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 3.10.0-957.27.2.el7 (RHEL7.6)
7 3.10.0-1062.18.1.el7 (RHEL7.7)
8 3.10.0-1127.el7 (RHEL7.8)
9 * Server known to build on the following kernels (others may also work):
10 2.6.32-696.18.7.el6 (RHEL6.9)
11 3.10.0-862.14.4.el7 (RHEL7.5)
12 4.4.120-92.70 (SLES12 SP2)
13 4.4.180-94.97 (SLES12 SP3)
14 4.4.0-142 (Ubuntu 16.04)
15 4.15.0-45 (Ubuntu 18.04)
16 vanilla linux 4.6.7 (ZFS only)
17 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
18 * Client primary unpatched kernels built and tested during release cycle:
19 3.10.0-957.27.2.el7 (RHEL7.6)
20 3.10.0-1062.18.1.el7 (RHEL7.7)
21 3.10.0-1127.el7 (RHEL7.8)
22 4.18.0-80.11.2.el8 (RHEL8.0)
23 4.4.180-94.97 (SLES12 SP3)
24 4.12.14-95.45 (SLES12 SP4)
25 4.15.0-45 (Ubuntu 18.04)
26 * Client known to build on unpatched kernels (others may also work):
27 2.6.32-696.18.7.el6 (RHEL6.9)
28 2.6.32-754.6.3.el6 (RHEL6.10)
29 3.10.0-862.14.4.el7 (RHEL7.5)
30 4.14.0-49.13.1.el7a (RHEL7.5)
31 4.14.0-115.2.2.el7a (RHEL7.6)
32 4.18.0-147.3.1.el8 (RHEL8.1)
33 4.4.120-92.70 (SLES12 SP2)
34 4.4.0-142 (Ubuntu 16.04)
35 vanilla linux 4.15.0 (ZFS only)
36 * Recommended e2fsprogs version: 1.45.2.wc1 or newer
37 * Recommended ZFS / SPL version: 0.7.13
38 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
39 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
40 stack overflow. For more information, please refer to bugzilla 17630.
41 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
42 the re-exporting nodes to be mounted with "32bitapi" mount option
43 * Note for Ubuntu 14.04.5:
44 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
45 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
49 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
50 for currently supported client and server kernel versions.
51 * Server primary kernels built and tested during release cycle:
52 3.10.0-957.27.2.el7 (RHEL7.6)
53 3.10.0-1062.9.1.el7 (RHEL7.7)
54 * Server known to build on the following kernels (others may also work):
55 2.6.32-696.18.7.el6 (RHEL6.9)
56 3.10.0-862.14.4.el7 (RHEL7.5)
57 4.4.120-92.70 (SLES12 SP2)
58 4.4.180-94.97 (SLES12 SP3)
59 4.4.0-142 (Ubuntu 16.04)
60 4.15.0-45 (Ubuntu 18.04)
61 vanilla linux 4.6.7 (ZFS only)
62 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
63 * Client primary unpatched kernels built and tested during release cycle:
64 3.10.0-957.27.2.el7 (RHEL7.6)
65 3.10.0-1062.9.1.el7 (RHEL7.7)
66 4.18.0-80.11.2.el8 (RHEL8.0)
67 4.4.180-94.97 (SLES12 SP3)
68 4.12.14-95.29 (SLES12 SP4)
69 4.15.0-45 (Ubuntu 18.04)
70 * Client known to build on unpatched kernels (others may also work):
71 2.6.32-696.18.7.el6 (RHEL6.9)
72 2.6.32-754.6.3.el6 (RHEL6.10)
73 3.10.0-862.14.4.el7 (RHEL7.5)
74 4.14.0-49.13.1.el7a (RHEL7.5)
75 4.14.0-115.2.2.el7a (RHEL7.6)
76 4.18.0-147.3.1.el8 (RHEL8.1)
77 4.4.120-92.70 (SLES12 SP2)
78 4.4.0-142 (Ubuntu 16.04)
79 vanilla linux 4.15.0 (ZFS only)
80 * Recommended e2fsprogs version: 1.45.2.wc1 or newer
81 * Recommended ZFS / SPL version: 0.7.13
82 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
83 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
84 stack overflow. For more information, please refer to bugzilla 17630.
85 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
86 the re-exporting nodes to be mounted with "32bitapi" mount option
87 * Note for Ubuntu 14.04.5:
88 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
89 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
93 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
94 for currently supported client and server kernel versions.
95 * Server primary kernels built and tested during release cycle:
96 3.10.0-957.27.2.el7 (RHEL7.6)
97 3.10.0-1062.1.1.el7 (RHEL7.7)
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 4.4.120-92.70 (SLES12 SP2)
102 4.4.180-94.97 (SLES12 SP3)
103 4.4.0-142 (Ubuntu 16.04)
104 4.15.0-45 (Ubuntu 18.04)
105 vanilla linux 4.6.7 (ZFS only)
106 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
107 * Client primary unpatched kernels built and tested during release cycle:
108 3.10.0-957.27.2.el7 (RHEL7.6)
109 3.10.0-1062.1.1.el7 (RHEL7.7)
110 4.18.0-80.7.1.el8 (RHEL8.0)
111 4.4.180-94.97 (SLES12 SP3)
112 4.12.14-95.29 (SLES12 SP4)
113 4.15.0-45 (Ubuntu 18.04)
114 * Client known to build on unpatched kernels (others may also work):
115 2.6.32-696.18.7.el6 (RHEL6.9)
116 2.6.32-754.6.3.el6 (RHEL6.10)
117 3.10.0-862.14.4.el7 (RHEL7.5)
118 4.14.0-49.13.1.el7a (RHEL7.5)
119 4.14.0-115.2.2.el7a (RHEL7.6)
120 4.4.120-92.70 (SLES12 SP2)
121 4.4.0-142 (Ubuntu 16.04)
122 vanilla linux 4.15.0 (ZFS only)
123 * Recommended e2fsprogs version: 1.45.2.wc1 or newer
124 * Recommended ZFS / SPL version: 0.7.13
125 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
126 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
127 stack overflow. For more information, please refer to bugzilla 17630.
128 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
129 the re-exporting nodes to be mounted with "32bitapi" mount option
130 * Note for Ubuntu 14.04.5:
131 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
132 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
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 the following kernels (others may also work):
139 2.6.32-696.18.7.el6 (RHEL6.9)
140 3.10.0-862.14.4.el7 (RHEL7.5)
141 3.10.0-957.10.1.el7 (RHEL7.6)
142 4.4.120-92.70 (SLES12 SP2)
143 4.4.162-94.69 (SLES12 SP3)
144 4.4.0-142 (Ubuntu 16.04)
145 4.15.0-45 (Ubuntu 18.04)
146 vanilla linux 4.6.7 (ZFS only)
147 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
148 * Client known to build on unpatched kernels (others may also work):
149 2.6.32-696.18.7.el6 (RHEL6.9)
150 2.6.32-754.6.3.el6 (RHEL6.10)
151 3.10.0-862.14.4.el7 (RHEL7.5)
152 3.10.0-957.10.1.el7 (RHEL7.6)
153 4.14.0-49.13.1.el7a (RHEL7.5)
154 4.14.0-115.2.2.el7a (RHEL7.6)
155 4.4.120-92.70 (SLES12 SP2)
156 4.4.162-94.69 (SLES12 SP3)
157 4.12.14-95.13 (SLES12 SP4)
158 4.4.0-142 (Ubuntu 16.04)
159 4.15.0-45 (Ubuntu 18.04)
160 vanilla linux 4.15.0 (ZFS only)
161 * Recommended e2fsprogs version: 1.44.5.wc1 or newer
162 * Recommended ZFS / SPL version: 0.7.13
163 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
164 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
165 stack overflow. For more information, please refer to bugzilla 17630.
166 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
167 the re-exporting nodes to be mounted with "32bitapi" mount option
168 * Note for Ubuntu 14.04.5:
169 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
170 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
175 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
176 for currently supported client and server kernel versions.
177 * Server known to build on the following kernels (others may also work):
178 2.6.32-696.18.7.el6 (RHEL6.9)
179 3.10.0-862.14.4.el7 (RHEL7.5)
180 3.10.0-957.10.1.el7 (RHEL7.6)
181 4.4.120-92.70 (SLES12 SP2)
182 4.4.162-94.69 (SLES12 SP3)
183 4.4.0-142 (Ubuntu 16.04)
184 4.15.0-45 (Ubuntu 18.04)
185 vanilla linux 4.6.7 (ZFS only)
186 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
187 * Client known to build on unpatched kernels (others may also work):
188 2.6.32-696.18.7.el6 (RHEL6.9)
189 2.6.32-754.6.3.el6 (RHEL6.10)
190 3.10.0-862.14.4.el7 (RHEL7.5)
191 3.10.0-957.10.1.el7 (RHEL7.6)
192 4.14.0-49.13.1.el7a (RHEL7.5)
193 4.14.0-115.2.2.el7a (RHEL7.6)
194 4.4.120-92.70 (SLES12 SP2)
195 4.4.162-94.69 (SLES12 SP3)
196 4.12.14-95.13 (SLES12 SP4)
197 4.4.0-142 (Ubuntu 16.04)
198 4.15.0-45 (Ubuntu 18.04)
199 vanilla linux 4.15.0 (ZFS only)
200 * Recommended e2fsprogs version: 1.44.5.wc1 or newer
201 * Recommended ZFS / SPL version: 0.7.13
202 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
203 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
204 stack overflow. For more information, please refer to bugzilla 17630.
205 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
206 the re-exporting nodes to be mounted with "32bitapi" mount option
207 * Note for Ubuntu 14.04.5:
208 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
209 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
213 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
214 for currently supported client and server kernel versions.
215 * Server known to build on the following kernels (others may also work):
216 2.6.32-696.18.7.el6 (RHEL6.9)
217 3.10.0-862.14.4.el7 (RHEL7.5)
218 3.10.0-957.el7 (RHEL7.6)
219 4.4.120-92.70 (SLES12 SP2)
220 4.4.162-94.69 (SLES12 SP3)
221 4.4.0-131 (Ubuntu 16.04)
222 4.15.0-32 (Ubuntu 18.04)
223 vanilla linux 4.6.7 (ZFS only)
224 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
225 * Client known to build on unpatched kernels (others may also work):
226 2.6.32-696.18.7.el6 (RHEL6.9)
227 2.6.32-754.6.3.el6 (RHEL6.10)
228 3.10.0-862.14.4.el7 (RHEL7.5)
229 3.10.0-957.el7 (RHEL7.6)
230 4.14.0-49.13.1.el7a (RHEL7.5)
231 4.14.0-115.2.2.el7a (RHEL7.6)
232 4.4.120-92.70 (SLES12 SP2)
233 4.4.162-94.69 (SLES12 SP3)
234 4.4.0-131 (Ubuntu 16.04)
235 4.15.0-32 (Ubuntu 18.04)
236 vanilla linux 4.15.0 (ZFS only)
237 * Recommended e2fsprogs version: 1.44.3.wc1 or newer
238 * Recommended ZFS / SPL version: 0.7.9
239 * Tested with ZFS / SPL version: 0.7.9
240 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
241 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
242 stack overflow. For more information, please refer to bugzilla 17630.
243 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
244 the re-exporting nodes to be mounted with "32bitapi" mount option
245 * Note for Ubuntu 14.04.5:
246 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
247 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
249 --------------------------------------------------------------------------------
252 2018-04-03 Intel Corporation
254 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
255 for currently supported client and server kernel versions.
256 * Server known to build on patched kernels:
257 2.6.32-431.29.2.el6 (RHEL6.5)
258 2.6.32-504.30.3.el6 (RHEL6.6)
259 2.6.32-573.26.1.el6 (RHEL6.7)
260 2.6.32-642.15.1.el6 (RHEL6.8)
261 2.6.32-696.18.7.el6 (RHEL6.9)
262 3.10.0-693.21.1.el7 (RHEL7.4)
263 3.0.101-0.47.71 (SLES11 SP3)
264 3.0.101-107 (SLES11 SP4)
265 3.12.74-60.64.40 (SLES12 SP1)
266 4.4.114-92.64 (SLES12 SP2)
267 4.4.114-94.11 (SLES12 SP3)
268 3.13.0-101 (Ubuntu 14.04, ZFS only)
269 4.4.0-85.108 (Ubuntu 14.04.5 LTS)
270 4.4.0-116 (Ubuntu 16.04)
271 vanilla linux 4.6.7 (ZFS only)
272 * Client known to build on unpatched kernels:
273 2.6.32-431.29.2.el6 (RHEL6.5)
274 2.6.32-504.30.3.el6 (RHEL6.6)
275 2.6.32-573.26.1.el6 (RHEL6.7)
276 2.6.32-642.15.1.el6 (RHEL6.8)
277 2.6.32-696.18.7.el6 (RHEL6.9)
278 3.10.0-693.21.1.el7 (RHEL7.4)
279 3.0.101-0.47.71 (SLES11 SP3)
280 3.0.101-107 (SLES11 SP4)
281 3.12.74-60.64.40 (SLES12 SP1)
282 4.4.114-92.64 (SLES12 SP2)
283 4.4.114-94.11 (SLES12 SP3)
284 3.13.0-101 (Ubuntu 14.04)
285 4.4.0-85.108 (Ubuntu 14.04.5 LTS)
286 4.4.0-116 (Ubuntu 16.04)
288 * Recommended e2fsprogs version: 1.42.13.wc6 or newer
289 * Recommended ZFS / SPL version: 0.7.6
290 * Tested with ZFS / SPL version: 0.7.6
291 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
292 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
293 stack overflow. For more information, please refer to bugzilla 17630.
294 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
295 the re-exporting nodes to be mounted with "32bitapi" mount option
296 * Note for Ubuntu 14.04.5:
297 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
298 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
300 --------------------------------------------------------------------------------
302 2017-07-05 Intel Corporation
304 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
305 for currently supported client and server kernel versions.
306 * Server known to build on patched kernels:
307 2.6.32-431.29.2.el6 (RHEL6.5)
308 2.6.32-504.30.3.el6 (RHEL6.6)
309 2.6.32-573.26.1.el6 (RHEL6.7)
310 2.6.32-642.15.1.el6 (RHEL6.8)
311 2.6.32-696.3.1.el6 (RHEL6.9)
312 3.10.0-514.21.1.el7 (RHEL7.3)
313 3.0.101-0.47.71 (SLES11 SP3)
314 3.0.101-100 (SLES11 SP4)
315 3.12.74-60.64.40 (SLES12 SP1)
316 4.4.59-92.17 (SLES12 SP2)
317 vanilla linux 4.6.7 (ZFS only)
318 * Client known to build on unpatched kernels:
319 2.6.32-431.29.2.el6 (RHEL6.5)
320 2.6.32-504.30.3.el6 (RHEL6.6)
321 2.6.32-573.26.1.el6 (RHEL6.7)
322 2.6.32-642.15.1.el6 (RHEL6.8)
323 2.6.32-696.3.1.el6 (RHEL6.9)
324 3.10.0-514.21.1.el7 (RHEL7.3)
325 3.0.101-0.47.71 (SLES11 SP3)
326 3.0.101-100 (SLES11 SP4)
327 3.12.74-60.64.40 (SLES12 SP1)
328 4.4.59-92.17 (SLES12 SP2)
330 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
331 * Recommended ZFS / SPL version: 0.6.5.9
332 * Tested with ZFS / SPL version: 0.6.5.9
333 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
334 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
335 stack overflow. For more information, please refer to bugzilla 17630.
336 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
337 the re-exporting nodes to be mounted with "32bitapi" mount option
339 --------------------------------------------------------------------------------
341 12-07-2016 Intel Corporation
343 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
344 for currently supported client and server kernel versions.
345 * Server known to build on patched kernels:
346 2.6.32-431.29.2.el6 (RHEL6.5)
347 2.6.32-504.30.3.el6 (RHEL6.6)
348 2.6.32-573.26.1.el6 (RHEL6.7)
349 2.6.32-642.11.1.el6 (RHEL6.8)
350 3.10.0-514.el7 (RHEL7.3)
351 3.0.101-0.47.71 (SLES11 SP3)
352 3.0.101-80 (SLES11 SP4)
353 3.12.62-60.64.8 (SLES12 SP1)
354 vanilla linux 4.5.7 (ZFS only)
355 * Client known to build on unpatched kernels:
356 2.6.32-431.29.2.el6 (RHEL6.5)
357 2.6.32-504.30.3.el6 (RHEL6.6)
358 2.6.32-573.26.1.el6 (RHEL6.7)
359 2.6.32-642.11.1.el6 (RHEL6.8)
360 3.10.0-514.el7 (RHEL7.3)
361 3.0.101-0.47.71 (SLES11 SP3)
362 3.0.101-80 (SLES11 SP4)
363 3.12.62-60.64.8 (SLES12 SP1)
365 * Recommended e2fsprogs version: 1.42.13.wc5 or newer
366 * Recommended ZFS / SPL version: 0.6.5.8
367 * Tested with ZFS / SPL version: 0.6.5.8
368 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
369 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
370 stack overflow. For more information, please refer to bugzilla 17630.
371 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
372 the re-exporting nodes to be mounted with "32bitapi" mount option
374 --------------------------------------------------------------------------------
376 02-29-2016 Intel Corporation
378 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
379 for currently supported client and server kernel versions.
380 * Server known to build on patched kernels:
381 2.6.32-431.29.2.el6 (RHEL6.5)
382 2.6.32-504.30.3.el6 (RHEL6.6)
383 2.6.32-573.12.1.el6 (RHEL6.7)
384 3.10.0-327.3.1.el7 (RHEL7.2)
385 3.0.101-0.47.71 (SLES11 SP3)
386 3.0.101-68 (SLES11 SP4)
387 vanilla linux 4.2.1 (ZFS only)
388 * Client known to build on unpatched kernels:
389 2.6.32-431.29.2.el6 (RHEL6.5)
390 2.6.32-504.30.3.el6 (RHEL6.6)
391 2.6.32-573.12.1.el6 (RHEL6.7)
392 3.10.0-327.3.1.el7 (RHEL7.2)
393 3.0.101-0.47.71 (SLES11 SP3)
394 3.0.101-68 (SLES11 SP4)
397 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
398 * Recommended ZFS / SPL version: 0.6.4.2
399 * Tested with ZFS / SPL version: 0.6.4.2
400 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
401 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
402 stack overflow. For more information, please refer to bugzilla 17630.
403 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
404 the re-exporting nodes to be mounted with "32bitapi" mount option
406 --------------------------------------------------------------------------------
408 03-10-2015 Intel Corporation
410 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
411 for currently supported client and server kernel versions.
412 * Server known to build on patched kernels:
413 2.6.32-431.29.2.el6 (RHEL6.5)
414 2.6.32-504.8.1.el6 (RHEL6.6)
415 3.0.101-0.46 (SLES11 SP3)
416 * Client known to build on unpatched kernels:
417 2.6.32-431.29.2.el6 (RHEL6.5)
418 2.6.32-504.8.1.el6 (RHEL6.6)
419 3.10.0-123.20.1.el7 (RHEL7)
420 3.0.101-0.46 (SLES11 SP3)
421 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
422 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
423 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
424 stack overflow. For more information, please refer to bugzilla 17630.
425 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
426 the re-exporting nodes to be mounted with "32bitapi" mount option
428 Severity : enhancement
430 Description: control OST-index in IDIF via ROCOMPAT flag.
431 Details : Introduce new flag OBD_ROCOMPAT_IDX_IN_IDIF that is stored in the
432 last_rcvd file. For new formatted OST device, it will be auto set;
433 for the case of upgrading from old OST device, you can enable it
434 via the lproc interface osd-ldiskfs.index_in_idif. With such flag
435 enabled, for new created OST-object, its IDIF-in-LMA will contain
436 the OST-index; for the existing OST-object, the OSD will convert
437 old format IDIF as new format IDIF with OST-index stored in the
438 LMA EA when accessing such OST-object or via OI scrub. Once such
439 flag is enabled, it cannot be reverted back, so the system cannot
440 be downgraded to the orignal incompatible version.
