3 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
4 for currently supported client and server kernel versions.
5 * Server primary kernels built and tested during release cycle:
6 3.10.0-1062.18.1.el7 (RHEL7.7)
7 3.10.0-1127.19.1.el7 (RHEL7.8)
8 * Other server kernels known to build and work at some point (others may also work):
9 3.10.0-862.14.4.el7 (RHEL7.5)
10 3.10.0-957.27.2.el7 (RHEL7.6)
11 4.18.0-80.11.2.el8 (RHEL8.0)
12 4.18.0-147.8.1.el8 (RHEL8.1)
13 4.18.0-193.6.3.el8 (RHEL8.2)
14 4.4.120-92.70 (SLES12 SP2)
15 4.4.180-94.100 (SLES12 SP3)
16 4.4.0-131 (Ubuntu 16.04)
17 4.15.0-32 (Ubuntu 18.04)
18 vanilla linux 4.6.7 (ZFS only)
19 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
20 * Client primary kernels built and tested during release cycle:
21 3.10.0-1062.18.1.el7 (RHEL7.7)
22 3.10.0-1127.19.1.el7 (RHEL7.8)
23 3.10.0-1160.2.1.el7 (RHEL7.9)
24 4.12.14-95.48 (SLES12 SP4)
25 4.12.14-122.20 (SLES12 SP5)
26 4.12.14-197.37 (SLES15 SP1)
27 4.15.0-48 (Ubuntu 18.04)
28 * Other clients known to build on these kernels at some point (others may also work):
29 3.10.0-862.14.4.el7 (RHEL7.5)
30 3.10.0-957.27.2.el7 (RHEL7.6)
31 4.14.0-49.13.1.el7a (RHEL7.5)
32 4.14.0-115.2.2.el7a (RHEL7.6)
33 4.18.0-80.11.2.el8 (RHEL8.0)
34 4.18.0-147.8.1.el8 (RHEL8.1)
35 4.18.0-193.6.3.el8 (RHEL8.2)
36 4.4.120-92.70 (SLES12 SP2)
37 4.4.180-94.100 (SLES12 SP3)
38 4.4.0-131 (Ubuntu 16.04)
39 5.4.0-37 (Ubuntu 20.04)
40 5.4.0 (vanilla kernel.org)
41 * Recommended e2fsprogs version: 1.45.6.wc1 or newer
42 * Recommended ZFS version: 0.8.4
43 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
44 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
45 stack overflow. For more information, please refer to bugzilla 17630.
46 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
47 the re-exporting nodes to be mounted with "32bitapi" mount option
49 --------------------------------------------------------------------------------
53 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
54 for currently supported client and server kernel versions.
55 * Server primary kernels built and tested during release cycle:
56 3.10.0-1062.1.1.el7 (RHEL7.7)
57 * Other server kernels known to build and work at some point (others may also work):
58 3.10.0-862.14.4.el7 (RHEL7.5)
59 3.10.0-957.27.2.el7 (RHEL7.6)
60 4.18.0-80.7.1.el8 (RHEL8.0)
61 4.4.120-92.70 (SLES12 SP2)
62 4.4.180-94.100 (SLES12 SP3)
63 4.4.0-131 (Ubuntu 16.04)
64 4.15.0-32 (Ubuntu 18.04)
65 vanilla linux 4.6.7 (ZFS only)
66 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
67 * Client primary unpatched kernels built and tested during release cycle:
68 3.10.0-1062.1.1.el7 (RHEL7.7)
69 4.12.14-95.29 (SLES12 SP4)
70 4.15.0-48 (Ubuntu 18.04)
71 * Other clients known to build on unpatched kernels at some point (others may also work):
72 3.10.0-862.14.4.el7 (RHEL7.5)
73 3.10.0-957.27.2.el7 (RHEL7.6)
74 4.14.0-49.13.1.el7a (RHEL7.5)
75 4.14.0-115.2.2.el7a (RHEL7.6)
76 4.18.0-80.7.1.el8 (RHEL8.0)
77 4.4.120-92.70 (SLES12 SP2)
78 4.4.180-94.100 (SLES12 SP3)
79 4.4.0-131 (Ubuntu 16.04)
80 vanilla linux 4.15.0 (ZFS only)
81 * Recommended e2fsprogs version: 1.45.2.wc1 or newer
82 * Recommended ZFS / SPL version: 0.7.13
83 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
84 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
85 stack overflow. For more information, please refer to bugzilla 17630.
86 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
87 the re-exporting nodes to be mounted with "32bitapi" mount option
88 * Note for Ubuntu 14.04.5:
89 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
90 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
92 --------------------------------------------------------------------------------
96 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
97 for currently supported client and server kernel versions.
98 * Server known to build on the following kernels (others may also work):
99 2.6.32-696.18.7.el6 (RHEL6.9)
100 3.10.0-862.14.4.el7 (RHEL7.5)
101 3.10.0-957.el7 (RHEL7.6)
102 4.4.120-92.70 (SLES12 SP2)
103 4.4.162-94.69 (SLES12 SP3)
104 4.4.0-131 (Ubuntu 16.04)
105 4.15.0-32 (Ubuntu 18.04)
106 vanilla linux 4.6.7 (ZFS only)
107 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
108 * Client known to build on unpatched kernels (others may also work):
109 2.6.32-696.18.7.el6 (RHEL6.9)
110 2.6.32-754.6.3.el6 (RHEL6.10)
111 3.10.0-862.14.4.el7 (RHEL7.5)
112 3.10.0-957.el7 (RHEL7.6)
113 4.14.0-49.13.1.el7a (RHEL7.5)
114 4.14.0-115.2.2.el7a (RHEL7.6)
115 4.4.120-92.70 (SLES12 SP2)
116 4.4.162-94.69 (SLES12 SP3)
117 4.4.0-131 (Ubuntu 16.04)
118 4.15.0-32 (Ubuntu 18.04)
119 vanilla linux 4.15.0 (ZFS only)
120 * Recommended e2fsprogs version: 1.44.3.wc1 or newer
121 * Recommended ZFS / SPL version: 0.7.9
122 * Tested with ZFS / SPL version: 0.7.9
123 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
124 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
125 stack overflow. For more information, please refer to bugzilla 17630.
126 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
127 the re-exporting nodes to be mounted with "32bitapi" mount option
128 * Note for Ubuntu 14.04.5:
129 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
130 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
132 --------------------------------------------------------------------------------
134 2018-04-03 Intel Corporation
136 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
137 for currently supported client and server kernel versions.
138 * Server known to build on patched kernels:
139 2.6.32-431.29.2.el6 (RHEL6.5)
140 2.6.32-504.30.3.el6 (RHEL6.6)
141 2.6.32-573.26.1.el6 (RHEL6.7)
142 2.6.32-642.15.1.el6 (RHEL6.8)
143 2.6.32-696.18.7.el6 (RHEL6.9)
144 3.10.0-693.21.1.el7 (RHEL7.4)
145 3.0.101-0.47.71 (SLES11 SP3)
146 3.0.101-107 (SLES11 SP4)
147 3.12.74-60.64.40 (SLES12 SP1)
148 4.4.114-92.64 (SLES12 SP2)
149 4.4.114-94.11 (SLES12 SP3)
150 3.13.0-101 (Ubuntu 14.04, ZFS only)
151 4.4.0-85.108 (Ubuntu 14.04.5 LTS)
152 4.4.0-116 (Ubuntu 16.04)
153 vanilla linux 4.6.7 (ZFS only)
154 * Client known to build on unpatched kernels:
155 2.6.32-431.29.2.el6 (RHEL6.5)
156 2.6.32-504.30.3.el6 (RHEL6.6)
157 2.6.32-573.26.1.el6 (RHEL6.7)
158 2.6.32-642.15.1.el6 (RHEL6.8)
159 2.6.32-696.18.7.el6 (RHEL6.9)
160 3.10.0-693.21.1.el7 (RHEL7.4)
161 3.0.101-0.47.71 (SLES11 SP3)
162 3.0.101-107 (SLES11 SP4)
163 3.12.74-60.64.40 (SLES12 SP1)
164 4.4.114-92.64 (SLES12 SP2)
165 4.4.114-94.11 (SLES12 SP3)
166 3.13.0-101 (Ubuntu 14.04)
167 4.4.0-85.108 (Ubuntu 14.04.5 LTS)
168 4.4.0-116 (Ubuntu 16.04)
170 * Recommended e2fsprogs version: 1.42.13.wc6 or newer
171 * Recommended ZFS / SPL version: 0.7.6
172 * Tested with ZFS / SPL version: 0.7.6
173 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
174 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
175 stack overflow. For more information, please refer to bugzilla 17630.
176 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
177 the re-exporting nodes to be mounted with "32bitapi" mount option
178 * Note for Ubuntu 14.04.5:
179 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
180 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
182 --------------------------------------------------------------------------------
184 2017-07-05 Intel Corporation
186 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
187 for currently supported client and server kernel versions.
188 * Server known to build on patched kernels:
189 2.6.32-431.29.2.el6 (RHEL6.5)
190 2.6.32-504.30.3.el6 (RHEL6.6)
191 2.6.32-573.26.1.el6 (RHEL6.7)
192 2.6.32-642.15.1.el6 (RHEL6.8)
193 2.6.32-696.3.1.el6 (RHEL6.9)
194 3.10.0-514.21.1.el7 (RHEL7.3)
195 3.0.101-0.47.71 (SLES11 SP3)
196 3.0.101-100 (SLES11 SP4)
197 3.12.74-60.64.40 (SLES12 SP1)
198 4.4.59-92.17 (SLES12 SP2)
199 vanilla linux 4.6.7 (ZFS only)
200 * Client known to build on unpatched kernels:
201 2.6.32-431.29.2.el6 (RHEL6.5)
202 2.6.32-504.30.3.el6 (RHEL6.6)
203 2.6.32-573.26.1.el6 (RHEL6.7)
204 2.6.32-642.15.1.el6 (RHEL6.8)
205 2.6.32-696.3.1.el6 (RHEL6.9)
206 3.10.0-514.21.1.el7 (RHEL7.3)
207 3.0.101-0.47.71 (SLES11 SP3)
208 3.0.101-100 (SLES11 SP4)
209 3.12.74-60.64.40 (SLES12 SP1)
210 4.4.59-92.17 (SLES12 SP2)
212 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
213 * Recommended ZFS / SPL version: 0.6.5.9
214 * Tested with ZFS / SPL version: 0.6.5.9
215 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
216 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
217 stack overflow. For more information, please refer to bugzilla 17630.
218 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
219 the re-exporting nodes to be mounted with "32bitapi" mount option
221 --------------------------------------------------------------------------------
223 12-07-2016 Intel Corporation
225 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
226 for currently supported client and server kernel versions.
227 * Server known to build on patched kernels:
228 2.6.32-431.29.2.el6 (RHEL6.5)
229 2.6.32-504.30.3.el6 (RHEL6.6)
230 2.6.32-573.26.1.el6 (RHEL6.7)
231 2.6.32-642.11.1.el6 (RHEL6.8)
232 3.10.0-514.el7 (RHEL7.3)
233 3.0.101-0.47.71 (SLES11 SP3)
234 3.0.101-80 (SLES11 SP4)
235 3.12.62-60.64.8 (SLES12 SP1)
236 vanilla linux 4.5.7 (ZFS only)
237 * Client known to build on unpatched kernels:
238 2.6.32-431.29.2.el6 (RHEL6.5)
239 2.6.32-504.30.3.el6 (RHEL6.6)
240 2.6.32-573.26.1.el6 (RHEL6.7)
241 2.6.32-642.11.1.el6 (RHEL6.8)
242 3.10.0-514.el7 (RHEL7.3)
243 3.0.101-0.47.71 (SLES11 SP3)
244 3.0.101-80 (SLES11 SP4)
245 3.12.62-60.64.8 (SLES12 SP1)
247 * Recommended e2fsprogs version: 1.42.13.wc5 or newer
248 * Recommended ZFS / SPL version: 0.6.5.8
249 * Tested with ZFS / SPL version: 0.6.5.8
250 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
251 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
252 stack overflow. For more information, please refer to bugzilla 17630.
253 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
254 the re-exporting nodes to be mounted with "32bitapi" mount option
256 --------------------------------------------------------------------------------
258 02-29-2016 Intel Corporation
260 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
261 for currently supported client and server kernel versions.
262 * Server known to build on patched kernels:
263 2.6.32-431.29.2.el6 (RHEL6.5)
264 2.6.32-504.30.3.el6 (RHEL6.6)
265 2.6.32-573.12.1.el6 (RHEL6.7)
266 3.10.0-327.3.1.el7 (RHEL7.2)
267 3.0.101-0.47.71 (SLES11 SP3)
268 3.0.101-68 (SLES11 SP4)
269 vanilla linux 4.2.1 (ZFS only)
270 * Client known to build on unpatched kernels:
271 2.6.32-431.29.2.el6 (RHEL6.5)
272 2.6.32-504.30.3.el6 (RHEL6.6)
273 2.6.32-573.12.1.el6 (RHEL6.7)
274 3.10.0-327.3.1.el7 (RHEL7.2)
275 3.0.101-0.47.71 (SLES11 SP3)
276 3.0.101-68 (SLES11 SP4)
279 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
280 * Recommended ZFS / SPL version: 0.6.4.2
281 * Tested with ZFS / SPL version: 0.6.4.2
282 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
283 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
284 stack overflow. For more information, please refer to bugzilla 17630.
285 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
286 the re-exporting nodes to be mounted with "32bitapi" mount option
288 --------------------------------------------------------------------------------
290 03-10-2015 Intel Corporation
292 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
293 for currently supported client and server kernel versions.
294 * Server known to build on patched kernels:
295 2.6.32-431.29.2.el6 (RHEL6.5)
296 2.6.32-504.8.1.el6 (RHEL6.6)
297 3.0.101-0.46 (SLES11 SP3)
298 * Client known to build on unpatched kernels:
299 2.6.32-431.29.2.el6 (RHEL6.5)
300 2.6.32-504.8.1.el6 (RHEL6.6)
301 3.10.0-123.20.1.el7 (RHEL7)
302 3.0.101-0.46 (SLES11 SP3)
303 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
304 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
305 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
306 stack overflow. For more information, please refer to bugzilla 17630.
307 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
308 the re-exporting nodes to be mounted with "32bitapi" mount option
310 Severity : enhancement
312 Description: control OST-index in IDIF via ROCOMPAT flag.
313 Details : Introduce new flag OBD_ROCOMPAT_IDX_IN_IDIF that is stored in the
314 last_rcvd file. For new formatted OST device, it will be auto set;
315 for the case of upgrading from old OST device, you can enable it
316 via the lproc interface osd-ldiskfs.index_in_idif. With such flag
317 enabled, for new created OST-object, its IDIF-in-LMA will contain
318 the OST-index; for the existing OST-object, the OSD will convert
319 old format IDIF as new format IDIF with OST-index stored in the
320 LMA EA when accessing such OST-object or via OI scrub. Once such
321 flag is enabled, it cannot be reverted back, so the system cannot
322 be downgraded to the orignal incompatible version.
324 --------------------------------------------------------------------------------
326 07-30-2014 Intel Corporation
328 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
329 for currently supported client and server kernel versions.
330 * Server known to build on patched kernels:
331 2.6.32-431.20.3.el6 (RHEL6.5)
332 3.0.101-0.31 (SLES11 SP3)
333 * Client known to build on unpatched kernels:
334 2.6.32-431.20.3.el6 (RHEL6.5)
335 3.0.101-0.31 (SLES11 SP3)
338 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
339 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
340 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
341 stack overflow. For more information, please refer to bugzilla 17630.
342 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
343 the re-exporting nodes to be mounted with "32bitapi" mount option
345 --------------------------------------------------------------------------------
347 10-31-2013 Intel Corporation
349 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
350 * Server builds for kernels:
351 2.6.32-358.18.1.el6 (RHEL6.4)
352 * Client builds for unpatched kernels:
353 2.6.32-358.18.1.el6 (RHEL6.4)
354 3.0.92-0.8 (SLES11 SP3)
356 * Recommended e2fsprogs version: 1.42.7.wc2 or newer
357 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
358 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
359 stack overflow. For more information, please refer to bugzilla 17630.
360 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
361 the re-exporting nodes to be mounted with "32bitapi" mount option
363 --------------------------------------------------------------------------------
365 05-30-2013 Intel Corporation
367 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
368 * Server builds for kernels:
369 2.6.32-358.6.2.el6 (RHEL6.4)
370 * Client builds for unpatched kernels:
371 2.6.32-358.6.2.el6 (RHEL6.4)
372 3.0.74-0.6.6 (SLES11 SP2)
374 * Recommended e2fsprogs version: 1.42.7.wc1 or newer
375 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
376 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
377 stack overflow. For more information, please refer to bugzilla 17630.
378 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
379 the re-exporting nodes to be mounted with "32bitapi" mount option
381 --------------------------------------------------------------------------------
383 9-30-2012 Whamcloud, Inc.
385 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
386 * Server support for kernels:
387 2.6.32-279.5.1.el6 (RHEL6)
388 * Client support for unpatched kernels:
389 2.6.32-279.5.1.el6 (RHEL6)
391 2.6.18-238.19.1.el5 (RHEL5)
392 2.6.16.60-0.69.1 (SLES10)
393 * Recommended e2fsprogs version: 1.41.90.wc4
394 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
395 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
396 stack overflow. For more information, please refer to bugzilla 17630.
397 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
398 the re-exporting nodes to be mounted with "32bitapi" mount option
400 Severity : enhancement
402 Description: Update RHEL6.3 kernel to 2.6.32-279.5.1.el6.
404 Severity : enhancement
406 Description: Update RHEL6.3 kernel to 2.6.32-279.2.1.el6.
408 Severity : enhancement
410 Description: Update RHEL6.2 kernel to 2.6.32-220.17.1.el6
412 Severity : enhancement
414 Description: Update RHEL6.2 kernel to 2.6.32-220.13.1.el6
416 Severity : enhancement
418 Description: Update RHEL6.2 kernel to 2.6.32-220.7.1.el6
420 --------------------------------------------------------------------------------
422 04-30-2012 Whamcloud, Inc.
424 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
425 * Server support for kernels:
426 2.6.32-220.4.2.el6 (RHEL6)
427 * Client support for unpatched kernels:
428 2.6.32-220.4.2.el6 (RHEL6)
430 2.6.18-238.19.1.el5 (RHEL5)
431 2.6.16.60-0.69.1 (SLES10)
432 * Recommended e2fsprogs version: 1.41.90.wc4
433 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
434 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
435 stack overflow. For more information, please refer to bugzilla 17630.
436 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
437 the re-exporting nodes to be mounted with "32bitapi" mount option
439 Severity : enhancement
441 Description: Update RHEL6.2 kernel to 2.6.32-220.4.2.el6
443 Severity : enhancement
445 Description: Update RHEL6.2 kernel to 2.6.32-220.4.1.el6
447 Severity : enhancement
449 Description: Update RHEL6.2 kernel to 2.6.32-220.el6
451 Severity : enhancement
453 Description: Update RHEL6.1 kernel to 2.6.32-131.17.1.el6
455 --------------------------------------------------------------------------------
457 09-30-2011 Whamcloud, Inc.
459 * Server support for kernels:
460 2.6.18-238.19.1.el5 (RHEL5)
461 2.6.32-131.6.1.el6 (RHEL6)
462 * Client support for unpatched kernels:
463 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
464 2.6.18-238.19.1.el5 (RHEL5)
465 2.6.32-131.6.1.el6 (RHEL6)
466 2.6.29.4-167.fc11 (FC11)
467 2.6.16.60-0.69.1 (SLES10)
469 * Recommended e2fsprogs version: 1.41.90.wc3
470 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
471 Lustre filesystem with 4K stack may cause a stack overflow. For more
472 information, please refer to bugzilla 17630.
473 * NFSv4 reexport requires Lustre client on the re-exporting nodes to be
474 mounted with "32bitapi" mount option
476 Severity : enhancement
477 Description: Update RHEL6.1 kernel to 2.6.32-131.6.1.el6
479 Severity : enhancement
480 Description: Update RHEL5.6 kernel to 2.6.18-238.19.1.el5
482 Severity : enhancement
483 Description: Update RHEL5.6 kernel to 2.6.18-238.9.1.el5
486 Frequency : only with SLES10
488 Description: Use OFED "KMP" provided by Novell
489 Details : SLES10 SP3 ships with OFED in a separate "KMP" package.
490 Lustre is now built against this package. That means you need to
491 install the ofed-kmp package from Novell for the patchless client.
492 Note that the ofed-kmp that Novell ships may not exactly match the
493 kernel version but should still be compatible.
495 Severity : enhancement
497 Description: Update RHEL5.5 kernel to 2.6.18-194.17.1.el5
498 Update OEL5.5 kernel to 2.6.18-194.17.1.0.1.el5.
500 --------------------------------------------------------------------------------
502 2010-07-15 Oracle, Inc.
504 * Server support for kernels:
505 2.6.18-164.11.1.el5 (RHEL 5)
506 2.6.18-164.11.1.0.1.el5 (OEL 5)
507 * Client support for unpatched kernels:
508 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
509 2.6.18-164.11.1.el5 (RHEL 5),
510 2.6.18-164.11.1.0.1.el5 (OEL 5)
511 2.6.16.60-0.42.8 (SLES 10),
513 2.6.29.4-167.fc11 (FC11)
514 * Recommended e2fsprogs version: 1.41.10-sun2
516 Severity : enhancement
518 Description: Distribute lfs_migrate script with lustre
519 Details : The lfs_migrate mentioned in the manual is now included in the
520 lustre tree in lustre/scripts and will distributed in lustre RPMs.
522 Severity : enhancement
524 Description: Failover nid list added to import proc file.
526 Severity : enhancement
528 Description: "som_preview" mount option required to use SOM
529 Details : The clients will need to use som_preview mount option
533 Description: Lustre does not do 1MB IOs to HW RAID
534 Details : Bump MAX_PHYS/HW_SEGMENTS and SG_ALL to 256 in the RHEL5 kernel.
536 Severity : enhancement
538 Description: lfs quota output is not very convenient for awk/sed-parsing
539 Details : Some positions in lfs quota output table could be empty or
540 non-empty which made it hard to parse it with scripts, now
541 a dash is put instead of space where there is not supposed
546 Description: Update to OFED 1.5.1
547 Details : OFED 1.5.1 will be used in environments that don't use
548 vendor supplied OFED.
552 Description: Remove lnet. prefix from lctl params display
553 Details : Also renamed lnet.memused to lnet_memused to
558 Description: Use conf_param -d to remove permanant settings.
562 Description: Add list_param -R
563 Details : lctl list_param now supports recursive mode.
566 Bugzilla : 21829, 22312
567 Description: Fix llobdstat.
568 Details : 1. Now updating the write_bytes counter
569 2. Port script fixes from 1.8.
573 Description: decrease the usage of memory on clients.
574 Details : 1. On clients, recycle dentries and inodes unused.
575 2. Delete the code related to ll_deathrow(att 6215 in bug 1443). It
580 Description: llobdstat fix and enhancement
581 Details : add a counter to set a limit to how many samples will be returned
582 fix a wildcard in the path to limit to obdfilter stats only
586 Description: lfs check only allowed for root
587 Details : Code cleanups and added a regresion test of non-root use of
592 Description: MDS Oopses in ptlrpc_dispatch_difficult_reply()/__list_add()
593 Details : Fix hr->hr_index overload.
597 Description: Fix FIEMAP functionality.
601 Description: FC11 patchless client support added
605 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and
606 OEL5.4 kernel to 2.6.18-164.11.1.0.1.el5.
610 Description: New acc-sm script and various test-framework enhancments.
614 Description: Previously landed patch was reverted.
618 Description: Update SLES10 kernel to 2.6.16.60-0.42.8 (patchless)
622 Description: Add ldiskfs maxdirsize mount option.
626 Description: Fix mkfs.lustre support for 16TB LUNs.
630 Description: Print mmp_check_interval
631 Details : print mmp_check_interval and make it possible to abort
632 mount operation in case it takes too long.
636 Description: Add cascading_rw to lustre mpi tests.
640 Description: Store SOM data in EA.
641 Details : Also several SOM related cleanups.
645 Description: Add refcount to md_open_data
649 Description: Use refcount on lov_request_set to prevent memory corruption.
653 Description: Statahead: use parent dir's i_mutex for synchronization.
657 Description: loadgen improvements.
661 Description: Enable peer health detection in ksocklnd and ko2iblnd.
665 Description: Control DCACHE_LUSTRE_INVALID flag with MDS_INODE_LOOKCUP lock
669 Description: Fix race on obd_devcie:md_stats
673 Description: Fix SLES 11 client builds.
674 Details : Resolved uninitalized pointer gcc warnings.
678 Description: Allow non-root access for "lfs check".
679 Details : Added a check in obd_class_ioctl() for OBD_IOC_PING_TARGET.
683 Description: Clear imp_force_reconnect correctly in ptlrpc_connect_interpret()
685 Severity : enhancement
687 Description: Update kernel to RHEL5.4 2.6.18-164.6.1.el5 and
688 OEL5 2.6.18-164.6.1.0.1.el5(Both in-kernel OFED enabled).
692 Description: Sanity test_27p fails.
693 Details : OBD_FAIL_OST_ENOSPC has been moved to filter_statfs().
694 Sanity tests 27 were cleaned up and improved.
698 Description: Unify req format on client/servers
699 Details : Use new req_capsule API [almost] everywhere instead of old PTLRPC
700 buffers and swabbers approach..
705 Description: drop join file stuff.
707 Severity : enhancement
709 Description: Add 2.6.27 and SLES11 (patchless client) support
713 Description: Fix in ptlrpc_expire_one_request() to print the signed time
716 Severity : enhancement
718 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
721 Severity : enhancement
723 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
725 Severity : enhancement
727 Description: correctly handle big reply message.
728 Details : send LNet event if reply is bigger than buffer and adjust this buffer
733 Description: Drop unnecessary __GFP_NOMEMALLOC flag from filter_get_page()
735 Severity : enhancement
737 Description: Add support for OEL5.
741 Description: Conf-sanity.sh 50g test - deactivated OST should not cause a panic.
743 Severity : enhancement
745 Description: "weak-modules" support
746 Details : Implement "weak-modules" support which enables kernel modules
747 to be used with any kernel that implements the same kABI. In
748 order to achieve this modules are now installed in
749 /lib/modules/$(uname -r)/updates/kernel on all distributions.
