaking to MDS.
1198 Description: add refcount protection for osc callbacks, avoid panic on shutdown
1202 Description: sanity test 65a fails if stripecount of -1 is set
1203 Details : handle -1 striping on filesystem in ll_dirstripe_verify
1208 Description: ASSERTION(CheckWriteback(page,cmd)) failed
1209 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
1212 Severity : enhancement
1214 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
1218 Description: do not process already freed flock
1219 Details : flock can possibly be freed by another thread before it reaches
1220 to ldlm_flock_completion_ast.
1224 Description: LBUG during stress test
1225 Details : Need properly lock accesses the flock deadlock detection list.
1229 Description: oops in page fault handler
1230 Details : kernel page fault handler can return two special 'pages' in error
1231 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
1235 Description: timeout with invalidate import.
1236 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
1237 should be handled by ptlrpcd. This produce long age waiting and
1238 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
1240 Severity : enhancement
1242 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
1246 Description: enable MGS and MDT services start separately
1247 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
1248 a co-located MGS without starting the MGS, which is a complement
1249 to 'nosvc' mount option.
1253 Frequency : after recovery
1254 Description: precreate to many object's after del orphan.
1255 Details : del orphan st in oscc last_id == next_id and this triger growing
1256 count of precreated objects. Set flag LOW to skip increase count
1257 of precreated objects.
1261 Frequency : rare, on clear nid stats
1262 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
1263 Details : when clean nid stats sometimes try destroy live entry,
1264 and this produce panic in free.
1268 Description: Stack overflow during MDS log replay
1269 ease stack pressure by using a thread dealing llog_process.
1273 Description: wait until IO finished before start new when do lock cancel.
1274 Details : VM protocol want old IO finished before start new, in this case
1275 need wait until PG_writeback is cleared until check dirty flag and
1276 call writepages in lock cancel callback.
1278 Severity : enhancement
1280 Description: using special macro for print time and cleanup in includes.
1284 Description: mds_mfd_close() ASSERTION(rc == 0)
1285 Details : In mds_mfd_close(), we need protect inode's writecount change
1286 within its orphan write semaphore to prevent possible races.
1290 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
1294 Frequency : rare, on shutdown ost
1295 Description: don't hit live lock with umount ost.
1296 Details : shrink_dcache_parent can be in long loop with destroy dentries,
1297 use shrink_dcache_sb instead.
1301 Description: don't panic with use echo client
1302 Details : echo client pass NULL as client nid pointer and this produce null
1303 pointer dereference.
1307 Description: fix build on ppc32
1308 Details : compile code with -m64 flag produce wrong object file for ppc32.
1312 Description: add message levels for liblustreapi
1316 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
1317 Details : Move the 'good_osts' check before the 'total_bavail' check. This
1318 will result in an -EAGAIN and in the exit call path we call
1319 alloc_rr() which will with increasing aggressiveness attempt to
1320 aquire precreated objects on the minimum number of required OSCs.
1324 Description: Use old size assignment to avoid deadlock
1325 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
1326 the scheduling while holding a spinlock. We do not need locking
1327 for size in ll_update_inode() because size is only updated from
1328 the MDS for directories or files without objects, so there is no
1329 other place to do the update, and concurrent access to such inodes
1330 are protected by the inode lock.
1334 Description: resolve "_IOWR redefined" build error on SLES10
1338 Description: dump the memory debugging after all modules are unloaded to
1339 suppress false negative in conf_sanity test 39
1341 Severity : enhancement
1343 Description: build kernel-ib packages for OFED 1.3 in our release cycle
1348 Description: fix SLES kernel versioning
1349 Details : the kernel version for our SLES 10 kernel did not include a "-"
1350 before the "smp" at the end. while this was not a problem in
1351 general, it did mean that software trying to use the kernel
1352 version to try to detect a vendor specific kernel would fail.
1353 this was most evident by the OFED build scripts.
1357 Description: Don't update lov_desc members until making sure they are valid
1358 Details : When updating lov_desc members via proc fs, need fix their
1359 validities before doing the real update.
1363 Description: don't put request into delay list while invalidate in flight.
1364 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
1365 import in flight. this produce timeout for invalidate and sometimes
1366 can cause stale data.
1370 Frequency : on ppc only
1371 Description: not convert ost objects for directory because it's not exist.
1372 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
1376 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
1378 Severity : enhancement
1380 Description: Update to support 2.6.22.14 vanilla kernel.
1384 Frequency : rare, on recovery
1385 Description: read procfs can produce deadlock in some situation
1386 Details : Holding lprocfs lock which send rpc can produce block for destroy
1387 obd objects and this also block reconnect with -EALREADY. This
1388 isn't fix all lprocfs bugs - but make it rare.
1390 Severity : enhancement
1392 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
1395 Frequency : frequent on X2 node
1397 Description: mdc_set_open_replay_data LBUG
1398 Details : Set replay data for requests that are eligible for replay.
1402 Description: lustre_mgs: operation 101 on unconnected MGS
1403 Details : When MGC is disconnected from MGS long enough, MGS will evict the
1404 MGC, and late on MGC cannot successfully connect to MGS and a lot
1405 of the error messages complaining that MGS is not connected.
1409 Frequency : on network error
1410 Description: panic with double free request if network error
1411 Details : mdc_finish_enqueue is finish request if any network error ocuring,
1412 but it's true only for synchronus enqueue, for async enqueue
1413 (via ptlrpcd) this incorrect and ptlrpcd want finish request
1416 Severity : enhancement
1418 Description: client-side metadata stat-ahead during readdir(directory readahead)
1419 Details : perform client-side metadata stat-ahead when the client detects
1420 readdir and sequential stat of dir entries therein
1423 Frequency : on start mds
1425 Description: Implement get_info(last_id) in obdfilter.
1428 Frequency : occasional
1430 Description: Correctly check stale fid, not start epoch if ost not support SOM
1431 Details : open with flag O_CREATE need set old fid in op_fid3 because
1432 op_fid2 was overwritten with new generated fid, but mds can answer
1433 with one of these two fids and both is not stale. Setattr
1434 incorrectly started an epoch and assume will be called
1435 done_writing, but without SOM done_writing ever being called.
1438 Frequency : rare, depends on device drivers and load
1440 Description: MDS or OSS nodes crash due to stack overflow
1441 Details : Code changes in 1.8.0 increased the stack usage of some functions.
1442 In some cases, in conjunction with device drivers that use a lot
1443 of stack the MDS (or possibly OSS) service threads could overflow
1444 the stack. One change which was identified to consume additional
1445 stack has been reworked to avoid the extra stack usage.
1448 Frequency : occasional
1450 Description: Do not fail import if osc_interpret_create gets -EAGAIN
1451 Details : If osc_interpret_create got -EAGAIN it immediately exits and
1452 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
1453 oscc_has_objects and see OSC has no objests and call
1454 oscc_internal_create to resend create request.
1456 Severity : enhancement
1458 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
1460 Severity : enhancement
1462 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
1465 Frequency : very rare
1467 Description: Fix replay if there is an un-replied request and open
1468 Details : In some cases, older replay request will revert the
1469 mcd->mcd_last_xid on MDS which is used to record the client's
1470 latest sent request.
1472 Severity : enhancement
1474 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
1476 Severity : enhancement
1478 Description: Add rhel5 support to HEAD.
1480 Serverity : enhancement
1482 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
1487 Description: Don't allow skipping OSTs if index has been specified.
1488 Details : Don't allow skipping OSTs if index has been specified, make
1489 locking in internal create lots better.
1493 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
1494 Details : access to bitfield in structure is always rounded to long
1495 and this produce problem with not atomic change any bit.
1499 Description: Lustre make rpms failed.
1500 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
1501 builds Lustre rpms from tarball.
1504 Frequency : rare on shutdown ost
1506 Description: If llog cancel was not send before clean_exports phase, this can
1507 produce deadlock in llog code.
1508 Details : If llog thread has last reference to obd and call class_import_put
1509 this produce deadlock because llog_cleanup_commit_master wait when
1510 last llog_commit_thread exited, but this never success because was
1511 called from llog_commit_thread.
1515 Description: allow userland application know is lost one of stripes.
1516 Details : fill lvb_blocks with error code on ost and return it to
1517 application if error flag found.
1521 Description: NULL lov_tgts causing MDS oops
1522 Details : more safe checks for NULL lov_tgts for avoid oops.
1524 Severity : enhancement
1526 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
1530 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
1531 Details : If an OST is down the MDS will hang indefinitely in
1532 obd_statfs_rqset() waiting for the statfs data. While for
1533 MDS QOS usage of statfs, it should not stuck in waiting.
1535 Severity : enhancement
1537 Description: remote_acl support
1538 Details : Support ACL-based permission check for remote user.
1539 Support setfacl/getfacl for remote user with the utils
1540 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
1541 the system "{s,g}etfacl" utils.
1543 Severity : enhancement
1545 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
1547 Severity : enhancement
1549 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
1553 Description: make mgs_setparam() handle fsname containing dash
1554 Details : fsname containing a dash does not work with lctl conf_param
1556 Severity : enhancement
1558 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
1560 Severity : enhancement
1562 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
1566 Frequency : rare, at shutdown
1567 Description: access already free / zero obd_namespace.
1568 Details : if client_disconnect_export was called without force flag set,
1569 and exist connect request in flight, this can produce access to
1570 NULL pointer (or already free pointer) when connect_interpret
1571 store ocd flags in obd_namespace.
1575 Frequency : only at startup
1576 Description: not alloc memory with spinlock held.
1577 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
1578 if any spinlock held.
1580 Severity : enhancement
1582 Description: make lustre randomly fail allocating memory
1583 Details : Make lustre randomly failed allocating memory for testing purpose.
1585 Severity : enhancement
1587 Description: lost problems with lov objid file
1588 Details : Fixes some scability and access to not inited memory problems
1589 in work with lov objdid file.
1594 Description: lfs find does not continue on file error
1595 Details : Continue other files processing when a file/dir is absent.
1599 Description: Inconsistent usage of lustre_pack_reply()
1600 Details : Standardize the usage of lustre_pack_reply() such that it
1601 always generate a CERROR on failure.
1604 Frequency : occasional
1606 Description: MDS hang or stay in waiting lock
1607 Details : If client receive lock with CBPENDING flag ldlm need send lock
1608 cancel as separate rpc, to avoid situation when cancel request
1609 can't processed due all i/o threads stay in wait lock.
1613 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1617 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1621 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1625 Description: Update to RHEL4 latest kernel.
1627 Severity : enhancement
1629 Description: Build SLES10 patchless client fails
1630 Details : The configure was broken by run ./configure with
1631 --with-linux-obj=.... argument for patchless client. When the
1632 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1633 can't search header adequately. Use absolute path such as
1634 -I($LINUX)/include instead.
1638 Description: interrupt oig_wait produce painc on resend.
1639 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1640 private set, and this produce situation when rq_ptlrpcd_data not
1641 copyed to new allocated request and triggered LBUG on assert
1642 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1643 ptlrpcd set if request is changed and can be safety changed to use
1648 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1649 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1650 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1651 these macros and replace them with their difinitions instead.
1655 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1656 Details : Update which_patch & target file for SLES10 latest kernel.
1659 Description: Remove client patches from SLES 10 kernel.
1660 Details : This causes SLES 10 clients to behave as patchless clients
1661 even on a Lustre-patched (server) kernel.
1663 Severity : enhancement
1665 Description: self-adjustable client's lru lists
1666 Details : use adaptive algorithm for managing client cached locks lru
1667 lists according to current server load, other client's work
1668 pattern, memory activities, etc. Both, server and client
1669 side namespaces provide number of proc tunables for controlling
1672 Severity : enhancement
1674 Description: light-weight GSS support
1675 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
1676 and reduce performance overhead.
1678 Severity : enhancement
1680 Description: Linux keyring support
1681 Details : Support using service of Linux keyring for Lustre GSS internal
1682 context refresh/cache mechanism.
1686 Description: Fix errors in lfs documentation
1687 Details : Fixes man pages
1691 Description: don't use GFP_* in generic Lustre code.
1692 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
1696 Description: obdclass is limited by single OBD_ALLOC(idarray)
1697 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
1701 Description: warnings with build patchless client with vanila 2.6.19 and up
1702 Details : change old ctl_table style and replace ctl_table/ctl_table_header
1703 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
1707 Frequency : Only for SLES
1708 Description: sanity_quota fail test_1
1709 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
1710 file, which makes TSTID[2] inunique.
1714 Frequency : Only for SLES9
1715 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
1716 Details : Update target/ChangeLog/which_patch .
1720 Frequency : Only for test_10 in sanity-quota.sh
1721 Description: a bug in quota test code
1722 Details : checking a test flag in a wrong place
1726 Frequency : Only for test_13 in sanity-quota.sh
1727 Description: a bug in quota test script
1728 Details : shouldn't check overfull allocated quota
1732 Description: jbd statistics
1733 Details : Port older jbd statistics patch for sles10
1735 Severity : enhancement
1737 Description: Add printing inode info into message about error in writepage.
1742 Description: reply_lock_interpret crash due to race with it and lock cancel.
1743 Details : Do not replay locks that are being cancelled. Do not reference
1744 locks by their address during replay, just by their handle.
1748 Frequency : When flocks are used.
1749 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
1750 Details : Flock locks might destroy just granted lock if it could be merged
1751 with another existing flock, this is done in completion handler,
1752 so teach ldlm_cli_enquque_fini that this is a valid case for
1758 Description: Oops in read and write path when failing to allocate lock.
1759 Details : Check if lock allocation failed and return error back.
1763 Description: lstripe command fails for valid OST index
1764 Details : The stripe offset is compared to lov->desc.ld_tgt_count
1765 instead of lov->desc.ld_active_tgt_count.
1769 Description: sanity.sh failed test 103
1770 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
1775 Description: df doesn't work properly if diskfs blocksize != 4K
1776 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
1780 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
1781 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
1783 Severity : enhancement
1785 Description: Adaptive timeouts
1786 Details : RPC timeouts adapt to changing server load and network
1787 conditions to reduce resend attempts and improve recovery time.
1791 Description: llapi_file_create() does not allow some changes
1792 Details : add llapi_file_open() that allows specifying the mode and
1793 open flags, and also returns an open file handle.
1797 Description: merge and cleanup kernel patches.
1798 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
1802 Description: Add journal checksum support.(Kernel part)
1803 Details : The journal checksum feature adds two new flags i.e
1804 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
1805 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
1806 indicates that the commit block contains the checksum for
1807 the blocks described by the descriptor blocks. Now commit
1808 record can be sent to disk without waiting for descriptor
1809 blocks to be written to disk. This behavior is controlled
1810 using JBD2_FEATURE_ASYNC_COMMIT flag.
1814 Description: OSS needs mutliple precreate threads
1815 Details : Add ability to start more than one create thread per OSS.
1819 Description: Sanity.sh test_65k failed: setstripe should have succeeded
1820 Details : execute lfs setstripe on client
1824 Description: mds_obd_create error creating tmp object
1825 Details : When the user sets quota on root, llog will be affected and can't
1826 create files and write files.
1829 Frequency : Always on ia64 patchless client, and possibly others.
1831 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
1832 Details : This allows the patchless client to be loaded on architectures
1833 without this export.
1837 Description: RedHat Update kernel for RHEL5
1838 Details : Add the kernel config file for RHEL5.
1842 Description: RedHat Update kernel for RHEL5
1843 Details : Modify the kernel config file more closer RHEL5.
1847 Description: Build failure against Centos5 (RHEL5)
1848 Details : Define PAGE_SIZE when it isn't present.
1852 Description: test_7 on sanity-quota.sh can't be used on separate machines
1853 Details : add facet to handle it
1857 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
1858 Details : with NFS, the anon dentry's parent was set to itself in
1859 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
1860 corresponding dentry other than use rec->ur_name.
1862 Severity : enhancement
1864 Description: lfs setstripe enhancement
1865 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
1869 Description: enable data checksumming by default
1870 Details : enable checksum by default, allow --disable-checksum
1871 configure option and "-o nochecksum" mount option. Checksums
1872 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
1876 Description: lustre support for RHEL5
1877 Details : Add support for RHEL5.
1881 Description: Client eviction due to failover config
1882 Details : after a connection loss, the lustre client should attempt to
1883 reconnect to the last active server first before trying the
1884 other potential connections.
1888 Description: when mds and osts use different quota unit(32bit and 64bit),
1889 quota will be released repeatly.
1890 Details : void sending multiple quota reqs to mds, which will keep the
1891 status between the reqs.
1895 Description: osts not allocated evenly to files
1896 Details : change the condition to increase offset_idx
1899 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
1901 Description: Data corruption for OSTs that are formatted larger than 2TB
1903 Details : When generating the bio request for lustre file writes the
1904 sector number would overflow a temporary variable before being
1905 used for the IO. The data reads correctly from Lustre (which
1906 will overflow in a similar manner) but other file data or
1907 filesystem metadata may be corrupted in some cases.
1911 Description: Tune the kernel for good SCSI performance.
1912 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
1913 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
1918 Description: rewrite ext2-derived code in obdclass/uuid.c
1919 Details : rewrite inherited code (uuid parsing code from ext2 utils)
1920 from scratch preserving functionality.
1924 Description: "lfs find -obd UUID" prints directories
1925 Details : "lfs find -obd UUID" will return all directory names instead
1926 of just file names. It is incorrect because the directories
1927 do not reside on the OSTs.
1931 Description: use i_size_read and i_size_write in 2.6 port
1932 Details : replace inode->i_size access with i_size_read/write()
1935 Frequency : when removing large files
1937 Description: scheduling issue during removal of large Lustre files
1938 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
1939 It causes scheduling issues when removing large files (17TB in the
1943 Frequency : only with liblustre clients on XT3
1945 Description: evictions taking too long
1946 Details : allow llrd to evict clients directly on OSTs
1949 Frequency : only on ppc
1951 Description: /proc/fs/lustre/devices broken on ppc
1952 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
1953 We should make sure the type of 'index' is loff_t and then cast
1954 explicitly as needed below. Do not assign an explicitly cast
1958 Frequency : only for rhel5
1960 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
1961 Details : Modify the target file & which_kernel.
1963 Severity : enhancement
1965 Description: omit set fsid for export NFS
1966 Details : fix set/restore device id for avoid EMFILE error and mark lustre
1967 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
1971 Frequency : Always, for kernels after 2.6.16
1972 Description: Fix warning idr_remove called for id=.. which is not allocated.
1973 Details : Last kernels save old s_dev before kill super and not allow
1974 to restore from callback - restore it before call kill_anon_super.
1978 Description: buffer overruns could theoretically occur
1979 Details : llapi_semantic_traverse() modifies the "path" argument by
1980 appending values to the end of the origin string, and a
1981 overrun may occur. Adding buffer overrun check in liblustreapi.
1985 Description: Fix error on 'ls .' at the top of the Lustre mount.
1986 Details : Don't revalidate dentry if it is a root dentry.
1990 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
1991 Details : Modify vm-tunables-rhel4.patch.
1995 Description: Kernel config for 2.6.18-vanilla.
1996 Details : Modify targets/2.6-vanilla.target.in.
1997 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
1998 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
1999 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
2000 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
2004 Description: improve handling recoverable errors
2005 Details : if request processig with error which can be recoverable on server
2006 request should be resend, otherwise page released from cache and
2011 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
2012 Details : Modify target file & which_patch.
2013 A flaw was found in the IA32 system call emulation provided
2014 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
2015 value could be stored in the %RAX register, which could trigger an
2016 out-of-bounds system call table access. An untrusted local user
2017 could exploit this flaw to run code in the kernel
2018 (ie a root privilege escalation). (CVE-2007-4573).
2022 Description: O_DIRECT bypasses client statistics.
2023 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
2024 read_ahead_stats were updated. copied the stats section from
2025 osc_send_oap_rpc() into async_internal().
2029 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
2033 Description: change order of libsysio includes
2034 Details : '#include sysio.h' should always come before '#include xtio.h'
2038 Description: handle "serious error: objid * already exists" more gracefully
2039 Details : If LAST_ID value on disk is smaller than the objects existing in
2040 the O/0/d* directories, it indicates disk corruption and causes an
2041 LBUG(). If the object is 0-length, then we should use the existing
2042 object. This will help to avoid a full fsck in most cases.
2044 Severity : enhancement
2046 Description: adapt the lustre_config script to support the upgrade case
2047 Details : Add "-u" option for lustre_config script to support upgrading 1.4
2048 server targets to 1.6 in parallel.
2052 Description: To avoid grant space > avaible space when the disk is almost
2053 full. Without this patch you might see the error "grant XXXX >
2054 available" or some LBUG about grant, when the disk is almost
2056 Details : In filter_check_grant, for non_grant cache write, we should
2057 check the left space by if (*left > ungranted + bytes), instead
2058 of (*left > ungranted), because only we are sure the left space
2059 is enough for another "bytes", then the ungrant space should be
2060 increase. In client, we should update cl_avail_grant only there
2061 is OBD_MD_FLGRANT in the reply.
2065 Description: Update RHEL 4 kernel to fix local root privilege escalation.
2066 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
2067 described in CVE-2007-4573. This problem could allow untrusted
2068 local users to gain root access.
