3 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
4 for currently supported client and server kernel versions.
5 * Server known to build on the following kernels (others may also work):
6 2.6.32-696.18.7.el6 (RHEL6.9)
7 3.10.0-862.14.4.el7 (RHEL7.5)
8 4.4.120-92.70 (SLES12 SP2)
9 4.4.155-94.50 (SLES12 SP3)
10 4.4.0-131 (Ubuntu 16.04)
11 4.15.0-32 (Ubuntu 18.04)
12 vanilla linux 4.6.7 (ZFS only)
13 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
14 * Client known to build on unpatched kernels (others may also work):
15 2.6.32-696.18.7.el6 (RHEL6.9)
16 2.6.32-754.6.3.el6 (RHEL6.10)
17 3.10.0-862.14.4.el7 (RHEL7.5)
18 4.14.0-49.13.1.el7a (RHEL7.5)
19 4.4.120-92.70 (SLES12 SP2)
20 4.4.155-94.50 (SLES12 SP3)
21 4.4.0-131 (Ubuntu 16.04)
22 4.15.0-32 (Ubuntu 18.04)
23 vanilla linux 4.15.0 (ZFS only)
24 * Recommended e2fsprogs version: 1.44.3.wc1 or newer
25 * Recommended ZFS / SPL version: 0.7.9
26 * Tested with ZFS / SPL version: 0.7.9
27 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
28 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
29 stack overflow. For more information, please refer to bugzilla 17630.
30 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
31 the re-exporting nodes to be mounted with "32bitapi" mount option
32 * Note for Ubuntu 14.04.5:
33 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
34 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
36 --------------------------------------------------------------------------------
39 2018-04-03 Intel Corporation
41 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
42 for currently supported client and server kernel versions.
43 * Server known to build on patched kernels:
44 2.6.32-431.29.2.el6 (RHEL6.5)
45 2.6.32-504.30.3.el6 (RHEL6.6)
46 2.6.32-573.26.1.el6 (RHEL6.7)
47 2.6.32-642.15.1.el6 (RHEL6.8)
48 2.6.32-696.18.7.el6 (RHEL6.9)
49 3.10.0-693.21.1.el7 (RHEL7.4)
50 3.0.101-0.47.71 (SLES11 SP3)
51 3.0.101-107 (SLES11 SP4)
52 3.12.74-60.64.40 (SLES12 SP1)
53 4.4.114-92.64 (SLES12 SP2)
54 4.4.114-94.11 (SLES12 SP3)
55 3.13.0-101 (Ubuntu 14.04, ZFS only)
56 4.4.0-85.108 (Ubuntu 14.04.5 LTS)
57 4.4.0-116 (Ubuntu 16.04)
58 vanilla linux 4.6.7 (ZFS only)
59 * Client known to build on unpatched kernels:
60 2.6.32-431.29.2.el6 (RHEL6.5)
61 2.6.32-504.30.3.el6 (RHEL6.6)
62 2.6.32-573.26.1.el6 (RHEL6.7)
63 2.6.32-642.15.1.el6 (RHEL6.8)
64 2.6.32-696.18.7.el6 (RHEL6.9)
65 3.10.0-693.21.1.el7 (RHEL7.4)
66 3.0.101-0.47.71 (SLES11 SP3)
67 3.0.101-107 (SLES11 SP4)
68 3.12.74-60.64.40 (SLES12 SP1)
69 4.4.114-92.64 (SLES12 SP2)
70 4.4.114-94.11 (SLES12 SP3)
71 3.13.0-101 (Ubuntu 14.04)
72 4.4.0-85.108 (Ubuntu 14.04.5 LTS)
73 4.4.0-116 (Ubuntu 16.04)
75 * Recommended e2fsprogs version: 1.42.13.wc6 or newer
76 * Recommended ZFS / SPL version: 0.7.6
77 * Tested with ZFS / SPL version: 0.7.6
78 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
79 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
80 stack overflow. For more information, please refer to bugzilla 17630.
81 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
82 the re-exporting nodes to be mounted with "32bitapi" mount option
83 * Note for Ubuntu 14.04.5:
84 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
85 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
87 --------------------------------------------------------------------------------
89 2017-07-05 Intel Corporation
91 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
92 for currently supported client and server kernel versions.
93 * Server known to build on patched kernels:
94 2.6.32-431.29.2.el6 (RHEL6.5)
95 2.6.32-504.30.3.el6 (RHEL6.6)
96 2.6.32-573.26.1.el6 (RHEL6.7)
97 2.6.32-642.15.1.el6 (RHEL6.8)
98 2.6.32-696.3.1.el6 (RHEL6.9)
99 3.10.0-514.21.1.el7 (RHEL7.3)
100 3.0.101-0.47.71 (SLES11 SP3)
101 3.0.101-100 (SLES11 SP4)
102 3.12.74-60.64.40 (SLES12 SP1)
103 4.4.59-92.17 (SLES12 SP2)
104 vanilla linux 4.6.7 (ZFS only)
105 * Client known to build on unpatched kernels:
106 2.6.32-431.29.2.el6 (RHEL6.5)
107 2.6.32-504.30.3.el6 (RHEL6.6)
108 2.6.32-573.26.1.el6 (RHEL6.7)
109 2.6.32-642.15.1.el6 (RHEL6.8)
110 2.6.32-696.3.1.el6 (RHEL6.9)
111 3.10.0-514.21.1.el7 (RHEL7.3)
112 3.0.101-0.47.71 (SLES11 SP3)
113 3.0.101-100 (SLES11 SP4)
114 3.12.74-60.64.40 (SLES12 SP1)
115 4.4.59-92.17 (SLES12 SP2)
117 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
118 * Recommended ZFS / SPL version: 0.6.5.9
119 * Tested with ZFS / SPL version: 0.6.5.9
120 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
121 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
122 stack overflow. For more information, please refer to bugzilla 17630.
123 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
124 the re-exporting nodes to be mounted with "32bitapi" mount option
126 --------------------------------------------------------------------------------
128 12-07-2016 Intel Corporation
130 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
131 for currently supported client and server kernel versions.
132 * Server known to build on patched kernels:
133 2.6.32-431.29.2.el6 (RHEL6.5)
134 2.6.32-504.30.3.el6 (RHEL6.6)
135 2.6.32-573.26.1.el6 (RHEL6.7)
136 2.6.32-642.11.1.el6 (RHEL6.8)
137 3.10.0-514.el7 (RHEL7.3)
138 3.0.101-0.47.71 (SLES11 SP3)
139 3.0.101-80 (SLES11 SP4)
140 3.12.62-60.64.8 (SLES12 SP1)
141 vanilla linux 4.5.7 (ZFS only)
142 * Client known to build on unpatched kernels:
143 2.6.32-431.29.2.el6 (RHEL6.5)
144 2.6.32-504.30.3.el6 (RHEL6.6)
145 2.6.32-573.26.1.el6 (RHEL6.7)
146 2.6.32-642.11.1.el6 (RHEL6.8)
147 3.10.0-514.el7 (RHEL7.3)
148 3.0.101-0.47.71 (SLES11 SP3)
149 3.0.101-80 (SLES11 SP4)
150 3.12.62-60.64.8 (SLES12 SP1)
152 * Recommended e2fsprogs version: 1.42.13.wc5 or newer
153 * Recommended ZFS / SPL version: 0.6.5.8
154 * Tested with ZFS / SPL version: 0.6.5.8
155 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
156 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
157 stack overflow. For more information, please refer to bugzilla 17630.
158 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
159 the re-exporting nodes to be mounted with "32bitapi" mount option
161 --------------------------------------------------------------------------------
163 02-29-2016 Intel Corporation
165 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
166 for currently supported client and server kernel versions.
167 * Server known to build on patched kernels:
168 2.6.32-431.29.2.el6 (RHEL6.5)
169 2.6.32-504.30.3.el6 (RHEL6.6)
170 2.6.32-573.12.1.el6 (RHEL6.7)
171 3.10.0-327.3.1.el7 (RHEL7.2)
172 3.0.101-0.47.71 (SLES11 SP3)
173 3.0.101-68 (SLES11 SP4)
174 vanilla linux 4.2.1 (ZFS only)
175 * Client known to build on unpatched kernels:
176 2.6.32-431.29.2.el6 (RHEL6.5)
177 2.6.32-504.30.3.el6 (RHEL6.6)
178 2.6.32-573.12.1.el6 (RHEL6.7)
179 3.10.0-327.3.1.el7 (RHEL7.2)
180 3.0.101-0.47.71 (SLES11 SP3)
181 3.0.101-68 (SLES11 SP4)
184 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
185 * Recommended ZFS / SPL version: 0.6.4.2
186 * Tested with ZFS / SPL version: 0.6.4.2
187 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
188 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
189 stack overflow. For more information, please refer to bugzilla 17630.
190 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
191 the re-exporting nodes to be mounted with "32bitapi" mount option
193 --------------------------------------------------------------------------------
195 03-10-2015 Intel Corporation
197 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
198 for currently supported client and server kernel versions.
199 * Server known to build on patched kernels:
200 2.6.32-431.29.2.el6 (RHEL6.5)
201 2.6.32-504.8.1.el6 (RHEL6.6)
202 3.0.101-0.46 (SLES11 SP3)
203 * Client known to build on unpatched kernels:
204 2.6.32-431.29.2.el6 (RHEL6.5)
205 2.6.32-504.8.1.el6 (RHEL6.6)
206 3.10.0-123.20.1.el7 (RHEL7)
207 3.0.101-0.46 (SLES11 SP3)
208 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
209 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
210 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
211 stack overflow. For more information, please refer to bugzilla 17630.
212 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
213 the re-exporting nodes to be mounted with "32bitapi" mount option
215 Severity : enhancement
217 Description: control OST-index in IDIF via ROCOMPAT flag.
218 Details : Introduce new flag OBD_ROCOMPAT_IDX_IN_IDIF that is stored in the
219 last_rcvd file. For new formatted OST device, it will be auto set;
220 for the case of upgrading from old OST device, you can enable it
221 via the lproc interface osd-ldiskfs.index_in_idif. With such flag
222 enabled, for new created OST-object, its IDIF-in-LMA will contain
223 the OST-index; for the existing OST-object, the OSD will convert
224 old format IDIF as new format IDIF with OST-index stored in the
225 LMA EA when accessing such OST-object or via OI scrub. Once such
226 flag is enabled, it cannot be reverted back, so the system cannot
227 be downgraded to the orignal incompatible version.
229 --------------------------------------------------------------------------------
231 07-30-2014 Intel Corporation
233 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
234 for currently supported client and server kernel versions.
235 * Server known to build on patched kernels:
236 2.6.32-431.20.3.el6 (RHEL6.5)
237 3.0.101-0.31 (SLES11 SP3)
238 * Client known to build on unpatched kernels:
239 2.6.32-431.20.3.el6 (RHEL6.5)
240 3.0.101-0.31 (SLES11 SP3)
243 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
244 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
245 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
246 stack overflow. For more information, please refer to bugzilla 17630.
247 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
248 the re-exporting nodes to be mounted with "32bitapi" mount option
250 --------------------------------------------------------------------------------
252 10-31-2013 Intel Corporation
254 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
255 * Server builds for kernels:
256 2.6.32-358.18.1.el6 (RHEL6.4)
257 * Client builds for unpatched kernels:
258 2.6.32-358.18.1.el6 (RHEL6.4)
259 3.0.92-0.8 (SLES11 SP3)
261 * Recommended e2fsprogs version: 1.42.7.wc2 or newer
262 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
263 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
264 stack overflow. For more information, please refer to bugzilla 17630.
265 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
266 the re-exporting nodes to be mounted with "32bitapi" mount option
268 --------------------------------------------------------------------------------
270 05-30-2013 Intel Corporation
272 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
273 * Server builds for kernels:
274 2.6.32-358.6.2.el6 (RHEL6.4)
275 * Client builds for unpatched kernels:
276 2.6.32-358.6.2.el6 (RHEL6.4)
277 3.0.74-0.6.6 (SLES11 SP2)
279 * Recommended e2fsprogs version: 1.42.7.wc1 or newer
280 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
281 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
282 stack overflow. For more information, please refer to bugzilla 17630.
283 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
284 the re-exporting nodes to be mounted with "32bitapi" mount option
286 --------------------------------------------------------------------------------
288 9-30-2012 Whamcloud, Inc.
290 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
291 * Server support for kernels:
292 2.6.32-279.5.1.el6 (RHEL6)
293 * Client support for unpatched kernels:
294 2.6.32-279.5.1.el6 (RHEL6)
296 2.6.18-238.19.1.el5 (RHEL5)
297 2.6.16.60-0.69.1 (SLES10)
298 * Recommended e2fsprogs version: 1.41.90.wc4
299 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
300 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
301 stack overflow. For more information, please refer to bugzilla 17630.
302 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
303 the re-exporting nodes to be mounted with "32bitapi" mount option
305 Severity : enhancement
307 Description: Update RHEL6.3 kernel to 2.6.32-279.5.1.el6.
309 Severity : enhancement
311 Description: Update RHEL6.3 kernel to 2.6.32-279.2.1.el6.
313 Severity : enhancement
315 Description: Update RHEL6.2 kernel to 2.6.32-220.17.1.el6
317 Severity : enhancement
319 Description: Update RHEL6.2 kernel to 2.6.32-220.13.1.el6
321 Severity : enhancement
323 Description: Update RHEL6.2 kernel to 2.6.32-220.7.1.el6
325 --------------------------------------------------------------------------------
327 04-30-2012 Whamcloud, Inc.
329 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
330 * Server support for kernels:
331 2.6.32-220.4.2.el6 (RHEL6)
332 * Client support for unpatched kernels:
333 2.6.32-220.4.2.el6 (RHEL6)
335 2.6.18-238.19.1.el5 (RHEL5)
336 2.6.16.60-0.69.1 (SLES10)
337 * Recommended e2fsprogs version: 1.41.90.wc4
338 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
339 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
340 stack overflow. For more information, please refer to bugzilla 17630.
341 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
342 the re-exporting nodes to be mounted with "32bitapi" mount option
344 Severity : enhancement
346 Description: Update RHEL6.2 kernel to 2.6.32-220.4.2.el6
348 Severity : enhancement
350 Description: Update RHEL6.2 kernel to 2.6.32-220.4.1.el6
352 Severity : enhancement
354 Description: Update RHEL6.2 kernel to 2.6.32-220.el6
356 Severity : enhancement
358 Description: Update RHEL6.1 kernel to 2.6.32-131.17.1.el6
360 --------------------------------------------------------------------------------
362 09-30-2011 Whamcloud, Inc.
364 * Server support for kernels:
365 2.6.18-238.19.1.el5 (RHEL5)
366 2.6.32-131.6.1.el6 (RHEL6)
367 * Client support for unpatched kernels:
368 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
369 2.6.18-238.19.1.el5 (RHEL5)
370 2.6.32-131.6.1.el6 (RHEL6)
371 2.6.29.4-167.fc11 (FC11)
372 2.6.16.60-0.69.1 (SLES10)
374 * Recommended e2fsprogs version: 1.41.90.wc3
375 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
376 Lustre filesystem with 4K stack may cause a stack overflow. For more
377 information, please refer to bugzilla 17630.
378 * NFSv4 reexport requires Lustre client on the re-exporting nodes to be
379 mounted with "32bitapi" mount option
381 Severity : enhancement
382 Description: Update RHEL6.1 kernel to 2.6.32-131.6.1.el6
384 Severity : enhancement
385 Description: Update RHEL5.6 kernel to 2.6.18-238.19.1.el5
387 Severity : enhancement
388 Description: Update RHEL5.6 kernel to 2.6.18-238.9.1.el5
391 Frequency : only with SLES10
393 Description: Use OFED "KMP" provided by Novell
394 Details : SLES10 SP3 ships with OFED in a separate "KMP" package.
395 Lustre is now built against this package. That means you need to
396 install the ofed-kmp package from Novell for the patchless client.
397 Note that the ofed-kmp that Novell ships may not exactly match the
398 kernel version but should still be compatible.
400 Severity : enhancement
402 Description: Update RHEL5.5 kernel to 2.6.18-194.17.1.el5
403 Update OEL5.5 kernel to 2.6.18-194.17.1.0.1.el5.
405 --------------------------------------------------------------------------------
407 2010-07-15 Oracle, Inc.
409 * Server support for kernels:
410 2.6.18-164.11.1.el5 (RHEL 5)
411 2.6.18-164.11.1.0.1.el5 (OEL 5)
412 * Client support for unpatched kernels:
413 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
414 2.6.18-164.11.1.el5 (RHEL 5),
415 2.6.18-164.11.1.0.1.el5 (OEL 5)
416 2.6.16.60-0.42.8 (SLES 10),
418 2.6.29.4-167.fc11 (FC11)
419 * Recommended e2fsprogs version: 1.41.10-sun2
421 Severity : enhancement
423 Description: Distribute lfs_migrate script with lustre
424 Details : The lfs_migrate mentioned in the manual is now included in the
425 lustre tree in lustre/scripts and will distributed in lustre RPMs.
427 Severity : enhancement
429 Description: Failover nid list added to import proc file.
431 Severity : enhancement
433 Description: "som_preview" mount option required to use SOM
434 Details : The clients will need to use som_preview mount option
438 Description: Lustre does not do 1MB IOs to HW RAID
439 Details : Bump MAX_PHYS/HW_SEGMENTS and SG_ALL to 256 in the RHEL5 kernel.
441 Severity : enhancement
443 Description: lfs quota output is not very convenient for awk/sed-parsing
444 Details : Some positions in lfs quota output table could be empty or
445 non-empty which made it hard to parse it with scripts, now
446 a dash is put instead of space where there is not supposed
451 Description: Update to OFED 1.5.1
452 Details : OFED 1.5.1 will be used in environments that don't use
453 vendor supplied OFED.
457 Description: Remove lnet. prefix from lctl params display
458 Details : Also renamed lnet.memused to lnet_memused to
463 Description: Use conf_param -d to remove permanant settings.
467 Description: Add list_param -R
468 Details : lctl list_param now supports recursive mode.
471 Bugzilla : 21829, 22312
472 Description: Fix llobdstat.
473 Details : 1. Now updating the write_bytes counter
474 2. Port script fixes from 1.8.
478 Description: decrease the usage of memory on clients.
479 Details : 1. On clients, recycle dentries and inodes unused.
480 2. Delete the code related to ll_deathrow(att 6215 in bug 1443). It
485 Description: llobdstat fix and enhancement
486 Details : add a counter to set a limit to how many samples will be returned
487 fix a wildcard in the path to limit to obdfilter stats only
491 Description: lfs check only allowed for root
492 Details : Code cleanups and added a regresion test of non-root use of
497 Description: MDS Oopses in ptlrpc_dispatch_difficult_reply()/__list_add()
498 Details : Fix hr->hr_index overload.
502 Description: Fix FIEMAP functionality.
506 Description: FC11 patchless client support added
510 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and
511 OEL5.4 kernel to 2.6.18-164.11.1.0.1.el5.
515 Description: New acc-sm script and various test-framework enhancments.
519 Description: Previously landed patch was reverted.
523 Description: Update SLES10 kernel to 2.6.16.60-0.42.8 (patchless)
527 Description: Add ldiskfs maxdirsize mount option.
531 Description: Fix mkfs.lustre support for 16TB LUNs.
535 Description: Print mmp_check_interval
536 Details : print mmp_check_interval and make it possible to abort
537 mount operation in case it takes too long.
541 Description: Add cascading_rw to lustre mpi tests.
545 Description: Store SOM data in EA.
546 Details : Also several SOM related cleanups.
550 Description: Add refcount to md_open_data
554 Description: Use refcount on lov_request_set to prevent memory corruption.
558 Description: Statahead: use parent dir's i_mutex for synchronization.
562 Description: loadgen improvements.
566 Description: Enable peer health detection in ksocklnd and ko2iblnd.
570 Description: Control DCACHE_LUSTRE_INVALID flag with MDS_INODE_LOOKCUP lock
574 Description: Fix race on obd_devcie:md_stats
578 Description: Fix SLES 11 client builds.
579 Details : Resolved uninitalized pointer gcc warnings.
583 Description: Allow non-root access for "lfs check".
584 Details : Added a check in obd_class_ioctl() for OBD_IOC_PING_TARGET.
588 Description: Clear imp_force_reconnect correctly in ptlrpc_connect_interpret()
590 Severity : enhancement
592 Description: Update kernel to RHEL5.4 2.6.18-164.6.1.el5 and
593 OEL5 2.6.18-164.6.1.0.1.el5(Both in-kernel OFED enabled).
597 Description: Sanity test_27p fails.
598 Details : OBD_FAIL_OST_ENOSPC has been moved to filter_statfs().
599 Sanity tests 27 were cleaned up and improved.
603 Description: Unify req format on client/servers
604 Details : Use new req_capsule API [almost] everywhere instead of old PTLRPC
605 buffers and swabbers approach..
610 Description: drop join file stuff.
612 Severity : enhancement
614 Description: Add 2.6.27 and SLES11 (patchless client) support
618 Description: Fix in ptlrpc_expire_one_request() to print the signed time
621 Severity : enhancement
623 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
626 Severity : enhancement
628 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
630 Severity : enhancement
632 Description: correctly handle big reply message.
633 Details : send LNet event if reply is bigger than buffer and adjust this buffer
638 Description: Drop unnecessary __GFP_NOMEMALLOC flag from filter_get_page()
640 Severity : enhancement
642 Description: Add support for OEL5.
646 Description: Conf-sanity.sh 50g test - deactivated OST should not cause a panic.
648 Severity : enhancement
650 Description: "weak-modules" support
651 Details : Implement "weak-modules" support which enables kernel modules
652 to be used with any kernel that implements the same kABI. In
653 order to achieve this modules are now installed in
654 /lib/modules/$(uname -r)/updates/kernel on all distributions.
658 Description: client could not reconnect to OST because of an active request.
659 Details : The client now retries to reconnect to the same server, if a
660 connect request failed with EBUSY or -EAGAIN. Test case 35b
661 conf-sanity.sh has been created.
665 Description: Don't allow make backward step on assignin osc next id.
666 Details : race between allocation next id and ll_sync thread can be cause
667 of set wrong osc next id and can be kill valid ost objects.
671 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
675 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
679 Description: Error handling in osc_statfs_interpret() has been improved.
680 Details : Check in osc_statfs_interpret() for EBADR.
684 Description: Do not send statfs() requests to OSTs disabled by administrator.
685 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
687 Severity : enhancement
689 Description: Added sanity-quota test to for lfs(1) to work with numeric uid/gid.
691 Severity : enhancement
693 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
695 Severity : enhancement
697 Description: Enhancement for lfs(1) command to use numeric uid/gid.
701 Description: Improved cleanup procedure for conf-sanity test_46a.
705 Description: Do not start bulk transfer, if client tried to reconnect.
706 Details : Do not start bulk transfer, if client tried to reconnect,
707 but failed due to active rpcs. Check for exp_abort_active_req
708 has been implemented.
712 Description: Prevent inconsistences between linux and lustre mount structures.
713 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
714 Make the sleep interruptible.
716 Severity : enhancement
718 Description: provide server to client comms path
719 Details : server to client communications path via new LDLM_SET_INFO rpc
723 Description: truncate starts GFP_FS allocation under transaction causing deadlock
724 Details : ldiskfs_truncate calls grab_cache_page which may start page
725 allocation under an open transaction. This may lead to
726 calling prune_icache with consequent lustre reentrance.
730 Description: send correctly lsm on open replay
731 Details : MDS is trust to LSM size on replay open, but client can set wrong size
736 Description: enable client interface failover
737 Details : When a child reconnects from another NID, properly update export
738 nid hash position and ldlm reverse import.
740 Severity : enhancement
742 Description: Communicate OST degraded/readonly state via statfs to MDS
743 Details : Flags in the statfs returned from OSTs indicate whether the
744 OST is in a degraded RAID state, or if the filesystem has
745 turned read-only after a filesystem error is detected.
750 Description: don't panic if EPROTO was hit when reading symlink
751 Details : correctly handling request reference in error cases.
753 Severity : enhancement
755 Description: Add LustreNetLink, a kernel-userspace communcation path.
757 Severity : enhancement
759 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
764 Description: access to llog context before init.
765 Details : move handling CATALOGS file at osc layer and forbid access to llog
769 Frequency : rare, only when client is writing to OST on local node
771 Description: Avoid deadlock for local client writes
772 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
773 memory freeing context. This allows OST threads to set the
774 PF_MEMALLOC flag on task structures in order to allocate memory
775 from reserved pools and complete IO.
776 Use GFP_HIGHUSER for OST allocations for non-local client writes,
777 so that the OST threads generate memory pressure and allow
778 inactive pages to be reclaimed.
780 Severity : enhancement
782 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
787 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
788 Details : move ldlm namespace creation in setup phase to avoid grab
789 _lprocfs_lock with cli_sem held.
793 Description: Temporarily disable grant shrink.
794 Details : Disable the feature for debugging.
797 Frequency : only when formatting test filesystems
799 Description: Unable to run several mkfs.lustre on loop devices at the same time
800 Details : mkfs.lustre returns error 256 on the concurrent loop devices
801 formatting. The solution is to proper handle the error.
803 Severity : enhancement
805 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
807 Severity : enhancement
809 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
814 Description: resolve race between obd_disconnect and class_disconnect_exports
815 Details : if obd_disconnect will be called to already disconnected export he
816 forget release one reference and osc module can't unloaded.