753 Description: client could not reconnect to OST because of an active request.
754 Details : The client now retries to reconnect to the same server, if a
755 connect request failed with EBUSY or -EAGAIN. Test case 35b
756 conf-sanity.sh has been created.
760 Description: Don't allow make backward step on assignin osc next id.
761 Details : race between allocation next id and ll_sync thread can be cause
762 of set wrong osc next id and can be kill valid ost objects.
766 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
770 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
774 Description: Error handling in osc_statfs_interpret() has been improved.
775 Details : Check in osc_statfs_interpret() for EBADR.
779 Description: Do not send statfs() requests to OSTs disabled by administrator.
780 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
782 Severity : enhancement
784 Description: Added sanity-quota test to for lfs(1) to work with numeric uid/gid.
786 Severity : enhancement
788 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
790 Severity : enhancement
792 Description: Enhancement for lfs(1) command to use numeric uid/gid.
796 Description: Improved cleanup procedure for conf-sanity test_46a.
800 Description: Do not start bulk transfer, if client tried to reconnect.
801 Details : Do not start bulk transfer, if client tried to reconnect,
802 but failed due to active rpcs. Check for exp_abort_active_req
803 has been implemented.
807 Description: Prevent inconsistences between linux and lustre mount structures.
808 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
809 Make the sleep interruptible.
811 Severity : enhancement
813 Description: provide server to client comms path
814 Details : server to client communications path via new LDLM_SET_INFO rpc
818 Description: truncate starts GFP_FS allocation under transaction causing deadlock
819 Details : ldiskfs_truncate calls grab_cache_page which may start page
820 allocation under an open transaction. This may lead to
821 calling prune_icache with consequent lustre reentrance.
825 Description: send correctly lsm on open replay
826 Details : MDS is trust to LSM size on replay open, but client can set wrong size
831 Description: enable client interface failover
832 Details : When a child reconnects from another NID, properly update export
833 nid hash position and ldlm reverse import.
835 Severity : enhancement
837 Description: Communicate OST degraded/readonly state via statfs to MDS
838 Details : Flags in the statfs returned from OSTs indicate whether the
839 OST is in a degraded RAID state, or if the filesystem has
840 turned read-only after a filesystem error is detected.
845 Description: don't panic if EPROTO was hit when reading symlink
846 Details : correctly handling request reference in error cases.
848 Severity : enhancement
850 Description: Add LustreNetLink, a kernel-userspace communcation path.
852 Severity : enhancement
854 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
859 Description: access to llog context before init.
860 Details : move handling CATALOGS file at osc layer and forbid access to llog
864 Frequency : rare, only when client is writing to OST on local node
866 Description: Avoid deadlock for local client writes
867 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
868 memory freeing context. This allows OST threads to set the
869 PF_MEMALLOC flag on task structures in order to allocate memory
870 from reserved pools and complete IO.
871 Use GFP_HIGHUSER for OST allocations for non-local client writes,
872 so that the OST threads generate memory pressure and allow
873 inactive pages to be reclaimed.
875 Severity : enhancement
877 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
882 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
883 Details : move ldlm namespace creation in setup phase to avoid grab
884 _lprocfs_lock with cli_sem held.
888 Description: Temporarily disable grant shrink.
889 Details : Disable the feature for debugging.
892 Frequency : only when formatting test filesystems
894 Description: Unable to run several mkfs.lustre on loop devices at the same time
895 Details : mkfs.lustre returns error 256 on the concurrent loop devices
896 formatting. The solution is to proper handle the error.
898 Severity : enhancement
900 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
902 Severity : enhancement
904 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
909 Description: resolve race between obd_disconnect and class_disconnect_exports
910 Details : if obd_disconnect will be called to already disconnected export he
911 forget release one reference and osc module can't unloaded.
913 Severity : enhancement
915 Description: Allow tuning service thread via /proc
916 Details : For each service a new
917 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
918 created that can be used to set min/max thread counts, and get the
919 current number of running threads.
924 Descriptoin: don't return error if have partially created objects for file.
925 Details : lov_update_create_set uses set->set_success as index for created
926 objects, so if some requests will be failed, they will have hole
927 at end of array and qos_shrink_lsm for allocate correct lsm.
929 Severity : enhancement
931 Description: Update OFED support to 1.4.1
936 Description: Reduce small size read RPC
937 Details : Set read-ahead limit for every file and only do read-ahead when
938 available read-ahead pages are bigger than 1M to avoid small size
941 Severity : enhancement
943 Description: move AT tunable parameters for more consistent usage
944 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
946 Severity : enhancement
948 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
949 Details : allow skip disconnected ost for send statfs request and hide error
955 Description: fix racy locking of mballoc block bitmaps causing BUG
956 Details : The locking of the mballoc buddy bitmap and the in-memory
957 block bitmap was using two different spin locks in some
958 cases. This made it possible to incorrectly access the
959 mballoc bitmap while another process was modifying it,
960 causing a sanity assertion to fail. While no on-disk corruption
961 was reported, there was some risk of this happening.
964 Frequency : rare, on llog test 6
966 Descriptoin: don't allow connect to already connected import
967 Details : allowing connect to already connected import is hide connecting problem.
972 Descriptoin: allow kill process which wait statahead result
973 Details : for some reasons 'ls' can stick in waiting result from statahead,
974 in this case need way for kill this process.
976 Severity : enhancement
978 Description: Add state history info file, enhance import info file
979 Details : Track import connection state changes in a new osc/mdc proc file;
980 add overview-type data to the osc/mdc import proc file.
982 Severity : enhancement
984 Description: MDS create should not wait for statfs RPC while holding DLM lock.
986 Severity : enhancement
988 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
993 Descriptoin: OST grant too much space to client even there are not enough space.
994 Details : Client will shrink its grant cache to OST if there are no write
995 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
996 this grant cache if there are already not enough avaible space
997 (left_space < total_clients * 32M).
1000 Frequency : start MDS on uncleanly shutdowned MDS device
1002 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
1003 Details : stay in waiting mds<>ost recovery finished produce random bugs
1004 due race between two ll_sync thread for one lov target. send
1005 ACTIVATE event only if connect realy finished and import have
1009 Frequency : rare, connect and disconnect target at same time
1011 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
1012 Details : don't call obd_disconnect under lov_lock. this long time
1013 operation and can block ptlrpcd which answer to connect request.
1018 Descriptoin: don't lose wakeup for imp_recovery_waitq
1019 Details : recover_import_no_retry or invalidate_import and import_close can
1020 both sleep on imp_recovery_waitq, but we was send only one wakeup
1024 Frequency : always with long access acl
1026 Descriptoin: mds can't pack reply with long acl.
1027 Details : mds don't control size of acl but they limited by reint/getattr
1030 Severity : enhancement
1032 Description: Update to SLES10 kernel-2.6.16.60-0.33.
1034 Severity : enhancement
1036 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
1039 Frequency : start MDS on uncleanly shutdowned MDS device
1041 Descriptoin: aborting recovery hang on MDS
1042 Details : don't throttle destroy RPCs for the MDT.
1045 Frequency : on remount
1047 Description: external journal device not working after the remount
1048 Details : clear dev_rdonly flag for external journal devices in
1054 Description: shutdown vs evict race
1055 Details : client_disconnect_export vs connect request race.
1056 if client will evicted at this time - we start invalidate
1057 thread without referece to import and import can be freed
1061 Frequency : rare, need acl's on inode.
1063 Description: client can't handle ost additional correctly
1064 Details : if ost was added after client connected to mds client can have
1065 hit lnet_try_match_md ... to big messages to wide striped files.
1066 in this case need teach client to handle config events about add
1067 lov target and update client max ea size at that event.
1069 Severity : enhancement
1071 Description: Changelogs
1072 Details : Changelogs are a lightweight mechanism to track filesystem
1073 metadata and namespace changes. The changelog is recorded
1074 permanently on the MDTs, and is periodically "consumed" / purged
1075 when records are no longer needed.
1077 Severity : enhancement
1079 Description: compact fld format with extents
1080 Details : Store range of seq rather than every seq in FLD. Seq
1081 controller update FLD rather than clients. In Case of CMD, mdt0
1082 has FLD, all other metadata server act as non persistent proxy
1083 for FLD queries and cache fld entries in fld cache.
1088 Description: don't skip ost target if they assigned to file
1089 Details : Drop slow OSCs if we can, but not for requested start idx.
1090 This means "if OSC is slow and it is not the requested
1091 start OST, then it can be skipped, otherwise skip it only
1092 if it is inactive/recovering/out-of-space.
1096 Description: more cleanup in mds_lov
1097 Details : not send LOV EA under replay, we can't know about they size at this
1098 time. Don't allow client connect to mds before any ost connected,
1099 for avoid problems with LOV EA size and returning EIO to client.
1101 Severity : enhancement
1103 Description: Interoperability at server side (Disk interoperability)
1105 Severity : enhancement
1107 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
1109 Severity : enhancement
1111 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
1113 Severity : enhancement
1115 Description: New client IO stack (CLIO).
1117 Severity : enhancement
1119 Description: Commit on sharing. Eliminate inter-client dependencies between
1120 uncommitted transactions by doing transaction commits.
1121 Thereby clients may recovery independently.
1124 Frequency : Create a symlink file with a very long name
1126 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
1127 Details : If there is no extra space in the request for early cancels,
1128 ldlm_req_handles_avail() returns 0 instead of a negative value.
1130 Severity : enhancement
1132 Description: Add /proc entry for import status
1133 Details : The mdc, osc, and mgc import directories now have
1134 an import directory that contains useful import data for debugging
1135 connection problems.
1137 Severity : enhancement
1139 Description: Re-disable certain /proc logging
1140 Details : Enable and disable client's offset_stats, extents_stats and
1141 extents_stats_per_process stats logging on the fly.
1144 Frequency : Only on FC kernels 2.6.22+
1146 Description: oops in statahead
1147 Details : Do not drop reference count for the dentry from VFS when lookup,
1148 VFS will do that by itself.
1150 Severity : enhancement
1152 Description: Generic /proc file permissions
1153 Details : Set /Proc file permissions in a more generic way to enable non-
1154 root users operate on some /proc files.
1158 Description: Hitting mdc_commit_close() ASSERTION
1159 Details : Properly handle request reference release in
1160 ll_release_openhandle().
1164 Description: quota recovery deadlock during mds failover
1165 Details : This patch includes att18982, att18236, att18237 in bz14840.
1167 1. fix osts hang when mds does failover with quotaon
1168 2. prevent watchdog storm when osts threads wait for the
1173 Frequency : only patchless client
1174 Description: add workaround for race between add/remove dentry from hash
1176 Severity : enhancement
1178 Description: Allow OST glimpses to return PW locks
1182 Description: LBUG when llog conf file is full
1183 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
1187 Description: Prevent import from entering FULL state when server in recovery
1191 Description: service mount cannot take device name with ":"
1192 Details : Only when device name contains ":/" will mount treat it as
1198 Description: replace ptlrpcd with the statahead thread to interpret the async
1199 statahead RPC callback
1203 Frequency : on recovery
1204 Description: I/O failures after umount during fail back
1205 Details : if client reconnected to restarted server we need join to recovery
1206 instead of find server handler is changed and process self
1207 eviction with cancel all locks.
1209 Severity : enhancement
1211 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
1213 Severity : enhancement
1215 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
1217 Severity : enhancement
1219 Description: Upcall on Lustre log has been dumped
1220 Details : Allow for a user mode script to be called once a Lustre log has
1221 been dumped. It passes the filename of the dumped log to the
1222 script, the location of the script can be specified via
1223 /proc/sys/lnet/debug_log_upcall.
1228 Description: avoid idr_remove called for id which is not allocated.
1229 Details : Move assigment s_dev for clustered nfs to end of initialization,
1230 to avoid problem with error handling.
1235 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
1236 Details : When connection is reused this not moved from CONN_UNUSED_HASH
1237 into CONN_USED_HASH and this prodice warning when put connection
1238 again in unused hash.
1241 Severity : enhancement
1243 Description: File striping can now be set to use an arbitrary pool of OSTs.
1245 Severity : enhancement
1247 Description: Export bytes_read/bytes_write count on OSC/OST.
1251 Description: Early reply size mismatch, MGC loses connection
1252 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
1253 the connect flags are properly negotiated.
1258 Description: quotas are not honored with O_DIRECT
1259 Details : all writes with the flag O_DIRECT will use grants which leads to
1260 this problem. Now using OBD_BRW_SYNC to guard this.
1264 Description: add quota statistics
1265 Details : 1. sort out quota proc entries and proc code.
1266 2. add quota statistics
1268 Severity : enhancement
1270 Description: enable quota support for HEAD.
1274 Description: Properly propagate oinfo flags from lov to osc for statfs
1275 Details : restore missing copy oi_flags to lov requests.
1277 Severity : enhancement
1279 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
1280 : l_getgroups(8), lst(8), routerstat(8)
1281 Details : included man pages for llobdstat(8), llstat(8),
1282 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
1284 Severity : enhancement
1286 Description: configure's --enable-quota should check the
1287 : kernel .config for CONFIG_QUOTA
1288 Details : configure is terminated if --enable-quota is passed but
1289 : no quota support is in kernel
1293 Description: Remove portals compatibility
1294 Details : Remove portals compatibility, not interoperable with releases
1299 Description: Resolve device initialization race
1300 Details : Prevent proc handler from accessing devices added to the
1301 obd_devs array but yet be intialized.
1303 Severity : enhancement
1305 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
1307 Severity : enhancement
1309 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
1314 Description: Using wrong pointer in osc_brw_prep_request
1315 Details : Access to array[-1] can produce panic if kernel compiled with
1316 CONFIG_PAGE_ALLOC enabled
1320 Description: Client runs out of low memory
1321 Details : Consider only lowmem when counting initial number of llap pages
1325 Description: *optional* service tags registration
1326 Details : if the "service tags" package is installed on a Lustre node
1327 When the filesystem is mounted, a local-node service tag will
1328 be created. See http://inventory.sun.com/ for more information
1329 about the Service Tags asset management system.
1333 Description: Kernel BUG tries to release flock
1334 Details : Lustre does not destroy flock lock before last reference goes
1335 away. So always drop flock locks when client is evicted and
1336 perform unlock regardless of successfulness of speaking to MDS.
1340 Description: add refcount protection for osc callbacks, avoid panic on shutdown
1344 Description: sanity test 65a fails if stripecount of -1 is set
1345 Details : handle -1 striping on filesystem in ll_dirstripe_verify
1350 Description: ASSERTION(CheckWriteback(page,cmd)) failed
1351 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
1354 Severity : enhancement
1356 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
1360 Description: do not process already freed flock
1361 Details : flock can possibly be freed by another thread before it reaches
1362 to ldlm_flock_completion_ast.
1366 Description: LBUG during stress test
1367 Details : Need properly lock accesses the flock deadlock detection list.
1371 Description: oops in page fault handler
1372 Details : kernel page fault handler can return two special 'pages' in error
1373 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
1377 Description: timeout with invalidate import.
1378 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
1379 should be handled by ptlrpcd. This produce long age waiting and
1380 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
1382 Severity : enhancement
1384 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
1388 Description: enable MGS and MDT services start separately
1389 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
1390 a co-located MGS without starting the MGS, which is a complement
1391 to 'nosvc' mount option.
1395 Frequency : after recovery
1396 Description: precreate to many object's after del orphan.
1397 Details : del orphan st in oscc last_id == next_id and this triger growing
1398 count of precreated objects. Set flag LOW to skip increase count
1399 of precreated objects.
1403 Frequency : rare, on clear nid stats
1404 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
1405 Details : when clean nid stats sometimes try destroy live entry,
1406 and this produce panic in free.
1410 Description: Stack overflow during MDS log replay
1411 ease stack pressure by using a thread dealing llog_process.
1415 Description: wait until IO finished before start new when do lock cancel.
1416 Details : VM protocol want old IO finished before start new, in this case
1417 need wait until PG_writeback is cleared until check dirty flag and
1418 call writepages in lock cancel callback.
1420 Severity : enhancement
1422 Description: using special macro for print time and cleanup in includes.
1426 Description: mds_mfd_close() ASSERTION(rc == 0)
1427 Details : In mds_mfd_close(), we need protect inode's writecount change
1428 within its orphan write semaphore to prevent possible races.
1432 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
1436 Frequency : rare, on shutdown ost
1437 Description: don't hit live lock with umount ost.
1438 Details : shrink_dcache_parent can be in long loop with destroy dentries,
1439 use shrink_dcache_sb instead.
1443 Description: don't panic with use echo client
1444 Details : echo client pass NULL as client nid pointer and this produce null
1445 pointer dereference.
1449 Description: fix build on ppc32
1450 Details : compile code with -m64 flag produce wrong object file for ppc32.
1454 Description: add message levels for liblustreapi
1458 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
1459 Details : Move the 'good_osts' check before the 'total_bavail' check. This
1460 will result in an -EAGAIN and in the exit call path we call
1461 alloc_rr() which will with increasing aggressiveness attempt to
1462 aquire precreated objects on the minimum number of required OSCs.
1466 Description: Use old size assignment to avoid deadlock
1467 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
1468 the scheduling while holding a spinlock. We do not need locking
1469 for size in ll_update_inode() because size is only updated from
1470 the MDS for directories or files without objects, so there is no
1471 other place to do the update, and concurrent access to such inodes
1472 are protected by the inode lock.
1476 Description: resolve "_IOWR redefined" build error on SLES10
1480 Description: dump the memory debugging after all modules are unloaded to
1481 suppress false negative in conf_sanity test 39
1483 Severity : enhancement
1485 Description: build kernel-ib packages for OFED 1.3 in our release cycle
1490 Description: fix SLES kernel versioning
1491 Details : the kernel version for our SLES 10 kernel did not include a "-"
1492 before the "smp" at the end. while this was not a problem in
1493 general, it did mean that software trying to use the kernel
1494 version to try to detect a vendor specific kernel would fail.
1495 this was most evident by the OFED build scripts.
1499 Description: Don't update lov_desc members until making sure they are valid
1500 Details : When updating lov_desc members via proc fs, need fix their
1501 validities before doing the real update.
1505 Description: don't put request into delay list while invalidate in flight.
1506 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
1507 import in flight. this produce timeout for invalidate and sometimes
1508 can cause stale data.
1512 Frequency : on ppc only
1513 Description: not convert ost objects for directory because it's not exist.
1514 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
1518 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
1520 Severity : enhancement
1522 Description: Update to support 2.6.22.14 vanilla kernel.
1526 Frequency : rare, on recovery
1527 Description: read procfs can produce deadlock in some situation
1528 Details : Holding lprocfs lock which send rpc can produce block for destroy
1529 obd objects and this also block reconnect with -EALREADY. This
1530 isn't fix all lprocfs bugs - but make it rare.
1532 Severity : enhancement
1534 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
1537 Frequency : frequent on X2 node
1539 Description: mdc_set_open_replay_data LBUG
1540 Details : Set replay data for requests that are eligible for replay.
1544 Description: lustre_mgs: operation 101 on unconnected MGS
1545 Details : When MGC is disconnected from MGS long enough, MGS will evict the
1546 MGC, and late on MGC cannot successfully connect to MGS and a lot
1547 of the error messages complaining that MGS is not connected.
1551 Frequency : on network error
1552 Description: panic with double free request if network error
1553 Details : mdc_finish_enqueue is finish request if any network error ocuring,
1554 but it's true only for synchronus enqueue, for async enqueue
1555 (via ptlrpcd) this incorrect and ptlrpcd want finish request
1558 Severity : enhancement
1560 Description: client-side metadata stat-ahead during readdir(directory readahead)
1561 Details : perform client-side metadata stat-ahead when the client detects
1562 readdir and sequential stat of dir entries therein
1565 Frequency : on start mds
1567 Description: Implement get_info(last_id) in obdfilter.
1570 Frequency : occasional
1572 Description: Correctly check stale fid, not start epoch if ost not support SOM
1573 Details : open with flag O_CREATE need set old fid in op_fid3 because
1574 op_fid2 was overwritten with new generated fid, but mds can answer
1575 with one of these two fids and both is not stale. Setattr
1576 incorrectly started an epoch and assume will be called
1577 done_writing, but without SOM done_writing ever being called.
1580 Frequency : rare, depends on device drivers and load
1582 Description: MDS or OSS nodes crash due to stack overflow
1583 Details : Code changes in 1.8.0 increased the stack usage of some functions.
1584 In some cases, in conjunction with device drivers that use a lot
1585 of stack the MDS (or possibly OSS) service threads could overflow
1586 the stack. One change which was identified to consume additional
1587 stack has been reworked to avoid the extra stack usage.
1590 Frequency : occasional
1592 Description: Do not fail import if osc_interpret_create gets -EAGAIN
1593 Details : If osc_interpret_create got -EAGAIN it immediately exits and
1594 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
1595 oscc_has_objects and see OSC has no objests and call
1596 oscc_internal_create to resend create request.
1598 Severity : enhancement
1600 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
1602 Severity : enhancement
1604 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
1607 Frequency : very rare
1609 Description: Fix replay if there is an un-replied request and open
1610 Details : In some cases, older replay request will revert the
1611 mcd->mcd_last_xid on MDS which is used to record the client's
1612 latest sent request.
1614 Severity : enhancement
1616 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
1618 Severity : enhancement
1620 Description: Add rhel5 support to HEAD.
1622 Serverity : enhancement
1624 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
1629 Description: Don't allow skipping OSTs if index has been specified.
1630 Details : Don't allow skipping OSTs if index has been specified, make
1631 locking in internal create lots better.
1635 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
1636 Details : access to bitfield in structure is always rounded to long
1637 and this produce problem with not atomic change any bit.
1641 Description: Lustre make rpms failed.
1642 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
1643 builds Lustre rpms from tarball.
1646 Frequency : rare on shutdown ost
1648 Description: If llog cancel was not send before clean_exports phase, this can
1649 produce deadlock in llog code.
1650 Details : If llog thread has last reference to obd and call class_import_put
1651 this produce deadlock because llog_cleanup_commit_master wait when
1652 last llog_commit_thread exited, but this never success because was
1653 called from llog_commit_thread.
1657 Description: allow userland application know is lost one of stripes.
1658 Details : fill lvb_blocks with error code on ost and return it to
1659 application if error flag found.
1663 Description: NULL lov_tgts causing MDS oops
1664 Details : more safe checks for NULL lov_tgts for avoid oops.
1666 Severity : enhancement
1668 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
1672 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
1673 Details : If an OST is down the MDS will hang indefinitely in
1674 obd_statfs_rqset() waiting for the statfs data. While for
1675 MDS QOS usage of statfs, it should not stuck in waiting.
1677 Severity : enhancement
1679 Description: remote_acl support
1680 Details : Support ACL-based permission check for remote user.
1681 Support setfacl/getfacl for remote user with the utils
1682 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
1683 the system "{s,g}etfacl" utils.
1685 Severity : enhancement
1687 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
1689 Severity : enhancement
1691 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
1695 Description: make mgs_setparam() handle fsname containing dash
1696 Details : fsname containing a dash does not work with lctl conf_param
1698 Severity : enhancement
1700 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
1702 Severity : enhancement
1704 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
1708 Frequency : rare, at shutdown
1709 Description: access already free / zero obd_namespace.
1710 Details : if client_disconnect_export was called without force flag set,
1711 and exist connect request in flight, this can produce access to
1712 NULL pointer (or already free pointer) when connect_interpret
1713 store ocd flags in obd_namespace.
1717 Frequency : only at startup
1718 Description: not alloc memory with spinlock held.
1719 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
1720 if any spinlock held.
1722 Severity : enhancement
1724 Description: make lustre randomly fail allocating memory
1725 Details : Make lustre randomly failed allocating memory for testing purpose.
1727 Severity : enhancement
1729 Description: lost problems with lov objid file
1730 Details : Fixes some scability and access to not inited memory problems
1731 in work with lov objdid file.
1736 Description: lfs find does not continue on file error
1737 Details : Continue other files processing when a file/dir is absent.
1741 Description: Inconsistent usage of lustre_pack_reply()
1742 Details : Standardize the usage of lustre_pack_reply() such that it
1743 always generate a CERROR on failure.
1746 Frequency : occasional
1748 Description: MDS hang or stay in waiting lock
1749 Details : If client receive lock with CBPENDING flag ldlm need send lock
1750 cancel as separate rpc, to avoid situation when cancel request
1751 can't processed due all i/o threads stay in wait lock.
1755 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1759 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1763 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1767 Description: Update to RHEL4 latest kernel.
1769 Severity : enhancement
1771 Description: Build SLES10 patchless client fails
1772 Details : The configure was broken by run ./configure with
1773 --with-linux-obj=.... argument for patchless client. When the
1774 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1775 can't search header adequately. Use absolute path such as
1776 -I($LINUX)/include instead.
1780 Description: interrupt oig_wait produce painc on resend.
1781 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1782 private set, and this produce situation when rq_ptlrpcd_data not
1783 copyed to new allocated request and triggered LBUG on assert
1784 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1785 ptlrpcd set if request is changed and can be safety changed to use
1790 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1791 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1792 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1793 these macros and replace them with their difinitions instead.
1797 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1798 Details : Update which_patch & target file for SLES10 latest kernel.
1801 Description: Remove client patches from SLES 10 kernel.
1802 Details : This causes SLES 10 clients to behave as patchless clients
1803 even on a Lustre-patched (server) kernel.
1805 Severity : enhancement
1807 Description: self-adjustable client's lru lists
1808 Details : use adaptive algorithm for managing client cached locks lru
1809 lists according to current server load, other client's work
1810 pattern, memory activities, etc. Both, server and client
1811 side namespaces provide number of proc tunables for controlling
1814 Severity : enhancement
1816 Description: light-weight GSS support
1817 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
1818 and reduce performance overhead.