2071 Frequency : when using O_DIRECT and quotas
2073 Description: Incorrect file ownership on O_DIRECT output files
2074 Details : block usage reported by 'lfs quota' does not take into account
2075 files that have been written with O_DIRECT.
2080 Description: touch file failed when fs is not full
2081 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
2082 otherwise we can get ENOSP while fs is not full.
2086 Description: parallel lock callbacks
2087 Details : Instead of sending blocking and completion callbacks as separated
2088 requests, adding them to a set and sending in parallel.
2091 Frequency : only for Cray XT3
2092 Bugzilla : 12829/13455
2093 Description: Changing primary group doesn't change the group lustre assigns to
2095 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
2096 overridden with the primary group provided by the group upcall,
2097 whereas the supplied fsgid can be trusted if it is in the list of
2098 supplementary groups returned by the group upcall.
2100 Severity : enhancement
2102 Description: Allow masking D_WARNING, D_ERROR messages from console
2103 Details : Console messages can now be disabled via lnet.printk.
2107 Description: User code with malformed file open parameter crashes client node
2108 Details : Before packing join_file req, all the related reference should be
2109 checked carefully in case some malformed flags cause fake
2110 join_file req on client.
2114 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
2115 Details : ldlm_completion_ast() assumes that a lock is granted when the req
2116 mode is equal to the granted mode. However, it should also check
2117 that LDLM_FL_CP_REQD is not set.
2121 Description: Heavy nfs access might result in deadlocks
2122 Details : After ELC code landed, it is now improper to enqueue any mds
2123 locks under och_sem, because enqueue might want to decide to
2124 cancel open locks for same inode we are holding och_sem for.
2128 Description: Client eviction while running blogbench
2129 Details : A lot of unlink operations with concurrent I/O can lead to a
2130 deadlock causing evictions. To address the problem, the number of
2131 oustanding OST_DESTROY requests is now throttled to
2132 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
2133 callbacks are processed in priority.
2137 Description: enable ACLs on MDS by default
2138 Details : ACLs must be enabled on MDS by default.
2141 Frequency : PPC/PPC64 only
2143 Description: conflicts between asm-ppc64/types.h and lustre_types.h
2144 Details : fix duplicated definitions between asm-ppc64/types.h and
2145 lustre_types.h on PPC.
2148 Frequency : PPC/PPC64 only
2150 Description: asm-ppc/segment.h does not exist
2151 Details : fix compile issue on PPC.
2155 Description: better handle error messages in extents code
2158 Frequency : RHEL4 only
2160 Description: mkfs is very slow on IA64/RHEL4
2161 Details : A performance regression has been discovered in the MPT Fusion
2162 driver between versions 3.02.73rh and 3.02.99.00rh. As a
2163 consequence, we have downgraded the MPT Fusion driver in the RHEL4
2164 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
2166 Severity : enhancement
2168 Description: SNMP support enhancement
2169 Details : Adding total number of sampled request for an MDS node in snmp
2172 Severity : enhancement
2174 Description: Optimize ldlm waiting list processing for PR extent locks
2175 Details : When processing waiting list for read extent lock and meeting read
2176 lock that is same or wider to it that is not contended, skip
2177 processing rest of the list and immediatelly return current
2178 status of conflictness, since we are guaranteed there are no
2179 conflicting locks in the rest of the list.
2183 Description: Time out and refuse to reconnect
2184 Details : When the failover node is the primary node, it is possible
2185 to have two identical connections in imp_conn_list. We must
2186 compare not conn's pointers but NIDs, otherwise we can defeat
2187 connection throttling.
2191 Description: port llog fixes from b1_6 into HEAD
2192 Details : Port llog reference couting and some llog cleanups from b1_6
2193 (bug 10800) into HEAD, for protect from panic and access to already
2194 free llog structures.
2198 Description: Detect stride IO mode in read-ahead
2199 Details : When a client does stride read, read-ahead should detect that and
2200 read-ahead pages according to the detected stride pattern.
2204 Description: data checksumming impacts single node performance
2205 Details : add support for several checksum algorithm. Currently, only CRC32
2206 and Adler-32 are supported. The checksum type can be changed on
2207 the fly via /proc/fs/lustre/osc/*/checksum_type.
2211 Description: use adler32 for page checksums
2212 Details : when available, use the Adler-32 algorithm instead of CRC32 for
2217 Description: build for x2 fails
2218 Details : fix compile issue on Cray systems.
2222 Description: Properly match for duplicate locks
2223 Details : Due to different lock order from skiplists code, we need to
2224 traverse entire list for now
2227 Frequency : only on PPC/SLES10
2229 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
2230 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
2231 be defined. Add a hack in mkfs_lustre.c to work around this compile
2236 Description: LASSERT on MDS when client holding flock lock dies
2237 Details : ldlm pool logic depends on number of granted locks equal to
2238 number of released locks which is not true for flock locks, so
2239 just exclude such locks from consideration.
2243 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
2244 Details : Use fsfilt_sync() for both the whole filesystem sync and
2245 individual file sync to eliminate dangerous inode locking
2246 with I_LOCK that can lead to a deadlock.
2250 Description: performance in 1.6.3
2251 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
2252 sure that 1MB requests are not fragmented by the block layer.
2254 Severity : enhancement
2256 Description: organize the server-side client stats on per-nid basis
2257 Details : Change the structure of stats under obdfilter and mds to
2265 The "uuid"s file would list the uuids of _active_ exports.
2266 And the clear entry is to clear all stats and stale nids.
2268 Severity : enhancement
2270 Description: eliminate client locks in face of contention
2271 Details : file contention detection and lockless i/o implementation
2272 for contended files.
2276 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
2279 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
2281 Description: MDS may lose file striping (and hence file data) in some cases
2282 Details : If there are additional extended attributes stored on the MDS,
2283 in particular ACLs, SELinux, or user attributes (if user_xattr
2284 is specified for the client mount options) then there is a risk
2285 of attribute loss. Additionally, the Lustre file striping
2286 needs to be larger than default (e.g. striped over all OSTs),
2287 and an additional attribute must be stored initially in the
2288 inode and then increase in size enough to be moved to the
2289 external attribute block (e.g. ACL growing in size) for file
2294 Description: skiplist implementation simplification
2295 Details : skiplists are used to group compatible locks on granted list
2296 that was implemented as tracking first and last lock of each
2297 lock group the patch changes that to using doubly linked lists
2301 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
2302 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
2303 because LASSERT on both the data supplied by a client, and the
2304 data on disk is dangerous and incorrect.
2306 Severity : enhancement
2308 Description: Slow truncate/writes to huge files at high offsets.
2309 Details : Directly associate cached pages to lock that protect those pages,
2310 this allows us to quickly find what pages to write and remove
2311 once lock callback is received.
2315 Description: more ldlm soft lockups
2316 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
2317 starve other threads from the resource lock for a long time in
2318 case of long waiting queue, so change the debug level from
2319 D_OTHER to the less frequently used D_INFO.
2321 Severity : enhancement
2323 Description: add -gid, -group, -uid, -user options to lfs find
2327 Description: Hung threads in invalidate_inode_pages2_range
2328 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
2329 once one is completed. As a result, some RPCs are stuck in the
2330 queue and are never sent.
2334 Description: filter threads hungs on waiting journal commit
2335 Details : Cleanup filter group llog code, then only filter group llog will
2336 be only created in the MDS/OST syncing process.
2340 Description: Procfs and llog threads access destoryed import sometimes.
2341 Details : Sync the import destoryed process with procfs and llog threads by
2342 the import refcount and semaphore.
2344 Severity : enhancement
2346 Description: openlock cache of b1_6 port to HEAD
2351 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
2352 Details : If insertion of an extent fails, then discard the inode
2353 preallocation and free data blocks else it can lead to duplicate
2358 Description: don't always update ctime in ext3_xattr_set_handle()
2359 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
2360 In some cases the ctime should not be updated, for example for
2361 2.0->1.8 compatibility it is necessary to delete an xattr and it
2362 should not update the ctime.
2366 Bugzilla : 15713/16362
2367 Description: Assertion in iopen_connect_dentry in 1.6.3
2368 Details : looking up an inode via iopen with the wrong generation number can
2369 populate the dcache with a disconneced dentry while the inode
2370 number is in the process of being reallocated. This causes an
2371 assertion failure in iopen since the inode's dentry list contains
2372 both a connected and disconnected dentry.
2376 Description: assertion failure in ldlm_handle2lock()
2377 Details : fix a race between class_handle_unhash() and class_handle2object()
2378 introduced in lustre 1.6.5 by bug 13622.
2383 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
2384 Details : remove the limit of 256 scsi disks in the sd_iostat patch
2389 Description: oops in sd_iostats_seq_show()
2390 Details : unloading/reloading the scsi low level driver triggers a kernel
2391 bug when trying to access the sd iostat file.
2396 Description: Kernel panics during QLogic driver reload
2397 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
2398 patch, causing memory corruption.
2403 Description: journal_dev option does not work in b1_6
2404 Details : pass mount option during pre-mount.
2406 Severity : enhancement
2408 Description: Add a FIEMAP(FIle Extent MAP) ioctl
2409 Details : FIEMAP ioctl will allow an application to efficiently fetch the
2410 extent information of a file. It can be used to map logical blocks
2411 in a file to physical blocks in the block device.
2415 Description: LDLM soft lockups - improvement
2416 Details : It is be possible to send the lock handle along with each read
2417 or write request because the client is already doing a lock match
2418 itself so there isn't any reason the OST should have to re-do that
2422 Frequency : only on Cray X2
2424 Description: X2 build failures
2425 Details : fix build failures on Cray X2.
2429 Description: xid & resent requests
2430 Details : Initialize RPC XID from clock at startup (randomly if clock is
2433 Severity : enhancement
2435 Description: Add lustre_start utility to start or stop multiple Lustre servers
2440 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
2441 Details : In case of memory pressure, list_del() can be called twice on
2442 req->rq_history_list, causing a kernel oops.
2446 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
2447 Details : fix stack overflow in the distributed lock manager by defering
2448 export eviction after a failed AST to the elt thread instead of
2449 handling it in the dlm interpret routine.
2453 Description: Convert some comments to new format.
2454 Details : Update documenting comments to match doxygen conventions.
2458 Description: Grammar fixes.
2459 Details : A couple of trivial spelling fixes.
2463 Description: OSD_COUNTERS-mandatory
2464 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
2465 to simplify the code.
2469 Description: simplify cmm_device freeing logic.
2470 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
2474 Description: Add lockdep support to dt_object_operations locking interface.
2475 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
2476 parameter indicating lock ordering. Update mdd code to use new
2481 Description: Introduce failloc constants for lockless IO tests.
2482 Details : Add two new failloc constants to test lockless IO. Only one of
2483 them in implemented---another is checked in yet to be landed
2488 Description: Add lockdep support for inode mutex.
2489 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
2490 in the situations where Lustre has to lock more than one inode
2495 Description: Add optional invariants checking support.
2496 Details : Add new LINVRNT() macro, optional on new --enable-invariants
2497 configure switch. This macro is to be used for consistency and
2498 sanity checks that are too expensive to be left in `production'
2503 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
2504 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
2509 Description: Add lockdep support for ldlm_lock and ldlm_resource.
2510 Details : Use spin_lock_nested() in (the only) situation where more than
2511 one ldlm_lock is locked simultaneously. Also, fix possible
2512 dead-lock in ldlm_lock_change_resource() by enforcing particular
2517 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
2518 Details : Instead of specifying each ldlm_lock call-back through separate
2519 parameter, wrap them into struct ldlm_callback_suite.
2523 Description: Kill join_lru obd method and its callers.
2524 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
2525 in memory---a requirement implemented through ->o_join_lru() obd
2526 method in HEAD. Remove this method and its users.
2530 Description: Add asynchronous ldlm ENQUEUE completion handler.
2531 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
2532 case that stock ldlm_completion_ast() cannot handle as it waits
2533 until lock is granted. Introduce new ldlm_completion_ast_async()
2534 for this. Also comment ldlm_completion_ast().
2538 Description: ldlm_error <-> errno conversion.
2539 Details : Add functions to map (rather arbitrary) between LDLM error codes
2540 and standard errno values. CLIO needs this to prevent LDLM specific
2541 constants from escaping ldlm and osc.
2545 Description: Kill unused ldlm_handle2lock_ns() function.
2546 Details : Kill unused ldlm_handle2lock_ns() function.
2550 Description: Add lu_ref support to ldlm_lock
2551 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
2552 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
2553 and ldlm_resource. LDLM interface has to be changed, because code that
2554 releases a reference on a lock, has to "know" what reference this is.
2555 In the most frequent case
2557 lock = ldlm_handle2lock(handle);
2559 LDLM_LOCK_PUT(lock);
2561 no changes are required. When any other reference (received _not_ from
2562 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
2563 instead of LDLM_LOCK_PUT().
2565 Arguably, changes are pervasive, and interface requires some discipline
2566 for proper use. On the other hand, it was very instrumental in finding
2567 a few leaked lock references.
2571 Description: Add ldlm_lock_addref_try().
2572 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
2573 attempts to addref a lock that might be being canceled
2578 Description: Add ldlm_weigh_callback().
2579 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
2580 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
2581 hard-coded `number of pages' logic.
2585 Description: Add lockdep annotations to llog code.
2586 Details : Use appropriately tagged _nested() locking calls in the places
2587 where llog takes more than one ->lgh_lock lock.
2591 Description: Add loi_kms_set().
2592 Details : Wrap kms updates into a helper function.
2596 Description: Constify instances of struct lsm_operations.
2597 Details : Constify instances of struct lsm_operations.
2601 Description: lu_conf support.
2602 Details : On a server, a file system object is uniquely identified
2603 by a fid, which is sufficient to locate and load all object
2604 state (inode). On a client, on the other hand, more data are
2605 necessary instantiate an object. Change lu_object_find() and
2606 friends to take additional `lu_conf' argument describing object.
2607 Typically this includes layout information.
2611 Description: lu_context fixes.
2612 Details : Introduce new lu_context functions that are needed on the client
2613 side, where some system threads (ptlrpcd) are shared by multiple
2614 modules, and so cannot be stopped during module shutdown.
2618 Description: Add start and stop methods to lu_device_type_operations.
2619 Details : Introduce two new methods in lu_device_type_operations, that are
2620 invoked when first instance of a given type is created and last one
2621 is destroyed respectively. This is need by CLIO.
2625 Description: Add lu_ref support to struct lu_device.
2626 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
2627 track leaked references.
2631 Description: Introduce lu_kmem_descr.
2632 Details : lu_kmem_descr and its companion interface allow to create
2633 and destroy a number of kmem caches at once.
2637 Description: Fix lu_object finalization race.
2638 Details : Fix a race between lu_object_find() finding an object and its
2639 concurrent finalization. This race is (most likely) not possible
2640 on the server, but might happen on the client.
2644 Description: Introduce lu_ref interface.
2645 Details : lu_ref is a debugging module allowing to track references to
2646 a given object. It is quite cpu expensive, and has to be
2647 explicitly enabled with --enable-lu_ref. See usage description
2652 Description: Factor lu_site procfs stats into a separate function.
2653 Details : Separate lu_site stats printing code into a separate function
2654 that can be reused on a client.
2658 Description: Constify instances of struct {lu,dt,md}_device_operations.
2659 Details : Constify instances of struct {lu,dt,md}_device_operations.
2663 Description: Introduce struct md_site and move meta-data specific parts of
2664 struct lu_site here.
2665 Details : Move md-specific fields out of struct lu_site into special struct
2666 md_site, so that lu_site can be used on a client.
2670 Description: Kill mdd_lov_destroy().
2671 Details : Remove unused mdd code.
2675 Description: Add st_block checking to multistat.c.
2676 Details : Add st_block checking to multistat.c.
2680 Description: Add lu_ref support to struct obd_device.
2681 Details : Add lu_ref tracking to obd_device.
2685 Description: Kill obd_set_fail_loc().
2686 Details : Remove unused code.
2690 Description: Add special type for ptlrpc_request interpret functions.
2691 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
2692 there. Actual usage will be in CLIO.
2696 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
2697 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
2698 doesn't work with lockdep.
2702 Description: Add rwv.c test program.
2703 Details : New testing program exercising readv(2) and writev(2) (Qian).
2707 Description: Add sendfile.c test program.
2708 Details : New testing program exercising sendfile(2) (Jay).
2712 Description: Ratelimit a message that can be very frequent.
2713 Details : Ratelimit a memory allocation failure message that can
2718 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2719 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2723 Description: Liblustre build fixes.
2724 Details : Liblustre build fixes.
2728 Description: libcfs: add cfs_{need,cond}_resched() interface.
2729 Details : libcfs: add cfs_{need,cond}_resched() definition and
2730 implementations for Linux, NT, and liblustre.
2732 Severity : enhancement
2734 Description: More exported tunables for mballoc
2735 Details : Add support for tunable preallocation window and new tunables for
2736 large/small requests
2740 Description: Detect corruption of block bitmap and checking for preallocations
2741 Details : Checks validity of on-disk block bitmap. Also it does better
2742 checking of number of applied preallocations. When corruption is
2743 found, it turns filesystem readonly to prevent further corruptions.
2747 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
2748 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
2749 window to 0. If it does hit cache miss, and read-pattern is still
2750 stride-io mode, does not reset the stride window, but also does not
2751 increase the stride window length in this case.
2755 Frequency : only for big-endian servers
2756 Description: Check if system is big-endian while mounting fs with extents feature
2757 Details : Mounting a filesystem with extents feature will fail on big-endian
2758 systems since ext3-based ldiskfs is not supported on big-endian
2759 systems. This can be over-riden with "bigendian_extents" mount option.
2761 Severity : enhancement
2763 Description: The root squash functionality
2764 Details : A security feature, which is to prevent users from being able
2765 to mount lustre on their desktop, run as root, and delete
2766 all of the files in the filesystem. The goal is accomplished by
2767 remapping user id (UID) and group id (GID) of the root user to
2768 a UID and GID specified by the system administartor via Lustre
2769 configuration management server (MGS). The functionality also
2770 allows to specify sets of clients for which the remapping does
2775 Description: Excessive recovery window
2776 Details : With AT enabled, the recovery window can be excessively long (6000+
2777 seconds). To address this problem, we no longer use
2778 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
2779 timeout no longer depends on the service time, it is set to
2780 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
2781 time via pb_service_time.
2785 Description: Watchdog triggered on MDS failover
2786 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
2787 the OSTs know that the MDS "UUID" can be reused for the same export
2788 from a different NID, so we do not need to wait for the export to be
2792 Frequency : rare, only if using MMP with Linux RAID
2794 Description: MMP doesn't work with Linux RAID
2795 Details : While using HA for Lustre servers with Linux RAID, it is possible
2796 that MMP will not detect multiple mounts. To make this work we
2797 need to unplug the device queue in RAID when the MMP block is being
2798 written. Also while reading the MMP block, we should read it from
2799 disk and not the cached one.
2801 Severity : enhancement
2803 Description: open file using fid
2804 Details : A file can be opened using just its fid, like
2805 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
2808 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
2811 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
2812 Details : add proper locking for creation of jbd_4k slab cache
2816 Description: MMP check in ext3_remount() fails without displaying any error
2817 Details : When multiple mount protection fails during remount, proper error
2820 Severity : enhancement
2822 Description: Allow stripe size to be up to 4G-64k
2823 Details : Fix math logic to allow large stripe sizes.
2827 Description: add check for >8TB ldiskfs filesystems
2828 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
2829 Don't allow >8TB ldiskfs filesystems to be mounted without
2830 force_over_8tb mount option
2835 Descriptoin: OSS complains "too many blocks" and crashes
2836 Details : In filter_commitrw_write():
2837 1. lquota_chkquota();
2838 2. obdo_from_inode();
2839 3. lquota_pending_commit();
2840 "2." has changed uid and gid in this case, then caused
2841 lquota_chkquota() and lquota_pending_commit() to use
2842 different uid and gid. That is the root of the bug.
2846 Descriptoin: getstripe doesn't report stripe count = -1 sanely
2847 Details : Detect and print out the 'stripe count = -1' case properly.
2851 Description: Out or order replies might be lost on replay
2852 Details : In ptlrpc_retain_replayable_request if we cannot find retained
2853 request with tid smaller than one currently being added, add it
2854 to the start, not end of the list.
2856 --------------------------------------------------------------------------------
2858 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
2860 * Support for kernels:
2861 2.6.5-7.283 (SLES 9),
2862 2.6.9-55.EL (RHEL 4),
2863 2.6.16.46-0.14 (SLES 10),
2864 2.6.18.8 vanilla (kernel.org)
2865 * Client support for unpatched kernels:
2866 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2867 2.6.16 - 2.6.22 vanilla (kernel.org)
2868 * Due to recently discovered recovery problems, we do not recommend
2869 using patchless RHEL 4 clients with this or any earlier release.
2870 * Recommended e2fsprogs version: 1.39.cfs8
2871 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2872 * Starting with this release, the ldiskfs backing filesystem required
2873 by Lustre is now in its own package, lustre-ldiskfs. This package
2874 should be installed. It is versioned separately from Lustre and
2875 may be released separately in future.