818 Severity : enhancement
820 Description: Allow tuning service thread via /proc
821 Details : For each service a new
822 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
823 created that can be used to set min/max thread counts, and get the
824 current number of running threads.
829 Descriptoin: don't return error if have partially created objects for file.
830 Details : lov_update_create_set uses set->set_success as index for created
831 objects, so if some requests will be failed, they will have hole
832 at end of array and qos_shrink_lsm for allocate correct lsm.
834 Severity : enhancement
836 Description: Update OFED support to 1.4.1
841 Description: Reduce small size read RPC
842 Details : Set read-ahead limit for every file and only do read-ahead when
843 available read-ahead pages are bigger than 1M to avoid small size
846 Severity : enhancement
848 Description: move AT tunable parameters for more consistent usage
849 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
851 Severity : enhancement
853 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
854 Details : allow skip disconnected ost for send statfs request and hide error
860 Description: fix racy locking of mballoc block bitmaps causing BUG
861 Details : The locking of the mballoc buddy bitmap and the in-memory
862 block bitmap was using two different spin locks in some
863 cases. This made it possible to incorrectly access the
864 mballoc bitmap while another process was modifying it,
865 causing a sanity assertion to fail. While no on-disk corruption
866 was reported, there was some risk of this happening.
869 Frequency : rare, on llog test 6
871 Descriptoin: don't allow connect to already connected import
872 Details : allowing connect to already connected import is hide connecting problem.
877 Descriptoin: allow kill process which wait statahead result
878 Details : for some reasons 'ls' can stick in waiting result from statahead,
879 in this case need way for kill this process.
881 Severity : enhancement
883 Description: Add state history info file, enhance import info file
884 Details : Track import connection state changes in a new osc/mdc proc file;
885 add overview-type data to the osc/mdc import proc file.
887 Severity : enhancement
889 Description: MDS create should not wait for statfs RPC while holding DLM lock.
891 Severity : enhancement
893 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
898 Descriptoin: OST grant too much space to client even there are not enough space.
899 Details : Client will shrink its grant cache to OST if there are no write
900 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
901 this grant cache if there are already not enough avaible space
902 (left_space < total_clients * 32M).
905 Frequency : start MDS on uncleanly shutdowned MDS device
907 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
908 Details : stay in waiting mds<>ost recovery finished produce random bugs
909 due race between two ll_sync thread for one lov target. send
910 ACTIVATE event only if connect realy finished and import have
914 Frequency : rare, connect and disconnect target at same time
916 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
917 Details : don't call obd_disconnect under lov_lock. this long time
918 operation and can block ptlrpcd which answer to connect request.
923 Descriptoin: don't lose wakeup for imp_recovery_waitq
924 Details : recover_import_no_retry or invalidate_import and import_close can
925 both sleep on imp_recovery_waitq, but we was send only one wakeup
929 Frequency : always with long access acl
931 Descriptoin: mds can't pack reply with long acl.
932 Details : mds don't control size of acl but they limited by reint/getattr
935 Severity : enhancement
937 Description: Update to SLES10 kernel-2.6.16.60-0.33.
939 Severity : enhancement
941 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
944 Frequency : start MDS on uncleanly shutdowned MDS device
946 Descriptoin: aborting recovery hang on MDS
947 Details : don't throttle destroy RPCs for the MDT.
950 Frequency : on remount
952 Description: external journal device not working after the remount
953 Details : clear dev_rdonly flag for external journal devices in
959 Description: shutdown vs evict race
960 Details : client_disconnect_export vs connect request race.
961 if client will evicted at this time - we start invalidate
962 thread without referece to import and import can be freed
966 Frequency : rare, need acl's on inode.
968 Description: client can't handle ost additional correctly
969 Details : if ost was added after client connected to mds client can have
970 hit lnet_try_match_md ... to big messages to wide striped files.
971 in this case need teach client to handle config events about add
972 lov target and update client max ea size at that event.
974 Severity : enhancement
976 Description: Changelogs
977 Details : Changelogs are a lightweight mechanism to track filesystem
978 metadata and namespace changes. The changelog is recorded
979 permanently on the MDTs, and is periodically "consumed" / purged
980 when records are no longer needed.
982 Severity : enhancement
984 Description: compact fld format with extents
985 Details : Store range of seq rather than every seq in FLD. Seq
986 controller update FLD rather than clients. In Case of CMD, mdt0
987 has FLD, all other metadata server act as non persistent proxy
988 for FLD queries and cache fld entries in fld cache.
993 Description: don't skip ost target if they assigned to file
994 Details : Drop slow OSCs if we can, but not for requested start idx.
995 This means "if OSC is slow and it is not the requested
996 start OST, then it can be skipped, otherwise skip it only
997 if it is inactive/recovering/out-of-space.
1001 Description: more cleanup in mds_lov
1002 Details : not send LOV EA under replay, we can't know about they size at this
1003 time. Don't allow client connect to mds before any ost connected,
1004 for avoid problems with LOV EA size and returning EIO to client.
1006 Severity : enhancement
1008 Description: Interoperability at server side (Disk interoperability)
1010 Severity : enhancement
1012 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
1014 Severity : enhancement
1016 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
1018 Severity : enhancement
1020 Description: New client IO stack (CLIO).
1022 Severity : enhancement
1024 Description: Commit on sharing. Eliminate inter-client dependencies between
1025 uncommitted transactions by doing transaction commits.
1026 Thereby clients may recovery independently.
1029 Frequency : Create a symlink file with a very long name
1031 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
1032 Details : If there is no extra space in the request for early cancels,
1033 ldlm_req_handles_avail() returns 0 instead of a negative value.
1035 Severity : enhancement
1037 Description: Add /proc entry for import status
1038 Details : The mdc, osc, and mgc import directories now have
1039 an import directory that contains useful import data for debugging
1040 connection problems.
1042 Severity : enhancement
1044 Description: Re-disable certain /proc logging
1045 Details : Enable and disable client's offset_stats, extents_stats and
1046 extents_stats_per_process stats logging on the fly.
1049 Frequency : Only on FC kernels 2.6.22+
1051 Description: oops in statahead
1052 Details : Do not drop reference count for the dentry from VFS when lookup,
1053 VFS will do that by itself.
1055 Severity : enhancement
1057 Description: Generic /proc file permissions
1058 Details : Set /Proc file permissions in a more generic way to enable non-
1059 root users operate on some /proc files.
1063 Description: Hitting mdc_commit_close() ASSERTION
1064 Details : Properly handle request reference release in
1065 ll_release_openhandle().
1069 Description: quota recovery deadlock during mds failover
1070 Details : This patch includes att18982, att18236, att18237 in bz14840.
1072 1. fix osts hang when mds does failover with quotaon
1073 2. prevent watchdog storm when osts threads wait for the
1078 Frequency : only patchless client
1079 Description: add workaround for race between add/remove dentry from hash
1081 Severity : enhancement
1083 Description: Allow OST glimpses to return PW locks
1087 Description: LBUG when llog conf file is full
1088 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
1092 Description: Prevent import from entering FULL state when server in recovery
1096 Description: service mount cannot take device name with ":"
1097 Details : Only when device name contains ":/" will mount treat it as
1103 Description: replace ptlrpcd with the statahead thread to interpret the async
1104 statahead RPC callback
1108 Frequency : on recovery
1109 Description: I/O failures after umount during fail back
1110 Details : if client reconnected to restarted server we need join to recovery
1111 instead of find server handler is changed and process self
1112 eviction with cancel all locks.
1114 Severity : enhancement
1116 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
1118 Severity : enhancement
1120 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
1122 Severity : enhancement
1124 Description: Upcall on Lustre log has been dumped
1125 Details : Allow for a user mode script to be called once a Lustre log has
1126 been dumped. It passes the filename of the dumped log to the
1127 script, the location of the script can be specified via
1128 /proc/sys/lnet/debug_log_upcall.
1133 Description: avoid idr_remove called for id which is not allocated.
1134 Details : Move assigment s_dev for clustered nfs to end of initialization,
1135 to avoid problem with error handling.
1140 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
1141 Details : When connection is reused this not moved from CONN_UNUSED_HASH
1142 into CONN_USED_HASH and this prodice warning when put connection
1143 again in unused hash.
1146 Severity : enhancement
1148 Description: File striping can now be set to use an arbitrary pool of OSTs.
1150 Severity : enhancement
1152 Description: Export bytes_read/bytes_write count on OSC/OST.
1156 Description: Early reply size mismatch, MGC loses connection
1157 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
1158 the connect flags are properly negotiated.
1163 Description: quotas are not honored with O_DIRECT
1164 Details : all writes with the flag O_DIRECT will use grants which leads to
1165 this problem. Now using OBD_BRW_SYNC to guard this.
1169 Description: add quota statistics
1170 Details : 1. sort out quota proc entries and proc code.
1171 2. add quota statistics
1173 Severity : enhancement
1175 Description: enable quota support for HEAD.
1179 Description: Properly propagate oinfo flags from lov to osc for statfs
1180 Details : restore missing copy oi_flags to lov requests.
1182 Severity : enhancement
1184 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
1185 : l_getgroups(8), lst(8), routerstat(8)
1186 Details : included man pages for llobdstat(8), llstat(8),
1187 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
1189 Severity : enhancement
1191 Description: configure's --enable-quota should check the
1192 : kernel .config for CONFIG_QUOTA
1193 Details : configure is terminated if --enable-quota is passed but
1194 : no quota support is in kernel
1198 Description: Remove portals compatibility
1199 Details : Remove portals compatibility, not interoperable with releases
1204 Description: Resolve device initialization race
1205 Details : Prevent proc handler from accessing devices added to the
1206 obd_devs array but yet be intialized.
1208 Severity : enhancement
1210 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
1212 Severity : enhancement
1214 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
1219 Description: Using wrong pointer in osc_brw_prep_request
1220 Details : Access to array[-1] can produce panic if kernel compiled with
1221 CONFIG_PAGE_ALLOC enabled
1225 Description: Client runs out of low memory
1226 Details : Consider only lowmem when counting initial number of llap pages
1230 Description: *optional* service tags registration
1231 Details : if the "service tags" package is installed on a Lustre node
1232 When the filesystem is mounted, a local-node service tag will
1233 be created. See http://inventory.sun.com/ for more information
1234 about the Service Tags asset management system.
1238 Description: Kernel BUG tries to release flock
1239 Details : Lustre does not destroy flock lock before last reference goes
1240 away. So always drop flock locks when client is evicted and
1241 perform unlock regardless of successfulness of speaking to MDS.
1245 Description: add refcount protection for osc callbacks, avoid panic on shutdown
1249 Description: sanity test 65a fails if stripecount of -1 is set
1250 Details : handle -1 striping on filesystem in ll_dirstripe_verify
1255 Description: ASSERTION(CheckWriteback(page,cmd)) failed
1256 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
1259 Severity : enhancement
1261 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
1265 Description: do not process already freed flock
1266 Details : flock can possibly be freed by another thread before it reaches
1267 to ldlm_flock_completion_ast.
1271 Description: LBUG during stress test
1272 Details : Need properly lock accesses the flock deadlock detection list.
1276 Description: oops in page fault handler
1277 Details : kernel page fault handler can return two special 'pages' in error
1278 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
1282 Description: timeout with invalidate import.
1283 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
1284 should be handled by ptlrpcd. This produce long age waiting and
1285 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
1287 Severity : enhancement
1289 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
1293 Description: enable MGS and MDT services start separately
1294 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
1295 a co-located MGS without starting the MGS, which is a complement
1296 to 'nosvc' mount option.
1300 Frequency : after recovery
1301 Description: precreate to many object's after del orphan.
1302 Details : del orphan st in oscc last_id == next_id and this triger growing
1303 count of precreated objects. Set flag LOW to skip increase count
1304 of precreated objects.
1308 Frequency : rare, on clear nid stats
1309 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
1310 Details : when clean nid stats sometimes try destroy live entry,
1311 and this produce panic in free.
1315 Description: Stack overflow during MDS log replay
1316 ease stack pressure by using a thread dealing llog_process.
1320 Description: wait until IO finished before start new when do lock cancel.
1321 Details : VM protocol want old IO finished before start new, in this case
1322 need wait until PG_writeback is cleared until check dirty flag and
1323 call writepages in lock cancel callback.
1325 Severity : enhancement
1327 Description: using special macro for print time and cleanup in includes.
1331 Description: mds_mfd_close() ASSERTION(rc == 0)
1332 Details : In mds_mfd_close(), we need protect inode's writecount change
1333 within its orphan write semaphore to prevent possible races.
1337 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
1341 Frequency : rare, on shutdown ost
1342 Description: don't hit live lock with umount ost.
1343 Details : shrink_dcache_parent can be in long loop with destroy dentries,
1344 use shrink_dcache_sb instead.
1348 Description: don't panic with use echo client
1349 Details : echo client pass NULL as client nid pointer and this produce null
1350 pointer dereference.
1354 Description: fix build on ppc32
1355 Details : compile code with -m64 flag produce wrong object file for ppc32.
1359 Description: add message levels for liblustreapi
1363 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
1364 Details : Move the 'good_osts' check before the 'total_bavail' check. This
1365 will result in an -EAGAIN and in the exit call path we call
1366 alloc_rr() which will with increasing aggressiveness attempt to
1367 aquire precreated objects on the minimum number of required OSCs.
1371 Description: Use old size assignment to avoid deadlock
1372 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
1373 the scheduling while holding a spinlock. We do not need locking
1374 for size in ll_update_inode() because size is only updated from
1375 the MDS for directories or files without objects, so there is no
1376 other place to do the update, and concurrent access to such inodes
1377 are protected by the inode lock.
1381 Description: resolve "_IOWR redefined" build error on SLES10
1385 Description: dump the memory debugging after all modules are unloaded to
1386 suppress false negative in conf_sanity test 39
1388 Severity : enhancement
1390 Description: build kernel-ib packages for OFED 1.3 in our release cycle
1395 Description: fix SLES kernel versioning
1396 Details : the kernel version for our SLES 10 kernel did not include a "-"
1397 before the "smp" at the end. while this was not a problem in
1398 general, it did mean that software trying to use the kernel
1399 version to try to detect a vendor specific kernel would fail.
1400 this was most evident by the OFED build scripts.
1404 Description: Don't update lov_desc members until making sure they are valid
1405 Details : When updating lov_desc members via proc fs, need fix their
1406 validities before doing the real update.
1410 Description: don't put request into delay list while invalidate in flight.
1411 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
1412 import in flight. this produce timeout for invalidate and sometimes
1413 can cause stale data.
1417 Frequency : on ppc only
1418 Description: not convert ost objects for directory because it's not exist.
1419 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
1423 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
1425 Severity : enhancement
1427 Description: Update to support 2.6.22.14 vanilla kernel.
1431 Frequency : rare, on recovery
1432 Description: read procfs can produce deadlock in some situation
1433 Details : Holding lprocfs lock which send rpc can produce block for destroy
1434 obd objects and this also block reconnect with -EALREADY. This
1435 isn't fix all lprocfs bugs - but make it rare.
1437 Severity : enhancement
1439 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
1442 Frequency : frequent on X2 node
1444 Description: mdc_set_open_replay_data LBUG
1445 Details : Set replay data for requests that are eligible for replay.
1449 Description: lustre_mgs: operation 101 on unconnected MGS
1450 Details : When MGC is disconnected from MGS long enough, MGS will evict the
1451 MGC, and late on MGC cannot successfully connect to MGS and a lot
1452 of the error messages complaining that MGS is not connected.
1456 Frequency : on network error
1457 Description: panic with double free request if network error
1458 Details : mdc_finish_enqueue is finish request if any network error ocuring,
1459 but it's true only for synchronus enqueue, for async enqueue
1460 (via ptlrpcd) this incorrect and ptlrpcd want finish request
1463 Severity : enhancement
1465 Description: client-side metadata stat-ahead during readdir(directory readahead)
1466 Details : perform client-side metadata stat-ahead when the client detects
1467 readdir and sequential stat of dir entries therein
1470 Frequency : on start mds
1472 Description: Implement get_info(last_id) in obdfilter.
1475 Frequency : occasional
1477 Description: Correctly check stale fid, not start epoch if ost not support SOM
1478 Details : open with flag O_CREATE need set old fid in op_fid3 because
1479 op_fid2 was overwritten with new generated fid, but mds can answer
1480 with one of these two fids and both is not stale. Setattr
1481 incorrectly started an epoch and assume will be called
1482 done_writing, but without SOM done_writing ever being called.
1485 Frequency : rare, depends on device drivers and load
1487 Description: MDS or OSS nodes crash due to stack overflow
1488 Details : Code changes in 1.8.0 increased the stack usage of some functions.
1489 In some cases, in conjunction with device drivers that use a lot
1490 of stack the MDS (or possibly OSS) service threads could overflow
1491 the stack. One change which was identified to consume additional
1492 stack has been reworked to avoid the extra stack usage.
1495 Frequency : occasional
1497 Description: Do not fail import if osc_interpret_create gets -EAGAIN
1498 Details : If osc_interpret_create got -EAGAIN it immediately exits and
1499 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
1500 oscc_has_objects and see OSC has no objests and call
1501 oscc_internal_create to resend create request.
1503 Severity : enhancement
1505 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
1507 Severity : enhancement
1509 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
1512 Frequency : very rare
1514 Description: Fix replay if there is an un-replied request and open
1515 Details : In some cases, older replay request will revert the
1516 mcd->mcd_last_xid on MDS which is used to record the client's
1517 latest sent request.
1519 Severity : enhancement
1521 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
1523 Severity : enhancement
1525 Description: Add rhel5 support to HEAD.
1527 Serverity : enhancement
1529 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
1534 Description: Don't allow skipping OSTs if index has been specified.
1535 Details : Don't allow skipping OSTs if index has been specified, make
1536 locking in internal create lots better.
1540 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
1541 Details : access to bitfield in structure is always rounded to long
1542 and this produce problem with not atomic change any bit.
1546 Description: Lustre make rpms failed.
1547 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
1548 builds Lustre rpms from tarball.
1551 Frequency : rare on shutdown ost
1553 Description: If llog cancel was not send before clean_exports phase, this can
1554 produce deadlock in llog code.
1555 Details : If llog thread has last reference to obd and call class_import_put
1556 this produce deadlock because llog_cleanup_commit_master wait when
1557 last llog_commit_thread exited, but this never success because was
1558 called from llog_commit_thread.
1562 Description: allow userland application know is lost one of stripes.
1563 Details : fill lvb_blocks with error code on ost and return it to
1564 application if error flag found.
1568 Description: NULL lov_tgts causing MDS oops
1569 Details : more safe checks for NULL lov_tgts for avoid oops.
1571 Severity : enhancement
1573 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
1577 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
1578 Details : If an OST is down the MDS will hang indefinitely in
1579 obd_statfs_rqset() waiting for the statfs data. While for
1580 MDS QOS usage of statfs, it should not stuck in waiting.
1582 Severity : enhancement
1584 Description: remote_acl support
1585 Details : Support ACL-based permission check for remote user.
1586 Support setfacl/getfacl for remote user with the utils
1587 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
1588 the system "{s,g}etfacl" utils.
1590 Severity : enhancement
1592 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
1594 Severity : enhancement
1596 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
1600 Description: make mgs_setparam() handle fsname containing dash
1601 Details : fsname containing a dash does not work with lctl conf_param
1603 Severity : enhancement
1605 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
1607 Severity : enhancement
1609 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
1613 Frequency : rare, at shutdown
1614 Description: access already free / zero obd_namespace.
1615 Details : if client_disconnect_export was called without force flag set,
1616 and exist connect request in flight, this can produce access to
1617 NULL pointer (or already free pointer) when connect_interpret
1618 store ocd flags in obd_namespace.
1622 Frequency : only at startup
1623 Description: not alloc memory with spinlock held.
1624 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
1625 if any spinlock held.
1627 Severity : enhancement
1629 Description: make lustre randomly fail allocating memory
1630 Details : Make lustre randomly failed allocating memory for testing purpose.
1632 Severity : enhancement
1634 Description: lost problems with lov objid file
1635 Details : Fixes some scability and access to not inited memory problems
1636 in work with lov objdid file.
1641 Description: lfs find does not continue on file error
1642 Details : Continue other files processing when a file/dir is absent.
1646 Description: Inconsistent usage of lustre_pack_reply()
1647 Details : Standardize the usage of lustre_pack_reply() such that it
1648 always generate a CERROR on failure.
1651 Frequency : occasional
1653 Description: MDS hang or stay in waiting lock
1654 Details : If client receive lock with CBPENDING flag ldlm need send lock
1655 cancel as separate rpc, to avoid situation when cancel request
1656 can't processed due all i/o threads stay in wait lock.
1660 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1664 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1668 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1672 Description: Update to RHEL4 latest kernel.
1674 Severity : enhancement
1676 Description: Build SLES10 patchless client fails
1677 Details : The configure was broken by run ./configure with
1678 --with-linux-obj=.... argument for patchless client. When the
1679 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1680 can't search header adequately. Use absolute path such as
1681 -I($LINUX)/include instead.
1685 Description: interrupt oig_wait produce painc on resend.
1686 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1687 private set, and this produce situation when rq_ptlrpcd_data not
1688 copyed to new allocated request and triggered LBUG on assert
1689 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1690 ptlrpcd set if request is changed and can be safety changed to use
1695 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1696 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1697 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1698 these macros and replace them with their difinitions instead.
1702 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1703 Details : Update which_patch & target file for SLES10 latest kernel.
1706 Description: Remove client patches from SLES 10 kernel.
1707 Details : This causes SLES 10 clients to behave as patchless clients
1708 even on a Lustre-patched (server) kernel.
1710 Severity : enhancement
1712 Description: self-adjustable client's lru lists
1713 Details : use adaptive algorithm for managing client cached locks lru
1714 lists according to current server load, other client's work
1715 pattern, memory activities, etc. Both, server and client
1716 side namespaces provide number of proc tunables for controlling
1719 Severity : enhancement
1721 Description: light-weight GSS support
1722 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
1723 and reduce performance overhead.
1725 Severity : enhancement
1727 Description: Linux keyring support
1728 Details : Support using service of Linux keyring for Lustre GSS internal
1729 context refresh/cache mechanism.
1733 Description: Fix errors in lfs documentation
1734 Details : Fixes man pages
1738 Description: don't use GFP_* in generic Lustre code.
1739 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
1743 Description: obdclass is limited by single OBD_ALLOC(idarray)
1744 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
1748 Description: warnings with build patchless client with vanila 2.6.19 and up
1749 Details : change old ctl_table style and replace ctl_table/ctl_table_header
1750 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
1754 Frequency : Only for SLES
1755 Description: sanity_quota fail test_1
1756 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
1757 file, which makes TSTID[2] inunique.
1761 Frequency : Only for SLES9
1762 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
1763 Details : Update target/ChangeLog/which_patch .
1767 Frequency : Only for test_10 in sanity-quota.sh
1768 Description: a bug in quota test code
1769 Details : checking a test flag in a wrong place
1773 Frequency : Only for test_13 in sanity-quota.sh
1774 Description: a bug in quota test script
1775 Details : shouldn't check overfull allocated quota
1779 Description: jbd statistics
1780 Details : Port older jbd statistics patch for sles10
1782 Severity : enhancement
1784 Description: Add printing inode info into message about error in writepage.
1789 Description: reply_lock_interpret crash due to race with it and lock cancel.
1790 Details : Do not replay locks that are being cancelled. Do not reference
1791 locks by their address during replay, just by their handle.
1795 Frequency : When flocks are used.
1796 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
1797 Details : Flock locks might destroy just granted lock if it could be merged
1798 with another existing flock, this is done in completion handler,
1799 so teach ldlm_cli_enquque_fini that this is a valid case for
1805 Description: Oops in read and write path when failing to allocate lock.
1806 Details : Check if lock allocation failed and return error back.
1810 Description: lstripe command fails for valid OST index
1811 Details : The stripe offset is compared to lov->desc.ld_tgt_count
1812 instead of lov->desc.ld_active_tgt_count.
1816 Description: sanity.sh failed test 103
1817 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
1822 Description: df doesn't work properly if diskfs blocksize != 4K
1823 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
1827 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
1828 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
1830 Severity : enhancement
1832 Description: Adaptive timeouts
1833 Details : RPC timeouts adapt to changing server load and network
1834 conditions to reduce resend attempts and improve recovery time.
1838 Description: llapi_file_create() does not allow some changes
1839 Details : add llapi_file_open() that allows specifying the mode and
1840 open flags, and also returns an open file handle.
1844 Description: merge and cleanup kernel patches.
1845 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
1849 Description: Add journal checksum support.(Kernel part)
1850 Details : The journal checksum feature adds two new flags i.e
1851 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
1852 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
1853 indicates that the commit block contains the checksum for
1854 the blocks described by the descriptor blocks. Now commit
1855 record can be sent to disk without waiting for descriptor
1856 blocks to be written to disk. This behavior is controlled
1857 using JBD2_FEATURE_ASYNC_COMMIT flag.
1861 Description: OSS needs mutliple precreate threads
1862 Details : Add ability to start more than one create thread per OSS.
1866 Description: Sanity.sh test_65k failed: setstripe should have succeeded
1867 Details : execute lfs setstripe on client
1871 Description: mds_obd_create error creating tmp object
1872 Details : When the user sets quota on root, llog will be affected and can't
1873 create files and write files.
1876 Frequency : Always on ia64 patchless client, and possibly others.
1878 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
1879 Details : This allows the patchless client to be loaded on architectures
1880 without this export.