1820 Severity : enhancement
1822 Description: Linux keyring support
1823 Details : Support using service of Linux keyring for Lustre GSS internal
1824 context refresh/cache mechanism.
1828 Description: Fix errors in lfs documentation
1829 Details : Fixes man pages
1833 Description: don't use GFP_* in generic Lustre code.
1834 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
1838 Description: obdclass is limited by single OBD_ALLOC(idarray)
1839 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
1843 Description: warnings with build patchless client with vanila 2.6.19 and up
1844 Details : change old ctl_table style and replace ctl_table/ctl_table_header
1845 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
1849 Frequency : Only for SLES
1850 Description: sanity_quota fail test_1
1851 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
1852 file, which makes TSTID[2] inunique.
1856 Frequency : Only for SLES9
1857 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
1858 Details : Update target/ChangeLog/which_patch .
1862 Frequency : Only for test_10 in sanity-quota.sh
1863 Description: a bug in quota test code
1864 Details : checking a test flag in a wrong place
1868 Frequency : Only for test_13 in sanity-quota.sh
1869 Description: a bug in quota test script
1870 Details : shouldn't check overfull allocated quota
1874 Description: jbd statistics
1875 Details : Port older jbd statistics patch for sles10
1877 Severity : enhancement
1879 Description: Add printing inode info into message about error in writepage.
1884 Description: reply_lock_interpret crash due to race with it and lock cancel.
1885 Details : Do not replay locks that are being cancelled. Do not reference
1886 locks by their address during replay, just by their handle.
1890 Frequency : When flocks are used.
1891 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
1892 Details : Flock locks might destroy just granted lock if it could be merged
1893 with another existing flock, this is done in completion handler,
1894 so teach ldlm_cli_enquque_fini that this is a valid case for
1900 Description: Oops in read and write path when failing to allocate lock.
1901 Details : Check if lock allocation failed and return error back.
1905 Description: lstripe command fails for valid OST index
1906 Details : The stripe offset is compared to lov->desc.ld_tgt_count
1907 instead of lov->desc.ld_active_tgt_count.
1911 Description: sanity.sh failed test 103
1912 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
1917 Description: df doesn't work properly if diskfs blocksize != 4K
1918 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
1922 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
1923 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
1925 Severity : enhancement
1927 Description: Adaptive timeouts
1928 Details : RPC timeouts adapt to changing server load and network
1929 conditions to reduce resend attempts and improve recovery time.
1933 Description: llapi_file_create() does not allow some changes
1934 Details : add llapi_file_open() that allows specifying the mode and
1935 open flags, and also returns an open file handle.
1939 Description: merge and cleanup kernel patches.
1940 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
1944 Description: Add journal checksum support.(Kernel part)
1945 Details : The journal checksum feature adds two new flags i.e
1946 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
1947 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
1948 indicates that the commit block contains the checksum for
1949 the blocks described by the descriptor blocks. Now commit
1950 record can be sent to disk without waiting for descriptor
1951 blocks to be written to disk. This behavior is controlled
1952 using JBD2_FEATURE_ASYNC_COMMIT flag.
1956 Description: OSS needs mutliple precreate threads
1957 Details : Add ability to start more than one create thread per OSS.
1961 Description: Sanity.sh test_65k failed: setstripe should have succeeded
1962 Details : execute lfs setstripe on client
1966 Description: mds_obd_create error creating tmp object
1967 Details : When the user sets quota on root, llog will be affected and can't
1968 create files and write files.
1971 Frequency : Always on ia64 patchless client, and possibly others.
1973 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
1974 Details : This allows the patchless client to be loaded on architectures
1975 without this export.
1979 Description: RedHat Update kernel for RHEL5
1980 Details : Add the kernel config file for RHEL5.
1984 Description: RedHat Update kernel for RHEL5
1985 Details : Modify the kernel config file more closer RHEL5.
1989 Description: Build failure against Centos5 (RHEL5)
1990 Details : Define PAGE_SIZE when it isn't present.
1994 Description: test_7 on sanity-quota.sh can't be used on separate machines
1995 Details : add facet to handle it
1999 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
2000 Details : with NFS, the anon dentry's parent was set to itself in
2001 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
2002 corresponding dentry other than use rec->ur_name.
2004 Severity : enhancement
2006 Description: lfs setstripe enhancement
2007 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
2011 Description: enable data checksumming by default
2012 Details : enable checksum by default, allow --disable-checksum
2013 configure option and "-o nochecksum" mount option. Checksums
2014 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
2018 Description: lustre support for RHEL5
2019 Details : Add support for RHEL5.
2023 Description: Client eviction due to failover config
2024 Details : after a connection loss, the lustre client should attempt to
2025 reconnect to the last active server first before trying the
2026 other potential connections.
2030 Description: when mds and osts use different quota unit(32bit and 64bit),
2031 quota will be released repeatly.
2032 Details : void sending multiple quota reqs to mds, which will keep the
2033 status between the reqs.
2037 Description: osts not allocated evenly to files
2038 Details : change the condition to increase offset_idx
2041 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
2043 Description: Data corruption for OSTs that are formatted larger than 2TB
2045 Details : When generating the bio request for lustre file writes the
2046 sector number would overflow a temporary variable before being
2047 used for the IO. The data reads correctly from Lustre (which
2048 will overflow in a similar manner) but other file data or
2049 filesystem metadata may be corrupted in some cases.
2053 Description: Tune the kernel for good SCSI performance.
2054 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
2055 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
2060 Description: rewrite ext2-derived code in obdclass/uuid.c
2061 Details : rewrite inherited code (uuid parsing code from ext2 utils)
2062 from scratch preserving functionality.
2066 Description: "lfs find -obd UUID" prints directories
2067 Details : "lfs find -obd UUID" will return all directory names instead
2068 of just file names. It is incorrect because the directories
2069 do not reside on the OSTs.
2073 Description: use i_size_read and i_size_write in 2.6 port
2074 Details : replace inode->i_size access with i_size_read/write()
2077 Frequency : when removing large files
2079 Description: scheduling issue during removal of large Lustre files
2080 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
2081 It causes scheduling issues when removing large files (17TB in the
2085 Frequency : only with liblustre clients on XT3
2087 Description: evictions taking too long
2088 Details : allow llrd to evict clients directly on OSTs
2091 Frequency : only on ppc
2093 Description: /proc/fs/lustre/devices broken on ppc
2094 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
2095 We should make sure the type of 'index' is loff_t and then cast
2096 explicitly as needed below. Do not assign an explicitly cast
2100 Frequency : only for rhel5
2102 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
2103 Details : Modify the target file & which_kernel.
2105 Severity : enhancement
2107 Description: omit set fsid for export NFS
2108 Details : fix set/restore device id for avoid EMFILE error and mark lustre
2109 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
2113 Frequency : Always, for kernels after 2.6.16
2114 Description: Fix warning idr_remove called for id=.. which is not allocated.
2115 Details : Last kernels save old s_dev before kill super and not allow
2116 to restore from callback - restore it before call kill_anon_super.
2120 Description: buffer overruns could theoretically occur
2121 Details : llapi_semantic_traverse() modifies the "path" argument by
2122 appending values to the end of the origin string, and a
2123 overrun may occur. Adding buffer overrun check in liblustreapi.
2127 Description: Fix error on 'ls .' at the top of the Lustre mount.
2128 Details : Don't revalidate dentry if it is a root dentry.
2132 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
2133 Details : Modify vm-tunables-rhel4.patch.
2137 Description: Kernel config for 2.6.18-vanilla.
2138 Details : Modify targets/2.6-vanilla.target.in.
2139 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
2140 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
2141 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
2142 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
2146 Description: improve handling recoverable errors
2147 Details : if request processig with error which can be recoverable on server
2148 request should be resend, otherwise page released from cache and
2153 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
2154 Details : Modify target file & which_patch.
2155 A flaw was found in the IA32 system call emulation provided
2156 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
2157 value could be stored in the %RAX register, which could trigger an
2158 out-of-bounds system call table access. An untrusted local user
2159 could exploit this flaw to run code in the kernel
2160 (ie a root privilege escalation). (CVE-2007-4573).
2164 Description: O_DIRECT bypasses client statistics.
2165 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
2166 read_ahead_stats were updated. copied the stats section from
2167 osc_send_oap_rpc() into async_internal().
2171 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
2175 Description: change order of libsysio includes
2176 Details : '#include sysio.h' should always come before '#include xtio.h'
2180 Description: handle "serious error: objid * already exists" more gracefully
2181 Details : If LAST_ID value on disk is smaller than the objects existing in
2182 the O/0/d* directories, it indicates disk corruption and causes an
2183 LBUG(). If the object is 0-length, then we should use the existing
2184 object. This will help to avoid a full fsck in most cases.
2186 Severity : enhancement
2188 Description: adapt the lustre_config script to support the upgrade case
2189 Details : Add "-u" option for lustre_config script to support upgrading 1.4
2190 server targets to 1.6 in parallel.
2194 Description: To avoid grant space > avaible space when the disk is almost
2195 full. Without this patch you might see the error "grant XXXX >
2196 available" or some LBUG about grant, when the disk is almost
2198 Details : In filter_check_grant, for non_grant cache write, we should
2199 check the left space by if (*left > ungranted + bytes), instead
2200 of (*left > ungranted), because only we are sure the left space
2201 is enough for another "bytes", then the ungrant space should be
2202 increase. In client, we should update cl_avail_grant only there
2203 is OBD_MD_FLGRANT in the reply.
2207 Description: Update RHEL 4 kernel to fix local root privilege escalation.
2208 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
2209 described in CVE-2007-4573. This problem could allow untrusted
2210 local users to gain root access.
2213 Frequency : when using O_DIRECT and quotas
2215 Description: Incorrect file ownership on O_DIRECT output files
2216 Details : block usage reported by 'lfs quota' does not take into account
2217 files that have been written with O_DIRECT.
2222 Description: touch file failed when fs is not full
2223 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
2224 otherwise we can get ENOSP while fs is not full.
2228 Description: parallel lock callbacks
2229 Details : Instead of sending blocking and completion callbacks as separated
2230 requests, adding them to a set and sending in parallel.
2233 Frequency : only for Cray XT3
2234 Bugzilla : 12829/13455
2235 Description: Changing primary group doesn't change the group lustre assigns to
2237 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
2238 overridden with the primary group provided by the group upcall,
2239 whereas the supplied fsgid can be trusted if it is in the list of
2240 supplementary groups returned by the group upcall.
2242 Severity : enhancement
2244 Description: Allow masking D_WARNING, D_ERROR messages from console
2245 Details : Console messages can now be disabled via lnet.printk.
2249 Description: User code with malformed file open parameter crashes client node
2250 Details : Before packing join_file req, all the related reference should be
2251 checked carefully in case some malformed flags cause fake
2252 join_file req on client.
2256 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
2257 Details : ldlm_completion_ast() assumes that a lock is granted when the req
2258 mode is equal to the granted mode. However, it should also check
2259 that LDLM_FL_CP_REQD is not set.
2263 Description: Heavy nfs access might result in deadlocks
2264 Details : After ELC code landed, it is now improper to enqueue any mds
2265 locks under och_sem, because enqueue might want to decide to
2266 cancel open locks for same inode we are holding och_sem for.
2270 Description: Client eviction while running blogbench
2271 Details : A lot of unlink operations with concurrent I/O can lead to a
2272 deadlock causing evictions. To address the problem, the number of
2273 oustanding OST_DESTROY requests is now throttled to
2274 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
2275 callbacks are processed in priority.
2279 Description: enable ACLs on MDS by default
2280 Details : ACLs must be enabled on MDS by default.
2283 Frequency : PPC/PPC64 only
2285 Description: conflicts between asm-ppc64/types.h and lustre_types.h
2286 Details : fix duplicated definitions between asm-ppc64/types.h and
2287 lustre_types.h on PPC.
2290 Frequency : PPC/PPC64 only
2292 Description: asm-ppc/segment.h does not exist
2293 Details : fix compile issue on PPC.
2297 Description: better handle error messages in extents code
2300 Frequency : RHEL4 only
2302 Description: mkfs is very slow on IA64/RHEL4
2303 Details : A performance regression has been discovered in the MPT Fusion
2304 driver between versions 3.02.73rh and 3.02.99.00rh. As a
2305 consequence, we have downgraded the MPT Fusion driver in the RHEL4
2306 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
2308 Severity : enhancement
2310 Description: SNMP support enhancement
2311 Details : Adding total number of sampled request for an MDS node in snmp
2314 Severity : enhancement
2316 Description: Optimize ldlm waiting list processing for PR extent locks
2317 Details : When processing waiting list for read extent lock and meeting read
2318 lock that is same or wider to it that is not contended, skip
2319 processing rest of the list and immediatelly return current
2320 status of conflictness, since we are guaranteed there are no
2321 conflicting locks in the rest of the list.
2325 Description: Time out and refuse to reconnect
2326 Details : When the failover node is the primary node, it is possible
2327 to have two identical connections in imp_conn_list. We must
2328 compare not conn's pointers but NIDs, otherwise we can defeat
2329 connection throttling.
2333 Description: port llog fixes from b1_6 into HEAD
2334 Details : Port llog reference couting and some llog cleanups from b1_6
2335 (bug 10800) into HEAD, for protect from panic and access to already
2336 free llog structures.
2340 Description: Detect stride IO mode in read-ahead
2341 Details : When a client does stride read, read-ahead should detect that and
2342 read-ahead pages according to the detected stride pattern.
2346 Description: data checksumming impacts single node performance
2347 Details : add support for several checksum algorithm. Currently, only CRC32
2348 and Adler-32 are supported. The checksum type can be changed on
2349 the fly via /proc/fs/lustre/osc/*/checksum_type.
2353 Description: use adler32 for page checksums
2354 Details : when available, use the Adler-32 algorithm instead of CRC32 for
2359 Description: build for x2 fails
2360 Details : fix compile issue on Cray systems.
2364 Description: Properly match for duplicate locks
2365 Details : Due to different lock order from skiplists code, we need to
2366 traverse entire list for now
2369 Frequency : only on PPC/SLES10
2371 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
2372 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
2373 be defined. Add a hack in mkfs_lustre.c to work around this compile
2378 Description: LASSERT on MDS when client holding flock lock dies
2379 Details : ldlm pool logic depends on number of granted locks equal to
2380 number of released locks which is not true for flock locks, so
2381 just exclude such locks from consideration.
2385 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
2386 Details : Use fsfilt_sync() for both the whole filesystem sync and
2387 individual file sync to eliminate dangerous inode locking
2388 with I_LOCK that can lead to a deadlock.
2392 Description: performance in 1.6.3
2393 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
2394 sure that 1MB requests are not fragmented by the block layer.
2396 Severity : enhancement
2398 Description: organize the server-side client stats on per-nid basis
2399 Details : Change the structure of stats under obdfilter and mds to
2407 The "uuid"s file would list the uuids of _active_ exports.
2408 And the clear entry is to clear all stats and stale nids.
2410 Severity : enhancement
2412 Description: eliminate client locks in face of contention
2413 Details : file contention detection and lockless i/o implementation
2414 for contended files.
2418 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
2421 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
2423 Description: MDS may lose file striping (and hence file data) in some cases
2424 Details : If there are additional extended attributes stored on the MDS,
2425 in particular ACLs, SELinux, or user attributes (if user_xattr
2426 is specified for the client mount options) then there is a risk
2427 of attribute loss. Additionally, the Lustre file striping
2428 needs to be larger than default (e.g. striped over all OSTs),
2429 and an additional attribute must be stored initially in the
2430 inode and then increase in size enough to be moved to the
2431 external attribute block (e.g. ACL growing in size) for file
2436 Description: skiplist implementation simplification
2437 Details : skiplists are used to group compatible locks on granted list
2438 that was implemented as tracking first and last lock of each
2439 lock group the patch changes that to using doubly linked lists
2443 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
2444 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
2445 because LASSERT on both the data supplied by a client, and the
2446 data on disk is dangerous and incorrect.
2448 Severity : enhancement
2450 Description: Slow truncate/writes to huge files at high offsets.
2451 Details : Directly associate cached pages to lock that protect those pages,
2452 this allows us to quickly find what pages to write and remove
2453 once lock callback is received.
2457 Description: more ldlm soft lockups
2458 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
2459 starve other threads from the resource lock for a long time in
2460 case of long waiting queue, so change the debug level from
2461 D_OTHER to the less frequently used D_INFO.
2463 Severity : enhancement
2465 Description: add -gid, -group, -uid, -user options to lfs find
2469 Description: Hung threads in invalidate_inode_pages2_range
2470 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
2471 once one is completed. As a result, some RPCs are stuck in the
2472 queue and are never sent.
2476 Description: filter threads hungs on waiting journal commit
2477 Details : Cleanup filter group llog code, then only filter group llog will
2478 be only created in the MDS/OST syncing process.
2482 Description: Procfs and llog threads access destoryed import sometimes.
2483 Details : Sync the import destoryed process with procfs and llog threads by
2484 the import refcount and semaphore.
2486 Severity : enhancement
2488 Description: openlock cache of b1_6 port to HEAD
2493 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
2494 Details : If insertion of an extent fails, then discard the inode
2495 preallocation and free data blocks else it can lead to duplicate
2500 Description: don't always update ctime in ext3_xattr_set_handle()
2501 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
2502 In some cases the ctime should not be updated, for example for
2503 2.0->1.8 compatibility it is necessary to delete an xattr and it
2504 should not update the ctime.
2508 Bugzilla : 15713/16362
2509 Description: Assertion in iopen_connect_dentry in 1.6.3
2510 Details : looking up an inode via iopen with the wrong generation number can
2511 populate the dcache with a disconneced dentry while the inode
2512 number is in the process of being reallocated. This causes an
2513 assertion failure in iopen since the inode's dentry list contains
2514 both a connected and disconnected dentry.
2518 Description: assertion failure in ldlm_handle2lock()
2519 Details : fix a race between class_handle_unhash() and class_handle2object()
2520 introduced in lustre 1.6.5 by bug 13622.
2525 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
2526 Details : remove the limit of 256 scsi disks in the sd_iostat patch
2531 Description: oops in sd_iostats_seq_show()
2532 Details : unloading/reloading the scsi low level driver triggers a kernel
2533 bug when trying to access the sd iostat file.
2538 Description: Kernel panics during QLogic driver reload
2539 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
2540 patch, causing memory corruption.
2545 Description: journal_dev option does not work in b1_6
2546 Details : pass mount option during pre-mount.
2548 Severity : enhancement
2550 Description: Add a FIEMAP(FIle Extent MAP) ioctl
2551 Details : FIEMAP ioctl will allow an application to efficiently fetch the
2552 extent information of a file. It can be used to map logical blocks
2553 in a file to physical blocks in the block device.
2557 Description: LDLM soft lockups - improvement
2558 Details : It is be possible to send the lock handle along with each read
2559 or write request because the client is already doing a lock match
2560 itself so there isn't any reason the OST should have to re-do that
2564 Frequency : only on Cray X2
2566 Description: X2 build failures
2567 Details : fix build failures on Cray X2.
2571 Description: xid & resent requests
2572 Details : Initialize RPC XID from clock at startup (randomly if clock is
2575 Severity : enhancement
2577 Description: Add lustre_start utility to start or stop multiple Lustre servers
2582 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
2583 Details : In case of memory pressure, list_del() can be called twice on
2584 req->rq_history_list, causing a kernel oops.
2588 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
2589 Details : fix stack overflow in the distributed lock manager by defering
2590 export eviction after a failed AST to the elt thread instead of
2591 handling it in the dlm interpret routine.
2595 Description: Convert some comments to new format.
2596 Details : Update documenting comments to match doxygen conventions.
2600 Description: Grammar fixes.
2601 Details : A couple of trivial spelling fixes.
2605 Description: OSD_COUNTERS-mandatory
2606 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
2607 to simplify the code.
2611 Description: simplify cmm_device freeing logic.
2612 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
2616 Description: Add lockdep support to dt_object_operations locking interface.
2617 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
2618 parameter indicating lock ordering. Update mdd code to use new
2623 Description: Introduce failloc constants for lockless IO tests.
2624 Details : Add two new failloc constants to test lockless IO. Only one of
2625 them in implemented---another is checked in yet to be landed
2630 Description: Add lockdep support for inode mutex.
2631 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
2632 in the situations where Lustre has to lock more than one inode
2637 Description: Add optional invariants checking support.
2638 Details : Add new LINVRNT() macro, optional on new --enable-invariants
2639 configure switch. This macro is to be used for consistency and
2640 sanity checks that are too expensive to be left in `production'
2645 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
2646 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
2651 Description: Add lockdep support for ldlm_lock and ldlm_resource.
2652 Details : Use spin_lock_nested() in (the only) situation where more than
2653 one ldlm_lock is locked simultaneously. Also, fix possible
2654 dead-lock in ldlm_lock_change_resource() by enforcing particular
2659 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
2660 Details : Instead of specifying each ldlm_lock call-back through separate
2661 parameter, wrap them into struct ldlm_callback_suite.
2665 Description: Kill join_lru obd method and its callers.
2666 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
2667 in memory---a requirement implemented through ->o_join_lru() obd
2668 method in HEAD. Remove this method and its users.
2672 Description: Add asynchronous ldlm ENQUEUE completion handler.
2673 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
2674 case that stock ldlm_completion_ast() cannot handle as it waits
2675 until lock is granted. Introduce new ldlm_completion_ast_async()
2676 for this. Also comment ldlm_completion_ast().
2680 Description: ldlm_error <-> errno conversion.
2681 Details : Add functions to map (rather arbitrary) between LDLM error codes
2682 and standard errno values. CLIO needs this to prevent LDLM specific
2683 constants from escaping ldlm and osc.
2687 Description: Kill unused ldlm_handle2lock_ns() function.
2688 Details : Kill unused ldlm_handle2lock_ns() function.
2692 Description: Add lu_ref support to ldlm_lock
2693 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
2694 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
2695 and ldlm_resource. LDLM interface has to be changed, because code that
2696 releases a reference on a lock, has to "know" what reference this is.
2697 In the most frequent case
2699 lock = ldlm_handle2lock(handle);
2701 LDLM_LOCK_PUT(lock);
2703 no changes are required. When any other reference (received _not_ from
2704 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
2705 instead of LDLM_LOCK_PUT().
2707 Arguably, changes are pervasive, and interface requires some discipline
2708 for proper use. On the other hand, it was very instrumental in finding
2709 a few leaked lock references.
2713 Description: Add ldlm_lock_addref_try().
2714 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
2715 attempts to addref a lock that might be being canceled
2720 Description: Add ldlm_weigh_callback().
2721 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
2722 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
2723 hard-coded `number of pages' logic.
2727 Description: Add lockdep annotations to llog code.
2728 Details : Use appropriately tagged _nested() locking calls in the places
2729 where llog takes more than one ->lgh_lock lock.
2733 Description: Add loi_kms_set().
2734 Details : Wrap kms updates into a helper function.
2738 Description: Constify instances of struct lsm_operations.
2739 Details : Constify instances of struct lsm_operations.
2743 Description: lu_conf support.
2744 Details : On a server, a file system object is uniquely identified
2745 by a fid, which is sufficient to locate and load all object
2746 state (inode). On a client, on the other hand, more data are
2747 necessary instantiate an object. Change lu_object_find() and
2748 friends to take additional `lu_conf' argument describing object.
2749 Typically this includes layout information.
2753 Description: lu_context fixes.
2754 Details : Introduce new lu_context functions that are needed on the client
2755 side, where some system threads (ptlrpcd) are shared by multiple
2756 modules, and so cannot be stopped during module shutdown.
2760 Description: Add start and stop methods to lu_device_type_operations.
2761 Details : Introduce two new methods in lu_device_type_operations, that are
2762 invoked when first instance of a given type is created and last one
2763 is destroyed respectively. This is need by CLIO.
2767 Description: Add lu_ref support to struct lu_device.
2768 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
2769 track leaked references.
2773 Description: Introduce lu_kmem_descr.
2774 Details : lu_kmem_descr and its companion interface allow to create
2775 and destroy a number of kmem caches at once.
2779 Description: Fix lu_object finalization race.
2780 Details : Fix a race between lu_object_find() finding an object and its
2781 concurrent finalization. This race is (most likely) not possible
2782 on the server, but might happen on the client.
2786 Description: Introduce lu_ref interface.
2787 Details : lu_ref is a debugging module allowing to track references to
2788 a given object. It is quite cpu expensive, and has to be
2789 explicitly enabled with --enable-lu_ref. See usage description
2794 Description: Factor lu_site procfs stats into a separate function.
2795 Details : Separate lu_site stats printing code into a separate function
2796 that can be reused on a client.
2800 Description: Constify instances of struct {lu,dt,md}_device_operations.
2801 Details : Constify instances of struct {lu,dt,md}_device_operations.
2805 Description: Introduce struct md_site and move meta-data specific parts of
2806 struct lu_site here.
2807 Details : Move md-specific fields out of struct lu_site into special struct
2808 md_site, so that lu_site can be used on a client.
2812 Description: Kill mdd_lov_destroy().
2813 Details : Remove unused mdd code.
2817 Description: Add st_block checking to multistat.c.
2818 Details : Add st_block checking to multistat.c.
2822 Description: Add lu_ref support to struct obd_device.
2823 Details : Add lu_ref tracking to obd_device.
2827 Description: Kill obd_set_fail_loc().
2828 Details : Remove unused code.
2832 Description: Add special type for ptlrpc_request interpret functions.
2833 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
2834 there. Actual usage will be in CLIO.
2838 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
2839 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
2840 doesn't work with lockdep.
2844 Description: Add rwv.c test program.
2845 Details : New testing program exercising readv(2) and writev(2) (Qian).
2849 Description: Add sendfile.c test program.
2850 Details : New testing program exercising sendfile(2) (Jay).
2854 Description: Ratelimit a message that can be very frequent.
2855 Details : Ratelimit a memory allocation failure message that can
2860 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2861 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2865 Description: Liblustre build fixes.
2866 Details : Liblustre build fixes.
2870 Description: libcfs: add cfs_{need,cond}_resched() interface.
2871 Details : libcfs: add cfs_{need,cond}_resched() definition and
2872 implementations for Linux, NT, and liblustre.