2879 Description: block reactivating mgc import until all deactivates complete
2880 Details : Fix race when failing back MDT/MGS to itself (testing)
2883 Frequency : at statup only
2885 Description: mds_lov_synchronize race leads to various problems
2886 Details : simultaneous MDT->OST connections at startup can cause the
2887 sync to abort, leaving the OSC in a bad state.
2889 Severity : enhancement
2891 Description: add optional extra BUILD_VERSION info
2892 Details : add a new environment variable (namely LUSTRE_VERS) which allows
2893 to override the lustre version.
2896 Frequency : 2.6.18 servers only
2898 Description: ll_kern_mount() doesn't release the module reference
2899 Details : The ldiskfs module reference count never drops down to 0
2900 because ll_kern_mount() doesn't release the module reference.
2905 Description: server LBUG when using old ost_num_threads parameter
2906 Details : Accept the old ost_num_threads parameter but warn that it
2907 is deprecated, and fix an off-by-one error that caused an LBUG.
2912 Description: Transient SCSI error results in persistent IO issue
2913 Details : iobuf->dr_error is not reinitialized to 0 between two
2917 Frequency : sometimes when underlying device returns I/O errors
2919 Description: OSTs not going read-only during write failures
2920 Details : OSTs are not remounted read-only when the journal commit threads
2921 get I/O errors because fsfilt_ext3 calls journal_start/stop()
2922 instead of the ext3 wrappers.
2926 Description: poor connect scaling with increasing client count
2927 Details : Don't run filter_grant_sanity_check for more than 100 exports
2928 to improve scaling for large numbers of clients.
2931 Frequency : SLES10 only
2933 Description: sanity-quota.sh quotacheck failed: rc = -22
2934 Details : Quotas cannot be enabled on SLES10.
2937 Frequency : liblustre clients only
2939 Description: getdirentries does not give error when run on compute nodes
2940 Details : getdirentries does not fail when the size specified as an argument
2941 is too small to contain at least one entry
2943 Severity : enhancement
2945 Description: Add LNET router traceability for debug purposes
2946 Details : If a checksum failure occurs with a router as part of the
2947 IO path, the NID of the last router that forwarded the bulk data
2948 is printed so it can be identified.
2953 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
2954 Details : Due to a race condition, liblustre clients were occasionally
2955 evicted incorrectly.
2957 Severity : enhancement
2959 Description: lfs setstripe use optional parameters instead of postional
2962 Severity : enhancement
2964 Description: Nanosecond timestamp support for ldiskfs
2965 Details : The on-disk ldiskfs filesystem has added support for nanosecond
2966 resolution timestamps. There is not yet support for this at
2967 the Lustre filesystem level.
2970 Frequency : during server recovery
2972 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
2973 Details : request with rq_no_resend flag not awake l_wait_event if they get
2977 Frequency : nfs export on patchless client
2979 Description: connectathon hang when test nfs export over patchless client
2980 Details : Disconnected dentry cannot be found with lookup, so we do not need
2981 to unhash it or make it invalid
2984 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
2985 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
2986 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
2991 Description: open req refcounting wrong on reconnect
2992 Details : If reconnect happened between getting open reply from server and
2993 call to mdc_set_replay_data in ll_file_open, we will schedule
2994 replay for unreferenced request that we are about to free.
2995 Subsequent close will crash in variety of ways.
2996 Check that request is still eligible for replay in
2997 mdc_set_replay_data().
3002 Description: disable writes to filesystem when reading health_check file
3003 Details : the default for reading the health_check proc file has changed
3004 to NOT do a journal transaction and write to disk, because this
3005 can cause reads of the /proc file to hang and block HA state
3006 checking on a healthy but otherwise heavily loaded system. It
3007 is possible to return to the previous behaviour during configure
3008 with --enable-health-write.
3010 Severity : enhancement
3012 Description: 64-bit inode version
3013 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
3014 to the inode. This will be required for version-based recovery.
3019 Description: MDS fails to start if a duplicate client export is detected
3020 Details : in some rare cases it was possible for a client to connect to
3021 an MDS multiple times. Upon recovery the MDS would detect this
3022 and fail during startup. Handle this more gracefully.
3024 Severity : enhancement
3026 Description: Add -o localflock option to simulate old noflock
3028 Details : This will achieve local-only flock/fcntl locks
3034 Description: log_commit_thread vs filter_destroy race leads to crash
3035 Details : Take import reference before releasing llog record semaphore
3040 Description: Wrong request locking in request set processing
3041 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
3042 imp_delayed_list, in this place should be used imp_lock.
3045 Frequency : when reconnection
3047 Description: Grant Leak when osc reconnect to OST
3048 Details : When osc reconnect ost, OST(filter) should check whether it
3049 should grant more space to client by comparing fed_grant and
3050 cl_avail_grant, and return the granted space to client instead
3051 of "new granted" space, because client will call osc_init_grant
3052 to update the client grant space info.
3055 Frequency : when client reconnect to OST
3057 Description: Grant Leak when osc do resend and replay bulk write
3058 Details : When osc reconnect to OST, OST(filter)should clear grant info of
3059 bulk write request, because the grant info while be sync between
3060 OSC and OST when reconnect, and we should ignore the grant info
3061 these of resend/replay write req.
3066 Description: Grant space more than available left space sometimes.
3067 Details : When then OST is about to be full, if two bulk writing from
3068 different clients came to OST. Accord the avaliable space of the
3069 OST, the first req should be permitted, and the second one
3070 should be denied by ENOSPC. But if the seconde arrived before
3071 the first one is committed. The OST might wrongly permit second
3072 writing, which will cause grant space > available space.
3075 Frequency : when client is evicted
3077 Description: Grant might be wrongly erased when osc is evicted by OST
3078 Details : when the import is evicted by server, it will fork another
3079 thread ptlrpc_invalidate_import_thread to invalidate the
3080 import, where the grant will be set to 0. While the original
3081 thread will update the grant it got when connecting. So if
3082 the former happened latter, the grant will be wrongly errased
3083 because of this race.
3088 Description: Checking Stale with correct fid
3089 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
3090 to check whether it is stale, because sometimes, we want the
3091 enqueue happened anyway, and op_data.fid2 will not be initialized.
3093 Severity : enhancement
3095 Description: update patchless client
3096 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
3099 Frequency : only with 2.4 kernel
3101 Description: random memory corruption
3102 Details : size of struct ll_inode_info is to big for union inode.u and this
3103 can be cause of random memory corruption.
3108 Description: Memory leak in recovery
3109 Details : Lov_mds_md was not free in an error handler in mds_create_object.
3110 It should also check obd_fail before fsfilt_start, otherwise if
3111 fsfilt_start return -EROFS,(failover mds during mds recovery).
3112 then the req will return with repmsg->transno = 0 and rc = EROFS.
3113 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
3114 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
3115 be freed no matter whether fsfilt_commit success or not.
3118 Frequency : only with huge count clients
3120 Description: Prevents from taking the superblock lock in llap_from_page for
3122 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
3123 for llap_from_page call in ll_removepage prevents from taking the
3124 superblock lock for a soon died page.
3129 Description: Not check open intent error before release open handle
3130 Details : in some rare cases, the open intent error is not checked before
3131 release open handle, which may cause
3132 ASSERTION(open_req->rq_transno != 0), because it tries to release
3133 the failed open handle.
3138 Description: Set cat log bitmap only after create log success.
3139 Details : in some rare cases, the cat log bitmap is set too early. and it
3140 should be set only after create log success.
3144 Description: Accessing a block bevice can re-enable I/O when Lustre is
3145 tearing down a device.
3146 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
3151 Description: service threads may hog cpus when there are a lot of requests
3152 Details : Insert cond_resched to give other threads a chance to use some CPU
3157 Description: the cat log was not initialized in recovery
3158 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
3159 unlink log on mds will not be initialized until mds post
3160 recovery. And also in mds post recovery, the unlink log will
3161 initialization will be done asynchronausly, so there will be race
3162 between add unlink log and unlink log initialization.
3166 Description: brw_stats were being printed incorrectly
3167 Details : brw_stats were being printed as log2 but all of them were not
3168 recorded as log2. Also remove some code duplication arising from
3169 filter_tally_{read,write}.
3173 Frequency : rare, only in recovery.
3174 Description: ASSERTION(req->rq_type != LI_POISON) failed
3175 Details : imp_lock should be held while iterating over imp_sending_list for
3176 prevent destroy request after get timeout in ptlrpc_queue_wait.
3180 Description: replay-single.sh test 52 fails
3181 Details : A lock's skiplist need to be cleanup when it being unlinked
3182 from its resource list.
3186 Description: Short directio read returns full requested size rather than
3188 Details : Direct I/O operations should return actual amount of bytes
3189 transferred rather than requested size.
3191 Severity : enhancement
3193 Description: metadata RPC reduction (e.g. for rm performance)
3194 Details : decrease the amount of synchronous RPC between clients and servers
3195 by canceling conflicing lock before the operation on the client
3196 and packing thier handles into the main operation RPC to server.
3198 Severity : enhancement
3200 Description: Async OSC create to avoid the blocking unnecessarily.
3201 Details : If an OST has no remaining object, system will block on the
3202 creation when it needs to create a new object on this OST. Now,
3203 ways use pre-created objects when available, instead of blocking on
3204 an empty osc while others are not empty. If we must block, we
3205 block for the shortest possible period of time.
3209 Description: improve handling recoverable errors
3210 Details : if request processig with error which can be recoverable on server
3211 request should be resend, otherwise page released from cache and
3214 Severity : enhancement
3216 Description: refine locking for avoid write wrong info into lov_objid file
3217 Details : fix possible races with add new target and write/update data in
3220 --------------------------------------------------------------------------------
3222 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
3227 Frequency : on some architectures
3229 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
3230 Details : Uninitialized flags sometimes cause configuration commands to
3234 Frequency : patchless clients only
3236 Description: missing __iget() symbol export
3237 Details : The __iget() symbol export is missing. To avoid the need for
3238 this on patchless clients the deathrow inode reaper is turned
3239 off, and we depend on the VM to clean up old inodes. This
3240 dependency was during via the fix for bug 12181.
3242 --------------------------------------------------------------------------------
3244 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
3246 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
3247 INTEROPERATE with older versions automatically. In many cases a
3248 special upgrade step is needed. Please read the
3249 user documentation before upgrading any part of a live system.
3250 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
3251 version will not interoperate with 1.6 betas before beta5 (1.5.95).
3252 * WARNING: Lustre configuration and startup changes are required with
3253 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
3257 Severity : enhancement
3259 Description: MountConf
3260 Details : Lustre configuration is now managed via mkfs and mount
3261 commands instead of lmc and lconf. New obd types (MGS, MGC)
3262 are added for dynamic configuration management. See
3263 https://mail.clusterfs.com/wikis/lustre/MountConf for
3266 Severity : enhancement
3268 Description: dynamic OST addition
3269 Details : OSTs can now be added to a live filesystem
3271 Severity : enhancement
3273 Description: startup order invariance
3274 Details : MDTs and OSTs can be started in any order. Clients only
3275 require the MDT to complete startup.
3277 Severity : enhancement
3279 Description: parallel, asynchronous orphan cleanup
3280 Details : orphan cleanup is now performed in separate threads for each
3281 OST, allowing parallel non-blocking operation.
3283 Severity : enhancement
3285 Description: optimized stripe assignment
3286 Details : stripe assignments are now made based on ost space available,
3287 ost previous usage, and OSS previous usage, in order to try
3288 to optimize storage space and networking resources.
3290 Severity : enhancement
3292 Description: Permanently set tunables
3293 Details : All writable /proc/fs/lustre tunables can now be permanently
3294 set on a per-server basis, at mkfs time or on a live system.
3296 Severity : enhancement
3298 Description: Lustre message v2
3299 Details : Add lustre message format v2.
3301 Severity : enhancement
3303 Description: client OST exclusion list
3304 Details : Clients can be started with a list of OSTs that should be
3305 declared "inactive" for known non-responsive OSTs.
3309 Description: SPEC SFS validation failure on NFS v2 over lustre.
3310 Details : Changes the blocksize for regular files to be 2x RPC size,
3311 and not depend on stripe size.
3313 Severity : enhancement
3315 Description: Multiple MD RPCs in flight.
3316 Details : Further unserialise some read-only MDS RPCs - learn about intents.
3317 To avoid overly-overloading MDS, introduce a limit on number of
3318 MDS RPCs in flight for a single client and add /proc controls
3319 to adjust this limit.
3321 Severity : enhancement
3323 Description: client read/write statistics
3324 Details : Add client read/write call usage stats for performance
3325 analysis of user processes.
3326 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
3327 file access. extents_stats shows chunk size distribution.
3328 extents_stats_per_process show chunk size distribution per
3331 Severity : enhancement
3333 Description: mds statistics
3334 Details : Add detailed mds operations statistics in
3335 /proc/fs/lustre/mds/*/stats.
3339 Description: Failure of copying files with lustre special EAs.
3340 Details : Client side always return success for setxattr call for lustre
3341 special xattr (currently only "trusted.lov").
3346 Description: Refcount LNET uuids
3347 Details : The global LNET uuid list grew linearly with every startup;
3348 refcount repeated list entries instead of always adding to
3351 Severity : enhancement
3353 Description: Dynamic service threads
3354 Details : Within a small range, start extra service threads
3355 automatically when the request queue builds up.
3358 Frequency : mixed-endian client/server environments
3360 Description: mixed-endian crashes
3361 Details : The new msg_v2 system had some failures in mixed-endian
3364 Severity : enhancement
3366 Description: Easy OST removal
3367 Details : OSTs can be permanently deactivated with e.g. 'lctl
3368 conf_param lustre-OST0001.osc.active=0'
3370 Severity : enhancement
3372 Description: MGS proc entries
3373 Details : Added basic proc entries for the MGS showing what filesystems
3376 Severity : enhancement
3378 Description: provide MGS failover
3379 Details : Added config lock reacquisition after MGS server failover.
3381 Severity : enhancement
3383 Description: add Linux 2.4 support
3384 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
3388 Description: a large application tries to do I/O to the same resource and dies
3389 in the middle of it.
3390 Details : Check the req->rq_arrival time after the call to
3391 ost_brw_lock_get(), but before we do anything about
3392 processing it & sending the BULK transfer request. This
3393 should help move old stale pending locks off the queue as
3394 quickly as obd_timeout.
3397 Frequency : when an incorrect nid is specified during startup
3399 Description: ptlrpc connect to non-existant node causes kernel crash
3400 Details : LNET can't be re-entered from an event callback, which
3401 happened when we expire a message after the export has been
3402 cleaned up. Instead, hand the zombie cleanup off to another
3405 Severity : enhancement
3407 Description: plain/inodebits lock performance improvement
3408 Details : Grouping plain/inodebits in granted list by their request modes
3409 and bits policy, thus improving the performance of search through
3413 Frequency : only if OST filesystem is corrupted
3415 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3416 Details : for a short time RPCs with bulk IO are in the replay list,
3417 but replay of bulk IOs is unimplemented. If the OST filesystem
3418 is corrupted due to disk cache incoherency and then replay is
3419 started it is possible to trip an assertion. Avoid putting
3420 committed RPCs into the replay list at all to avoid this issue.
3423 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3425 Description: System hang on startup
3426 Details : This bug allowed the liblustre (e.g. catamount) client to
3427 return to the app before handling all startup RPCs. This
3428 could leave the node unresponsive to lustre network traffic
3429 and manifested as a server ptllnd timeout.
3431 Severity : enhancement
3433 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3434 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
3435 boolean to control whether to print peer debug info when a
3436 client's RPC times out.
3439 Frequency : only for kernels with patches from Lustre below 1.4.3
3441 Description: Remove old rdonly API
3442 Details : Remove old rdonly API which unused from at least lustre 1.4.3
3445 Frequency : only for devices with external journals
3447 Description: Set external device read-only also
3448 Details : During a commanded failover stop, we set the disk device
3449 read-only while the server shuts down. We now also set any
3450 external journal device read-only at the same time.
3453 Frequency : when upgrading from 1.4 while trying to change parameters
3455 Description: The wrong (new) MDC name was used when setting parameters for
3456 upgraded MDT's. Also allows changing of OSC (and MDC)
3457 parameters if --writeconf is specified at tunefs upgrade time.
3460 Frequency : when setting specific ost indicies
3462 Description: QOS code breaks on skipped indicies
3463 Details : Add checks for missing OST indicies in the QOS code, so OSTs
3464 created with --index need not be sequential.
3466 Severity : enhancement
3468 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3469 Details : The uninit_groups feature works in conjunction with the kernel
3470 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3471 up the pass1 processing of e2fsck. This is a read-only feature
3472 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3473 mount filesystems that have had this feature enabled.
3475 Severity : enhancement
3477 Description: Improve multi-block allocation algorithm to avoid fragmentation
3478 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3479 allocation locality and avoid filesystem fragmentation.
3481 ------------------------------------------------------------------------------
3483 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
3485 * Support for kernels:
3486 2.4.21-47.0.1.EL (RHEL 3)
3487 2.6.5-7.283 (SLES 9)
3488 2.6.9-42.0.10.EL (RHEL 4)
3489 2.6.12.6 vanilla (kernel.org)
3490 2.6.16.27-0.9 (SLES 10)
3491 * Recommended e2fsprogs version: 1.39.cfs5
3493 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3497 Frequency : occasional, depends on client load and configuration
3498 Bugzilla : 12181, 12203
3499 Description: data loss for recently-modified files
3501 Details : In some cases it is possible that recently written or created
3502 files may not be written to disk in a timely manner (this should
3503 normally be within 30s unless client IO load is very high).
3504 The problem appears as zero-length files or files that are a
3505 multiple of 1MB in size after a client crash or client eviction
3506 that are missing data at the end of the file.
3508 This problem is more likely to be hit on clients where files are
3509 repeatedly created and unlinked in the same directory, clients
3510 have a large amount of RAM, have many CPUs, the filesystem has
3511 many OSTs, the clients are rebooted frequently, and/or the files
3512 are not accessed by other nodes after being written.
3514 The presence of the problem can be detected by looking at
3515 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
3516 smaller than the second (nr_unused) then dirty files will not
3517 be flushed automatically to disk. "sync; sleep 10" should be
3518 run several times on the node before unmounting it to update
3519 Lustre (this is also safe to run on nodes without this problem).
3521 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
3522 that can cause this same problem, so customers using that kernel
3523 also need to update the kernel in addition to Lustre. In order
3524 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
3526 It is normal that files written just before a client crash (less
3527 than 30s) may not yet have been flushed to disk, even for local
3531 Frequency : frequent on thin XT3 nodes
3533 Description: UUID collision on thin XT3 Linux nodes
3534 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
3535 randomly, since we relied on an insufficiently-seeded PRNG.
3540 Description: OSS hangs after "All ost request buffers busy"
3541 Details : A deadlock between quota and journal operations caused OSS
3542 hangs after printing "All ost request buffers busy."
3545 Frequency : always on liblustre builds
3547 Description: Cleanup compiler warnings on liblustre
3550 Frequency : always on liblustre builds on XT3
3552 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
3556 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
3557 > 32000 subdirectories
3558 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
3559 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
3560 subdirectory count crosses 32000. This will aid e2fsck to
3561 correctly handle more than 32000 subdirectories.
3564 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3566 Description: System hang on startup
3567 Details : This bug allowed the liblustre (e.g. catamount) client to
3568 return to the app before handling all startup RPCs. This
3569 could leave the node unresponsive to lustre network traffic
3570 and manifested as a server ptllnd timeout.
3572 Severity : enhancement
3574 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3575 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
3576 boolean to control whether to print peer debug info when a
3577 client's RPC times out.
3582 Description: make O_SYNC working on 2.6 kernels
3583 Details : 2.6 kernels use different method for mark pages for write,
3584 so need add a code to lustre for O_SYNC work.
3589 Description: Failure to close file and release space on NFS
3590 Details : Put inode details into lock acquired in ll_intent_file_open.
3591 Use mdc_intent_lock in ll_intent_open to properly
3592 detect all kind of errors unhandled by mdc_enqueue.
3597 Description: proc file read during shutdown sometimes raced obd removal,
3599 Details : Add lock to prevent obd access after proc file removal.
3602 Frequency : Only for files larger than 4GB on 32-bit clients.
3604 Description: improperly doing page alignment of locks
3605 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
3606 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3611 Description: RPCs being resent when they shouldn't be
3612 Details : Some RPCs that should not be resent are being resent. This
3613 can cause inconsistencies in the RPC state machine. Do not
3614 resend such requests.
3617 Frequency : rare, only with NFS export
3619 Description: Crash on NFS re-export node
3620 Details : under very unusual load conditions an assertion is hit in
3621 ll_intent_file_open()
3624 Frequency : only if OST filesystem is corrupted
3626 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3627 Details : for a short time RPCs with bulk IO are in the replay list,
3628 but replay of bulk IOs is unimplemented. If the OST filesystem
3629 is corrupted due to disk cache incoherency and then replay is
3630 started it is possible to trip an assertion. Avoid putting
3631 committed RPCs into the replay list at all to avoid this issue.