1884 Description: RedHat Update kernel for RHEL5
1885 Details : Add the kernel config file for RHEL5.
1889 Description: RedHat Update kernel for RHEL5
1890 Details : Modify the kernel config file more closer RHEL5.
1894 Description: Build failure against Centos5 (RHEL5)
1895 Details : Define PAGE_SIZE when it isn't present.
1899 Description: test_7 on sanity-quota.sh can't be used on separate machines
1900 Details : add facet to handle it
1904 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
1905 Details : with NFS, the anon dentry's parent was set to itself in
1906 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
1907 corresponding dentry other than use rec->ur_name.
1909 Severity : enhancement
1911 Description: lfs setstripe enhancement
1912 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
1916 Description: enable data checksumming by default
1917 Details : enable checksum by default, allow --disable-checksum
1918 configure option and "-o nochecksum" mount option. Checksums
1919 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
1923 Description: lustre support for RHEL5
1924 Details : Add support for RHEL5.
1928 Description: Client eviction due to failover config
1929 Details : after a connection loss, the lustre client should attempt to
1930 reconnect to the last active server first before trying the
1931 other potential connections.
1935 Description: when mds and osts use different quota unit(32bit and 64bit),
1936 quota will be released repeatly.
1937 Details : void sending multiple quota reqs to mds, which will keep the
1938 status between the reqs.
1942 Description: osts not allocated evenly to files
1943 Details : change the condition to increase offset_idx
1946 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
1948 Description: Data corruption for OSTs that are formatted larger than 2TB
1950 Details : When generating the bio request for lustre file writes the
1951 sector number would overflow a temporary variable before being
1952 used for the IO. The data reads correctly from Lustre (which
1953 will overflow in a similar manner) but other file data or
1954 filesystem metadata may be corrupted in some cases.
1958 Description: Tune the kernel for good SCSI performance.
1959 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
1960 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
1965 Description: rewrite ext2-derived code in obdclass/uuid.c
1966 Details : rewrite inherited code (uuid parsing code from ext2 utils)
1967 from scratch preserving functionality.
1971 Description: "lfs find -obd UUID" prints directories
1972 Details : "lfs find -obd UUID" will return all directory names instead
1973 of just file names. It is incorrect because the directories
1974 do not reside on the OSTs.
1978 Description: use i_size_read and i_size_write in 2.6 port
1979 Details : replace inode->i_size access with i_size_read/write()
1982 Frequency : when removing large files
1984 Description: scheduling issue during removal of large Lustre files
1985 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
1986 It causes scheduling issues when removing large files (17TB in the
1990 Frequency : only with liblustre clients on XT3
1992 Description: evictions taking too long
1993 Details : allow llrd to evict clients directly on OSTs
1996 Frequency : only on ppc
1998 Description: /proc/fs/lustre/devices broken on ppc
1999 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
2000 We should make sure the type of 'index' is loff_t and then cast
2001 explicitly as needed below. Do not assign an explicitly cast
2005 Frequency : only for rhel5
2007 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
2008 Details : Modify the target file & which_kernel.
2010 Severity : enhancement
2012 Description: omit set fsid for export NFS
2013 Details : fix set/restore device id for avoid EMFILE error and mark lustre
2014 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
2018 Frequency : Always, for kernels after 2.6.16
2019 Description: Fix warning idr_remove called for id=.. which is not allocated.
2020 Details : Last kernels save old s_dev before kill super and not allow
2021 to restore from callback - restore it before call kill_anon_super.
2025 Description: buffer overruns could theoretically occur
2026 Details : llapi_semantic_traverse() modifies the "path" argument by
2027 appending values to the end of the origin string, and a
2028 overrun may occur. Adding buffer overrun check in liblustreapi.
2032 Description: Fix error on 'ls .' at the top of the Lustre mount.
2033 Details : Don't revalidate dentry if it is a root dentry.
2037 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
2038 Details : Modify vm-tunables-rhel4.patch.
2042 Description: Kernel config for 2.6.18-vanilla.
2043 Details : Modify targets/2.6-vanilla.target.in.
2044 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
2045 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
2046 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
2047 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
2051 Description: improve handling recoverable errors
2052 Details : if request processig with error which can be recoverable on server
2053 request should be resend, otherwise page released from cache and
2058 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
2059 Details : Modify target file & which_patch.
2060 A flaw was found in the IA32 system call emulation provided
2061 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
2062 value could be stored in the %RAX register, which could trigger an
2063 out-of-bounds system call table access. An untrusted local user
2064 could exploit this flaw to run code in the kernel
2065 (ie a root privilege escalation). (CVE-2007-4573).
2069 Description: O_DIRECT bypasses client statistics.
2070 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
2071 read_ahead_stats were updated. copied the stats section from
2072 osc_send_oap_rpc() into async_internal().
2076 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
2080 Description: change order of libsysio includes
2081 Details : '#include sysio.h' should always come before '#include xtio.h'
2085 Description: handle "serious error: objid * already exists" more gracefully
2086 Details : If LAST_ID value on disk is smaller than the objects existing in
2087 the O/0/d* directories, it indicates disk corruption and causes an
2088 LBUG(). If the object is 0-length, then we should use the existing
2089 object. This will help to avoid a full fsck in most cases.
2091 Severity : enhancement
2093 Description: adapt the lustre_config script to support the upgrade case
2094 Details : Add "-u" option for lustre_config script to support upgrading 1.4
2095 server targets to 1.6 in parallel.
2099 Description: To avoid grant space > avaible space when the disk is almost
2100 full. Without this patch you might see the error "grant XXXX >
2101 available" or some LBUG about grant, when the disk is almost
2103 Details : In filter_check_grant, for non_grant cache write, we should
2104 check the left space by if (*left > ungranted + bytes), instead
2105 of (*left > ungranted), because only we are sure the left space
2106 is enough for another "bytes", then the ungrant space should be
2107 increase. In client, we should update cl_avail_grant only there
2108 is OBD_MD_FLGRANT in the reply.
2112 Description: Update RHEL 4 kernel to fix local root privilege escalation.
2113 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
2114 described in CVE-2007-4573. This problem could allow untrusted
2115 local users to gain root access.
2118 Frequency : when using O_DIRECT and quotas
2120 Description: Incorrect file ownership on O_DIRECT output files
2121 Details : block usage reported by 'lfs quota' does not take into account
2122 files that have been written with O_DIRECT.
2127 Description: touch file failed when fs is not full
2128 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
2129 otherwise we can get ENOSP while fs is not full.
2133 Description: parallel lock callbacks
2134 Details : Instead of sending blocking and completion callbacks as separated
2135 requests, adding them to a set and sending in parallel.
2138 Frequency : only for Cray XT3
2139 Bugzilla : 12829/13455
2140 Description: Changing primary group doesn't change the group lustre assigns to
2142 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
2143 overridden with the primary group provided by the group upcall,
2144 whereas the supplied fsgid can be trusted if it is in the list of
2145 supplementary groups returned by the group upcall.
2147 Severity : enhancement
2149 Description: Allow masking D_WARNING, D_ERROR messages from console
2150 Details : Console messages can now be disabled via lnet.printk.
2154 Description: User code with malformed file open parameter crashes client node
2155 Details : Before packing join_file req, all the related reference should be
2156 checked carefully in case some malformed flags cause fake
2157 join_file req on client.
2161 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
2162 Details : ldlm_completion_ast() assumes that a lock is granted when the req
2163 mode is equal to the granted mode. However, it should also check
2164 that LDLM_FL_CP_REQD is not set.
2168 Description: Heavy nfs access might result in deadlocks
2169 Details : After ELC code landed, it is now improper to enqueue any mds
2170 locks under och_sem, because enqueue might want to decide to
2171 cancel open locks for same inode we are holding och_sem for.
2175 Description: Client eviction while running blogbench
2176 Details : A lot of unlink operations with concurrent I/O can lead to a
2177 deadlock causing evictions. To address the problem, the number of
2178 oustanding OST_DESTROY requests is now throttled to
2179 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
2180 callbacks are processed in priority.
2184 Description: enable ACLs on MDS by default
2185 Details : ACLs must be enabled on MDS by default.
2188 Frequency : PPC/PPC64 only
2190 Description: conflicts between asm-ppc64/types.h and lustre_types.h
2191 Details : fix duplicated definitions between asm-ppc64/types.h and
2192 lustre_types.h on PPC.
2195 Frequency : PPC/PPC64 only
2197 Description: asm-ppc/segment.h does not exist
2198 Details : fix compile issue on PPC.
2202 Description: better handle error messages in extents code
2205 Frequency : RHEL4 only
2207 Description: mkfs is very slow on IA64/RHEL4
2208 Details : A performance regression has been discovered in the MPT Fusion
2209 driver between versions 3.02.73rh and 3.02.99.00rh. As a
2210 consequence, we have downgraded the MPT Fusion driver in the RHEL4
2211 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
2213 Severity : enhancement
2215 Description: SNMP support enhancement
2216 Details : Adding total number of sampled request for an MDS node in snmp
2219 Severity : enhancement
2221 Description: Optimize ldlm waiting list processing for PR extent locks
2222 Details : When processing waiting list for read extent lock and meeting read
2223 lock that is same or wider to it that is not contended, skip
2224 processing rest of the list and immediatelly return current
2225 status of conflictness, since we are guaranteed there are no
2226 conflicting locks in the rest of the list.
2230 Description: Time out and refuse to reconnect
2231 Details : When the failover node is the primary node, it is possible
2232 to have two identical connections in imp_conn_list. We must
2233 compare not conn's pointers but NIDs, otherwise we can defeat
2234 connection throttling.
2238 Description: port llog fixes from b1_6 into HEAD
2239 Details : Port llog reference couting and some llog cleanups from b1_6
2240 (bug 10800) into HEAD, for protect from panic and access to already
2241 free llog structures.
2245 Description: Detect stride IO mode in read-ahead
2246 Details : When a client does stride read, read-ahead should detect that and
2247 read-ahead pages according to the detected stride pattern.
2251 Description: data checksumming impacts single node performance
2252 Details : add support for several checksum algorithm. Currently, only CRC32
2253 and Adler-32 are supported. The checksum type can be changed on
2254 the fly via /proc/fs/lustre/osc/*/checksum_type.
2258 Description: use adler32 for page checksums
2259 Details : when available, use the Adler-32 algorithm instead of CRC32 for
2264 Description: build for x2 fails
2265 Details : fix compile issue on Cray systems.
2269 Description: Properly match for duplicate locks
2270 Details : Due to different lock order from skiplists code, we need to
2271 traverse entire list for now
2274 Frequency : only on PPC/SLES10
2276 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
2277 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
2278 be defined. Add a hack in mkfs_lustre.c to work around this compile
2283 Description: LASSERT on MDS when client holding flock lock dies
2284 Details : ldlm pool logic depends on number of granted locks equal to
2285 number of released locks which is not true for flock locks, so
2286 just exclude such locks from consideration.
2290 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
2291 Details : Use fsfilt_sync() for both the whole filesystem sync and
2292 individual file sync to eliminate dangerous inode locking
2293 with I_LOCK that can lead to a deadlock.
2297 Description: performance in 1.6.3
2298 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
2299 sure that 1MB requests are not fragmented by the block layer.
2301 Severity : enhancement
2303 Description: organize the server-side client stats on per-nid basis
2304 Details : Change the structure of stats under obdfilter and mds to
2312 The "uuid"s file would list the uuids of _active_ exports.
2313 And the clear entry is to clear all stats and stale nids.
2315 Severity : enhancement
2317 Description: eliminate client locks in face of contention
2318 Details : file contention detection and lockless i/o implementation
2319 for contended files.
2323 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
2326 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
2328 Description: MDS may lose file striping (and hence file data) in some cases
2329 Details : If there are additional extended attributes stored on the MDS,
2330 in particular ACLs, SELinux, or user attributes (if user_xattr
2331 is specified for the client mount options) then there is a risk
2332 of attribute loss. Additionally, the Lustre file striping
2333 needs to be larger than default (e.g. striped over all OSTs),
2334 and an additional attribute must be stored initially in the
2335 inode and then increase in size enough to be moved to the
2336 external attribute block (e.g. ACL growing in size) for file
2341 Description: skiplist implementation simplification
2342 Details : skiplists are used to group compatible locks on granted list
2343 that was implemented as tracking first and last lock of each
2344 lock group the patch changes that to using doubly linked lists
2348 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
2349 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
2350 because LASSERT on both the data supplied by a client, and the
2351 data on disk is dangerous and incorrect.
2353 Severity : enhancement
2355 Description: Slow truncate/writes to huge files at high offsets.
2356 Details : Directly associate cached pages to lock that protect those pages,
2357 this allows us to quickly find what pages to write and remove
2358 once lock callback is received.
2362 Description: more ldlm soft lockups
2363 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
2364 starve other threads from the resource lock for a long time in
2365 case of long waiting queue, so change the debug level from
2366 D_OTHER to the less frequently used D_INFO.
2368 Severity : enhancement
2370 Description: add -gid, -group, -uid, -user options to lfs find
2374 Description: Hung threads in invalidate_inode_pages2_range
2375 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
2376 once one is completed. As a result, some RPCs are stuck in the
2377 queue and are never sent.
2381 Description: filter threads hungs on waiting journal commit
2382 Details : Cleanup filter group llog code, then only filter group llog will
2383 be only created in the MDS/OST syncing process.
2387 Description: Procfs and llog threads access destoryed import sometimes.
2388 Details : Sync the import destoryed process with procfs and llog threads by
2389 the import refcount and semaphore.
2391 Severity : enhancement
2393 Description: openlock cache of b1_6 port to HEAD
2398 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
2399 Details : If insertion of an extent fails, then discard the inode
2400 preallocation and free data blocks else it can lead to duplicate
2405 Description: don't always update ctime in ext3_xattr_set_handle()
2406 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
2407 In some cases the ctime should not be updated, for example for
2408 2.0->1.8 compatibility it is necessary to delete an xattr and it
2409 should not update the ctime.
2413 Bugzilla : 15713/16362
2414 Description: Assertion in iopen_connect_dentry in 1.6.3
2415 Details : looking up an inode via iopen with the wrong generation number can
2416 populate the dcache with a disconneced dentry while the inode
2417 number is in the process of being reallocated. This causes an
2418 assertion failure in iopen since the inode's dentry list contains
2419 both a connected and disconnected dentry.
2423 Description: assertion failure in ldlm_handle2lock()
2424 Details : fix a race between class_handle_unhash() and class_handle2object()
2425 introduced in lustre 1.6.5 by bug 13622.
2430 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
2431 Details : remove the limit of 256 scsi disks in the sd_iostat patch
2436 Description: oops in sd_iostats_seq_show()
2437 Details : unloading/reloading the scsi low level driver triggers a kernel
2438 bug when trying to access the sd iostat file.
2443 Description: Kernel panics during QLogic driver reload
2444 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
2445 patch, causing memory corruption.
2450 Description: journal_dev option does not work in b1_6
2451 Details : pass mount option during pre-mount.
2453 Severity : enhancement
2455 Description: Add a FIEMAP(FIle Extent MAP) ioctl
2456 Details : FIEMAP ioctl will allow an application to efficiently fetch the
2457 extent information of a file. It can be used to map logical blocks
2458 in a file to physical blocks in the block device.
2462 Description: LDLM soft lockups - improvement
2463 Details : It is be possible to send the lock handle along with each read
2464 or write request because the client is already doing a lock match
2465 itself so there isn't any reason the OST should have to re-do that
2469 Frequency : only on Cray X2
2471 Description: X2 build failures
2472 Details : fix build failures on Cray X2.
2476 Description: xid & resent requests
2477 Details : Initialize RPC XID from clock at startup (randomly if clock is
2480 Severity : enhancement
2482 Description: Add lustre_start utility to start or stop multiple Lustre servers
2487 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
2488 Details : In case of memory pressure, list_del() can be called twice on
2489 req->rq_history_list, causing a kernel oops.
2493 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
2494 Details : fix stack overflow in the distributed lock manager by defering
2495 export eviction after a failed AST to the elt thread instead of
2496 handling it in the dlm interpret routine.
2500 Description: Convert some comments to new format.
2501 Details : Update documenting comments to match doxygen conventions.
2505 Description: Grammar fixes.
2506 Details : A couple of trivial spelling fixes.
2510 Description: OSD_COUNTERS-mandatory
2511 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
2512 to simplify the code.
2516 Description: simplify cmm_device freeing logic.
2517 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
2521 Description: Add lockdep support to dt_object_operations locking interface.
2522 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
2523 parameter indicating lock ordering. Update mdd code to use new
2528 Description: Introduce failloc constants for lockless IO tests.
2529 Details : Add two new failloc constants to test lockless IO. Only one of
2530 them in implemented---another is checked in yet to be landed
2535 Description: Add lockdep support for inode mutex.
2536 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
2537 in the situations where Lustre has to lock more than one inode
2542 Description: Add optional invariants checking support.
2543 Details : Add new LINVRNT() macro, optional on new --enable-invariants
2544 configure switch. This macro is to be used for consistency and
2545 sanity checks that are too expensive to be left in `production'
2550 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
2551 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
2556 Description: Add lockdep support for ldlm_lock and ldlm_resource.
2557 Details : Use spin_lock_nested() in (the only) situation where more than
2558 one ldlm_lock is locked simultaneously. Also, fix possible
2559 dead-lock in ldlm_lock_change_resource() by enforcing particular
2564 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
2565 Details : Instead of specifying each ldlm_lock call-back through separate
2566 parameter, wrap them into struct ldlm_callback_suite.
2570 Description: Kill join_lru obd method and its callers.
2571 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
2572 in memory---a requirement implemented through ->o_join_lru() obd
2573 method in HEAD. Remove this method and its users.
2577 Description: Add asynchronous ldlm ENQUEUE completion handler.
2578 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
2579 case that stock ldlm_completion_ast() cannot handle as it waits
2580 until lock is granted. Introduce new ldlm_completion_ast_async()
2581 for this. Also comment ldlm_completion_ast().
2585 Description: ldlm_error <-> errno conversion.
2586 Details : Add functions to map (rather arbitrary) between LDLM error codes
2587 and standard errno values. CLIO needs this to prevent LDLM specific
2588 constants from escaping ldlm and osc.
2592 Description: Kill unused ldlm_handle2lock_ns() function.
2593 Details : Kill unused ldlm_handle2lock_ns() function.
2597 Description: Add lu_ref support to ldlm_lock
2598 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
2599 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
2600 and ldlm_resource. LDLM interface has to be changed, because code that
2601 releases a reference on a lock, has to "know" what reference this is.
2602 In the most frequent case
2604 lock = ldlm_handle2lock(handle);
2606 LDLM_LOCK_PUT(lock);
2608 no changes are required. When any other reference (received _not_ from
2609 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
2610 instead of LDLM_LOCK_PUT().
2612 Arguably, changes are pervasive, and interface requires some discipline
2613 for proper use. On the other hand, it was very instrumental in finding
2614 a few leaked lock references.
2618 Description: Add ldlm_lock_addref_try().
2619 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
2620 attempts to addref a lock that might be being canceled
2625 Description: Add ldlm_weigh_callback().
2626 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
2627 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
2628 hard-coded `number of pages' logic.
2632 Description: Add lockdep annotations to llog code.
2633 Details : Use appropriately tagged _nested() locking calls in the places
2634 where llog takes more than one ->lgh_lock lock.
2638 Description: Add loi_kms_set().
2639 Details : Wrap kms updates into a helper function.
2643 Description: Constify instances of struct lsm_operations.
2644 Details : Constify instances of struct lsm_operations.
2648 Description: lu_conf support.
2649 Details : On a server, a file system object is uniquely identified
2650 by a fid, which is sufficient to locate and load all object
2651 state (inode). On a client, on the other hand, more data are
2652 necessary instantiate an object. Change lu_object_find() and
2653 friends to take additional `lu_conf' argument describing object.
2654 Typically this includes layout information.
2658 Description: lu_context fixes.
2659 Details : Introduce new lu_context functions that are needed on the client
2660 side, where some system threads (ptlrpcd) are shared by multiple
2661 modules, and so cannot be stopped during module shutdown.
2665 Description: Add start and stop methods to lu_device_type_operations.
2666 Details : Introduce two new methods in lu_device_type_operations, that are
2667 invoked when first instance of a given type is created and last one
2668 is destroyed respectively. This is need by CLIO.
2672 Description: Add lu_ref support to struct lu_device.
2673 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
2674 track leaked references.
2678 Description: Introduce lu_kmem_descr.
2679 Details : lu_kmem_descr and its companion interface allow to create
2680 and destroy a number of kmem caches at once.
2684 Description: Fix lu_object finalization race.
2685 Details : Fix a race between lu_object_find() finding an object and its
2686 concurrent finalization. This race is (most likely) not possible
2687 on the server, but might happen on the client.
2691 Description: Introduce lu_ref interface.
2692 Details : lu_ref is a debugging module allowing to track references to
2693 a given object. It is quite cpu expensive, and has to be
2694 explicitly enabled with --enable-lu_ref. See usage description
2699 Description: Factor lu_site procfs stats into a separate function.
2700 Details : Separate lu_site stats printing code into a separate function
2701 that can be reused on a client.
2705 Description: Constify instances of struct {lu,dt,md}_device_operations.
2706 Details : Constify instances of struct {lu,dt,md}_device_operations.
2710 Description: Introduce struct md_site and move meta-data specific parts of
2711 struct lu_site here.
2712 Details : Move md-specific fields out of struct lu_site into special struct
2713 md_site, so that lu_site can be used on a client.
2717 Description: Kill mdd_lov_destroy().
2718 Details : Remove unused mdd code.
2722 Description: Add st_block checking to multistat.c.
2723 Details : Add st_block checking to multistat.c.
2727 Description: Add lu_ref support to struct obd_device.
2728 Details : Add lu_ref tracking to obd_device.
2732 Description: Kill obd_set_fail_loc().
2733 Details : Remove unused code.
2737 Description: Add special type for ptlrpc_request interpret functions.
2738 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
2739 there. Actual usage will be in CLIO.
2743 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
2744 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
2745 doesn't work with lockdep.
2749 Description: Add rwv.c test program.
2750 Details : New testing program exercising readv(2) and writev(2) (Qian).
2754 Description: Add sendfile.c test program.
2755 Details : New testing program exercising sendfile(2) (Jay).
2759 Description: Ratelimit a message that can be very frequent.
2760 Details : Ratelimit a memory allocation failure message that can
2765 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2766 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2770 Description: Liblustre build fixes.
2771 Details : Liblustre build fixes.
2775 Description: libcfs: add cfs_{need,cond}_resched() interface.
2776 Details : libcfs: add cfs_{need,cond}_resched() definition and
2777 implementations for Linux, NT, and liblustre.
2779 Severity : enhancement
2781 Description: More exported tunables for mballoc
2782 Details : Add support for tunable preallocation window and new tunables for
2783 large/small requests
2787 Description: Detect corruption of block bitmap and checking for preallocations
2788 Details : Checks validity of on-disk block bitmap. Also it does better
2789 checking of number of applied preallocations. When corruption is
2790 found, it turns filesystem readonly to prevent further corruptions.
2794 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
2795 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
2796 window to 0. If it does hit cache miss, and read-pattern is still
2797 stride-io mode, does not reset the stride window, but also does not
2798 increase the stride window length in this case.
2802 Frequency : only for big-endian servers
2803 Description: Check if system is big-endian while mounting fs with extents feature
2804 Details : Mounting a filesystem with extents feature will fail on big-endian
2805 systems since ext3-based ldiskfs is not supported on big-endian
2806 systems. This can be over-riden with "bigendian_extents" mount option.
2808 Severity : enhancement
2810 Description: The root squash functionality
2811 Details : A security feature, which is to prevent users from being able
2812 to mount lustre on their desktop, run as root, and delete
2813 all of the files in the filesystem. The goal is accomplished by
2814 remapping user id (UID) and group id (GID) of the root user to
2815 a UID and GID specified by the system administartor via Lustre
2816 configuration management server (MGS). The functionality also
2817 allows to specify sets of clients for which the remapping does
2822 Description: Excessive recovery window
2823 Details : With AT enabled, the recovery window can be excessively long (6000+
2824 seconds). To address this problem, we no longer use
2825 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
2826 timeout no longer depends on the service time, it is set to
2827 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
2828 time via pb_service_time.
2832 Description: Watchdog triggered on MDS failover
2833 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
2834 the OSTs know that the MDS "UUID" can be reused for the same export
2835 from a different NID, so we do not need to wait for the export to be
2839 Frequency : rare, only if using MMP with Linux RAID
2841 Description: MMP doesn't work with Linux RAID
2842 Details : While using HA for Lustre servers with Linux RAID, it is possible
2843 that MMP will not detect multiple mounts. To make this work we
2844 need to unplug the device queue in RAID when the MMP block is being
2845 written. Also while reading the MMP block, we should read it from
2846 disk and not the cached one.
2848 Severity : enhancement
2850 Description: open file using fid
2851 Details : A file can be opened using just its fid, like
2852 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
2855 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
2858 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
2859 Details : add proper locking for creation of jbd_4k slab cache
2863 Description: MMP check in ext3_remount() fails without displaying any error
2864 Details : When multiple mount protection fails during remount, proper error
2867 Severity : enhancement
2869 Description: Allow stripe size to be up to 4G-64k
2870 Details : Fix math logic to allow large stripe sizes.
2874 Description: add check for >8TB ldiskfs filesystems
2875 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
2876 Don't allow >8TB ldiskfs filesystems to be mounted without
2877 force_over_8tb mount option
2882 Descriptoin: OSS complains "too many blocks" and crashes
2883 Details : In filter_commitrw_write():
2884 1. lquota_chkquota();
2885 2. obdo_from_inode();
2886 3. lquota_pending_commit();
2887 "2." has changed uid and gid in this case, then caused
2888 lquota_chkquota() and lquota_pending_commit() to use
2889 different uid and gid. That is the root of the bug.