2874 Severity : enhancement
2876 Description: More exported tunables for mballoc
2877 Details : Add support for tunable preallocation window and new tunables for
2878 large/small requests
2882 Description: Detect corruption of block bitmap and checking for preallocations
2883 Details : Checks validity of on-disk block bitmap. Also it does better
2884 checking of number of applied preallocations. When corruption is
2885 found, it turns filesystem readonly to prevent further corruptions.
2889 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
2890 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
2891 window to 0. If it does hit cache miss, and read-pattern is still
2892 stride-io mode, does not reset the stride window, but also does not
2893 increase the stride window length in this case.
2897 Frequency : only for big-endian servers
2898 Description: Check if system is big-endian while mounting fs with extents feature
2899 Details : Mounting a filesystem with extents feature will fail on big-endian
2900 systems since ext3-based ldiskfs is not supported on big-endian
2901 systems. This can be over-riden with "bigendian_extents" mount option.
2903 Severity : enhancement
2905 Description: The root squash functionality
2906 Details : A security feature, which is to prevent users from being able
2907 to mount lustre on their desktop, run as root, and delete
2908 all of the files in the filesystem. The goal is accomplished by
2909 remapping user id (UID) and group id (GID) of the root user to
2910 a UID and GID specified by the system administartor via Lustre
2911 configuration management server (MGS). The functionality also
2912 allows to specify sets of clients for which the remapping does
2917 Description: Excessive recovery window
2918 Details : With AT enabled, the recovery window can be excessively long (6000+
2919 seconds). To address this problem, we no longer use
2920 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
2921 timeout no longer depends on the service time, it is set to
2922 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
2923 time via pb_service_time.
2927 Description: Watchdog triggered on MDS failover
2928 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
2929 the OSTs know that the MDS "UUID" can be reused for the same export
2930 from a different NID, so we do not need to wait for the export to be
2934 Frequency : rare, only if using MMP with Linux RAID
2936 Description: MMP doesn't work with Linux RAID
2937 Details : While using HA for Lustre servers with Linux RAID, it is possible
2938 that MMP will not detect multiple mounts. To make this work we
2939 need to unplug the device queue in RAID when the MMP block is being
2940 written. Also while reading the MMP block, we should read it from
2941 disk and not the cached one.
2943 Severity : enhancement
2945 Description: open file using fid
2946 Details : A file can be opened using just its fid, like
2947 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
2950 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
2953 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
2954 Details : add proper locking for creation of jbd_4k slab cache
2958 Description: MMP check in ext3_remount() fails without displaying any error
2959 Details : When multiple mount protection fails during remount, proper error
2962 Severity : enhancement
2964 Description: Allow stripe size to be up to 4G-64k
2965 Details : Fix math logic to allow large stripe sizes.
2969 Description: add check for >8TB ldiskfs filesystems
2970 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
2971 Don't allow >8TB ldiskfs filesystems to be mounted without
2972 force_over_8tb mount option
2977 Descriptoin: OSS complains "too many blocks" and crashes
2978 Details : In filter_commitrw_write():
2979 1. lquota_chkquota();
2980 2. obdo_from_inode();
2981 3. lquota_pending_commit();
2982 "2." has changed uid and gid in this case, then caused
2983 lquota_chkquota() and lquota_pending_commit() to use
2984 different uid and gid. That is the root of the bug.
2988 Descriptoin: getstripe doesn't report stripe count = -1 sanely
2989 Details : Detect and print out the 'stripe count = -1' case properly.
2993 Description: Out or order replies might be lost on replay
2994 Details : In ptlrpc_retain_replayable_request if we cannot find retained
2995 request with tid smaller than one currently being added, add it
2996 to the start, not end of the list.
2998 --------------------------------------------------------------------------------
3000 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
3002 * Support for kernels:
3003 2.6.5-7.283 (SLES 9),
3004 2.6.9-55.EL (RHEL 4),
3005 2.6.16.46-0.14 (SLES 10),
3006 2.6.18.8 vanilla (kernel.org)
3007 * Client support for unpatched kernels:
3008 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
3009 2.6.16 - 2.6.22 vanilla (kernel.org)
3010 * Due to recently discovered recovery problems, we do not recommend
3011 using patchless RHEL 4 clients with this or any earlier release.
3012 * Recommended e2fsprogs version: 1.39.cfs8
3013 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
3014 * Starting with this release, the ldiskfs backing filesystem required
3015 by Lustre is now in its own package, lustre-ldiskfs. This package
3016 should be installed. It is versioned separately from Lustre and
3017 may be released separately in future.
3021 Description: block reactivating mgc import until all deactivates complete
3022 Details : Fix race when failing back MDT/MGS to itself (testing)
3025 Frequency : at statup only
3027 Description: mds_lov_synchronize race leads to various problems
3028 Details : simultaneous MDT->OST connections at startup can cause the
3029 sync to abort, leaving the OSC in a bad state.
3031 Severity : enhancement
3033 Description: add optional extra BUILD_VERSION info
3034 Details : add a new environment variable (namely LUSTRE_VERS) which allows
3035 to override the lustre version.
3038 Frequency : 2.6.18 servers only
3040 Description: ll_kern_mount() doesn't release the module reference
3041 Details : The ldiskfs module reference count never drops down to 0
3042 because ll_kern_mount() doesn't release the module reference.
3047 Description: server LBUG when using old ost_num_threads parameter
3048 Details : Accept the old ost_num_threads parameter but warn that it
3049 is deprecated, and fix an off-by-one error that caused an LBUG.
3054 Description: Transient SCSI error results in persistent IO issue
3055 Details : iobuf->dr_error is not reinitialized to 0 between two
3059 Frequency : sometimes when underlying device returns I/O errors
3061 Description: OSTs not going read-only during write failures
3062 Details : OSTs are not remounted read-only when the journal commit threads
3063 get I/O errors because fsfilt_ext3 calls journal_start/stop()
3064 instead of the ext3 wrappers.
3068 Description: poor connect scaling with increasing client count
3069 Details : Don't run filter_grant_sanity_check for more than 100 exports
3070 to improve scaling for large numbers of clients.
3073 Frequency : SLES10 only
3075 Description: sanity-quota.sh quotacheck failed: rc = -22
3076 Details : Quotas cannot be enabled on SLES10.
3079 Frequency : liblustre clients only
3081 Description: getdirentries does not give error when run on compute nodes
3082 Details : getdirentries does not fail when the size specified as an argument
3083 is too small to contain at least one entry
3085 Severity : enhancement
3087 Description: Add LNET router traceability for debug purposes
3088 Details : If a checksum failure occurs with a router as part of the
3089 IO path, the NID of the last router that forwarded the bulk data
3090 is printed so it can be identified.
3095 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
3096 Details : Due to a race condition, liblustre clients were occasionally
3097 evicted incorrectly.
3099 Severity : enhancement
3101 Description: lfs setstripe use optional parameters instead of postional
3104 Severity : enhancement
3106 Description: Nanosecond timestamp support for ldiskfs
3107 Details : The on-disk ldiskfs filesystem has added support for nanosecond
3108 resolution timestamps. There is not yet support for this at
3109 the Lustre filesystem level.
3112 Frequency : during server recovery
3114 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
3115 Details : request with rq_no_resend flag not awake l_wait_event if they get
3119 Frequency : nfs export on patchless client
3121 Description: connectathon hang when test nfs export over patchless client
3122 Details : Disconnected dentry cannot be found with lookup, so we do not need
3123 to unhash it or make it invalid
3126 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
3127 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
3128 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
3133 Description: open req refcounting wrong on reconnect
3134 Details : If reconnect happened between getting open reply from server and
3135 call to mdc_set_replay_data in ll_file_open, we will schedule
3136 replay for unreferenced request that we are about to free.
3137 Subsequent close will crash in variety of ways.
3138 Check that request is still eligible for replay in
3139 mdc_set_replay_data().
3144 Description: disable writes to filesystem when reading health_check file
3145 Details : the default for reading the health_check proc file has changed
3146 to NOT do a journal transaction and write to disk, because this
3147 can cause reads of the /proc file to hang and block HA state
3148 checking on a healthy but otherwise heavily loaded system. It
3149 is possible to return to the previous behaviour during configure
3150 with --enable-health-write.
3152 Severity : enhancement
3154 Description: 64-bit inode version
3155 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
3156 to the inode. This will be required for version-based recovery.
3161 Description: MDS fails to start if a duplicate client export is detected
3162 Details : in some rare cases it was possible for a client to connect to
3163 an MDS multiple times. Upon recovery the MDS would detect this
3164 and fail during startup. Handle this more gracefully.
3166 Severity : enhancement
3168 Description: Add -o localflock option to simulate old noflock
3170 Details : This will achieve local-only flock/fcntl locks
3176 Description: log_commit_thread vs filter_destroy race leads to crash
3177 Details : Take import reference before releasing llog record semaphore
3182 Description: Wrong request locking in request set processing
3183 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
3184 imp_delayed_list, in this place should be used imp_lock.
3187 Frequency : when reconnection
3189 Description: Grant Leak when osc reconnect to OST
3190 Details : When osc reconnect ost, OST(filter) should check whether it
3191 should grant more space to client by comparing fed_grant and
3192 cl_avail_grant, and return the granted space to client instead
3193 of "new granted" space, because client will call osc_init_grant
3194 to update the client grant space info.
3197 Frequency : when client reconnect to OST
3199 Description: Grant Leak when osc do resend and replay bulk write
3200 Details : When osc reconnect to OST, OST(filter)should clear grant info of
3201 bulk write request, because the grant info while be sync between
3202 OSC and OST when reconnect, and we should ignore the grant info
3203 these of resend/replay write req.
3208 Description: Grant space more than available left space sometimes.
3209 Details : When then OST is about to be full, if two bulk writing from
3210 different clients came to OST. Accord the avaliable space of the
3211 OST, the first req should be permitted, and the second one
3212 should be denied by ENOSPC. But if the seconde arrived before
3213 the first one is committed. The OST might wrongly permit second
3214 writing, which will cause grant space > available space.
3217 Frequency : when client is evicted
3219 Description: Grant might be wrongly erased when osc is evicted by OST
3220 Details : when the import is evicted by server, it will fork another
3221 thread ptlrpc_invalidate_import_thread to invalidate the
3222 import, where the grant will be set to 0. While the original
3223 thread will update the grant it got when connecting. So if
3224 the former happened latter, the grant will be wrongly errased
3225 because of this race.
3230 Description: Checking Stale with correct fid
3231 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
3232 to check whether it is stale, because sometimes, we want the
3233 enqueue happened anyway, and op_data.fid2 will not be initialized.
3235 Severity : enhancement
3237 Description: update patchless client
3238 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
3241 Frequency : only with 2.4 kernel
3243 Description: random memory corruption
3244 Details : size of struct ll_inode_info is to big for union inode.u and this
3245 can be cause of random memory corruption.
3250 Description: Memory leak in recovery
3251 Details : Lov_mds_md was not free in an error handler in mds_create_object.
3252 It should also check obd_fail before fsfilt_start, otherwise if
3253 fsfilt_start return -EROFS,(failover mds during mds recovery).
3254 then the req will return with repmsg->transno = 0 and rc = EROFS.
3255 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
3256 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
3257 be freed no matter whether fsfilt_commit success or not.
3260 Frequency : only with huge count clients
3262 Description: Prevents from taking the superblock lock in llap_from_page for
3264 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
3265 for llap_from_page call in ll_removepage prevents from taking the
3266 superblock lock for a soon died page.
3271 Description: Not check open intent error before release open handle
3272 Details : in some rare cases, the open intent error is not checked before
3273 release open handle, which may cause
3274 ASSERTION(open_req->rq_transno != 0), because it tries to release
3275 the failed open handle.
3280 Description: Set cat log bitmap only after create log success.
3281 Details : in some rare cases, the cat log bitmap is set too early. and it
3282 should be set only after create log success.
3286 Description: Accessing a block bevice can re-enable I/O when Lustre is
3287 tearing down a device.
3288 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
3293 Description: service threads may hog cpus when there are a lot of requests
3294 Details : Insert cond_resched to give other threads a chance to use some CPU
3299 Description: the cat log was not initialized in recovery
3300 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
3301 unlink log on mds will not be initialized until mds post
3302 recovery. And also in mds post recovery, the unlink log will
3303 initialization will be done asynchronausly, so there will be race
3304 between add unlink log and unlink log initialization.
3308 Description: brw_stats were being printed incorrectly
3309 Details : brw_stats were being printed as log2 but all of them were not
3310 recorded as log2. Also remove some code duplication arising from
3311 filter_tally_{read,write}.
3315 Frequency : rare, only in recovery.
3316 Description: ASSERTION(req->rq_type != LI_POISON) failed
3317 Details : imp_lock should be held while iterating over imp_sending_list for
3318 prevent destroy request after get timeout in ptlrpc_queue_wait.
3322 Description: replay-single.sh test 52 fails
3323 Details : A lock's skiplist need to be cleanup when it being unlinked
3324 from its resource list.
3328 Description: Short directio read returns full requested size rather than
3330 Details : Direct I/O operations should return actual amount of bytes
3331 transferred rather than requested size.
3333 Severity : enhancement
3335 Description: metadata RPC reduction (e.g. for rm performance)
3336 Details : decrease the amount of synchronous RPC between clients and servers
3337 by canceling conflicing lock before the operation on the client
3338 and packing thier handles into the main operation RPC to server.
3340 Severity : enhancement
3342 Description: Async OSC create to avoid the blocking unnecessarily.
3343 Details : If an OST has no remaining object, system will block on the
3344 creation when it needs to create a new object on this OST. Now,
3345 ways use pre-created objects when available, instead of blocking on
3346 an empty osc while others are not empty. If we must block, we
3347 block for the shortest possible period of time.
3351 Description: improve handling recoverable errors
3352 Details : if request processig with error which can be recoverable on server
3353 request should be resend, otherwise page released from cache and
3356 Severity : enhancement
3358 Description: refine locking for avoid write wrong info into lov_objid file
3359 Details : fix possible races with add new target and write/update data in
3362 --------------------------------------------------------------------------------
3364 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
3369 Frequency : on some architectures
3371 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
3372 Details : Uninitialized flags sometimes cause configuration commands to
3376 Frequency : patchless clients only
3378 Description: missing __iget() symbol export
3379 Details : The __iget() symbol export is missing. To avoid the need for
3380 this on patchless clients the deathrow inode reaper is turned
3381 off, and we depend on the VM to clean up old inodes. This
3382 dependency was during via the fix for bug 12181.
3384 --------------------------------------------------------------------------------
3386 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
3388 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
3389 INTEROPERATE with older versions automatically. In many cases a
3390 special upgrade step is needed. Please read the
3391 user documentation before upgrading any part of a live system.
3392 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
3393 version will not interoperate with 1.6 betas before beta5 (1.5.95).
3394 * WARNING: Lustre configuration and startup changes are required with
3395 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
3399 Severity : enhancement
3401 Description: MountConf
3402 Details : Lustre configuration is now managed via mkfs and mount
3403 commands instead of lmc and lconf. New obd types (MGS, MGC)
3404 are added for dynamic configuration management. See
3405 https://mail.clusterfs.com/wikis/lustre/MountConf for
3408 Severity : enhancement
3410 Description: dynamic OST addition
3411 Details : OSTs can now be added to a live filesystem
3413 Severity : enhancement
3415 Description: startup order invariance
3416 Details : MDTs and OSTs can be started in any order. Clients only
3417 require the MDT to complete startup.
3419 Severity : enhancement
3421 Description: parallel, asynchronous orphan cleanup
3422 Details : orphan cleanup is now performed in separate threads for each
3423 OST, allowing parallel non-blocking operation.
3425 Severity : enhancement
3427 Description: optimized stripe assignment
3428 Details : stripe assignments are now made based on ost space available,
3429 ost previous usage, and OSS previous usage, in order to try
3430 to optimize storage space and networking resources.
3432 Severity : enhancement
3434 Description: Permanently set tunables
3435 Details : All writable /proc/fs/lustre tunables can now be permanently
3436 set on a per-server basis, at mkfs time or on a live system.
3438 Severity : enhancement
3440 Description: Lustre message v2
3441 Details : Add lustre message format v2.
3443 Severity : enhancement
3445 Description: client OST exclusion list
3446 Details : Clients can be started with a list of OSTs that should be
3447 declared "inactive" for known non-responsive OSTs.
3451 Description: SPEC SFS validation failure on NFS v2 over lustre.
3452 Details : Changes the blocksize for regular files to be 2x RPC size,
3453 and not depend on stripe size.
3455 Severity : enhancement
3457 Description: Multiple MD RPCs in flight.
3458 Details : Further unserialise some read-only MDS RPCs - learn about intents.
3459 To avoid overly-overloading MDS, introduce a limit on number of
3460 MDS RPCs in flight for a single client and add /proc controls
3461 to adjust this limit.
3463 Severity : enhancement
3465 Description: client read/write statistics
3466 Details : Add client read/write call usage stats for performance
3467 analysis of user processes.
3468 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
3469 file access. extents_stats shows chunk size distribution.
3470 extents_stats_per_process show chunk size distribution per
3473 Severity : enhancement
3475 Description: mds statistics
3476 Details : Add detailed mds operations statistics in
3477 /proc/fs/lustre/mds/*/stats.
3481 Description: Failure of copying files with lustre special EAs.
3482 Details : Client side always return success for setxattr call for lustre
3483 special xattr (currently only "trusted.lov").
3488 Description: Refcount LNET uuids
3489 Details : The global LNET uuid list grew linearly with every startup;
3490 refcount repeated list entries instead of always adding to
3493 Severity : enhancement
3495 Description: Dynamic service threads
3496 Details : Within a small range, start extra service threads
3497 automatically when the request queue builds up.
3500 Frequency : mixed-endian client/server environments
3502 Description: mixed-endian crashes
3503 Details : The new msg_v2 system had some failures in mixed-endian
3506 Severity : enhancement
3508 Description: Easy OST removal
3509 Details : OSTs can be permanently deactivated with e.g. 'lctl
3510 conf_param lustre-OST0001.osc.active=0'
3512 Severity : enhancement
3514 Description: MGS proc entries
3515 Details : Added basic proc entries for the MGS showing what filesystems
3518 Severity : enhancement
3520 Description: provide MGS failover
3521 Details : Added config lock reacquisition after MGS server failover.
3523 Severity : enhancement
3525 Description: add Linux 2.4 support
3526 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
3530 Description: a large application tries to do I/O to the same resource and dies
3531 in the middle of it.
3532 Details : Check the req->rq_arrival time after the call to
3533 ost_brw_lock_get(), but before we do anything about
3534 processing it & sending the BULK transfer request. This
3535 should help move old stale pending locks off the queue as
3536 quickly as obd_timeout.
3539 Frequency : when an incorrect nid is specified during startup
3541 Description: ptlrpc connect to non-existant node causes kernel crash
3542 Details : LNET can't be re-entered from an event callback, which
3543 happened when we expire a message after the export has been
3544 cleaned up. Instead, hand the zombie cleanup off to another
3547 Severity : enhancement
3549 Description: plain/inodebits lock performance improvement
3550 Details : Grouping plain/inodebits in granted list by their request modes
3551 and bits policy, thus improving the performance of search through
3555 Frequency : only if OST filesystem is corrupted
3557 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3558 Details : for a short time RPCs with bulk IO are in the replay list,
3559 but replay of bulk IOs is unimplemented. If the OST filesystem
3560 is corrupted due to disk cache incoherency and then replay is
3561 started it is possible to trip an assertion. Avoid putting
3562 committed RPCs into the replay list at all to avoid this issue.
3565 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3567 Description: System hang on startup
3568 Details : This bug allowed the liblustre (e.g. catamount) client to
3569 return to the app before handling all startup RPCs. This
3570 could leave the node unresponsive to lustre network traffic
3571 and manifested as a server ptllnd timeout.
3573 Severity : enhancement
3575 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3576 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
3577 boolean to control whether to print peer debug info when a
3578 client's RPC times out.
3581 Frequency : only for kernels with patches from Lustre below 1.4.3
3583 Description: Remove old rdonly API
3584 Details : Remove old rdonly API which unused from at least lustre 1.4.3
3587 Frequency : only for devices with external journals
3589 Description: Set external device read-only also
3590 Details : During a commanded failover stop, we set the disk device
3591 read-only while the server shuts down. We now also set any
3592 external journal device read-only at the same time.
3595 Frequency : when upgrading from 1.4 while trying to change parameters
3597 Description: The wrong (new) MDC name was used when setting parameters for
3598 upgraded MDT's. Also allows changing of OSC (and MDC)
3599 parameters if --writeconf is specified at tunefs upgrade time.
3602 Frequency : when setting specific ost indicies
3604 Description: QOS code breaks on skipped indicies
3605 Details : Add checks for missing OST indicies in the QOS code, so OSTs
3606 created with --index need not be sequential.
3608 Severity : enhancement
3610 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3611 Details : The uninit_groups feature works in conjunction with the kernel
3612 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3613 up the pass1 processing of e2fsck. This is a read-only feature
3614 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3615 mount filesystems that have had this feature enabled.
3617 Severity : enhancement
3619 Description: Improve multi-block allocation algorithm to avoid fragmentation
3620 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3621 allocation locality and avoid filesystem fragmentation.
3623 ------------------------------------------------------------------------------
3625 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
3627 * Support for kernels:
3628 2.4.21-47.0.1.EL (RHEL 3)
3629 2.6.5-7.283 (SLES 9)
3630 2.6.9-42.0.10.EL (RHEL 4)
3631 2.6.12.6 vanilla (kernel.org)
3632 2.6.16.27-0.9 (SLES 10)
3633 * Recommended e2fsprogs version: 1.39.cfs5
3635 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3639 Frequency : occasional, depends on client load and configuration
3640 Bugzilla : 12181, 12203
3641 Description: data loss for recently-modified files
3643 Details : In some cases it is possible that recently written or created
3644 files may not be written to disk in a timely manner (this should
3645 normally be within 30s unless client IO load is very high).
3646 The problem appears as zero-length files or files that are a
3647 multiple of 1MB in size after a client crash or client eviction
3648 that are missing data at the end of the file.
3650 This problem is more likely to be hit on clients where files are
3651 repeatedly created and unlinked in the same directory, clients
3652 have a large amount of RAM, have many CPUs, the filesystem has
3653 many OSTs, the clients are rebooted frequently, and/or the files
3654 are not accessed by other nodes after being written.
3656 The presence of the problem can be detected by looking at
3657 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
3658 smaller than the second (nr_unused) then dirty files will not
3659 be flushed automatically to disk. "sync; sleep 10" should be
3660 run several times on the node before unmounting it to update
3661 Lustre (this is also safe to run on nodes without this problem).
3663 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
3664 that can cause this same problem, so customers using that kernel
3665 also need to update the kernel in addition to Lustre. In order
3666 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
3668 It is normal that files written just before a client crash (less
3669 than 30s) may not yet have been flushed to disk, even for local
3673 Frequency : frequent on thin XT3 nodes
3675 Description: UUID collision on thin XT3 Linux nodes
3676 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
3677 randomly, since we relied on an insufficiently-seeded PRNG.
3682 Description: OSS hangs after "All ost request buffers busy"
3683 Details : A deadlock between quota and journal operations caused OSS
3684 hangs after printing "All ost request buffers busy."
3687 Frequency : always on liblustre builds
3689 Description: Cleanup compiler warnings on liblustre
3692 Frequency : always on liblustre builds on XT3
3694 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
3698 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
3699 > 32000 subdirectories
3700 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
3701 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
3702 subdirectory count crosses 32000. This will aid e2fsck to
3703 correctly handle more than 32000 subdirectories.
3706 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3708 Description: System hang on startup
3709 Details : This bug allowed the liblustre (e.g. catamount) client to
3710 return to the app before handling all startup RPCs. This
3711 could leave the node unresponsive to lustre network traffic
3712 and manifested as a server ptllnd timeout.
3714 Severity : enhancement
3716 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3717 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
3718 boolean to control whether to print peer debug info when a
3719 client's RPC times out.
3724 Description: make O_SYNC working on 2.6 kernels
3725 Details : 2.6 kernels use different method for mark pages for write,
3726 so need add a code to lustre for O_SYNC work.
3731 Description: Failure to close file and release space on NFS
3732 Details : Put inode details into lock acquired in ll_intent_file_open.
3733 Use mdc_intent_lock in ll_intent_open to properly
3734 detect all kind of errors unhandled by mdc_enqueue.
3739 Description: proc file read during shutdown sometimes raced obd removal,
3741 Details : Add lock to prevent obd access after proc file removal.
3744 Frequency : Only for files larger than 4GB on 32-bit clients.
3746 Description: improperly doing page alignment of locks
3747 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
3748 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3753 Description: RPCs being resent when they shouldn't be
3754 Details : Some RPCs that should not be resent are being resent. This
3755 can cause inconsistencies in the RPC state machine. Do not
3756 resend such requests.
3759 Frequency : rare, only with NFS export
3761 Description: Crash on NFS re-export node
3762 Details : under very unusual load conditions an assertion is hit in
3763 ll_intent_file_open()
3766 Frequency : only if OST filesystem is corrupted
3768 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3769 Details : for a short time RPCs with bulk IO are in the replay list,
3770 but replay of bulk IOs is unimplemented. If the OST filesystem
3771 is corrupted due to disk cache incoherency and then replay is
3772 started it is possible to trip an assertion. Avoid putting
3773 committed RPCs into the replay list at all to avoid this issue.
3778 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3779 Details : Large single O_DIRECT read and write calls can fail to allocate
3780 a sufficiently large buffer to process the request. In case of
3781 allocation failure the allocation is retried with a smaller
3782 buffer and broken into smaller requests.
3784 Severity : enhancement
3786 Description: Add -o localflock option to simulate old noflock behaviour.
3787 Details : This will achieve local-only flock/fcntl locks coherentness.
3792 Description: versioning check is incomplete
3793 Details : Checking the version difference of client vs. server, report
3794 error if the gap is too big.