442 --------------------------------------------------------------------------------
444 07-30-2014 Intel Corporation
446 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
447 for currently supported client and server kernel versions.
448 * Server known to build on patched kernels:
449 2.6.32-431.20.3.el6 (RHEL6.5)
450 3.0.101-0.31 (SLES11 SP3)
451 * Client known to build on unpatched kernels:
452 2.6.32-431.20.3.el6 (RHEL6.5)
453 3.0.101-0.31 (SLES11 SP3)
456 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
457 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
458 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
459 stack overflow. For more information, please refer to bugzilla 17630.
460 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
461 the re-exporting nodes to be mounted with "32bitapi" mount option
463 --------------------------------------------------------------------------------
465 10-31-2013 Intel Corporation
467 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
468 * Server builds for kernels:
469 2.6.32-358.18.1.el6 (RHEL6.4)
470 * Client builds for unpatched kernels:
471 2.6.32-358.18.1.el6 (RHEL6.4)
472 3.0.92-0.8 (SLES11 SP3)
474 * Recommended e2fsprogs version: 1.42.7.wc2 or newer
475 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
476 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
477 stack overflow. For more information, please refer to bugzilla 17630.
478 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
479 the re-exporting nodes to be mounted with "32bitapi" mount option
481 --------------------------------------------------------------------------------
483 05-30-2013 Intel Corporation
485 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
486 * Server builds for kernels:
487 2.6.32-358.6.2.el6 (RHEL6.4)
488 * Client builds for unpatched kernels:
489 2.6.32-358.6.2.el6 (RHEL6.4)
490 3.0.74-0.6.6 (SLES11 SP2)
492 * Recommended e2fsprogs version: 1.42.7.wc1 or newer
493 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
494 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
495 stack overflow. For more information, please refer to bugzilla 17630.
496 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
497 the re-exporting nodes to be mounted with "32bitapi" mount option
499 --------------------------------------------------------------------------------
501 9-30-2012 Whamcloud, Inc.
503 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
504 * Server support for kernels:
505 2.6.32-279.5.1.el6 (RHEL6)
506 * Client support for unpatched kernels:
507 2.6.32-279.5.1.el6 (RHEL6)
509 2.6.18-238.19.1.el5 (RHEL5)
510 2.6.16.60-0.69.1 (SLES10)
511 * Recommended e2fsprogs version: 1.41.90.wc4
512 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
513 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
514 stack overflow. For more information, please refer to bugzilla 17630.
515 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
516 the re-exporting nodes to be mounted with "32bitapi" mount option
518 Severity : enhancement
520 Description: Update RHEL6.3 kernel to 2.6.32-279.5.1.el6.
522 Severity : enhancement
524 Description: Update RHEL6.3 kernel to 2.6.32-279.2.1.el6.
526 Severity : enhancement
528 Description: Update RHEL6.2 kernel to 2.6.32-220.17.1.el6
530 Severity : enhancement
532 Description: Update RHEL6.2 kernel to 2.6.32-220.13.1.el6
534 Severity : enhancement
536 Description: Update RHEL6.2 kernel to 2.6.32-220.7.1.el6
538 --------------------------------------------------------------------------------
540 04-30-2012 Whamcloud, Inc.
542 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
543 * Server support for kernels:
544 2.6.32-220.4.2.el6 (RHEL6)
545 * Client support for unpatched kernels:
546 2.6.32-220.4.2.el6 (RHEL6)
548 2.6.18-238.19.1.el5 (RHEL5)
549 2.6.16.60-0.69.1 (SLES10)
550 * Recommended e2fsprogs version: 1.41.90.wc4
551 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
552 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
553 stack overflow. For more information, please refer to bugzilla 17630.
554 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
555 the re-exporting nodes to be mounted with "32bitapi" mount option
557 Severity : enhancement
559 Description: Update RHEL6.2 kernel to 2.6.32-220.4.2.el6
561 Severity : enhancement
563 Description: Update RHEL6.2 kernel to 2.6.32-220.4.1.el6
565 Severity : enhancement
567 Description: Update RHEL6.2 kernel to 2.6.32-220.el6
569 Severity : enhancement
571 Description: Update RHEL6.1 kernel to 2.6.32-131.17.1.el6
573 --------------------------------------------------------------------------------
575 09-30-2011 Whamcloud, Inc.
577 * Server support for kernels:
578 2.6.18-238.19.1.el5 (RHEL5)
579 2.6.32-131.6.1.el6 (RHEL6)
580 * Client support for unpatched kernels:
581 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
582 2.6.18-238.19.1.el5 (RHEL5)
583 2.6.32-131.6.1.el6 (RHEL6)
584 2.6.29.4-167.fc11 (FC11)
585 2.6.16.60-0.69.1 (SLES10)
587 * Recommended e2fsprogs version: 1.41.90.wc3
588 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
589 Lustre filesystem with 4K stack may cause a stack overflow. For more
590 information, please refer to bugzilla 17630.
591 * NFSv4 reexport requires Lustre client on the re-exporting nodes to be
592 mounted with "32bitapi" mount option
594 Severity : enhancement
595 Description: Update RHEL6.1 kernel to 2.6.32-131.6.1.el6
597 Severity : enhancement
598 Description: Update RHEL5.6 kernel to 2.6.18-238.19.1.el5
600 Severity : enhancement
601 Description: Update RHEL5.6 kernel to 2.6.18-238.9.1.el5
604 Frequency : only with SLES10
606 Description: Use OFED "KMP" provided by Novell
607 Details : SLES10 SP3 ships with OFED in a separate "KMP" package.
608 Lustre is now built against this package. That means you need to
609 install the ofed-kmp package from Novell for the patchless client.
610 Note that the ofed-kmp that Novell ships may not exactly match the
611 kernel version but should still be compatible.
613 Severity : enhancement
615 Description: Update RHEL5.5 kernel to 2.6.18-194.17.1.el5
616 Update OEL5.5 kernel to 2.6.18-194.17.1.0.1.el5.
618 --------------------------------------------------------------------------------
620 2010-07-15 Oracle, Inc.
622 * Server support for kernels:
623 2.6.18-164.11.1.el5 (RHEL 5)
624 2.6.18-164.11.1.0.1.el5 (OEL 5)
625 * Client support for unpatched kernels:
626 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
627 2.6.18-164.11.1.el5 (RHEL 5),
628 2.6.18-164.11.1.0.1.el5 (OEL 5)
629 2.6.16.60-0.42.8 (SLES 10),
631 2.6.29.4-167.fc11 (FC11)
632 * Recommended e2fsprogs version: 1.41.10-sun2
634 Severity : enhancement
636 Description: Distribute lfs_migrate script with lustre
637 Details : The lfs_migrate mentioned in the manual is now included in the
638 lustre tree in lustre/scripts and will distributed in lustre RPMs.
640 Severity : enhancement
642 Description: Failover nid list added to import proc file.
644 Severity : enhancement
646 Description: "som_preview" mount option required to use SOM
647 Details : The clients will need to use som_preview mount option
651 Description: Lustre does not do 1MB IOs to HW RAID
652 Details : Bump MAX_PHYS/HW_SEGMENTS and SG_ALL to 256 in the RHEL5 kernel.
654 Severity : enhancement
656 Description: lfs quota output is not very convenient for awk/sed-parsing
657 Details : Some positions in lfs quota output table could be empty or
658 non-empty which made it hard to parse it with scripts, now
659 a dash is put instead of space where there is not supposed
664 Description: Update to OFED 1.5.1
665 Details : OFED 1.5.1 will be used in environments that don't use
666 vendor supplied OFED.
670 Description: Remove lnet. prefix from lctl params display
671 Details : Also renamed lnet.memused to lnet_memused to
676 Description: Use conf_param -d to remove permanant settings.
680 Description: Add list_param -R
681 Details : lctl list_param now supports recursive mode.
684 Bugzilla : 21829, 22312
685 Description: Fix llobdstat.
686 Details : 1. Now updating the write_bytes counter
687 2. Port script fixes from 1.8.
691 Description: decrease the usage of memory on clients.
692 Details : 1. On clients, recycle dentries and inodes unused.
693 2. Delete the code related to ll_deathrow(att 6215 in bug 1443). It
698 Description: llobdstat fix and enhancement
699 Details : add a counter to set a limit to how many samples will be returned
700 fix a wildcard in the path to limit to obdfilter stats only
704 Description: lfs check only allowed for root
705 Details : Code cleanups and added a regresion test of non-root use of
710 Description: MDS Oopses in ptlrpc_dispatch_difficult_reply()/__list_add()
711 Details : Fix hr->hr_index overload.
715 Description: Fix FIEMAP functionality.
719 Description: FC11 patchless client support added
723 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and
724 OEL5.4 kernel to 2.6.18-164.11.1.0.1.el5.
728 Description: New acc-sm script and various test-framework enhancments.
732 Description: Previously landed patch was reverted.
736 Description: Update SLES10 kernel to 2.6.16.60-0.42.8 (patchless)
740 Description: Add ldiskfs maxdirsize mount option.
744 Description: Fix mkfs.lustre support for 16TB LUNs.
748 Description: Print mmp_check_interval
749 Details : print mmp_check_interval and make it possible to abort
750 mount operation in case it takes too long.
754 Description: Add cascading_rw to lustre mpi tests.
758 Description: Store SOM data in EA.
759 Details : Also several SOM related cleanups.
763 Description: Add refcount to md_open_data
767 Description: Use refcount on lov_request_set to prevent memory corruption.
771 Description: Statahead: use parent dir's i_mutex for synchronization.
775 Description: loadgen improvements.
779 Description: Enable peer health detection in ksocklnd and ko2iblnd.
783 Description: Control DCACHE_LUSTRE_INVALID flag with MDS_INODE_LOOKCUP lock
787 Description: Fix race on obd_devcie:md_stats
791 Description: Fix SLES 11 client builds.
792 Details : Resolved uninitalized pointer gcc warnings.
796 Description: Allow non-root access for "lfs check".
797 Details : Added a check in obd_class_ioctl() for OBD_IOC_PING_TARGET.
801 Description: Clear imp_force_reconnect correctly in ptlrpc_connect_interpret()
803 Severity : enhancement
805 Description: Update kernel to RHEL5.4 2.6.18-164.6.1.el5 and
806 OEL5 2.6.18-164.6.1.0.1.el5(Both in-kernel OFED enabled).
810 Description: Sanity test_27p fails.
811 Details : OBD_FAIL_OST_ENOSPC has been moved to filter_statfs().
812 Sanity tests 27 were cleaned up and improved.
816 Description: Unify req format on client/servers
817 Details : Use new req_capsule API [almost] everywhere instead of old PTLRPC
818 buffers and swabbers approach..
823 Description: drop join file stuff.
825 Severity : enhancement
827 Description: Add 2.6.27 and SLES11 (patchless client) support
831 Description: Fix in ptlrpc_expire_one_request() to print the signed time
834 Severity : enhancement
836 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
839 Severity : enhancement
841 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
843 Severity : enhancement
845 Description: correctly handle big reply message.
846 Details : send LNet event if reply is bigger than buffer and adjust this buffer
851 Description: Drop unnecessary __GFP_NOMEMALLOC flag from filter_get_page()
853 Severity : enhancement
855 Description: Add support for OEL5.
859 Description: Conf-sanity.sh 50g test - deactivated OST should not cause a panic.
861 Severity : enhancement
863 Description: "weak-modules" support
864 Details : Implement "weak-modules" support which enables kernel modules
865 to be used with any kernel that implements the same kABI. In
866 order to achieve this modules are now installed in
867 /lib/modules/$(uname -r)/updates/kernel on all distributions.
871 Description: client could not reconnect to OST because of an active request.
872 Details : The client now retries to reconnect to the same server, if a
873 connect request failed with EBUSY or -EAGAIN. Test case 35b
874 conf-sanity.sh has been created.
878 Description: Don't allow make backward step on assignin osc next id.
879 Details : race between allocation next id and ll_sync thread can be cause
880 of set wrong osc next id and can be kill valid ost objects.
884 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
888 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
892 Description: Error handling in osc_statfs_interpret() has been improved.
893 Details : Check in osc_statfs_interpret() for EBADR.
897 Description: Do not send statfs() requests to OSTs disabled by administrator.
898 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
900 Severity : enhancement
902 Description: Added sanity-quota test to for lfs(1) to work with numeric uid/gid.
904 Severity : enhancement
906 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
908 Severity : enhancement
910 Description: Enhancement for lfs(1) command to use numeric uid/gid.
914 Description: Improved cleanup procedure for conf-sanity test_46a.
918 Description: Do not start bulk transfer, if client tried to reconnect.
919 Details : Do not start bulk transfer, if client tried to reconnect,
920 but failed due to active rpcs. Check for exp_abort_active_req
921 has been implemented.
925 Description: Prevent inconsistences between linux and lustre mount structures.
926 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
927 Make the sleep interruptible.
929 Severity : enhancement
931 Description: provide server to client comms path
932 Details : server to client communications path via new LDLM_SET_INFO rpc
936 Description: truncate starts GFP_FS allocation under transaction causing deadlock
937 Details : ldiskfs_truncate calls grab_cache_page which may start page
938 allocation under an open transaction. This may lead to
939 calling prune_icache with consequent lustre reentrance.
943 Description: send correctly lsm on open replay
944 Details : MDS is trust to LSM size on replay open, but client can set wrong size
949 Description: enable client interface failover
950 Details : When a child reconnects from another NID, properly update export
951 nid hash position and ldlm reverse import.
953 Severity : enhancement
955 Description: Communicate OST degraded/readonly state via statfs to MDS
956 Details : Flags in the statfs returned from OSTs indicate whether the
957 OST is in a degraded RAID state, or if the filesystem has
958 turned read-only after a filesystem error is detected.
963 Description: don't panic if EPROTO was hit when reading symlink
964 Details : correctly handling request reference in error cases.
966 Severity : enhancement
968 Description: Add LustreNetLink, a kernel-userspace communcation path.
970 Severity : enhancement
972 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
977 Description: access to llog context before init.
978 Details : move handling CATALOGS file at osc layer and forbid access to llog
982 Frequency : rare, only when client is writing to OST on local node
984 Description: Avoid deadlock for local client writes
985 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
986 memory freeing context. This allows OST threads to set the
987 PF_MEMALLOC flag on task structures in order to allocate memory
988 from reserved pools and complete IO.
989 Use GFP_HIGHUSER for OST allocations for non-local client writes,
990 so that the OST threads generate memory pressure and allow
991 inactive pages to be reclaimed.
993 Severity : enhancement
995 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
1000 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
1001 Details : move ldlm namespace creation in setup phase to avoid grab
1002 _lprocfs_lock with cli_sem held.
1006 Description: Temporarily disable grant shrink.
1007 Details : Disable the feature for debugging.
1010 Frequency : only when formatting test filesystems
1012 Description: Unable to run several mkfs.lustre on loop devices at the same time
1013 Details : mkfs.lustre returns error 256 on the concurrent loop devices
1014 formatting. The solution is to proper handle the error.
1016 Severity : enhancement
1018 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
1020 Severity : enhancement
1022 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
1027 Description: resolve race between obd_disconnect and class_disconnect_exports
1028 Details : if obd_disconnect will be called to already disconnected export he
1029 forget release one reference and osc module can't unloaded.
1031 Severity : enhancement
1033 Description: Allow tuning service thread via /proc
1034 Details : For each service a new
1035 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
1036 created that can be used to set min/max thread counts, and get the
1037 current number of running threads.
1042 Descriptoin: don't return error if have partially created objects for file.
1043 Details : lov_update_create_set uses set->set_success as index for created
1044 objects, so if some requests will be failed, they will have hole
1045 at end of array and qos_shrink_lsm for allocate correct lsm.
1047 Severity : enhancement
1049 Description: Update OFED support to 1.4.1
1054 Description: Reduce small size read RPC
1055 Details : Set read-ahead limit for every file and only do read-ahead when
1056 available read-ahead pages are bigger than 1M to avoid small size
1059 Severity : enhancement
1061 Description: move AT tunable parameters for more consistent usage
1062 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
1064 Severity : enhancement
1066 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
1067 Details : allow skip disconnected ost for send statfs request and hide error
1073 Description: fix racy locking of mballoc block bitmaps causing BUG
1074 Details : The locking of the mballoc buddy bitmap and the in-memory
1075 block bitmap was using two different spin locks in some
1076 cases. This made it possible to incorrectly access the
1077 mballoc bitmap while another process was modifying it,
1078 causing a sanity assertion to fail. While no on-disk corruption
1079 was reported, there was some risk of this happening.
1082 Frequency : rare, on llog test 6
1084 Descriptoin: don't allow connect to already connected import
1085 Details : allowing connect to already connected import is hide connecting problem.
1090 Descriptoin: allow kill process which wait statahead result
1091 Details : for some reasons 'ls' can stick in waiting result from statahead,
1092 in this case need way for kill this process.
1094 Severity : enhancement
1096 Description: Add state history info file, enhance import info file
1097 Details : Track import connection state changes in a new osc/mdc proc file;
1098 add overview-type data to the osc/mdc import proc file.
1100 Severity : enhancement
1102 Description: MDS create should not wait for statfs RPC while holding DLM lock.
1104 Severity : enhancement
1106 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
1111 Descriptoin: OST grant too much space to client even there are not enough space.
1112 Details : Client will shrink its grant cache to OST if there are no write
1113 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
1114 this grant cache if there are already not enough avaible space
1115 (left_space < total_clients * 32M).
1118 Frequency : start MDS on uncleanly shutdowned MDS device
1120 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
1121 Details : stay in waiting mds<>ost recovery finished produce random bugs
1122 due race between two ll_sync thread for one lov target. send
1123 ACTIVATE event only if connect realy finished and import have
1127 Frequency : rare, connect and disconnect target at same time
1129 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
1130 Details : don't call obd_disconnect under lov_lock. this long time
1131 operation and can block ptlrpcd which answer to connect request.
1136 Descriptoin: don't lose wakeup for imp_recovery_waitq
1137 Details : recover_import_no_retry or invalidate_import and import_close can
1138 both sleep on imp_recovery_waitq, but we was send only one wakeup
1142 Frequency : always with long access acl
1144 Descriptoin: mds can't pack reply with long acl.
1145 Details : mds don't control size of acl but they limited by reint/getattr
1148 Severity : enhancement
1150 Description: Update to SLES10 kernel-2.6.16.60-0.33.
1152 Severity : enhancement
1154 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
1157 Frequency : start MDS on uncleanly shutdowned MDS device
1159 Descriptoin: aborting recovery hang on MDS
1160 Details : don't throttle destroy RPCs for the MDT.
1163 Frequency : on remount
1165 Description: external journal device not working after the remount
1166 Details : clear dev_rdonly flag for external journal devices in
1172 Description: shutdown vs evict race
1173 Details : client_disconnect_export vs connect request race.
1174 if client will evicted at this time - we start invalidate
1175 thread without referece to import and import can be freed
1179 Frequency : rare, need acl's on inode.
1181 Description: client can't handle ost additional correctly
1182 Details : if ost was added after client connected to mds client can have
1183 hit lnet_try_match_md ... to big messages to wide striped files.
1184 in this case need teach client to handle config events about add
1185 lov target and update client max ea size at that event.
1187 Severity : enhancement
1189 Description: Changelogs
1190 Details : Changelogs are a lightweight mechanism to track filesystem
1191 metadata and namespace changes. The changelog is recorded
1192 permanently on the MDTs, and is periodically "consumed" / purged
1193 when records are no longer needed.
1195 Severity : enhancement
1197 Description: compact fld format with extents
1198 Details : Store range of seq rather than every seq in FLD. Seq
1199 controller update FLD rather than clients. In Case of CMD, mdt0
1200 has FLD, all other metadata server act as non persistent proxy
1201 for FLD queries and cache fld entries in fld cache.