2893 Descriptoin: getstripe doesn't report stripe count = -1 sanely
2894 Details : Detect and print out the 'stripe count = -1' case properly.
2898 Description: Out or order replies might be lost on replay
2899 Details : In ptlrpc_retain_replayable_request if we cannot find retained
2900 request with tid smaller than one currently being added, add it
2901 to the start, not end of the list.
2903 --------------------------------------------------------------------------------
2905 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
2907 * Support for kernels:
2908 2.6.5-7.283 (SLES 9),
2909 2.6.9-55.EL (RHEL 4),
2910 2.6.16.46-0.14 (SLES 10),
2911 2.6.18.8 vanilla (kernel.org)
2912 * Client support for unpatched kernels:
2913 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2914 2.6.16 - 2.6.22 vanilla (kernel.org)
2915 * Due to recently discovered recovery problems, we do not recommend
2916 using patchless RHEL 4 clients with this or any earlier release.
2917 * Recommended e2fsprogs version: 1.39.cfs8
2918 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2919 * Starting with this release, the ldiskfs backing filesystem required
2920 by Lustre is now in its own package, lustre-ldiskfs. This package
2921 should be installed. It is versioned separately from Lustre and
2922 may be released separately in future.
2926 Description: block reactivating mgc import until all deactivates complete
2927 Details : Fix race when failing back MDT/MGS to itself (testing)
2930 Frequency : at statup only
2932 Description: mds_lov_synchronize race leads to various problems
2933 Details : simultaneous MDT->OST connections at startup can cause the
2934 sync to abort, leaving the OSC in a bad state.
2936 Severity : enhancement
2938 Description: add optional extra BUILD_VERSION info
2939 Details : add a new environment variable (namely LUSTRE_VERS) which allows
2940 to override the lustre version.
2943 Frequency : 2.6.18 servers only
2945 Description: ll_kern_mount() doesn't release the module reference
2946 Details : The ldiskfs module reference count never drops down to 0
2947 because ll_kern_mount() doesn't release the module reference.
2952 Description: server LBUG when using old ost_num_threads parameter
2953 Details : Accept the old ost_num_threads parameter but warn that it
2954 is deprecated, and fix an off-by-one error that caused an LBUG.
2959 Description: Transient SCSI error results in persistent IO issue
2960 Details : iobuf->dr_error is not reinitialized to 0 between two
2964 Frequency : sometimes when underlying device returns I/O errors
2966 Description: OSTs not going read-only during write failures
2967 Details : OSTs are not remounted read-only when the journal commit threads
2968 get I/O errors because fsfilt_ext3 calls journal_start/stop()
2969 instead of the ext3 wrappers.
2973 Description: poor connect scaling with increasing client count
2974 Details : Don't run filter_grant_sanity_check for more than 100 exports
2975 to improve scaling for large numbers of clients.
2978 Frequency : SLES10 only
2980 Description: sanity-quota.sh quotacheck failed: rc = -22
2981 Details : Quotas cannot be enabled on SLES10.
2984 Frequency : liblustre clients only
2986 Description: getdirentries does not give error when run on compute nodes
2987 Details : getdirentries does not fail when the size specified as an argument
2988 is too small to contain at least one entry
2990 Severity : enhancement
2992 Description: Add LNET router traceability for debug purposes
2993 Details : If a checksum failure occurs with a router as part of the
2994 IO path, the NID of the last router that forwarded the bulk data
2995 is printed so it can be identified.
3000 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
3001 Details : Due to a race condition, liblustre clients were occasionally
3002 evicted incorrectly.
3004 Severity : enhancement
3006 Description: lfs setstripe use optional parameters instead of postional
3009 Severity : enhancement
3011 Description: Nanosecond timestamp support for ldiskfs
3012 Details : The on-disk ldiskfs filesystem has added support for nanosecond
3013 resolution timestamps. There is not yet support for this at
3014 the Lustre filesystem level.
3017 Frequency : during server recovery
3019 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
3020 Details : request with rq_no_resend flag not awake l_wait_event if they get
3024 Frequency : nfs export on patchless client
3026 Description: connectathon hang when test nfs export over patchless client
3027 Details : Disconnected dentry cannot be found with lookup, so we do not need
3028 to unhash it or make it invalid
3031 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
3032 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
3033 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
3038 Description: open req refcounting wrong on reconnect
3039 Details : If reconnect happened between getting open reply from server and
3040 call to mdc_set_replay_data in ll_file_open, we will schedule
3041 replay for unreferenced request that we are about to free.
3042 Subsequent close will crash in variety of ways.
3043 Check that request is still eligible for replay in
3044 mdc_set_replay_data().
3049 Description: disable writes to filesystem when reading health_check file
3050 Details : the default for reading the health_check proc file has changed
3051 to NOT do a journal transaction and write to disk, because this
3052 can cause reads of the /proc file to hang and block HA state
3053 checking on a healthy but otherwise heavily loaded system. It
3054 is possible to return to the previous behaviour during configure
3055 with --enable-health-write.
3057 Severity : enhancement
3059 Description: 64-bit inode version
3060 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
3061 to the inode. This will be required for version-based recovery.
3066 Description: MDS fails to start if a duplicate client export is detected
3067 Details : in some rare cases it was possible for a client to connect to
3068 an MDS multiple times. Upon recovery the MDS would detect this
3069 and fail during startup. Handle this more gracefully.
3071 Severity : enhancement
3073 Description: Add -o localflock option to simulate old noflock
3075 Details : This will achieve local-only flock/fcntl locks
3081 Description: log_commit_thread vs filter_destroy race leads to crash
3082 Details : Take import reference before releasing llog record semaphore
3087 Description: Wrong request locking in request set processing
3088 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
3089 imp_delayed_list, in this place should be used imp_lock.
3092 Frequency : when reconnection
3094 Description: Grant Leak when osc reconnect to OST
3095 Details : When osc reconnect ost, OST(filter) should check whether it
3096 should grant more space to client by comparing fed_grant and
3097 cl_avail_grant, and return the granted space to client instead
3098 of "new granted" space, because client will call osc_init_grant
3099 to update the client grant space info.
3102 Frequency : when client reconnect to OST
3104 Description: Grant Leak when osc do resend and replay bulk write
3105 Details : When osc reconnect to OST, OST(filter)should clear grant info of
3106 bulk write request, because the grant info while be sync between
3107 OSC and OST when reconnect, and we should ignore the grant info
3108 these of resend/replay write req.
3113 Description: Grant space more than available left space sometimes.
3114 Details : When then OST is about to be full, if two bulk writing from
3115 different clients came to OST. Accord the avaliable space of the
3116 OST, the first req should be permitted, and the second one
3117 should be denied by ENOSPC. But if the seconde arrived before
3118 the first one is committed. The OST might wrongly permit second
3119 writing, which will cause grant space > available space.
3122 Frequency : when client is evicted
3124 Description: Grant might be wrongly erased when osc is evicted by OST
3125 Details : when the import is evicted by server, it will fork another
3126 thread ptlrpc_invalidate_import_thread to invalidate the
3127 import, where the grant will be set to 0. While the original
3128 thread will update the grant it got when connecting. So if
3129 the former happened latter, the grant will be wrongly errased
3130 because of this race.
3135 Description: Checking Stale with correct fid
3136 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
3137 to check whether it is stale, because sometimes, we want the
3138 enqueue happened anyway, and op_data.fid2 will not be initialized.
3140 Severity : enhancement
3142 Description: update patchless client
3143 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
3146 Frequency : only with 2.4 kernel
3148 Description: random memory corruption
3149 Details : size of struct ll_inode_info is to big for union inode.u and this
3150 can be cause of random memory corruption.
3155 Description: Memory leak in recovery
3156 Details : Lov_mds_md was not free in an error handler in mds_create_object.
3157 It should also check obd_fail before fsfilt_start, otherwise if
3158 fsfilt_start return -EROFS,(failover mds during mds recovery).
3159 then the req will return with repmsg->transno = 0 and rc = EROFS.
3160 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
3161 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
3162 be freed no matter whether fsfilt_commit success or not.
3165 Frequency : only with huge count clients
3167 Description: Prevents from taking the superblock lock in llap_from_page for
3169 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
3170 for llap_from_page call in ll_removepage prevents from taking the
3171 superblock lock for a soon died page.
3176 Description: Not check open intent error before release open handle
3177 Details : in some rare cases, the open intent error is not checked before
3178 release open handle, which may cause
3179 ASSERTION(open_req->rq_transno != 0), because it tries to release
3180 the failed open handle.
3185 Description: Set cat log bitmap only after create log success.
3186 Details : in some rare cases, the cat log bitmap is set too early. and it
3187 should be set only after create log success.
3191 Description: Accessing a block bevice can re-enable I/O when Lustre is
3192 tearing down a device.
3193 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
3198 Description: service threads may hog cpus when there are a lot of requests
3199 Details : Insert cond_resched to give other threads a chance to use some CPU
3204 Description: the cat log was not initialized in recovery
3205 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
3206 unlink log on mds will not be initialized until mds post
3207 recovery. And also in mds post recovery, the unlink log will
3208 initialization will be done asynchronausly, so there will be race
3209 between add unlink log and unlink log initialization.
3213 Description: brw_stats were being printed incorrectly
3214 Details : brw_stats were being printed as log2 but all of them were not
3215 recorded as log2. Also remove some code duplication arising from
3216 filter_tally_{read,write}.
3220 Frequency : rare, only in recovery.
3221 Description: ASSERTION(req->rq_type != LI_POISON) failed
3222 Details : imp_lock should be held while iterating over imp_sending_list for
3223 prevent destroy request after get timeout in ptlrpc_queue_wait.
3227 Description: replay-single.sh test 52 fails
3228 Details : A lock's skiplist need to be cleanup when it being unlinked
3229 from its resource list.
3233 Description: Short directio read returns full requested size rather than
3235 Details : Direct I/O operations should return actual amount of bytes
3236 transferred rather than requested size.
3238 Severity : enhancement
3240 Description: metadata RPC reduction (e.g. for rm performance)
3241 Details : decrease the amount of synchronous RPC between clients and servers
3242 by canceling conflicing lock before the operation on the client
3243 and packing thier handles into the main operation RPC to server.
3245 Severity : enhancement
3247 Description: Async OSC create to avoid the blocking unnecessarily.
3248 Details : If an OST has no remaining object, system will block on the
3249 creation when it needs to create a new object on this OST. Now,
3250 ways use pre-created objects when available, instead of blocking on
3251 an empty osc while others are not empty. If we must block, we
3252 block for the shortest possible period of time.
3256 Description: improve handling recoverable errors
3257 Details : if request processig with error which can be recoverable on server
3258 request should be resend, otherwise page released from cache and
3261 Severity : enhancement
3263 Description: refine locking for avoid write wrong info into lov_objid file
3264 Details : fix possible races with add new target and write/update data in
3267 --------------------------------------------------------------------------------
3269 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
3274 Frequency : on some architectures
3276 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
3277 Details : Uninitialized flags sometimes cause configuration commands to
3281 Frequency : patchless clients only
3283 Description: missing __iget() symbol export
3284 Details : The __iget() symbol export is missing. To avoid the need for
3285 this on patchless clients the deathrow inode reaper is turned
3286 off, and we depend on the VM to clean up old inodes. This
3287 dependency was during via the fix for bug 12181.
3289 --------------------------------------------------------------------------------
3291 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
3293 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
3294 INTEROPERATE with older versions automatically. In many cases a
3295 special upgrade step is needed. Please read the
3296 user documentation before upgrading any part of a live system.
3297 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
3298 version will not interoperate with 1.6 betas before beta5 (1.5.95).
3299 * WARNING: Lustre configuration and startup changes are required with
3300 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
3304 Severity : enhancement
3306 Description: MountConf
3307 Details : Lustre configuration is now managed via mkfs and mount
3308 commands instead of lmc and lconf. New obd types (MGS, MGC)
3309 are added for dynamic configuration management. See
3310 https://mail.clusterfs.com/wikis/lustre/MountConf for
3313 Severity : enhancement
3315 Description: dynamic OST addition
3316 Details : OSTs can now be added to a live filesystem
3318 Severity : enhancement
3320 Description: startup order invariance
3321 Details : MDTs and OSTs can be started in any order. Clients only
3322 require the MDT to complete startup.
3324 Severity : enhancement
3326 Description: parallel, asynchronous orphan cleanup
3327 Details : orphan cleanup is now performed in separate threads for each
3328 OST, allowing parallel non-blocking operation.
3330 Severity : enhancement
3332 Description: optimized stripe assignment
3333 Details : stripe assignments are now made based on ost space available,
3334 ost previous usage, and OSS previous usage, in order to try
3335 to optimize storage space and networking resources.
3337 Severity : enhancement
3339 Description: Permanently set tunables
3340 Details : All writable /proc/fs/lustre tunables can now be permanently
3341 set on a per-server basis, at mkfs time or on a live system.
3343 Severity : enhancement
3345 Description: Lustre message v2
3346 Details : Add lustre message format v2.
3348 Severity : enhancement
3350 Description: client OST exclusion list
3351 Details : Clients can be started with a list of OSTs that should be
3352 declared "inactive" for known non-responsive OSTs.
3356 Description: SPEC SFS validation failure on NFS v2 over lustre.
3357 Details : Changes the blocksize for regular files to be 2x RPC size,
3358 and not depend on stripe size.
3360 Severity : enhancement
3362 Description: Multiple MD RPCs in flight.
3363 Details : Further unserialise some read-only MDS RPCs - learn about intents.
3364 To avoid overly-overloading MDS, introduce a limit on number of
3365 MDS RPCs in flight for a single client and add /proc controls
3366 to adjust this limit.
3368 Severity : enhancement
3370 Description: client read/write statistics
3371 Details : Add client read/write call usage stats for performance
3372 analysis of user processes.
3373 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
3374 file access. extents_stats shows chunk size distribution.
3375 extents_stats_per_process show chunk size distribution per
3378 Severity : enhancement
3380 Description: mds statistics
3381 Details : Add detailed mds operations statistics in
3382 /proc/fs/lustre/mds/*/stats.
3386 Description: Failure of copying files with lustre special EAs.
3387 Details : Client side always return success for setxattr call for lustre
3388 special xattr (currently only "trusted.lov").
3393 Description: Refcount LNET uuids
3394 Details : The global LNET uuid list grew linearly with every startup;
3395 refcount repeated list entries instead of always adding to
3398 Severity : enhancement
3400 Description: Dynamic service threads
3401 Details : Within a small range, start extra service threads
3402 automatically when the request queue builds up.
3405 Frequency : mixed-endian client/server environments
3407 Description: mixed-endian crashes
3408 Details : The new msg_v2 system had some failures in mixed-endian
3411 Severity : enhancement
3413 Description: Easy OST removal
3414 Details : OSTs can be permanently deactivated with e.g. 'lctl
3415 conf_param lustre-OST0001.osc.active=0'
3417 Severity : enhancement
3419 Description: MGS proc entries
3420 Details : Added basic proc entries for the MGS showing what filesystems
3423 Severity : enhancement
3425 Description: provide MGS failover
3426 Details : Added config lock reacquisition after MGS server failover.
3428 Severity : enhancement
3430 Description: add Linux 2.4 support
3431 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
3435 Description: a large application tries to do I/O to the same resource and dies
3436 in the middle of it.
3437 Details : Check the req->rq_arrival time after the call to
3438 ost_brw_lock_get(), but before we do anything about
3439 processing it & sending the BULK transfer request. This
3440 should help move old stale pending locks off the queue as
3441 quickly as obd_timeout.
3444 Frequency : when an incorrect nid is specified during startup
3446 Description: ptlrpc connect to non-existant node causes kernel crash
3447 Details : LNET can't be re-entered from an event callback, which
3448 happened when we expire a message after the export has been
3449 cleaned up. Instead, hand the zombie cleanup off to another
3452 Severity : enhancement
3454 Description: plain/inodebits lock performance improvement
3455 Details : Grouping plain/inodebits in granted list by their request modes
3456 and bits policy, thus improving the performance of search through
3460 Frequency : only if OST filesystem is corrupted
3462 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3463 Details : for a short time RPCs with bulk IO are in the replay list,
3464 but replay of bulk IOs is unimplemented. If the OST filesystem
3465 is corrupted due to disk cache incoherency and then replay is
3466 started it is possible to trip an assertion. Avoid putting
3467 committed RPCs into the replay list at all to avoid this issue.
3470 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3472 Description: System hang on startup
3473 Details : This bug allowed the liblustre (e.g. catamount) client to
3474 return to the app before handling all startup RPCs. This
3475 could leave the node unresponsive to lustre network traffic
3476 and manifested as a server ptllnd timeout.
3478 Severity : enhancement
3480 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3481 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
3482 boolean to control whether to print peer debug info when a
3483 client's RPC times out.
3486 Frequency : only for kernels with patches from Lustre below 1.4.3
3488 Description: Remove old rdonly API
3489 Details : Remove old rdonly API which unused from at least lustre 1.4.3
3492 Frequency : only for devices with external journals
3494 Description: Set external device read-only also
3495 Details : During a commanded failover stop, we set the disk device
3496 read-only while the server shuts down. We now also set any
3497 external journal device read-only at the same time.
3500 Frequency : when upgrading from 1.4 while trying to change parameters
3502 Description: The wrong (new) MDC name was used when setting parameters for
3503 upgraded MDT's. Also allows changing of OSC (and MDC)
3504 parameters if --writeconf is specified at tunefs upgrade time.
3507 Frequency : when setting specific ost indicies
3509 Description: QOS code breaks on skipped indicies
3510 Details : Add checks for missing OST indicies in the QOS code, so OSTs
3511 created with --index need not be sequential.
3513 Severity : enhancement
3515 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3516 Details : The uninit_groups feature works in conjunction with the kernel
3517 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3518 up the pass1 processing of e2fsck. This is a read-only feature
3519 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3520 mount filesystems that have had this feature enabled.
3522 Severity : enhancement
3524 Description: Improve multi-block allocation algorithm to avoid fragmentation
3525 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3526 allocation locality and avoid filesystem fragmentation.
3528 ------------------------------------------------------------------------------
3530 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
3532 * Support for kernels:
3533 2.4.21-47.0.1.EL (RHEL 3)
3534 2.6.5-7.283 (SLES 9)
3535 2.6.9-42.0.10.EL (RHEL 4)
3536 2.6.12.6 vanilla (kernel.org)
3537 2.6.16.27-0.9 (SLES 10)
3538 * Recommended e2fsprogs version: 1.39.cfs5
3540 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3544 Frequency : occasional, depends on client load and configuration
3545 Bugzilla : 12181, 12203
3546 Description: data loss for recently-modified files
3548 Details : In some cases it is possible that recently written or created
3549 files may not be written to disk in a timely manner (this should
3550 normally be within 30s unless client IO load is very high).
3551 The problem appears as zero-length files or files that are a
3552 multiple of 1MB in size after a client crash or client eviction
3553 that are missing data at the end of the file.
3555 This problem is more likely to be hit on clients where files are
3556 repeatedly created and unlinked in the same directory, clients
3557 have a large amount of RAM, have many CPUs, the filesystem has
3558 many OSTs, the clients are rebooted frequently, and/or the files
3559 are not accessed by other nodes after being written.
3561 The presence of the problem can be detected by looking at
3562 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
3563 smaller than the second (nr_unused) then dirty files will not
3564 be flushed automatically to disk. "sync; sleep 10" should be
3565 run several times on the node before unmounting it to update
3566 Lustre (this is also safe to run on nodes without this problem).
3568 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
3569 that can cause this same problem, so customers using that kernel
3570 also need to update the kernel in addition to Lustre. In order
3571 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
3573 It is normal that files written just before a client crash (less
3574 than 30s) may not yet have been flushed to disk, even for local
3578 Frequency : frequent on thin XT3 nodes
3580 Description: UUID collision on thin XT3 Linux nodes
3581 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
3582 randomly, since we relied on an insufficiently-seeded PRNG.
3587 Description: OSS hangs after "All ost request buffers busy"
3588 Details : A deadlock between quota and journal operations caused OSS
3589 hangs after printing "All ost request buffers busy."
3592 Frequency : always on liblustre builds
3594 Description: Cleanup compiler warnings on liblustre
3597 Frequency : always on liblustre builds on XT3
3599 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
3603 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
3604 > 32000 subdirectories
3605 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
3606 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
3607 subdirectory count crosses 32000. This will aid e2fsck to
3608 correctly handle more than 32000 subdirectories.
3611 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3613 Description: System hang on startup
3614 Details : This bug allowed the liblustre (e.g. catamount) client to
3615 return to the app before handling all startup RPCs. This
3616 could leave the node unresponsive to lustre network traffic
3617 and manifested as a server ptllnd timeout.
3619 Severity : enhancement
3621 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3622 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
3623 boolean to control whether to print peer debug info when a
3624 client's RPC times out.
3629 Description: make O_SYNC working on 2.6 kernels
3630 Details : 2.6 kernels use different method for mark pages for write,
3631 so need add a code to lustre for O_SYNC work.
3636 Description: Failure to close file and release space on NFS
3637 Details : Put inode details into lock acquired in ll_intent_file_open.
3638 Use mdc_intent_lock in ll_intent_open to properly
3639 detect all kind of errors unhandled by mdc_enqueue.
3644 Description: proc file read during shutdown sometimes raced obd removal,
3646 Details : Add lock to prevent obd access after proc file removal.
3649 Frequency : Only for files larger than 4GB on 32-bit clients.
3651 Description: improperly doing page alignment of locks
3652 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
3653 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3658 Description: RPCs being resent when they shouldn't be
3659 Details : Some RPCs that should not be resent are being resent. This
3660 can cause inconsistencies in the RPC state machine. Do not
3661 resend such requests.
3664 Frequency : rare, only with NFS export
3666 Description: Crash on NFS re-export node
3667 Details : under very unusual load conditions an assertion is hit in
3668 ll_intent_file_open()
3671 Frequency : only if OST filesystem is corrupted
3673 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3674 Details : for a short time RPCs with bulk IO are in the replay list,
3675 but replay of bulk IOs is unimplemented. If the OST filesystem
3676 is corrupted due to disk cache incoherency and then replay is
3677 started it is possible to trip an assertion. Avoid putting
3678 committed RPCs into the replay list at all to avoid this issue.
3683 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3684 Details : Large single O_DIRECT read and write calls can fail to allocate
3685 a sufficiently large buffer to process the request. In case of
3686 allocation failure the allocation is retried with a smaller
3687 buffer and broken into smaller requests.
3689 Severity : enhancement
3691 Description: Add -o localflock option to simulate old noflock behaviour.
3692 Details : This will achieve local-only flock/fcntl locks coherentness.
3697 Description: versioning check is incomplete
3698 Details : Checking the version difference of client vs. server, report
3699 error if the gap is too big.
3704 Description: add support PG_writeback bit
3705 Details : add support for PG_writeback bit for Lustre, for more carefull
3706 work with page cache in 2.6 kernel. This also fix some deadlocks
3707 and remove hack for work O_SYNC with 2.6 kernel.
3709 Severity : enhancement
3711 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3712 Details : The uninit_groups feature works in conjunction with the kernel
3713 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3714 up the pass1 processing of e2fsck. This is a read-only feature
3715 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3716 mount filesystems that have had this feature enabled.
3718 Severity : enhancement
3720 Description: Improve multi-block allocation algorithm to avoid fragmentation
3721 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3722 allocation locality and avoid filesystem fragmentation.
3724 ------------------------------------------------------------------------------
3726 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3728 * Support for kernels:
3729 2.6.9-42.0.3.EL (RHEL 4)
3730 2.6.5-7.276 (SLES 9)
3731 2.4.21-47.0.1.EL (RHEL 3)
3732 2.6.12.6 vanilla (kernel.org)
3733 2.6.16.21-0.8 (SLES10)
3734 * Recommended e2fsprogs version: 1.39.cfs2-0
3736 * The backwards-compatible /proc/sys/portals symlink has been removed
3737 in this release. Before upgrading, please ensure that you change
3738 any configuration scripts or /etc/sysctl.conf files that access
3739 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3740 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3741 in advance of the upgrade on any system running Lustre 1.4.6 or
3742 newer, since /proc/sys/lnet was added in that version.
3743 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3747 Frequency : only when quota is used
3749 Description: avoid scanning export list for quota master
3750 Details : Change the algorithms to avoid scanning export list in order
3751 to improve the efficiency.
3754 Frequency : MDS failover only, very rarely
3756 Description: "went back in time" messages on mds failover
3757 Details : The greatest transno may be lost when the current operation
3758 finishes with an error (transno==0) and the client's last_rcvd
3759 record is over-written. Save the greatest transno in the
3760 mds_last_transno for this case.
3763 Frequency : always for specific kernels and striping counts
3765 Description: client may get "Matching packet too big" without ACL support
3766 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
3767 when trying to access files in certain configurations. The
3768 clients should in fact be denied when mounting because they do
3769 not understand ACLs.
3772 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
3774 Description: many clients connecting with IO in progress causes connect timeouts
3775 Details : Avoid synchronous journal commits to avoid delays caused by many
3776 clients connecting/disconnecting when bulk IO is in progress.