3799 Description: add support PG_writeback bit
3800 Details : add support for PG_writeback bit for Lustre, for more carefull
3801 work with page cache in 2.6 kernel. This also fix some deadlocks
3802 and remove hack for work O_SYNC with 2.6 kernel.
3804 Severity : enhancement
3806 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3807 Details : The uninit_groups feature works in conjunction with the kernel
3808 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3809 up the pass1 processing of e2fsck. This is a read-only feature
3810 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3811 mount filesystems that have had this feature enabled.
3813 Severity : enhancement
3815 Description: Improve multi-block allocation algorithm to avoid fragmentation
3816 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3817 allocation locality and avoid filesystem fragmentation.
3819 ------------------------------------------------------------------------------
3821 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3823 * Support for kernels:
3824 2.6.9-42.0.3.EL (RHEL 4)
3825 2.6.5-7.276 (SLES 9)
3826 2.4.21-47.0.1.EL (RHEL 3)
3827 2.6.12.6 vanilla (kernel.org)
3828 2.6.16.21-0.8 (SLES10)
3829 * Recommended e2fsprogs version: 1.39.cfs2-0
3831 * The backwards-compatible /proc/sys/portals symlink has been removed
3832 in this release. Before upgrading, please ensure that you change
3833 any configuration scripts or /etc/sysctl.conf files that access
3834 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3835 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3836 in advance of the upgrade on any system running Lustre 1.4.6 or
3837 newer, since /proc/sys/lnet was added in that version.
3838 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3842 Frequency : only when quota is used
3844 Description: avoid scanning export list for quota master
3845 Details : Change the algorithms to avoid scanning export list in order
3846 to improve the efficiency.
3849 Frequency : MDS failover only, very rarely
3851 Description: "went back in time" messages on mds failover
3852 Details : The greatest transno may be lost when the current operation
3853 finishes with an error (transno==0) and the client's last_rcvd
3854 record is over-written. Save the greatest transno in the
3855 mds_last_transno for this case.
3858 Frequency : always for specific kernels and striping counts
3860 Description: client may get "Matching packet too big" without ACL support
3861 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
3862 when trying to access files in certain configurations. The
3863 clients should in fact be denied when mounting because they do
3864 not understand ACLs.
3867 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
3869 Description: many clients connecting with IO in progress causes connect timeouts
3870 Details : Avoid synchronous journal commits to avoid delays caused by many
3871 clients connecting/disconnecting when bulk IO is in progress.
3872 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
3873 OST_IO_PORTAL to avoid delays behind potentially many pending
3877 Frequency : occasionally with multiple writers to a single file
3879 Description: shared writes to file may result in wrong size reported by stat()
3880 Details : Allow growing of kms when extent lock is cancelled
3883 Frequency : always with random mmap IO to multi-striped file
3885 Description: mmap write might be lost if we are writing to a 'hole' in stripe
3886 Details : Only if the hole is at the end of OST object so that kms is too
3887 small. Fix is to increase kms accordingly in ll_nopage.
3890 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
3892 Description: writes to a missing object would leak memory on the OST
3893 Details : If there is an inconsistency between the MDS and OST filesystems,
3894 such that the MDS references an object that doesn't exist, writes
3895 to that object will leak memory due to incorrect cleanup in the
3896 error handling path, eventually running out of memory on the OST.
3901 Description: Creating too long symlink causes lustre errors
3902 Details : Check symlink and name lengths before sending requests to MDS.
3905 Frequency : only if flock is enabled (not on by default)
3907 Description: posix locks not released on fd closure on 2.6.9+
3908 Details : We failed to add posix locks to list of inode locks on 2.6.9+
3909 kernels, this caused such locks not to be released on fd close and
3910 then assertions on fs unmount about still used locks.
3913 Frequency : MDS failover only, very rarely
3915 Description: clients may get ASSERTION(granted_lock != NULL)
3916 Details : When request was taking a long time, and a client was resending
3917 a getattr by name lock request. The were multiple lock requests
3918 with the same client lock handle and
3919 mds_getattr_name->fixup_handle_for_resent_request found one of the
3920 lock handles but later failed with ASSERTION(granted_lock != NULL).
3925 Description: handle->h_buffer_credits > 0, assertion failure
3926 Details : h_buffer_credits is zero after truncate, causing assertion
3927 failure. This patch extends the transaction or creates a new
3931 Frequency : NFS re-export or patchless client
3932 Bugzilla : 11179, 10796
3933 Description: Crash on NFS re-export node (__d_move)
3934 Details : We do not want to hash the dentry if we don't have a lock.
3935 But if this dentry is later used in d_move, we'd hit uninitialised
3936 list head d_hash, so we just do this to init d_hash field but
3937 leave dentry unhashed.
3940 Frequency : NFS re-export or patchless client
3942 Description: NFS exports has problem with symbolic link
3943 Details : lustre client didn't properly install dentry when re-exported
3944 to NFS or running patchless client.
3947 Frequency : NFS re-export or patchless client
3949 Description: Various nfs/patchless fixes.
3950 Details : fixes reuse disconected alias for lookup process - this fixes
3951 warning "find_exported_dentry: npd != pd",
3952 fix permission error with open files at nfs.
3953 fix apply umask when do revalidate.
3956 Frequency : occasional
3958 Description: Crash on NFS re-export node
3959 Details : calling clear_page() on the wrong pointer triggered oops in
3960 generic_mapping_read().
3963 Frequency : rarely, using O_DIRECT IO
3965 Description: unaligned directio crashes client with LASSERT
3966 Details : check for unaligned buffers before trying any requests.
3969 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
3971 Description: stale data returned from RAID cache
3972 Details : If only a small amount of IO is done to the RAID device before
3973 reading it again it is possible to get stale data from the RAID
3974 cache instead of reading it from disk.
3977 Frequency : always for sles10 kernel
3979 Description: sles10 support
3980 Details : ll_follow_link: compile fixes and using of nd_set_link
3981 under newer kernels.
3984 Frequency : depends on arch, kernel and compiler version, always on sles10
3987 Description: recursive or deep enough symlinks cause stack overflow
3988 Details : getting rid of large stack-allocated variable in
3992 Frequency : depends on hardware
3994 Description: lustre write performance loss in the SLES10 kernel
3995 Details : the performance loss is caused by using of write barriers in the
3996 ext3 code. The SLES10 kernel turns barrier support on by
3997 default. The fix is to undo that change for ldiskfs.
3999 ------------------------------------------------------------------------------
4001 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
4003 * Support for kernels:
4004 2.6.9-42.0.3EL (RHEL 4)
4005 2.6.5-7.276 (SLES 9)
4006 2.4.21-47.0.1.EL (RHEL 3)
4007 2.6.12.6 vanilla (kernel.org)
4011 Frequency : quota enabled and large files being deleted
4013 Description: releasing more than 4GB of quota at once hangs OST
4014 Details : If a user deletes more than 4GB of files on a single OST it
4015 will cause the OST to spin in an infinite loop. Release
4016 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4021 Description: statfs data retrieved from /proc may be stale or zero
4022 Details : When reading per-device statfs data from /proc, in the
4023 {kbytes,files}_{total,free,avail} files, it may appear
4024 as zero or be out of date.
4027 Frequency : systems with MD RAID1 external journal devices
4029 Description: lconf's call to blkid is confused by RAID1 journal devices
4030 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4031 of returning all block devices that match the journal UUID.
4034 Frequency : always, for aggregate stripe size over 4GB
4036 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
4037 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
4038 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
4041 Frequency : always if "lfs find" used on a local file/directory
4043 Description: "lfs find" segfaults if used on a local file/directory
4044 Details : The case where a directory component was not specified wasn't
4045 handled correctly. Handle this properly.
4048 Frequency : always on ppc64
4050 Description: the write to an ext3 filesystem mounted with mballoc got stuck
4051 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4052 perform endianness conversion.
4055 Frequency : rarely (truncate to non-zero file size after write under load)
4056 Bugzilla : 10730, 10687
4057 Description: Files padded with zeros to next 4K multiple
4058 Details : With filesystems mounted using the "extents" option (2.6 kernels)
4059 it is possible that files that are truncated to a non-zero size
4060 immediately after being written are filled with zero bytes beyond
4061 the truncated size. No file data is lost.
4063 Severity : enhancement
4065 Description: Allow recovery/failover for liblustre clients.
4066 Details : liblustre clients were unaware of failover configurations until
4069 Severity : enhancement
4071 Description: user file locks should fail when not mounting with flock option
4072 Details : Set up an error-returning stub in ll_file_operations.lock field
4073 to prevent incorrect behaviour when client is mounted without
4074 flock option. Also, set up properly f_op->flock field for
4078 Frequency : always on ia64
4080 Description: "lfs df" loops on printing out MDS statfs information
4081 Details : The obd_ioctl_data was not initialized and in some systems
4082 this caused a failure during the ioctl that did not return
4083 an error. Initialize the struct and return an error on failure.
4086 Frequency : SLES 9 only
4088 Description: Error of copying files with lustre special EAs as root
4089 Details : Client side always return success for setxattr call for lustre
4090 special xattr (currently only "trusted.lov").
4093 Frequency : rarely on clusters with both ia64+i386 clients
4095 Description: ia64+i686 clients doing shared IO on the same file may LBUG
4096 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
4097 clients are doing concurrent writes to the same file it is
4098 possible that the ia64 clients may LASSERT because the OST
4099 extent locks are not PAGE_SIZE aligned. Ensure that grown
4100 locks are always aligned on the request boundary.
4103 Frequency : specific use, occasional
4105 Description: Overwriting in use executable truncates on-disk binary image
4106 Details : If one node attempts to overwrite an executable in use by
4107 another node, we now correctly return ETXTBSY instead of
4108 truncating the file.
4110 Severity : enhancement
4112 Description: Async OSC create to avoid the blocking unnecessarily.
4113 Details : If an OST has no remaining object, system will block on the creating
4114 when need to create a new object on this OST. Now, ways use
4115 pre-created objects when available, instead of blocking on an
4116 empty osc while others are not empty. If we must block, we block
4117 for the shortest possible period of time.
4122 Description: chmod on Lustre root is propagated to other clients
4123 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
4124 invalid by the follow_mount time.
4127 Frequency : liblustre clients only
4129 Description: Race in 'instant cancel' lock handling could lead to such locks
4130 never to be granted in case of SMP MDS
4131 Details : Do not destroy not yet granted but cbpending locks in
4135 Frequency : replay/resend of open
4137 Description: non null lock assetion failure in mds_intent_policy
4138 Details : Trying to replay/resend lockless open requests resulted in
4139 mds_open() returning 0 with no lock. Now it sets a flag if
4140 a lock is going to be returned.
4142 Severity : enhancement
4144 Description: Checksum enhancements
4145 Details : New checksum enhancements allow for resending RPCs that failed
4148 Severity : enhancement
4150 Description: Tunables on number of dirty pages in cacche
4151 Details : Allow to set limit on number of dirty pages cached.
4156 Description: client crash on unmount - lock still has references
4157 Details : In some error handling cases it was possible to leak a lock
4158 reference on a client while accessing a file. This was not
4159 harmful to the client during operation, but would cause the
4160 client to crash when the filesystem is unmounted.
4163 Frequency : specific case, rare
4165 Description: ETXTBSY on mds though file not in use by client
4166 Details : ETXTBSY is no longer incorrectly returned when attempting to
4167 chmod or chown a directory that the user previously tried to
4168 execute or a currently-executing binary.
4171 Frequency : extremely rare except on liblustre-based clients
4173 Description: Lustre space not freed when files are deleted
4174 Details : Clean up open-unlinked files after client eviction. Previously
4175 the unlink was skipped and the files remained as orphans.
4180 Description: OST failure "would be an LBUG" in waiting_locks_callback()
4181 Details : In some cases it was possible to send a blocking callback to
4182 a client doing a glimpse, even though that client didn't get
4183 a lock granted. When the glimpse lock is cancelled on the OST
4184 the freed lock is left on the waiting list and corrupted the list.
4187 Frequency : all core dumps
4189 Description: Broke core dumps to lustre
4190 Details : Negative dentry may be unhashed if parent does not have UPDATE
4191 lock, but some callers, e.g. do_coredump, expect dentry to be
4192 hashed after successful create, hash it in ll_create_it.
4194 ------------------------------------------------------------------------------
4196 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
4198 * Support for kernels:
4199 2.6.9-42.0.2.EL (RHEL 4)
4200 2.6.5-7.276 (SLES 9)
4201 2.4.21-40.EL (RHEL 3)
4202 2.6.12.6 vanilla (kernel.org)
4206 Frequency : always on RHEL 3
4208 Description: Number of open files grows over time
4209 Details : The number of open files grows over time, whether or not
4210 Lustre is started. This was due to a filp leak introduced
4211 by one of our kernel patches.
4213 ------------------------------------------------------------------------------
4215 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4217 * Support for kernels:
4218 2.6.9-42.EL (RHEL 4)
4219 2.6.5-7.276 (SLES 9)
4220 2.4.21-40.EL (RHEL 3)
4221 2.6.12.6 vanilla (kernel.org)
4226 Bugzilla : 5719, 9635, 9792, 9684
4227 Description: OST (or MDS) trips assertions in (re)connection under heavy load
4228 Details : If a server is under heavy load and cannot reply to new
4229 connection requests before the client resends the (re)connect,
4230 the connection handling code can behave badly if two service
4231 threads are concurrently handing separate (re)connections from
4232 the same client. Add better locking to the connection handling
4233 code, and ensure that only a single connection will be processed
4234 for a given client UUID, even if the lock is dropped.
4236 Severity : enhancement
4238 Description: add TCP zero-copy support to kernel
4239 Details : Add support to the kernel TCP stack to allow zero-copy bulk
4240 sends if the hardware supports scatter-gather and checksumming.
4241 This allows socklnd to do client-write and server-read more
4242 efficiently and reduce CPU utilization from skbuf copying.
4245 Frequency : only if NFS exporting from client
4247 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
4248 Details : If setting attributes on a file created under NFS that had
4249 never been opened it would be possible to oops the client
4250 if the file had no objects.
4253 Frequency : always for liblustre
4255 Description: liblustre client does MDS+OSTs setattr RPC for each write
4256 Details : When doing a write from a liblustre client, the client
4257 incorrectly issued an RPC to the MDS and each OST the file was
4258 striped over in order to update the timestamps. When writing
4259 with small chunks and many clients this could overwhelm the MDS
4260 with RPCs. In all cases it would slow down the write because
4261 these RPCs are unnecessary.
4263 Severity : enhancement
4265 Description: allow number of MDS service threads to be changed at module load
4266 Details : It is now possible to change the number of MDS service threads
4267 running. Adding "options mds mds_num_threads={N}" to the MDS's
4268 /etc/modprobe.conf will set the number of threads for the next
4269 time Lustre is restarted (assuming the "mds" module is also
4270 reloaded at that time). The default number of threads will
4271 stay the same, 32 for most systems.
4276 Description: OST crash if filesystem is unformatted or corrupt
4277 Details : If an OST is started on a device that has never been formatted
4278 or if the filesystem is corrupt and cannot even mount then the
4279 error handling cleanup routines would dereference a NULL pointer.
4284 Description: NULL pointer deref in llap_from_page.
4285 Details : get_cache_page_nowait can return a page with NULL (or otherwise
4286 incorrect) mapping if the page was truncated/reclaimed while it was
4287 searched for. Check for this condition and skip such pages when
4288 doing readahead. Introduce extra check to llap_from_page() to
4289 verify page->mapping->host is non-NULL (so page is not anonymous).
4292 Frequency : Sometimes when using sys_sendfile
4294 Description: "page not covered by a lock" warnings from ll_readpage
4295 Details : sendfile called ll_readpage without right page locks present.
4296 Now we introduced ll_file_sendfile that does necessary locking
4297 around call to generic_file_sendfile() much like we do in
4301 Frequency : with certain MDS communication failures at client mount time
4303 Description: NULL pointer deref after failed client mount
4304 Details : a client connection request may delayed by the network layer
4305 and not be sent until after the PTLRPC layer has timed out the
4306 request. If the client fails the mount immediately it will try
4307 to clean up before the network times out the request. Add a
4308 reference from the request import to the obd device and delay
4309 the cleanup until the network drops the request.
4312 Frequency : occasionally during client (re)connect
4314 Description: assertion failure during client (re)connect
4315 Details : processing a client connection request may be delayed by the
4316 client or server longer than the client connect timeout. This
4317 causes the client to resend the connection request. If the
4318 original connection request is replied in this interval, the
4319 client may trip an assertion failure in ptlrpc_connect_interpret()
4320 which thought it would be the only running connect process.
4323 Frequency : only with obd_echo servers and clients that are rebooted
4325 Description: kernel BUG accessing uninitialized data structure
4326 Details : When running an obd_echo server it did not start the ping_evictor
4327 thread, and when a client was evicted an uninitialized data
4328 structure was accessed. Start the ping_evictor in the RPC
4329 service startup instead of the OBD startup.
4331 Severity : enhancement
4332 Bugzilla : 10193 (patchless)
4333 Description: Remove dependency on various unexported kernel interfaces.
4334 Details : No longer need reparent_to_init, exit_mm, exit_files,
4335 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
4338 Frequency : rare (only users of deprecated and unsupported LDAP config)
4340 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
4341 Details : LDAP apparently contains 'lustreName' attributes instead of
4342 'name'. A simple remapping of the name is sufficient.
4345 Frequency : rare (only with non-default dump_on_timeout debug enabled)
4347 Description: waiting_locks_callback trips kernel BUG if client is evicted
4348 Details : Running with the dump_on_timeout debug flag turned on makes
4349 it possible that the waiting_locks_callback() can try to dump
4350 the Lustre kernel debug logs from an interrupt handler. Defer
4351 this log dumping to the expired_lock_main() thread.
4353 Severity : enhancement
4355 Description: Support NFS exporting on 2.6 kernels.
4356 Details : Implement non-rawops metadata methods for NFS server to use without
4357 changing NFS server code.
4360 Frequency : very rare (synthetic metadata workload only)
4362 Description: two racing renames might cause an MDS thread to deadlock
4363 Details : Running the "racer" program may cause one MDS thread to rename
4364 a file from being the source of a rename to being the target of
4365 a rename at exactly the same time that another thread is doing
4366 so, and the second thread has already enqueued these locks after
4367 doing a lookup of the target and is trying to relock them in
4368 order. Ensure that we don't try to re-lock the same resource.
4371 Frequency : only very large systems with liblustre clients
4373 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
4374 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
4375 all OSTs in parallel. This allows the eviction of stale liblustre
4376 clients to proceed much faster than if they were done in series,
4377 and also offers similar improvements for other set_info RPCs.
4382 Description: excessive CPU usage during initial read phase on client
4383 Details : During the initial read phase on a client, it would agressively
4384 retry readahead on the file, consuming too much CPU and impacting
4385 performance (since 1.4.5.8). Improve the readahead algorithm
4386 to avoid this, and also improve some other common cases (read
4387 of small files in particular, where "small" is files smaller than
4388 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
4393 Description: MDS crash when receiving packet with unknown intent.
4394 Details : Do not LBUG in unknown intent case, just return -EFAULT
4396 Severity : enhancement
4397 Bugzilla : 9293, 9385
4398 Description: MDS RPCs are serialised on client. This is unnecessary for some.
4399 Details : Do not serialize getattr (non-intent version) and statfs.
4402 Frequency : occasional, when OST network is overloaded/intermittent
4404 Description: client evicted by OST after bulk IO timeout
4405 Details : If a client sends a bulk IO request (read or write) the OST
4406 may evict the client if it is unresposive to its data GET/PUT
4407 request. This is incorrect if the network is overloaded (takes
4408 too long to transfer the RPC data) or dropped the OST GET/PUT
4409 request. There is no need to evict the client at all, since
4410 the pinger and/or lock callbacks will handle this, and the
4411 client can restart the bulk request.
4414 Frequency : Always when mmapping file with no objects
4416 Description: client crashes when mmapping file with no objects
4417 Details : Check that we actually have objects in a file before doing any
4418 operations on objects in ll_vm_open, ll_vm_close and
4424 Description: Request leak when working with deleted CWD
4425 Details : Introduce advanced request refcount tracking for requests
4426 referenced from lustre intent.
4428 Severity : Enhancement
4430 Description: Cache open file handles on client.
4431 Details : MDS now will return special lock along with openhandle, if
4432 requested and client is allowed to hold openhandle, even if unused,
4433 until such a lock is revoked. Helps NFS a lot, since NFS is opening
4434 closing files for every read/write openration.
4436 Severity : Enhancement
4438 Description: Cache open negative dentries on client when possible.
4439 Details : Guard negative dentries with UPDATE lock on parent dir, drop
4440 negative dentries on lock revocation.
4445 Description: Remounting a client read-only wasn't possible with a zconf mount
4446 Details : It wasn't possible to remount a client read-only with llmount.
4448 Severity : enhancement
4449 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
4450 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
4451 MPICH is included to add Lustre-specific ADIO interfaces.
4452 This is based closely on the UFS ADIO layer and only differs
4453 in file creation, in order to allow the OST striping to be set.
4454 This is user-contributed code and not supported by CFS.
4459 Description: extended inode attributes (immutable, append-only) work improperly
4460 when 2.4 and 2.6 kernels are used on client/server or vice versa
4461 Details : Introduce kernel-independent values for these flags.
4463 Severity : enhancement
4466 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
4467 Details : Many of the /proc/ tunables can only be tuned at a megabyte
4468 granularity. Now, Fractional MB granularity is be supported,
4469 this is very useful for low memory system.
4471 Severity : enhancement
4473 Description: Getattr by fid
4474 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
4475 locks, avoids extra getattr rpc requests to MDS, allows '/' to
4476 have locks and avoids getattr rpc requests for it on every stat.
4479 Frequency : Always, for filesystems larger than 2TB
4481 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4482 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
4483 filesystems larger than 2TB to be mounted with mballoc enabled.
4486 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
4488 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4489 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
4490 filesystems larger than 2TB is mounted then the kernel will
4491 silently corrupt the start of the filesystem. CONFIG_LBD is
4492 enabled for all CFS-supported kernels, but the possibility of
4493 this happening with a modified kernel config exists.
4495 Severity : enhancement
4497 Description: add client O_DIRECT support for 2.6 kernels
4498 Details : It is now possible to do O_DIRECT reads and writes to files
4499 in the Lustre client mountpoint on 2.6 kernel clients.
4501 Severity : enhancement
4503 Description: parallel glimpse, setattr, statfs, punch, destroy requests
4504 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
4505 parallel, not waiting for response from every OST before sending
4506 a rpc to the next OST.
4511 Description: setattr vs write race when updating file timestamps
4512 Details : Client processes that update a file timestamp into the past
4513 right after writing to the file (e.g. tar) it is possible that
4514 the updated file modification time can be reset to the current
4515 time due to a race between processing the setattr and write RPC.
4517 Severity : enhancement
4519 Description: Bring 'lfs find' closer in line with regular Linux find.
4520 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
4521 -print0 options and obtains all the needed info through the lustre
4524 Severity : enhancement
4526 Description: support up to 1024 configured devices on one node
4527 Details : change obd_dev array from statically allocated to dynamically
4528 allocated structs as they are first used to reduce memory usage
4533 Description: Flush dirty partially truncated pages during truncate
4534 Details : Immediatelly flush partially truncated pages in filter_setattr,
4535 this way we completely avoid having any pages in page cache on OST
4536 and can retire ugly workarounds during writes to flush such pages.
4541 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
4542 Details : protect inode from truncation within vfs_unlink() context
4543 just take a reference before calling vfs_unlink() and release it
4544 when parent's i_sem is free.
4549 Description: last_id value checked outside lock on OST caused LASSERT failure
4550 Details : If there were multiple MDS->OST object precreate requests in
4551 flight, it was possible that the OST's last object id was checked
4552 outside a lock and incorrectly tripped an assertion. Move checks
4553 inside locks, and discard old precreate requests.
4556 Frequency : always, if extents are used on OSTs
4558 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
4559 Details : OSTs using the extents format would not zero the high 16 bits of
4560 the index physical block number. This is not a problem for any
4561 OST filesystems smaller than 16TB, and no kernels support ext3
4562 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
4563 new/modified files) and can be fixed for existing filesystems
4564 with e2fsprogs-1.39-cfs1.
4569 Description: import connection selection may be incorrect if timer wraps
4570 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
4571 import connections to be ignored if the 32-bit jiffies counter
4572 wraps. Use a 64-bit jiffies counter.
4575 Frequency : very large clusters immediately after boot
4577 Description: LNET request buffers exhausted under heavy short-term load
4578 Details : If a large number of client requests are generated on a service
4579 that has previously never seen so many requests it is possible
4580 that the request buffer growth cannot keep up with the spike in
4581 demand. Instead of dropping incoming requests, they are held in
4582 the LND until the RPC service can accept more requests.
4585 Frequency : Sometimes during replay
4587 Description: Assertion failure in ll_local_open after replay.
4588 Details : If replay happened on an open request reply before we were able
4589 to set replay handler, reply will become not swabbed tripping the
4590 assertion in ll_local_open. Now we set the handler right after
4591 recognising of open request
4594 Frequency : very rare
4596 Description: kernel reports "badness in vsnprintf"
4597 Details : Reading from the "recovery_status" /proc file in small chunks
4598 may cause a negative length in lprocfs_obd_rd_recovery_status()
4599 call to vsnprintf() (which is otherwise harmless). Exit early
4600 if there is no more space in the output buffer.
4602 Severity : enhancement
4604 Description: clear OBD RPC statistics by writing to them
4605 Details : It is now possible to clear the OBD RPC statistics by writing
4606 to the "stats" file.
4611 Description: Client mtime is not the same on different clients after utimes
4612 Details : In some cases, the client was using the utimes() syscall on
4613 a file cached on another node. The clients now validate the
4614 ctime from the MDS + OSTs to determine which one is right.
4619 Description: Inability to activate failout mode
4620 Details : lconf script incorrectly assumed that in pythong string's numeric
4621 value is used in comparisons.
4624 Frequency : always with multiple stripes per file
4626 Description: Inefficient object allocation for mutli-stripe files
4627 Details : When selecting which OSTs to stripe files over, for files with
4628 a stripe count that divides evenly into the number of OSTs,
4629 the MDS is always picking the same starting OST for each file.