1206 Description: don't skip ost target if they assigned to file
1207 Details : Drop slow OSCs if we can, but not for requested start idx.
1208 This means "if OSC is slow and it is not the requested
1209 start OST, then it can be skipped, otherwise skip it only
1210 if it is inactive/recovering/out-of-space.
1214 Description: more cleanup in mds_lov
1215 Details : not send LOV EA under replay, we can't know about they size at this
1216 time. Don't allow client connect to mds before any ost connected,
1217 for avoid problems with LOV EA size and returning EIO to client.
1219 Severity : enhancement
1221 Description: Interoperability at server side (Disk interoperability)
1223 Severity : enhancement
1225 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
1227 Severity : enhancement
1229 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
1231 Severity : enhancement
1233 Description: New client IO stack (CLIO).
1235 Severity : enhancement
1237 Description: Commit on sharing. Eliminate inter-client dependencies between
1238 uncommitted transactions by doing transaction commits.
1239 Thereby clients may recovery independently.
1242 Frequency : Create a symlink file with a very long name
1244 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
1245 Details : If there is no extra space in the request for early cancels,
1246 ldlm_req_handles_avail() returns 0 instead of a negative value.
1248 Severity : enhancement
1250 Description: Add /proc entry for import status
1251 Details : The mdc, osc, and mgc import directories now have
1252 an import directory that contains useful import data for debugging
1253 connection problems.
1255 Severity : enhancement
1257 Description: Re-disable certain /proc logging
1258 Details : Enable and disable client's offset_stats, extents_stats and
1259 extents_stats_per_process stats logging on the fly.
1262 Frequency : Only on FC kernels 2.6.22+
1264 Description: oops in statahead
1265 Details : Do not drop reference count for the dentry from VFS when lookup,
1266 VFS will do that by itself.
1268 Severity : enhancement
1270 Description: Generic /proc file permissions
1271 Details : Set /Proc file permissions in a more generic way to enable non-
1272 root users operate on some /proc files.
1276 Description: Hitting mdc_commit_close() ASSERTION
1277 Details : Properly handle request reference release in
1278 ll_release_openhandle().
1282 Description: quota recovery deadlock during mds failover
1283 Details : This patch includes att18982, att18236, att18237 in bz14840.
1285 1. fix osts hang when mds does failover with quotaon
1286 2. prevent watchdog storm when osts threads wait for the
1291 Frequency : only patchless client
1292 Description: add workaround for race between add/remove dentry from hash
1294 Severity : enhancement
1296 Description: Allow OST glimpses to return PW locks
1300 Description: LBUG when llog conf file is full
1301 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
1305 Description: Prevent import from entering FULL state when server in recovery
1309 Description: service mount cannot take device name with ":"
1310 Details : Only when device name contains ":/" will mount treat it as
1316 Description: replace ptlrpcd with the statahead thread to interpret the async
1317 statahead RPC callback
1321 Frequency : on recovery
1322 Description: I/O failures after umount during fail back
1323 Details : if client reconnected to restarted server we need join to recovery
1324 instead of find server handler is changed and process self
1325 eviction with cancel all locks.
1327 Severity : enhancement
1329 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
1331 Severity : enhancement
1333 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
1335 Severity : enhancement
1337 Description: Upcall on Lustre log has been dumped
1338 Details : Allow for a user mode script to be called once a Lustre log has
1339 been dumped. It passes the filename of the dumped log to the
1340 script, the location of the script can be specified via
1341 /proc/sys/lnet/debug_log_upcall.
1346 Description: avoid idr_remove called for id which is not allocated.
1347 Details : Move assigment s_dev for clustered nfs to end of initialization,
1348 to avoid problem with error handling.
1353 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
1354 Details : When connection is reused this not moved from CONN_UNUSED_HASH
1355 into CONN_USED_HASH and this prodice warning when put connection
1356 again in unused hash.
1359 Severity : enhancement
1361 Description: File striping can now be set to use an arbitrary pool of OSTs.
1363 Severity : enhancement
1365 Description: Export bytes_read/bytes_write count on OSC/OST.
1369 Description: Early reply size mismatch, MGC loses connection
1370 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
1371 the connect flags are properly negotiated.
1376 Description: quotas are not honored with O_DIRECT
1377 Details : all writes with the flag O_DIRECT will use grants which leads to
1378 this problem. Now using OBD_BRW_SYNC to guard this.
1382 Description: add quota statistics
1383 Details : 1. sort out quota proc entries and proc code.
1384 2. add quota statistics
1386 Severity : enhancement
1388 Description: enable quota support for HEAD.
1392 Description: Properly propagate oinfo flags from lov to osc for statfs
1393 Details : restore missing copy oi_flags to lov requests.
1395 Severity : enhancement
1397 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
1398 : l_getgroups(8), lst(8), routerstat(8)
1399 Details : included man pages for llobdstat(8), llstat(8),
1400 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
1402 Severity : enhancement
1404 Description: configure's --enable-quota should check the
1405 : kernel .config for CONFIG_QUOTA
1406 Details : configure is terminated if --enable-quota is passed but
1407 : no quota support is in kernel
1411 Description: Remove portals compatibility
1412 Details : Remove portals compatibility, not interoperable with releases
1417 Description: Resolve device initialization race
1418 Details : Prevent proc handler from accessing devices added to the
1419 obd_devs array but yet be intialized.
1421 Severity : enhancement
1423 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
1425 Severity : enhancement
1427 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
1432 Description: Using wrong pointer in osc_brw_prep_request
1433 Details : Access to array[-1] can produce panic if kernel compiled with
1434 CONFIG_PAGE_ALLOC enabled
1438 Description: Client runs out of low memory
1439 Details : Consider only lowmem when counting initial number of llap pages
1443 Description: *optional* service tags registration
1444 Details : if the "service tags" package is installed on a Lustre node
1445 When the filesystem is mounted, a local-node service tag will
1446 be created. See http://inventory.sun.com/ for more information
1447 about the Service Tags asset management system.
1451 Description: Kernel BUG tries to release flock
1452 Details : Lustre does not destroy flock lock before last reference goes
1453 away. So always drop flock locks when client is evicted and
1454 perform unlock regardless of successfulness of speaking to MDS.
1458 Description: add refcount protection for osc callbacks, avoid panic on shutdown
1462 Description: sanity test 65a fails if stripecount of -1 is set
1463 Details : handle -1 striping on filesystem in ll_dirstripe_verify
1468 Description: ASSERTION(CheckWriteback(page,cmd)) failed
1469 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
1472 Severity : enhancement
1474 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
1478 Description: do not process already freed flock
1479 Details : flock can possibly be freed by another thread before it reaches
1480 to ldlm_flock_completion_ast.
1484 Description: LBUG during stress test
1485 Details : Need properly lock accesses the flock deadlock detection list.
1489 Description: oops in page fault handler
1490 Details : kernel page fault handler can return two special 'pages' in error
1491 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
1495 Description: timeout with invalidate import.
1496 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
1497 should be handled by ptlrpcd. This produce long age waiting and
1498 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
1500 Severity : enhancement
1502 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
1506 Description: enable MGS and MDT services start separately
1507 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
1508 a co-located MGS without starting the MGS, which is a complement
1509 to 'nosvc' mount option.
1513 Frequency : after recovery
1514 Description: precreate to many object's after del orphan.
1515 Details : del orphan st in oscc last_id == next_id and this triger growing
1516 count of precreated objects. Set flag LOW to skip increase count
1517 of precreated objects.
1521 Frequency : rare, on clear nid stats
1522 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
1523 Details : when clean nid stats sometimes try destroy live entry,
1524 and this produce panic in free.
1528 Description: Stack overflow during MDS log replay
1529 ease stack pressure by using a thread dealing llog_process.
1533 Description: wait until IO finished before start new when do lock cancel.
1534 Details : VM protocol want old IO finished before start new, in this case
1535 need wait until PG_writeback is cleared until check dirty flag and
1536 call writepages in lock cancel callback.
1538 Severity : enhancement
1540 Description: using special macro for print time and cleanup in includes.
1544 Description: mds_mfd_close() ASSERTION(rc == 0)
1545 Details : In mds_mfd_close(), we need protect inode's writecount change
1546 within its orphan write semaphore to prevent possible races.
1550 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
1554 Frequency : rare, on shutdown ost
1555 Description: don't hit live lock with umount ost.
1556 Details : shrink_dcache_parent can be in long loop with destroy dentries,
1557 use shrink_dcache_sb instead.
1561 Description: don't panic with use echo client
1562 Details : echo client pass NULL as client nid pointer and this produce null
1563 pointer dereference.
1567 Description: fix build on ppc32
1568 Details : compile code with -m64 flag produce wrong object file for ppc32.
1572 Description: add message levels for liblustreapi
1576 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
1577 Details : Move the 'good_osts' check before the 'total_bavail' check. This
1578 will result in an -EAGAIN and in the exit call path we call
1579 alloc_rr() which will with increasing aggressiveness attempt to
1580 aquire precreated objects on the minimum number of required OSCs.
1584 Description: Use old size assignment to avoid deadlock
1585 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
1586 the scheduling while holding a spinlock. We do not need locking
1587 for size in ll_update_inode() because size is only updated from
1588 the MDS for directories or files without objects, so there is no
1589 other place to do the update, and concurrent access to such inodes
1590 are protected by the inode lock.
1594 Description: resolve "_IOWR redefined" build error on SLES10
1598 Description: dump the memory debugging after all modules are unloaded to
1599 suppress false negative in conf_sanity test 39
1601 Severity : enhancement
1603 Description: build kernel-ib packages for OFED 1.3 in our release cycle
1608 Description: fix SLES kernel versioning
1609 Details : the kernel version for our SLES 10 kernel did not include a "-"
1610 before the "smp" at the end. while this was not a problem in
1611 general, it did mean that software trying to use the kernel
1612 version to try to detect a vendor specific kernel would fail.
1613 this was most evident by the OFED build scripts.
1617 Description: Don't update lov_desc members until making sure they are valid
1618 Details : When updating lov_desc members via proc fs, need fix their
1619 validities before doing the real update.
1623 Description: don't put request into delay list while invalidate in flight.
1624 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
1625 import in flight. this produce timeout for invalidate and sometimes
1626 can cause stale data.
1630 Frequency : on ppc only
1631 Description: not convert ost objects for directory because it's not exist.
1632 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
1636 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
1638 Severity : enhancement
1640 Description: Update to support 2.6.22.14 vanilla kernel.
1644 Frequency : rare, on recovery
1645 Description: read procfs can produce deadlock in some situation
1646 Details : Holding lprocfs lock which send rpc can produce block for destroy
1647 obd objects and this also block reconnect with -EALREADY. This
1648 isn't fix all lprocfs bugs - but make it rare.
1650 Severity : enhancement
1652 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
1655 Frequency : frequent on X2 node
1657 Description: mdc_set_open_replay_data LBUG
1658 Details : Set replay data for requests that are eligible for replay.
1662 Description: lustre_mgs: operation 101 on unconnected MGS
1663 Details : When MGC is disconnected from MGS long enough, MGS will evict the
1664 MGC, and late on MGC cannot successfully connect to MGS and a lot
1665 of the error messages complaining that MGS is not connected.
1669 Frequency : on network error
1670 Description: panic with double free request if network error
1671 Details : mdc_finish_enqueue is finish request if any network error ocuring,
1672 but it's true only for synchronus enqueue, for async enqueue
1673 (via ptlrpcd) this incorrect and ptlrpcd want finish request
1676 Severity : enhancement
1678 Description: client-side metadata stat-ahead during readdir(directory readahead)
1679 Details : perform client-side metadata stat-ahead when the client detects
1680 readdir and sequential stat of dir entries therein
1683 Frequency : on start mds
1685 Description: Implement get_info(last_id) in obdfilter.
1688 Frequency : occasional
1690 Description: Correctly check stale fid, not start epoch if ost not support SOM
1691 Details : open with flag O_CREATE need set old fid in op_fid3 because
1692 op_fid2 was overwritten with new generated fid, but mds can answer
1693 with one of these two fids and both is not stale. Setattr
1694 incorrectly started an epoch and assume will be called
1695 done_writing, but without SOM done_writing ever being called.
1698 Frequency : rare, depends on device drivers and load
1700 Description: MDS or OSS nodes crash due to stack overflow
1701 Details : Code changes in 1.8.0 increased the stack usage of some functions.
1702 In some cases, in conjunction with device drivers that use a lot
1703 of stack the MDS (or possibly OSS) service threads could overflow
1704 the stack. One change which was identified to consume additional
1705 stack has been reworked to avoid the extra stack usage.
1708 Frequency : occasional
1710 Description: Do not fail import if osc_interpret_create gets -EAGAIN
1711 Details : If osc_interpret_create got -EAGAIN it immediately exits and
1712 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
1713 oscc_has_objects and see OSC has no objests and call
1714 oscc_internal_create to resend create request.
1716 Severity : enhancement
1718 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
1720 Severity : enhancement
1722 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
1725 Frequency : very rare
1727 Description: Fix replay if there is an un-replied request and open
1728 Details : In some cases, older replay request will revert the
1729 mcd->mcd_last_xid on MDS which is used to record the client's
1730 latest sent request.
1732 Severity : enhancement
1734 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
1736 Severity : enhancement
1738 Description: Add rhel5 support to HEAD.
1740 Serverity : enhancement
1742 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
1747 Description: Don't allow skipping OSTs if index has been specified.
1748 Details : Don't allow skipping OSTs if index has been specified, make
1749 locking in internal create lots better.
1753 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
1754 Details : access to bitfield in structure is always rounded to long
1755 and this produce problem with not atomic change any bit.
1759 Description: Lustre make rpms failed.
1760 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
1761 builds Lustre rpms from tarball.
1764 Frequency : rare on shutdown ost
1766 Description: If llog cancel was not send before clean_exports phase, this can
1767 produce deadlock in llog code.
1768 Details : If llog thread has last reference to obd and call class_import_put
1769 this produce deadlock because llog_cleanup_commit_master wait when
1770 last llog_commit_thread exited, but this never success because was
1771 called from llog_commit_thread.
1775 Description: allow userland application know is lost one of stripes.
1776 Details : fill lvb_blocks with error code on ost and return it to
1777 application if error flag found.
1781 Description: NULL lov_tgts causing MDS oops
1782 Details : more safe checks for NULL lov_tgts for avoid oops.
1784 Severity : enhancement
1786 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
1790 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
1791 Details : If an OST is down the MDS will hang indefinitely in
1792 obd_statfs_rqset() waiting for the statfs data. While for
1793 MDS QOS usage of statfs, it should not stuck in waiting.
1795 Severity : enhancement
1797 Description: remote_acl support
1798 Details : Support ACL-based permission check for remote user.
1799 Support setfacl/getfacl for remote user with the utils
1800 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
1801 the system "{s,g}etfacl" utils.
1803 Severity : enhancement
1805 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
1807 Severity : enhancement
1809 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
1813 Description: make mgs_setparam() handle fsname containing dash
1814 Details : fsname containing a dash does not work with lctl conf_param
1816 Severity : enhancement
1818 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
1820 Severity : enhancement
1822 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
1826 Frequency : rare, at shutdown
1827 Description: access already free / zero obd_namespace.
1828 Details : if client_disconnect_export was called without force flag set,
1829 and exist connect request in flight, this can produce access to
1830 NULL pointer (or already free pointer) when connect_interpret
1831 store ocd flags in obd_namespace.
1835 Frequency : only at startup
1836 Description: not alloc memory with spinlock held.
1837 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
1838 if any spinlock held.
1840 Severity : enhancement
1842 Description: make lustre randomly fail allocating memory
1843 Details : Make lustre randomly failed allocating memory for testing purpose.
1845 Severity : enhancement
1847 Description: lost problems with lov objid file
1848 Details : Fixes some scability and access to not inited memory problems
1849 in work with lov objdid file.
1854 Description: lfs find does not continue on file error
1855 Details : Continue other files processing when a file/dir is absent.
1859 Description: Inconsistent usage of lustre_pack_reply()
1860 Details : Standardize the usage of lustre_pack_reply() such that it
1861 always generate a CERROR on failure.
1864 Frequency : occasional
1866 Description: MDS hang or stay in waiting lock
1867 Details : If client receive lock with CBPENDING flag ldlm need send lock
1868 cancel as separate rpc, to avoid situation when cancel request
1869 can't processed due all i/o threads stay in wait lock.
1873 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1877 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1881 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1885 Description: Update to RHEL4 latest kernel.
1887 Severity : enhancement
1889 Description: Build SLES10 patchless client fails
1890 Details : The configure was broken by run ./configure with
1891 --with-linux-obj=.... argument for patchless client. When the
1892 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1893 can't search header adequately. Use absolute path such as
1894 -I($LINUX)/include instead.
1898 Description: interrupt oig_wait produce painc on resend.
1899 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1900 private set, and this produce situation when rq_ptlrpcd_data not
1901 copyed to new allocated request and triggered LBUG on assert
1902 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1903 ptlrpcd set if request is changed and can be safety changed to use
1908 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1909 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1910 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1911 these macros and replace them with their difinitions instead.
1915 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1916 Details : Update which_patch & target file for SLES10 latest kernel.
1919 Description: Remove client patches from SLES 10 kernel.
1920 Details : This causes SLES 10 clients to behave as patchless clients
1921 even on a Lustre-patched (server) kernel.
1923 Severity : enhancement
1925 Description: self-adjustable client's lru lists
1926 Details : use adaptive algorithm for managing client cached locks lru
1927 lists according to current server load, other client's work
1928 pattern, memory activities, etc. Both, server and client
1929 side namespaces provide number of proc tunables for controlling
1932 Severity : enhancement
1934 Description: light-weight GSS support
1935 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
1936 and reduce performance overhead.
1938 Severity : enhancement
1940 Description: Linux keyring support
1941 Details : Support using service of Linux keyring for Lustre GSS internal
1942 context refresh/cache mechanism.
1946 Description: Fix errors in lfs documentation
1947 Details : Fixes man pages
1951 Description: don't use GFP_* in generic Lustre code.
1952 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
1956 Description: obdclass is limited by single OBD_ALLOC(idarray)
1957 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
1961 Description: warnings with build patchless client with vanila 2.6.19 and up
1962 Details : change old ctl_table style and replace ctl_table/ctl_table_header
1963 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
1967 Frequency : Only for SLES
1968 Description: sanity_quota fail test_1
1969 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
1970 file, which makes TSTID[2] inunique.
1974 Frequency : Only for SLES9
1975 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
1976 Details : Update target/ChangeLog/which_patch .
1980 Frequency : Only for test_10 in sanity-quota.sh
1981 Description: a bug in quota test code
1982 Details : checking a test flag in a wrong place
1986 Frequency : Only for test_13 in sanity-quota.sh
1987 Description: a bug in quota test script
1988 Details : shouldn't check overfull allocated quota
1992 Description: jbd statistics
1993 Details : Port older jbd statistics patch for sles10
1995 Severity : enhancement
1997 Description: Add printing inode info into message about error in writepage.
2002 Description: reply_lock_interpret crash due to race with it and lock cancel.
2003 Details : Do not replay locks that are being cancelled. Do not reference
2004 locks by their address during replay, just by their handle.
2008 Frequency : When flocks are used.
2009 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
2010 Details : Flock locks might destroy just granted lock if it could be merged
2011 with another existing flock, this is done in completion handler,
2012 so teach ldlm_cli_enquque_fini that this is a valid case for
2018 Description: Oops in read and write path when failing to allocate lock.
2019 Details : Check if lock allocation failed and return error back.
2023 Description: lstripe command fails for valid OST index
2024 Details : The stripe offset is compared to lov->desc.ld_tgt_count
2025 instead of lov->desc.ld_active_tgt_count.
2029 Description: sanity.sh failed test 103
2030 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
2035 Description: df doesn't work properly if diskfs blocksize != 4K
2036 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
2040 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
2041 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
2043 Severity : enhancement
2045 Description: Adaptive timeouts
2046 Details : RPC timeouts adapt to changing server load and network
2047 conditions to reduce resend attempts and improve recovery time.