3777 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
3778 OST_IO_PORTAL to avoid delays behind potentially many pending
3782 Frequency : occasionally with multiple writers to a single file
3784 Description: shared writes to file may result in wrong size reported by stat()
3785 Details : Allow growing of kms when extent lock is cancelled
3788 Frequency : always with random mmap IO to multi-striped file
3790 Description: mmap write might be lost if we are writing to a 'hole' in stripe
3791 Details : Only if the hole is at the end of OST object so that kms is too
3792 small. Fix is to increase kms accordingly in ll_nopage.
3795 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
3797 Description: writes to a missing object would leak memory on the OST
3798 Details : If there is an inconsistency between the MDS and OST filesystems,
3799 such that the MDS references an object that doesn't exist, writes
3800 to that object will leak memory due to incorrect cleanup in the
3801 error handling path, eventually running out of memory on the OST.
3806 Description: Creating too long symlink causes lustre errors
3807 Details : Check symlink and name lengths before sending requests to MDS.
3810 Frequency : only if flock is enabled (not on by default)
3812 Description: posix locks not released on fd closure on 2.6.9+
3813 Details : We failed to add posix locks to list of inode locks on 2.6.9+
3814 kernels, this caused such locks not to be released on fd close and
3815 then assertions on fs unmount about still used locks.
3818 Frequency : MDS failover only, very rarely
3820 Description: clients may get ASSERTION(granted_lock != NULL)
3821 Details : When request was taking a long time, and a client was resending
3822 a getattr by name lock request. The were multiple lock requests
3823 with the same client lock handle and
3824 mds_getattr_name->fixup_handle_for_resent_request found one of the
3825 lock handles but later failed with ASSERTION(granted_lock != NULL).
3830 Description: handle->h_buffer_credits > 0, assertion failure
3831 Details : h_buffer_credits is zero after truncate, causing assertion
3832 failure. This patch extends the transaction or creates a new
3836 Frequency : NFS re-export or patchless client
3837 Bugzilla : 11179, 10796
3838 Description: Crash on NFS re-export node (__d_move)
3839 Details : We do not want to hash the dentry if we don't have a lock.
3840 But if this dentry is later used in d_move, we'd hit uninitialised
3841 list head d_hash, so we just do this to init d_hash field but
3842 leave dentry unhashed.
3845 Frequency : NFS re-export or patchless client
3847 Description: NFS exports has problem with symbolic link
3848 Details : lustre client didn't properly install dentry when re-exported
3849 to NFS or running patchless client.
3852 Frequency : NFS re-export or patchless client
3854 Description: Various nfs/patchless fixes.
3855 Details : fixes reuse disconected alias for lookup process - this fixes
3856 warning "find_exported_dentry: npd != pd",
3857 fix permission error with open files at nfs.
3858 fix apply umask when do revalidate.
3861 Frequency : occasional
3863 Description: Crash on NFS re-export node
3864 Details : calling clear_page() on the wrong pointer triggered oops in
3865 generic_mapping_read().
3868 Frequency : rarely, using O_DIRECT IO
3870 Description: unaligned directio crashes client with LASSERT
3871 Details : check for unaligned buffers before trying any requests.
3874 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
3876 Description: stale data returned from RAID cache
3877 Details : If only a small amount of IO is done to the RAID device before
3878 reading it again it is possible to get stale data from the RAID
3879 cache instead of reading it from disk.
3882 Frequency : always for sles10 kernel
3884 Description: sles10 support
3885 Details : ll_follow_link: compile fixes and using of nd_set_link
3886 under newer kernels.
3889 Frequency : depends on arch, kernel and compiler version, always on sles10
3892 Description: recursive or deep enough symlinks cause stack overflow
3893 Details : getting rid of large stack-allocated variable in
3897 Frequency : depends on hardware
3899 Description: lustre write performance loss in the SLES10 kernel
3900 Details : the performance loss is caused by using of write barriers in the
3901 ext3 code. The SLES10 kernel turns barrier support on by
3902 default. The fix is to undo that change for ldiskfs.
3904 ------------------------------------------------------------------------------
3906 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
3908 * Support for kernels:
3909 2.6.9-42.0.3EL (RHEL 4)
3910 2.6.5-7.276 (SLES 9)
3911 2.4.21-47.0.1.EL (RHEL 3)
3912 2.6.12.6 vanilla (kernel.org)
3916 Frequency : quota enabled and large files being deleted
3918 Description: releasing more than 4GB of quota at once hangs OST
3919 Details : If a user deletes more than 4GB of files on a single OST it
3920 will cause the OST to spin in an infinite loop. Release
3921 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3926 Description: statfs data retrieved from /proc may be stale or zero
3927 Details : When reading per-device statfs data from /proc, in the
3928 {kbytes,files}_{total,free,avail} files, it may appear
3929 as zero or be out of date.
3932 Frequency : systems with MD RAID1 external journal devices
3934 Description: lconf's call to blkid is confused by RAID1 journal devices
3935 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3936 of returning all block devices that match the journal UUID.
3939 Frequency : always, for aggregate stripe size over 4GB
3941 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
3942 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
3943 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
3946 Frequency : always if "lfs find" used on a local file/directory
3948 Description: "lfs find" segfaults if used on a local file/directory
3949 Details : The case where a directory component was not specified wasn't
3950 handled correctly. Handle this properly.
3953 Frequency : always on ppc64
3955 Description: the write to an ext3 filesystem mounted with mballoc got stuck
3956 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3957 perform endianness conversion.
3960 Frequency : rarely (truncate to non-zero file size after write under load)
3961 Bugzilla : 10730, 10687
3962 Description: Files padded with zeros to next 4K multiple
3963 Details : With filesystems mounted using the "extents" option (2.6 kernels)
3964 it is possible that files that are truncated to a non-zero size
3965 immediately after being written are filled with zero bytes beyond
3966 the truncated size. No file data is lost.
3968 Severity : enhancement
3970 Description: Allow recovery/failover for liblustre clients.
3971 Details : liblustre clients were unaware of failover configurations until
3974 Severity : enhancement
3976 Description: user file locks should fail when not mounting with flock option
3977 Details : Set up an error-returning stub in ll_file_operations.lock field
3978 to prevent incorrect behaviour when client is mounted without
3979 flock option. Also, set up properly f_op->flock field for
3983 Frequency : always on ia64
3985 Description: "lfs df" loops on printing out MDS statfs information
3986 Details : The obd_ioctl_data was not initialized and in some systems
3987 this caused a failure during the ioctl that did not return
3988 an error. Initialize the struct and return an error on failure.
3991 Frequency : SLES 9 only
3993 Description: Error of copying files with lustre special EAs as root
3994 Details : Client side always return success for setxattr call for lustre
3995 special xattr (currently only "trusted.lov").
3998 Frequency : rarely on clusters with both ia64+i386 clients
4000 Description: ia64+i686 clients doing shared IO on the same file may LBUG
4001 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
4002 clients are doing concurrent writes to the same file it is
4003 possible that the ia64 clients may LASSERT because the OST
4004 extent locks are not PAGE_SIZE aligned. Ensure that grown
4005 locks are always aligned on the request boundary.
4008 Frequency : specific use, occasional
4010 Description: Overwriting in use executable truncates on-disk binary image
4011 Details : If one node attempts to overwrite an executable in use by
4012 another node, we now correctly return ETXTBSY instead of
4013 truncating the file.
4015 Severity : enhancement
4017 Description: Async OSC create to avoid the blocking unnecessarily.
4018 Details : If an OST has no remaining object, system will block on the creating
4019 when need to create a new object on this OST. Now, ways use
4020 pre-created objects when available, instead of blocking on an
4021 empty osc while others are not empty. If we must block, we block
4022 for the shortest possible period of time.
4027 Description: chmod on Lustre root is propagated to other clients
4028 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
4029 invalid by the follow_mount time.
4032 Frequency : liblustre clients only
4034 Description: Race in 'instant cancel' lock handling could lead to such locks
4035 never to be granted in case of SMP MDS
4036 Details : Do not destroy not yet granted but cbpending locks in
4040 Frequency : replay/resend of open
4042 Description: non null lock assetion failure in mds_intent_policy
4043 Details : Trying to replay/resend lockless open requests resulted in
4044 mds_open() returning 0 with no lock. Now it sets a flag if
4045 a lock is going to be returned.
4047 Severity : enhancement
4049 Description: Checksum enhancements
4050 Details : New checksum enhancements allow for resending RPCs that failed
4053 Severity : enhancement
4055 Description: Tunables on number of dirty pages in cacche
4056 Details : Allow to set limit on number of dirty pages cached.
4061 Description: client crash on unmount - lock still has references
4062 Details : In some error handling cases it was possible to leak a lock
4063 reference on a client while accessing a file. This was not
4064 harmful to the client during operation, but would cause the
4065 client to crash when the filesystem is unmounted.
4068 Frequency : specific case, rare
4070 Description: ETXTBSY on mds though file not in use by client
4071 Details : ETXTBSY is no longer incorrectly returned when attempting to
4072 chmod or chown a directory that the user previously tried to
4073 execute or a currently-executing binary.
4076 Frequency : extremely rare except on liblustre-based clients
4078 Description: Lustre space not freed when files are deleted
4079 Details : Clean up open-unlinked files after client eviction. Previously
4080 the unlink was skipped and the files remained as orphans.
4085 Description: OST failure "would be an LBUG" in waiting_locks_callback()
4086 Details : In some cases it was possible to send a blocking callback to
4087 a client doing a glimpse, even though that client didn't get
4088 a lock granted. When the glimpse lock is cancelled on the OST
4089 the freed lock is left on the waiting list and corrupted the list.
4092 Frequency : all core dumps
4094 Description: Broke core dumps to lustre
4095 Details : Negative dentry may be unhashed if parent does not have UPDATE
4096 lock, but some callers, e.g. do_coredump, expect dentry to be
4097 hashed after successful create, hash it in ll_create_it.
4099 ------------------------------------------------------------------------------
4101 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
4103 * Support for kernels:
4104 2.6.9-42.0.2.EL (RHEL 4)
4105 2.6.5-7.276 (SLES 9)
4106 2.4.21-40.EL (RHEL 3)
4107 2.6.12.6 vanilla (kernel.org)
4111 Frequency : always on RHEL 3
4113 Description: Number of open files grows over time
4114 Details : The number of open files grows over time, whether or not
4115 Lustre is started. This was due to a filp leak introduced
4116 by one of our kernel patches.
4118 ------------------------------------------------------------------------------
4120 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4122 * Support for kernels:
4123 2.6.9-42.EL (RHEL 4)
4124 2.6.5-7.276 (SLES 9)
4125 2.4.21-40.EL (RHEL 3)
4126 2.6.12.6 vanilla (kernel.org)
4131 Bugzilla : 5719, 9635, 9792, 9684
4132 Description: OST (or MDS) trips assertions in (re)connection under heavy load
4133 Details : If a server is under heavy load and cannot reply to new
4134 connection requests before the client resends the (re)connect,
4135 the connection handling code can behave badly if two service
4136 threads are concurrently handing separate (re)connections from
4137 the same client. Add better locking to the connection handling
4138 code, and ensure that only a single connection will be processed
4139 for a given client UUID, even if the lock is dropped.
4141 Severity : enhancement
4143 Description: add TCP zero-copy support to kernel
4144 Details : Add support to the kernel TCP stack to allow zero-copy bulk
4145 sends if the hardware supports scatter-gather and checksumming.
4146 This allows socklnd to do client-write and server-read more
4147 efficiently and reduce CPU utilization from skbuf copying.
4150 Frequency : only if NFS exporting from client
4152 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
4153 Details : If setting attributes on a file created under NFS that had
4154 never been opened it would be possible to oops the client
4155 if the file had no objects.
4158 Frequency : always for liblustre
4160 Description: liblustre client does MDS+OSTs setattr RPC for each write
4161 Details : When doing a write from a liblustre client, the client
4162 incorrectly issued an RPC to the MDS and each OST the file was
4163 striped over in order to update the timestamps. When writing
4164 with small chunks and many clients this could overwhelm the MDS
4165 with RPCs. In all cases it would slow down the write because
4166 these RPCs are unnecessary.
4168 Severity : enhancement
4170 Description: allow number of MDS service threads to be changed at module load
4171 Details : It is now possible to change the number of MDS service threads
4172 running. Adding "options mds mds_num_threads={N}" to the MDS's
4173 /etc/modprobe.conf will set the number of threads for the next
4174 time Lustre is restarted (assuming the "mds" module is also
4175 reloaded at that time). The default number of threads will
4176 stay the same, 32 for most systems.
4181 Description: OST crash if filesystem is unformatted or corrupt
4182 Details : If an OST is started on a device that has never been formatted
4183 or if the filesystem is corrupt and cannot even mount then the
4184 error handling cleanup routines would dereference a NULL pointer.
4189 Description: NULL pointer deref in llap_from_page.
4190 Details : get_cache_page_nowait can return a page with NULL (or otherwise
4191 incorrect) mapping if the page was truncated/reclaimed while it was
4192 searched for. Check for this condition and skip such pages when
4193 doing readahead. Introduce extra check to llap_from_page() to
4194 verify page->mapping->host is non-NULL (so page is not anonymous).
4197 Frequency : Sometimes when using sys_sendfile
4199 Description: "page not covered by a lock" warnings from ll_readpage
4200 Details : sendfile called ll_readpage without right page locks present.
4201 Now we introduced ll_file_sendfile that does necessary locking
4202 around call to generic_file_sendfile() much like we do in
4206 Frequency : with certain MDS communication failures at client mount time
4208 Description: NULL pointer deref after failed client mount
4209 Details : a client connection request may delayed by the network layer
4210 and not be sent until after the PTLRPC layer has timed out the
4211 request. If the client fails the mount immediately it will try
4212 to clean up before the network times out the request. Add a
4213 reference from the request import to the obd device and delay
4214 the cleanup until the network drops the request.
4217 Frequency : occasionally during client (re)connect
4219 Description: assertion failure during client (re)connect
4220 Details : processing a client connection request may be delayed by the
4221 client or server longer than the client connect timeout. This
4222 causes the client to resend the connection request. If the
4223 original connection request is replied in this interval, the
4224 client may trip an assertion failure in ptlrpc_connect_interpret()
4225 which thought it would be the only running connect process.
4228 Frequency : only with obd_echo servers and clients that are rebooted
4230 Description: kernel BUG accessing uninitialized data structure
4231 Details : When running an obd_echo server it did not start the ping_evictor
4232 thread, and when a client was evicted an uninitialized data
4233 structure was accessed. Start the ping_evictor in the RPC
4234 service startup instead of the OBD startup.
4236 Severity : enhancement
4237 Bugzilla : 10193 (patchless)
4238 Description: Remove dependency on various unexported kernel interfaces.
4239 Details : No longer need reparent_to_init, exit_mm, exit_files,
4240 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
4243 Frequency : rare (only users of deprecated and unsupported LDAP config)
4245 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
4246 Details : LDAP apparently contains 'lustreName' attributes instead of
4247 'name'. A simple remapping of the name is sufficient.
4250 Frequency : rare (only with non-default dump_on_timeout debug enabled)
4252 Description: waiting_locks_callback trips kernel BUG if client is evicted
4253 Details : Running with the dump_on_timeout debug flag turned on makes
4254 it possible that the waiting_locks_callback() can try to dump
4255 the Lustre kernel debug logs from an interrupt handler. Defer
4256 this log dumping to the expired_lock_main() thread.
4258 Severity : enhancement
4260 Description: Support NFS exporting on 2.6 kernels.
4261 Details : Implement non-rawops metadata methods for NFS server to use without
4262 changing NFS server code.
4265 Frequency : very rare (synthetic metadata workload only)
4267 Description: two racing renames might cause an MDS thread to deadlock
4268 Details : Running the "racer" program may cause one MDS thread to rename
4269 a file from being the source of a rename to being the target of
4270 a rename at exactly the same time that another thread is doing
4271 so, and the second thread has already enqueued these locks after
4272 doing a lookup of the target and is trying to relock them in
4273 order. Ensure that we don't try to re-lock the same resource.
4276 Frequency : only very large systems with liblustre clients
4278 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
4279 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
4280 all OSTs in parallel. This allows the eviction of stale liblustre
4281 clients to proceed much faster than if they were done in series,
4282 and also offers similar improvements for other set_info RPCs.
4287 Description: excessive CPU usage during initial read phase on client
4288 Details : During the initial read phase on a client, it would agressively
4289 retry readahead on the file, consuming too much CPU and impacting
4290 performance (since 1.4.5.8). Improve the readahead algorithm
4291 to avoid this, and also improve some other common cases (read
4292 of small files in particular, where "small" is files smaller than
4293 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
4298 Description: MDS crash when receiving packet with unknown intent.
4299 Details : Do not LBUG in unknown intent case, just return -EFAULT
4301 Severity : enhancement
4302 Bugzilla : 9293, 9385
4303 Description: MDS RPCs are serialised on client. This is unnecessary for some.
4304 Details : Do not serialize getattr (non-intent version) and statfs.
4307 Frequency : occasional, when OST network is overloaded/intermittent
4309 Description: client evicted by OST after bulk IO timeout
4310 Details : If a client sends a bulk IO request (read or write) the OST
4311 may evict the client if it is unresposive to its data GET/PUT
4312 request. This is incorrect if the network is overloaded (takes
4313 too long to transfer the RPC data) or dropped the OST GET/PUT
4314 request. There is no need to evict the client at all, since
4315 the pinger and/or lock callbacks will handle this, and the
4316 client can restart the bulk request.
4319 Frequency : Always when mmapping file with no objects
4321 Description: client crashes when mmapping file with no objects
4322 Details : Check that we actually have objects in a file before doing any
4323 operations on objects in ll_vm_open, ll_vm_close and
4329 Description: Request leak when working with deleted CWD
4330 Details : Introduce advanced request refcount tracking for requests
4331 referenced from lustre intent.
4333 Severity : Enhancement
4335 Description: Cache open file handles on client.
4336 Details : MDS now will return special lock along with openhandle, if
4337 requested and client is allowed to hold openhandle, even if unused,
4338 until such a lock is revoked. Helps NFS a lot, since NFS is opening
4339 closing files for every read/write openration.
4341 Severity : Enhancement
4343 Description: Cache open negative dentries on client when possible.
4344 Details : Guard negative dentries with UPDATE lock on parent dir, drop
4345 negative dentries on lock revocation.
4350 Description: Remounting a client read-only wasn't possible with a zconf mount
4351 Details : It wasn't possible to remount a client read-only with llmount.
4353 Severity : enhancement
4354 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
4355 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
4356 MPICH is included to add Lustre-specific ADIO interfaces.
4357 This is based closely on the UFS ADIO layer and only differs
4358 in file creation, in order to allow the OST striping to be set.
4359 This is user-contributed code and not supported by CFS.
4364 Description: extended inode attributes (immutable, append-only) work improperly
4365 when 2.4 and 2.6 kernels are used on client/server or vice versa
4366 Details : Introduce kernel-independent values for these flags.
4368 Severity : enhancement
4371 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
4372 Details : Many of the /proc/ tunables can only be tuned at a megabyte
4373 granularity. Now, Fractional MB granularity is be supported,
4374 this is very useful for low memory system.
4376 Severity : enhancement
4378 Description: Getattr by fid
4379 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
4380 locks, avoids extra getattr rpc requests to MDS, allows '/' to
4381 have locks and avoids getattr rpc requests for it on every stat.
4384 Frequency : Always, for filesystems larger than 2TB
4386 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4387 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
4388 filesystems larger than 2TB to be mounted with mballoc enabled.
4391 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
4393 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4394 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
4395 filesystems larger than 2TB is mounted then the kernel will
4396 silently corrupt the start of the filesystem. CONFIG_LBD is
4397 enabled for all CFS-supported kernels, but the possibility of
4398 this happening with a modified kernel config exists.
4400 Severity : enhancement
4402 Description: add client O_DIRECT support for 2.6 kernels
4403 Details : It is now possible to do O_DIRECT reads and writes to files
4404 in the Lustre client mountpoint on 2.6 kernel clients.
4406 Severity : enhancement
4408 Description: parallel glimpse, setattr, statfs, punch, destroy requests
4409 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
4410 parallel, not waiting for response from every OST before sending
4411 a rpc to the next OST.
4416 Description: setattr vs write race when updating file timestamps
4417 Details : Client processes that update a file timestamp into the past
4418 right after writing to the file (e.g. tar) it is possible that
4419 the updated file modification time can be reset to the current
4420 time due to a race between processing the setattr and write RPC.
4422 Severity : enhancement
4424 Description: Bring 'lfs find' closer in line with regular Linux find.
4425 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
4426 -print0 options and obtains all the needed info through the lustre
4429 Severity : enhancement
4431 Description: support up to 1024 configured devices on one node
4432 Details : change obd_dev array from statically allocated to dynamically
4433 allocated structs as they are first used to reduce memory usage
4438 Description: Flush dirty partially truncated pages during truncate
4439 Details : Immediatelly flush partially truncated pages in filter_setattr,
4440 this way we completely avoid having any pages in page cache on OST
4441 and can retire ugly workarounds during writes to flush such pages.
4446 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
4447 Details : protect inode from truncation within vfs_unlink() context
4448 just take a reference before calling vfs_unlink() and release it
4449 when parent's i_sem is free.
4454 Description: last_id value checked outside lock on OST caused LASSERT failure
4455 Details : If there were multiple MDS->OST object precreate requests in
4456 flight, it was possible that the OST's last object id was checked
4457 outside a lock and incorrectly tripped an assertion. Move checks
4458 inside locks, and discard old precreate requests.
4461 Frequency : always, if extents are used on OSTs
4463 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
4464 Details : OSTs using the extents format would not zero the high 16 bits of
4465 the index physical block number. This is not a problem for any
4466 OST filesystems smaller than 16TB, and no kernels support ext3
4467 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
4468 new/modified files) and can be fixed for existing filesystems
4469 with e2fsprogs-1.39-cfs1.
4474 Description: import connection selection may be incorrect if timer wraps
4475 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
4476 import connections to be ignored if the 32-bit jiffies counter
4477 wraps. Use a 64-bit jiffies counter.
4480 Frequency : very large clusters immediately after boot
4482 Description: LNET request buffers exhausted under heavy short-term load
4483 Details : If a large number of client requests are generated on a service
4484 that has previously never seen so many requests it is possible
4485 that the request buffer growth cannot keep up with the spike in
4486 demand. Instead of dropping incoming requests, they are held in
4487 the LND until the RPC service can accept more requests.
4490 Frequency : Sometimes during replay
4492 Description: Assertion failure in ll_local_open after replay.
4493 Details : If replay happened on an open request reply before we were able
4494 to set replay handler, reply will become not swabbed tripping the
4495 assertion in ll_local_open. Now we set the handler right after
4496 recognising of open request
4499 Frequency : very rare
4501 Description: kernel reports "badness in vsnprintf"
4502 Details : Reading from the "recovery_status" /proc file in small chunks
4503 may cause a negative length in lprocfs_obd_rd_recovery_status()
4504 call to vsnprintf() (which is otherwise harmless). Exit early
4505 if there is no more space in the output buffer.
4507 Severity : enhancement
4509 Description: clear OBD RPC statistics by writing to them
4510 Details : It is now possible to clear the OBD RPC statistics by writing
4511 to the "stats" file.
4516 Description: Client mtime is not the same on different clients after utimes
4517 Details : In some cases, the client was using the utimes() syscall on
4518 a file cached on another node. The clients now validate the
4519 ctime from the MDS + OSTs to determine which one is right.
4524 Description: Inability to activate failout mode
4525 Details : lconf script incorrectly assumed that in pythong string's numeric
4526 value is used in comparisons.
4529 Frequency : always with multiple stripes per file
4531 Description: Inefficient object allocation for mutli-stripe files
4532 Details : When selecting which OSTs to stripe files over, for files with
4533 a stripe count that divides evenly into the number of OSTs,
4534 the MDS is always picking the same starting OST for each file.
4535 Return the OST selection heuristic to the original design.
4540 Description: mount failures may take full timeout to return an error
4541 Details : Under some heavy load conditions it is possible that a
4542 failed mount can wait for the full obd_timeout interval,
4543 possibly several minutes, before reporting an error.
4544 Instead return an error as soon as the status is known.
4546 Frequency : quota enabled and large files being deleted
4548 Description: releasing more than 4GB of quota at once hangs OST
4549 Details : If a user deletes more than 4GB of files on a single OST it
4550 will cause the OST to spin in an infinite loop. Release
4551 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4556 Description: statfs data retrieved from /proc may be stale or zero
4557 Details : When reading per-device statfs data from /proc, in the
4558 {kbytes,files}_{total,free,avail} files, it may appear
4559 as zero or be out of date.
4562 Frequency : systems with MD RAID1 external journal devices
4564 Description: lconf's call to blkid is confused by RAID1 journal devices
4565 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4566 of returning all block devices that match the journal UUID.
4569 Frequency : always, for aggregate stripe size over 4GB
4571 Description: assertion fails when trying to use 4GB stripe size
4572 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
4573 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
4576 Frequency : always on ppc64
4578 Description: the first write on an ext3 filesystem with mballoc got stuck
4579 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4580 perform endianness conversion.
4582 ------------------------------------------------------------------------------
4584 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4586 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
4587 INTEROPERATE with older versions automatically. Please read the
4588 user documentation before upgrading any part of a live system.