3636 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3637 Details : Large single O_DIRECT read and write calls can fail to allocate
3638 a sufficiently large buffer to process the request. In case of
3639 allocation failure the allocation is retried with a smaller
3640 buffer and broken into smaller requests.
3642 Severity : enhancement
3644 Description: Add -o localflock option to simulate old noflock behaviour.
3645 Details : This will achieve local-only flock/fcntl locks coherentness.
3650 Description: versioning check is incomplete
3651 Details : Checking the version difference of client vs. server, report
3652 error if the gap is too big.
3657 Description: add support PG_writeback bit
3658 Details : add support for PG_writeback bit for Lustre, for more carefull
3659 work with page cache in 2.6 kernel. This also fix some deadlocks
3660 and remove hack for work O_SYNC with 2.6 kernel.
3662 Severity : enhancement
3664 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3665 Details : The uninit_groups feature works in conjunction with the kernel
3666 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3667 up the pass1 processing of e2fsck. This is a read-only feature
3668 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3669 mount filesystems that have had this feature enabled.
3671 Severity : enhancement
3673 Description: Improve multi-block allocation algorithm to avoid fragmentation
3674 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3675 allocation locality and avoid filesystem fragmentation.
3677 ------------------------------------------------------------------------------
3679 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3681 * Support for kernels:
3682 2.6.9-42.0.3.EL (RHEL 4)
3683 2.6.5-7.276 (SLES 9)
3684 2.4.21-47.0.1.EL (RHEL 3)
3685 2.6.12.6 vanilla (kernel.org)
3686 2.6.16.21-0.8 (SLES10)
3687 * Recommended e2fsprogs version: 1.39.cfs2-0
3689 * The backwards-compatible /proc/sys/portals symlink has been removed
3690 in this release. Before upgrading, please ensure that you change
3691 any configuration scripts or /etc/sysctl.conf files that access
3692 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3693 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3694 in advance of the upgrade on any system running Lustre 1.4.6 or
3695 newer, since /proc/sys/lnet was added in that version.
3696 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3700 Frequency : only when quota is used
3702 Description: avoid scanning export list for quota master
3703 Details : Change the algorithms to avoid scanning export list in order
3704 to improve the efficiency.
3707 Frequency : MDS failover only, very rarely
3709 Description: "went back in time" messages on mds failover
3710 Details : The greatest transno may be lost when the current operation
3711 finishes with an error (transno==0) and the client's last_rcvd
3712 record is over-written. Save the greatest transno in the
3713 mds_last_transno for this case.
3716 Frequency : always for specific kernels and striping counts
3718 Description: client may get "Matching packet too big" without ACL support
3719 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
3720 when trying to access files in certain configurations. The
3721 clients should in fact be denied when mounting because they do
3722 not understand ACLs.
3725 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
3727 Description: many clients connecting with IO in progress causes connect timeouts
3728 Details : Avoid synchronous journal commits to avoid delays caused by many
3729 clients connecting/disconnecting when bulk IO is in progress.
3730 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
3731 OST_IO_PORTAL to avoid delays behind potentially many pending
3735 Frequency : occasionally with multiple writers to a single file
3737 Description: shared writes to file may result in wrong size reported by stat()
3738 Details : Allow growing of kms when extent lock is cancelled
3741 Frequency : always with random mmap IO to multi-striped file
3743 Description: mmap write might be lost if we are writing to a 'hole' in stripe
3744 Details : Only if the hole is at the end of OST object so that kms is too
3745 small. Fix is to increase kms accordingly in ll_nopage.
3748 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
3750 Description: writes to a missing object would leak memory on the OST
3751 Details : If there is an inconsistency between the MDS and OST filesystems,
3752 such that the MDS references an object that doesn't exist, writes
3753 to that object will leak memory due to incorrect cleanup in the
3754 error handling path, eventually running out of memory on the OST.
3759 Description: Creating too long symlink causes lustre errors
3760 Details : Check symlink and name lengths before sending requests to MDS.
3763 Frequency : only if flock is enabled (not on by default)
3765 Description: posix locks not released on fd closure on 2.6.9+
3766 Details : We failed to add posix locks to list of inode locks on 2.6.9+
3767 kernels, this caused such locks not to be released on fd close and
3768 then assertions on fs unmount about still used locks.
3771 Frequency : MDS failover only, very rarely
3773 Description: clients may get ASSERTION(granted_lock != NULL)
3774 Details : When request was taking a long time, and a client was resending
3775 a getattr by name lock request. The were multiple lock requests
3776 with the same client lock handle and
3777 mds_getattr_name->fixup_handle_for_resent_request found one of the
3778 lock handles but later failed with ASSERTION(granted_lock != NULL).
3783 Description: handle->h_buffer_credits > 0, assertion failure
3784 Details : h_buffer_credits is zero after truncate, causing assertion
3785 failure. This patch extends the transaction or creates a new
3789 Frequency : NFS re-export or patchless client
3790 Bugzilla : 11179, 10796
3791 Description: Crash on NFS re-export node (__d_move)
3792 Details : We do not want to hash the dentry if we don't have a lock.
3793 But if this dentry is later used in d_move, we'd hit uninitialised
3794 list head d_hash, so we just do this to init d_hash field but
3795 leave dentry unhashed.
3798 Frequency : NFS re-export or patchless client
3800 Description: NFS exports has problem with symbolic link
3801 Details : lustre client didn't properly install dentry when re-exported
3802 to NFS or running patchless client.
3805 Frequency : NFS re-export or patchless client
3807 Description: Various nfs/patchless fixes.
3808 Details : fixes reuse disconected alias for lookup process - this fixes
3809 warning "find_exported_dentry: npd != pd",
3810 fix permission error with open files at nfs.
3811 fix apply umask when do revalidate.
3814 Frequency : occasional
3816 Description: Crash on NFS re-export node
3817 Details : calling clear_page() on the wrong pointer triggered oops in
3818 generic_mapping_read().
3821 Frequency : rarely, using O_DIRECT IO
3823 Description: unaligned directio crashes client with LASSERT
3824 Details : check for unaligned buffers before trying any requests.
3827 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
3829 Description: stale data returned from RAID cache
3830 Details : If only a small amount of IO is done to the RAID device before
3831 reading it again it is possible to get stale data from the RAID
3832 cache instead of reading it from disk.
3835 Frequency : always for sles10 kernel
3837 Description: sles10 support
3838 Details : ll_follow_link: compile fixes and using of nd_set_link
3839 under newer kernels.
3842 Frequency : depends on arch, kernel and compiler version, always on sles10
3845 Description: recursive or deep enough symlinks cause stack overflow
3846 Details : getting rid of large stack-allocated variable in
3850 Frequency : depends on hardware
3852 Description: lustre write performance loss in the SLES10 kernel
3853 Details : the performance loss is caused by using of write barriers in the
3854 ext3 code. The SLES10 kernel turns barrier support on by
3855 default. The fix is to undo that change for ldiskfs.
3857 ------------------------------------------------------------------------------
3859 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
3861 * Support for kernels:
3862 2.6.9-42.0.3EL (RHEL 4)
3863 2.6.5-7.276 (SLES 9)
3864 2.4.21-47.0.1.EL (RHEL 3)
3865 2.6.12.6 vanilla (kernel.org)
3869 Frequency : quota enabled and large files being deleted
3871 Description: releasing more than 4GB of quota at once hangs OST
3872 Details : If a user deletes more than 4GB of files on a single OST it
3873 will cause the OST to spin in an infinite loop. Release
3874 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3879 Description: statfs data retrieved from /proc may be stale or zero
3880 Details : When reading per-device statfs data from /proc, in the
3881 {kbytes,files}_{total,free,avail} files, it may appear
3882 as zero or be out of date.
3885 Frequency : systems with MD RAID1 external journal devices
3887 Description: lconf's call to blkid is confused by RAID1 journal devices
3888 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3889 of returning all block devices that match the journal UUID.
3892 Frequency : always, for aggregate stripe size over 4GB
3894 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
3895 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
3896 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
3899 Frequency : always if "lfs find" used on a local file/directory
3901 Description: "lfs find" segfaults if used on a local file/directory
3902 Details : The case where a directory component was not specified wasn't
3903 handled correctly. Handle this properly.
3906 Frequency : always on ppc64
3908 Description: the write to an ext3 filesystem mounted with mballoc got stuck
3909 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3910 perform endianness conversion.
3913 Frequency : rarely (truncate to non-zero file size after write under load)
3914 Bugzilla : 10730, 10687
3915 Description: Files padded with zeros to next 4K multiple
3916 Details : With filesystems mounted using the "extents" option (2.6 kernels)
3917 it is possible that files that are truncated to a non-zero size
3918 immediately after being written are filled with zero bytes beyond
3919 the truncated size. No file data is lost.
3921 Severity : enhancement
3923 Description: Allow recovery/failover for liblustre clients.
3924 Details : liblustre clients were unaware of failover configurations until
3927 Severity : enhancement
3929 Description: user file locks should fail when not mounting with flock option
3930 Details : Set up an error-returning stub in ll_file_operations.lock field
3931 to prevent incorrect behaviour when client is mounted without
3932 flock option. Also, set up properly f_op->flock field for
3936 Frequency : always on ia64
3938 Description: "lfs df" loops on printing out MDS statfs information
3939 Details : The obd_ioctl_data was not initialized and in some systems
3940 this caused a failure during the ioctl that did not return
3941 an error. Initialize the struct and return an error on failure.
3944 Frequency : SLES 9 only
3946 Description: Error of copying files with lustre special EAs as root
3947 Details : Client side always return success for setxattr call for lustre
3948 special xattr (currently only "trusted.lov").
3951 Frequency : rarely on clusters with both ia64+i386 clients
3953 Description: ia64+i686 clients doing shared IO on the same file may LBUG
3954 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
3955 clients are doing concurrent writes to the same file it is
3956 possible that the ia64 clients may LASSERT because the OST
3957 extent locks are not PAGE_SIZE aligned. Ensure that grown
3958 locks are always aligned on the request boundary.
3961 Frequency : specific use, occasional
3963 Description: Overwriting in use executable truncates on-disk binary image
3964 Details : If one node attempts to overwrite an executable in use by
3965 another node, we now correctly return ETXTBSY instead of
3966 truncating the file.
3968 Severity : enhancement
3970 Description: Async OSC create to avoid the blocking unnecessarily.
3971 Details : If an OST has no remaining object, system will block on the creating
3972 when need to create a new object on this OST. Now, ways use
3973 pre-created objects when available, instead of blocking on an
3974 empty osc while others are not empty. If we must block, we block
3975 for the shortest possible period of time.
3980 Description: chmod on Lustre root is propagated to other clients
3981 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
3982 invalid by the follow_mount time.
3985 Frequency : liblustre clients only
3987 Description: Race in 'instant cancel' lock handling could lead to such locks
3988 never to be granted in case of SMP MDS
3989 Details : Do not destroy not yet granted but cbpending locks in
3993 Frequency : replay/resend of open
3995 Description: non null lock assetion failure in mds_intent_policy
3996 Details : Trying to replay/resend lockless open requests resulted in
3997 mds_open() returning 0 with no lock. Now it sets a flag if
3998 a lock is going to be returned.
4000 Severity : enhancement
4002 Description: Checksum enhancements
4003 Details : New checksum enhancements allow for resending RPCs that failed
4006 Severity : enhancement
4008 Description: Tunables on number of dirty pages in cacche
4009 Details : Allow to set limit on number of dirty pages cached.
4014 Description: client crash on unmount - lock still has references
4015 Details : In some error handling cases it was possible to leak a lock
4016 reference on a client while accessing a file. This was not
4017 harmful to the client during operation, but would cause the
4018 client to crash when the filesystem is unmounted.
4021 Frequency : specific case, rare
4023 Description: ETXTBSY on mds though file not in use by client
4024 Details : ETXTBSY is no longer incorrectly returned when attempting to
4025 chmod or chown a directory that the user previously tried to
4026 execute or a currently-executing binary.
4029 Frequency : extremely rare except on liblustre-based clients
4031 Description: Lustre space not freed when files are deleted
4032 Details : Clean up open-unlinked files after client eviction. Previously
4033 the unlink was skipped and the files remained as orphans.
4038 Description: OST failure "would be an LBUG" in waiting_locks_callback()
4039 Details : In some cases it was possible to send a blocking callback to
4040 a client doing a glimpse, even though that client didn't get
4041 a lock granted. When the glimpse lock is cancelled on the OST
4042 the freed lock is left on the waiting list and corrupted the list.
4045 Frequency : all core dumps
4047 Description: Broke core dumps to lustre
4048 Details : Negative dentry may be unhashed if parent does not have UPDATE
4049 lock, but some callers, e.g. do_coredump, expect dentry to be
4050 hashed after successful create, hash it in ll_create_it.
4052 ------------------------------------------------------------------------------
4054 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
4056 * Support for kernels:
4057 2.6.9-42.0.2.EL (RHEL 4)
4058 2.6.5-7.276 (SLES 9)
4059 2.4.21-40.EL (RHEL 3)
4060 2.6.12.6 vanilla (kernel.org)
4064 Frequency : always on RHEL 3
4066 Description: Number of open files grows over time
4067 Details : The number of open files grows over time, whether or not
4068 Lustre is started. This was due to a filp leak introduced
4069 by one of our kernel patches.
4071 ------------------------------------------------------------------------------
4073 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4075 * Support for kernels:
4076 2.6.9-42.EL (RHEL 4)
4077 2.6.5-7.276 (SLES 9)
4078 2.4.21-40.EL (RHEL 3)
4079 2.6.12.6 vanilla (kernel.org)
4084 Bugzilla : 5719, 9635, 9792, 9684
4085 Description: OST (or MDS) trips assertions in (re)connection under heavy load
4086 Details : If a server is under heavy load and cannot reply to new
4087 connection requests before the client resends the (re)connect,
4088 the connection handling code can behave badly if two service
4089 threads are concurrently handing separate (re)connections from
4090 the same client. Add better locking to the connection handling
4091 code, and ensure that only a single connection will be processed
4092 for a given client UUID, even if the lock is dropped.
4094 Severity : enhancement
4096 Description: add TCP zero-copy support to kernel
4097 Details : Add support to the kernel TCP stack to allow zero-copy bulk
4098 sends if the hardware supports scatter-gather and checksumming.
4099 This allows socklnd to do client-write and server-read more
4100 efficiently and reduce CPU utilization from skbuf copying.
4103 Frequency : only if NFS exporting from client
4105 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
4106 Details : If setting attributes on a file created under NFS that had
4107 never been opened it would be possible to oops the client
4108 if the file had no objects.
4111 Frequency : always for liblustre
4113 Description: liblustre client does MDS+OSTs setattr RPC for each write
4114 Details : When doing a write from a liblustre client, the client
4115 incorrectly issued an RPC to the MDS and each OST the file was
4116 striped over in order to update the timestamps. When writing
4117 with small chunks and many clients this could overwhelm the MDS
4118 with RPCs. In all cases it would slow down the write because
4119 these RPCs are unnecessary.
4121 Severity : enhancement
4123 Description: allow number of MDS service threads to be changed at module load
4124 Details : It is now possible to change the number of MDS service threads
4125 running. Adding "options mds mds_num_threads={N}" to the MDS's
4126 /etc/modprobe.conf will set the number of threads for the next
4127 time Lustre is restarted (assuming the "mds" module is also
4128 reloaded at that time). The default number of threads will
4129 stay the same, 32 for most systems.
4134 Description: OST crash if filesystem is unformatted or corrupt
4135 Details : If an OST is started on a device that has never been formatted
4136 or if the filesystem is corrupt and cannot even mount then the
4137 error handling cleanup routines would dereference a NULL pointer.
4142 Description: NULL pointer deref in llap_from_page.
4143 Details : get_cache_page_nowait can return a page with NULL (or otherwise
4144 incorrect) mapping if the page was truncated/reclaimed while it was
4145 searched for. Check for this condition and skip such pages when
4146 doing readahead. Introduce extra check to llap_from_page() to
4147 verify page->mapping->host is non-NULL (so page is not anonymous).
4150 Frequency : Sometimes when using sys_sendfile
4152 Description: "page not covered by a lock" warnings from ll_readpage
4153 Details : sendfile called ll_readpage without right page locks present.
4154 Now we introduced ll_file_sendfile that does necessary locking
4155 around call to generic_file_sendfile() much like we do in
4159 Frequency : with certain MDS communication failures at client mount time
4161 Description: NULL pointer deref after failed client mount
4162 Details : a client connection request may delayed by the network layer
4163 and not be sent until after the PTLRPC layer has timed out the
4164 request. If the client fails the mount immediately it will try
4165 to clean up before the network times out the request. Add a
4166 reference from the request import to the obd device and delay
4167 the cleanup until the network drops the request.
4170 Frequency : occasionally during client (re)connect
4172 Description: assertion failure during client (re)connect
4173 Details : processing a client connection request may be delayed by the
4174 client or server longer than the client connect timeout. This
4175 causes the client to resend the connection request. If the
4176 original connection request is replied in this interval, the
4177 client may trip an assertion failure in ptlrpc_connect_interpret()
4178 which thought it would be the only running connect process.
4181 Frequency : only with obd_echo servers and clients that are rebooted
4183 Description: kernel BUG accessing uninitialized data structure
4184 Details : When running an obd_echo server it did not start the ping_evictor
4185 thread, and when a client was evicted an uninitialized data
4186 structure was accessed. Start the ping_evictor in the RPC
4187 service startup instead of the OBD startup.
4189 Severity : enhancement
4190 Bugzilla : 10193 (patchless)
4191 Description: Remove dependency on various unexported kernel interfaces.
4192 Details : No longer need reparent_to_init, exit_mm, exit_files,
4193 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
4196 Frequency : rare (only users of deprecated and unsupported LDAP config)
4198 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
4199 Details : LDAP apparently contains 'lustreName' attributes instead of
4200 'name'. A simple remapping of the name is sufficient.
4203 Frequency : rare (only with non-default dump_on_timeout debug enabled)
4205 Description: waiting_locks_callback trips kernel BUG if client is evicted
4206 Details : Running with the dump_on_timeout debug flag turned on makes
4207 it possible that the waiting_locks_callback() can try to dump
4208 the Lustre kernel debug logs from an interrupt handler. Defer
4209 this log dumping to the expired_lock_main() thread.
4211 Severity : enhancement
4213 Description: Support NFS exporting on 2.6 kernels.
4214 Details : Implement non-rawops metadata methods for NFS server to use without
4215 changing NFS server code.
4218 Frequency : very rare (synthetic metadata workload only)
4220 Description: two racing renames might cause an MDS thread to deadlock
4221 Details : Running the "racer" program may cause one MDS thread to rename
4222 a file from being the source of a rename to being the target of
4223 a rename at exactly the same time that another thread is doing
4224 so, and the second thread has already enqueued these locks after
4225 doing a lookup of the target and is trying to relock them in
4226 order. Ensure that we don't try to re-lock the same resource.
4229 Frequency : only very large systems with liblustre clients
4231 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
4232 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
4233 all OSTs in parallel. This allows the eviction of stale liblustre
4234 clients to proceed much faster than if they were done in series,
4235 and also offers similar improvements for other set_info RPCs.
4240 Description: excessive CPU usage during initial read phase on client
4241 Details : During the initial read phase on a client, it would agressively
4242 retry readahead on the file, consuming too much CPU and impacting
4243 performance (since 1.4.5.8). Improve the readahead algorithm
4244 to avoid this, and also improve some other common cases (read
4245 of small files in particular, where "small" is files smaller than
4246 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
4251 Description: MDS crash when receiving packet with unknown intent.
4252 Details : Do not LBUG in unknown intent case, just return -EFAULT
4254 Severity : enhancement
4255 Bugzilla : 9293, 9385
4256 Description: MDS RPCs are serialised on client. This is unnecessary for some.
4257 Details : Do not serialize getattr (non-intent version) and statfs.
4260 Frequency : occasional, when OST network is overloaded/intermittent
4262 Description: client evicted by OST after bulk IO timeout
4263 Details : If a client sends a bulk IO request (read or write) the OST
4264 may evict the client if it is unresposive to its data GET/PUT
4265 request. This is incorrect if the network is overloaded (takes
4266 too long to transfer the RPC data) or dropped the OST GET/PUT
4267 request. There is no need to evict the client at all, since
4268 the pinger and/or lock callbacks will handle this, and the
4269 client can restart the bulk request.
4272 Frequency : Always when mmapping file with no objects
4274 Description: client crashes when mmapping file with no objects
4275 Details : Check that we actually have objects in a file before doing any
4276 operations on objects in ll_vm_open, ll_vm_close and
4282 Description: Request leak when working with deleted CWD
4283 Details : Introduce advanced request refcount tracking for requests
4284 referenced from lustre intent.
4286 Severity : Enhancement
4288 Description: Cache open file handles on client.
4289 Details : MDS now will return special lock along with openhandle, if
4290 requested and client is allowed to hold openhandle, even if unused,
4291 until such a lock is revoked. Helps NFS a lot, since NFS is opening
4292 closing files for every read/write openration.