2051 Description: llapi_file_create() does not allow some changes
2052 Details : add llapi_file_open() that allows specifying the mode and
2053 open flags, and also returns an open file handle.
2057 Description: merge and cleanup kernel patches.
2058 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
2062 Description: Add journal checksum support.(Kernel part)
2063 Details : The journal checksum feature adds two new flags i.e
2064 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
2065 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
2066 indicates that the commit block contains the checksum for
2067 the blocks described by the descriptor blocks. Now commit
2068 record can be sent to disk without waiting for descriptor
2069 blocks to be written to disk. This behavior is controlled
2070 using JBD2_FEATURE_ASYNC_COMMIT flag.
2074 Description: OSS needs mutliple precreate threads
2075 Details : Add ability to start more than one create thread per OSS.
2079 Description: Sanity.sh test_65k failed: setstripe should have succeeded
2080 Details : execute lfs setstripe on client
2084 Description: mds_obd_create error creating tmp object
2085 Details : When the user sets quota on root, llog will be affected and can't
2086 create files and write files.
2089 Frequency : Always on ia64 patchless client, and possibly others.
2091 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
2092 Details : This allows the patchless client to be loaded on architectures
2093 without this export.
2097 Description: RedHat Update kernel for RHEL5
2098 Details : Add the kernel config file for RHEL5.
2102 Description: RedHat Update kernel for RHEL5
2103 Details : Modify the kernel config file more closer RHEL5.
2107 Description: Build failure against Centos5 (RHEL5)
2108 Details : Define PAGE_SIZE when it isn't present.
2112 Description: test_7 on sanity-quota.sh can't be used on separate machines
2113 Details : add facet to handle it
2117 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
2118 Details : with NFS, the anon dentry's parent was set to itself in
2119 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
2120 corresponding dentry other than use rec->ur_name.
2122 Severity : enhancement
2124 Description: lfs setstripe enhancement
2125 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
2129 Description: enable data checksumming by default
2130 Details : enable checksum by default, allow --disable-checksum
2131 configure option and "-o nochecksum" mount option. Checksums
2132 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
2136 Description: lustre support for RHEL5
2137 Details : Add support for RHEL5.
2141 Description: Client eviction due to failover config
2142 Details : after a connection loss, the lustre client should attempt to
2143 reconnect to the last active server first before trying the
2144 other potential connections.
2148 Description: when mds and osts use different quota unit(32bit and 64bit),
2149 quota will be released repeatly.
2150 Details : void sending multiple quota reqs to mds, which will keep the
2151 status between the reqs.
2155 Description: osts not allocated evenly to files
2156 Details : change the condition to increase offset_idx
2159 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
2161 Description: Data corruption for OSTs that are formatted larger than 2TB
2163 Details : When generating the bio request for lustre file writes the
2164 sector number would overflow a temporary variable before being
2165 used for the IO. The data reads correctly from Lustre (which
2166 will overflow in a similar manner) but other file data or
2167 filesystem metadata may be corrupted in some cases.
2171 Description: Tune the kernel for good SCSI performance.
2172 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
2173 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
2178 Description: rewrite ext2-derived code in obdclass/uuid.c
2179 Details : rewrite inherited code (uuid parsing code from ext2 utils)
2180 from scratch preserving functionality.
2184 Description: "lfs find -obd UUID" prints directories
2185 Details : "lfs find -obd UUID" will return all directory names instead
2186 of just file names. It is incorrect because the directories
2187 do not reside on the OSTs.
2191 Description: use i_size_read and i_size_write in 2.6 port
2192 Details : replace inode->i_size access with i_size_read/write()
2195 Frequency : when removing large files
2197 Description: scheduling issue during removal of large Lustre files
2198 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
2199 It causes scheduling issues when removing large files (17TB in the
2203 Frequency : only with liblustre clients on XT3
2205 Description: evictions taking too long
2206 Details : allow llrd to evict clients directly on OSTs
2209 Frequency : only on ppc
2211 Description: /proc/fs/lustre/devices broken on ppc
2212 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
2213 We should make sure the type of 'index' is loff_t and then cast
2214 explicitly as needed below. Do not assign an explicitly cast
2218 Frequency : only for rhel5
2220 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
2221 Details : Modify the target file & which_kernel.
2223 Severity : enhancement
2225 Description: omit set fsid for export NFS
2226 Details : fix set/restore device id for avoid EMFILE error and mark lustre
2227 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
2231 Frequency : Always, for kernels after 2.6.16
2232 Description: Fix warning idr_remove called for id=.. which is not allocated.
2233 Details : Last kernels save old s_dev before kill super and not allow
2234 to restore from callback - restore it before call kill_anon_super.
2238 Description: buffer overruns could theoretically occur
2239 Details : llapi_semantic_traverse() modifies the "path" argument by
2240 appending values to the end of the origin string, and a
2241 overrun may occur. Adding buffer overrun check in liblustreapi.
2245 Description: Fix error on 'ls .' at the top of the Lustre mount.
2246 Details : Don't revalidate dentry if it is a root dentry.
2250 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
2251 Details : Modify vm-tunables-rhel4.patch.
2255 Description: Kernel config for 2.6.18-vanilla.
2256 Details : Modify targets/2.6-vanilla.target.in.
2257 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
2258 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
2259 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
2260 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
2264 Description: improve handling recoverable errors
2265 Details : if request processig with error which can be recoverable on server
2266 request should be resend, otherwise page released from cache and
2271 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
2272 Details : Modify target file & which_patch.
2273 A flaw was found in the IA32 system call emulation provided
2274 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
2275 value could be stored in the %RAX register, which could trigger an
2276 out-of-bounds system call table access. An untrusted local user
2277 could exploit this flaw to run code in the kernel
2278 (ie a root privilege escalation). (CVE-2007-4573).
2282 Description: O_DIRECT bypasses client statistics.
2283 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
2284 read_ahead_stats were updated. copied the stats section from
2285 osc_send_oap_rpc() into async_internal().
2289 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
2293 Description: change order of libsysio includes
2294 Details : '#include sysio.h' should always come before '#include xtio.h'
2298 Description: handle "serious error: objid * already exists" more gracefully
2299 Details : If LAST_ID value on disk is smaller than the objects existing in
2300 the O/0/d* directories, it indicates disk corruption and causes an
2301 LBUG(). If the object is 0-length, then we should use the existing
2302 object. This will help to avoid a full fsck in most cases.
2304 Severity : enhancement
2306 Description: adapt the lustre_config script to support the upgrade case
2307 Details : Add "-u" option for lustre_config script to support upgrading 1.4
2308 server targets to 1.6 in parallel.
2312 Description: To avoid grant space > avaible space when the disk is almost
2313 full. Without this patch you might see the error "grant XXXX >
2314 available" or some LBUG about grant, when the disk is almost
2316 Details : In filter_check_grant, for non_grant cache write, we should
2317 check the left space by if (*left > ungranted + bytes), instead
2318 of (*left > ungranted), because only we are sure the left space
2319 is enough for another "bytes", then the ungrant space should be
2320 increase. In client, we should update cl_avail_grant only there
2321 is OBD_MD_FLGRANT in the reply.
2325 Description: Update RHEL 4 kernel to fix local root privilege escalation.
2326 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
2327 described in CVE-2007-4573. This problem could allow untrusted
2328 local users to gain root access.
2331 Frequency : when using O_DIRECT and quotas
2333 Description: Incorrect file ownership on O_DIRECT output files
2334 Details : block usage reported by 'lfs quota' does not take into account
2335 files that have been written with O_DIRECT.
2340 Description: touch file failed when fs is not full
2341 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
2342 otherwise we can get ENOSP while fs is not full.
2346 Description: parallel lock callbacks
2347 Details : Instead of sending blocking and completion callbacks as separated
2348 requests, adding them to a set and sending in parallel.
2351 Frequency : only for Cray XT3
2352 Bugzilla : 12829/13455
2353 Description: Changing primary group doesn't change the group lustre assigns to
2355 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
2356 overridden with the primary group provided by the group upcall,
2357 whereas the supplied fsgid can be trusted if it is in the list of
2358 supplementary groups returned by the group upcall.
2360 Severity : enhancement
2362 Description: Allow masking D_WARNING, D_ERROR messages from console
2363 Details : Console messages can now be disabled via lnet.printk.
2367 Description: User code with malformed file open parameter crashes client node
2368 Details : Before packing join_file req, all the related reference should be
2369 checked carefully in case some malformed flags cause fake
2370 join_file req on client.
2374 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
2375 Details : ldlm_completion_ast() assumes that a lock is granted when the req
2376 mode is equal to the granted mode. However, it should also check
2377 that LDLM_FL_CP_REQD is not set.
2381 Description: Heavy nfs access might result in deadlocks
2382 Details : After ELC code landed, it is now improper to enqueue any mds
2383 locks under och_sem, because enqueue might want to decide to
2384 cancel open locks for same inode we are holding och_sem for.
2388 Description: Client eviction while running blogbench
2389 Details : A lot of unlink operations with concurrent I/O can lead to a
2390 deadlock causing evictions. To address the problem, the number of
2391 oustanding OST_DESTROY requests is now throttled to
2392 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
2393 callbacks are processed in priority.
2397 Description: enable ACLs on MDS by default
2398 Details : ACLs must be enabled on MDS by default.
2401 Frequency : PPC/PPC64 only
2403 Description: conflicts between asm-ppc64/types.h and lustre_types.h
2404 Details : fix duplicated definitions between asm-ppc64/types.h and
2405 lustre_types.h on PPC.
2408 Frequency : PPC/PPC64 only
2410 Description: asm-ppc/segment.h does not exist
2411 Details : fix compile issue on PPC.
2415 Description: better handle error messages in extents code
2418 Frequency : RHEL4 only
2420 Description: mkfs is very slow on IA64/RHEL4
2421 Details : A performance regression has been discovered in the MPT Fusion
2422 driver between versions 3.02.73rh and 3.02.99.00rh. As a
2423 consequence, we have downgraded the MPT Fusion driver in the RHEL4
2424 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
2426 Severity : enhancement
2428 Description: SNMP support enhancement
2429 Details : Adding total number of sampled request for an MDS node in snmp
2432 Severity : enhancement
2434 Description: Optimize ldlm waiting list processing for PR extent locks
2435 Details : When processing waiting list for read extent lock and meeting read
2436 lock that is same or wider to it that is not contended, skip
2437 processing rest of the list and immediatelly return current
2438 status of conflictness, since we are guaranteed there are no
2439 conflicting locks in the rest of the list.
2443 Description: Time out and refuse to reconnect
2444 Details : When the failover node is the primary node, it is possible
2445 to have two identical connections in imp_conn_list. We must
2446 compare not conn's pointers but NIDs, otherwise we can defeat
2447 connection throttling.
2451 Description: port llog fixes from b1_6 into HEAD
2452 Details : Port llog reference couting and some llog cleanups from b1_6
2453 (bug 10800) into HEAD, for protect from panic and access to already
2454 free llog structures.
2458 Description: Detect stride IO mode in read-ahead
2459 Details : When a client does stride read, read-ahead should detect that and
2460 read-ahead pages according to the detected stride pattern.
2464 Description: data checksumming impacts single node performance
2465 Details : add support for several checksum algorithm. Currently, only CRC32
2466 and Adler-32 are supported. The checksum type can be changed on
2467 the fly via /proc/fs/lustre/osc/*/checksum_type.
2471 Description: use adler32 for page checksums
2472 Details : when available, use the Adler-32 algorithm instead of CRC32 for
2477 Description: build for x2 fails
2478 Details : fix compile issue on Cray systems.
2482 Description: Properly match for duplicate locks
2483 Details : Due to different lock order from skiplists code, we need to
2484 traverse entire list for now
2487 Frequency : only on PPC/SLES10
2489 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
2490 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
2491 be defined. Add a hack in mkfs_lustre.c to work around this compile
2496 Description: LASSERT on MDS when client holding flock lock dies
2497 Details : ldlm pool logic depends on number of granted locks equal to
2498 number of released locks which is not true for flock locks, so
2499 just exclude such locks from consideration.
2503 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
2504 Details : Use fsfilt_sync() for both the whole filesystem sync and
2505 individual file sync to eliminate dangerous inode locking
2506 with I_LOCK that can lead to a deadlock.
2510 Description: performance in 1.6.3
2511 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
2512 sure that 1MB requests are not fragmented by the block layer.
2514 Severity : enhancement
2516 Description: organize the server-side client stats on per-nid basis
2517 Details : Change the structure of stats under obdfilter and mds to
2525 The "uuid"s file would list the uuids of _active_ exports.
2526 And the clear entry is to clear all stats and stale nids.
2528 Severity : enhancement
2530 Description: eliminate client locks in face of contention
2531 Details : file contention detection and lockless i/o implementation
2532 for contended files.
2536 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
2539 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
2541 Description: MDS may lose file striping (and hence file data) in some cases
2542 Details : If there are additional extended attributes stored on the MDS,
2543 in particular ACLs, SELinux, or user attributes (if user_xattr
2544 is specified for the client mount options) then there is a risk
2545 of attribute loss. Additionally, the Lustre file striping
2546 needs to be larger than default (e.g. striped over all OSTs),
2547 and an additional attribute must be stored initially in the
2548 inode and then increase in size enough to be moved to the
2549 external attribute block (e.g. ACL growing in size) for file
2554 Description: skiplist implementation simplification
2555 Details : skiplists are used to group compatible locks on granted list
2556 that was implemented as tracking first and last lock of each
2557 lock group the patch changes that to using doubly linked lists
2561 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
2562 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
2563 because LASSERT on both the data supplied by a client, and the
2564 data on disk is dangerous and incorrect.
2566 Severity : enhancement
2568 Description: Slow truncate/writes to huge files at high offsets.
2569 Details : Directly associate cached pages to lock that protect those pages,
2570 this allows us to quickly find what pages to write and remove
2571 once lock callback is received.
2575 Description: more ldlm soft lockups
2576 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
2577 starve other threads from the resource lock for a long time in
2578 case of long waiting queue, so change the debug level from
2579 D_OTHER to the less frequently used D_INFO.
2581 Severity : enhancement
2583 Description: add -gid, -group, -uid, -user options to lfs find
2587 Description: Hung threads in invalidate_inode_pages2_range
2588 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
2589 once one is completed. As a result, some RPCs are stuck in the
2590 queue and are never sent.
2594 Description: filter threads hungs on waiting journal commit
2595 Details : Cleanup filter group llog code, then only filter group llog will
2596 be only created in the MDS/OST syncing process.
2600 Description: Procfs and llog threads access destoryed import sometimes.
2601 Details : Sync the import destoryed process with procfs and llog threads by
2602 the import refcount and semaphore.
2604 Severity : enhancement
2606 Description: openlock cache of b1_6 port to HEAD
2611 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
2612 Details : If insertion of an extent fails, then discard the inode
2613 preallocation and free data blocks else it can lead to duplicate
2618 Description: don't always update ctime in ext3_xattr_set_handle()
2619 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
2620 In some cases the ctime should not be updated, for example for
2621 2.0->1.8 compatibility it is necessary to delete an xattr and it
2622 should not update the ctime.
2626 Bugzilla : 15713/16362
2627 Description: Assertion in iopen_connect_dentry in 1.6.3
2628 Details : looking up an inode via iopen with the wrong generation number can
2629 populate the dcache with a disconneced dentry while the inode
2630 number is in the process of being reallocated. This causes an
2631 assertion failure in iopen since the inode's dentry list contains
2632 both a connected and disconnected dentry.
2636 Description: assertion failure in ldlm_handle2lock()
2637 Details : fix a race between class_handle_unhash() and class_handle2object()
2638 introduced in lustre 1.6.5 by bug 13622.
2643 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
2644 Details : remove the limit of 256 scsi disks in the sd_iostat patch
2649 Description: oops in sd_iostats_seq_show()
2650 Details : unloading/reloading the scsi low level driver triggers a kernel
2651 bug when trying to access the sd iostat file.
2656 Description: Kernel panics during QLogic driver reload
2657 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
2658 patch, causing memory corruption.
2663 Description: journal_dev option does not work in b1_6
2664 Details : pass mount option during pre-mount.
2666 Severity : enhancement
2668 Description: Add a FIEMAP(FIle Extent MAP) ioctl
2669 Details : FIEMAP ioctl will allow an application to efficiently fetch the
2670 extent information of a file. It can be used to map logical blocks
2671 in a file to physical blocks in the block device.
2675 Description: LDLM soft lockups - improvement
2676 Details : It is be possible to send the lock handle along with each read
2677 or write request because the client is already doing a lock match
2678 itself so there isn't any reason the OST should have to re-do that
2682 Frequency : only on Cray X2
2684 Description: X2 build failures
2685 Details : fix build failures on Cray X2.
2689 Description: xid & resent requests
2690 Details : Initialize RPC XID from clock at startup (randomly if clock is
2693 Severity : enhancement
2695 Description: Add lustre_start utility to start or stop multiple Lustre servers
2700 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
2701 Details : In case of memory pressure, list_del() can be called twice on
2702 req->rq_history_list, causing a kernel oops.
2706 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
2707 Details : fix stack overflow in the distributed lock manager by defering
2708 export eviction after a failed AST to the elt thread instead of
2709 handling it in the dlm interpret routine.
2713 Description: Convert some comments to new format.
2714 Details : Update documenting comments to match doxygen conventions.
2718 Description: Grammar fixes.
2719 Details : A couple of trivial spelling fixes.
2723 Description: OSD_COUNTERS-mandatory
2724 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
2725 to simplify the code.
2729 Description: simplify cmm_device freeing logic.
2730 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
2734 Description: Add lockdep support to dt_object_operations locking interface.
2735 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
2736 parameter indicating lock ordering. Update mdd code to use new
2741 Description: Introduce failloc constants for lockless IO tests.
2742 Details : Add two new failloc constants to test lockless IO. Only one of
2743 them in implemented---another is checked in yet to be landed
2748 Description: Add lockdep support for inode mutex.
2749 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
2750 in the situations where Lustre has to lock more than one inode
2755 Description: Add optional invariants checking support.
2756 Details : Add new LINVRNT() macro, optional on new --enable-invariants
2757 configure switch. This macro is to be used for consistency and
2758 sanity checks that are too expensive to be left in `production'
2763 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
2764 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
2769 Description: Add lockdep support for ldlm_lock and ldlm_resource.
2770 Details : Use spin_lock_nested() in (the only) situation where more than
2771 one ldlm_lock is locked simultaneously. Also, fix possible
2772 dead-lock in ldlm_lock_change_resource() by enforcing particular
2777 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
2778 Details : Instead of specifying each ldlm_lock call-back through separate
2779 parameter, wrap them into struct ldlm_callback_suite.
2783 Description: Kill join_lru obd method and its callers.
2784 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
2785 in memory---a requirement implemented through ->o_join_lru() obd
2786 method in HEAD. Remove this method and its users.
2790 Description: Add asynchronous ldlm ENQUEUE completion handler.
2791 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
2792 case that stock ldlm_completion_ast() cannot handle as it waits
2793 until lock is granted. Introduce new ldlm_completion_ast_async()
2794 for this. Also comment ldlm_completion_ast().
2798 Description: ldlm_error <-> errno conversion.
2799 Details : Add functions to map (rather arbitrary) between LDLM error codes
2800 and standard errno values. CLIO needs this to prevent LDLM specific
2801 constants from escaping ldlm and osc.
2805 Description: Kill unused ldlm_handle2lock_ns() function.
2806 Details : Kill unused ldlm_handle2lock_ns() function.
2810 Description: Add lu_ref support to ldlm_lock
2811 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
2812 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
2813 and ldlm_resource. LDLM interface has to be changed, because code that
2814 releases a reference on a lock, has to "know" what reference this is.