4589 * WARNING: Lustre networking configuration changes are required with
4590 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
4593 * Support for kernels:
4594 2.6.9-22.0.2.EL (RHEL 4)
4595 2.6.5-7.244 (SLES 9)
4596 2.6.12.6 vanilla (kernel.org)
4599 Severity : enhancement
4600 Bugzilla : 7981/8208
4601 Description: Introduced Lustre Networking (LNET)
4602 Details : LNET is new networking infrastructure for Lustre, it includes
4603 a reorganized network configuration mode (see the user
4604 documentation for full details) as well as support for routing
4605 between different network fabrics. Lustre Networking Devices
4606 (LNDs) for the supported network fabrics have also been
4607 created for this new infrastructure.
4609 Severity : enhancement
4610 Description: Introduced Access control lists
4611 Details : clients can set ACLs on files and directories in order to have
4612 more fine-grained permissions than the standard Unix UGO+RWX.
4613 The MDS must be started with the "-o acl" mount option.
4615 Severity : enhancement
4616 Description: Introduced filesystem quotas
4617 Details : Administrators may now establish per-user quotas on the
4620 Severity : enhancement
4622 Description: Configuration change for the XT3
4623 The PTLLND is now used to run Lustre over Portals on the XT3
4624 The configure option(s) --with-cray-portals are no longer used.
4625 Rather --with-portals=<path-to-portals-includes> is used to
4626 enable building on the XT3. In addition to enable XT3 specific
4627 features the option --enable-cray-xt3 must be used.
4632 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
4633 Details : A race between allocating a new llcd and re-getting the llcd_lock
4634 allowed another thread to grab newly-allocated llcd.
4636 Severity : enhancement
4638 Description: 2.6 OST async journal commit and locking fix to improve performance
4639 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
4640 2.6 kernels are now async as they already were in 2.4 kernels so
4641 that they can commit concurrently with the network bulk transfer.
4642 For block-allocated files the filter allocation semaphore is held
4643 to avoid filesystem fragmentation during allocation. BKL lock
4644 removed for 2.6 xattr operations where it is no longer needed.
4649 Description: lconf incorrectly determined whether two IP networks could talk
4650 Details : In some more complicated routing and multiple-network
4651 configurations, lconf will avoid trying to make a network
4652 connection to a disjoint part of the IP space. It was doing the
4653 math incorrectly for one set of cases.
4658 Description: Fix for potential infinite loop processing records in an llog.
4659 Details : If an llog record is corrupted/zeroed, it is possible to loop
4660 forever in llog_process(). Validate the llog record length
4661 and skip the remainder of the block on error.
4664 Frequency : occasional (liblustre only)
4666 Description: liblustre could not open files whose last component is a symlink
4667 Details : sysio_path_walk() would incorrectly pass the open intent to
4668 intermediate path components.
4671 Frequency : rare (liblustre only with non-standard tuning)
4672 Bugzilla : 7201 (7350)
4673 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4674 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4675 after it has been released, only do this for referenced locks
4677 Severity : enhancement
4679 Description: specifying an (invalid) directory default stripe_size of -1
4680 would reset the directory default striping
4681 Details : stripe_size -1 was used internally to signal directory stripe
4682 removal, now use "all default" to signal dir stripe removal
4683 as a directory striping of "all default" is not useful
4686 Frequency : common for large clusters running liblustre clients
4688 Description: doing an ls when liblustre clients are running is slow
4689 Details : sending a glimpse AST to a liblustre client waits for every AST
4690 to time out, as liblustre clients will not respond. Since they
4691 cannot cache data we refresh the OST lock LVB from disk instead.
4693 Severity : enhancement
4695 Description: doing an ls at the same time as file IO can be slow
4696 Details : enqueue and other "small" requests can be blocked behind many
4697 large IO requests. Create a new OST IO portal for non-IO
4698 requests so they can be processed faster.
4701 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4703 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4704 created file with mode 000
4705 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4706 allow an ftruncate() as a non-root user to a file with mode 000.
4707 Lustre now respects this flag to disable mode checks when
4708 truncating a file owned by the user
4711 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4714 Description: Revoking locks from clients that went dead or catatonic might take
4716 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4717 cancellation of such locks instant on servers without waiting for
4718 any reply from clients. Clients drops these locks when cancel
4719 notification from server is received without replying.
4722 Frequency : liblustre-only, when liblustre client dies or becomes busy
4724 Description: Doing ls on Linux clients can take a long time with active
4726 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4727 granting such locks to it in the first place if possible. Locks
4728 are taken by proxy on the OST during the read or write and
4729 dropped immediately afterward. Add connect flags handling, do
4730 not grant locks to liblustre clients for glimpse ASTs.
4732 Severity : enhancement
4734 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4735 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4736 sequential reads and behaves badly with applications doing random
4737 reads. Improve it by reading ahead at least read region, and
4738 avoiding excessive large RPC for small reads.
4740 Severity : enhancement
4742 Description: Creating more than 1000 files for a single job may cause a load
4743 imbalance on the OSTs if there are also a large number of OSTs.
4744 Details : qos_prep_create() uses an OST index reseed value that is an
4745 even multiple of the number of available OSTs so that if the
4746 reseed happens in the middle of the object allocation it will
4747 still utilize the OSTs as uniformly as possible.
4752 Description: OST or MDS may oops in ping_evictor_main()
4753 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
4754 but doesn't restart at beginning of obd_exports_timed list
4757 Severity : enhancement
4759 Description: improve by-nid export eviction on the MDS and OST
4760 Details : allow multiple exports with the same NID to be evicted at one
4761 time without re-searching the exports list.
4764 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
4766 Description: MDS may oops in groups_free()
4767 Details : in rare race conditions a newly allocated group_info struct is
4768 freed again, and this can be NULL. The 2.4 compatibility code
4769 for groups_free() checked for a NULL pointer, but 2.6 did not.
4772 Frequency : common for liblustre clients doing little filesystem IO
4773 Bugzilla : 9352, 7313
4774 Description: server may evict liblustre clients accessing contended locks
4775 Details : if a client is granted a lock or receives a completion AST
4776 with a blocking AST already set it would not reply to the AST
4777 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
4780 Frequency : lfs setstripe, only systems with more than 160 OSTs
4782 Description: unable to set striping with a starting offset beyond OST 160
4783 Details : llapi_create_file() incorrectly limited the starting stripe
4784 index to the maximum single-file stripe count.
4787 Frequency : LDAP users only
4789 Description: lconf did not handle in-kernel recovery with LDAP properly
4790 Details : lconf/LustreDB get_refs() is searching the wrong namespace
4792 Severity : enhancement
4794 Description: bind OST threads to NUMA nodes to improve performance
4795 Details : all OST threads are uniformly bound to CPUs on a single NUMA
4796 node and do their allocations there to localize memory access
4798 Severity : enhancement
4800 Description: llmount can determine client NID directly from Myrinet (GM)
4801 Details : the client NID code from gmnalnid was moved directly into
4802 llmount, removing the need to use this or specifying the
4803 client NID explicitly when mounting GM clients with zeroconf
4806 Frequency : if client is started with down MDS
4808 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
4809 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
4810 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
4811 ptlrpc_import_delayed_req() because we didn't check if the
4812 request was interrupted, and we also didn't break out of the
4813 event loop if there was no timeout
4818 Description: data loss during non-page-aligned writes to a single file from
4819 both multiple nodes and multiple threads on one node at same time
4820 Details : updates to KMS and lsm weren't protected by common lock. Resulting
4821 inconsistency led to false short-reads, that were cached and later
4822 used by ->prepare_write() to fill in partially written page,
4823 leading to data loss.
4826 Frequency : always, if lconf --abort_recovery used
4828 Description: lconf --abort_recovery fails with 'Operation not supported'
4829 Details : lconf was attempting to abort recovery on the MDT device and not
4832 Severity : enhancement
4834 Description: remove cleanup logs
4835 Details : replace lconf-generated cleanup logs with lustre internal
4836 cleanup routines. Eliminates the need for client-cleanup and
4839 Severity : enhancement
4841 Description: add support for EAs (user and system) on lustre filesystems
4842 Details : it is now possible to store extended attributes in the Lustre
4843 client filesystem, and with the user_xattr mount option it
4844 is possible to allow users to store EAs on their files also
4846 Severity : enhancement
4848 Description: Add possibility (config option) to show minimal available OST free
4850 Details : When compiled with --enable-mindf configure option, statfs(2)
4851 (and so, df) will return least minimal free space available from
4852 all OSTs as amount of free space on FS, instead of summary of
4853 free spaces of all OSTs.
4855 Severity : enhancement
4857 Description: do not expand extent locks acquired on OST-side
4858 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
4859 by server: they are not cached anyway.
4862 Frequency : when mmap is used/binaries executed from Lustre
4864 Description: Unmmap pages before throwing them away from read cache.
4865 Details : llap_shrink cache now attempts to unmap pages before discarding
4866 them (if unmapping failed - do not discard). SLES9 kernel has
4867 extra checks that trigger if this unmapping is not done first.
4872 Description: lconf didn't resolve symlinks before checking to see whether a
4873 given mountpoint was already in use
4876 Frequency : when migrating failover services
4877 Bugzilla : 6395, 9514
4878 Description: When migrating a subset of services from a node (e.g. failback
4879 from a failover service node) the remaining services would
4880 time out and evict clients.
4881 Details : lconf --force (implied by --failover) sets the global obd_timeout
4882 to 5 seconds in order to quickly disconnect, but this caused
4883 other RPCs to time out too quickly. Do not change the global
4884 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
4886 Severity : enhancement
4887 Frequency : if MDS is started with down OST
4888 Bugzilla : 9439,5706
4889 Description: Allow startup/shutdown of an MDS without depending on the
4890 availability of the OSTs.
4891 Details : Asynchronously call mds_lov_synchronize during MDS startup.
4892 Add appropriate locking and lov-osc refcounts for safe
4893 cleaning. Add osc abort_inflight calls in case the
4894 synchronize never started.
4897 Frequency : occasional (Cray XT3 only)
4899 Description: root not authorized to access files in CRAY_PORTALS environment
4900 Details : The client process capabilities were not honoured on the MDS in
4901 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
4902 been accessed by an authorized user then root was able to access
4903 the file on the local client also. The root user capabilities
4904 are now allowed on the MDS, as this environment has secure UID.
4907 Frequency : occasional
4909 Description: ldiskfs "too long searching" message happens too often
4910 Details : A debugging message (otherwise harmless) prints too often on
4911 the OST console. This has been reduced to only happen when
4912 there are fragmentation problems on the filesystem.
4917 Description: Division by zero in statfs when all OSCs are inactive
4918 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
4919 lov_statfs divides by value returned by lov_get_stripecnt().
4923 Bugzilla : 9489, 3273
4924 Description: First write from each client to each OST was only 4kB in size,
4925 to initialize client writeback cache, which caused sub-optimal
4926 RPCs and poor layout on disk for the first written file.
4927 Details : Clients now request an initial cache grant at (re)connect time
4928 and so that they can start streaming writes to the cache right
4929 away and always do full-sized RPCs if there is enough data.
4930 If the OST is rebooted the client also re-establishes its grant
4931 so that client cached writes will be honoured under the grant.
4936 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
4937 Details : Now I/O RPCs go to different portal number and (presumably) fast
4938 lock requests (and glimses) and other RPCs get their own service
4939 threads pool that should be able to service those RPCs
4942 Severity : enhancement
4944 Description: Ability to exchange lustre version between client and servers and
4945 issue warnings at client side if client is too old. Also for
4946 liblustre clients there is ability to refuse connection of too old
4948 Details : New 'version' field is added to connect data structure that is
4949 filled with version info. That info is later checked by server and
4953 Frequency : rare, liblustre only.
4954 Bugzilla : 9296, 9581
4955 Description: Two simultaneous writes from liblustre at offset within same page
4956 might proceed at the same time overwriting eachother with stale
4958 Details : I/O lock withing llu_file_prwv was released too early, before data
4959 actually was hitting the wire. Extended lock-holding time until
4960 server acknowledges receiving data.
4963 Frequency : extremely rare. Never observed in practice.
4965 Description: avoid generating lustre_handle cookie of 0.
4966 Details : class_handle_hash() generates handle cookies by incrementing
4967 global counter, and can hit 0 occasionaly (this is unlikely, but
4968 not impossible, because initial value of cookie counter is
4969 selected randonly). Value of 0 is used as a sentinel meaning
4970 "unassigned handle" --- avoid it. Also coalesce two critical
4971 sections in this function into one.
4973 Severity : enhancement
4975 Description: allow liblustre clients to delegate truncate locking to OST
4976 Details : To avoid overhead of locking, liblustre client instructs OST to
4977 take extent lock in ost_punch() on client's behalf. New connection
4978 flag is added to handle backward compatibility.
4980 Severity : enhancement
4981 Bugzilla : 4928, 7341, 9758
4982 Description: allow number of OST service threads to be specified
4983 Details : a module parameter allows the number of OST service threads
4984 to be specified via "options ost ost_num_threads={N}" in the
4985 OSS's /etc/modules.conf or /etc/modprobe.conf.
4989 Bugzilla : 6146, 9635, 9895
4990 Description: servers crash with bad pointer in target_handle_connect()
4991 Details : In rare cases when a client is reconnecting it was possible that
4992 the connection request was the last reference for that export.
4993 We would temporarily drop the export reference and get a new
4994 one, but this may have been the last reference and the export
4995 was just destroyed. Get new reference before dropping old one.
4997 Severity : enhancement
4998 Frequency : if client is started with failover MDS
5000 Description: Allow multiple MDS hostnames in the mount command
5001 Details : Try to read the configuration from all specified MDS
5002 hostnames during a client mount in case the "primary"
5005 Severity : enhancement
5007 Description: Stop sending data to evicted clients as soon as possible.
5008 Details : Check if the client we are about to send or are sending data to
5009 was evicted already. (Check is done every second of waiting,
5010 for which l_wait_event interface was extended to allow checking
5011 of exit condition at specified intervals).
5014 Frequency : rare, normally only when NFS exporting is done from client
5016 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
5018 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
5021 Frequency : always, if srand() is called before liblustre initialization
5023 Description: Liblustre uses system PRNG disturbing its usage by user application
5024 Details : Introduce internal to lustre fast and high-quality PRNG for
5025 lustre usage and make liblustre and some other places in generic
5026 lustre code to use it.
5028 Severity : enhancement
5029 Bugzilla : 9477, 9557, 9870
5030 Description: Verify that the MDS configuration logs are updated when xml is
5031 Details : Check if the .xml configuration logs are newer than the config
5032 logs stored on the MDS and report an error if this is the case.
5033 Request --write-conf, or allow starting with --old_conf.
5035 Severity : enhancement
5037 Description: Handle symlinks in the path when checking if Lustre is mounted.
5038 Details : Resolve intermediate symlinks when checking if a client has
5039 mounted a filesystem to avoid duplicate client mounts.
5044 Description: lconf can hit an error exception but still return success.
5045 Details : The lconf command catches the Command error exception at the top
5046 level script context and will exit with the associated exit
5047 status, but doesn't ensure that this exit status is non-zero.
5052 Description: failure of ptlrpc thread startup can cause oops
5053 Details : Starting a ptlrpc service thread can fail if there are a large
5054 number of threads or the server memory is very fragmented.
5055 Handle this without oopsing.
5058 Frequency : always, only if liblustre and non-default acceptor port was used
5060 Description: liblustre cannot connect to servers with non-default acceptor port
5061 Details : tcpnal_set_default_params() was not called and was therefore
5062 ignoring the environment varaible TCPNAL_PORT, as well as other
5063 TCPNAL_ environment variables
5068 Description: two objects could be created on the same OST for a single file
5069 Details : If an OST is down, in some cases it was possible to create two
5070 objects on a single OST for a single file. No problems other
5071 than potential performance impact and spurious error messages.
5075 Bugzilla : 5681, 9562
5076 Description: Client may oops in ll_unhash_aliases
5077 Details : Client dcache may become inconsistent in race condition.
5078 In some cases "getcwd" can fail if the current directory is
5084 Description: Inode refcounting problems in NFS export code
5085 Details : link_raw functions used to call d_instantiate without obtaining
5086 extra inode reference first.
5090 Bugzilla : 9942, 9903
5091 Description: Referencing freed requests leading to crash, memleaks with NFS.
5092 Details : We used to require that call to ll_revalidate_it was always
5093 followed by ll_lookup_it. Also with revalidate_special() it is
5094 possible to call ll_revalidate_it() twice for the same dentry
5095 even if first occurence returned success. This fix changes semantic
5096 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
5097 reference on a request referred from the intent.
5098 ll_intent_release() then releases such a request.
5101 Frequency : rare, normally benchmark loads only
5103 Description: unlinked inodes were kept in memory on the client
5104 Details : If a client is repeatedly creating and unlinking files it
5105 can accumulate a lot of stale inodes in the inode slab cache.
5106 If there is no other client load running this can cause the
5107 client node to run out of memory. Instead flush old inodes
5108 from client cache that have the same inode number as a new inode.
5111 Frequency : SLES9 2.6.5 kernel and long filenames only
5112 Bugzilla : 9969, 10379
5113 Description: utime reports stale NFS file handle
5114 Details : SLES9 uses out-of-dentry names in some cases, which confused
5115 the lustre dentry revalidation. Change it to always use the
5119 Frequency : rare, unless heavy write-truncate concurrency is continuous
5120 Bugzilla : 4180, 6984, 7171, 9963, 9331
5121 Description: OST becomes very slow and/or deadlocked during object unlink
5122 Details : filter_destroy() was holding onto the parent directory lock
5123 while truncating+unlinking objects. For very large objects this
5124 may block other threads for a long time and slow overall OST
5125 responsiveness. It may also be possible to get a lock ordering
5126 deadlock in this case, or run out of journal credits because of
5127 the combined truncate+unlink. Solution is to do object truncate
5128 first in one transaction without parent lock, and then do the
5129 final unlink in a new transaction with the parent lock. This
5130 reduces the lock hold time dramatically.
5133 Frequency : rare, 2.4 kernels only
5135 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
5136 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
5137 before unmounting their filesystem in order to avoid deadlock.
5138 The kernel_locked() function in 2.4 kernels only checks whether
5139 the kernel lock is held, not whether it is this process that is
5140 holding it as 2.6 kernels do.
5145 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
5146 Details : The client ptlrpc code may be trying to reconnect to a down
5147 server before a previous connection attempt has timed out.
5148 Increase the reconnect interval to be longer than the connection
5149 timeout interval to avoid sending duplicate connections to
5153 Frequency : echo_client brw_test command
5155 Description: fix echo_client to work with OST preallocated code
5156 Details : OST preallocation code (5137) didn't take echo_client IO path
5157 into account: echo_client calls filter methods outside of any
5158 OST thread and, hence, there is no per-thread preallocated
5159 pages and buffers to use. Solution: hijack pga pages for IO. As
5160 a byproduct, this avoids unnecessary data copying.
5164 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
5165 Description: Client can oops in mdc_commit_close() after open replay
5166 Details : It was possible for the MDS to return an open request with no
5167 transaction number in mds_finish_transno() if the client was
5168 evicted, but without actually returning an error. Clients
5169 would later try to replay that open and may trip an assertion
5170 Simplify the client close codepath, and always return an error
5171 from the MDS in case the open is not successful.
5174 Frequency : rare, 2.6 OSTs only
5176 Description: OST may deadlock under high load on fragmented files
5177 Details : If there was a heavy load and highly-fragmented OST filesystems
5178 it was possible to have all the OST threads deadlock waiting on
5179 allocation of biovecs, because the biovecs were not released
5180 until the entire RPC IO was completed. Instead, release biovecs
5181 as soon as they are complete to ensure forward IO progress.
5183 Severity : enhancement
5185 Description: Support for specifying external journal device at mount
5186 Details : If an OST or MDS device is formatted with an external journal
5187 device, this device major/minor is stored in the ext3 superblock
5188 and may not be valid for failover. Allow detecting and
5189 specifying the external journal at mount time.
5194 Description: Mounting an MDS with pending unlinked files may cause oops
5195 Details : target_finish_recovery() calls mds_postrecov() which returned
5196 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
5197 considers this an error and immediately begins cleaning up the
5198 lov, just after starting the mds_lov process
5200 Severity : enhancement
5202 Description: Implement 'lfs df' to report actual free space on per-OST basis
5203 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
5204 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
5205 usage of objects; '-h' to report in human readable format.
5207 ------------------------------------------------------------------------------
5209 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
5216 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
5217 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
5218 at startup during orphan unlinking. Instead, skip buddy update
5219 before setup, as it will be regenerated after recovery is complete.
5224 Description: If an OST is inactive, its locks might reference stale inodes.
5225 Details : lov_change_cbdata() must iterate over all namespaces, even if
5226 they are inactive to clear inode references from the lock.
5228 Severity : enhancement
5229 Frequency : occasional, if non-standard max_dirty_mb used
5231 Description: Client will block write RPCs if not enough grant
5232 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
5233 then the client will block writes while waiting for another RPC
5234 to complete instead of consuming its dirty limit. With change
5235 we get improved performance when max_dirty_mb is small.
5237 Severity : enhancement
5238 Bugzilla : 3389, 6253
5239 Description: Add support for supplementary groups on the MDS.
5240 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
5241 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
5242 lookups for user supplementary groups into a cache.
5246 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
5247 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
5248 for liblustre O_EXCL is enforced by MDS.
5252 Description: readdir never returns NULL in liblustre.
5253 Details : Corrected llu_iop_getdirentries logic, to return offset of next
5254 dentry in struct dirent.
5258 Frequency : liblustre only, depends on application IO pattern
5259 Description: liblustre clients evicted if not contacting servers
5260 Details : Don't put liblustre clients into the ping_evictor list, so
5261 they will not be evicted by the pinger ever.
5263 Severity : enhancement
5265 Description: Add ability to evict clients by NID from MDS.
5266 Details : By echoing "nid:$NID" string into
5267 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
5268 $NID would be instantly evicted from this MDS and from all active
5269 OSTs connected to it.
5273 Description: Do not query file size twice, somewhat slowing stat(2) calls.
5274 Details : lookup_it_finish() used to query file size from OSTs that was not
5279 Description: service threads change working directory to that of init
5280 Details : Starting lustre service threads may pin the working directory
5281 of the parent thread, making that filesystem busy. Threads
5282 now change to the working directory of init to avoid this.
5286 Frequency : during shutdown only
5287 Description: shutdown with a failed MDS or OST can cause unmount to hang
5288 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
5293 Frequency : 2.6 only
5294 Description: chmod/chown may include an extra supplementary group
5295 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
5296 supplementary group and if none is specified this is used.
5299 Bugzilla : 5479 (6816)
5301 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
5302 Details : If lookup is in progress on a file that is unlinked we might try
5303 to revalidate the inode and fail in revalidate after lookup is
5304 complete and ll_file_open() enqueues the open again but
5305 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
5308 Frequency : always, if lconf --abort_recovery used
5310 Description: lconf --abort_recovery fails with 'Operation not supported'
5311 Details : lconf was attempting to abort recovery on the MDT device and not
5314 ------------------------------------------------------------------------------
5316 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
5321 Frequency : rare (only unsupported configurations with a node running as an
5323 Bugzilla : 6514, 5137
5324 Description: Mounting a Lustre file system on a node running as an OST could
5326 Details : OSTs now preallocates memory needed to write out data at
5327 startup, instead of when needed, to avoid having to
5328 allocate memory in possibly low memory situations.
5329 Specifically, if the file system is mounted on on OST,
5330 memory pressure could force it to try to write out data,
5331 which it needed to allocate memory to do. Due to the low
5332 memory, it would be unable to do so and the node would
5333 become unresponsive.
5335 Severity : enhancement
5337 Description: Addition of lconf --service command line option
5338 Details : lconf now accepts a '--service <arg>' option, which is
5339 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
5341 Severity : enhancement
5343 Description: Failover mode is now the default for OSTs.
5344 Details : By default, OSTs will now run in failover mode. To return to
5345 the old behaviour, add '--failout' to the lmc line for OSTs.
5347 Severity : enhancement
5349 Description: Health checks are now provided for MDS and OSTs
5350 Details : Additional detailed health check information on MSD and OSTs
5351 is now provided through the procfs health_check value.
5354 Frequency : occasional, depends on IO load
5356 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
5357 after numerous create/delete cycles
5358 Details : The ext3 inode allocation policy would not allocate new inodes
5359 very well on the OSTs because there are no new directories
5360 being created. Instead we look for groups with free space if
5361 the parent directories are nearly full.
5366 Description: Network or server problems during mount may cause partially
5367 mounted clients instead of returning an error.
5368 Details : The config llog parsing code may overwrite the error return
5369 code during mount error handling, returning success instead
5375 Description: MDS can fail to allocate large reply buffers
5376 Details : After long uptimes the MDS can fail to allocate large reply
5377 buffers (e.g. zconf client mount config records) due to memory
5378 fragmentation or consumption by the buffer cache. Preallocate
5379 some large reply buffers so that these replies can be sent even
5380 under memory pressure.
5384 Frequency : rare (liblustre)
5385 Description: fsx running with liblustre complained that using truncate() to
5386 extend the file doesn't work. This patch corrects that issue.
5387 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
5388 ATTR_SIZE and lsm use in llu_setattr_raw.
5392 Frequency : rare, only 2.6 kernels
5393 Description: Unusual file access patterns on the MDS may result in inode
5394 data being lost in very rare circumstances.
5395 Details : Bad interaction between the ea-in-inode patch and the "no-read"
5396 code in the 2.6 kernel caused the inode and/or EA data not to
5397 be read from disk, causing single-file corruption.
5401 Frequency : rare, only 2.6 filesystems using extents
5402 Description: Heavy concurrent write and delete load may cause data corruption.
5403 Details : It was possible under high-load situations to have an extent
5404 metadata block in the block device cache from a just-unlinked
5405 file overwrite a newly-allocated data block. We now unmap any
5406 metadata buffers that alias just-allocated data blocks.