4294 Severity : Enhancement
4296 Description: Cache open negative dentries on client when possible.
4297 Details : Guard negative dentries with UPDATE lock on parent dir, drop
4298 negative dentries on lock revocation.
4303 Description: Remounting a client read-only wasn't possible with a zconf mount
4304 Details : It wasn't possible to remount a client read-only with llmount.
4306 Severity : enhancement
4307 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
4308 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
4309 MPICH is included to add Lustre-specific ADIO interfaces.
4310 This is based closely on the UFS ADIO layer and only differs
4311 in file creation, in order to allow the OST striping to be set.
4312 This is user-contributed code and not supported by CFS.
4317 Description: extended inode attributes (immutable, append-only) work improperly
4318 when 2.4 and 2.6 kernels are used on client/server or vice versa
4319 Details : Introduce kernel-independent values for these flags.
4321 Severity : enhancement
4324 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
4325 Details : Many of the /proc/ tunables can only be tuned at a megabyte
4326 granularity. Now, Fractional MB granularity is be supported,
4327 this is very useful for low memory system.
4329 Severity : enhancement
4331 Description: Getattr by fid
4332 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
4333 locks, avoids extra getattr rpc requests to MDS, allows '/' to
4334 have locks and avoids getattr rpc requests for it on every stat.
4337 Frequency : Always, for filesystems larger than 2TB
4339 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4340 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
4341 filesystems larger than 2TB to be mounted with mballoc enabled.
4344 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
4346 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4347 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
4348 filesystems larger than 2TB is mounted then the kernel will
4349 silently corrupt the start of the filesystem. CONFIG_LBD is
4350 enabled for all CFS-supported kernels, but the possibility of
4351 this happening with a modified kernel config exists.
4353 Severity : enhancement
4355 Description: add client O_DIRECT support for 2.6 kernels
4356 Details : It is now possible to do O_DIRECT reads and writes to files
4357 in the Lustre client mountpoint on 2.6 kernel clients.
4359 Severity : enhancement
4361 Description: parallel glimpse, setattr, statfs, punch, destroy requests
4362 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
4363 parallel, not waiting for response from every OST before sending
4364 a rpc to the next OST.
4369 Description: setattr vs write race when updating file timestamps
4370 Details : Client processes that update a file timestamp into the past
4371 right after writing to the file (e.g. tar) it is possible that
4372 the updated file modification time can be reset to the current
4373 time due to a race between processing the setattr and write RPC.
4375 Severity : enhancement
4377 Description: Bring 'lfs find' closer in line with regular Linux find.
4378 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
4379 -print0 options and obtains all the needed info through the lustre
4382 Severity : enhancement
4384 Description: support up to 1024 configured devices on one node
4385 Details : change obd_dev array from statically allocated to dynamically
4386 allocated structs as they are first used to reduce memory usage
4391 Description: Flush dirty partially truncated pages during truncate
4392 Details : Immediatelly flush partially truncated pages in filter_setattr,
4393 this way we completely avoid having any pages in page cache on OST
4394 and can retire ugly workarounds during writes to flush such pages.
4399 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
4400 Details : protect inode from truncation within vfs_unlink() context
4401 just take a reference before calling vfs_unlink() and release it
4402 when parent's i_sem is free.
4407 Description: last_id value checked outside lock on OST caused LASSERT failure
4408 Details : If there were multiple MDS->OST object precreate requests in
4409 flight, it was possible that the OST's last object id was checked
4410 outside a lock and incorrectly tripped an assertion. Move checks
4411 inside locks, and discard old precreate requests.
4414 Frequency : always, if extents are used on OSTs
4416 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
4417 Details : OSTs using the extents format would not zero the high 16 bits of
4418 the index physical block number. This is not a problem for any
4419 OST filesystems smaller than 16TB, and no kernels support ext3
4420 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
4421 new/modified files) and can be fixed for existing filesystems
4422 with e2fsprogs-1.39-cfs1.
4427 Description: import connection selection may be incorrect if timer wraps
4428 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
4429 import connections to be ignored if the 32-bit jiffies counter
4430 wraps. Use a 64-bit jiffies counter.
4433 Frequency : very large clusters immediately after boot
4435 Description: LNET request buffers exhausted under heavy short-term load
4436 Details : If a large number of client requests are generated on a service
4437 that has previously never seen so many requests it is possible
4438 that the request buffer growth cannot keep up with the spike in
4439 demand. Instead of dropping incoming requests, they are held in
4440 the LND until the RPC service can accept more requests.
4443 Frequency : Sometimes during replay
4445 Description: Assertion failure in ll_local_open after replay.
4446 Details : If replay happened on an open request reply before we were able
4447 to set replay handler, reply will become not swabbed tripping the
4448 assertion in ll_local_open. Now we set the handler right after
4449 recognising of open request
4452 Frequency : very rare
4454 Description: kernel reports "badness in vsnprintf"
4455 Details : Reading from the "recovery_status" /proc file in small chunks
4456 may cause a negative length in lprocfs_obd_rd_recovery_status()
4457 call to vsnprintf() (which is otherwise harmless). Exit early
4458 if there is no more space in the output buffer.
4460 Severity : enhancement
4462 Description: clear OBD RPC statistics by writing to them
4463 Details : It is now possible to clear the OBD RPC statistics by writing
4464 to the "stats" file.
4469 Description: Client mtime is not the same on different clients after utimes
4470 Details : In some cases, the client was using the utimes() syscall on
4471 a file cached on another node. The clients now validate the
4472 ctime from the MDS + OSTs to determine which one is right.
4477 Description: Inability to activate failout mode
4478 Details : lconf script incorrectly assumed that in pythong string's numeric
4479 value is used in comparisons.
4482 Frequency : always with multiple stripes per file
4484 Description: Inefficient object allocation for mutli-stripe files
4485 Details : When selecting which OSTs to stripe files over, for files with
4486 a stripe count that divides evenly into the number of OSTs,
4487 the MDS is always picking the same starting OST for each file.
4488 Return the OST selection heuristic to the original design.
4493 Description: mount failures may take full timeout to return an error
4494 Details : Under some heavy load conditions it is possible that a
4495 failed mount can wait for the full obd_timeout interval,
4496 possibly several minutes, before reporting an error.
4497 Instead return an error as soon as the status is known.
4499 Frequency : quota enabled and large files being deleted
4501 Description: releasing more than 4GB of quota at once hangs OST
4502 Details : If a user deletes more than 4GB of files on a single OST it
4503 will cause the OST to spin in an infinite loop. Release
4504 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4509 Description: statfs data retrieved from /proc may be stale or zero
4510 Details : When reading per-device statfs data from /proc, in the
4511 {kbytes,files}_{total,free,avail} files, it may appear
4512 as zero or be out of date.
4515 Frequency : systems with MD RAID1 external journal devices
4517 Description: lconf's call to blkid is confused by RAID1 journal devices
4518 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4519 of returning all block devices that match the journal UUID.
4522 Frequency : always, for aggregate stripe size over 4GB
4524 Description: assertion fails when trying to use 4GB stripe size
4525 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
4526 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
4529 Frequency : always on ppc64
4531 Description: the first write on an ext3 filesystem with mballoc got stuck
4532 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4533 perform endianness conversion.
4535 ------------------------------------------------------------------------------
4537 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4539 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
4540 INTEROPERATE with older versions automatically. Please read the
4541 user documentation before upgrading any part of a live system.
4542 * WARNING: Lustre networking configuration changes are required with
4543 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
4546 * Support for kernels:
4547 2.6.9-22.0.2.EL (RHEL 4)
4548 2.6.5-7.244 (SLES 9)
4549 2.6.12.6 vanilla (kernel.org)
4552 Severity : enhancement
4553 Bugzilla : 7981/8208
4554 Description: Introduced Lustre Networking (LNET)
4555 Details : LNET is new networking infrastructure for Lustre, it includes
4556 a reorganized network configuration mode (see the user
4557 documentation for full details) as well as support for routing
4558 between different network fabrics. Lustre Networking Devices
4559 (LNDs) for the supported network fabrics have also been
4560 created for this new infrastructure.
4562 Severity : enhancement
4563 Description: Introduced Access control lists
4564 Details : clients can set ACLs on files and directories in order to have
4565 more fine-grained permissions than the standard Unix UGO+RWX.
4566 The MDS must be started with the "-o acl" mount option.
4568 Severity : enhancement
4569 Description: Introduced filesystem quotas
4570 Details : Administrators may now establish per-user quotas on the
4573 Severity : enhancement
4575 Description: Configuration change for the XT3
4576 The PTLLND is now used to run Lustre over Portals on the XT3
4577 The configure option(s) --with-cray-portals are no longer used.
4578 Rather --with-portals=<path-to-portals-includes> is used to
4579 enable building on the XT3. In addition to enable XT3 specific
4580 features the option --enable-cray-xt3 must be used.
4585 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
4586 Details : A race between allocating a new llcd and re-getting the llcd_lock
4587 allowed another thread to grab newly-allocated llcd.
4589 Severity : enhancement
4591 Description: 2.6 OST async journal commit and locking fix to improve performance
4592 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
4593 2.6 kernels are now async as they already were in 2.4 kernels so
4594 that they can commit concurrently with the network bulk transfer.
4595 For block-allocated files the filter allocation semaphore is held
4596 to avoid filesystem fragmentation during allocation. BKL lock
4597 removed for 2.6 xattr operations where it is no longer needed.
4602 Description: lconf incorrectly determined whether two IP networks could talk
4603 Details : In some more complicated routing and multiple-network
4604 configurations, lconf will avoid trying to make a network
4605 connection to a disjoint part of the IP space. It was doing the
4606 math incorrectly for one set of cases.
4611 Description: Fix for potential infinite loop processing records in an llog.
4612 Details : If an llog record is corrupted/zeroed, it is possible to loop
4613 forever in llog_process(). Validate the llog record length
4614 and skip the remainder of the block on error.
4617 Frequency : occasional (liblustre only)
4619 Description: liblustre could not open files whose last component is a symlink
4620 Details : sysio_path_walk() would incorrectly pass the open intent to
4621 intermediate path components.
4624 Frequency : rare (liblustre only with non-standard tuning)
4625 Bugzilla : 7201 (7350)
4626 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4627 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4628 after it has been released, only do this for referenced locks
4630 Severity : enhancement
4632 Description: specifying an (invalid) directory default stripe_size of -1
4633 would reset the directory default striping
4634 Details : stripe_size -1 was used internally to signal directory stripe
4635 removal, now use "all default" to signal dir stripe removal
4636 as a directory striping of "all default" is not useful
4639 Frequency : common for large clusters running liblustre clients
4641 Description: doing an ls when liblustre clients are running is slow
4642 Details : sending a glimpse AST to a liblustre client waits for every AST
4643 to time out, as liblustre clients will not respond. Since they
4644 cannot cache data we refresh the OST lock LVB from disk instead.
4646 Severity : enhancement
4648 Description: doing an ls at the same time as file IO can be slow
4649 Details : enqueue and other "small" requests can be blocked behind many
4650 large IO requests. Create a new OST IO portal for non-IO
4651 requests so they can be processed faster.
4654 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4656 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4657 created file with mode 000
4658 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4659 allow an ftruncate() as a non-root user to a file with mode 000.
4660 Lustre now respects this flag to disable mode checks when
4661 truncating a file owned by the user
4664 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4667 Description: Revoking locks from clients that went dead or catatonic might take
4669 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4670 cancellation of such locks instant on servers without waiting for
4671 any reply from clients. Clients drops these locks when cancel
4672 notification from server is received without replying.
4675 Frequency : liblustre-only, when liblustre client dies or becomes busy
4677 Description: Doing ls on Linux clients can take a long time with active
4679 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4680 granting such locks to it in the first place if possible. Locks
4681 are taken by proxy on the OST during the read or write and
4682 dropped immediately afterward. Add connect flags handling, do
4683 not grant locks to liblustre clients for glimpse ASTs.
4685 Severity : enhancement
4687 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4688 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4689 sequential reads and behaves badly with applications doing random
4690 reads. Improve it by reading ahead at least read region, and
4691 avoiding excessive large RPC for small reads.
4693 Severity : enhancement
4695 Description: Creating more than 1000 files for a single job may cause a load
4696 imbalance on the OSTs if there are also a large number of OSTs.
4697 Details : qos_prep_create() uses an OST index reseed value that is an
4698 even multiple of the number of available OSTs so that if the
4699 reseed happens in the middle of the object allocation it will
4700 still utilize the OSTs as uniformly as possible.
4705 Description: OST or MDS may oops in ping_evictor_main()
4706 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
4707 but doesn't restart at beginning of obd_exports_timed list
4710 Severity : enhancement
4712 Description: improve by-nid export eviction on the MDS and OST
4713 Details : allow multiple exports with the same NID to be evicted at one
4714 time without re-searching the exports list.
4717 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
4719 Description: MDS may oops in groups_free()
4720 Details : in rare race conditions a newly allocated group_info struct is
4721 freed again, and this can be NULL. The 2.4 compatibility code
4722 for groups_free() checked for a NULL pointer, but 2.6 did not.
4725 Frequency : common for liblustre clients doing little filesystem IO
4726 Bugzilla : 9352, 7313
4727 Description: server may evict liblustre clients accessing contended locks
4728 Details : if a client is granted a lock or receives a completion AST
4729 with a blocking AST already set it would not reply to the AST
4730 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
4733 Frequency : lfs setstripe, only systems with more than 160 OSTs
4735 Description: unable to set striping with a starting offset beyond OST 160
4736 Details : llapi_create_file() incorrectly limited the starting stripe
4737 index to the maximum single-file stripe count.
4740 Frequency : LDAP users only
4742 Description: lconf did not handle in-kernel recovery with LDAP properly
4743 Details : lconf/LustreDB get_refs() is searching the wrong namespace
4745 Severity : enhancement
4747 Description: bind OST threads to NUMA nodes to improve performance
4748 Details : all OST threads are uniformly bound to CPUs on a single NUMA
4749 node and do their allocations there to localize memory access
4751 Severity : enhancement
4753 Description: llmount can determine client NID directly from Myrinet (GM)
4754 Details : the client NID code from gmnalnid was moved directly into
4755 llmount, removing the need to use this or specifying the
4756 client NID explicitly when mounting GM clients with zeroconf
4759 Frequency : if client is started with down MDS
4761 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
4762 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
4763 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
4764 ptlrpc_import_delayed_req() because we didn't check if the
4765 request was interrupted, and we also didn't break out of the
4766 event loop if there was no timeout
4771 Description: data loss during non-page-aligned writes to a single file from
4772 both multiple nodes and multiple threads on one node at same time
4773 Details : updates to KMS and lsm weren't protected by common lock. Resulting
4774 inconsistency led to false short-reads, that were cached and later
4775 used by ->prepare_write() to fill in partially written page,
4776 leading to data loss.
4779 Frequency : always, if lconf --abort_recovery used
4781 Description: lconf --abort_recovery fails with 'Operation not supported'
4782 Details : lconf was attempting to abort recovery on the MDT device and not
4785 Severity : enhancement
4787 Description: remove cleanup logs
4788 Details : replace lconf-generated cleanup logs with lustre internal
4789 cleanup routines. Eliminates the need for client-cleanup and
4792 Severity : enhancement
4794 Description: add support for EAs (user and system) on lustre filesystems
4795 Details : it is now possible to store extended attributes in the Lustre
4796 client filesystem, and with the user_xattr mount option it
4797 is possible to allow users to store EAs on their files also
4799 Severity : enhancement
4801 Description: Add possibility (config option) to show minimal available OST free
4803 Details : When compiled with --enable-mindf configure option, statfs(2)
4804 (and so, df) will return least minimal free space available from
4805 all OSTs as amount of free space on FS, instead of summary of
4806 free spaces of all OSTs.
4808 Severity : enhancement
4810 Description: do not expand extent locks acquired on OST-side
4811 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
4812 by server: they are not cached anyway.
4815 Frequency : when mmap is used/binaries executed from Lustre
4817 Description: Unmmap pages before throwing them away from read cache.
4818 Details : llap_shrink cache now attempts to unmap pages before discarding
4819 them (if unmapping failed - do not discard). SLES9 kernel has
4820 extra checks that trigger if this unmapping is not done first.
4825 Description: lconf didn't resolve symlinks before checking to see whether a
4826 given mountpoint was already in use
4829 Frequency : when migrating failover services
4830 Bugzilla : 6395, 9514
4831 Description: When migrating a subset of services from a node (e.g. failback
4832 from a failover service node) the remaining services would
4833 time out and evict clients.
4834 Details : lconf --force (implied by --failover) sets the global obd_timeout
4835 to 5 seconds in order to quickly disconnect, but this caused
4836 other RPCs to time out too quickly. Do not change the global
4837 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
4839 Severity : enhancement
4840 Frequency : if MDS is started with down OST
4841 Bugzilla : 9439,5706
4842 Description: Allow startup/shutdown of an MDS without depending on the
4843 availability of the OSTs.
4844 Details : Asynchronously call mds_lov_synchronize during MDS startup.
4845 Add appropriate locking and lov-osc refcounts for safe
4846 cleaning. Add osc abort_inflight calls in case the
4847 synchronize never started.
4850 Frequency : occasional (Cray XT3 only)
4852 Description: root not authorized to access files in CRAY_PORTALS environment
4853 Details : The client process capabilities were not honoured on the MDS in
4854 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
4855 been accessed by an authorized user then root was able to access
4856 the file on the local client also. The root user capabilities
4857 are now allowed on the MDS, as this environment has secure UID.
4860 Frequency : occasional
4862 Description: ldiskfs "too long searching" message happens too often
4863 Details : A debugging message (otherwise harmless) prints too often on
4864 the OST console. This has been reduced to only happen when
4865 there are fragmentation problems on the filesystem.
4870 Description: Division by zero in statfs when all OSCs are inactive
4871 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
4872 lov_statfs divides by value returned by lov_get_stripecnt().
4876 Bugzilla : 9489, 3273
4877 Description: First write from each client to each OST was only 4kB in size,
4878 to initialize client writeback cache, which caused sub-optimal
4879 RPCs and poor layout on disk for the first written file.
4880 Details : Clients now request an initial cache grant at (re)connect time
4881 and so that they can start streaming writes to the cache right
4882 away and always do full-sized RPCs if there is enough data.
4883 If the OST is rebooted the client also re-establishes its grant
4884 so that client cached writes will be honoured under the grant.
4889 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
4890 Details : Now I/O RPCs go to different portal number and (presumably) fast
4891 lock requests (and glimses) and other RPCs get their own service
4892 threads pool that should be able to service those RPCs
4895 Severity : enhancement
4897 Description: Ability to exchange lustre version between client and servers and
4898 issue warnings at client side if client is too old. Also for
4899 liblustre clients there is ability to refuse connection of too old
4901 Details : New 'version' field is added to connect data structure that is
4902 filled with version info. That info is later checked by server and
4906 Frequency : rare, liblustre only.
4907 Bugzilla : 9296, 9581
4908 Description: Two simultaneous writes from liblustre at offset within same page
4909 might proceed at the same time overwriting eachother with stale
4911 Details : I/O lock withing llu_file_prwv was released too early, before data
4912 actually was hitting the wire. Extended lock-holding time until
4913 server acknowledges receiving data.
4916 Frequency : extremely rare. Never observed in practice.
4918 Description: avoid generating lustre_handle cookie of 0.
4919 Details : class_handle_hash() generates handle cookies by incrementing
4920 global counter, and can hit 0 occasionaly (this is unlikely, but
4921 not impossible, because initial value of cookie counter is
4922 selected randonly). Value of 0 is used as a sentinel meaning
4923 "unassigned handle" --- avoid it. Also coalesce two critical
4924 sections in this function into one.
4926 Severity : enhancement
4928 Description: allow liblustre clients to delegate truncate locking to OST
4929 Details : To avoid overhead of locking, liblustre client instructs OST to
4930 take extent lock in ost_punch() on client's behalf. New connection
4931 flag is added to handle backward compatibility.
4933 Severity : enhancement
4934 Bugzilla : 4928, 7341, 9758
4935 Description: allow number of OST service threads to be specified
4936 Details : a module parameter allows the number of OST service threads
4937 to be specified via "options ost ost_num_threads={N}" in the
4938 OSS's /etc/modules.conf or /etc/modprobe.conf.
4942 Bugzilla : 6146, 9635, 9895
4943 Description: servers crash with bad pointer in target_handle_connect()
4944 Details : In rare cases when a client is reconnecting it was possible that
4945 the connection request was the last reference for that export.
4946 We would temporarily drop the export reference and get a new
4947 one, but this may have been the last reference and the export
4948 was just destroyed. Get new reference before dropping old one.
4950 Severity : enhancement
4951 Frequency : if client is started with failover MDS
4953 Description: Allow multiple MDS hostnames in the mount command
4954 Details : Try to read the configuration from all specified MDS
4955 hostnames during a client mount in case the "primary"
4958 Severity : enhancement
4960 Description: Stop sending data to evicted clients as soon as possible.
4961 Details : Check if the client we are about to send or are sending data to
4962 was evicted already. (Check is done every second of waiting,
4963 for which l_wait_event interface was extended to allow checking
4964 of exit condition at specified intervals).