2815 In the most frequent case
2817 lock = ldlm_handle2lock(handle);
2819 LDLM_LOCK_PUT(lock);
2821 no changes are required. When any other reference (received _not_ from
2822 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
2823 instead of LDLM_LOCK_PUT().
2825 Arguably, changes are pervasive, and interface requires some discipline
2826 for proper use. On the other hand, it was very instrumental in finding
2827 a few leaked lock references.
2831 Description: Add ldlm_lock_addref_try().
2832 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
2833 attempts to addref a lock that might be being canceled
2838 Description: Add ldlm_weigh_callback().
2839 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
2840 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
2841 hard-coded `number of pages' logic.
2845 Description: Add lockdep annotations to llog code.
2846 Details : Use appropriately tagged _nested() locking calls in the places
2847 where llog takes more than one ->lgh_lock lock.
2851 Description: Add loi_kms_set().
2852 Details : Wrap kms updates into a helper function.
2856 Description: Constify instances of struct lsm_operations.
2857 Details : Constify instances of struct lsm_operations.
2861 Description: lu_conf support.
2862 Details : On a server, a file system object is uniquely identified
2863 by a fid, which is sufficient to locate and load all object
2864 state (inode). On a client, on the other hand, more data are
2865 necessary instantiate an object. Change lu_object_find() and
2866 friends to take additional `lu_conf' argument describing object.
2867 Typically this includes layout information.
2871 Description: lu_context fixes.
2872 Details : Introduce new lu_context functions that are needed on the client
2873 side, where some system threads (ptlrpcd) are shared by multiple
2874 modules, and so cannot be stopped during module shutdown.
2878 Description: Add start and stop methods to lu_device_type_operations.
2879 Details : Introduce two new methods in lu_device_type_operations, that are
2880 invoked when first instance of a given type is created and last one
2881 is destroyed respectively. This is need by CLIO.
2885 Description: Add lu_ref support to struct lu_device.
2886 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
2887 track leaked references.
2891 Description: Introduce lu_kmem_descr.
2892 Details : lu_kmem_descr and its companion interface allow to create
2893 and destroy a number of kmem caches at once.
2897 Description: Fix lu_object finalization race.
2898 Details : Fix a race between lu_object_find() finding an object and its
2899 concurrent finalization. This race is (most likely) not possible
2900 on the server, but might happen on the client.
2904 Description: Introduce lu_ref interface.
2905 Details : lu_ref is a debugging module allowing to track references to
2906 a given object. It is quite cpu expensive, and has to be
2907 explicitly enabled with --enable-lu_ref. See usage description
2912 Description: Factor lu_site procfs stats into a separate function.
2913 Details : Separate lu_site stats printing code into a separate function
2914 that can be reused on a client.
2918 Description: Constify instances of struct {lu,dt,md}_device_operations.
2919 Details : Constify instances of struct {lu,dt,md}_device_operations.
2923 Description: Introduce struct md_site and move meta-data specific parts of
2924 struct lu_site here.
2925 Details : Move md-specific fields out of struct lu_site into special struct
2926 md_site, so that lu_site can be used on a client.
2930 Description: Kill mdd_lov_destroy().
2931 Details : Remove unused mdd code.
2935 Description: Add st_block checking to multistat.c.
2936 Details : Add st_block checking to multistat.c.
2940 Description: Add lu_ref support to struct obd_device.
2941 Details : Add lu_ref tracking to obd_device.
2945 Description: Kill obd_set_fail_loc().
2946 Details : Remove unused code.
2950 Description: Add special type for ptlrpc_request interpret functions.
2951 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
2952 there. Actual usage will be in CLIO.
2956 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
2957 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
2958 doesn't work with lockdep.
2962 Description: Add rwv.c test program.
2963 Details : New testing program exercising readv(2) and writev(2) (Qian).
2967 Description: Add sendfile.c test program.
2968 Details : New testing program exercising sendfile(2) (Jay).
2972 Description: Ratelimit a message that can be very frequent.
2973 Details : Ratelimit a memory allocation failure message that can
2978 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2979 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2983 Description: Liblustre build fixes.
2984 Details : Liblustre build fixes.
2988 Description: libcfs: add cfs_{need,cond}_resched() interface.
2989 Details : libcfs: add cfs_{need,cond}_resched() definition and
2990 implementations for Linux, NT, and liblustre.
2992 Severity : enhancement
2994 Description: More exported tunables for mballoc
2995 Details : Add support for tunable preallocation window and new tunables for
2996 large/small requests
3000 Description: Detect corruption of block bitmap and checking for preallocations
3001 Details : Checks validity of on-disk block bitmap. Also it does better
3002 checking of number of applied preallocations. When corruption is
3003 found, it turns filesystem readonly to prevent further corruptions.
3007 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
3008 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
3009 window to 0. If it does hit cache miss, and read-pattern is still
3010 stride-io mode, does not reset the stride window, but also does not
3011 increase the stride window length in this case.
3015 Frequency : only for big-endian servers
3016 Description: Check if system is big-endian while mounting fs with extents feature
3017 Details : Mounting a filesystem with extents feature will fail on big-endian
3018 systems since ext3-based ldiskfs is not supported on big-endian
3019 systems. This can be over-riden with "bigendian_extents" mount option.
3021 Severity : enhancement
3023 Description: The root squash functionality
3024 Details : A security feature, which is to prevent users from being able
3025 to mount lustre on their desktop, run as root, and delete
3026 all of the files in the filesystem. The goal is accomplished by
3027 remapping user id (UID) and group id (GID) of the root user to
3028 a UID and GID specified by the system administartor via Lustre
3029 configuration management server (MGS). The functionality also
3030 allows to specify sets of clients for which the remapping does
3035 Description: Excessive recovery window
3036 Details : With AT enabled, the recovery window can be excessively long (6000+
3037 seconds). To address this problem, we no longer use
3038 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
3039 timeout no longer depends on the service time, it is set to
3040 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
3041 time via pb_service_time.
3045 Description: Watchdog triggered on MDS failover
3046 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
3047 the OSTs know that the MDS "UUID" can be reused for the same export
3048 from a different NID, so we do not need to wait for the export to be
3052 Frequency : rare, only if using MMP with Linux RAID
3054 Description: MMP doesn't work with Linux RAID
3055 Details : While using HA for Lustre servers with Linux RAID, it is possible
3056 that MMP will not detect multiple mounts. To make this work we
3057 need to unplug the device queue in RAID when the MMP block is being
3058 written. Also while reading the MMP block, we should read it from
3059 disk and not the cached one.
3061 Severity : enhancement
3063 Description: open file using fid
3064 Details : A file can be opened using just its fid, like
3065 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
3068 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
3071 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
3072 Details : add proper locking for creation of jbd_4k slab cache
3076 Description: MMP check in ext3_remount() fails without displaying any error
3077 Details : When multiple mount protection fails during remount, proper error
3080 Severity : enhancement
3082 Description: Allow stripe size to be up to 4G-64k
3083 Details : Fix math logic to allow large stripe sizes.
3087 Description: add check for >8TB ldiskfs filesystems
3088 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
3089 Don't allow >8TB ldiskfs filesystems to be mounted without
3090 force_over_8tb mount option
3095 Descriptoin: OSS complains "too many blocks" and crashes
3096 Details : In filter_commitrw_write():
3097 1. lquota_chkquota();
3098 2. obdo_from_inode();
3099 3. lquota_pending_commit();
3100 "2." has changed uid and gid in this case, then caused
3101 lquota_chkquota() and lquota_pending_commit() to use
3102 different uid and gid. That is the root of the bug.
3106 Descriptoin: getstripe doesn't report stripe count = -1 sanely
3107 Details : Detect and print out the 'stripe count = -1' case properly.
3111 Description: Out or order replies might be lost on replay
3112 Details : In ptlrpc_retain_replayable_request if we cannot find retained
3113 request with tid smaller than one currently being added, add it
3114 to the start, not end of the list.
3116 --------------------------------------------------------------------------------
3118 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
3120 * Support for kernels:
3121 2.6.5-7.283 (SLES 9),
3122 2.6.9-55.EL (RHEL 4),
3123 2.6.16.46-0.14 (SLES 10),
3124 2.6.18.8 vanilla (kernel.org)
3125 * Client support for unpatched kernels:
3126 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
3127 2.6.16 - 2.6.22 vanilla (kernel.org)
3128 * Due to recently discovered recovery problems, we do not recommend
3129 using patchless RHEL 4 clients with this or any earlier release.
3130 * Recommended e2fsprogs version: 1.39.cfs8
3131 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
3132 * Starting with this release, the ldiskfs backing filesystem required
3133 by Lustre is now in its own package, lustre-ldiskfs. This package
3134 should be installed. It is versioned separately from Lustre and
3135 may be released separately in future.
3139 Description: block reactivating mgc import until all deactivates complete
3140 Details : Fix race when failing back MDT/MGS to itself (testing)
3143 Frequency : at statup only
3145 Description: mds_lov_synchronize race leads to various problems
3146 Details : simultaneous MDT->OST connections at startup can cause the
3147 sync to abort, leaving the OSC in a bad state.
3149 Severity : enhancement
3151 Description: add optional extra BUILD_VERSION info
3152 Details : add a new environment variable (namely LUSTRE_VERS) which allows
3153 to override the lustre version.
3156 Frequency : 2.6.18 servers only
3158 Description: ll_kern_mount() doesn't release the module reference
3159 Details : The ldiskfs module reference count never drops down to 0
3160 because ll_kern_mount() doesn't release the module reference.
3165 Description: server LBUG when using old ost_num_threads parameter
3166 Details : Accept the old ost_num_threads parameter but warn that it
3167 is deprecated, and fix an off-by-one error that caused an LBUG.
3172 Description: Transient SCSI error results in persistent IO issue
3173 Details : iobuf->dr_error is not reinitialized to 0 between two
3177 Frequency : sometimes when underlying device returns I/O errors
3179 Description: OSTs not going read-only during write failures
3180 Details : OSTs are not remounted read-only when the journal commit threads
3181 get I/O errors because fsfilt_ext3 calls journal_start/stop()
3182 instead of the ext3 wrappers.
3186 Description: poor connect scaling with increasing client count
3187 Details : Don't run filter_grant_sanity_check for more than 100 exports
3188 to improve scaling for large numbers of clients.
3191 Frequency : SLES10 only
3193 Description: sanity-quota.sh quotacheck failed: rc = -22
3194 Details : Quotas cannot be enabled on SLES10.
3197 Frequency : liblustre clients only
3199 Description: getdirentries does not give error when run on compute nodes
3200 Details : getdirentries does not fail when the size specified as an argument
3201 is too small to contain at least one entry
3203 Severity : enhancement
3205 Description: Add LNET router traceability for debug purposes
3206 Details : If a checksum failure occurs with a router as part of the
3207 IO path, the NID of the last router that forwarded the bulk data
3208 is printed so it can be identified.
3213 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
3214 Details : Due to a race condition, liblustre clients were occasionally
3215 evicted incorrectly.
3217 Severity : enhancement
3219 Description: lfs setstripe use optional parameters instead of postional
3222 Severity : enhancement
3224 Description: Nanosecond timestamp support for ldiskfs
3225 Details : The on-disk ldiskfs filesystem has added support for nanosecond
3226 resolution timestamps. There is not yet support for this at
3227 the Lustre filesystem level.
3230 Frequency : during server recovery
3232 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
3233 Details : request with rq_no_resend flag not awake l_wait_event if they get
3237 Frequency : nfs export on patchless client
3239 Description: connectathon hang when test nfs export over patchless client
3240 Details : Disconnected dentry cannot be found with lookup, so we do not need
3241 to unhash it or make it invalid
3244 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
3245 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
3246 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
3251 Description: open req refcounting wrong on reconnect
3252 Details : If reconnect happened between getting open reply from server and
3253 call to mdc_set_replay_data in ll_file_open, we will schedule
3254 replay for unreferenced request that we are about to free.
3255 Subsequent close will crash in variety of ways.
3256 Check that request is still eligible for replay in
3257 mdc_set_replay_data().
3262 Description: disable writes to filesystem when reading health_check file
3263 Details : the default for reading the health_check proc file has changed
3264 to NOT do a journal transaction and write to disk, because this
3265 can cause reads of the /proc file to hang and block HA state
3266 checking on a healthy but otherwise heavily loaded system. It
3267 is possible to return to the previous behaviour during configure
3268 with --enable-health-write.
3270 Severity : enhancement
3272 Description: 64-bit inode version
3273 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
3274 to the inode. This will be required for version-based recovery.
3279 Description: MDS fails to start if a duplicate client export is detected
3280 Details : in some rare cases it was possible for a client to connect to
3281 an MDS multiple times. Upon recovery the MDS would detect this
3282 and fail during startup. Handle this more gracefully.
3284 Severity : enhancement
3286 Description: Add -o localflock option to simulate old noflock
3288 Details : This will achieve local-only flock/fcntl locks
3294 Description: log_commit_thread vs filter_destroy race leads to crash
3295 Details : Take import reference before releasing llog record semaphore
3300 Description: Wrong request locking in request set processing
3301 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
3302 imp_delayed_list, in this place should be used imp_lock.
3305 Frequency : when reconnection
3307 Description: Grant Leak when osc reconnect to OST
3308 Details : When osc reconnect ost, OST(filter) should check whether it
3309 should grant more space to client by comparing fed_grant and
3310 cl_avail_grant, and return the granted space to client instead
3311 of "new granted" space, because client will call osc_init_grant
3312 to update the client grant space info.
3315 Frequency : when client reconnect to OST
3317 Description: Grant Leak when osc do resend and replay bulk write
3318 Details : When osc reconnect to OST, OST(filter)should clear grant info of
3319 bulk write request, because the grant info while be sync between
3320 OSC and OST when reconnect, and we should ignore the grant info
3321 these of resend/replay write req.
3326 Description: Grant space more than available left space sometimes.
3327 Details : When then OST is about to be full, if two bulk writing from
3328 different clients came to OST. Accord the avaliable space of the
3329 OST, the first req should be permitted, and the second one
3330 should be denied by ENOSPC. But if the seconde arrived before
3331 the first one is committed. The OST might wrongly permit second
3332 writing, which will cause grant space > available space.
3335 Frequency : when client is evicted
3337 Description: Grant might be wrongly erased when osc is evicted by OST
3338 Details : when the import is evicted by server, it will fork another
3339 thread ptlrpc_invalidate_import_thread to invalidate the
3340 import, where the grant will be set to 0. While the original
3341 thread will update the grant it got when connecting. So if
3342 the former happened latter, the grant will be wrongly errased
3343 because of this race.
3348 Description: Checking Stale with correct fid
3349 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
3350 to check whether it is stale, because sometimes, we want the
3351 enqueue happened anyway, and op_data.fid2 will not be initialized.
3353 Severity : enhancement
3355 Description: update patchless client
3356 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
3359 Frequency : only with 2.4 kernel
3361 Description: random memory corruption
3362 Details : size of struct ll_inode_info is to big for union inode.u and this
3363 can be cause of random memory corruption.
3368 Description: Memory leak in recovery
3369 Details : Lov_mds_md was not free in an error handler in mds_create_object.
3370 It should also check obd_fail before fsfilt_start, otherwise if
3371 fsfilt_start return -EROFS,(failover mds during mds recovery).
3372 then the req will return with repmsg->transno = 0 and rc = EROFS.
3373 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
3374 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
3375 be freed no matter whether fsfilt_commit success or not.
3378 Frequency : only with huge count clients
3380 Description: Prevents from taking the superblock lock in llap_from_page for
3382 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
3383 for llap_from_page call in ll_removepage prevents from taking the
3384 superblock lock for a soon died page.
3389 Description: Not check open intent error before release open handle
3390 Details : in some rare cases, the open intent error is not checked before
3391 release open handle, which may cause
3392 ASSERTION(open_req->rq_transno != 0), because it tries to release
3393 the failed open handle.
3398 Description: Set cat log bitmap only after create log success.
3399 Details : in some rare cases, the cat log bitmap is set too early. and it
3400 should be set only after create log success.
3404 Description: Accessing a block bevice can re-enable I/O when Lustre is
3405 tearing down a device.
3406 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
3411 Description: service threads may hog cpus when there are a lot of requests
3412 Details : Insert cond_resched to give other threads a chance to use some CPU
3417 Description: the cat log was not initialized in recovery
3418 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
3419 unlink log on mds will not be initialized until mds post
3420 recovery. And also in mds post recovery, the unlink log will
3421 initialization will be done asynchronausly, so there will be race
3422 between add unlink log and unlink log initialization.
3426 Description: brw_stats were being printed incorrectly
3427 Details : brw_stats were being printed as log2 but all of them were not
3428 recorded as log2. Also remove some code duplication arising from
3429 filter_tally_{read,write}.
3433 Frequency : rare, only in recovery.
3434 Description: ASSERTION(req->rq_type != LI_POISON) failed
3435 Details : imp_lock should be held while iterating over imp_sending_list for
3436 prevent destroy request after get timeout in ptlrpc_queue_wait.
3440 Description: replay-single.sh test 52 fails
3441 Details : A lock's skiplist need to be cleanup when it being unlinked
3442 from its resource list.
3446 Description: Short directio read returns full requested size rather than
3448 Details : Direct I/O operations should return actual amount of bytes
3449 transferred rather than requested size.
3451 Severity : enhancement
3453 Description: metadata RPC reduction (e.g. for rm performance)
3454 Details : decrease the amount of synchronous RPC between clients and servers
3455 by canceling conflicing lock before the operation on the client
3456 and packing thier handles into the main operation RPC to server.
3458 Severity : enhancement
3460 Description: Async OSC create to avoid the blocking unnecessarily.
3461 Details : If an OST has no remaining object, system will block on the
3462 creation when it needs to create a new object on this OST. Now,
3463 ways use pre-created objects when available, instead of blocking on
3464 an empty osc while others are not empty. If we must block, we
3465 block for the shortest possible period of time.
3469 Description: improve handling recoverable errors
3470 Details : if request processig with error which can be recoverable on server
3471 request should be resend, otherwise page released from cache and
3474 Severity : enhancement
3476 Description: refine locking for avoid write wrong info into lov_objid file
3477 Details : fix possible races with add new target and write/update data in
3480 --------------------------------------------------------------------------------
3482 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
3487 Frequency : on some architectures
3489 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
3490 Details : Uninitialized flags sometimes cause configuration commands to
3494 Frequency : patchless clients only
3496 Description: missing __iget() symbol export
3497 Details : The __iget() symbol export is missing. To avoid the need for
3498 this on patchless clients the deathrow inode reaper is turned
3499 off, and we depend on the VM to clean up old inodes. This
3500 dependency was during via the fix for bug 12181.
3502 --------------------------------------------------------------------------------
3504 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
3506 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
3507 INTEROPERATE with older versions automatically. In many cases a
3508 special upgrade step is needed. Please read the
3509 user documentation before upgrading any part of a live system.
3510 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
3511 version will not interoperate with 1.6 betas before beta5 (1.5.95).
3512 * WARNING: Lustre configuration and startup changes are required with
3513 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
3517 Severity : enhancement
3519 Description: MountConf
3520 Details : Lustre configuration is now managed via mkfs and mount
3521 commands instead of lmc and lconf. New obd types (MGS, MGC)
3522 are added for dynamic configuration management. See
3523 https://mail.clusterfs.com/wikis/lustre/MountConf for
3526 Severity : enhancement
3528 Description: dynamic OST addition
3529 Details : OSTs can now be added to a live filesystem
3531 Severity : enhancement
3533 Description: startup order invariance
3534 Details : MDTs and OSTs can be started in any order. Clients only
3535 require the MDT to complete startup.
3537 Severity : enhancement
3539 Description: parallel, asynchronous orphan cleanup
3540 Details : orphan cleanup is now performed in separate threads for each
3541 OST, allowing parallel non-blocking operation.
3543 Severity : enhancement
3545 Description: optimized stripe assignment
3546 Details : stripe assignments are now made based on ost space available,
3547 ost previous usage, and OSS previous usage, in order to try
3548 to optimize storage space and networking resources.