5410 Frequency : filesystems with default stripe_count larger than 77
5411 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
5412 Details : lconf specifies an inode size of 4096 bytes when the default
5413 stripe_count is larger than 77. This conflicts with the default
5414 inode density of 1 per 4096 bytes. Allocate smaller inodes in
5415 this case to avoid pinning too much memory for large EAs.
5417 ------------------------------------------------------------------------------
5419 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
5424 Frequency : rare (extremely heavy IO load with hundreds of clients)
5426 Description: Client is evicted, gets IO error writing to file
5427 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
5428 caused clients to be evicted for AST timeouts. The fixes in
5429 bug 5192 mean we no longer need to have such short AST timeouts
5430 so ldlm_timeout has been increased.
5433 Frequency : occasional during --force or --failover shutdown under load
5434 Bugzilla : 5949, 4834
5435 Description: Server oops/LBUG if stopped with --force or --failover under load
5436 Details : a collection of import/export refcount and cleanup ordering
5437 issues fixed for safer force cleanup
5440 Frequency : only filesystems larger than 120 OSTs
5441 Bugzilla : 5990, 6223
5442 Description: lfs getstripe would oops on a very large filesystem
5443 Details : lov_getconfig used kfree on vmalloc'd memory
5446 Frequency : only filesystems exporting via NFS to Solaris 10 clients
5447 Bugzilla : 6242, 6243
5448 Description: reading from files that had been truncated to a non-zero size
5449 but never opened returned no data
5450 Details : ll_file_read() reads zeros from no-object files to EOF
5455 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
5456 Details : There's little harm in aborting MDS/OSS recovery and letting it
5457 try again, so I removed the LASSERT and return an error instead.
5459 Severity : enhancement
5461 Description: New debugging infrastructure for tracking down data corruption
5462 Details : The I/O checksum code was replaced to: (a) control it at runtime,
5463 (b) cover more of the client-side code path, and (c) try to narrow
5464 down where problems occurred
5468 Bugzilla : 3819, 4364, 4397, 6313
5469 Description: Racing close and eviction MDS could cause assertion in mds_close
5470 Details : It was possible to get multiple mfd references during close and
5471 client eviction, leading to one thread referencing a freed mfd.
5473 Severity: : enhancement
5474 Bugzilla : 3262, 6359
5475 Description: Attempts to reconnect to servers are now more aggressive.
5476 Details : This builds on the enhanced upcall-less recovery that was added
5477 in 1.4.2. When trying to reconnect to servers, clients will
5478 now try each server in the failover group every 10 seconds. By
5479 default, clients would previously try one server every 25 seconds.
5484 Description: After recovery, certain operations trigger a failed
5485 assertion on a client.
5486 Details : Failing over an mds, using lconf -d --failover, while a
5487 client was doing a readdir() call would cause the client to
5488 LBUG after recovery completed and the readdir() was resent.
5490 Severity : enhancement
5492 Description: Default groups are now added by lconf
5493 Details : You can now run lconf --group <servicename> without having to
5494 manually add groups with lmc.
5497 Frequency : occasional
5499 Description: Nodes with an elan id of 0 trigger a failed assertion
5502 Frequency : always when accessing e.g. tty/console device nodes
5504 Description: tty and some other devices nodes cannot be used on lustre
5505 Details : file's private_data field is used by device data and lustre
5506 values in there got lost. New field was added to struct file to
5507 store fs-specific private data.
5510 Frequency : when exporting Lustre via NFS
5512 Description: NFSD failed occasionally when looking up a path component
5513 Details : NFSD is looking up ".." which was broken in ext3 directories
5514 that had grown large enough to become hashed.
5517 Frequency : Clusters with multiple interfaces not on the same subnet
5519 Description: Nodes will repeatedly try to reconnect to an interface which it
5520 cannot reach and report an error to the log.
5521 Details : Extra peer list entries will be created by lconf with some peers
5522 unreachable. lconf now validates the peer before adding it.
5525 Frequency : Only if a default stripe is set on the filesystem root.
5527 Description: Setting a default stripe on the filesystem root prevented the
5528 filesystem from being remounted.
5529 Details : The client was sending extra request flags in the root getattr
5530 request and did not allocate a reply buffer for the dir EA.
5533 Frequency : occasional, higher if lots of files are accessed by one client
5534 Bugzilla : 6159, 6097
5535 Description: Client trips assertion regarding lsm mismatch/magic
5536 Details : While revalidating inodes the VFS looks up inodes with ifind()
5537 and in rare cases can find an inode that is being freed.
5538 The ll_test_inode() code will free the lsm during ifind()
5539 when it finds an existing inode and then the VFS later attaches
5540 this free lsm to a new inode.
5544 Bugzilla : 6422, 7030
5545 Description: MDS deadlock between mkdir and client eviction
5546 Details : Creating a new file via mkdir or mknod (starting a transaction
5547 and getting the ns lock) can deadlock with client eviction
5548 (gets ns lock and trying to finish a synchronous transaction).
5551 Frequency : occasional
5552 Description: While starting a server, the fsfilt_ext3 module could not be
5554 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
5555 kernels. Previously, lconf would still use the ext3 name
5556 when trying to load modules. Now, it will correctly use
5557 ext3 on 2.4 and ldiskfs on 2.6.
5559 Severity : enhancement
5560 Description: The default stripe count has been changed to 1
5561 Details : The interpretation of the default stripe count (0, to lfs
5562 or lmc) has been changed to mean striping across a single
5563 OST, rather than all available. For general usage we have
5564 found a stripe count of 1 or 2 works best.
5566 Severity : enhancement
5567 Description: Add support for compiling against Cray portals.
5568 Details : Conditional compiling for some areas that are different
5572 Frequency : occasional
5573 Bugzilla : 6409, 6834
5574 Description: Creating files with an explicit stripe count may lead to
5575 a failed assertion on the MDS
5576 Details : If some OSTs are full or unavailable, creating files may
5577 trigger a failed assertion on the MDS. Now, Lustre will
5578 try to use other servers or return an error to the
5582 Frequency : occasional
5584 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
5585 caused lock timeout during readahead (since 1.4.2).
5586 Details : Processes doing readahead might match a lock that hasn't been
5587 granted yet if there are overlapping and conflicting lock
5588 requests. The readahead process waits on ungranted lock
5589 (original lock is CBPENDING), while OST waits for that process
5590 to cancel CBPENDING read lock and eventually evicts client.
5592 Severity : enhancement
5594 Description: Initial enabling of flock support for clients
5595 Details : Implements fcntl advisory locking and file status functions.
5596 This feature is provided as an optional mount flag (default
5597 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
5598 locking are implemented yet, and those that are are not guaranteed
5599 to be completely correct in production environments.
5600 mount -t lustre -o [flock|noflock] ...
5603 Frequency : occasional
5605 Description: OSTs running 2.4 kernels but with extents enabled might trip an
5606 assertion in the ext3 JBD (journaling) layer.
5607 Details : The b_committed_data struct is protected by the big kernel lock
5608 in 2.4 kernels, serializing journal_commit_transaction() and
5609 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
5610 access to this struct. In 2.6 kernels there is finer grained
5611 locking to improve SMP performance of the JBD layer.
5615 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
5620 Description: merge and cleanup kernel patches.
5623 -----------------------------------------------------------------------------
5625 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
5627 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
5628 versions of Lustre. Please update all servers and clients to
5629 version 1.4.2 or later at the same time. You must also run
5630 "lconf --write-conf {config}.xml" on the MDS while it is stopped
5631 to update the configuration logs.
5633 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
5634 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
5635 - swab llog records of type '0' so we get proper header size/idx (5861)
5636 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
5637 - fix rename of one directory over another leaking an inode (5953)
5638 - avoid SetPageDirty on 2.6 (5981)
5639 - don't re-add just-being-destroyed locks to the waiting list (5653)
5640 - when creating new directories, inherit the parent's custom
5641 striping settings if present parent (3048)
5642 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
5643 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
5644 - don't hold client locks on temporary worklist from l_lru (5666)
5645 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
5646 - automatically evict dead clients (5921)
5647 - Update file size properly in create+truncate+fstat case (6196)
5648 - Do not unhash mountpoint dentries, do not allow removal of
5650 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
5651 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
5652 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
5653 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
5654 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
5655 - Correctly handle reads of files with no objects (6243)
5656 - lctl recover will also mark a device active if deactivate used (5933)
5658 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
5659 - allow --write-conf on an MDS with different nettype than client (5619)
5660 - don't write config llogs to MDS for mounts not from that MDS (5617)
5661 - lconf should create multiple TCP connections from a client (5201)
5662 - init scripts are now turned off by default; run chkconfig --on
5663 lustre and chkconfig --on lustrefs to use them
5664 - upcalls are no longer needed for clients to recover to failover
5666 - add --abort-recovery option to lconf to abort recovery on device
5668 - add support for an arbitrary number of OSTs (3026)
5669 - Quota support protocol changes.
5670 - forward compatibility changes to wire structs (6007)
5671 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5672 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5673 - improved "lustre status" script
5674 - initialize blocksize for non-regular files (6062)
5675 - added --disable-server and --disable-client configure options (5782)
5676 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5677 - Vanilla 2.4.29 support
5678 - increase maximum number of obd devices to 520 (6242)
5679 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5680 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5681 SLES 9 2.6.5 kernels
5682 - limit stripes per file to 160 (the maximum EA size) (6093)
5684 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5687 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5688 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5689 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5690 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5691 - small ext3 extents cleanups and fixes (5733)
5692 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5693 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5694 - avoid needless client->OST connect, fix handle mismatch (5317)
5695 - fix DLM error path that led to out-of-sync client, long delays (5779)
5696 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5697 - fix several locking issues related to i_size (5492,5624,5654,5672)
5698 - don't move pending lock onto export if it is already evicted (5683)
5699 - fix kernel oops when creating .foo in unlinked directory (5548)
5700 - fix deadlock in obdfilter statistics vs. object create (5811)
5701 - use time_{before,after} to avoid timer jiffies wrap (5882)
5702 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5703 - Do not leak request if server was not able to process it (5154)
5704 - If mds_open unable to find parent dir, make that negative lookup(5154)
5705 - don't create new directories with extent-mapping (5909, 5936)
5707 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5708 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5709 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5710 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5711 - support for RHEL4 kernel on i686 (5773)
5712 - provide error messages when incompatible logs are encountered (5898)
5714 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5715 * version 1.4.0.10 (1.4.1 release candidate 1)
5717 - don't keep a lock reference when lock is not granted (4238)
5718 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5719 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5720 - drop import inflight refcount on signal_completed_replay error (5255)
5721 - unlock page after async write error during send (3677)
5722 - handle missing objects in filter_preprw_read properly (5265)
5723 - no transno return for symlink open, don't save no-trasno open (3440)
5724 - don't try to complete elan receive that already failed (4012)
5725 - free RPC server reply state on error (5406)
5726 - clean up thread from ptlrpc_start_thread() on error (5160)
5727 - readahead could read extra page into cache that wasn't ejected (5388)
5728 - prevent races in class_attach/setup/cleanup/detach (5260)
5729 - don't dereference de->d_inode after l_dput of de (5458)
5730 - use "int" for stripe value returned from lock_to_stripe (5544)
5731 - mballoc allocation and error-checking fixes in 2.6 (5504)
5732 - block device patches to fix I/O request sizes in 2.6 (5482)
5733 - look up hostnames for IB nals (5602)
5734 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5735 - don't start multiple acceptors for the same port (5277)
5736 - fix incorrect LASSERT in mds_getattr_name (5635)
5737 - export a proc file for general "ping" checking (5628)
5738 - fix "lfs check" to not block when the MDS is down (5628)
5740 - service request history (4965)
5741 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5742 - create an "evict_client" /proc entry on OSTs, like the MDS has
5743 - fix mount usage message, return errors per mount(8) (5168)
5744 - change grep [] to grep "[]" in tests so they work in more UMLs
5745 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5746 - remove ancient LOV_MAGIC_V0 EA support (5047)
5747 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5748 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5749 - allow readahead allocations to fail when low on memory (5383)
5750 - mmap locking landed again, after considerable improvement (2828)
5751 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
5753 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
5756 - send OST transaction number in read/write reply to free req (4966)
5757 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5758 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5759 - the watchdog thread now runs as interruptible (5246)
5760 - flock/lockf fixes (but it's still disabled, pending 5135)
5761 - don't use EXT3 constants in llite code (5094)
5762 - memory shortage at startup could cause assertion (5176)
5764 - reorganization of lov code
5765 - single portals codebase
5767 - add extents/mballoc support (5025)
5768 - direct I/O reads in the obdfilter (4048)
5769 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
5771 tbd Cluster File Systems, Inc. <info@clusterfs.com>
5774 - send OST transaction number in read/write reply to free req (4966)
5775 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5776 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5777 - the watchdog thread now runs as interruptible (5246)
5778 - handle missing objects in filter_preprw_read properly (5265)
5779 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5780 - drop import inflight refcount on signal_completed_replay error (5255)
5781 - unlock page after async write error during send (3677)
5782 - return original error code on reconstructed replies (3761)
5783 - no transno return for symlink open, don't save no-trasno open (3440)
5785 - add pid to ldlm debugging output (4922)
5786 - bump the watchdog timeouts -- we can't handle 30sec yet
5787 - extra debugging for orphan dentry/inode bug (5259)
5789 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
5792 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
5793 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
5794 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
5795 - fix NULL dereference in /proc/sys/portals/routes (4827)
5796 - allow failed mdc_close() operations to be interrupted (4561)
5797 - stop precreate on OST before MDS would time out on it (4778)
5798 - don't send partial-page writes before EOF from client (4410)
5799 - discard client grant for sub-page writes on large-page clients (4520)
5800 - don't free dentries not owned by NFS code, check generation (4806)
5801 - fix lsm leak if mds_create_objects() fails (4801)
5802 - limit debug_daemon file size, always print CERROR messages (4789)
5803 - use transno after validating reply (3892)
5804 - process timed out requests if import state changes (3754)
5805 - update mtime on OST during writes, return in glimpse (4829)
5806 - add mkfsoptions to LDAP (4679)
5807 - use ->max_readahead method instead of zapping global ra (5039)
5808 - don't interrupt __l_wait_event() during strace
5810 - add software watchdogs to catch hung threads quickly (4941)
5811 - make lustrefs init script start after nfs is mounted
5812 - fix CWARN/ERROR duplication (4930)
5813 - return async write errors to application if possible (2248)
5814 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
5815 - print NAL number in %x format (4645)
5816 - update barely-supported suse-2.4.21-171 series (4842)
5817 - support for sles 9 %post scripts
5818 - support for building 2.6 kernel-source packages
5819 - support for sles km_* packages
5821 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
5824 - ignore -ENOENT errors in osc_destroy (3639)
5825 - notify osc create thread that OSC is being cleaned up (4600)
5826 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
5827 - reconstruct ost_handle() like mds_handle() (4657)
5828 - create a new thread to do import eviction to avoid deadlock (3969)
5829 - let lconf resolve symlinked-to devices (4629)
5830 - don't unlink "objects" from directory with default EA (4554)
5831 - hold socknal file ref over connect in case target is down (4394)
5832 - allow more than 32000 subdirectories in a single directory (3244)
5833 - fix blocks count for O_DIRECT writes (3751)
5834 - OST returns ENOSPC from object create when no space left (4539)
5835 - don't send truncate RPC if file size isn't changing (4410)
5836 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
5837 - bind to privileged port in socknal and tcpnal (3689)
5839 - rate limit CERROR/CWARN console message to avoid overload (4519)
5840 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
5841 - basic mmap support (3918)
5842 - kernel patch series update from b1_4 (4711)
5844 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
5847 - avoid crash during MDS cleanup with OST shut down (2775)
5848 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
5849 - don't use bad inodes on the MDS (3744)
5850 - dynamic object preallocation to improve recovery speed (4236)
5851 - don't hold spinlock over lock dumping or change debug flags (4401)
5852 - don't zero obd_dev when it is force cleaned (3651)
5853 - print grants to console if they go negative (4431)
5854 - "lctl deactivate" will stop automatic recovery attempts (3406)
5855 - look for existing locks in ldlm_handle_enqueue() (3764)
5856 - don't resolve lock handle twice in recovery avoiding race (4401)
5857 - revalidate should check working dir is a directory (4134)
5859 - don't always mark "slow" obdfilter messages as errors (4418)
5861 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
5864 - don't close LustreDB during write_conf until it is done (3860)
5865 - fix typo in lconf for_each_profile (3821)
5866 - allow dumping logs from multiple threads at one time (3820)
5867 - don't allow multiple threads in OSC recovery (3812)
5868 - fix debug_size parameters (3864)
5869 - fix mds_postrecov to initialize import for llog ctxt (3121)
5870 - replace config semaphore with spinlock (3306)
5871 - be sure to send a reply for a CANCEL rpc with bad export (3863)
5872 - don't allow enqueue to complete on a destroyed export (3822)
5873 - down write_lock before checking llog header bitmap (3825)
5874 - recover from lock replay timeout (3764)
5875 - up llog sem before sending rpc (3652)
5876 - reduce ns lock hold times when setting kms (3267)
5877 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
5878 - fix NULL deref and obd_dev leak on setup error (3312)
5879 - replace some LBUG about llog ops with error handling (3841)
5880 - don't match INVALID dentries from d_lookup and spin (3784)
5881 - hold dcache_lock while marking dentries INVALID and hashing (4255)
5882 - fix invalid assertion in ptlrpc_set_wait (3880)
5884 - add libwrap support for the TCP acceptor (3996)
5885 - add /proc/sys/portals/routes for non-root route listing (3994)
5886 - allow setting MDS UUID in .xml (2580)
5887 - print the stack of a process that LBUGs (4228)
5889 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
5892 - don't cleanup request in ll_file_open() on failed MDS open (3430)
5893 - make sure to unset replay flag from failed open requests (3440)
5894 - if default stripe count is 0, use OST count for inode size (3636)
5895 - update parent mtime/ctime on client for create/unlink (2611)
5896 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
5897 - free recovery state on server during a forced cleanup (3571)
5898 - unregister_reply for resent reqs (3063)
5899 - loop back devices mounting and status check on 2.6 (3563)
5900 - fix resource-creation race that can provoke i_size == 0 (3513)
5901 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
5902 - more debugging for page-accounting assertion (3746)
5903 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
5904 - avoid deadlock after precreation failure (3758)
5905 - fix race and lock order deadlock in orphan handling (3450, 3750)
5906 - add validity checks when grabbing inodes from l_ast_data (3599)
5908 - add /proc/.../recovery_status to obdfilter (3428)
5909 - lightweight CDEBUG infrastructure, debug daemon (3668)
5910 - change default OSC RPC parameters to be better on small clusters
5911 - turn off OST read cache for files smaller than 32MB
5912 - install man pages and include them in rpms (3100)
5913 - add new init script for (un)mounting lustre filesystems (2593)
5914 - run chkconfig in %post for init scripts (3701)
5915 - drop scimac NAL (unmaintained)
5917 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
5920 - clean kiobufs before and after use (3485)
5921 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
5922 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
5923 - change init script to fail more clearly if not run as root (1528)
5924 - allow clients to reconnect during replay (1742)
5925 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
5926 - don't do DNS lookups on NIDs too small for IP addresses (3442)
5927 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
5928 - fix cond_resched (3554)
5929 - only evict unfinished clients after recovery (3515)
5930 - allow bulk resend, prevent data loss (3570)
5931 - dynamic ptlrpc request buffer allocation (2102)
5932 - don't allow unlinking open directory if it isn't empty (2904)
5933 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
5934 - remove extraneous obd dereference causing LASSERT failure (3334)
5935 - don't use get_cycles() when creating temp. files on the mds (3156)
5936 - hold i_sem when setting i_size in ll_extent_lock() (3564)
5937 - handle EEXIST for set-stripe, set proper directory name (3336)
5939 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
5940 - fix ksocknal_fmb_callback() error messages (2918)
5942 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
5945 - don't copy lvb into (possibly NULL) reply on error (2983)
5946 - don't deref dentry after dput, don't free lvb on error (2922)
5947 - use the kms to determine writeback rpc length (2947)
5948 - increment oti_logcookies when osc is inactive (2948)
5949 - update client's i_blocks count via lvb messages (2543)
5950 - handle intent open/close of special files properly (1557)
5951 - mount MDS with errors=remount-ro, like obdfilter (2009)
5952 - initialize lock handle to avoid ASSERT on error cleanup (3057)
5953 - don't use cancelling-locks' kms values (2947)
5954 - use highest lock extent for kms, not last one (2925)
5955 - don't dereference ERR_PTR() dentry in error handling path (3107)
5956 - fix thread race in portals_debug_dumplog() (3122)
5957 - create lprocfs device entries at setup instead of at attach (1519)
5958 - common AST error handler, don't evict client on completion race (3145)
5959 - zero nameidata in detach_mnt in 2.6 (3118)
5960 - verify d_inode after revalidate_special is valid in 2.6 (3116)
5961 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
5962 - initialize RPC timeout timer earlier for 2.6 (3219)
5963 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
5964 - print nal/nid for unknown nid (3258)
5965 - additional checks for oscc recovery before doing precreate (3284)
5966 - fix ll_extent_lock() error return code for 64-bit systems (3043)
5967 - don't crash in mdc_close for bad permissions on open (3285)
5968 - zero i_rdev for non-device files (3147)
5969 - clear page->private before handing to FS, better assertion (3119)
5970 - tune the read pipeline (3236)
5971 - fix incorrect decref of invalidated dentry (2350)
5972 - provide read-ahead stats and refine rpc in flight stats (3328)
5973 - don't hold journal transaction open across create RPC (3313)
5974 - update atime on MDS at close time (3265)
5975 - close LDAP connection when recovering to avoid server load (3315)
5976 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
5977 - don't leak open file on MDS after open resend (3325)
5978 - serialize filter_precreate and filter_destroy_precreated (3329)
5979 - loop device shouldn't call sync_dev() for nul device (3092)
5980 - clear page cache after eviction (2766)
5981 - resynchronize MDS->OST in background (2824)
5982 - refuse to mount the same filesystem twice on same mountpoint (3394)
5983 - allow llmount to create routes for mounting behind routers (3320)
5984 - push lock cancellation to blocking thread for glimpse ASTs (3409)
5985 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
5986 - fix rare problem with rename on htree directories (3417)
5988 - allow default OST striping configuration per directory (1414)
5989 - fix compilation for qswnal for 2.6 kernels (3125)
5990 - increase maximum number of MDS request buffers for large systems
5991 - change liblustreapi to be useful for external progs like lfsck (3098)
5992 - increase local configuration timeout for slow disks (3353)
5993 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
5995 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5998 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
5999 - don't overwrite extent policy data in reply if lock was blocked (2901)
6000 - drop filter export grants atomically with removal from device (2663)
6001 - del obd_self_export from work_list in class_disconnect_exports (2908)
6002 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
6003 - swab reply message in mdc_close, other PPC fixes (2464)
6004 - fix destroying of named logs (2325)
6005 - overwrite old logs when running lconf --write_conf (2264)
6006 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
6007 - fix race in target_handle_connect (2898)
6008 - mds_reint_create() should take same inode create lock (2926)
6009 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
6010 - don't close files for self_export to avoid uninitialized obd (2936)
6011 - allow MDS with the same name as client node (2939)
6012 - hold dentry reference for closed log files for unlink (2325)
6013 - reserve space for all logs during transactions (2059)
6014 - don't evict page beyond end of stripe extent (2925)
6015 - don't oops on a deleted current working directory (2399)
6016 - handle hard links to targets without a parent properly (2517)
6017 - don't dereference NULL lock when racing during eviction (2867)
6018 - don't grow lock extents when lots of conflicting locks (2919)
6020 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
6023 - account for cache space usage on clients to avoid data loss (974)
6024 - lfsck support in lustre kernel code (2349)
6025 - reduce journal credits needed for BRW writes (2370)
6026 - orphan handling to avoid losing space on client/server crashes
6027 - ptlrpcd can be blocked, stopping ALL progress (2477)
6028 - use lock value blocks to assist in proper KMS, faster stat (1021)
6029 - takes i_sem instead of DLM locks internally on obdfilter (2720)
6030 - recovery for initial connections (2355)
6031 - fixes for mds_cleanup_orphans (1934)
6032 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
6033 - block all file creations until orphan recovery completes (1901)
6034 - client remove rq_connection from request struct (2423)
6035 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
6036 - recovery timer race (2670)
6037 - mdc_close recovey bug (2532)