4630 Return the OST selection heuristic to the original design.
4635 Description: mount failures may take full timeout to return an error
4636 Details : Under some heavy load conditions it is possible that a
4637 failed mount can wait for the full obd_timeout interval,
4638 possibly several minutes, before reporting an error.
4639 Instead return an error as soon as the status is known.
4641 Frequency : quota enabled and large files being deleted
4643 Description: releasing more than 4GB of quota at once hangs OST
4644 Details : If a user deletes more than 4GB of files on a single OST it
4645 will cause the OST to spin in an infinite loop. Release
4646 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4651 Description: statfs data retrieved from /proc may be stale or zero
4652 Details : When reading per-device statfs data from /proc, in the
4653 {kbytes,files}_{total,free,avail} files, it may appear
4654 as zero or be out of date.
4657 Frequency : systems with MD RAID1 external journal devices
4659 Description: lconf's call to blkid is confused by RAID1 journal devices
4660 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4661 of returning all block devices that match the journal UUID.
4664 Frequency : always, for aggregate stripe size over 4GB
4666 Description: assertion fails when trying to use 4GB stripe size
4667 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
4668 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
4671 Frequency : always on ppc64
4673 Description: the first write on an ext3 filesystem with mballoc got stuck
4674 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4675 perform endianness conversion.
4677 ------------------------------------------------------------------------------
4679 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4681 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
4682 INTEROPERATE with older versions automatically. Please read the
4683 user documentation before upgrading any part of a live system.
4684 * WARNING: Lustre networking configuration changes are required with
4685 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
4688 * Support for kernels:
4689 2.6.9-22.0.2.EL (RHEL 4)
4690 2.6.5-7.244 (SLES 9)
4691 2.6.12.6 vanilla (kernel.org)
4694 Severity : enhancement
4695 Bugzilla : 7981/8208
4696 Description: Introduced Lustre Networking (LNET)
4697 Details : LNET is new networking infrastructure for Lustre, it includes
4698 a reorganized network configuration mode (see the user
4699 documentation for full details) as well as support for routing
4700 between different network fabrics. Lustre Networking Devices
4701 (LNDs) for the supported network fabrics have also been
4702 created for this new infrastructure.
4704 Severity : enhancement
4705 Description: Introduced Access control lists
4706 Details : clients can set ACLs on files and directories in order to have
4707 more fine-grained permissions than the standard Unix UGO+RWX.
4708 The MDS must be started with the "-o acl" mount option.
4710 Severity : enhancement
4711 Description: Introduced filesystem quotas
4712 Details : Administrators may now establish per-user quotas on the
4715 Severity : enhancement
4717 Description: Configuration change for the XT3
4718 The PTLLND is now used to run Lustre over Portals on the XT3
4719 The configure option(s) --with-cray-portals are no longer used.
4720 Rather --with-portals=<path-to-portals-includes> is used to
4721 enable building on the XT3. In addition to enable XT3 specific
4722 features the option --enable-cray-xt3 must be used.
4727 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
4728 Details : A race between allocating a new llcd and re-getting the llcd_lock
4729 allowed another thread to grab newly-allocated llcd.
4731 Severity : enhancement
4733 Description: 2.6 OST async journal commit and locking fix to improve performance
4734 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
4735 2.6 kernels are now async as they already were in 2.4 kernels so
4736 that they can commit concurrently with the network bulk transfer.
4737 For block-allocated files the filter allocation semaphore is held
4738 to avoid filesystem fragmentation during allocation. BKL lock
4739 removed for 2.6 xattr operations where it is no longer needed.
4744 Description: lconf incorrectly determined whether two IP networks could talk
4745 Details : In some more complicated routing and multiple-network
4746 configurations, lconf will avoid trying to make a network
4747 connection to a disjoint part of the IP space. It was doing the
4748 math incorrectly for one set of cases.
4753 Description: Fix for potential infinite loop processing records in an llog.
4754 Details : If an llog record is corrupted/zeroed, it is possible to loop
4755 forever in llog_process(). Validate the llog record length
4756 and skip the remainder of the block on error.
4759 Frequency : occasional (liblustre only)
4761 Description: liblustre could not open files whose last component is a symlink
4762 Details : sysio_path_walk() would incorrectly pass the open intent to
4763 intermediate path components.
4766 Frequency : rare (liblustre only with non-standard tuning)
4767 Bugzilla : 7201 (7350)
4768 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4769 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4770 after it has been released, only do this for referenced locks
4772 Severity : enhancement
4774 Description: specifying an (invalid) directory default stripe_size of -1
4775 would reset the directory default striping
4776 Details : stripe_size -1 was used internally to signal directory stripe
4777 removal, now use "all default" to signal dir stripe removal
4778 as a directory striping of "all default" is not useful
4781 Frequency : common for large clusters running liblustre clients
4783 Description: doing an ls when liblustre clients are running is slow
4784 Details : sending a glimpse AST to a liblustre client waits for every AST
4785 to time out, as liblustre clients will not respond. Since they
4786 cannot cache data we refresh the OST lock LVB from disk instead.
4788 Severity : enhancement
4790 Description: doing an ls at the same time as file IO can be slow
4791 Details : enqueue and other "small" requests can be blocked behind many
4792 large IO requests. Create a new OST IO portal for non-IO
4793 requests so they can be processed faster.
4796 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4798 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4799 created file with mode 000
4800 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4801 allow an ftruncate() as a non-root user to a file with mode 000.
4802 Lustre now respects this flag to disable mode checks when
4803 truncating a file owned by the user
4806 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4809 Description: Revoking locks from clients that went dead or catatonic might take
4811 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4812 cancellation of such locks instant on servers without waiting for
4813 any reply from clients. Clients drops these locks when cancel
4814 notification from server is received without replying.
4817 Frequency : liblustre-only, when liblustre client dies or becomes busy
4819 Description: Doing ls on Linux clients can take a long time with active
4821 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4822 granting such locks to it in the first place if possible. Locks
4823 are taken by proxy on the OST during the read or write and
4824 dropped immediately afterward. Add connect flags handling, do
4825 not grant locks to liblustre clients for glimpse ASTs.
4827 Severity : enhancement
4829 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4830 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4831 sequential reads and behaves badly with applications doing random
4832 reads. Improve it by reading ahead at least read region, and
4833 avoiding excessive large RPC for small reads.
4835 Severity : enhancement
4837 Description: Creating more than 1000 files for a single job may cause a load
4838 imbalance on the OSTs if there are also a large number of OSTs.
4839 Details : qos_prep_create() uses an OST index reseed value that is an
4840 even multiple of the number of available OSTs so that if the
4841 reseed happens in the middle of the object allocation it will
4842 still utilize the OSTs as uniformly as possible.
4847 Description: OST or MDS may oops in ping_evictor_main()
4848 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
4849 but doesn't restart at beginning of obd_exports_timed list
4852 Severity : enhancement
4854 Description: improve by-nid export eviction on the MDS and OST
4855 Details : allow multiple exports with the same NID to be evicted at one
4856 time without re-searching the exports list.
4859 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
4861 Description: MDS may oops in groups_free()
4862 Details : in rare race conditions a newly allocated group_info struct is
4863 freed again, and this can be NULL. The 2.4 compatibility code
4864 for groups_free() checked for a NULL pointer, but 2.6 did not.
4867 Frequency : common for liblustre clients doing little filesystem IO
4868 Bugzilla : 9352, 7313
4869 Description: server may evict liblustre clients accessing contended locks
4870 Details : if a client is granted a lock or receives a completion AST
4871 with a blocking AST already set it would not reply to the AST
4872 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
4875 Frequency : lfs setstripe, only systems with more than 160 OSTs
4877 Description: unable to set striping with a starting offset beyond OST 160
4878 Details : llapi_create_file() incorrectly limited the starting stripe
4879 index to the maximum single-file stripe count.
4882 Frequency : LDAP users only
4884 Description: lconf did not handle in-kernel recovery with LDAP properly
4885 Details : lconf/LustreDB get_refs() is searching the wrong namespace
4887 Severity : enhancement
4889 Description: bind OST threads to NUMA nodes to improve performance
4890 Details : all OST threads are uniformly bound to CPUs on a single NUMA
4891 node and do their allocations there to localize memory access
4893 Severity : enhancement
4895 Description: llmount can determine client NID directly from Myrinet (GM)
4896 Details : the client NID code from gmnalnid was moved directly into
4897 llmount, removing the need to use this or specifying the
4898 client NID explicitly when mounting GM clients with zeroconf
4901 Frequency : if client is started with down MDS
4903 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
4904 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
4905 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
4906 ptlrpc_import_delayed_req() because we didn't check if the
4907 request was interrupted, and we also didn't break out of the
4908 event loop if there was no timeout
4913 Description: data loss during non-page-aligned writes to a single file from
4914 both multiple nodes and multiple threads on one node at same time
4915 Details : updates to KMS and lsm weren't protected by common lock. Resulting
4916 inconsistency led to false short-reads, that were cached and later
4917 used by ->prepare_write() to fill in partially written page,
4918 leading to data loss.
4921 Frequency : always, if lconf --abort_recovery used
4923 Description: lconf --abort_recovery fails with 'Operation not supported'
4924 Details : lconf was attempting to abort recovery on the MDT device and not
4927 Severity : enhancement
4929 Description: remove cleanup logs
4930 Details : replace lconf-generated cleanup logs with lustre internal
4931 cleanup routines. Eliminates the need for client-cleanup and
4934 Severity : enhancement
4936 Description: add support for EAs (user and system) on lustre filesystems
4937 Details : it is now possible to store extended attributes in the Lustre
4938 client filesystem, and with the user_xattr mount option it
4939 is possible to allow users to store EAs on their files also
4941 Severity : enhancement
4943 Description: Add possibility (config option) to show minimal available OST free
4945 Details : When compiled with --enable-mindf configure option, statfs(2)
4946 (and so, df) will return least minimal free space available from
4947 all OSTs as amount of free space on FS, instead of summary of
4948 free spaces of all OSTs.
4950 Severity : enhancement
4952 Description: do not expand extent locks acquired on OST-side
4953 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
4954 by server: they are not cached anyway.
4957 Frequency : when mmap is used/binaries executed from Lustre
4959 Description: Unmmap pages before throwing them away from read cache.
4960 Details : llap_shrink cache now attempts to unmap pages before discarding
4961 them (if unmapping failed - do not discard). SLES9 kernel has
4962 extra checks that trigger if this unmapping is not done first.
4967 Description: lconf didn't resolve symlinks before checking to see whether a
4968 given mountpoint was already in use
4971 Frequency : when migrating failover services
4972 Bugzilla : 6395, 9514
4973 Description: When migrating a subset of services from a node (e.g. failback
4974 from a failover service node) the remaining services would
4975 time out and evict clients.
4976 Details : lconf --force (implied by --failover) sets the global obd_timeout
4977 to 5 seconds in order to quickly disconnect, but this caused
4978 other RPCs to time out too quickly. Do not change the global
4979 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
4981 Severity : enhancement
4982 Frequency : if MDS is started with down OST
4983 Bugzilla : 9439,5706
4984 Description: Allow startup/shutdown of an MDS without depending on the
4985 availability of the OSTs.
4986 Details : Asynchronously call mds_lov_synchronize during MDS startup.
4987 Add appropriate locking and lov-osc refcounts for safe
4988 cleaning. Add osc abort_inflight calls in case the
4989 synchronize never started.
4992 Frequency : occasional (Cray XT3 only)
4994 Description: root not authorized to access files in CRAY_PORTALS environment
4995 Details : The client process capabilities were not honoured on the MDS in
4996 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
4997 been accessed by an authorized user then root was able to access
4998 the file on the local client also. The root user capabilities
4999 are now allowed on the MDS, as this environment has secure UID.
5002 Frequency : occasional
5004 Description: ldiskfs "too long searching" message happens too often
5005 Details : A debugging message (otherwise harmless) prints too often on
5006 the OST console. This has been reduced to only happen when
5007 there are fragmentation problems on the filesystem.
5012 Description: Division by zero in statfs when all OSCs are inactive
5013 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
5014 lov_statfs divides by value returned by lov_get_stripecnt().
5018 Bugzilla : 9489, 3273
5019 Description: First write from each client to each OST was only 4kB in size,
5020 to initialize client writeback cache, which caused sub-optimal
5021 RPCs and poor layout on disk for the first written file.
5022 Details : Clients now request an initial cache grant at (re)connect time
5023 and so that they can start streaming writes to the cache right
5024 away and always do full-sized RPCs if there is enough data.
5025 If the OST is rebooted the client also re-establishes its grant
5026 so that client cached writes will be honoured under the grant.
5031 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
5032 Details : Now I/O RPCs go to different portal number and (presumably) fast
5033 lock requests (and glimses) and other RPCs get their own service
5034 threads pool that should be able to service those RPCs
5037 Severity : enhancement
5039 Description: Ability to exchange lustre version between client and servers and
5040 issue warnings at client side if client is too old. Also for
5041 liblustre clients there is ability to refuse connection of too old
5043 Details : New 'version' field is added to connect data structure that is
5044 filled with version info. That info is later checked by server and
5048 Frequency : rare, liblustre only.
5049 Bugzilla : 9296, 9581
5050 Description: Two simultaneous writes from liblustre at offset within same page
5051 might proceed at the same time overwriting eachother with stale
5053 Details : I/O lock withing llu_file_prwv was released too early, before data
5054 actually was hitting the wire. Extended lock-holding time until
5055 server acknowledges receiving data.
5058 Frequency : extremely rare. Never observed in practice.
5060 Description: avoid generating lustre_handle cookie of 0.
5061 Details : class_handle_hash() generates handle cookies by incrementing
5062 global counter, and can hit 0 occasionaly (this is unlikely, but
5063 not impossible, because initial value of cookie counter is
5064 selected randonly). Value of 0 is used as a sentinel meaning
5065 "unassigned handle" --- avoid it. Also coalesce two critical
5066 sections in this function into one.
5068 Severity : enhancement
5070 Description: allow liblustre clients to delegate truncate locking to OST
5071 Details : To avoid overhead of locking, liblustre client instructs OST to
5072 take extent lock in ost_punch() on client's behalf. New connection
5073 flag is added to handle backward compatibility.
5075 Severity : enhancement
5076 Bugzilla : 4928, 7341, 9758
5077 Description: allow number of OST service threads to be specified
5078 Details : a module parameter allows the number of OST service threads
5079 to be specified via "options ost ost_num_threads={N}" in the
5080 OSS's /etc/modules.conf or /etc/modprobe.conf.
5084 Bugzilla : 6146, 9635, 9895
5085 Description: servers crash with bad pointer in target_handle_connect()
5086 Details : In rare cases when a client is reconnecting it was possible that
5087 the connection request was the last reference for that export.
5088 We would temporarily drop the export reference and get a new
5089 one, but this may have been the last reference and the export
5090 was just destroyed. Get new reference before dropping old one.
5092 Severity : enhancement
5093 Frequency : if client is started with failover MDS
5095 Description: Allow multiple MDS hostnames in the mount command
5096 Details : Try to read the configuration from all specified MDS
5097 hostnames during a client mount in case the "primary"
5100 Severity : enhancement
5102 Description: Stop sending data to evicted clients as soon as possible.
5103 Details : Check if the client we are about to send or are sending data to
5104 was evicted already. (Check is done every second of waiting,
5105 for which l_wait_event interface was extended to allow checking
5106 of exit condition at specified intervals).
5109 Frequency : rare, normally only when NFS exporting is done from client
5111 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
5113 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
5116 Frequency : always, if srand() is called before liblustre initialization
5118 Description: Liblustre uses system PRNG disturbing its usage by user application
5119 Details : Introduce internal to lustre fast and high-quality PRNG for
5120 lustre usage and make liblustre and some other places in generic
5121 lustre code to use it.
5123 Severity : enhancement
5124 Bugzilla : 9477, 9557, 9870
5125 Description: Verify that the MDS configuration logs are updated when xml is
5126 Details : Check if the .xml configuration logs are newer than the config
5127 logs stored on the MDS and report an error if this is the case.
5128 Request --write-conf, or allow starting with --old_conf.
5130 Severity : enhancement
5132 Description: Handle symlinks in the path when checking if Lustre is mounted.
5133 Details : Resolve intermediate symlinks when checking if a client has
5134 mounted a filesystem to avoid duplicate client mounts.
5139 Description: lconf can hit an error exception but still return success.
5140 Details : The lconf command catches the Command error exception at the top
5141 level script context and will exit with the associated exit
5142 status, but doesn't ensure that this exit status is non-zero.
5147 Description: failure of ptlrpc thread startup can cause oops
5148 Details : Starting a ptlrpc service thread can fail if there are a large
5149 number of threads or the server memory is very fragmented.
5150 Handle this without oopsing.
5153 Frequency : always, only if liblustre and non-default acceptor port was used
5155 Description: liblustre cannot connect to servers with non-default acceptor port
5156 Details : tcpnal_set_default_params() was not called and was therefore
5157 ignoring the environment varaible TCPNAL_PORT, as well as other
5158 TCPNAL_ environment variables
5163 Description: two objects could be created on the same OST for a single file
5164 Details : If an OST is down, in some cases it was possible to create two
5165 objects on a single OST for a single file. No problems other
5166 than potential performance impact and spurious error messages.
5170 Bugzilla : 5681, 9562
5171 Description: Client may oops in ll_unhash_aliases
5172 Details : Client dcache may become inconsistent in race condition.
5173 In some cases "getcwd" can fail if the current directory is
5179 Description: Inode refcounting problems in NFS export code
5180 Details : link_raw functions used to call d_instantiate without obtaining
5181 extra inode reference first.
5185 Bugzilla : 9942, 9903
5186 Description: Referencing freed requests leading to crash, memleaks with NFS.
5187 Details : We used to require that call to ll_revalidate_it was always
5188 followed by ll_lookup_it. Also with revalidate_special() it is
5189 possible to call ll_revalidate_it() twice for the same dentry
5190 even if first occurence returned success. This fix changes semantic
5191 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
5192 reference on a request referred from the intent.
5193 ll_intent_release() then releases such a request.
5196 Frequency : rare, normally benchmark loads only
5198 Description: unlinked inodes were kept in memory on the client
5199 Details : If a client is repeatedly creating and unlinking files it
5200 can accumulate a lot of stale inodes in the inode slab cache.
5201 If there is no other client load running this can cause the
5202 client node to run out of memory. Instead flush old inodes
5203 from client cache that have the same inode number as a new inode.
5206 Frequency : SLES9 2.6.5 kernel and long filenames only
5207 Bugzilla : 9969, 10379
5208 Description: utime reports stale NFS file handle
5209 Details : SLES9 uses out-of-dentry names in some cases, which confused
5210 the lustre dentry revalidation. Change it to always use the
5214 Frequency : rare, unless heavy write-truncate concurrency is continuous
5215 Bugzilla : 4180, 6984, 7171, 9963, 9331
5216 Description: OST becomes very slow and/or deadlocked during object unlink
5217 Details : filter_destroy() was holding onto the parent directory lock
5218 while truncating+unlinking objects. For very large objects this
5219 may block other threads for a long time and slow overall OST
5220 responsiveness. It may also be possible to get a lock ordering
5221 deadlock in this case, or run out of journal credits because of
5222 the combined truncate+unlink. Solution is to do object truncate
5223 first in one transaction without parent lock, and then do the
5224 final unlink in a new transaction with the parent lock. This
5225 reduces the lock hold time dramatically.
5228 Frequency : rare, 2.4 kernels only
5230 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
5231 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
5232 before unmounting their filesystem in order to avoid deadlock.
5233 The kernel_locked() function in 2.4 kernels only checks whether
5234 the kernel lock is held, not whether it is this process that is
5235 holding it as 2.6 kernels do.
5240 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
5241 Details : The client ptlrpc code may be trying to reconnect to a down
5242 server before a previous connection attempt has timed out.
5243 Increase the reconnect interval to be longer than the connection
5244 timeout interval to avoid sending duplicate connections to
5248 Frequency : echo_client brw_test command
5250 Description: fix echo_client to work with OST preallocated code
5251 Details : OST preallocation code (5137) didn't take echo_client IO path
5252 into account: echo_client calls filter methods outside of any
5253 OST thread and, hence, there is no per-thread preallocated
5254 pages and buffers to use. Solution: hijack pga pages for IO. As
5255 a byproduct, this avoids unnecessary data copying.
5259 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
5260 Description: Client can oops in mdc_commit_close() after open replay
5261 Details : It was possible for the MDS to return an open request with no
5262 transaction number in mds_finish_transno() if the client was
5263 evicted, but without actually returning an error. Clients
5264 would later try to replay that open and may trip an assertion
5265 Simplify the client close codepath, and always return an error
5266 from the MDS in case the open is not successful.
5269 Frequency : rare, 2.6 OSTs only
5271 Description: OST may deadlock under high load on fragmented files
5272 Details : If there was a heavy load and highly-fragmented OST filesystems
5273 it was possible to have all the OST threads deadlock waiting on
5274 allocation of biovecs, because the biovecs were not released
5275 until the entire RPC IO was completed. Instead, release biovecs
5276 as soon as they are complete to ensure forward IO progress.
5278 Severity : enhancement
5280 Description: Support for specifying external journal device at mount
5281 Details : If an OST or MDS device is formatted with an external journal
5282 device, this device major/minor is stored in the ext3 superblock
5283 and may not be valid for failover. Allow detecting and
5284 specifying the external journal at mount time.
5289 Description: Mounting an MDS with pending unlinked files may cause oops
5290 Details : target_finish_recovery() calls mds_postrecov() which returned
5291 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
5292 considers this an error and immediately begins cleaning up the
5293 lov, just after starting the mds_lov process
5295 Severity : enhancement
5297 Description: Implement 'lfs df' to report actual free space on per-OST basis
5298 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
5299 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
5300 usage of objects; '-h' to report in human readable format.
5302 ------------------------------------------------------------------------------
5304 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
5311 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
5312 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
5313 at startup during orphan unlinking. Instead, skip buddy update
5314 before setup, as it will be regenerated after recovery is complete.
5319 Description: If an OST is inactive, its locks might reference stale inodes.
5320 Details : lov_change_cbdata() must iterate over all namespaces, even if
5321 they are inactive to clear inode references from the lock.
5323 Severity : enhancement
5324 Frequency : occasional, if non-standard max_dirty_mb used
5326 Description: Client will block write RPCs if not enough grant
5327 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
5328 then the client will block writes while waiting for another RPC
5329 to complete instead of consuming its dirty limit. With change
5330 we get improved performance when max_dirty_mb is small.
5332 Severity : enhancement
5333 Bugzilla : 3389, 6253
5334 Description: Add support for supplementary groups on the MDS.
5335 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
5336 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
5337 lookups for user supplementary groups into a cache.
5341 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
5342 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
5343 for liblustre O_EXCL is enforced by MDS.
5347 Description: readdir never returns NULL in liblustre.
5348 Details : Corrected llu_iop_getdirentries logic, to return offset of next
5349 dentry in struct dirent.
5353 Frequency : liblustre only, depends on application IO pattern
5354 Description: liblustre clients evicted if not contacting servers
5355 Details : Don't put liblustre clients into the ping_evictor list, so
5356 they will not be evicted by the pinger ever.
5358 Severity : enhancement
5360 Description: Add ability to evict clients by NID from MDS.
5361 Details : By echoing "nid:$NID" string into
5362 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
5363 $NID would be instantly evicted from this MDS and from all active
5364 OSTs connected to it.
5368 Description: Do not query file size twice, somewhat slowing stat(2) calls.
5369 Details : lookup_it_finish() used to query file size from OSTs that was not
5374 Description: service threads change working directory to that of init
5375 Details : Starting lustre service threads may pin the working directory
5376 of the parent thread, making that filesystem busy. Threads
5377 now change to the working directory of init to avoid this.
5381 Frequency : during shutdown only
5382 Description: shutdown with a failed MDS or OST can cause unmount to hang
5383 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
5388 Frequency : 2.6 only
5389 Description: chmod/chown may include an extra supplementary group
5390 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
5391 supplementary group and if none is specified this is used.
5394 Bugzilla : 5479 (6816)
5396 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
5397 Details : If lookup is in progress on a file that is unlinked we might try
5398 to revalidate the inode and fail in revalidate after lookup is
5399 complete and ll_file_open() enqueues the open again but
5400 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
5403 Frequency : always, if lconf --abort_recovery used
5405 Description: lconf --abort_recovery fails with 'Operation not supported'
5406 Details : lconf was attempting to abort recovery on the MDT device and not
5409 ------------------------------------------------------------------------------
5411 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
5416 Frequency : rare (only unsupported configurations with a node running as an
5418 Bugzilla : 6514, 5137
5419 Description: Mounting a Lustre file system on a node running as an OST could
5421 Details : OSTs now preallocates memory needed to write out data at
5422 startup, instead of when needed, to avoid having to
5423 allocate memory in possibly low memory situations.
5424 Specifically, if the file system is mounted on on OST,
5425 memory pressure could force it to try to write out data,
5426 which it needed to allocate memory to do. Due to the low
5427 memory, it would be unable to do so and the node would
5428 become unresponsive.
5430 Severity : enhancement
5432 Description: Addition of lconf --service command line option
5433 Details : lconf now accepts a '--service <arg>' option, which is
5434 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
5436 Severity : enhancement
5438 Description: Failover mode is now the default for OSTs.
5439 Details : By default, OSTs will now run in failover mode. To return to
5440 the old behaviour, add '--failout' to the lmc line for OSTs.
5442 Severity : enhancement
5444 Description: Health checks are now provided for MDS and OSTs
5445 Details : Additional detailed health check information on MSD and OSTs
5446 is now provided through the procfs health_check value.
5449 Frequency : occasional, depends on IO load
5451 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
5452 after numerous create/delete cycles
5453 Details : The ext3 inode allocation policy would not allocate new inodes
5454 very well on the OSTs because there are no new directories
5455 being created. Instead we look for groups with free space if
5456 the parent directories are nearly full.
5461 Description: Network or server problems during mount may cause partially
5462 mounted clients instead of returning an error.