3550 Severity : enhancement
3552 Description: Permanently set tunables
3553 Details : All writable /proc/fs/lustre tunables can now be permanently
3554 set on a per-server basis, at mkfs time or on a live system.
3556 Severity : enhancement
3558 Description: Lustre message v2
3559 Details : Add lustre message format v2.
3561 Severity : enhancement
3563 Description: client OST exclusion list
3564 Details : Clients can be started with a list of OSTs that should be
3565 declared "inactive" for known non-responsive OSTs.
3569 Description: SPEC SFS validation failure on NFS v2 over lustre.
3570 Details : Changes the blocksize for regular files to be 2x RPC size,
3571 and not depend on stripe size.
3573 Severity : enhancement
3575 Description: Multiple MD RPCs in flight.
3576 Details : Further unserialise some read-only MDS RPCs - learn about intents.
3577 To avoid overly-overloading MDS, introduce a limit on number of
3578 MDS RPCs in flight for a single client and add /proc controls
3579 to adjust this limit.
3581 Severity : enhancement
3583 Description: client read/write statistics
3584 Details : Add client read/write call usage stats for performance
3585 analysis of user processes.
3586 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
3587 file access. extents_stats shows chunk size distribution.
3588 extents_stats_per_process show chunk size distribution per
3591 Severity : enhancement
3593 Description: mds statistics
3594 Details : Add detailed mds operations statistics in
3595 /proc/fs/lustre/mds/*/stats.
3599 Description: Failure of copying files with lustre special EAs.
3600 Details : Client side always return success for setxattr call for lustre
3601 special xattr (currently only "trusted.lov").
3606 Description: Refcount LNET uuids
3607 Details : The global LNET uuid list grew linearly with every startup;
3608 refcount repeated list entries instead of always adding to
3611 Severity : enhancement
3613 Description: Dynamic service threads
3614 Details : Within a small range, start extra service threads
3615 automatically when the request queue builds up.
3618 Frequency : mixed-endian client/server environments
3620 Description: mixed-endian crashes
3621 Details : The new msg_v2 system had some failures in mixed-endian
3624 Severity : enhancement
3626 Description: Easy OST removal
3627 Details : OSTs can be permanently deactivated with e.g. 'lctl
3628 conf_param lustre-OST0001.osc.active=0'
3630 Severity : enhancement
3632 Description: MGS proc entries
3633 Details : Added basic proc entries for the MGS showing what filesystems
3636 Severity : enhancement
3638 Description: provide MGS failover
3639 Details : Added config lock reacquisition after MGS server failover.
3641 Severity : enhancement
3643 Description: add Linux 2.4 support
3644 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
3648 Description: a large application tries to do I/O to the same resource and dies
3649 in the middle of it.
3650 Details : Check the req->rq_arrival time after the call to
3651 ost_brw_lock_get(), but before we do anything about
3652 processing it & sending the BULK transfer request. This
3653 should help move old stale pending locks off the queue as
3654 quickly as obd_timeout.
3657 Frequency : when an incorrect nid is specified during startup
3659 Description: ptlrpc connect to non-existant node causes kernel crash
3660 Details : LNET can't be re-entered from an event callback, which
3661 happened when we expire a message after the export has been
3662 cleaned up. Instead, hand the zombie cleanup off to another
3665 Severity : enhancement
3667 Description: plain/inodebits lock performance improvement
3668 Details : Grouping plain/inodebits in granted list by their request modes
3669 and bits policy, thus improving the performance of search through
3673 Frequency : only if OST filesystem is corrupted
3675 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3676 Details : for a short time RPCs with bulk IO are in the replay list,
3677 but replay of bulk IOs is unimplemented. If the OST filesystem
3678 is corrupted due to disk cache incoherency and then replay is
3679 started it is possible to trip an assertion. Avoid putting
3680 committed RPCs into the replay list at all to avoid this issue.
3683 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3685 Description: System hang on startup
3686 Details : This bug allowed the liblustre (e.g. catamount) client to
3687 return to the app before handling all startup RPCs. This
3688 could leave the node unresponsive to lustre network traffic
3689 and manifested as a server ptllnd timeout.
3691 Severity : enhancement
3693 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3694 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
3695 boolean to control whether to print peer debug info when a
3696 client's RPC times out.
3699 Frequency : only for kernels with patches from Lustre below 1.4.3
3701 Description: Remove old rdonly API
3702 Details : Remove old rdonly API which unused from at least lustre 1.4.3
3705 Frequency : only for devices with external journals
3707 Description: Set external device read-only also
3708 Details : During a commanded failover stop, we set the disk device
3709 read-only while the server shuts down. We now also set any
3710 external journal device read-only at the same time.
3713 Frequency : when upgrading from 1.4 while trying to change parameters
3715 Description: The wrong (new) MDC name was used when setting parameters for
3716 upgraded MDT's. Also allows changing of OSC (and MDC)
3717 parameters if --writeconf is specified at tunefs upgrade time.
3720 Frequency : when setting specific ost indicies
3722 Description: QOS code breaks on skipped indicies
3723 Details : Add checks for missing OST indicies in the QOS code, so OSTs
3724 created with --index need not be sequential.
3726 Severity : enhancement
3728 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3729 Details : The uninit_groups feature works in conjunction with the kernel
3730 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3731 up the pass1 processing of e2fsck. This is a read-only feature
3732 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3733 mount filesystems that have had this feature enabled.
3735 Severity : enhancement
3737 Description: Improve multi-block allocation algorithm to avoid fragmentation
3738 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3739 allocation locality and avoid filesystem fragmentation.
3741 ------------------------------------------------------------------------------
3743 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
3745 * Support for kernels:
3746 2.4.21-47.0.1.EL (RHEL 3)
3747 2.6.5-7.283 (SLES 9)
3748 2.6.9-42.0.10.EL (RHEL 4)
3749 2.6.12.6 vanilla (kernel.org)
3750 2.6.16.27-0.9 (SLES 10)
3751 * Recommended e2fsprogs version: 1.39.cfs5
3753 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3757 Frequency : occasional, depends on client load and configuration
3758 Bugzilla : 12181, 12203
3759 Description: data loss for recently-modified files
3761 Details : In some cases it is possible that recently written or created
3762 files may not be written to disk in a timely manner (this should
3763 normally be within 30s unless client IO load is very high).
3764 The problem appears as zero-length files or files that are a
3765 multiple of 1MB in size after a client crash or client eviction
3766 that are missing data at the end of the file.
3768 This problem is more likely to be hit on clients where files are
3769 repeatedly created and unlinked in the same directory, clients
3770 have a large amount of RAM, have many CPUs, the filesystem has
3771 many OSTs, the clients are rebooted frequently, and/or the files
3772 are not accessed by other nodes after being written.
3774 The presence of the problem can be detected by looking at
3775 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
3776 smaller than the second (nr_unused) then dirty files will not
3777 be flushed automatically to disk. "sync; sleep 10" should be
3778 run several times on the node before unmounting it to update
3779 Lustre (this is also safe to run on nodes without this problem).
3781 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
3782 that can cause this same problem, so customers using that kernel
3783 also need to update the kernel in addition to Lustre. In order
3784 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
3786 It is normal that files written just before a client crash (less
3787 than 30s) may not yet have been flushed to disk, even for local
3791 Frequency : frequent on thin XT3 nodes
3793 Description: UUID collision on thin XT3 Linux nodes
3794 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
3795 randomly, since we relied on an insufficiently-seeded PRNG.
3800 Description: OSS hangs after "All ost request buffers busy"
3801 Details : A deadlock between quota and journal operations caused OSS
3802 hangs after printing "All ost request buffers busy."
3805 Frequency : always on liblustre builds
3807 Description: Cleanup compiler warnings on liblustre
3810 Frequency : always on liblustre builds on XT3
3812 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
3816 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
3817 > 32000 subdirectories
3818 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
3819 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
3820 subdirectory count crosses 32000. This will aid e2fsck to
3821 correctly handle more than 32000 subdirectories.
3824 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3826 Description: System hang on startup
3827 Details : This bug allowed the liblustre (e.g. catamount) client to
3828 return to the app before handling all startup RPCs. This
3829 could leave the node unresponsive to lustre network traffic
3830 and manifested as a server ptllnd timeout.
3832 Severity : enhancement
3834 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3835 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
3836 boolean to control whether to print peer debug info when a
3837 client's RPC times out.
3842 Description: make O_SYNC working on 2.6 kernels
3843 Details : 2.6 kernels use different method for mark pages for write,
3844 so need add a code to lustre for O_SYNC work.
3849 Description: Failure to close file and release space on NFS
3850 Details : Put inode details into lock acquired in ll_intent_file_open.
3851 Use mdc_intent_lock in ll_intent_open to properly
3852 detect all kind of errors unhandled by mdc_enqueue.
3857 Description: proc file read during shutdown sometimes raced obd removal,
3859 Details : Add lock to prevent obd access after proc file removal.
3862 Frequency : Only for files larger than 4GB on 32-bit clients.
3864 Description: improperly doing page alignment of locks
3865 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
3866 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3871 Description: RPCs being resent when they shouldn't be
3872 Details : Some RPCs that should not be resent are being resent. This
3873 can cause inconsistencies in the RPC state machine. Do not
3874 resend such requests.
3877 Frequency : rare, only with NFS export
3879 Description: Crash on NFS re-export node
3880 Details : under very unusual load conditions an assertion is hit in
3881 ll_intent_file_open()
3884 Frequency : only if OST filesystem is corrupted
3886 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3887 Details : for a short time RPCs with bulk IO are in the replay list,
3888 but replay of bulk IOs is unimplemented. If the OST filesystem
3889 is corrupted due to disk cache incoherency and then replay is
3890 started it is possible to trip an assertion. Avoid putting
3891 committed RPCs into the replay list at all to avoid this issue.
3896 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3897 Details : Large single O_DIRECT read and write calls can fail to allocate
3898 a sufficiently large buffer to process the request. In case of
3899 allocation failure the allocation is retried with a smaller
3900 buffer and broken into smaller requests.
3902 Severity : enhancement
3904 Description: Add -o localflock option to simulate old noflock behaviour.
3905 Details : This will achieve local-only flock/fcntl locks coherentness.
3910 Description: versioning check is incomplete
3911 Details : Checking the version difference of client vs. server, report
3912 error if the gap is too big.
3917 Description: add support PG_writeback bit
3918 Details : add support for PG_writeback bit for Lustre, for more carefull
3919 work with page cache in 2.6 kernel. This also fix some deadlocks
3920 and remove hack for work O_SYNC with 2.6 kernel.
3922 Severity : enhancement
3924 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3925 Details : The uninit_groups feature works in conjunction with the kernel
3926 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3927 up the pass1 processing of e2fsck. This is a read-only feature
3928 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3929 mount filesystems that have had this feature enabled.
3931 Severity : enhancement
3933 Description: Improve multi-block allocation algorithm to avoid fragmentation
3934 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3935 allocation locality and avoid filesystem fragmentation.
3937 ------------------------------------------------------------------------------
3939 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3941 * Support for kernels:
3942 2.6.9-42.0.3.EL (RHEL 4)
3943 2.6.5-7.276 (SLES 9)
3944 2.4.21-47.0.1.EL (RHEL 3)
3945 2.6.12.6 vanilla (kernel.org)
3946 2.6.16.21-0.8 (SLES10)
3947 * Recommended e2fsprogs version: 1.39.cfs2-0
3949 * The backwards-compatible /proc/sys/portals symlink has been removed
3950 in this release. Before upgrading, please ensure that you change
3951 any configuration scripts or /etc/sysctl.conf files that access
3952 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3953 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3954 in advance of the upgrade on any system running Lustre 1.4.6 or
3955 newer, since /proc/sys/lnet was added in that version.
3956 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3960 Frequency : only when quota is used
3962 Description: avoid scanning export list for quota master
3963 Details : Change the algorithms to avoid scanning export list in order
3964 to improve the efficiency.
3967 Frequency : MDS failover only, very rarely
3969 Description: "went back in time" messages on mds failover
3970 Details : The greatest transno may be lost when the current operation
3971 finishes with an error (transno==0) and the client's last_rcvd
3972 record is over-written. Save the greatest transno in the
3973 mds_last_transno for this case.
3976 Frequency : always for specific kernels and striping counts
3978 Description: client may get "Matching packet too big" without ACL support
3979 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
3980 when trying to access files in certain configurations. The
3981 clients should in fact be denied when mounting because they do
3982 not understand ACLs.
3985 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
3987 Description: many clients connecting with IO in progress causes connect timeouts
3988 Details : Avoid synchronous journal commits to avoid delays caused by many
3989 clients connecting/disconnecting when bulk IO is in progress.
3990 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
3991 OST_IO_PORTAL to avoid delays behind potentially many pending
3995 Frequency : occasionally with multiple writers to a single file
3997 Description: shared writes to file may result in wrong size reported by stat()
3998 Details : Allow growing of kms when extent lock is cancelled
4001 Frequency : always with random mmap IO to multi-striped file
4003 Description: mmap write might be lost if we are writing to a 'hole' in stripe
4004 Details : Only if the hole is at the end of OST object so that kms is too
4005 small. Fix is to increase kms accordingly in ll_nopage.
4008 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
4010 Description: writes to a missing object would leak memory on the OST
4011 Details : If there is an inconsistency between the MDS and OST filesystems,
4012 such that the MDS references an object that doesn't exist, writes
4013 to that object will leak memory due to incorrect cleanup in the
4014 error handling path, eventually running out of memory on the OST.
4019 Description: Creating too long symlink causes lustre errors
4020 Details : Check symlink and name lengths before sending requests to MDS.
4023 Frequency : only if flock is enabled (not on by default)
4025 Description: posix locks not released on fd closure on 2.6.9+
4026 Details : We failed to add posix locks to list of inode locks on 2.6.9+
4027 kernels, this caused such locks not to be released on fd close and
4028 then assertions on fs unmount about still used locks.
4031 Frequency : MDS failover only, very rarely
4033 Description: clients may get ASSERTION(granted_lock != NULL)
4034 Details : When request was taking a long time, and a client was resending
4035 a getattr by name lock request. The were multiple lock requests
4036 with the same client lock handle and
4037 mds_getattr_name->fixup_handle_for_resent_request found one of the
4038 lock handles but later failed with ASSERTION(granted_lock != NULL).
4043 Description: handle->h_buffer_credits > 0, assertion failure
4044 Details : h_buffer_credits is zero after truncate, causing assertion
4045 failure. This patch extends the transaction or creates a new
4049 Frequency : NFS re-export or patchless client
4050 Bugzilla : 11179, 10796
4051 Description: Crash on NFS re-export node (__d_move)
4052 Details : We do not want to hash the dentry if we don't have a lock.
4053 But if this dentry is later used in d_move, we'd hit uninitialised
4054 list head d_hash, so we just do this to init d_hash field but
4055 leave dentry unhashed.
4058 Frequency : NFS re-export or patchless client
4060 Description: NFS exports has problem with symbolic link
4061 Details : lustre client didn't properly install dentry when re-exported
4062 to NFS or running patchless client.
4065 Frequency : NFS re-export or patchless client
4067 Description: Various nfs/patchless fixes.
4068 Details : fixes reuse disconected alias for lookup process - this fixes
4069 warning "find_exported_dentry: npd != pd",
4070 fix permission error with open files at nfs.
4071 fix apply umask when do revalidate.
4074 Frequency : occasional
4076 Description: Crash on NFS re-export node
4077 Details : calling clear_page() on the wrong pointer triggered oops in
4078 generic_mapping_read().
4081 Frequency : rarely, using O_DIRECT IO
4083 Description: unaligned directio crashes client with LASSERT
4084 Details : check for unaligned buffers before trying any requests.
4087 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
4089 Description: stale data returned from RAID cache
4090 Details : If only a small amount of IO is done to the RAID device before
4091 reading it again it is possible to get stale data from the RAID
4092 cache instead of reading it from disk.
4095 Frequency : always for sles10 kernel
4097 Description: sles10 support
4098 Details : ll_follow_link: compile fixes and using of nd_set_link
4099 under newer kernels.
4102 Frequency : depends on arch, kernel and compiler version, always on sles10
4105 Description: recursive or deep enough symlinks cause stack overflow
4106 Details : getting rid of large stack-allocated variable in
4110 Frequency : depends on hardware
4112 Description: lustre write performance loss in the SLES10 kernel
4113 Details : the performance loss is caused by using of write barriers in the
4114 ext3 code. The SLES10 kernel turns barrier support on by
4115 default. The fix is to undo that change for ldiskfs.
4117 ------------------------------------------------------------------------------
4119 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
4121 * Support for kernels:
4122 2.6.9-42.0.3EL (RHEL 4)
4123 2.6.5-7.276 (SLES 9)
4124 2.4.21-47.0.1.EL (RHEL 3)
4125 2.6.12.6 vanilla (kernel.org)
4129 Frequency : quota enabled and large files being deleted
4131 Description: releasing more than 4GB of quota at once hangs OST
4132 Details : If a user deletes more than 4GB of files on a single OST it
4133 will cause the OST to spin in an infinite loop. Release
4134 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4139 Description: statfs data retrieved from /proc may be stale or zero
4140 Details : When reading per-device statfs data from /proc, in the
4141 {kbytes,files}_{total,free,avail} files, it may appear
4142 as zero or be out of date.
4145 Frequency : systems with MD RAID1 external journal devices
4147 Description: lconf's call to blkid is confused by RAID1 journal devices
4148 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4149 of returning all block devices that match the journal UUID.
4152 Frequency : always, for aggregate stripe size over 4GB
4154 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
4155 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
4156 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
4159 Frequency : always if "lfs find" used on a local file/directory
4161 Description: "lfs find" segfaults if used on a local file/directory
4162 Details : The case where a directory component was not specified wasn't
4163 handled correctly. Handle this properly.
4166 Frequency : always on ppc64
4168 Description: the write to an ext3 filesystem mounted with mballoc got stuck
4169 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4170 perform endianness conversion.
4173 Frequency : rarely (truncate to non-zero file size after write under load)
4174 Bugzilla : 10730, 10687
4175 Description: Files padded with zeros to next 4K multiple
4176 Details : With filesystems mounted using the "extents" option (2.6 kernels)
4177 it is possible that files that are truncated to a non-zero size
4178 immediately after being written are filled with zero bytes beyond
4179 the truncated size. No file data is lost.
4181 Severity : enhancement
4183 Description: Allow recovery/failover for liblustre clients.
4184 Details : liblustre clients were unaware of failover configurations until
4187 Severity : enhancement
4189 Description: user file locks should fail when not mounting with flock option
4190 Details : Set up an error-returning stub in ll_file_operations.lock field
4191 to prevent incorrect behaviour when client is mounted without
4192 flock option. Also, set up properly f_op->flock field for
4196 Frequency : always on ia64
4198 Description: "lfs df" loops on printing out MDS statfs information
4199 Details : The obd_ioctl_data was not initialized and in some systems
4200 this caused a failure during the ioctl that did not return
4201 an error. Initialize the struct and return an error on failure.
4204 Frequency : SLES 9 only
4206 Description: Error of copying files with lustre special EAs as root
4207 Details : Client side always return success for setxattr call for lustre
4208 special xattr (currently only "trusted.lov").
4211 Frequency : rarely on clusters with both ia64+i386 clients
4213 Description: ia64+i686 clients doing shared IO on the same file may LBUG
4214 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
4215 clients are doing concurrent writes to the same file it is
4216 possible that the ia64 clients may LASSERT because the OST
4217 extent locks are not PAGE_SIZE aligned. Ensure that grown
4218 locks are always aligned on the request boundary.
4221 Frequency : specific use, occasional
4223 Description: Overwriting in use executable truncates on-disk binary image
4224 Details : If one node attempts to overwrite an executable in use by
4225 another node, we now correctly return ETXTBSY instead of
4226 truncating the file.