4967 Frequency : rare, normally only when NFS exporting is done from client
4969 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
4971 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
4974 Frequency : always, if srand() is called before liblustre initialization
4976 Description: Liblustre uses system PRNG disturbing its usage by user application
4977 Details : Introduce internal to lustre fast and high-quality PRNG for
4978 lustre usage and make liblustre and some other places in generic
4979 lustre code to use it.
4981 Severity : enhancement
4982 Bugzilla : 9477, 9557, 9870
4983 Description: Verify that the MDS configuration logs are updated when xml is
4984 Details : Check if the .xml configuration logs are newer than the config
4985 logs stored on the MDS and report an error if this is the case.
4986 Request --write-conf, or allow starting with --old_conf.
4988 Severity : enhancement
4990 Description: Handle symlinks in the path when checking if Lustre is mounted.
4991 Details : Resolve intermediate symlinks when checking if a client has
4992 mounted a filesystem to avoid duplicate client mounts.
4997 Description: lconf can hit an error exception but still return success.
4998 Details : The lconf command catches the Command error exception at the top
4999 level script context and will exit with the associated exit
5000 status, but doesn't ensure that this exit status is non-zero.
5005 Description: failure of ptlrpc thread startup can cause oops
5006 Details : Starting a ptlrpc service thread can fail if there are a large
5007 number of threads or the server memory is very fragmented.
5008 Handle this without oopsing.
5011 Frequency : always, only if liblustre and non-default acceptor port was used
5013 Description: liblustre cannot connect to servers with non-default acceptor port
5014 Details : tcpnal_set_default_params() was not called and was therefore
5015 ignoring the environment varaible TCPNAL_PORT, as well as other
5016 TCPNAL_ environment variables
5021 Description: two objects could be created on the same OST for a single file
5022 Details : If an OST is down, in some cases it was possible to create two
5023 objects on a single OST for a single file. No problems other
5024 than potential performance impact and spurious error messages.
5028 Bugzilla : 5681, 9562
5029 Description: Client may oops in ll_unhash_aliases
5030 Details : Client dcache may become inconsistent in race condition.
5031 In some cases "getcwd" can fail if the current directory is
5037 Description: Inode refcounting problems in NFS export code
5038 Details : link_raw functions used to call d_instantiate without obtaining
5039 extra inode reference first.
5043 Bugzilla : 9942, 9903
5044 Description: Referencing freed requests leading to crash, memleaks with NFS.
5045 Details : We used to require that call to ll_revalidate_it was always
5046 followed by ll_lookup_it. Also with revalidate_special() it is
5047 possible to call ll_revalidate_it() twice for the same dentry
5048 even if first occurence returned success. This fix changes semantic
5049 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
5050 reference on a request referred from the intent.
5051 ll_intent_release() then releases such a request.
5054 Frequency : rare, normally benchmark loads only
5056 Description: unlinked inodes were kept in memory on the client
5057 Details : If a client is repeatedly creating and unlinking files it
5058 can accumulate a lot of stale inodes in the inode slab cache.
5059 If there is no other client load running this can cause the
5060 client node to run out of memory. Instead flush old inodes
5061 from client cache that have the same inode number as a new inode.
5064 Frequency : SLES9 2.6.5 kernel and long filenames only
5065 Bugzilla : 9969, 10379
5066 Description: utime reports stale NFS file handle
5067 Details : SLES9 uses out-of-dentry names in some cases, which confused
5068 the lustre dentry revalidation. Change it to always use the
5072 Frequency : rare, unless heavy write-truncate concurrency is continuous
5073 Bugzilla : 4180, 6984, 7171, 9963, 9331
5074 Description: OST becomes very slow and/or deadlocked during object unlink
5075 Details : filter_destroy() was holding onto the parent directory lock
5076 while truncating+unlinking objects. For very large objects this
5077 may block other threads for a long time and slow overall OST
5078 responsiveness. It may also be possible to get a lock ordering
5079 deadlock in this case, or run out of journal credits because of
5080 the combined truncate+unlink. Solution is to do object truncate
5081 first in one transaction without parent lock, and then do the
5082 final unlink in a new transaction with the parent lock. This
5083 reduces the lock hold time dramatically.
5086 Frequency : rare, 2.4 kernels only
5088 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
5089 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
5090 before unmounting their filesystem in order to avoid deadlock.
5091 The kernel_locked() function in 2.4 kernels only checks whether
5092 the kernel lock is held, not whether it is this process that is
5093 holding it as 2.6 kernels do.
5098 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
5099 Details : The client ptlrpc code may be trying to reconnect to a down
5100 server before a previous connection attempt has timed out.
5101 Increase the reconnect interval to be longer than the connection
5102 timeout interval to avoid sending duplicate connections to
5106 Frequency : echo_client brw_test command
5108 Description: fix echo_client to work with OST preallocated code
5109 Details : OST preallocation code (5137) didn't take echo_client IO path
5110 into account: echo_client calls filter methods outside of any
5111 OST thread and, hence, there is no per-thread preallocated
5112 pages and buffers to use. Solution: hijack pga pages for IO. As
5113 a byproduct, this avoids unnecessary data copying.
5117 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
5118 Description: Client can oops in mdc_commit_close() after open replay
5119 Details : It was possible for the MDS to return an open request with no
5120 transaction number in mds_finish_transno() if the client was
5121 evicted, but without actually returning an error. Clients
5122 would later try to replay that open and may trip an assertion
5123 Simplify the client close codepath, and always return an error
5124 from the MDS in case the open is not successful.
5127 Frequency : rare, 2.6 OSTs only
5129 Description: OST may deadlock under high load on fragmented files
5130 Details : If there was a heavy load and highly-fragmented OST filesystems
5131 it was possible to have all the OST threads deadlock waiting on
5132 allocation of biovecs, because the biovecs were not released
5133 until the entire RPC IO was completed. Instead, release biovecs
5134 as soon as they are complete to ensure forward IO progress.
5136 Severity : enhancement
5138 Description: Support for specifying external journal device at mount
5139 Details : If an OST or MDS device is formatted with an external journal
5140 device, this device major/minor is stored in the ext3 superblock
5141 and may not be valid for failover. Allow detecting and
5142 specifying the external journal at mount time.
5147 Description: Mounting an MDS with pending unlinked files may cause oops
5148 Details : target_finish_recovery() calls mds_postrecov() which returned
5149 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
5150 considers this an error and immediately begins cleaning up the
5151 lov, just after starting the mds_lov process
5153 Severity : enhancement
5155 Description: Implement 'lfs df' to report actual free space on per-OST basis
5156 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
5157 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
5158 usage of objects; '-h' to report in human readable format.
5160 ------------------------------------------------------------------------------
5162 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
5169 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
5170 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
5171 at startup during orphan unlinking. Instead, skip buddy update
5172 before setup, as it will be regenerated after recovery is complete.
5177 Description: If an OST is inactive, its locks might reference stale inodes.
5178 Details : lov_change_cbdata() must iterate over all namespaces, even if
5179 they are inactive to clear inode references from the lock.
5181 Severity : enhancement
5182 Frequency : occasional, if non-standard max_dirty_mb used
5184 Description: Client will block write RPCs if not enough grant
5185 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
5186 then the client will block writes while waiting for another RPC
5187 to complete instead of consuming its dirty limit. With change
5188 we get improved performance when max_dirty_mb is small.
5190 Severity : enhancement
5191 Bugzilla : 3389, 6253
5192 Description: Add support for supplementary groups on the MDS.
5193 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
5194 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
5195 lookups for user supplementary groups into a cache.
5199 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
5200 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
5201 for liblustre O_EXCL is enforced by MDS.
5205 Description: readdir never returns NULL in liblustre.
5206 Details : Corrected llu_iop_getdirentries logic, to return offset of next
5207 dentry in struct dirent.
5211 Frequency : liblustre only, depends on application IO pattern
5212 Description: liblustre clients evicted if not contacting servers
5213 Details : Don't put liblustre clients into the ping_evictor list, so
5214 they will not be evicted by the pinger ever.
5216 Severity : enhancement
5218 Description: Add ability to evict clients by NID from MDS.
5219 Details : By echoing "nid:$NID" string into
5220 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
5221 $NID would be instantly evicted from this MDS and from all active
5222 OSTs connected to it.
5226 Description: Do not query file size twice, somewhat slowing stat(2) calls.
5227 Details : lookup_it_finish() used to query file size from OSTs that was not
5232 Description: service threads change working directory to that of init
5233 Details : Starting lustre service threads may pin the working directory
5234 of the parent thread, making that filesystem busy. Threads
5235 now change to the working directory of init to avoid this.
5239 Frequency : during shutdown only
5240 Description: shutdown with a failed MDS or OST can cause unmount to hang
5241 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
5246 Frequency : 2.6 only
5247 Description: chmod/chown may include an extra supplementary group
5248 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
5249 supplementary group and if none is specified this is used.
5252 Bugzilla : 5479 (6816)
5254 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
5255 Details : If lookup is in progress on a file that is unlinked we might try
5256 to revalidate the inode and fail in revalidate after lookup is
5257 complete and ll_file_open() enqueues the open again but
5258 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
5261 Frequency : always, if lconf --abort_recovery used
5263 Description: lconf --abort_recovery fails with 'Operation not supported'
5264 Details : lconf was attempting to abort recovery on the MDT device and not
5267 ------------------------------------------------------------------------------
5269 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
5274 Frequency : rare (only unsupported configurations with a node running as an
5276 Bugzilla : 6514, 5137
5277 Description: Mounting a Lustre file system on a node running as an OST could
5279 Details : OSTs now preallocates memory needed to write out data at
5280 startup, instead of when needed, to avoid having to
5281 allocate memory in possibly low memory situations.
5282 Specifically, if the file system is mounted on on OST,
5283 memory pressure could force it to try to write out data,
5284 which it needed to allocate memory to do. Due to the low
5285 memory, it would be unable to do so and the node would
5286 become unresponsive.
5288 Severity : enhancement
5290 Description: Addition of lconf --service command line option
5291 Details : lconf now accepts a '--service <arg>' option, which is
5292 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
5294 Severity : enhancement
5296 Description: Failover mode is now the default for OSTs.
5297 Details : By default, OSTs will now run in failover mode. To return to
5298 the old behaviour, add '--failout' to the lmc line for OSTs.
5300 Severity : enhancement
5302 Description: Health checks are now provided for MDS and OSTs
5303 Details : Additional detailed health check information on MSD and OSTs
5304 is now provided through the procfs health_check value.
5307 Frequency : occasional, depends on IO load
5309 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
5310 after numerous create/delete cycles
5311 Details : The ext3 inode allocation policy would not allocate new inodes
5312 very well on the OSTs because there are no new directories
5313 being created. Instead we look for groups with free space if
5314 the parent directories are nearly full.
5319 Description: Network or server problems during mount may cause partially
5320 mounted clients instead of returning an error.
5321 Details : The config llog parsing code may overwrite the error return
5322 code during mount error handling, returning success instead
5328 Description: MDS can fail to allocate large reply buffers
5329 Details : After long uptimes the MDS can fail to allocate large reply
5330 buffers (e.g. zconf client mount config records) due to memory
5331 fragmentation or consumption by the buffer cache. Preallocate
5332 some large reply buffers so that these replies can be sent even
5333 under memory pressure.
5337 Frequency : rare (liblustre)
5338 Description: fsx running with liblustre complained that using truncate() to
5339 extend the file doesn't work. This patch corrects that issue.
5340 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
5341 ATTR_SIZE and lsm use in llu_setattr_raw.
5345 Frequency : rare, only 2.6 kernels
5346 Description: Unusual file access patterns on the MDS may result in inode
5347 data being lost in very rare circumstances.
5348 Details : Bad interaction between the ea-in-inode patch and the "no-read"
5349 code in the 2.6 kernel caused the inode and/or EA data not to
5350 be read from disk, causing single-file corruption.
5354 Frequency : rare, only 2.6 filesystems using extents
5355 Description: Heavy concurrent write and delete load may cause data corruption.
5356 Details : It was possible under high-load situations to have an extent
5357 metadata block in the block device cache from a just-unlinked
5358 file overwrite a newly-allocated data block. We now unmap any
5359 metadata buffers that alias just-allocated data blocks.
5363 Frequency : filesystems with default stripe_count larger than 77
5364 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
5365 Details : lconf specifies an inode size of 4096 bytes when the default
5366 stripe_count is larger than 77. This conflicts with the default
5367 inode density of 1 per 4096 bytes. Allocate smaller inodes in
5368 this case to avoid pinning too much memory for large EAs.
5370 ------------------------------------------------------------------------------
5372 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
5377 Frequency : rare (extremely heavy IO load with hundreds of clients)
5379 Description: Client is evicted, gets IO error writing to file
5380 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
5381 caused clients to be evicted for AST timeouts. The fixes in
5382 bug 5192 mean we no longer need to have such short AST timeouts
5383 so ldlm_timeout has been increased.
5386 Frequency : occasional during --force or --failover shutdown under load
5387 Bugzilla : 5949, 4834
5388 Description: Server oops/LBUG if stopped with --force or --failover under load
5389 Details : a collection of import/export refcount and cleanup ordering
5390 issues fixed for safer force cleanup
5393 Frequency : only filesystems larger than 120 OSTs
5394 Bugzilla : 5990, 6223
5395 Description: lfs getstripe would oops on a very large filesystem
5396 Details : lov_getconfig used kfree on vmalloc'd memory
5399 Frequency : only filesystems exporting via NFS to Solaris 10 clients
5400 Bugzilla : 6242, 6243
5401 Description: reading from files that had been truncated to a non-zero size
5402 but never opened returned no data
5403 Details : ll_file_read() reads zeros from no-object files to EOF
5408 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
5409 Details : There's little harm in aborting MDS/OSS recovery and letting it
5410 try again, so I removed the LASSERT and return an error instead.
5412 Severity : enhancement
5414 Description: New debugging infrastructure for tracking down data corruption
5415 Details : The I/O checksum code was replaced to: (a) control it at runtime,
5416 (b) cover more of the client-side code path, and (c) try to narrow
5417 down where problems occurred
5421 Bugzilla : 3819, 4364, 4397, 6313
5422 Description: Racing close and eviction MDS could cause assertion in mds_close
5423 Details : It was possible to get multiple mfd references during close and
5424 client eviction, leading to one thread referencing a freed mfd.
5426 Severity: : enhancement
5427 Bugzilla : 3262, 6359
5428 Description: Attempts to reconnect to servers are now more aggressive.
5429 Details : This builds on the enhanced upcall-less recovery that was added
5430 in 1.4.2. When trying to reconnect to servers, clients will
5431 now try each server in the failover group every 10 seconds. By
5432 default, clients would previously try one server every 25 seconds.
5437 Description: After recovery, certain operations trigger a failed
5438 assertion on a client.
5439 Details : Failing over an mds, using lconf -d --failover, while a
5440 client was doing a readdir() call would cause the client to
5441 LBUG after recovery completed and the readdir() was resent.
5443 Severity : enhancement
5445 Description: Default groups are now added by lconf
5446 Details : You can now run lconf --group <servicename> without having to
5447 manually add groups with lmc.
5450 Frequency : occasional
5452 Description: Nodes with an elan id of 0 trigger a failed assertion
5455 Frequency : always when accessing e.g. tty/console device nodes
5457 Description: tty and some other devices nodes cannot be used on lustre
5458 Details : file's private_data field is used by device data and lustre
5459 values in there got lost. New field was added to struct file to
5460 store fs-specific private data.
5463 Frequency : when exporting Lustre via NFS
5465 Description: NFSD failed occasionally when looking up a path component
5466 Details : NFSD is looking up ".." which was broken in ext3 directories
5467 that had grown large enough to become hashed.
5470 Frequency : Clusters with multiple interfaces not on the same subnet
5472 Description: Nodes will repeatedly try to reconnect to an interface which it
5473 cannot reach and report an error to the log.
5474 Details : Extra peer list entries will be created by lconf with some peers
5475 unreachable. lconf now validates the peer before adding it.
5478 Frequency : Only if a default stripe is set on the filesystem root.
5480 Description: Setting a default stripe on the filesystem root prevented the
5481 filesystem from being remounted.
5482 Details : The client was sending extra request flags in the root getattr
5483 request and did not allocate a reply buffer for the dir EA.
5486 Frequency : occasional, higher if lots of files are accessed by one client
5487 Bugzilla : 6159, 6097
5488 Description: Client trips assertion regarding lsm mismatch/magic
5489 Details : While revalidating inodes the VFS looks up inodes with ifind()
5490 and in rare cases can find an inode that is being freed.
5491 The ll_test_inode() code will free the lsm during ifind()
5492 when it finds an existing inode and then the VFS later attaches
5493 this free lsm to a new inode.
5497 Bugzilla : 6422, 7030
5498 Description: MDS deadlock between mkdir and client eviction
5499 Details : Creating a new file via mkdir or mknod (starting a transaction
5500 and getting the ns lock) can deadlock with client eviction
5501 (gets ns lock and trying to finish a synchronous transaction).
5504 Frequency : occasional
5505 Description: While starting a server, the fsfilt_ext3 module could not be
5507 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
5508 kernels. Previously, lconf would still use the ext3 name
5509 when trying to load modules. Now, it will correctly use
5510 ext3 on 2.4 and ldiskfs on 2.6.
5512 Severity : enhancement
5513 Description: The default stripe count has been changed to 1
5514 Details : The interpretation of the default stripe count (0, to lfs
5515 or lmc) has been changed to mean striping across a single
5516 OST, rather than all available. For general usage we have
5517 found a stripe count of 1 or 2 works best.
5519 Severity : enhancement
5520 Description: Add support for compiling against Cray portals.
5521 Details : Conditional compiling for some areas that are different
5525 Frequency : occasional
5526 Bugzilla : 6409, 6834
5527 Description: Creating files with an explicit stripe count may lead to
5528 a failed assertion on the MDS
5529 Details : If some OSTs are full or unavailable, creating files may
5530 trigger a failed assertion on the MDS. Now, Lustre will
5531 try to use other servers or return an error to the
5535 Frequency : occasional
5537 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
5538 caused lock timeout during readahead (since 1.4.2).
5539 Details : Processes doing readahead might match a lock that hasn't been
5540 granted yet if there are overlapping and conflicting lock
5541 requests. The readahead process waits on ungranted lock
5542 (original lock is CBPENDING), while OST waits for that process
5543 to cancel CBPENDING read lock and eventually evicts client.
5545 Severity : enhancement
5547 Description: Initial enabling of flock support for clients
5548 Details : Implements fcntl advisory locking and file status functions.
5549 This feature is provided as an optional mount flag (default
5550 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
5551 locking are implemented yet, and those that are are not guaranteed
5552 to be completely correct in production environments.
5553 mount -t lustre -o [flock|noflock] ...
5556 Frequency : occasional
5558 Description: OSTs running 2.4 kernels but with extents enabled might trip an
5559 assertion in the ext3 JBD (journaling) layer.
5560 Details : The b_committed_data struct is protected by the big kernel lock
5561 in 2.4 kernels, serializing journal_commit_transaction() and
5562 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
5563 access to this struct. In 2.6 kernels there is finer grained
5564 locking to improve SMP performance of the JBD layer.
5568 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
5573 Description: merge and cleanup kernel patches.
5576 -----------------------------------------------------------------------------
5578 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
5580 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
5581 versions of Lustre. Please update all servers and clients to
5582 version 1.4.2 or later at the same time. You must also run
5583 "lconf --write-conf {config}.xml" on the MDS while it is stopped
5584 to update the configuration logs.
5586 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
5587 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
5588 - swab llog records of type '0' so we get proper header size/idx (5861)
5589 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
5590 - fix rename of one directory over another leaking an inode (5953)
5591 - avoid SetPageDirty on 2.6 (5981)
5592 - don't re-add just-being-destroyed locks to the waiting list (5653)
5593 - when creating new directories, inherit the parent's custom
5594 striping settings if present parent (3048)
5595 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
5596 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
5597 - don't hold client locks on temporary worklist from l_lru (5666)
5598 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
5599 - automatically evict dead clients (5921)
5600 - Update file size properly in create+truncate+fstat case (6196)
5601 - Do not unhash mountpoint dentries, do not allow removal of
5603 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
5604 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
5605 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
5606 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
5607 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
5608 - Correctly handle reads of files with no objects (6243)
5609 - lctl recover will also mark a device active if deactivate used (5933)
5611 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
5612 - allow --write-conf on an MDS with different nettype than client (5619)
5613 - don't write config llogs to MDS for mounts not from that MDS (5617)
5614 - lconf should create multiple TCP connections from a client (5201)
5615 - init scripts are now turned off by default; run chkconfig --on
5616 lustre and chkconfig --on lustrefs to use them
5617 - upcalls are no longer needed for clients to recover to failover
5619 - add --abort-recovery option to lconf to abort recovery on device
5621 - add support for an arbitrary number of OSTs (3026)
5622 - Quota support protocol changes.