5463 Details : The config llog parsing code may overwrite the error return
5464 code during mount error handling, returning success instead
5470 Description: MDS can fail to allocate large reply buffers
5471 Details : After long uptimes the MDS can fail to allocate large reply
5472 buffers (e.g. zconf client mount config records) due to memory
5473 fragmentation or consumption by the buffer cache. Preallocate
5474 some large reply buffers so that these replies can be sent even
5475 under memory pressure.
5479 Frequency : rare (liblustre)
5480 Description: fsx running with liblustre complained that using truncate() to
5481 extend the file doesn't work. This patch corrects that issue.
5482 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
5483 ATTR_SIZE and lsm use in llu_setattr_raw.
5487 Frequency : rare, only 2.6 kernels
5488 Description: Unusual file access patterns on the MDS may result in inode
5489 data being lost in very rare circumstances.
5490 Details : Bad interaction between the ea-in-inode patch and the "no-read"
5491 code in the 2.6 kernel caused the inode and/or EA data not to
5492 be read from disk, causing single-file corruption.
5496 Frequency : rare, only 2.6 filesystems using extents
5497 Description: Heavy concurrent write and delete load may cause data corruption.
5498 Details : It was possible under high-load situations to have an extent
5499 metadata block in the block device cache from a just-unlinked
5500 file overwrite a newly-allocated data block. We now unmap any
5501 metadata buffers that alias just-allocated data blocks.
5505 Frequency : filesystems with default stripe_count larger than 77
5506 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
5507 Details : lconf specifies an inode size of 4096 bytes when the default
5508 stripe_count is larger than 77. This conflicts with the default
5509 inode density of 1 per 4096 bytes. Allocate smaller inodes in
5510 this case to avoid pinning too much memory for large EAs.
5512 ------------------------------------------------------------------------------
5514 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
5519 Frequency : rare (extremely heavy IO load with hundreds of clients)
5521 Description: Client is evicted, gets IO error writing to file
5522 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
5523 caused clients to be evicted for AST timeouts. The fixes in
5524 bug 5192 mean we no longer need to have such short AST timeouts
5525 so ldlm_timeout has been increased.
5528 Frequency : occasional during --force or --failover shutdown under load
5529 Bugzilla : 5949, 4834
5530 Description: Server oops/LBUG if stopped with --force or --failover under load
5531 Details : a collection of import/export refcount and cleanup ordering
5532 issues fixed for safer force cleanup
5535 Frequency : only filesystems larger than 120 OSTs
5536 Bugzilla : 5990, 6223
5537 Description: lfs getstripe would oops on a very large filesystem
5538 Details : lov_getconfig used kfree on vmalloc'd memory
5541 Frequency : only filesystems exporting via NFS to Solaris 10 clients
5542 Bugzilla : 6242, 6243
5543 Description: reading from files that had been truncated to a non-zero size
5544 but never opened returned no data
5545 Details : ll_file_read() reads zeros from no-object files to EOF
5550 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
5551 Details : There's little harm in aborting MDS/OSS recovery and letting it
5552 try again, so I removed the LASSERT and return an error instead.
5554 Severity : enhancement
5556 Description: New debugging infrastructure for tracking down data corruption
5557 Details : The I/O checksum code was replaced to: (a) control it at runtime,
5558 (b) cover more of the client-side code path, and (c) try to narrow
5559 down where problems occurred
5563 Bugzilla : 3819, 4364, 4397, 6313
5564 Description: Racing close and eviction MDS could cause assertion in mds_close
5565 Details : It was possible to get multiple mfd references during close and
5566 client eviction, leading to one thread referencing a freed mfd.
5568 Severity: : enhancement
5569 Bugzilla : 3262, 6359
5570 Description: Attempts to reconnect to servers are now more aggressive.
5571 Details : This builds on the enhanced upcall-less recovery that was added
5572 in 1.4.2. When trying to reconnect to servers, clients will
5573 now try each server in the failover group every 10 seconds. By
5574 default, clients would previously try one server every 25 seconds.
5579 Description: After recovery, certain operations trigger a failed
5580 assertion on a client.
5581 Details : Failing over an mds, using lconf -d --failover, while a
5582 client was doing a readdir() call would cause the client to
5583 LBUG after recovery completed and the readdir() was resent.
5585 Severity : enhancement
5587 Description: Default groups are now added by lconf
5588 Details : You can now run lconf --group <servicename> without having to
5589 manually add groups with lmc.
5592 Frequency : occasional
5594 Description: Nodes with an elan id of 0 trigger a failed assertion
5597 Frequency : always when accessing e.g. tty/console device nodes
5599 Description: tty and some other devices nodes cannot be used on lustre
5600 Details : file's private_data field is used by device data and lustre
5601 values in there got lost. New field was added to struct file to
5602 store fs-specific private data.
5605 Frequency : when exporting Lustre via NFS
5607 Description: NFSD failed occasionally when looking up a path component
5608 Details : NFSD is looking up ".." which was broken in ext3 directories
5609 that had grown large enough to become hashed.
5612 Frequency : Clusters with multiple interfaces not on the same subnet
5614 Description: Nodes will repeatedly try to reconnect to an interface which it
5615 cannot reach and report an error to the log.
5616 Details : Extra peer list entries will be created by lconf with some peers
5617 unreachable. lconf now validates the peer before adding it.
5620 Frequency : Only if a default stripe is set on the filesystem root.
5622 Description: Setting a default stripe on the filesystem root prevented the
5623 filesystem from being remounted.
5624 Details : The client was sending extra request flags in the root getattr
5625 request and did not allocate a reply buffer for the dir EA.
5628 Frequency : occasional, higher if lots of files are accessed by one client
5629 Bugzilla : 6159, 6097
5630 Description: Client trips assertion regarding lsm mismatch/magic
5631 Details : While revalidating inodes the VFS looks up inodes with ifind()
5632 and in rare cases can find an inode that is being freed.
5633 The ll_test_inode() code will free the lsm during ifind()
5634 when it finds an existing inode and then the VFS later attaches
5635 this free lsm to a new inode.
5639 Bugzilla : 6422, 7030
5640 Description: MDS deadlock between mkdir and client eviction
5641 Details : Creating a new file via mkdir or mknod (starting a transaction
5642 and getting the ns lock) can deadlock with client eviction
5643 (gets ns lock and trying to finish a synchronous transaction).
5646 Frequency : occasional
5647 Description: While starting a server, the fsfilt_ext3 module could not be
5649 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
5650 kernels. Previously, lconf would still use the ext3 name
5651 when trying to load modules. Now, it will correctly use
5652 ext3 on 2.4 and ldiskfs on 2.6.
5654 Severity : enhancement
5655 Description: The default stripe count has been changed to 1
5656 Details : The interpretation of the default stripe count (0, to lfs
5657 or lmc) has been changed to mean striping across a single
5658 OST, rather than all available. For general usage we have
5659 found a stripe count of 1 or 2 works best.
5661 Severity : enhancement
5662 Description: Add support for compiling against Cray portals.
5663 Details : Conditional compiling for some areas that are different
5667 Frequency : occasional
5668 Bugzilla : 6409, 6834
5669 Description: Creating files with an explicit stripe count may lead to
5670 a failed assertion on the MDS
5671 Details : If some OSTs are full or unavailable, creating files may
5672 trigger a failed assertion on the MDS. Now, Lustre will
5673 try to use other servers or return an error to the
5677 Frequency : occasional
5679 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
5680 caused lock timeout during readahead (since 1.4.2).
5681 Details : Processes doing readahead might match a lock that hasn't been
5682 granted yet if there are overlapping and conflicting lock
5683 requests. The readahead process waits on ungranted lock
5684 (original lock is CBPENDING), while OST waits for that process
5685 to cancel CBPENDING read lock and eventually evicts client.
5687 Severity : enhancement
5689 Description: Initial enabling of flock support for clients
5690 Details : Implements fcntl advisory locking and file status functions.
5691 This feature is provided as an optional mount flag (default
5692 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
5693 locking are implemented yet, and those that are are not guaranteed
5694 to be completely correct in production environments.
5695 mount -t lustre -o [flock|noflock] ...
5698 Frequency : occasional
5700 Description: OSTs running 2.4 kernels but with extents enabled might trip an
5701 assertion in the ext3 JBD (journaling) layer.
5702 Details : The b_committed_data struct is protected by the big kernel lock
5703 in 2.4 kernels, serializing journal_commit_transaction() and
5704 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
5705 access to this struct. In 2.6 kernels there is finer grained
5706 locking to improve SMP performance of the JBD layer.
5710 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
5715 Description: merge and cleanup kernel patches.
5718 -----------------------------------------------------------------------------
5720 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
5722 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
5723 versions of Lustre. Please update all servers and clients to
5724 version 1.4.2 or later at the same time. You must also run
5725 "lconf --write-conf {config}.xml" on the MDS while it is stopped
5726 to update the configuration logs.
5728 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
5729 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
5730 - swab llog records of type '0' so we get proper header size/idx (5861)
5731 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
5732 - fix rename of one directory over another leaking an inode (5953)
5733 - avoid SetPageDirty on 2.6 (5981)
5734 - don't re-add just-being-destroyed locks to the waiting list (5653)
5735 - when creating new directories, inherit the parent's custom
5736 striping settings if present parent (3048)
5737 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
5738 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
5739 - don't hold client locks on temporary worklist from l_lru (5666)
5740 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
5741 - automatically evict dead clients (5921)
5742 - Update file size properly in create+truncate+fstat case (6196)
5743 - Do not unhash mountpoint dentries, do not allow removal of
5745 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
5746 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
5747 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
5748 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
5749 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
5750 - Correctly handle reads of files with no objects (6243)
5751 - lctl recover will also mark a device active if deactivate used (5933)
5753 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
5754 - allow --write-conf on an MDS with different nettype than client (5619)
5755 - don't write config llogs to MDS for mounts not from that MDS (5617)
5756 - lconf should create multiple TCP connections from a client (5201)
5757 - init scripts are now turned off by default; run chkconfig --on
5758 lustre and chkconfig --on lustrefs to use them
5759 - upcalls are no longer needed for clients to recover to failover
5761 - add --abort-recovery option to lconf to abort recovery on device
5763 - add support for an arbitrary number of OSTs (3026)
5764 - Quota support protocol changes.
5765 - forward compatibility changes to wire structs (6007)
5766 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5767 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5768 - improved "lustre status" script
5769 - initialize blocksize for non-regular files (6062)
5770 - added --disable-server and --disable-client configure options (5782)
5771 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5772 - Vanilla 2.4.29 support
5773 - increase maximum number of obd devices to 520 (6242)
5774 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5775 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5776 SLES 9 2.6.5 kernels
5777 - limit stripes per file to 160 (the maximum EA size) (6093)
5779 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5782 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5783 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5784 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5785 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5786 - small ext3 extents cleanups and fixes (5733)
5787 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5788 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5789 - avoid needless client->OST connect, fix handle mismatch (5317)
5790 - fix DLM error path that led to out-of-sync client, long delays (5779)
5791 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5792 - fix several locking issues related to i_size (5492,5624,5654,5672)
5793 - don't move pending lock onto export if it is already evicted (5683)
5794 - fix kernel oops when creating .foo in unlinked directory (5548)
5795 - fix deadlock in obdfilter statistics vs. object create (5811)
5796 - use time_{before,after} to avoid timer jiffies wrap (5882)
5797 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5798 - Do not leak request if server was not able to process it (5154)
5799 - If mds_open unable to find parent dir, make that negative lookup(5154)
5800 - don't create new directories with extent-mapping (5909, 5936)
5802 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5803 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5804 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5805 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5806 - support for RHEL4 kernel on i686 (5773)
5807 - provide error messages when incompatible logs are encountered (5898)
5809 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5810 * version 1.4.0.10 (1.4.1 release candidate 1)
5812 - don't keep a lock reference when lock is not granted (4238)
5813 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5814 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5815 - drop import inflight refcount on signal_completed_replay error (5255)
5816 - unlock page after async write error during send (3677)
5817 - handle missing objects in filter_preprw_read properly (5265)
5818 - no transno return for symlink open, don't save no-trasno open (3440)
5819 - don't try to complete elan receive that already failed (4012)
5820 - free RPC server reply state on error (5406)
5821 - clean up thread from ptlrpc_start_thread() on error (5160)
5822 - readahead could read extra page into cache that wasn't ejected (5388)
5823 - prevent races in class_attach/setup/cleanup/detach (5260)
5824 - don't dereference de->d_inode after l_dput of de (5458)
5825 - use "int" for stripe value returned from lock_to_stripe (5544)
5826 - mballoc allocation and error-checking fixes in 2.6 (5504)
5827 - block device patches to fix I/O request sizes in 2.6 (5482)
5828 - look up hostnames for IB nals (5602)
5829 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5830 - don't start multiple acceptors for the same port (5277)
5831 - fix incorrect LASSERT in mds_getattr_name (5635)
5832 - export a proc file for general "ping" checking (5628)
5833 - fix "lfs check" to not block when the MDS is down (5628)
5835 - service request history (4965)
5836 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5837 - create an "evict_client" /proc entry on OSTs, like the MDS has
5838 - fix mount usage message, return errors per mount(8) (5168)
5839 - change grep [] to grep "[]" in tests so they work in more UMLs
5840 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5841 - remove ancient LOV_MAGIC_V0 EA support (5047)
5842 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5843 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5844 - allow readahead allocations to fail when low on memory (5383)
5845 - mmap locking landed again, after considerable improvement (2828)
5846 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
5848 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
5851 - send OST transaction number in read/write reply to free req (4966)
5852 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5853 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5854 - the watchdog thread now runs as interruptible (5246)
5855 - flock/lockf fixes (but it's still disabled, pending 5135)
5856 - don't use EXT3 constants in llite code (5094)
5857 - memory shortage at startup could cause assertion (5176)
5859 - reorganization of lov code
5860 - single portals codebase
5862 - add extents/mballoc support (5025)
5863 - direct I/O reads in the obdfilter (4048)
5864 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
5866 tbd Cluster File Systems, Inc. <info@clusterfs.com>
5869 - send OST transaction number in read/write reply to free req (4966)
5870 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5871 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5872 - the watchdog thread now runs as interruptible (5246)
5873 - handle missing objects in filter_preprw_read properly (5265)
5874 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5875 - drop import inflight refcount on signal_completed_replay error (5255)
5876 - unlock page after async write error during send (3677)
5877 - return original error code on reconstructed replies (3761)
5878 - no transno return for symlink open, don't save no-trasno open (3440)
5880 - add pid to ldlm debugging output (4922)
5881 - bump the watchdog timeouts -- we can't handle 30sec yet
5882 - extra debugging for orphan dentry/inode bug (5259)
5884 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
5887 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
5888 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
5889 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
5890 - fix NULL dereference in /proc/sys/portals/routes (4827)
5891 - allow failed mdc_close() operations to be interrupted (4561)
5892 - stop precreate on OST before MDS would time out on it (4778)
5893 - don't send partial-page writes before EOF from client (4410)
5894 - discard client grant for sub-page writes on large-page clients (4520)
5895 - don't free dentries not owned by NFS code, check generation (4806)
5896 - fix lsm leak if mds_create_objects() fails (4801)
5897 - limit debug_daemon file size, always print CERROR messages (4789)
5898 - use transno after validating reply (3892)
5899 - process timed out requests if import state changes (3754)
5900 - update mtime on OST during writes, return in glimpse (4829)
5901 - add mkfsoptions to LDAP (4679)
5902 - use ->max_readahead method instead of zapping global ra (5039)
5903 - don't interrupt __l_wait_event() during strace
5905 - add software watchdogs to catch hung threads quickly (4941)
5906 - make lustrefs init script start after nfs is mounted
5907 - fix CWARN/ERROR duplication (4930)
5908 - return async write errors to application if possible (2248)
5909 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
5910 - print NAL number in %x format (4645)
5911 - update barely-supported suse-2.4.21-171 series (4842)
5912 - support for sles 9 %post scripts
5913 - support for building 2.6 kernel-source packages
5914 - support for sles km_* packages
5916 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
5919 - ignore -ENOENT errors in osc_destroy (3639)
5920 - notify osc create thread that OSC is being cleaned up (4600)
5921 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
5922 - reconstruct ost_handle() like mds_handle() (4657)
5923 - create a new thread to do import eviction to avoid deadlock (3969)
5924 - let lconf resolve symlinked-to devices (4629)
5925 - don't unlink "objects" from directory with default EA (4554)
5926 - hold socknal file ref over connect in case target is down (4394)
5927 - allow more than 32000 subdirectories in a single directory (3244)
5928 - fix blocks count for O_DIRECT writes (3751)
5929 - OST returns ENOSPC from object create when no space left (4539)
5930 - don't send truncate RPC if file size isn't changing (4410)
5931 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
5932 - bind to privileged port in socknal and tcpnal (3689)
5934 - rate limit CERROR/CWARN console message to avoid overload (4519)
5935 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
5936 - basic mmap support (3918)
5937 - kernel patch series update from b1_4 (4711)
5939 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
5942 - avoid crash during MDS cleanup with OST shut down (2775)
5943 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
5944 - don't use bad inodes on the MDS (3744)
5945 - dynamic object preallocation to improve recovery speed (4236)
5946 - don't hold spinlock over lock dumping or change debug flags (4401)
5947 - don't zero obd_dev when it is force cleaned (3651)
5948 - print grants to console if they go negative (4431)
5949 - "lctl deactivate" will stop automatic recovery attempts (3406)
5950 - look for existing locks in ldlm_handle_enqueue() (3764)
5951 - don't resolve lock handle twice in recovery avoiding race (4401)
5952 - revalidate should check working dir is a directory (4134)
5954 - don't always mark "slow" obdfilter messages as errors (4418)
5956 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
5959 - don't close LustreDB during write_conf until it is done (3860)
5960 - fix typo in lconf for_each_profile (3821)
5961 - allow dumping logs from multiple threads at one time (3820)
5962 - don't allow multiple threads in OSC recovery (3812)
5963 - fix debug_size parameters (3864)
5964 - fix mds_postrecov to initialize import for llog ctxt (3121)
5965 - replace config semaphore with spinlock (3306)
5966 - be sure to send a reply for a CANCEL rpc with bad export (3863)
5967 - don't allow enqueue to complete on a destroyed export (3822)
5968 - down write_lock before checking llog header bitmap (3825)
5969 - recover from lock replay timeout (3764)
5970 - up llog sem before sending rpc (3652)
5971 - reduce ns lock hold times when setting kms (3267)
5972 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
5973 - fix NULL deref and obd_dev leak on setup error (3312)
5974 - replace some LBUG about llog ops with error handling (3841)
5975 - don't match INVALID dentries from d_lookup and spin (3784)
5976 - hold dcache_lock while marking dentries INVALID and hashing (4255)
5977 - fix invalid assertion in ptlrpc_set_wait (3880)
5979 - add libwrap support for the TCP acceptor (3996)
5980 - add /proc/sys/portals/routes for non-root route listing (3994)
5981 - allow setting MDS UUID in .xml (2580)
5982 - print the stack of a process that LBUGs (4228)
5984 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
5987 - don't cleanup request in ll_file_open() on failed MDS open (3430)
5988 - make sure to unset replay flag from failed open requests (3440)
5989 - if default stripe count is 0, use OST count for inode size (3636)
5990 - update parent mtime/ctime on client for create/unlink (2611)
5991 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
5992 - free recovery state on server during a forced cleanup (3571)
5993 - unregister_reply for resent reqs (3063)
5994 - loop back devices mounting and status check on 2.6 (3563)
5995 - fix resource-creation race that can provoke i_size == 0 (3513)
5996 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
5997 - more debugging for page-accounting assertion (3746)
5998 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
5999 - avoid deadlock after precreation failure (3758)
6000 - fix race and lock order deadlock in orphan handling (3450, 3750)
6001 - add validity checks when grabbing inodes from l_ast_data (3599)
6003 - add /proc/.../recovery_status to obdfilter (3428)
6004 - lightweight CDEBUG infrastructure, debug daemon (3668)
6005 - change default OSC RPC parameters to be better on small clusters
6006 - turn off OST read cache for files smaller than 32MB
6007 - install man pages and include them in rpms (3100)
6008 - add new init script for (un)mounting lustre filesystems (2593)
6009 - run chkconfig in %post for init scripts (3701)
6010 - drop scimac NAL (unmaintained)
6012 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
6015 - clean kiobufs before and after use (3485)
6016 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
6017 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
6018 - change init script to fail more clearly if not run as root (1528)
6019 - allow clients to reconnect during replay (1742)
6020 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
6021 - don't do DNS lookups on NIDs too small for IP addresses (3442)
6022 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
6023 - fix cond_resched (3554)
6024 - only evict unfinished clients after recovery (3515)
6025 - allow bulk resend, prevent data loss (3570)
6026 - dynamic ptlrpc request buffer allocation (2102)
6027 - don't allow unlinking open directory if it isn't empty (2904)
6028 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
6029 - remove extraneous obd dereference causing LASSERT failure (3334)
6030 - don't use get_cycles() when creating temp. files on the mds (3156)
6031 - hold i_sem when setting i_size in ll_extent_lock() (3564)
6032 - handle EEXIST for set-stripe, set proper directory name (3336)
6034 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
6035 - fix ksocknal_fmb_callback() error messages (2918)
6037 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
6040 - don't copy lvb into (possibly NULL) reply on error (2983)
6041 - don't deref dentry after dput, don't free lvb on error (2922)
6042 - use the kms to determine writeback rpc length (2947)
6043 - increment oti_logcookies when osc is inactive (2948)
6044 - update client's i_blocks count via lvb messages (2543)
6045 - handle intent open/close of special files properly (1557)
6046 - mount MDS with errors=remount-ro, like obdfilter (2009)
6047 - initialize lock handle to avoid ASSERT on error cleanup (3057)
6048 - don't use cancelling-locks' kms values (2947)
6049 - use highest lock extent for kms, not last one (2925)
6050 - don't dereference ERR_PTR() dentry in error handling path (3107)
6051 - fix thread race in portals_debug_dumplog() (3122)
6052 - create lprocfs device entries at setup instead of at attach (1519)
6053 - common AST error handler, don't evict client on completion race (3145)
6054 - zero nameidata in detach_mnt in 2.6 (3118)
6055 - verify d_inode after revalidate_special is valid in 2.6 (3116)
6056 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
6057 - initialize RPC timeout timer earlier for 2.6 (3219)
6058 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
6059 - print nal/nid for unknown nid (3258)
6060 - additional checks for oscc recovery before doing precreate (3284)
6061 - fix ll_extent_lock() error return code for 64-bit systems (3043)
6062 - don't crash in mdc_close for bad permissions on open (3285)
6063 - zero i_rdev for non-device files (3147)
6064 - clear page->private before handing to FS, better assertion (3119)
6065 - tune the read pipeline (3236)
6066 - fix incorrect decref of invalidated dentry (2350)
6067 - provide read-ahead stats and refine rpc in flight stats (3328)
6068 - don't hold journal transaction open across create RPC (3313)
6069 - update atime on MDS at close time (3265)
6070 - close LDAP connection when recovering to avoid server load (3315)
6071 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
6072 - don't leak open file on MDS after open resend (3325)
6073 - serialize filter_precreate and filter_destroy_precreated (3329)
6074 - loop device shouldn't call sync_dev() for nul device (3092)
6075 - clear page cache after eviction (2766)
6076 - resynchronize MDS->OST in background (2824)
6077 - refuse to mount the same filesystem twice on same mountpoint (3394)
6078 - allow llmount to create routes for mounting behind routers (3320)
6079 - push lock cancellation to blocking thread for glimpse ASTs (3409)
6080 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
6081 - fix rare problem with rename on htree directories (3417)
6083 - allow default OST striping configuration per directory (1414)
6084 - fix compilation for qswnal for 2.6 kernels (3125)
6085 - increase maximum number of MDS request buffers for large systems
6086 - change liblustreapi to be useful for external progs like lfsck (3098)
6087 - increase local configuration timeout for slow disks (3353)
6088 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
6090 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
6093 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
6094 - don't overwrite extent policy data in reply if lock was blocked (2901)
6095 - drop filter export grants atomically with removal from device (2663)
6096 - del obd_self_export from work_list in class_disconnect_exports (2908)
6097 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
6098 - swab reply message in mdc_close, other PPC fixes (2464)
6099 - fix destroying of named logs (2325)
6100 - overwrite old logs when running lconf --write_conf (2264)
6101 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
6102 - fix race in target_handle_connect (2898)
6103 - mds_reint_create() should take same inode create lock (2926)
6104 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
6105 - don't close files for self_export to avoid uninitialized obd (2936)
6106 - allow MDS with the same name as client node (2939)
6107 - hold dentry reference for closed log files for unlink (2325)
6108 - reserve space for all logs during transactions (2059)
6109 - don't evict page beyond end of stripe extent (2925)
6110 - don't oops on a deleted current working directory (2399)
6111 - handle hard links to targets without a parent properly (2517)
6112 - don't dereference NULL lock when racing during eviction (2867)
6113 - don't grow lock extents when lots of conflicting locks (2919)
6115 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
6118 - account for cache space usage on clients to avoid data loss (974)