4228 Severity : enhancement
4230 Description: Async OSC create to avoid the blocking unnecessarily.
4231 Details : If an OST has no remaining object, system will block on the creating
4232 when need to create a new object on this OST. Now, ways use
4233 pre-created objects when available, instead of blocking on an
4234 empty osc while others are not empty. If we must block, we block
4235 for the shortest possible period of time.
4240 Description: chmod on Lustre root is propagated to other clients
4241 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
4242 invalid by the follow_mount time.
4245 Frequency : liblustre clients only
4247 Description: Race in 'instant cancel' lock handling could lead to such locks
4248 never to be granted in case of SMP MDS
4249 Details : Do not destroy not yet granted but cbpending locks in
4253 Frequency : replay/resend of open
4255 Description: non null lock assetion failure in mds_intent_policy
4256 Details : Trying to replay/resend lockless open requests resulted in
4257 mds_open() returning 0 with no lock. Now it sets a flag if
4258 a lock is going to be returned.
4260 Severity : enhancement
4262 Description: Checksum enhancements
4263 Details : New checksum enhancements allow for resending RPCs that failed
4266 Severity : enhancement
4268 Description: Tunables on number of dirty pages in cacche
4269 Details : Allow to set limit on number of dirty pages cached.
4274 Description: client crash on unmount - lock still has references
4275 Details : In some error handling cases it was possible to leak a lock
4276 reference on a client while accessing a file. This was not
4277 harmful to the client during operation, but would cause the
4278 client to crash when the filesystem is unmounted.
4281 Frequency : specific case, rare
4283 Description: ETXTBSY on mds though file not in use by client
4284 Details : ETXTBSY is no longer incorrectly returned when attempting to
4285 chmod or chown a directory that the user previously tried to
4286 execute or a currently-executing binary.
4289 Frequency : extremely rare except on liblustre-based clients
4291 Description: Lustre space not freed when files are deleted
4292 Details : Clean up open-unlinked files after client eviction. Previously
4293 the unlink was skipped and the files remained as orphans.
4298 Description: OST failure "would be an LBUG" in waiting_locks_callback()
4299 Details : In some cases it was possible to send a blocking callback to
4300 a client doing a glimpse, even though that client didn't get
4301 a lock granted. When the glimpse lock is cancelled on the OST
4302 the freed lock is left on the waiting list and corrupted the list.
4305 Frequency : all core dumps
4307 Description: Broke core dumps to lustre
4308 Details : Negative dentry may be unhashed if parent does not have UPDATE
4309 lock, but some callers, e.g. do_coredump, expect dentry to be
4310 hashed after successful create, hash it in ll_create_it.
4312 ------------------------------------------------------------------------------
4314 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
4316 * Support for kernels:
4317 2.6.9-42.0.2.EL (RHEL 4)
4318 2.6.5-7.276 (SLES 9)
4319 2.4.21-40.EL (RHEL 3)
4320 2.6.12.6 vanilla (kernel.org)
4324 Frequency : always on RHEL 3
4326 Description: Number of open files grows over time
4327 Details : The number of open files grows over time, whether or not
4328 Lustre is started. This was due to a filp leak introduced
4329 by one of our kernel patches.
4331 ------------------------------------------------------------------------------
4333 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4335 * Support for kernels:
4336 2.6.9-42.EL (RHEL 4)
4337 2.6.5-7.276 (SLES 9)
4338 2.4.21-40.EL (RHEL 3)
4339 2.6.12.6 vanilla (kernel.org)
4344 Bugzilla : 5719, 9635, 9792, 9684
4345 Description: OST (or MDS) trips assertions in (re)connection under heavy load
4346 Details : If a server is under heavy load and cannot reply to new
4347 connection requests before the client resends the (re)connect,
4348 the connection handling code can behave badly if two service
4349 threads are concurrently handing separate (re)connections from
4350 the same client. Add better locking to the connection handling
4351 code, and ensure that only a single connection will be processed
4352 for a given client UUID, even if the lock is dropped.
4354 Severity : enhancement
4356 Description: add TCP zero-copy support to kernel
4357 Details : Add support to the kernel TCP stack to allow zero-copy bulk
4358 sends if the hardware supports scatter-gather and checksumming.
4359 This allows socklnd to do client-write and server-read more
4360 efficiently and reduce CPU utilization from skbuf copying.
4363 Frequency : only if NFS exporting from client
4365 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
4366 Details : If setting attributes on a file created under NFS that had
4367 never been opened it would be possible to oops the client
4368 if the file had no objects.
4371 Frequency : always for liblustre
4373 Description: liblustre client does MDS+OSTs setattr RPC for each write
4374 Details : When doing a write from a liblustre client, the client
4375 incorrectly issued an RPC to the MDS and each OST the file was
4376 striped over in order to update the timestamps. When writing
4377 with small chunks and many clients this could overwhelm the MDS
4378 with RPCs. In all cases it would slow down the write because
4379 these RPCs are unnecessary.
4381 Severity : enhancement
4383 Description: allow number of MDS service threads to be changed at module load
4384 Details : It is now possible to change the number of MDS service threads
4385 running. Adding "options mds mds_num_threads={N}" to the MDS's
4386 /etc/modprobe.conf will set the number of threads for the next
4387 time Lustre is restarted (assuming the "mds" module is also
4388 reloaded at that time). The default number of threads will
4389 stay the same, 32 for most systems.
4394 Description: OST crash if filesystem is unformatted or corrupt
4395 Details : If an OST is started on a device that has never been formatted
4396 or if the filesystem is corrupt and cannot even mount then the
4397 error handling cleanup routines would dereference a NULL pointer.
4402 Description: NULL pointer deref in llap_from_page.
4403 Details : get_cache_page_nowait can return a page with NULL (or otherwise
4404 incorrect) mapping if the page was truncated/reclaimed while it was
4405 searched for. Check for this condition and skip such pages when
4406 doing readahead. Introduce extra check to llap_from_page() to
4407 verify page->mapping->host is non-NULL (so page is not anonymous).
4410 Frequency : Sometimes when using sys_sendfile
4412 Description: "page not covered by a lock" warnings from ll_readpage
4413 Details : sendfile called ll_readpage without right page locks present.
4414 Now we introduced ll_file_sendfile that does necessary locking
4415 around call to generic_file_sendfile() much like we do in
4419 Frequency : with certain MDS communication failures at client mount time
4421 Description: NULL pointer deref after failed client mount
4422 Details : a client connection request may delayed by the network layer
4423 and not be sent until after the PTLRPC layer has timed out the
4424 request. If the client fails the mount immediately it will try
4425 to clean up before the network times out the request. Add a
4426 reference from the request import to the obd device and delay
4427 the cleanup until the network drops the request.
4430 Frequency : occasionally during client (re)connect
4432 Description: assertion failure during client (re)connect
4433 Details : processing a client connection request may be delayed by the
4434 client or server longer than the client connect timeout. This
4435 causes the client to resend the connection request. If the
4436 original connection request is replied in this interval, the
4437 client may trip an assertion failure in ptlrpc_connect_interpret()
4438 which thought it would be the only running connect process.
4441 Frequency : only with obd_echo servers and clients that are rebooted
4443 Description: kernel BUG accessing uninitialized data structure
4444 Details : When running an obd_echo server it did not start the ping_evictor
4445 thread, and when a client was evicted an uninitialized data
4446 structure was accessed. Start the ping_evictor in the RPC
4447 service startup instead of the OBD startup.
4449 Severity : enhancement
4450 Bugzilla : 10193 (patchless)
4451 Description: Remove dependency on various unexported kernel interfaces.
4452 Details : No longer need reparent_to_init, exit_mm, exit_files,
4453 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
4456 Frequency : rare (only users of deprecated and unsupported LDAP config)
4458 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
4459 Details : LDAP apparently contains 'lustreName' attributes instead of
4460 'name'. A simple remapping of the name is sufficient.
4463 Frequency : rare (only with non-default dump_on_timeout debug enabled)
4465 Description: waiting_locks_callback trips kernel BUG if client is evicted
4466 Details : Running with the dump_on_timeout debug flag turned on makes
4467 it possible that the waiting_locks_callback() can try to dump
4468 the Lustre kernel debug logs from an interrupt handler. Defer
4469 this log dumping to the expired_lock_main() thread.
4471 Severity : enhancement
4473 Description: Support NFS exporting on 2.6 kernels.
4474 Details : Implement non-rawops metadata methods for NFS server to use without
4475 changing NFS server code.
4478 Frequency : very rare (synthetic metadata workload only)
4480 Description: two racing renames might cause an MDS thread to deadlock
4481 Details : Running the "racer" program may cause one MDS thread to rename
4482 a file from being the source of a rename to being the target of
4483 a rename at exactly the same time that another thread is doing
4484 so, and the second thread has already enqueued these locks after
4485 doing a lookup of the target and is trying to relock them in
4486 order. Ensure that we don't try to re-lock the same resource.
4489 Frequency : only very large systems with liblustre clients
4491 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
4492 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
4493 all OSTs in parallel. This allows the eviction of stale liblustre
4494 clients to proceed much faster than if they were done in series,
4495 and also offers similar improvements for other set_info RPCs.
4500 Description: excessive CPU usage during initial read phase on client
4501 Details : During the initial read phase on a client, it would agressively
4502 retry readahead on the file, consuming too much CPU and impacting
4503 performance (since 1.4.5.8). Improve the readahead algorithm
4504 to avoid this, and also improve some other common cases (read
4505 of small files in particular, where "small" is files smaller than
4506 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
4511 Description: MDS crash when receiving packet with unknown intent.
4512 Details : Do not LBUG in unknown intent case, just return -EFAULT
4514 Severity : enhancement
4515 Bugzilla : 9293, 9385
4516 Description: MDS RPCs are serialised on client. This is unnecessary for some.
4517 Details : Do not serialize getattr (non-intent version) and statfs.
4520 Frequency : occasional, when OST network is overloaded/intermittent
4522 Description: client evicted by OST after bulk IO timeout
4523 Details : If a client sends a bulk IO request (read or write) the OST
4524 may evict the client if it is unresposive to its data GET/PUT
4525 request. This is incorrect if the network is overloaded (takes
4526 too long to transfer the RPC data) or dropped the OST GET/PUT
4527 request. There is no need to evict the client at all, since
4528 the pinger and/or lock callbacks will handle this, and the
4529 client can restart the bulk request.
4532 Frequency : Always when mmapping file with no objects
4534 Description: client crashes when mmapping file with no objects
4535 Details : Check that we actually have objects in a file before doing any
4536 operations on objects in ll_vm_open, ll_vm_close and
4542 Description: Request leak when working with deleted CWD
4543 Details : Introduce advanced request refcount tracking for requests
4544 referenced from lustre intent.
4546 Severity : Enhancement
4548 Description: Cache open file handles on client.
4549 Details : MDS now will return special lock along with openhandle, if
4550 requested and client is allowed to hold openhandle, even if unused,
4551 until such a lock is revoked. Helps NFS a lot, since NFS is opening
4552 closing files for every read/write openration.
4554 Severity : Enhancement
4556 Description: Cache open negative dentries on client when possible.
4557 Details : Guard negative dentries with UPDATE lock on parent dir, drop
4558 negative dentries on lock revocation.
4563 Description: Remounting a client read-only wasn't possible with a zconf mount
4564 Details : It wasn't possible to remount a client read-only with llmount.
4566 Severity : enhancement
4567 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
4568 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
4569 MPICH is included to add Lustre-specific ADIO interfaces.
4570 This is based closely on the UFS ADIO layer and only differs
4571 in file creation, in order to allow the OST striping to be set.
4572 This is user-contributed code and not supported by CFS.
4577 Description: extended inode attributes (immutable, append-only) work improperly
4578 when 2.4 and 2.6 kernels are used on client/server or vice versa
4579 Details : Introduce kernel-independent values for these flags.
4581 Severity : enhancement
4584 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
4585 Details : Many of the /proc/ tunables can only be tuned at a megabyte
4586 granularity. Now, Fractional MB granularity is be supported,
4587 this is very useful for low memory system.
4589 Severity : enhancement
4591 Description: Getattr by fid
4592 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
4593 locks, avoids extra getattr rpc requests to MDS, allows '/' to
4594 have locks and avoids getattr rpc requests for it on every stat.
4597 Frequency : Always, for filesystems larger than 2TB
4599 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4600 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
4601 filesystems larger than 2TB to be mounted with mballoc enabled.
4604 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
4606 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4607 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
4608 filesystems larger than 2TB is mounted then the kernel will
4609 silently corrupt the start of the filesystem. CONFIG_LBD is
4610 enabled for all CFS-supported kernels, but the possibility of
4611 this happening with a modified kernel config exists.
4613 Severity : enhancement
4615 Description: add client O_DIRECT support for 2.6 kernels
4616 Details : It is now possible to do O_DIRECT reads and writes to files
4617 in the Lustre client mountpoint on 2.6 kernel clients.
4619 Severity : enhancement
4621 Description: parallel glimpse, setattr, statfs, punch, destroy requests
4622 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
4623 parallel, not waiting for response from every OST before sending
4624 a rpc to the next OST.
4629 Description: setattr vs write race when updating file timestamps
4630 Details : Client processes that update a file timestamp into the past
4631 right after writing to the file (e.g. tar) it is possible that
4632 the updated file modification time can be reset to the current
4633 time due to a race between processing the setattr and write RPC.
4635 Severity : enhancement
4637 Description: Bring 'lfs find' closer in line with regular Linux find.
4638 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
4639 -print0 options and obtains all the needed info through the lustre
4642 Severity : enhancement
4644 Description: support up to 1024 configured devices on one node
4645 Details : change obd_dev array from statically allocated to dynamically
4646 allocated structs as they are first used to reduce memory usage
4651 Description: Flush dirty partially truncated pages during truncate
4652 Details : Immediatelly flush partially truncated pages in filter_setattr,
4653 this way we completely avoid having any pages in page cache on OST
4654 and can retire ugly workarounds during writes to flush such pages.
4659 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
4660 Details : protect inode from truncation within vfs_unlink() context
4661 just take a reference before calling vfs_unlink() and release it
4662 when parent's i_sem is free.
4667 Description: last_id value checked outside lock on OST caused LASSERT failure
4668 Details : If there were multiple MDS->OST object precreate requests in
4669 flight, it was possible that the OST's last object id was checked
4670 outside a lock and incorrectly tripped an assertion. Move checks
4671 inside locks, and discard old precreate requests.
4674 Frequency : always, if extents are used on OSTs
4676 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
4677 Details : OSTs using the extents format would not zero the high 16 bits of
4678 the index physical block number. This is not a problem for any
4679 OST filesystems smaller than 16TB, and no kernels support ext3
4680 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
4681 new/modified files) and can be fixed for existing filesystems
4682 with e2fsprogs-1.39-cfs1.
4687 Description: import connection selection may be incorrect if timer wraps
4688 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
4689 import connections to be ignored if the 32-bit jiffies counter
4690 wraps. Use a 64-bit jiffies counter.
4693 Frequency : very large clusters immediately after boot
4695 Description: LNET request buffers exhausted under heavy short-term load
4696 Details : If a large number of client requests are generated on a service
4697 that has previously never seen so many requests it is possible
4698 that the request buffer growth cannot keep up with the spike in
4699 demand. Instead of dropping incoming requests, they are held in
4700 the LND until the RPC service can accept more requests.
4703 Frequency : Sometimes during replay
4705 Description: Assertion failure in ll_local_open after replay.
4706 Details : If replay happened on an open request reply before we were able
4707 to set replay handler, reply will become not swabbed tripping the
4708 assertion in ll_local_open. Now we set the handler right after
4709 recognising of open request
4712 Frequency : very rare
4714 Description: kernel reports "badness in vsnprintf"
4715 Details : Reading from the "recovery_status" /proc file in small chunks
4716 may cause a negative length in lprocfs_obd_rd_recovery_status()
4717 call to vsnprintf() (which is otherwise harmless). Exit early
4718 if there is no more space in the output buffer.
4720 Severity : enhancement
4722 Description: clear OBD RPC statistics by writing to them
4723 Details : It is now possible to clear the OBD RPC statistics by writing
4724 to the "stats" file.
4729 Description: Client mtime is not the same on different clients after utimes
4730 Details : In some cases, the client was using the utimes() syscall on
4731 a file cached on another node. The clients now validate the
4732 ctime from the MDS + OSTs to determine which one is right.
4737 Description: Inability to activate failout mode
4738 Details : lconf script incorrectly assumed that in pythong string's numeric
4739 value is used in comparisons.
4742 Frequency : always with multiple stripes per file
4744 Description: Inefficient object allocation for mutli-stripe files
4745 Details : When selecting which OSTs to stripe files over, for files with
4746 a stripe count that divides evenly into the number of OSTs,
4747 the MDS is always picking the same starting OST for each file.
4748 Return the OST selection heuristic to the original design.
4753 Description: mount failures may take full timeout to return an error
4754 Details : Under some heavy load conditions it is possible that a
4755 failed mount can wait for the full obd_timeout interval,
4756 possibly several minutes, before reporting an error.
4757 Instead return an error as soon as the status is known.
4759 Frequency : quota enabled and large files being deleted
4761 Description: releasing more than 4GB of quota at once hangs OST
4762 Details : If a user deletes more than 4GB of files on a single OST it
4763 will cause the OST to spin in an infinite loop. Release
4764 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4769 Description: statfs data retrieved from /proc may be stale or zero
4770 Details : When reading per-device statfs data from /proc, in the
4771 {kbytes,files}_{total,free,avail} files, it may appear
4772 as zero or be out of date.
4775 Frequency : systems with MD RAID1 external journal devices
4777 Description: lconf's call to blkid is confused by RAID1 journal devices
4778 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4779 of returning all block devices that match the journal UUID.
4782 Frequency : always, for aggregate stripe size over 4GB
4784 Description: assertion fails when trying to use 4GB stripe size
4785 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
4786 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
4789 Frequency : always on ppc64
4791 Description: the first write on an ext3 filesystem with mballoc got stuck
4792 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4793 perform endianness conversion.
4795 ------------------------------------------------------------------------------
4797 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4799 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
4800 INTEROPERATE with older versions automatically. Please read the
4801 user documentation before upgrading any part of a live system.
4802 * WARNING: Lustre networking configuration changes are required with
4803 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
4806 * Support for kernels:
4807 2.6.9-22.0.2.EL (RHEL 4)
4808 2.6.5-7.244 (SLES 9)
4809 2.6.12.6 vanilla (kernel.org)
4812 Severity : enhancement
4813 Bugzilla : 7981/8208
4814 Description: Introduced Lustre Networking (LNET)
4815 Details : LNET is new networking infrastructure for Lustre, it includes
4816 a reorganized network configuration mode (see the user
4817 documentation for full details) as well as support for routing
4818 between different network fabrics. Lustre Networking Devices
4819 (LNDs) for the supported network fabrics have also been
4820 created for this new infrastructure.
4822 Severity : enhancement
4823 Description: Introduced Access control lists
4824 Details : clients can set ACLs on files and directories in order to have
4825 more fine-grained permissions than the standard Unix UGO+RWX.
4826 The MDS must be started with the "-o acl" mount option.
4828 Severity : enhancement
4829 Description: Introduced filesystem quotas
4830 Details : Administrators may now establish per-user quotas on the
4833 Severity : enhancement
4835 Description: Configuration change for the XT3
4836 The PTLLND is now used to run Lustre over Portals on the XT3
4837 The configure option(s) --with-cray-portals are no longer used.
4838 Rather --with-portals=<path-to-portals-includes> is used to
4839 enable building on the XT3. In addition to enable XT3 specific
4840 features the option --enable-cray-xt3 must be used.
4845 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
4846 Details : A race between allocating a new llcd and re-getting the llcd_lock
4847 allowed another thread to grab newly-allocated llcd.
4849 Severity : enhancement
4851 Description: 2.6 OST async journal commit and locking fix to improve performance
4852 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
4853 2.6 kernels are now async as they already were in 2.4 kernels so
4854 that they can commit concurrently with the network bulk transfer.