5623 - forward compatibility changes to wire structs (6007)
5624 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5625 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5626 - improved "lustre status" script
5627 - initialize blocksize for non-regular files (6062)
5628 - added --disable-server and --disable-client configure options (5782)
5629 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5630 - Vanilla 2.4.29 support
5631 - increase maximum number of obd devices to 520 (6242)
5632 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5633 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5634 SLES 9 2.6.5 kernels
5635 - limit stripes per file to 160 (the maximum EA size) (6093)
5637 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5640 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5641 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5642 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5643 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5644 - small ext3 extents cleanups and fixes (5733)
5645 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5646 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5647 - avoid needless client->OST connect, fix handle mismatch (5317)
5648 - fix DLM error path that led to out-of-sync client, long delays (5779)
5649 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5650 - fix several locking issues related to i_size (5492,5624,5654,5672)
5651 - don't move pending lock onto export if it is already evicted (5683)
5652 - fix kernel oops when creating .foo in unlinked directory (5548)
5653 - fix deadlock in obdfilter statistics vs. object create (5811)
5654 - use time_{before,after} to avoid timer jiffies wrap (5882)
5655 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5656 - Do not leak request if server was not able to process it (5154)
5657 - If mds_open unable to find parent dir, make that negative lookup(5154)
5658 - don't create new directories with extent-mapping (5909, 5936)
5660 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5661 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5662 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5663 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5664 - support for RHEL4 kernel on i686 (5773)
5665 - provide error messages when incompatible logs are encountered (5898)
5667 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5668 * version 1.4.0.10 (1.4.1 release candidate 1)
5670 - don't keep a lock reference when lock is not granted (4238)
5671 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5672 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5673 - drop import inflight refcount on signal_completed_replay error (5255)
5674 - unlock page after async write error during send (3677)
5675 - handle missing objects in filter_preprw_read properly (5265)
5676 - no transno return for symlink open, don't save no-trasno open (3440)
5677 - don't try to complete elan receive that already failed (4012)
5678 - free RPC server reply state on error (5406)
5679 - clean up thread from ptlrpc_start_thread() on error (5160)
5680 - readahead could read extra page into cache that wasn't ejected (5388)
5681 - prevent races in class_attach/setup/cleanup/detach (5260)
5682 - don't dereference de->d_inode after l_dput of de (5458)
5683 - use "int" for stripe value returned from lock_to_stripe (5544)
5684 - mballoc allocation and error-checking fixes in 2.6 (5504)
5685 - block device patches to fix I/O request sizes in 2.6 (5482)
5686 - look up hostnames for IB nals (5602)
5687 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5688 - don't start multiple acceptors for the same port (5277)
5689 - fix incorrect LASSERT in mds_getattr_name (5635)
5690 - export a proc file for general "ping" checking (5628)
5691 - fix "lfs check" to not block when the MDS is down (5628)
5693 - service request history (4965)
5694 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5695 - create an "evict_client" /proc entry on OSTs, like the MDS has
5696 - fix mount usage message, return errors per mount(8) (5168)
5697 - change grep [] to grep "[]" in tests so they work in more UMLs
5698 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5699 - remove ancient LOV_MAGIC_V0 EA support (5047)
5700 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5701 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5702 - allow readahead allocations to fail when low on memory (5383)
5703 - mmap locking landed again, after considerable improvement (2828)
5704 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
5706 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
5709 - send OST transaction number in read/write reply to free req (4966)
5710 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5711 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5712 - the watchdog thread now runs as interruptible (5246)
5713 - flock/lockf fixes (but it's still disabled, pending 5135)
5714 - don't use EXT3 constants in llite code (5094)
5715 - memory shortage at startup could cause assertion (5176)
5717 - reorganization of lov code
5718 - single portals codebase
5720 - add extents/mballoc support (5025)
5721 - direct I/O reads in the obdfilter (4048)
5722 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
5724 tbd Cluster File Systems, Inc. <info@clusterfs.com>
5727 - send OST transaction number in read/write reply to free req (4966)
5728 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5729 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5730 - the watchdog thread now runs as interruptible (5246)
5731 - handle missing objects in filter_preprw_read properly (5265)
5732 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5733 - drop import inflight refcount on signal_completed_replay error (5255)
5734 - unlock page after async write error during send (3677)
5735 - return original error code on reconstructed replies (3761)
5736 - no transno return for symlink open, don't save no-trasno open (3440)
5738 - add pid to ldlm debugging output (4922)
5739 - bump the watchdog timeouts -- we can't handle 30sec yet
5740 - extra debugging for orphan dentry/inode bug (5259)
5742 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
5745 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
5746 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
5747 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
5748 - fix NULL dereference in /proc/sys/portals/routes (4827)
5749 - allow failed mdc_close() operations to be interrupted (4561)
5750 - stop precreate on OST before MDS would time out on it (4778)
5751 - don't send partial-page writes before EOF from client (4410)
5752 - discard client grant for sub-page writes on large-page clients (4520)
5753 - don't free dentries not owned by NFS code, check generation (4806)
5754 - fix lsm leak if mds_create_objects() fails (4801)
5755 - limit debug_daemon file size, always print CERROR messages (4789)
5756 - use transno after validating reply (3892)
5757 - process timed out requests if import state changes (3754)
5758 - update mtime on OST during writes, return in glimpse (4829)
5759 - add mkfsoptions to LDAP (4679)
5760 - use ->max_readahead method instead of zapping global ra (5039)
5761 - don't interrupt __l_wait_event() during strace
5763 - add software watchdogs to catch hung threads quickly (4941)
5764 - make lustrefs init script start after nfs is mounted
5765 - fix CWARN/ERROR duplication (4930)
5766 - return async write errors to application if possible (2248)
5767 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
5768 - print NAL number in %x format (4645)
5769 - update barely-supported suse-2.4.21-171 series (4842)
5770 - support for sles 9 %post scripts
5771 - support for building 2.6 kernel-source packages
5772 - support for sles km_* packages
5774 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
5777 - ignore -ENOENT errors in osc_destroy (3639)
5778 - notify osc create thread that OSC is being cleaned up (4600)
5779 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
5780 - reconstruct ost_handle() like mds_handle() (4657)
5781 - create a new thread to do import eviction to avoid deadlock (3969)
5782 - let lconf resolve symlinked-to devices (4629)
5783 - don't unlink "objects" from directory with default EA (4554)
5784 - hold socknal file ref over connect in case target is down (4394)
5785 - allow more than 32000 subdirectories in a single directory (3244)
5786 - fix blocks count for O_DIRECT writes (3751)
5787 - OST returns ENOSPC from object create when no space left (4539)
5788 - don't send truncate RPC if file size isn't changing (4410)
5789 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
5790 - bind to privileged port in socknal and tcpnal (3689)
5792 - rate limit CERROR/CWARN console message to avoid overload (4519)
5793 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
5794 - basic mmap support (3918)
5795 - kernel patch series update from b1_4 (4711)
5797 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
5800 - avoid crash during MDS cleanup with OST shut down (2775)
5801 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
5802 - don't use bad inodes on the MDS (3744)
5803 - dynamic object preallocation to improve recovery speed (4236)
5804 - don't hold spinlock over lock dumping or change debug flags (4401)
5805 - don't zero obd_dev when it is force cleaned (3651)
5806 - print grants to console if they go negative (4431)
5807 - "lctl deactivate" will stop automatic recovery attempts (3406)
5808 - look for existing locks in ldlm_handle_enqueue() (3764)
5809 - don't resolve lock handle twice in recovery avoiding race (4401)
5810 - revalidate should check working dir is a directory (4134)
5812 - don't always mark "slow" obdfilter messages as errors (4418)
5814 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
5817 - don't close LustreDB during write_conf until it is done (3860)
5818 - fix typo in lconf for_each_profile (3821)
5819 - allow dumping logs from multiple threads at one time (3820)
5820 - don't allow multiple threads in OSC recovery (3812)
5821 - fix debug_size parameters (3864)
5822 - fix mds_postrecov to initialize import for llog ctxt (3121)
5823 - replace config semaphore with spinlock (3306)
5824 - be sure to send a reply for a CANCEL rpc with bad export (3863)
5825 - don't allow enqueue to complete on a destroyed export (3822)
5826 - down write_lock before checking llog header bitmap (3825)
5827 - recover from lock replay timeout (3764)
5828 - up llog sem before sending rpc (3652)
5829 - reduce ns lock hold times when setting kms (3267)
5830 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
5831 - fix NULL deref and obd_dev leak on setup error (3312)
5832 - replace some LBUG about llog ops with error handling (3841)
5833 - don't match INVALID dentries from d_lookup and spin (3784)
5834 - hold dcache_lock while marking dentries INVALID and hashing (4255)
5835 - fix invalid assertion in ptlrpc_set_wait (3880)
5837 - add libwrap support for the TCP acceptor (3996)
5838 - add /proc/sys/portals/routes for non-root route listing (3994)
5839 - allow setting MDS UUID in .xml (2580)
5840 - print the stack of a process that LBUGs (4228)
5842 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
5845 - don't cleanup request in ll_file_open() on failed MDS open (3430)
5846 - make sure to unset replay flag from failed open requests (3440)
5847 - if default stripe count is 0, use OST count for inode size (3636)
5848 - update parent mtime/ctime on client for create/unlink (2611)
5849 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
5850 - free recovery state on server during a forced cleanup (3571)
5851 - unregister_reply for resent reqs (3063)
5852 - loop back devices mounting and status check on 2.6 (3563)
5853 - fix resource-creation race that can provoke i_size == 0 (3513)
5854 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
5855 - more debugging for page-accounting assertion (3746)
5856 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
5857 - avoid deadlock after precreation failure (3758)
5858 - fix race and lock order deadlock in orphan handling (3450, 3750)
5859 - add validity checks when grabbing inodes from l_ast_data (3599)
5861 - add /proc/.../recovery_status to obdfilter (3428)
5862 - lightweight CDEBUG infrastructure, debug daemon (3668)
5863 - change default OSC RPC parameters to be better on small clusters
5864 - turn off OST read cache for files smaller than 32MB
5865 - install man pages and include them in rpms (3100)
5866 - add new init script for (un)mounting lustre filesystems (2593)
5867 - run chkconfig in %post for init scripts (3701)
5868 - drop scimac NAL (unmaintained)
5870 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
5873 - clean kiobufs before and after use (3485)
5874 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
5875 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
5876 - change init script to fail more clearly if not run as root (1528)
5877 - allow clients to reconnect during replay (1742)
5878 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
5879 - don't do DNS lookups on NIDs too small for IP addresses (3442)
5880 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
5881 - fix cond_resched (3554)
5882 - only evict unfinished clients after recovery (3515)
5883 - allow bulk resend, prevent data loss (3570)
5884 - dynamic ptlrpc request buffer allocation (2102)
5885 - don't allow unlinking open directory if it isn't empty (2904)
5886 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
5887 - remove extraneous obd dereference causing LASSERT failure (3334)
5888 - don't use get_cycles() when creating temp. files on the mds (3156)
5889 - hold i_sem when setting i_size in ll_extent_lock() (3564)
5890 - handle EEXIST for set-stripe, set proper directory name (3336)
5892 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
5893 - fix ksocknal_fmb_callback() error messages (2918)
5895 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
5898 - don't copy lvb into (possibly NULL) reply on error (2983)
5899 - don't deref dentry after dput, don't free lvb on error (2922)
5900 - use the kms to determine writeback rpc length (2947)
5901 - increment oti_logcookies when osc is inactive (2948)
5902 - update client's i_blocks count via lvb messages (2543)
5903 - handle intent open/close of special files properly (1557)
5904 - mount MDS with errors=remount-ro, like obdfilter (2009)
5905 - initialize lock handle to avoid ASSERT on error cleanup (3057)
5906 - don't use cancelling-locks' kms values (2947)
5907 - use highest lock extent for kms, not last one (2925)
5908 - don't dereference ERR_PTR() dentry in error handling path (3107)
5909 - fix thread race in portals_debug_dumplog() (3122)
5910 - create lprocfs device entries at setup instead of at attach (1519)
5911 - common AST error handler, don't evict client on completion race (3145)
5912 - zero nameidata in detach_mnt in 2.6 (3118)
5913 - verify d_inode after revalidate_special is valid in 2.6 (3116)
5914 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
5915 - initialize RPC timeout timer earlier for 2.6 (3219)
5916 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
5917 - print nal/nid for unknown nid (3258)
5918 - additional checks for oscc recovery before doing precreate (3284)
5919 - fix ll_extent_lock() error return code for 64-bit systems (3043)
5920 - don't crash in mdc_close for bad permissions on open (3285)
5921 - zero i_rdev for non-device files (3147)
5922 - clear page->private before handing to FS, better assertion (3119)
5923 - tune the read pipeline (3236)
5924 - fix incorrect decref of invalidated dentry (2350)
5925 - provide read-ahead stats and refine rpc in flight stats (3328)
5926 - don't hold journal transaction open across create RPC (3313)
5927 - update atime on MDS at close time (3265)
5928 - close LDAP connection when recovering to avoid server load (3315)
5929 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
5930 - don't leak open file on MDS after open resend (3325)
5931 - serialize filter_precreate and filter_destroy_precreated (3329)
5932 - loop device shouldn't call sync_dev() for nul device (3092)
5933 - clear page cache after eviction (2766)
5934 - resynchronize MDS->OST in background (2824)
5935 - refuse to mount the same filesystem twice on same mountpoint (3394)
5936 - allow llmount to create routes for mounting behind routers (3320)
5937 - push lock cancellation to blocking thread for glimpse ASTs (3409)
5938 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
5939 - fix rare problem with rename on htree directories (3417)
5941 - allow default OST striping configuration per directory (1414)
5942 - fix compilation for qswnal for 2.6 kernels (3125)
5943 - increase maximum number of MDS request buffers for large systems
5944 - change liblustreapi to be useful for external progs like lfsck (3098)
5945 - increase local configuration timeout for slow disks (3353)
5946 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
5948 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5951 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
5952 - don't overwrite extent policy data in reply if lock was blocked (2901)
5953 - drop filter export grants atomically with removal from device (2663)
5954 - del obd_self_export from work_list in class_disconnect_exports (2908)
5955 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
5956 - swab reply message in mdc_close, other PPC fixes (2464)
5957 - fix destroying of named logs (2325)
5958 - overwrite old logs when running lconf --write_conf (2264)
5959 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
5960 - fix race in target_handle_connect (2898)
5961 - mds_reint_create() should take same inode create lock (2926)
5962 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
5963 - don't close files for self_export to avoid uninitialized obd (2936)
5964 - allow MDS with the same name as client node (2939)
5965 - hold dentry reference for closed log files for unlink (2325)
5966 - reserve space for all logs during transactions (2059)
5967 - don't evict page beyond end of stripe extent (2925)
5968 - don't oops on a deleted current working directory (2399)
5969 - handle hard links to targets without a parent properly (2517)
5970 - don't dereference NULL lock when racing during eviction (2867)
5971 - don't grow lock extents when lots of conflicting locks (2919)
5973 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
5976 - account for cache space usage on clients to avoid data loss (974)
5977 - lfsck support in lustre kernel code (2349)
5978 - reduce journal credits needed for BRW writes (2370)
5979 - orphan handling to avoid losing space on client/server crashes
5980 - ptlrpcd can be blocked, stopping ALL progress (2477)
5981 - use lock value blocks to assist in proper KMS, faster stat (1021)
5982 - takes i_sem instead of DLM locks internally on obdfilter (2720)
5983 - recovery for initial connections (2355)
5984 - fixes for mds_cleanup_orphans (1934)
5985 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
5986 - block all file creations until orphan recovery completes (1901)
5987 - client remove rq_connection from request struct (2423)
5988 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
5989 - recovery timer race (2670)
5990 - mdc_close recovey bug (2532)
5991 - ptlrpc cleanup bug (2710)
5992 - mds timeout on local locks (2588)
5993 - namespace lock held during RPCs (2431)
5994 - handle interrupted sync write properly (2503)
5995 - don't try to handle a message that hasn't been replied to (2699)
5996 - client assert failure during cleanup after abort recovery (2701)
5997 - leak mdc device after failed mount (2712)
5998 - ptlrpc_check_set allows timedout requests to complete (2714)
5999 - wait for inflight reqs when ptlrpcd finishes (2710)
6000 - make sure unregistered services are removed from the srv_list
6001 - reset bulk XID's when resending them (caught by 1138 test)
6002 - unregister_bulk after timeout
6003 - fix lconf error (2694)
6004 - handle write after unfinished setstripe, stripe-only getstripe (2388)
6005 - readahead locks pages, leaves pending causing memory pressure (2673)
6006 - increase OST request buffers to 4096 on large machines (2729)
6007 - fix up permission of existing directories in simple_mkdir (2661)
6008 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
6009 - don't assign transno to errored transactions (2742)
6010 - don't delete objects on OST if given a bogus objid from MDS (2751)
6011 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
6012 - if rq_no_resend, then timeout request after recovery (2432)
6013 - fix MDS llog_logid record size, 64-bit array alignment (2733)
6014 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
6015 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
6016 - MDS recovery shouldn't delete objects that it has given out (2730)
6017 - if enqueue arrives after completion, don't clobber LVB (2819)
6018 - don't unlock pages twice when trigger_group_io returns error (2814)
6019 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
6020 - don't write pages to disk if there was an error (1450)
6021 - don't ping imports that have recovery disabled (2676)
6022 - take buffered bytes into account when balancing socknal conn (2817)
6023 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
6024 - reconnect unlink llog connection after MDS reconnects to OST (2816)
6025 - remove little-endian swabbing of llog records (1987)
6026 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
6027 - retry reposting request buffers if they fail (1191)
6028 - grow extent at grant time to avoid granting a revoked lock (2809)
6029 - lock revoke doesn't evict page if covered by a second lock (2765)
6030 - disable VM readahead to avoid reading outside lock extents (2805)
6032 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
6033 - updated kernel patches for hp-2.4.20 kernel (2681)
6035 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
6038 - fix truncated write corruption (2366)
6039 - fix for failed assertion in iopen_connect_dentry (1792,2517)
6041 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
6042 - ensure len(uuid) < 37 in lmc (1171)
6043 - fix ia64 OOPS in llog_test (2255)
6044 - zero end of page at obdfilter for partial page writes (2648)
6045 - don't leave stale dentries around after renames (bug 2428)
6046 - fix timeouts when evicting a client with a single lock held (2642)
6047 - set deadline for the initial HELLO message to drain (2634)
6048 - print out dotted-quad IP addresses in the socknal (2302)
6050 - additional debugging for MDS client eviction problem (2443)
6051 - fix mkfsoptions support for osts (2603, 2604)
6053 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
6056 - add series for the vanilla 2.6.0 kernel
6057 - add series for the vanilla 2.4.24 kernel
6058 - add series for a cray x86/64 UL kernel drop
6059 - fix xattr patches for the vanilla 2.4.19 series
6061 - generate true UUIDs in lmc (1171)
6062 - have portals stack dumping break in UML (2466)
6063 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
6064 - allocate with _NOFS in ldlm to avoid deadlock (1933)
6065 - wake callback waiting threads on client eviction (2460)
6066 - Add --ptldebug and --subsystem to lmc (1719)
6067 - update assertion to allow safe interrupt allocation
6068 - set rq_no_resend for cancel requests (2432)
6069 - recalculate ptlrpcd timeout after resend (2494)
6070 - call vfs_rmdir when removing pending directories (2368)
6071 - fix renaming a file to itself (2429)
6072 - lmc creates a default one-stripe lov (2454)
6073 - expand procfs space to handle large clusters (2326)
6074 - increase UML stack to avoid overflow
6075 - update lconf's list of debug and subsystem masks
6076 - fix lfs find --obd (2510)
6077 - /proc tunable for disabling filter read caching (2591)
6078 - stop rpm packages from altering slapd.conf (2301)
6079 - disable nagle in the socknal under 0conf (2578)
6080 - choose mds inode size based on stripe count (2572)
6081 - fix kernel-source rpm problems (2516)
6083 - add --disable-doc to avoid pdf generation (2421)
6084 - update documentation, tests, type-os, comments
6085 - avoid format warnings on ia64
6086 - remove the TOE NAL
6087 - tiny code cleanups by removing unused fields
6089 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
6092 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
6093 - remove the most blatant lies from BUILDING file (2371)
6094 - change default debug level to reasonable production setting
6095 - reduce client side cache size to reduce cache flush time
6096 - reduce max RPCs in flight to avoid unnecessary file fragmentation
6097 - make TCP zerocopy and pinger support enabled by default (2476)