6119 - lfsck support in lustre kernel code (2349)
6120 - reduce journal credits needed for BRW writes (2370)
6121 - orphan handling to avoid losing space on client/server crashes
6122 - ptlrpcd can be blocked, stopping ALL progress (2477)
6123 - use lock value blocks to assist in proper KMS, faster stat (1021)
6124 - takes i_sem instead of DLM locks internally on obdfilter (2720)
6125 - recovery for initial connections (2355)
6126 - fixes for mds_cleanup_orphans (1934)
6127 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
6128 - block all file creations until orphan recovery completes (1901)
6129 - client remove rq_connection from request struct (2423)
6130 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
6131 - recovery timer race (2670)
6132 - mdc_close recovey bug (2532)
6133 - ptlrpc cleanup bug (2710)
6134 - mds timeout on local locks (2588)
6135 - namespace lock held during RPCs (2431)
6136 - handle interrupted sync write properly (2503)
6137 - don't try to handle a message that hasn't been replied to (2699)
6138 - client assert failure during cleanup after abort recovery (2701)
6139 - leak mdc device after failed mount (2712)
6140 - ptlrpc_check_set allows timedout requests to complete (2714)
6141 - wait for inflight reqs when ptlrpcd finishes (2710)
6142 - make sure unregistered services are removed from the srv_list
6143 - reset bulk XID's when resending them (caught by 1138 test)
6144 - unregister_bulk after timeout
6145 - fix lconf error (2694)
6146 - handle write after unfinished setstripe, stripe-only getstripe (2388)
6147 - readahead locks pages, leaves pending causing memory pressure (2673)
6148 - increase OST request buffers to 4096 on large machines (2729)
6149 - fix up permission of existing directories in simple_mkdir (2661)
6150 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
6151 - don't assign transno to errored transactions (2742)
6152 - don't delete objects on OST if given a bogus objid from MDS (2751)
6153 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
6154 - if rq_no_resend, then timeout request after recovery (2432)
6155 - fix MDS llog_logid record size, 64-bit array alignment (2733)
6156 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
6157 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
6158 - MDS recovery shouldn't delete objects that it has given out (2730)
6159 - if enqueue arrives after completion, don't clobber LVB (2819)
6160 - don't unlock pages twice when trigger_group_io returns error (2814)
6161 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
6162 - don't write pages to disk if there was an error (1450)
6163 - don't ping imports that have recovery disabled (2676)
6164 - take buffered bytes into account when balancing socknal conn (2817)
6165 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
6166 - reconnect unlink llog connection after MDS reconnects to OST (2816)
6167 - remove little-endian swabbing of llog records (1987)
6168 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
6169 - retry reposting request buffers if they fail (1191)
6170 - grow extent at grant time to avoid granting a revoked lock (2809)
6171 - lock revoke doesn't evict page if covered by a second lock (2765)
6172 - disable VM readahead to avoid reading outside lock extents (2805)
6174 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
6175 - updated kernel patches for hp-2.4.20 kernel (2681)
6177 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
6180 - fix truncated write corruption (2366)
6181 - fix for failed assertion in iopen_connect_dentry (1792,2517)
6183 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
6184 - ensure len(uuid) < 37 in lmc (1171)
6185 - fix ia64 OOPS in llog_test (2255)
6186 - zero end of page at obdfilter for partial page writes (2648)
6187 - don't leave stale dentries around after renames (bug 2428)
6188 - fix timeouts when evicting a client with a single lock held (2642)
6189 - set deadline for the initial HELLO message to drain (2634)
6190 - print out dotted-quad IP addresses in the socknal (2302)
6192 - additional debugging for MDS client eviction problem (2443)
6193 - fix mkfsoptions support for osts (2603, 2604)
6195 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
6198 - add series for the vanilla 2.6.0 kernel
6199 - add series for the vanilla 2.4.24 kernel
6200 - add series for a cray x86/64 UL kernel drop
6201 - fix xattr patches for the vanilla 2.4.19 series
6203 - generate true UUIDs in lmc (1171)
6204 - have portals stack dumping break in UML (2466)
6205 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
6206 - allocate with _NOFS in ldlm to avoid deadlock (1933)
6207 - wake callback waiting threads on client eviction (2460)
6208 - Add --ptldebug and --subsystem to lmc (1719)
6209 - update assertion to allow safe interrupt allocation
6210 - set rq_no_resend for cancel requests (2432)
6211 - recalculate ptlrpcd timeout after resend (2494)
6212 - call vfs_rmdir when removing pending directories (2368)
6213 - fix renaming a file to itself (2429)
6214 - lmc creates a default one-stripe lov (2454)
6215 - expand procfs space to handle large clusters (2326)
6216 - increase UML stack to avoid overflow
6217 - update lconf's list of debug and subsystem masks
6218 - fix lfs find --obd (2510)
6219 - /proc tunable for disabling filter read caching (2591)
6220 - stop rpm packages from altering slapd.conf (2301)
6221 - disable nagle in the socknal under 0conf (2578)
6222 - choose mds inode size based on stripe count (2572)
6223 - fix kernel-source rpm problems (2516)
6225 - add --disable-doc to avoid pdf generation (2421)
6226 - update documentation, tests, type-os, comments
6227 - avoid format warnings on ia64
6228 - remove the TOE NAL
6229 - tiny code cleanups by removing unused fields
6231 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
6234 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
6235 - remove the most blatant lies from BUILDING file (2371)
6236 - change default debug level to reasonable production setting
6237 - reduce client side cache size to reduce cache flush time
6238 - reduce max RPCs in flight to avoid unnecessary file fragmentation
6239 - make TCP zerocopy and pinger support enabled by default (2476)
6240 - sync writes completed after process exits caused crashes (2319)
6241 - maintain correct mount count on the MDS (2356)
6242 - backout 1557, because 2316 wasn't really fixed
6243 - better file I/O statistics gathering in /proc
6244 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
6245 - another kernel patch to fix zero-copy TCP function export
6246 - don't take duplicate lock when processing re-sent getattr (2420)
6247 - lctl uses obd_self_export instead of creating new conn (2353)
6248 - MDS/OST recovery case which requires object creation asserted (2425)
6249 - move lfs from /usr/sbin to /usr/bin in packages
6250 - fix race between mds_client_add and mds_client_free (2417)
6251 - use kmalloc instead of slabs in portals (2430)
6252 - don't create duplicate records when a failover MDS is present (2442)
6253 - remove unnecessary mount age check (2332)
6254 - don't remove directory inodes from locks prematurely (2451)
6255 - don't break if MDS service name is the same as hostname (2103)
6256 - fix races in client write RPC generation when cache full (2482)
6258 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
6261 - remove now-unused request->rq_obd (278)
6262 - if an allocation fails, print out how much memory we've used (1933)
6263 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
6264 - add the "configurable stack size" patch to most series files (1256)
6265 - ability to write large log records, for 100+ OST configs (2306)
6266 - fix NULL deref when filter_prep fails (2314)
6267 - fix operator precedence error in filter_sync
6268 - dynamic allocation of socknal TX descriptors (2315)
6269 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
6270 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
6271 - ensure that CWARN messages in Portals always get to the syslog
6272 - __init/__exit are not for prototype decls (ldlm_init/exit)
6273 - x86-64 compile warning fixes
6274 - fix gateway LMC keyword conflict (2318)
6275 - fix MDS lock inversions in getattr/reint paths (1844)
6276 - fix a rare lock re-ordering bug, which caused deadlock (2322)
6277 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
6278 - DLM race condition prevented some lock evictions (2328)
6279 - ENOMEM detection and retry on socknal sends (2230)
6280 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
6281 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
6282 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
6283 - better error messages when a client is rejected during recovery (1505)
6284 - avoid cancelling locks which were never granted, after failure (2330)
6285 - fix i_sem/journal inversion in mds_client_add (2333)
6286 - fix truncate/getattr lock cycle deadlock (2334)
6287 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
6288 - fix two rare exit paths which could leak an l_lock() ref (2321)
6289 - fencepost error in MDS/OST orphan recovery (2226)
6290 - make log record alignment 8 bytes (1988)
6291 - lstripe now fails when requested offset > ost_count (2237)
6292 - ensure that all kernel series have a complete list.h (1607)
6293 - fix crashes in special-file operations (2316)
6294 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
6296 - allow configurable automake binary, for testing new versions
6297 - small update to the lfs documentation
6299 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
6301 * fix negative export reference count in fsfilt_sync (2312)
6303 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
6304 * release candidate 0.9.1
6306 - orphans are moved into the PENDING directory for possible recovery
6307 - replayed opens now open by fid for orphan/rename safety (1042)
6308 - last close of an orphan inode generates a transno (683)
6309 - chdir() and mount() now pin the directory entry (1020)
6310 - avoid CERROR in normal ll_setattr_raw() error case (1500)
6311 - discard very old requests without processing them (1502)
6312 - remove some common, well-understood CERRORs (1505)
6313 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
6314 - clear "grant" flags in OST replies until OST grant code lands (1644)
6315 - fix read performance by not clobbering i_blksize on client (1598)
6316 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
6317 - make LRU size a /proc tunable, clears locks when reduced (707)
6318 - fix some lprocfs rot that prevented ptlbd from loading (1732)
6319 - server locks take references on exports now (1558)
6320 - build fixes for 2.4.20-rh trees (1663)
6321 - return an error from lov_create if all OSCs are inactive (1751)
6322 - fix import levels when a reconnect happens without a timeout (1597)
6323 - exit early from mds_open if we get a lookup error (1749)
6324 - partial page read at EOF wouldn't wait for disk before sending (1642)
6325 - avoid NULL deref in obdfilter when reading page past EOF (1592)
6326 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
6327 - fix LBUG in ll_it_open_error with rc = -2 (1861)
6328 - write/truncate lock inversion (1639)
6329 - Don't auto-load obdclass, portals modules during cleanup (1495)
6330 - fix timestamps from jumping to "now" (1763)
6331 - extra journal assertions (1648)
6332 - add an extra multiunlink test (1771)
6333 - fix read_record/write_record API (1776)
6334 - fix leak of offset_extent, possible incorrect i_size later (1772)
6335 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
6336 - Debugging for the kqswnal_get_idle_tx problems (1820)
6337 - Allow recovery to be attempted multiple times (1536)
6338 - Write out MDS last_rcvd file after it is first created (1600)
6339 - Fix tx_descriptor leak in failed transmit situations (1827)
6340 - ext3 journaling fixes for assertion failure after IO error (1871)
6341 - class_export_put() on freed export after completion AST error (1896)
6342 - Fix revalidate looping in VFS (1322)
6343 - Don't access a freed export during MDS_REINT timeout (1521)
6344 - Add open-unlink recovery support on the MDS (1673,1764)
6345 - Return an error if no MDS data was read from last_rcvd (1946)
6346 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
6347 - Don't setup a disk device that doesn't match exported UUID (317)
6348 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
6349 - avoid committing NULL handle in force close
6350 - local.sh is now a one-stripe LOV configuration
6351 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
6352 - fix ext3 htree duplicate directory entry corruption (1516)
6353 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
6354 - update to new LOV EA format (2097)
6355 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
6357 2003-06-15 Phil Schwan <phil@clusterfs.com>
6360 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
6361 - per-import/export recovery handling (958, 931, 959)
6362 - multiple last-rcvd slots, for serving multiple FSes (949)
6363 - connections are again shared between multiple imp/exports (963, 964)
6364 - "umount -f" would hang if any requests needed to be sent (393, 978)
6365 - avoid pinning large req buffer by copying for queued messages (989)
6366 - add "uuid" to "lctl device" command to help upcalls (991)
6367 - "open" RPCs with transnos would confuse recovery counters (1037)
6368 - do proper endian conversion of all wire messages (288, 340, 891)
6369 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
6370 - call waiting locks callback from LDLM recovery thread (1127, 1151)
6371 - fix ptlrpc_connection leak in target_handle_connect (1174)
6372 - fix import refcounting bug in OST and MDS cleanup (1134)
6373 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
6374 - fix very unlikely obd_types race condition (501)
6375 - remove osc_open hack for echo_client (1187)
6376 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
6377 - a failure in read_inode2 leads to deadlock (1139)
6378 - cancel ack-locks as soon as transaction is committed (1072)
6379 - fix major leaks and crashes in the bulk I/O path (937, 1057)
6380 - make sure to commitrw after any preprw to avoid deadlock (1162)
6381 - failing to execute a file in a lustre FS would lock inode (1203)
6382 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
6383 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
6384 - obd_uuid2tgt would walk off the end of the list (1255)
6385 - on IA64 the osc would give portals incorrect bulk size (1258)
6386 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
6387 - fix lock inversion caused by new llite matching code (1282)
6388 - limit the number of dirty pages on a client to 10MB (1286)
6389 - timed out locks were not being corrected cancelled (1289)
6390 - fix O_DIRECT above 4GB on IA-32 (1292)
6391 * major user-visible changes
6392 - fail out/fail over policy now controlled by the upcall (993)
6394 - add OBD_PING to check server availability and failure (954)
6395 - lustre messages are now sent in sending host order (288, 340, 891)
6396 - add eadatalen to MDS getattr reply (340)
6397 - OST read replies may contain second buffer, with per-page status (593)
6399 2003-03-11 Phil Schwan <phil@clusterfs.com>
6402 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
6403 - failed open()s could cause deadlock; fixed (867, 869)
6404 - stop cancelling OST locks when files are closed (481)
6405 - overlapping XID spaces caused network corruption (851, 853)
6406 - fix unsafe fsfilt counter arithmetic; change to atomic_t
6407 - setattr_raw added, to do single-RPC, server-side setattrs
6408 - lmc/lconf syntax change for OST UUIDs
6409 - fix crashy race condition between ptlrpc_free_req and osc_close
6410 - don't use request in mdc_enqueue if we hit a timeout (889)
6411 - don't set the inode i_size for regular files from the MDS (896)
6412 - handle out of order completion AST (842)
6413 - don't LBUG if a lock request times out after receiving AST (913)
6414 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
6415 - if a bad lock AST arrives, send an error instead of dropping entirely
6416 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
6417 - fix leak in bulk IO when only partially completed (899, 900, 926)
6418 - fix O_DIRECT for ia64 (55)
6419 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
6420 - C-z after timeout could hang a process forever; fixed (977)
6422 - client-side I/O cache (678, 924, 929, 941, 970)
6424 - READPAGE and SETATTRs which don't take server-side locks get
6427 2003-02-11 Phil Schwan <phil@clusterfs.com>
6430 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
6431 - Fix fsfilt_extN_readpage() to read a full page of directory
6432 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
6433 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
6434 - Fix imbalanced LOV object allocation and out-of-bound access (469)
6435 - Most intent operations were removed, in favour of a new RPC mode
6436 that does a single RPC to the server and bypasses most of the VFS
6437 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
6438 - Aggressively cancel local locks on DLM servers
6439 - mds_reint_unlink sends EA to the client if it's the last nlink.
6440 client uses that EA to unlink OST objects.
6441 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
6442 - recursive symlinks were fixed (439)
6443 - fixed NULL deref in DEBUG_REQ
6444 - filter_update_lastobjid no longer calls sync, which annoyed extN
6445 - fixed multi-client small-writes to a single file problem (445)
6446 - fixed mtime updates during file writes (607)
6447 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
6448 - fixed bug in obd_brw_read/write() (under guise of testing 367)
6449 - fixed Linux OST size reporting problem (444, 656)
6450 - OST now updates object mtime with writes or setattr (607, 619)
6451 - client verifies file size before zeroing page past EOF (445)
6452 - OST now writes last allocated objid to disk with allocation (108)
6453 - LOV on echo now works (409)
6455 - mds_reint_unlink sends a new buffer, with the EA included. this
6456 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
6457 set if a regular file was being unlinked, and it was the last link
6458 - use PtlGet from the target for bulk writes (315)
6459 - OST now updates object mtime with writes or setattr (607, 619)
6460 - LDLM now has a grant-time callback to revalidate locked items, if
6462 - Many MDS operations were reorganized to combat race conditions
6464 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
6465 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
6467 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
6470 - Fully reactivate OST imports after reconnection (512, others)
6471 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
6472 - More graceful error handling for truncating on dead OST (515)
6473 - Don't error out unless we're actually accessing dead stripes (474)
6474 - Fix garbage sizes when stripes are missing (410)
6475 - LRU counters were broken, causing constant lock purge (433, 432)
6476 - garbage on read from stripes with failed OSTs (441)
6477 - mark OSCs as active before reconnecting during recovery (438)
6478 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
6479 - lfind did not preserve OST order in output (443)
6480 - symlinks cause hung clients, incorrect data (439)
6481 - stop dereferencing request after dropping refcount (457)
6482 - don't LASSERT(spin_is_locked) on non-SMP (455)
6483 - fixes for many rename() bugs
6484 - fstat didn't correctly synchronize attributes (399)
6485 - server must handle lock cancellation during blocking AST prep (487)
6486 - bulk descriptors were free()d too soon (511)
6487 - fix paths in lconf, which would load incorrect modules (451, 507)
6488 - fix confusing lconf 'host not found' error message (386)
6489 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
6490 - fix race condition in mdc_blocking_ast() for inode access (526)
6491 - fix lov_unpackmd() unpacking wrong number of stripes (537)
6492 - fix lov_set_osc_active() marking wrong OSC inactive (440)
6493 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
6494 - fix multiple writes of stripe MD to MDS (358, maybe 519)
6495 - fix lstripe in several ways (kernel side) (527)
6496 - fix request leak in ldlm_cli_enqueue (262)
6497 - incorrect OSC was marked inactive after OST failure
6498 - call mds_fs_cleanup before unmounting filesystem (524)
6499 - fix races between taking ns_lock and ldlm_lock_change_resource
6500 - fix races updating LOV export open file list
6501 - fix lov_enqueue error path, avoid decref-ing bad lock handle
6502 - fix recovery NULL deref in ldlm_cli_cancel_unused
6503 - fix some DLM races by using new hash table for lock handles (419)
6504 - permit the client to specify desired inodes, at replay
6505 - duplicate requests when we queue them for replay reintegration
6506 - fix last_rcvd offset calculation
6507 - sync after each recovered transaction, so we always make progress
6508 - never, not always, ERESTART requests without transnos
6509 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
6511 - skip replay if the MDS says that the client is already connected
6512 - don't check for a recovery-enabled export to match lctl's UUID
6513 - don't INC_USE_COUNT for phantom exports
6514 - don't crash when cleaning up phantom exports (567)
6515 - don't double-finish or set replay data for errored mdc_open requests
6516 - abort requests when they time out, so we don't get old replies
6517 - send/receive replies for AST messages again
6518 - if the client says that it doesn't have the lock, cancel it on the
6520 - if we timeout during I/O, don't try to cancel an in-use lock; instead
6521 mark it as destroyed, it will all work out when decref is called
6522 - fix module use counts (22, 581)
6524 - ASTs now expect a reply (server cancels lock on error reply)
6526 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
6529 - fix many simultaneous client startup (392)
6530 - fix dentry->d_it clobbering
6531 - credentials weren't being shipped for readdir/getattr operations
6532 - remove invalid assertions triggered during some concurrent MD
6534 - proper Lustre versions added (336, 389)
6535 - fix memory leak for create error case (398)
6536 - fix LOV locking bug that would get cli/srv out of sync
6537 - fix echo client over LOV (409)
6538 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
6539 - fix double-O_EXCL intent crash (424)
6540 - avoid sending multiple lock CANCELs (352)
6542 - MDS can do multi-client recovery (modulo bugs in new code)
6544 - many updates, edits, cleanups
6546 2002-11-18 Phil Schwan <phil@clusterfs.com>
6549 - fix null d_it dereference (346)
6550 - fix full OST/dbench hang (333)
6551 - fix permission problem with file removal (286)
6552 - fix removal of OSCs from LOV when they fail
6553 - fix NULL deref during bulk timeout (214)
6554 - fix problems related to multiple filesystems on one MDS (241)
6555 - fixed serious subtle metadata locking bugs
6556 - free locks on clients when inodes are removed due to memory
6558 - fix inode pointer in lock data (285)
6559 - partial support for multiple MDS on a single host (241)
6560 - data locks weren't cancelled at clear_inode time (290, 311)
6561 - intent locks could lead to unbounded lock growth (205)
6562 - added a maximum lock count, an LRU list, and a flusher
6563 - fix multiple rename (365)
6564 - properly abstracted the echo client
6565 - OSC locked 1 byte too many; fixed
6566 - rewrote brw callback code:
6567 - fixed recovery bugs related to LOVs (306)
6568 - fixed too-many-pages-in-one-write crash (191)
6569 - fixed (again) crash in sync_io_timeout (214)
6570 - probably fixed callback-related race (385)
6572 - Add capability to MDS protocol
6573 - LDLM cancellations and callbacks on different portals
6575 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
6578 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
6579 - defer open object destruction to close time (601981, 138)
6580 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
6581 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
6582 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
6583 - don't call obd_setattr in ll_file_release for destroyed objects
6584 * protocol change to lustre_msg: move |version| and add |flags|
6585 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
6586 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
6587 * added replay of create, unlink, link and rename operations during
6588 MDS failover; recovery should be much more robust now
6589 * remove failed OSCs from LOVs (only lov_create uses this so far)
6590 * the lustre-HOWTO was brought (more) up to date (582544)
6592 2002-10-23 Phil Schwan <phil@clusterfs.com>
6595 - in-use dentries weren't being reused properly (617851)
6596 - prevent multiple LDLM setup (599178)
6597 - fix LOV size calculations for truncate (617853)
6598 - fix client handling of MDS intent errors (POSIX)
6599 - fix permission bug in lovstripe.c test (624321)
6600 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
6601 - truncate past end of file could corrupt data
6602 - proper cleanup after timeouts, crashes, etc (592524, 550815)
6603 - a race in recovery could return ETIMEDOUT to apps (623947)
6604 - building outside the source directory was fixed
6605 * the lustre-HOWTO was brought (more) up to date (582544)
6606 * major progress was made on recovery functionality
6608 2002-10-10 Phil Schwan <phil@clusterfs.com>
6611 - recovery deadlock fix
6612 - rm -rf causes LBUG fix (617817)
6613 - file open by multiple tasks fix (618962)
6614 - directory permissions bugs (602707 and 620007)
6615 - journal_stop fixed with locking (611313)
6616 - O_APPEND failures resolved (618273, perhaps 614459)
6617 - lconf PATH fix (619770)
6618 - IA64 build fix (621450)
6619 - RPC buffer sizes scale with amount of memory
6621 2002-10-01 Phil Schwan <phil@clusterfs.com>
6624 - locks would be cancelled without throwing away data pages,
6625 resulting in inconsistent data (605627)
6626 - inode attributes were not always being refreshed (605627, 612449)
6627 - lconf now continues to cleanup after lctl reports an error
6628 - MDS now enforces user permissions (602707)
6629 - lprocfs cleanup fixed, but not yet enabled (614157)
6630 - fixed infinite server hang, should a client not respond to an AST
6631 - avoid going into recovery if user calls readlink() with a buffer
6632 that's too small (613941)
6633 - AST RPCs no longer require replies (614867) -- this may be changed
6634 - don't crash server if client sends an IOV that's too big (611336)
6635 - fixed lock conversion deadlock (611892)
6636 - fixed the following of symlinks (614622)
6637 * recovery: the server can remove locks from a client that dies, other
6638 clients can make progress
6639 * more extN patch fixes
6640 * compile-time configurable ptlrpc buffer allocations
6642 - collaborative read cache document
6643 - Lustre Lite Performance CDR document-in-progress
6645 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6648 - fix typo in patch-2.4.18
6650 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6653 - clear ptlrpc request each time in handle_incoming_request()
6654 - unlink of files now destroys the object on the OST
6656 2002-09-19 Peter Braam <braam@clusterfs.com>
6658 * add hard link support
6659 * change obdfile creation method
6660 * kernel patch changed
6662 2002-09-19 Peter Braam <braam@clusterfs.com>
6665 - stack overflow bug in extN fixed
6667 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
6669 * documentation updates
6670 - add man pages for config tools
6671 - update tests/README to describe testing with new config tools
6672 - finish metadata API descriptions
6673 * bug fixes and cleanups
6674 - statfs workaround for 16TB limit
6675 - LOV stripe allocation improved, can stripe on subset of OSTs
6676 - LOV file size/IO offset was wrong for files > 4GB in size
6677 - object EA data was being dropped, caused files to be unreadable
6678 - memory overflow with non-LOV OST caused memory corruption
6679 - fixed regression tests to work with new config tools, obdfilter
6680 - fixed bug when directory size became larger than 1 block
6681 - fixed bug (for single client case) when PWD was deleted
6682 - invalidate local directory pages when doing intent-based ops
6683 - avoid LDLM oops when lock callback contained bad data
6685 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
6687 * documentation updates
6688 * bug fixes and cleanups
6689 - configuration tools
6692 - 64-bit compile warnings
6693 - 64-bit internal statfs data
6695 * test_brw on persistent OST devices
6697 * lprocfs (disabled)
6699 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
6701 * documentation updates
6702 * bug fixes and cleanups
6703 * configuration tools
6705 2002-08-30 Peter J. Braam <braam@clusterfs.com>
6708 * many small fixes to 0_5_4
6709 * io/network handling
6710 * thinkos in MDS operations
6712 2002-08-24 Peter J. Braam <braam@clusterfs.com>
6715 * crucial basic fixes to 0.5.3
6716 * IOR, Iozone work over Elan
6719 2002-08-07 Phil Schwan <phil@clusterfs.com>
6720 * version 0_5_3, our first alpha
6721 * we use the new Portals iovs
6722 * documentation updates
6723 * bug fixes and cleanups
6724 * small changes in the DLM wire protocol
6726 2002-07-25 Peter J. Braam <braam@clusterfs.com>
6727 * version 0_5_1 with some initial stability,
6728 * locking on MD and file I/O.
6729 * documentation updates
6730 * several bug fixes since 0.5.0
6731 * small changes in wire protocol
6733 2002-07-18 Phil Schwan <phil@clusterfs.com>
6735 * delivered as Lustre Light Alpha
6736 * fixed a crash after handling invalid MDS requests
6737 * fixed directory pages for architectures with non-4k pages sizes
6739 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
6740 * release version v0_4_4
6741 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
6743 * Fixes a number of interruption problems with OST operations.
6744 * Update documentation for portals header changes
6745 * Move all wire protocol structs/defines to lustre_idl.h
6746 * Fixes symlink length bug.
6747 * Add tcpdump to repository.
6749 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
6750 * release version v0_4_3
6751 * Fixes statfs for inodes on extN.
6752 * Fixes bug in runtests which would delete /etc/hosts.
6753 * Use 64-bit object IDs wherever possible (not into VFS though)
6754 Remove ost_get_info, which is unused by lustre, and out of date.
6756 2002-07-03 Peter Braam <braam@clusterfs.com>
6757 * release version v0_4_2 Fixes a lookup error (type not passed)
6758 * move forward to head of Portals
6759 * move forward to latest Lustre kernel
6761 2002-06-25 Peter Braam <braam@clusterfs.com>
6762 * release version v0_4_1. Hopefully stable on single node use.