6038 - ptlrpc cleanup bug (2710)
6039 - mds timeout on local locks (2588)
6040 - namespace lock held during RPCs (2431)
6041 - handle interrupted sync write properly (2503)
6042 - don't try to handle a message that hasn't been replied to (2699)
6043 - client assert failure during cleanup after abort recovery (2701)
6044 - leak mdc device after failed mount (2712)
6045 - ptlrpc_check_set allows timedout requests to complete (2714)
6046 - wait for inflight reqs when ptlrpcd finishes (2710)
6047 - make sure unregistered services are removed from the srv_list
6048 - reset bulk XID's when resending them (caught by 1138 test)
6049 - unregister_bulk after timeout
6050 - fix lconf error (2694)
6051 - handle write after unfinished setstripe, stripe-only getstripe (2388)
6052 - readahead locks pages, leaves pending causing memory pressure (2673)
6053 - increase OST request buffers to 4096 on large machines (2729)
6054 - fix up permission of existing directories in simple_mkdir (2661)
6055 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
6056 - don't assign transno to errored transactions (2742)
6057 - don't delete objects on OST if given a bogus objid from MDS (2751)
6058 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
6059 - if rq_no_resend, then timeout request after recovery (2432)
6060 - fix MDS llog_logid record size, 64-bit array alignment (2733)
6061 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
6062 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
6063 - MDS recovery shouldn't delete objects that it has given out (2730)
6064 - if enqueue arrives after completion, don't clobber LVB (2819)
6065 - don't unlock pages twice when trigger_group_io returns error (2814)
6066 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
6067 - don't write pages to disk if there was an error (1450)
6068 - don't ping imports that have recovery disabled (2676)
6069 - take buffered bytes into account when balancing socknal conn (2817)
6070 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
6071 - reconnect unlink llog connection after MDS reconnects to OST (2816)
6072 - remove little-endian swabbing of llog records (1987)
6073 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
6074 - retry reposting request buffers if they fail (1191)
6075 - grow extent at grant time to avoid granting a revoked lock (2809)
6076 - lock revoke doesn't evict page if covered by a second lock (2765)
6077 - disable VM readahead to avoid reading outside lock extents (2805)
6079 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
6080 - updated kernel patches for hp-2.4.20 kernel (2681)
6082 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
6085 - fix truncated write corruption (2366)
6086 - fix for failed assertion in iopen_connect_dentry (1792,2517)
6088 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
6089 - ensure len(uuid) < 37 in lmc (1171)
6090 - fix ia64 OOPS in llog_test (2255)
6091 - zero end of page at obdfilter for partial page writes (2648)
6092 - don't leave stale dentries around after renames (bug 2428)
6093 - fix timeouts when evicting a client with a single lock held (2642)
6094 - set deadline for the initial HELLO message to drain (2634)
6095 - print out dotted-quad IP addresses in the socknal (2302)
6097 - additional debugging for MDS client eviction problem (2443)
6098 - fix mkfsoptions support for osts (2603, 2604)
6100 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
6103 - add series for the vanilla 2.6.0 kernel
6104 - add series for the vanilla 2.4.24 kernel
6105 - add series for a cray x86/64 UL kernel drop
6106 - fix xattr patches for the vanilla 2.4.19 series
6108 - generate true UUIDs in lmc (1171)
6109 - have portals stack dumping break in UML (2466)
6110 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
6111 - allocate with _NOFS in ldlm to avoid deadlock (1933)
6112 - wake callback waiting threads on client eviction (2460)
6113 - Add --ptldebug and --subsystem to lmc (1719)
6114 - update assertion to allow safe interrupt allocation
6115 - set rq_no_resend for cancel requests (2432)
6116 - recalculate ptlrpcd timeout after resend (2494)
6117 - call vfs_rmdir when removing pending directories (2368)
6118 - fix renaming a file to itself (2429)
6119 - lmc creates a default one-stripe lov (2454)
6120 - expand procfs space to handle large clusters (2326)
6121 - increase UML stack to avoid overflow
6122 - update lconf's list of debug and subsystem masks
6123 - fix lfs find --obd (2510)
6124 - /proc tunable for disabling filter read caching (2591)
6125 - stop rpm packages from altering slapd.conf (2301)
6126 - disable nagle in the socknal under 0conf (2578)
6127 - choose mds inode size based on stripe count (2572)
6128 - fix kernel-source rpm problems (2516)
6130 - add --disable-doc to avoid pdf generation (2421)
6131 - update documentation, tests, type-os, comments
6132 - avoid format warnings on ia64
6133 - remove the TOE NAL
6134 - tiny code cleanups by removing unused fields
6136 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
6139 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
6140 - remove the most blatant lies from BUILDING file (2371)
6141 - change default debug level to reasonable production setting
6142 - reduce client side cache size to reduce cache flush time
6143 - reduce max RPCs in flight to avoid unnecessary file fragmentation
6144 - make TCP zerocopy and pinger support enabled by default (2476)
6145 - sync writes completed after process exits caused crashes (2319)
6146 - maintain correct mount count on the MDS (2356)
6147 - backout 1557, because 2316 wasn't really fixed
6148 - better file I/O statistics gathering in /proc
6149 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
6150 - another kernel patch to fix zero-copy TCP function export
6151 - don't take duplicate lock when processing re-sent getattr (2420)
6152 - lctl uses obd_self_export instead of creating new conn (2353)
6153 - MDS/OST recovery case which requires object creation asserted (2425)
6154 - move lfs from /usr/sbin to /usr/bin in packages
6155 - fix race between mds_client_add and mds_client_free (2417)
6156 - use kmalloc instead of slabs in portals (2430)
6157 - don't create duplicate records when a failover MDS is present (2442)
6158 - remove unnecessary mount age check (2332)
6159 - don't remove directory inodes from locks prematurely (2451)
6160 - don't break if MDS service name is the same as hostname (2103)
6161 - fix races in client write RPC generation when cache full (2482)
6163 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
6166 - remove now-unused request->rq_obd (278)
6167 - if an allocation fails, print out how much memory we've used (1933)
6168 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
6169 - add the "configurable stack size" patch to most series files (1256)
6170 - ability to write large log records, for 100+ OST configs (2306)
6171 - fix NULL deref when filter_prep fails (2314)
6172 - fix operator precedence error in filter_sync
6173 - dynamic allocation of socknal TX descriptors (2315)
6174 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
6175 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
6176 - ensure that CWARN messages in Portals always get to the syslog
6177 - __init/__exit are not for prototype decls (ldlm_init/exit)
6178 - x86-64 compile warning fixes
6179 - fix gateway LMC keyword conflict (2318)
6180 - fix MDS lock inversions in getattr/reint paths (1844)
6181 - fix a rare lock re-ordering bug, which caused deadlock (2322)
6182 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
6183 - DLM race condition prevented some lock evictions (2328)
6184 - ENOMEM detection and retry on socknal sends (2230)
6185 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
6186 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
6187 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
6188 - better error messages when a client is rejected during recovery (1505)
6189 - avoid cancelling locks which were never granted, after failure (2330)
6190 - fix i_sem/journal inversion in mds_client_add (2333)
6191 - fix truncate/getattr lock cycle deadlock (2334)
6192 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
6193 - fix two rare exit paths which could leak an l_lock() ref (2321)
6194 - fencepost error in MDS/OST orphan recovery (2226)
6195 - make log record alignment 8 bytes (1988)
6196 - lstripe now fails when requested offset > ost_count (2237)
6197 - ensure that all kernel series have a complete list.h (1607)
6198 - fix crashes in special-file operations (2316)
6199 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
6201 - allow configurable automake binary, for testing new versions
6202 - small update to the lfs documentation
6204 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
6206 * fix negative export reference count in fsfilt_sync (2312)
6208 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
6209 * release candidate 0.9.1
6211 - orphans are moved into the PENDING directory for possible recovery
6212 - replayed opens now open by fid for orphan/rename safety (1042)
6213 - last close of an orphan inode generates a transno (683)
6214 - chdir() and mount() now pin the directory entry (1020)
6215 - avoid CERROR in normal ll_setattr_raw() error case (1500)
6216 - discard very old requests without processing them (1502)
6217 - remove some common, well-understood CERRORs (1505)
6218 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
6219 - clear "grant" flags in OST replies until OST grant code lands (1644)
6220 - fix read performance by not clobbering i_blksize on client (1598)
6221 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
6222 - make LRU size a /proc tunable, clears locks when reduced (707)
6223 - fix some lprocfs rot that prevented ptlbd from loading (1732)
6224 - server locks take references on exports now (1558)
6225 - build fixes for 2.4.20-rh trees (1663)
6226 - return an error from lov_create if all OSCs are inactive (1751)
6227 - fix import levels when a reconnect happens without a timeout (1597)
6228 - exit early from mds_open if we get a lookup error (1749)
6229 - partial page read at EOF wouldn't wait for disk before sending (1642)
6230 - avoid NULL deref in obdfilter when reading page past EOF (1592)
6231 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
6232 - fix LBUG in ll_it_open_error with rc = -2 (1861)
6233 - write/truncate lock inversion (1639)
6234 - Don't auto-load obdclass, portals modules during cleanup (1495)
6235 - fix timestamps from jumping to "now" (1763)
6236 - extra journal assertions (1648)
6237 - add an extra multiunlink test (1771)
6238 - fix read_record/write_record API (1776)
6239 - fix leak of offset_extent, possible incorrect i_size later (1772)
6240 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
6241 - Debugging for the kqswnal_get_idle_tx problems (1820)
6242 - Allow recovery to be attempted multiple times (1536)
6243 - Write out MDS last_rcvd file after it is first created (1600)
6244 - Fix tx_descriptor leak in failed transmit situations (1827)
6245 - ext3 journaling fixes for assertion failure after IO error (1871)
6246 - class_export_put() on freed export after completion AST error (1896)
6247 - Fix revalidate looping in VFS (1322)
6248 - Don't access a freed export during MDS_REINT timeout (1521)
6249 - Add open-unlink recovery support on the MDS (1673,1764)
6250 - Return an error if no MDS data was read from last_rcvd (1946)
6251 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
6252 - Don't setup a disk device that doesn't match exported UUID (317)
6253 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
6254 - avoid committing NULL handle in force close
6255 - local.sh is now a one-stripe LOV configuration
6256 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
6257 - fix ext3 htree duplicate directory entry corruption (1516)
6258 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
6259 - update to new LOV EA format (2097)
6260 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
6262 2003-06-15 Phil Schwan <phil@clusterfs.com>
6265 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
6266 - per-import/export recovery handling (958, 931, 959)
6267 - multiple last-rcvd slots, for serving multiple FSes (949)
6268 - connections are again shared between multiple imp/exports (963, 964)
6269 - "umount -f" would hang if any requests needed to be sent (393, 978)
6270 - avoid pinning large req buffer by copying for queued messages (989)
6271 - add "uuid" to "lctl device" command to help upcalls (991)
6272 - "open" RPCs with transnos would confuse recovery counters (1037)
6273 - do proper endian conversion of all wire messages (288, 340, 891)
6274 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
6275 - call waiting locks callback from LDLM recovery thread (1127, 1151)
6276 - fix ptlrpc_connection leak in target_handle_connect (1174)
6277 - fix import refcounting bug in OST and MDS cleanup (1134)
6278 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
6279 - fix very unlikely obd_types race condition (501)
6280 - remove osc_open hack for echo_client (1187)
6281 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
6282 - a failure in read_inode2 leads to deadlock (1139)
6283 - cancel ack-locks as soon as transaction is committed (1072)
6284 - fix major leaks and crashes in the bulk I/O path (937, 1057)
6285 - make sure to commitrw after any preprw to avoid deadlock (1162)
6286 - failing to execute a file in a lustre FS would lock inode (1203)
6287 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
6288 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
6289 - obd_uuid2tgt would walk off the end of the list (1255)
6290 - on IA64 the osc would give portals incorrect bulk size (1258)
6291 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
6292 - fix lock inversion caused by new llite matching code (1282)
6293 - limit the number of dirty pages on a client to 10MB (1286)
6294 - timed out locks were not being corrected cancelled (1289)
6295 - fix O_DIRECT above 4GB on IA-32 (1292)
6296 * major user-visible changes
6297 - fail out/fail over policy now controlled by the upcall (993)
6299 - add OBD_PING to check server availability and failure (954)
6300 - lustre messages are now sent in sending host order (288, 340, 891)
6301 - add eadatalen to MDS getattr reply (340)
6302 - OST read replies may contain second buffer, with per-page status (593)
6304 2003-03-11 Phil Schwan <phil@clusterfs.com>
6307 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
6308 - failed open()s could cause deadlock; fixed (867, 869)
6309 - stop cancelling OST locks when files are closed (481)
6310 - overlapping XID spaces caused network corruption (851, 853)
6311 - fix unsafe fsfilt counter arithmetic; change to atomic_t
6312 - setattr_raw added, to do single-RPC, server-side setattrs
6313 - lmc/lconf syntax change for OST UUIDs
6314 - fix crashy race condition between ptlrpc_free_req and osc_close
6315 - don't use request in mdc_enqueue if we hit a timeout (889)
6316 - don't set the inode i_size for regular files from the MDS (896)
6317 - handle out of order completion AST (842)
6318 - don't LBUG if a lock request times out after receiving AST (913)
6319 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
6320 - if a bad lock AST arrives, send an error instead of dropping entirely
6321 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
6322 - fix leak in bulk IO when only partially completed (899, 900, 926)
6323 - fix O_DIRECT for ia64 (55)
6324 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
6325 - C-z after timeout could hang a process forever; fixed (977)
6327 - client-side I/O cache (678, 924, 929, 941, 970)
6329 - READPAGE and SETATTRs which don't take server-side locks get
6332 2003-02-11 Phil Schwan <phil@clusterfs.com>
6335 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
6336 - Fix fsfilt_extN_readpage() to read a full page of directory
6337 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
6338 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
6339 - Fix imbalanced LOV object allocation and out-of-bound access (469)
6340 - Most intent operations were removed, in favour of a new RPC mode
6341 that does a single RPC to the server and bypasses most of the VFS
6342 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
6343 - Aggressively cancel local locks on DLM servers
6344 - mds_reint_unlink sends EA to the client if it's the last nlink.
6345 client uses that EA to unlink OST objects.
6346 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
6347 - recursive symlinks were fixed (439)
6348 - fixed NULL deref in DEBUG_REQ
6349 - filter_update_lastobjid no longer calls sync, which annoyed extN
6350 - fixed multi-client small-writes to a single file problem (445)
6351 - fixed mtime updates during file writes (607)
6352 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
6353 - fixed bug in obd_brw_read/write() (under guise of testing 367)
6354 - fixed Linux OST size reporting problem (444, 656)
6355 - OST now updates object mtime with writes or setattr (607, 619)
6356 - client verifies file size before zeroing page past EOF (445)
6357 - OST now writes last allocated objid to disk with allocation (108)
6358 - LOV on echo now works (409)
6360 - mds_reint_unlink sends a new buffer, with the EA included. this
6361 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
6362 set if a regular file was being unlinked, and it was the last link
6363 - use PtlGet from the target for bulk writes (315)
6364 - OST now updates object mtime with writes or setattr (607, 619)
6365 - LDLM now has a grant-time callback to revalidate locked items, if
6367 - Many MDS operations were reorganized to combat race conditions
6369 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
6370 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
6372 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
6375 - Fully reactivate OST imports after reconnection (512, others)
6376 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
6377 - More graceful error handling for truncating on dead OST (515)
6378 - Don't error out unless we're actually accessing dead stripes (474)
6379 - Fix garbage sizes when stripes are missing (410)
6380 - LRU counters were broken, causing constant lock purge (433, 432)
6381 - garbage on read from stripes with failed OSTs (441)
6382 - mark OSCs as active before reconnecting during recovery (438)
6383 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
6384 - lfind did not preserve OST order in output (443)
6385 - symlinks cause hung clients, incorrect data (439)
6386 - stop dereferencing request after dropping refcount (457)
6387 - don't LASSERT(spin_is_locked) on non-SMP (455)
6388 - fixes for many rename() bugs
6389 - fstat didn't correctly synchronize attributes (399)
6390 - server must handle lock cancellation during blocking AST prep (487)
6391 - bulk descriptors were free()d too soon (511)
6392 - fix paths in lconf, which would load incorrect modules (451, 507)
6393 - fix confusing lconf 'host not found' error message (386)
6394 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
6395 - fix race condition in mdc_blocking_ast() for inode access (526)
6396 - fix lov_unpackmd() unpacking wrong number of stripes (537)
6397 - fix lov_set_osc_active() marking wrong OSC inactive (440)
6398 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
6399 - fix multiple writes of stripe MD to MDS (358, maybe 519)
6400 - fix lstripe in several ways (kernel side) (527)
6401 - fix request leak in ldlm_cli_enqueue (262)
6402 - incorrect OSC was marked inactive after OST failure
6403 - call mds_fs_cleanup before unmounting filesystem (524)
6404 - fix races between taking ns_lock and ldlm_lock_change_resource
6405 - fix races updating LOV export open file list
6406 - fix lov_enqueue error path, avoid decref-ing bad lock handle
6407 - fix recovery NULL deref in ldlm_cli_cancel_unused
6408 - fix some DLM races by using new hash table for lock handles (419)
6409 - permit the client to specify desired inodes, at replay
6410 - duplicate requests when we queue them for replay reintegration
6411 - fix last_rcvd offset calculation
6412 - sync after each recovered transaction, so we always make progress
6413 - never, not always, ERESTART requests without transnos
6414 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
6416 - skip replay if the MDS says that the client is already connected
6417 - don't check for a recovery-enabled export to match lctl's UUID
6418 - don't INC_USE_COUNT for phantom exports
6419 - don't crash when cleaning up phantom exports (567)
6420 - don't double-finish or set replay data for errored mdc_open requests
6421 - abort requests when they time out, so we don't get old replies
6422 - send/receive replies for AST messages again
6423 - if the client says that it doesn't have the lock, cancel it on the
6425 - if we timeout during I/O, don't try to cancel an in-use lock; instead
6426 mark it as destroyed, it will all work out when decref is called
6427 - fix module use counts (22, 581)
6429 - ASTs now expect a reply (server cancels lock on error reply)
6431 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
6434 - fix many simultaneous client startup (392)
6435 - fix dentry->d_it clobbering
6436 - credentials weren't being shipped for readdir/getattr operations
6437 - remove invalid assertions triggered during some concurrent MD
6439 - proper Lustre versions added (336, 389)
6440 - fix memory leak for create error case (398)
6441 - fix LOV locking bug that would get cli/srv out of sync
6442 - fix echo client over LOV (409)
6443 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
6444 - fix double-O_EXCL intent crash (424)
6445 - avoid sending multiple lock CANCELs (352)
6447 - MDS can do multi-client recovery (modulo bugs in new code)
6449 - many updates, edits, cleanups
6451 2002-11-18 Phil Schwan <phil@clusterfs.com>
6454 - fix null d_it dereference (346)
6455 - fix full OST/dbench hang (333)
6456 - fix permission problem with file removal (286)
6457 - fix removal of OSCs from LOV when they fail
6458 - fix NULL deref during bulk timeout (214)
6459 - fix problems related to multiple filesystems on one MDS (241)
6460 - fixed serious subtle metadata locking bugs
6461 - free locks on clients when inodes are removed due to memory
6463 - fix inode pointer in lock data (285)
6464 - partial support for multiple MDS on a single host (241)
6465 - data locks weren't cancelled at clear_inode time (290, 311)
6466 - intent locks could lead to unbounded lock growth (205)
6467 - added a maximum lock count, an LRU list, and a flusher
6468 - fix multiple rename (365)
6469 - properly abstracted the echo client
6470 - OSC locked 1 byte too many; fixed
6471 - rewrote brw callback code:
6472 - fixed recovery bugs related to LOVs (306)
6473 - fixed too-many-pages-in-one-write crash (191)
6474 - fixed (again) crash in sync_io_timeout (214)
6475 - probably fixed callback-related race (385)
6477 - Add capability to MDS protocol
6478 - LDLM cancellations and callbacks on different portals
6480 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
6483 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
6484 - defer open object destruction to close time (601981, 138)
6485 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
6486 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
6487 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
6488 - don't call obd_setattr in ll_file_release for destroyed objects
6489 * protocol change to lustre_msg: move |version| and add |flags|
6490 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
6491 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
6492 * added replay of create, unlink, link and rename operations during
6493 MDS failover; recovery should be much more robust now
6494 * remove failed OSCs from LOVs (only lov_create uses this so far)
6495 * the lustre-HOWTO was brought (more) up to date (582544)
6497 2002-10-23 Phil Schwan <phil@clusterfs.com>
6500 - in-use dentries weren't being reused properly (617851)
6501 - prevent multiple LDLM setup (599178)
6502 - fix LOV size calculations for truncate (617853)
6503 - fix client handling of MDS intent errors (POSIX)
6504 - fix permission bug in lovstripe.c test (624321)
6505 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
6506 - truncate past end of file could corrupt data
6507 - proper cleanup after timeouts, crashes, etc (592524, 550815)
6508 - a race in recovery could return ETIMEDOUT to apps (623947)
6509 - building outside the source directory was fixed
6510 * the lustre-HOWTO was brought (more) up to date (582544)
6511 * major progress was made on recovery functionality
6513 2002-10-10 Phil Schwan <phil@clusterfs.com>
6516 - recovery deadlock fix
6517 - rm -rf causes LBUG fix (617817)
6518 - file open by multiple tasks fix (618962)
6519 - directory permissions bugs (602707 and 620007)
6520 - journal_stop fixed with locking (611313)
6521 - O_APPEND failures resolved (618273, perhaps 614459)
6522 - lconf PATH fix (619770)
6523 - IA64 build fix (621450)
6524 - RPC buffer sizes scale with amount of memory
6526 2002-10-01 Phil Schwan <phil@clusterfs.com>
6529 - locks would be cancelled without throwing away data pages,
6530 resulting in inconsistent data (605627)
6531 - inode attributes were not always being refreshed (605627, 612449)
6532 - lconf now continues to cleanup after lctl reports an error
6533 - MDS now enforces user permissions (602707)
6534 - lprocfs cleanup fixed, but not yet enabled (614157)
6535 - fixed infinite server hang, should a client not respond to an AST
6536 - avoid going into recovery if user calls readlink() with a buffer
6537 that's too small (613941)
6538 - AST RPCs no longer require replies (614867) -- this may be changed
6539 - don't crash server if client sends an IOV that's too big (611336)
6540 - fixed lock conversion deadlock (611892)
6541 - fixed the following of symlinks (614622)
6542 * recovery: the server can remove locks from a client that dies, other
6543 clients can make progress
6544 * more extN patch fixes
6545 * compile-time configurable ptlrpc buffer allocations
6547 - collaborative read cache document
6548 - Lustre Lite Performance CDR document-in-progress
6550 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6553 - fix typo in patch-2.4.18
6555 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6558 - clear ptlrpc request each time in handle_incoming_request()
6559 - unlink of files now destroys the object on the OST
6561 2002-09-19 Peter Braam <braam@clusterfs.com>
6563 * add hard link support
6564 * change obdfile creation method
6565 * kernel patch changed
6567 2002-09-19 Peter Braam <braam@clusterfs.com>
6570 - stack overflow bug in extN fixed
6572 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
6574 * documentation updates
6575 - add man pages for config tools
6576 - update tests/README to describe testing with new config tools
6577 - finish metadata API descriptions
6578 * bug fixes and cleanups
6579 - statfs workaround for 16TB limit
6580 - LOV stripe allocation improved, can stripe on subset of OSTs
6581 - LOV file size/IO offset was wrong for files > 4GB in size
6582 - object EA data was being dropped, caused files to be unreadable
6583 - memory overflow with non-LOV OST caused memory corruption
6584 - fixed regression tests to work with new config tools, obdfilter
6585 - fixed bug when directory size became larger than 1 block
6586 - fixed bug (for single client case) when PWD was deleted
6587 - invalidate local directory pages when doing intent-based ops
6588 - avoid LDLM oops when lock callback contained bad data
6590 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
6592 * documentation updates
6593 * bug fixes and cleanups
6594 - configuration tools
6597 - 64-bit compile warnings
6598 - 64-bit internal statfs data
6600 * test_brw on persistent OST devices
6602 * lprocfs (disabled)
6604 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
6606 * documentation updates
6607 * bug fixes and cleanups
6608 * configuration tools
6610 2002-08-30 Peter J. Braam <braam@clusterfs.com>
6613 * many small fixes to 0_5_4
6614 * io/network handling
6615 * thinkos in MDS operations
6617 2002-08-24 Peter J. Braam <braam@clusterfs.com>
6620 * crucial basic fixes to 0.5.3
6621 * IOR, Iozone work over Elan
6624 2002-08-07 Phil Schwan <phil@clusterfs.com>
6625 * version 0_5_3, our first alpha
6626 * we use the new Portals iovs
6627 * documentation updates
6628 * bug fixes and cleanups
6629 * small changes in the DLM wire protocol
6631 2002-07-25 Peter J. Braam <braam@clusterfs.com>
6632 * version 0_5_1 with some initial stability,
6633 * locking on MD and file I/O.
6634 * documentation updates
6635 * several bug fixes since 0.5.0
6636 * small changes in wire protocol
6638 2002-07-18 Phil Schwan <phil@clusterfs.com>
6640 * delivered as Lustre Light Alpha
6641 * fixed a crash after handling invalid MDS requests
6642 * fixed directory pages for architectures with non-4k pages sizes
6644 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
6645 * release version v0_4_4
6646 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
6648 * Fixes a number of interruption problems with OST operations.
6649 * Update documentation for portals header changes
6650 * Move all wire protocol structs/defines to lustre_idl.h
6651 * Fixes symlink length bug.
6652 * Add tcpdump to repository.
6654 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
6655 * release version v0_4_3
6656 * Fixes statfs for inodes on extN.
6657 * Fixes bug in runtests which would delete /etc/hosts.
6658 * Use 64-bit object IDs wherever possible (not into VFS though)
6659 Remove ost_get_info, which is unused by lustre, and out of date.
6661 2002-07-03 Peter Braam <braam@clusterfs.com>
6662 * release version v0_4_2 Fixes a lookup error (type not passed)
6663 * move forward to head of Portals
6664 * move forward to latest Lustre kernel
6666 2002-06-25 Peter Braam <braam@clusterfs.com>
6667 * release version v0_4_1. Hopefully stable on single node use.