6098 - sync writes completed after process exits caused crashes (2319)
6099 - maintain correct mount count on the MDS (2356)
6100 - backout 1557, because 2316 wasn't really fixed
6101 - better file I/O statistics gathering in /proc
6102 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
6103 - another kernel patch to fix zero-copy TCP function export
6104 - don't take duplicate lock when processing re-sent getattr (2420)
6105 - lctl uses obd_self_export instead of creating new conn (2353)
6106 - MDS/OST recovery case which requires object creation asserted (2425)
6107 - move lfs from /usr/sbin to /usr/bin in packages
6108 - fix race between mds_client_add and mds_client_free (2417)
6109 - use kmalloc instead of slabs in portals (2430)
6110 - don't create duplicate records when a failover MDS is present (2442)
6111 - remove unnecessary mount age check (2332)
6112 - don't remove directory inodes from locks prematurely (2451)
6113 - don't break if MDS service name is the same as hostname (2103)
6114 - fix races in client write RPC generation when cache full (2482)
6116 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
6119 - remove now-unused request->rq_obd (278)
6120 - if an allocation fails, print out how much memory we've used (1933)
6121 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
6122 - add the "configurable stack size" patch to most series files (1256)
6123 - ability to write large log records, for 100+ OST configs (2306)
6124 - fix NULL deref when filter_prep fails (2314)
6125 - fix operator precedence error in filter_sync
6126 - dynamic allocation of socknal TX descriptors (2315)
6127 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
6128 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
6129 - ensure that CWARN messages in Portals always get to the syslog
6130 - __init/__exit are not for prototype decls (ldlm_init/exit)
6131 - x86-64 compile warning fixes
6132 - fix gateway LMC keyword conflict (2318)
6133 - fix MDS lock inversions in getattr/reint paths (1844)
6134 - fix a rare lock re-ordering bug, which caused deadlock (2322)
6135 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
6136 - DLM race condition prevented some lock evictions (2328)
6137 - ENOMEM detection and retry on socknal sends (2230)
6138 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
6139 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
6140 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
6141 - better error messages when a client is rejected during recovery (1505)
6142 - avoid cancelling locks which were never granted, after failure (2330)
6143 - fix i_sem/journal inversion in mds_client_add (2333)
6144 - fix truncate/getattr lock cycle deadlock (2334)
6145 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
6146 - fix two rare exit paths which could leak an l_lock() ref (2321)
6147 - fencepost error in MDS/OST orphan recovery (2226)
6148 - make log record alignment 8 bytes (1988)
6149 - lstripe now fails when requested offset > ost_count (2237)
6150 - ensure that all kernel series have a complete list.h (1607)
6151 - fix crashes in special-file operations (2316)
6152 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
6154 - allow configurable automake binary, for testing new versions
6155 - small update to the lfs documentation
6157 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
6159 * fix negative export reference count in fsfilt_sync (2312)
6161 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
6162 * release candidate 0.9.1
6164 - orphans are moved into the PENDING directory for possible recovery
6165 - replayed opens now open by fid for orphan/rename safety (1042)
6166 - last close of an orphan inode generates a transno (683)
6167 - chdir() and mount() now pin the directory entry (1020)
6168 - avoid CERROR in normal ll_setattr_raw() error case (1500)
6169 - discard very old requests without processing them (1502)
6170 - remove some common, well-understood CERRORs (1505)
6171 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
6172 - clear "grant" flags in OST replies until OST grant code lands (1644)
6173 - fix read performance by not clobbering i_blksize on client (1598)
6174 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
6175 - make LRU size a /proc tunable, clears locks when reduced (707)
6176 - fix some lprocfs rot that prevented ptlbd from loading (1732)
6177 - server locks take references on exports now (1558)
6178 - build fixes for 2.4.20-rh trees (1663)
6179 - return an error from lov_create if all OSCs are inactive (1751)
6180 - fix import levels when a reconnect happens without a timeout (1597)
6181 - exit early from mds_open if we get a lookup error (1749)
6182 - partial page read at EOF wouldn't wait for disk before sending (1642)
6183 - avoid NULL deref in obdfilter when reading page past EOF (1592)
6184 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
6185 - fix LBUG in ll_it_open_error with rc = -2 (1861)
6186 - write/truncate lock inversion (1639)
6187 - Don't auto-load obdclass, portals modules during cleanup (1495)
6188 - fix timestamps from jumping to "now" (1763)
6189 - extra journal assertions (1648)
6190 - add an extra multiunlink test (1771)
6191 - fix read_record/write_record API (1776)
6192 - fix leak of offset_extent, possible incorrect i_size later (1772)
6193 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
6194 - Debugging for the kqswnal_get_idle_tx problems (1820)
6195 - Allow recovery to be attempted multiple times (1536)
6196 - Write out MDS last_rcvd file after it is first created (1600)
6197 - Fix tx_descriptor leak in failed transmit situations (1827)
6198 - ext3 journaling fixes for assertion failure after IO error (1871)
6199 - class_export_put() on freed export after completion AST error (1896)
6200 - Fix revalidate looping in VFS (1322)
6201 - Don't access a freed export during MDS_REINT timeout (1521)
6202 - Add open-unlink recovery support on the MDS (1673,1764)
6203 - Return an error if no MDS data was read from last_rcvd (1946)
6204 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
6205 - Don't setup a disk device that doesn't match exported UUID (317)
6206 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
6207 - avoid committing NULL handle in force close
6208 - local.sh is now a one-stripe LOV configuration
6209 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
6210 - fix ext3 htree duplicate directory entry corruption (1516)
6211 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
6212 - update to new LOV EA format (2097)
6213 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
6215 2003-06-15 Phil Schwan <phil@clusterfs.com>
6218 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
6219 - per-import/export recovery handling (958, 931, 959)
6220 - multiple last-rcvd slots, for serving multiple FSes (949)
6221 - connections are again shared between multiple imp/exports (963, 964)
6222 - "umount -f" would hang if any requests needed to be sent (393, 978)
6223 - avoid pinning large req buffer by copying for queued messages (989)
6224 - add "uuid" to "lctl device" command to help upcalls (991)
6225 - "open" RPCs with transnos would confuse recovery counters (1037)
6226 - do proper endian conversion of all wire messages (288, 340, 891)
6227 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
6228 - call waiting locks callback from LDLM recovery thread (1127, 1151)
6229 - fix ptlrpc_connection leak in target_handle_connect (1174)
6230 - fix import refcounting bug in OST and MDS cleanup (1134)
6231 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
6232 - fix very unlikely obd_types race condition (501)
6233 - remove osc_open hack for echo_client (1187)
6234 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
6235 - a failure in read_inode2 leads to deadlock (1139)
6236 - cancel ack-locks as soon as transaction is committed (1072)
6237 - fix major leaks and crashes in the bulk I/O path (937, 1057)
6238 - make sure to commitrw after any preprw to avoid deadlock (1162)
6239 - failing to execute a file in a lustre FS would lock inode (1203)
6240 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
6241 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
6242 - obd_uuid2tgt would walk off the end of the list (1255)
6243 - on IA64 the osc would give portals incorrect bulk size (1258)
6244 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
6245 - fix lock inversion caused by new llite matching code (1282)
6246 - limit the number of dirty pages on a client to 10MB (1286)
6247 - timed out locks were not being corrected cancelled (1289)
6248 - fix O_DIRECT above 4GB on IA-32 (1292)
6249 * major user-visible changes
6250 - fail out/fail over policy now controlled by the upcall (993)
6252 - add OBD_PING to check server availability and failure (954)
6253 - lustre messages are now sent in sending host order (288, 340, 891)
6254 - add eadatalen to MDS getattr reply (340)
6255 - OST read replies may contain second buffer, with per-page status (593)
6257 2003-03-11 Phil Schwan <phil@clusterfs.com>
6260 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
6261 - failed open()s could cause deadlock; fixed (867, 869)
6262 - stop cancelling OST locks when files are closed (481)
6263 - overlapping XID spaces caused network corruption (851, 853)
6264 - fix unsafe fsfilt counter arithmetic; change to atomic_t
6265 - setattr_raw added, to do single-RPC, server-side setattrs
6266 - lmc/lconf syntax change for OST UUIDs
6267 - fix crashy race condition between ptlrpc_free_req and osc_close
6268 - don't use request in mdc_enqueue if we hit a timeout (889)
6269 - don't set the inode i_size for regular files from the MDS (896)
6270 - handle out of order completion AST (842)
6271 - don't LBUG if a lock request times out after receiving AST (913)
6272 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
6273 - if a bad lock AST arrives, send an error instead of dropping entirely
6274 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
6275 - fix leak in bulk IO when only partially completed (899, 900, 926)
6276 - fix O_DIRECT for ia64 (55)
6277 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
6278 - C-z after timeout could hang a process forever; fixed (977)
6280 - client-side I/O cache (678, 924, 929, 941, 970)
6282 - READPAGE and SETATTRs which don't take server-side locks get
6285 2003-02-11 Phil Schwan <phil@clusterfs.com>
6288 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
6289 - Fix fsfilt_extN_readpage() to read a full page of directory
6290 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
6291 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
6292 - Fix imbalanced LOV object allocation and out-of-bound access (469)
6293 - Most intent operations were removed, in favour of a new RPC mode
6294 that does a single RPC to the server and bypasses most of the VFS
6295 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
6296 - Aggressively cancel local locks on DLM servers
6297 - mds_reint_unlink sends EA to the client if it's the last nlink.
6298 client uses that EA to unlink OST objects.
6299 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
6300 - recursive symlinks were fixed (439)
6301 - fixed NULL deref in DEBUG_REQ
6302 - filter_update_lastobjid no longer calls sync, which annoyed extN
6303 - fixed multi-client small-writes to a single file problem (445)
6304 - fixed mtime updates during file writes (607)
6305 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
6306 - fixed bug in obd_brw_read/write() (under guise of testing 367)
6307 - fixed Linux OST size reporting problem (444, 656)
6308 - OST now updates object mtime with writes or setattr (607, 619)
6309 - client verifies file size before zeroing page past EOF (445)
6310 - OST now writes last allocated objid to disk with allocation (108)
6311 - LOV on echo now works (409)
6313 - mds_reint_unlink sends a new buffer, with the EA included. this
6314 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
6315 set if a regular file was being unlinked, and it was the last link
6316 - use PtlGet from the target for bulk writes (315)
6317 - OST now updates object mtime with writes or setattr (607, 619)
6318 - LDLM now has a grant-time callback to revalidate locked items, if
6320 - Many MDS operations were reorganized to combat race conditions
6322 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
6323 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
6325 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
6328 - Fully reactivate OST imports after reconnection (512, others)
6329 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
6330 - More graceful error handling for truncating on dead OST (515)
6331 - Don't error out unless we're actually accessing dead stripes (474)
6332 - Fix garbage sizes when stripes are missing (410)
6333 - LRU counters were broken, causing constant lock purge (433, 432)
6334 - garbage on read from stripes with failed OSTs (441)
6335 - mark OSCs as active before reconnecting during recovery (438)
6336 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
6337 - lfind did not preserve OST order in output (443)
6338 - symlinks cause hung clients, incorrect data (439)
6339 - stop dereferencing request after dropping refcount (457)
6340 - don't LASSERT(spin_is_locked) on non-SMP (455)
6341 - fixes for many rename() bugs
6342 - fstat didn't correctly synchronize attributes (399)
6343 - server must handle lock cancellation during blocking AST prep (487)
6344 - bulk descriptors were free()d too soon (511)
6345 - fix paths in lconf, which would load incorrect modules (451, 507)
6346 - fix confusing lconf 'host not found' error message (386)
6347 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
6348 - fix race condition in mdc_blocking_ast() for inode access (526)
6349 - fix lov_unpackmd() unpacking wrong number of stripes (537)
6350 - fix lov_set_osc_active() marking wrong OSC inactive (440)
6351 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
6352 - fix multiple writes of stripe MD to MDS (358, maybe 519)
6353 - fix lstripe in several ways (kernel side) (527)
6354 - fix request leak in ldlm_cli_enqueue (262)
6355 - incorrect OSC was marked inactive after OST failure
6356 - call mds_fs_cleanup before unmounting filesystem (524)
6357 - fix races between taking ns_lock and ldlm_lock_change_resource
6358 - fix races updating LOV export open file list
6359 - fix lov_enqueue error path, avoid decref-ing bad lock handle
6360 - fix recovery NULL deref in ldlm_cli_cancel_unused
6361 - fix some DLM races by using new hash table for lock handles (419)
6362 - permit the client to specify desired inodes, at replay
6363 - duplicate requests when we queue them for replay reintegration
6364 - fix last_rcvd offset calculation
6365 - sync after each recovered transaction, so we always make progress
6366 - never, not always, ERESTART requests without transnos
6367 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
6369 - skip replay if the MDS says that the client is already connected
6370 - don't check for a recovery-enabled export to match lctl's UUID
6371 - don't INC_USE_COUNT for phantom exports
6372 - don't crash when cleaning up phantom exports (567)
6373 - don't double-finish or set replay data for errored mdc_open requests
6374 - abort requests when they time out, so we don't get old replies
6375 - send/receive replies for AST messages again
6376 - if the client says that it doesn't have the lock, cancel it on the
6378 - if we timeout during I/O, don't try to cancel an in-use lock; instead
6379 mark it as destroyed, it will all work out when decref is called
6380 - fix module use counts (22, 581)
6382 - ASTs now expect a reply (server cancels lock on error reply)
6384 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
6387 - fix many simultaneous client startup (392)
6388 - fix dentry->d_it clobbering
6389 - credentials weren't being shipped for readdir/getattr operations
6390 - remove invalid assertions triggered during some concurrent MD
6392 - proper Lustre versions added (336, 389)
6393 - fix memory leak for create error case (398)
6394 - fix LOV locking bug that would get cli/srv out of sync
6395 - fix echo client over LOV (409)
6396 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
6397 - fix double-O_EXCL intent crash (424)
6398 - avoid sending multiple lock CANCELs (352)
6400 - MDS can do multi-client recovery (modulo bugs in new code)
6402 - many updates, edits, cleanups
6404 2002-11-18 Phil Schwan <phil@clusterfs.com>
6407 - fix null d_it dereference (346)
6408 - fix full OST/dbench hang (333)
6409 - fix permission problem with file removal (286)
6410 - fix removal of OSCs from LOV when they fail
6411 - fix NULL deref during bulk timeout (214)
6412 - fix problems related to multiple filesystems on one MDS (241)
6413 - fixed serious subtle metadata locking bugs
6414 - free locks on clients when inodes are removed due to memory
6416 - fix inode pointer in lock data (285)
6417 - partial support for multiple MDS on a single host (241)
6418 - data locks weren't cancelled at clear_inode time (290, 311)
6419 - intent locks could lead to unbounded lock growth (205)
6420 - added a maximum lock count, an LRU list, and a flusher
6421 - fix multiple rename (365)
6422 - properly abstracted the echo client
6423 - OSC locked 1 byte too many; fixed
6424 - rewrote brw callback code:
6425 - fixed recovery bugs related to LOVs (306)
6426 - fixed too-many-pages-in-one-write crash (191)
6427 - fixed (again) crash in sync_io_timeout (214)
6428 - probably fixed callback-related race (385)
6430 - Add capability to MDS protocol
6431 - LDLM cancellations and callbacks on different portals
6433 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
6436 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
6437 - defer open object destruction to close time (601981, 138)
6438 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
6439 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
6440 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
6441 - don't call obd_setattr in ll_file_release for destroyed objects
6442 * protocol change to lustre_msg: move |version| and add |flags|
6443 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
6444 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
6445 * added replay of create, unlink, link and rename operations during
6446 MDS failover; recovery should be much more robust now
6447 * remove failed OSCs from LOVs (only lov_create uses this so far)
6448 * the lustre-HOWTO was brought (more) up to date (582544)
6450 2002-10-23 Phil Schwan <phil@clusterfs.com>
6453 - in-use dentries weren't being reused properly (617851)
6454 - prevent multiple LDLM setup (599178)
6455 - fix LOV size calculations for truncate (617853)
6456 - fix client handling of MDS intent errors (POSIX)
6457 - fix permission bug in lovstripe.c test (624321)
6458 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
6459 - truncate past end of file could corrupt data
6460 - proper cleanup after timeouts, crashes, etc (592524, 550815)
6461 - a race in recovery could return ETIMEDOUT to apps (623947)
6462 - building outside the source directory was fixed
6463 * the lustre-HOWTO was brought (more) up to date (582544)
6464 * major progress was made on recovery functionality
6466 2002-10-10 Phil Schwan <phil@clusterfs.com>
6469 - recovery deadlock fix
6470 - rm -rf causes LBUG fix (617817)
6471 - file open by multiple tasks fix (618962)
6472 - directory permissions bugs (602707 and 620007)
6473 - journal_stop fixed with locking (611313)
6474 - O_APPEND failures resolved (618273, perhaps 614459)
6475 - lconf PATH fix (619770)
6476 - IA64 build fix (621450)
6477 - RPC buffer sizes scale with amount of memory
6479 2002-10-01 Phil Schwan <phil@clusterfs.com>
6482 - locks would be cancelled without throwing away data pages,
6483 resulting in inconsistent data (605627)
6484 - inode attributes were not always being refreshed (605627, 612449)
6485 - lconf now continues to cleanup after lctl reports an error
6486 - MDS now enforces user permissions (602707)
6487 - lprocfs cleanup fixed, but not yet enabled (614157)
6488 - fixed infinite server hang, should a client not respond to an AST
6489 - avoid going into recovery if user calls readlink() with a buffer
6490 that's too small (613941)
6491 - AST RPCs no longer require replies (614867) -- this may be changed
6492 - don't crash server if client sends an IOV that's too big (611336)
6493 - fixed lock conversion deadlock (611892)
6494 - fixed the following of symlinks (614622)
6495 * recovery: the server can remove locks from a client that dies, other
6496 clients can make progress
6497 * more extN patch fixes
6498 * compile-time configurable ptlrpc buffer allocations
6500 - collaborative read cache document
6501 - Lustre Lite Performance CDR document-in-progress
6503 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6506 - fix typo in patch-2.4.18
6508 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6511 - clear ptlrpc request each time in handle_incoming_request()
6512 - unlink of files now destroys the object on the OST
6514 2002-09-19 Peter Braam <braam@clusterfs.com>
6516 * add hard link support
6517 * change obdfile creation method
6518 * kernel patch changed
6520 2002-09-19 Peter Braam <braam@clusterfs.com>
6523 - stack overflow bug in extN fixed
6525 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
6527 * documentation updates
6528 - add man pages for config tools
6529 - update tests/README to describe testing with new config tools
6530 - finish metadata API descriptions
6531 * bug fixes and cleanups
6532 - statfs workaround for 16TB limit
6533 - LOV stripe allocation improved, can stripe on subset of OSTs
6534 - LOV file size/IO offset was wrong for files > 4GB in size
6535 - object EA data was being dropped, caused files to be unreadable
6536 - memory overflow with non-LOV OST caused memory corruption
6537 - fixed regression tests to work with new config tools, obdfilter
6538 - fixed bug when directory size became larger than 1 block
6539 - fixed bug (for single client case) when PWD was deleted
6540 - invalidate local directory pages when doing intent-based ops
6541 - avoid LDLM oops when lock callback contained bad data
6543 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
6545 * documentation updates
6546 * bug fixes and cleanups
6547 - configuration tools
6550 - 64-bit compile warnings
6551 - 64-bit internal statfs data
6553 * test_brw on persistent OST devices
6555 * lprocfs (disabled)
6557 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
6559 * documentation updates
6560 * bug fixes and cleanups
6561 * configuration tools
6563 2002-08-30 Peter J. Braam <braam@clusterfs.com>
6566 * many small fixes to 0_5_4
6567 * io/network handling
6568 * thinkos in MDS operations
6570 2002-08-24 Peter J. Braam <braam@clusterfs.com>
6573 * crucial basic fixes to 0.5.3
6574 * IOR, Iozone work over Elan
6577 2002-08-07 Phil Schwan <phil@clusterfs.com>
6578 * version 0_5_3, our first alpha
6579 * we use the new Portals iovs
6580 * documentation updates
6581 * bug fixes and cleanups
6582 * small changes in the DLM wire protocol
6584 2002-07-25 Peter J. Braam <braam@clusterfs.com>
6585 * version 0_5_1 with some initial stability,
6586 * locking on MD and file I/O.
6587 * documentation updates
6588 * several bug fixes since 0.5.0
6589 * small changes in wire protocol
6591 2002-07-18 Phil Schwan <phil@clusterfs.com>
6593 * delivered as Lustre Light Alpha
6594 * fixed a crash after handling invalid MDS requests
6595 * fixed directory pages for architectures with non-4k pages sizes
6597 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
6598 * release version v0_4_4
6599 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
6601 * Fixes a number of interruption problems with OST operations.
6602 * Update documentation for portals header changes
6603 * Move all wire protocol structs/defines to lustre_idl.h
6604 * Fixes symlink length bug.
6605 * Add tcpdump to repository.
6607 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
6608 * release version v0_4_3
6609 * Fixes statfs for inodes on extN.
6610 * Fixes bug in runtests which would delete /etc/hosts.
6611 * Use 64-bit object IDs wherever possible (not into VFS though)
6612 Remove ost_get_info, which is unused by lustre, and out of date.
6614 2002-07-03 Peter Braam <braam@clusterfs.com>
6615 * release version v0_4_2 Fixes a lookup error (type not passed)
6616 * move forward to head of Portals
6617 * move forward to latest Lustre kernel
6619 2002-06-25 Peter Braam <braam@clusterfs.com>
6620 * release version v0_4_1. Hopefully stable on single node use.
git://git.whamcloud.com / fs / lustre-release.git / blob