4855 For block-allocated files the filter allocation semaphore is held
4856 to avoid filesystem fragmentation during allocation. BKL lock
4857 removed for 2.6 xattr operations where it is no longer needed.
4862 Description: lconf incorrectly determined whether two IP networks could talk
4863 Details : In some more complicated routing and multiple-network
4864 configurations, lconf will avoid trying to make a network
4865 connection to a disjoint part of the IP space. It was doing the
4866 math incorrectly for one set of cases.
4871 Description: Fix for potential infinite loop processing records in an llog.
4872 Details : If an llog record is corrupted/zeroed, it is possible to loop
4873 forever in llog_process(). Validate the llog record length
4874 and skip the remainder of the block on error.
4877 Frequency : occasional (liblustre only)
4879 Description: liblustre could not open files whose last component is a symlink
4880 Details : sysio_path_walk() would incorrectly pass the open intent to
4881 intermediate path components.
4884 Frequency : rare (liblustre only with non-standard tuning)
4885 Bugzilla : 7201 (7350)
4886 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4887 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4888 after it has been released, only do this for referenced locks
4890 Severity : enhancement
4892 Description: specifying an (invalid) directory default stripe_size of -1
4893 would reset the directory default striping
4894 Details : stripe_size -1 was used internally to signal directory stripe
4895 removal, now use "all default" to signal dir stripe removal
4896 as a directory striping of "all default" is not useful
4899 Frequency : common for large clusters running liblustre clients
4901 Description: doing an ls when liblustre clients are running is slow
4902 Details : sending a glimpse AST to a liblustre client waits for every AST
4903 to time out, as liblustre clients will not respond. Since they
4904 cannot cache data we refresh the OST lock LVB from disk instead.
4906 Severity : enhancement
4908 Description: doing an ls at the same time as file IO can be slow
4909 Details : enqueue and other "small" requests can be blocked behind many
4910 large IO requests. Create a new OST IO portal for non-IO
4911 requests so they can be processed faster.
4914 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4916 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4917 created file with mode 000
4918 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4919 allow an ftruncate() as a non-root user to a file with mode 000.
4920 Lustre now respects this flag to disable mode checks when
4921 truncating a file owned by the user
4924 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4927 Description: Revoking locks from clients that went dead or catatonic might take
4929 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4930 cancellation of such locks instant on servers without waiting for
4931 any reply from clients. Clients drops these locks when cancel
4932 notification from server is received without replying.
4935 Frequency : liblustre-only, when liblustre client dies or becomes busy
4937 Description: Doing ls on Linux clients can take a long time with active
4939 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4940 granting such locks to it in the first place if possible. Locks
4941 are taken by proxy on the OST during the read or write and
4942 dropped immediately afterward. Add connect flags handling, do
4943 not grant locks to liblustre clients for glimpse ASTs.
4945 Severity : enhancement
4947 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4948 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4949 sequential reads and behaves badly with applications doing random
4950 reads. Improve it by reading ahead at least read region, and
4951 avoiding excessive large RPC for small reads.
4953 Severity : enhancement
4955 Description: Creating more than 1000 files for a single job may cause a load
4956 imbalance on the OSTs if there are also a large number of OSTs.
4957 Details : qos_prep_create() uses an OST index reseed value that is an
4958 even multiple of the number of available OSTs so that if the
4959 reseed happens in the middle of the object allocation it will
4960 still utilize the OSTs as uniformly as possible.
4965 Description: OST or MDS may oops in ping_evictor_main()
4966 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
4967 but doesn't restart at beginning of obd_exports_timed list
4970 Severity : enhancement
4972 Description: improve by-nid export eviction on the MDS and OST
4973 Details : allow multiple exports with the same NID to be evicted at one
4974 time without re-searching the exports list.
4977 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
4979 Description: MDS may oops in groups_free()
4980 Details : in rare race conditions a newly allocated group_info struct is
4981 freed again, and this can be NULL. The 2.4 compatibility code
4982 for groups_free() checked for a NULL pointer, but 2.6 did not.
4985 Frequency : common for liblustre clients doing little filesystem IO
4986 Bugzilla : 9352, 7313
4987 Description: server may evict liblustre clients accessing contended locks
4988 Details : if a client is granted a lock or receives a completion AST
4989 with a blocking AST already set it would not reply to the AST
4990 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
4993 Frequency : lfs setstripe, only systems with more than 160 OSTs
4995 Description: unable to set striping with a starting offset beyond OST 160
4996 Details : llapi_create_file() incorrectly limited the starting stripe
4997 index to the maximum single-file stripe count.
5000 Frequency : LDAP users only
5002 Description: lconf did not handle in-kernel recovery with LDAP properly
5003 Details : lconf/LustreDB get_refs() is searching the wrong namespace
5005 Severity : enhancement
5007 Description: bind OST threads to NUMA nodes to improve performance
5008 Details : all OST threads are uniformly bound to CPUs on a single NUMA
5009 node and do their allocations there to localize memory access
5011 Severity : enhancement
5013 Description: llmount can determine client NID directly from Myrinet (GM)
5014 Details : the client NID code from gmnalnid was moved directly into
5015 llmount, removing the need to use this or specifying the
5016 client NID explicitly when mounting GM clients with zeroconf
5019 Frequency : if client is started with down MDS
5021 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
5022 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
5023 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
5024 ptlrpc_import_delayed_req() because we didn't check if the
5025 request was interrupted, and we also didn't break out of the
5026 event loop if there was no timeout
5031 Description: data loss during non-page-aligned writes to a single file from
5032 both multiple nodes and multiple threads on one node at same time
5033 Details : updates to KMS and lsm weren't protected by common lock. Resulting
5034 inconsistency led to false short-reads, that were cached and later
5035 used by ->prepare_write() to fill in partially written page,
5036 leading to data loss.
5039 Frequency : always, if lconf --abort_recovery used
5041 Description: lconf --abort_recovery fails with 'Operation not supported'
5042 Details : lconf was attempting to abort recovery on the MDT device and not
5045 Severity : enhancement
5047 Description: remove cleanup logs
5048 Details : replace lconf-generated cleanup logs with lustre internal
5049 cleanup routines. Eliminates the need for client-cleanup and
5052 Severity : enhancement
5054 Description: add support for EAs (user and system) on lustre filesystems
5055 Details : it is now possible to store extended attributes in the Lustre
5056 client filesystem, and with the user_xattr mount option it
5057 is possible to allow users to store EAs on their files also
5059 Severity : enhancement
5061 Description: Add possibility (config option) to show minimal available OST free
5063 Details : When compiled with --enable-mindf configure option, statfs(2)
5064 (and so, df) will return least minimal free space available from
5065 all OSTs as amount of free space on FS, instead of summary of
5066 free spaces of all OSTs.
5068 Severity : enhancement
5070 Description: do not expand extent locks acquired on OST-side
5071 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
5072 by server: they are not cached anyway.
5075 Frequency : when mmap is used/binaries executed from Lustre
5077 Description: Unmmap pages before throwing them away from read cache.
5078 Details : llap_shrink cache now attempts to unmap pages before discarding
5079 them (if unmapping failed - do not discard). SLES9 kernel has
5080 extra checks that trigger if this unmapping is not done first.
5085 Description: lconf didn't resolve symlinks before checking to see whether a
5086 given mountpoint was already in use
5089 Frequency : when migrating failover services
5090 Bugzilla : 6395, 9514
5091 Description: When migrating a subset of services from a node (e.g. failback
5092 from a failover service node) the remaining services would
5093 time out and evict clients.
5094 Details : lconf --force (implied by --failover) sets the global obd_timeout
5095 to 5 seconds in order to quickly disconnect, but this caused
5096 other RPCs to time out too quickly. Do not change the global
5097 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
5099 Severity : enhancement
5100 Frequency : if MDS is started with down OST
5101 Bugzilla : 9439,5706
5102 Description: Allow startup/shutdown of an MDS without depending on the
5103 availability of the OSTs.
5104 Details : Asynchronously call mds_lov_synchronize during MDS startup.
5105 Add appropriate locking and lov-osc refcounts for safe
5106 cleaning. Add osc abort_inflight calls in case the
5107 synchronize never started.
5110 Frequency : occasional (Cray XT3 only)
5112 Description: root not authorized to access files in CRAY_PORTALS environment
5113 Details : The client process capabilities were not honoured on the MDS in
5114 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
5115 been accessed by an authorized user then root was able to access
5116 the file on the local client also. The root user capabilities
5117 are now allowed on the MDS, as this environment has secure UID.
5120 Frequency : occasional
5122 Description: ldiskfs "too long searching" message happens too often
5123 Details : A debugging message (otherwise harmless) prints too often on
5124 the OST console. This has been reduced to only happen when
5125 there are fragmentation problems on the filesystem.
5130 Description: Division by zero in statfs when all OSCs are inactive
5131 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
5132 lov_statfs divides by value returned by lov_get_stripecnt().
5136 Bugzilla : 9489, 3273
5137 Description: First write from each client to each OST was only 4kB in size,
5138 to initialize client writeback cache, which caused sub-optimal
5139 RPCs and poor layout on disk for the first written file.
5140 Details : Clients now request an initial cache grant at (re)connect time
5141 and so that they can start streaming writes to the cache right
5142 away and always do full-sized RPCs if there is enough data.
5143 If the OST is rebooted the client also re-establishes its grant
5144 so that client cached writes will be honoured under the grant.
5149 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
5150 Details : Now I/O RPCs go to different portal number and (presumably) fast
5151 lock requests (and glimses) and other RPCs get their own service
5152 threads pool that should be able to service those RPCs
5155 Severity : enhancement
5157 Description: Ability to exchange lustre version between client and servers and
5158 issue warnings at client side if client is too old. Also for
5159 liblustre clients there is ability to refuse connection of too old
5161 Details : New 'version' field is added to connect data structure that is
5162 filled with version info. That info is later checked by server and
5166 Frequency : rare, liblustre only.
5167 Bugzilla : 9296, 9581
5168 Description: Two simultaneous writes from liblustre at offset within same page
5169 might proceed at the same time overwriting eachother with stale
5171 Details : I/O lock withing llu_file_prwv was released too early, before data
5172 actually was hitting the wire. Extended lock-holding time until
5173 server acknowledges receiving data.
5176 Frequency : extremely rare. Never observed in practice.
5178 Description: avoid generating lustre_handle cookie of 0.
5179 Details : class_handle_hash() generates handle cookies by incrementing
5180 global counter, and can hit 0 occasionaly (this is unlikely, but
5181 not impossible, because initial value of cookie counter is
5182 selected randonly). Value of 0 is used as a sentinel meaning
5183 "unassigned handle" --- avoid it. Also coalesce two critical
5184 sections in this function into one.
5186 Severity : enhancement
5188 Description: allow liblustre clients to delegate truncate locking to OST
5189 Details : To avoid overhead of locking, liblustre client instructs OST to
5190 take extent lock in ost_punch() on client's behalf. New connection
5191 flag is added to handle backward compatibility.
5193 Severity : enhancement
5194 Bugzilla : 4928, 7341, 9758
5195 Description: allow number of OST service threads to be specified
5196 Details : a module parameter allows the number of OST service threads
5197 to be specified via "options ost ost_num_threads={N}" in the
5198 OSS's /etc/modules.conf or /etc/modprobe.conf.
5202 Bugzilla : 6146, 9635, 9895
5203 Description: servers crash with bad pointer in target_handle_connect()
5204 Details : In rare cases when a client is reconnecting it was possible that
5205 the connection request was the last reference for that export.
5206 We would temporarily drop the export reference and get a new
5207 one, but this may have been the last reference and the export
5208 was just destroyed. Get new reference before dropping old one.
5210 Severity : enhancement
5211 Frequency : if client is started with failover MDS
5213 Description: Allow multiple MDS hostnames in the mount command
5214 Details : Try to read the configuration from all specified MDS
5215 hostnames during a client mount in case the "primary"
5218 Severity : enhancement
5220 Description: Stop sending data to evicted clients as soon as possible.
5221 Details : Check if the client we are about to send or are sending data to
5222 was evicted already. (Check is done every second of waiting,
5223 for which l_wait_event interface was extended to allow checking
5224 of exit condition at specified intervals).
5227 Frequency : rare, normally only when NFS exporting is done from client
5229 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
5231 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
5234 Frequency : always, if srand() is called before liblustre initialization
5236 Description: Liblustre uses system PRNG disturbing its usage by user application
5237 Details : Introduce internal to lustre fast and high-quality PRNG for
5238 lustre usage and make liblustre and some other places in generic
5239 lustre code to use it.
5241 Severity : enhancement
5242 Bugzilla : 9477, 9557, 9870
5243 Description: Verify that the MDS configuration logs are updated when xml is
5244 Details : Check if the .xml configuration logs are newer than the config
5245 logs stored on the MDS and report an error if this is the case.
5246 Request --write-conf, or allow starting with --old_conf.
5248 Severity : enhancement
5250 Description: Handle symlinks in the path when checking if Lustre is mounted.
5251 Details : Resolve intermediate symlinks when checking if a client has
5252 mounted a filesystem to avoid duplicate client mounts.
5257 Description: lconf can hit an error exception but still return success.
5258 Details : The lconf command catches the Command error exception at the top
5259 level script context and will exit with the associated exit
5260 status, but doesn't ensure that this exit status is non-zero.
5265 Description: failure of ptlrpc thread startup can cause oops
5266 Details : Starting a ptlrpc service thread can fail if there are a large
5267 number of threads or the server memory is very fragmented.
5268 Handle this without oopsing.
5271 Frequency : always, only if liblustre and non-default acceptor port was used
5273 Description: liblustre cannot connect to servers with non-default acceptor port
5274 Details : tcpnal_set_default_params() was not called and was therefore
5275 ignoring the environment varaible TCPNAL_PORT, as well as other
5276 TCPNAL_ environment variables
5281 Description: two objects could be created on the same OST for a single file
5282 Details : If an OST is down, in some cases it was possible to create two
5283 objects on a single OST for a single file. No problems other
5284 than potential performance impact and spurious error messages.
5288 Bugzilla : 5681, 9562
5289 Description: Client may oops in ll_unhash_aliases
5290 Details : Client dcache may become inconsistent in race condition.
5291 In some cases "getcwd" can fail if the current directory is
5297 Description: Inode refcounting problems in NFS export code
5298 Details : link_raw functions used to call d_instantiate without obtaining
5299 extra inode reference first.
5303 Bugzilla : 9942, 9903
5304 Description: Referencing freed requests leading to crash, memleaks with NFS.
5305 Details : We used to require that call to ll_revalidate_it was always
5306 followed by ll_lookup_it. Also with revalidate_special() it is
5307 possible to call ll_revalidate_it() twice for the same dentry
5308 even if first occurence returned success. This fix changes semantic
5309 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
5310 reference on a request referred from the intent.
5311 ll_intent_release() then releases such a request.
5314 Frequency : rare, normally benchmark loads only
5316 Description: unlinked inodes were kept in memory on the client
5317 Details : If a client is repeatedly creating and unlinking files it
5318 can accumulate a lot of stale inodes in the inode slab cache.
5319 If there is no other client load running this can cause the
5320 client node to run out of memory. Instead flush old inodes
5321 from client cache that have the same inode number as a new inode.
5324 Frequency : SLES9 2.6.5 kernel and long filenames only
5325 Bugzilla : 9969, 10379
5326 Description: utime reports stale NFS file handle
5327 Details : SLES9 uses out-of-dentry names in some cases, which confused
5328 the lustre dentry revalidation. Change it to always use the
5332 Frequency : rare, unless heavy write-truncate concurrency is continuous
5333 Bugzilla : 4180, 6984, 7171, 9963, 9331
5334 Description: OST becomes very slow and/or deadlocked during object unlink
5335 Details : filter_destroy() was holding onto the parent directory lock
5336 while truncating+unlinking objects. For very large objects this
5337 may block other threads for a long time and slow overall OST
5338 responsiveness. It may also be possible to get a lock ordering
5339 deadlock in this case, or run out of journal credits because of
5340 the combined truncate+unlink. Solution is to do object truncate
5341 first in one transaction without parent lock, and then do the
5342 final unlink in a new transaction with the parent lock. This
5343 reduces the lock hold time dramatically.
5346 Frequency : rare, 2.4 kernels only
5348 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
5349 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
5350 before unmounting their filesystem in order to avoid deadlock.
5351 The kernel_locked() function in 2.4 kernels only checks whether
5352 the kernel lock is held, not whether it is this process that is
5353 holding it as 2.6 kernels do.
5358 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
5359 Details : The client ptlrpc code may be trying to reconnect to a down
5360 server before a previous connection attempt has timed out.
5361 Increase the reconnect interval to be longer than the connection
5362 timeout interval to avoid sending duplicate connections to
5366 Frequency : echo_client brw_test command
5368 Description: fix echo_client to work with OST preallocated code
5369 Details : OST preallocation code (5137) didn't take echo_client IO path
5370 into account: echo_client calls filter methods outside of any
5371 OST thread and, hence, there is no per-thread preallocated
5372 pages and buffers to use. Solution: hijack pga pages for IO. As
5373 a byproduct, this avoids unnecessary data copying.
5377 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
5378 Description: Client can oops in mdc_commit_close() after open replay
5379 Details : It was possible for the MDS to return an open request with no
5380 transaction number in mds_finish_transno() if the client was
5381 evicted, but without actually returning an error. Clients
5382 would later try to replay that open and may trip an assertion
5383 Simplify the client close codepath, and always return an error
5384 from the MDS in case the open is not successful.
5387 Frequency : rare, 2.6 OSTs only
5389 Description: OST may deadlock under high load on fragmented files
5390 Details : If there was a heavy load and highly-fragmented OST filesystems
5391 it was possible to have all the OST threads deadlock waiting on
5392 allocation of biovecs, because the biovecs were not released
5393 until the entire RPC IO was completed. Instead, release biovecs
5394 as soon as they are complete to ensure forward IO progress.
5396 Severity : enhancement
5398 Description: Support for specifying external journal device at mount
5399 Details : If an OST or MDS device is formatted with an external journal
5400 device, this device major/minor is stored in the ext3 superblock
5401 and may not be valid for failover. Allow detecting and
5402 specifying the external journal at mount time.
5407 Description: Mounting an MDS with pending unlinked files may cause oops
5408 Details : target_finish_recovery() calls mds_postrecov() which returned
5409 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
5410 considers this an error and immediately begins cleaning up the
5411 lov, just after starting the mds_lov process
5413 Severity : enhancement
5415 Description: Implement 'lfs df' to report actual free space on per-OST basis
5416 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
5417 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
5418 usage of objects; '-h' to report in human readable format.
5420 ------------------------------------------------------------------------------
5422 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
5429 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
5430 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
5431 at startup during orphan unlinking. Instead, skip buddy update
5432 before setup, as it will be regenerated after recovery is complete.
5437 Description: If an OST is inactive, its locks might reference stale inodes.
5438 Details : lov_change_cbdata() must iterate over all namespaces, even if
5439 they are inactive to clear inode references from the lock.
5441 Severity : enhancement
5442 Frequency : occasional, if non-standard max_dirty_mb used
5444 Description: Client will block write RPCs if not enough grant
5445 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
5446 then the client will block writes while waiting for another RPC
5447 to complete instead of consuming its dirty limit. With change
5448 we get improved performance when max_dirty_mb is small.
5450 Severity : enhancement
5451 Bugzilla : 3389, 6253
5452 Description: Add support for supplementary groups on the MDS.
5453 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
5454 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
5455 lookups for user supplementary groups into a cache.
5459 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
5460 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
5461 for liblustre O_EXCL is enforced by MDS.
5465 Description: readdir never returns NULL in liblustre.
5466 Details : Corrected llu_iop_getdirentries logic, to return offset of next
5467 dentry in struct dirent.
5471 Frequency : liblustre only, depends on application IO pattern
5472 Description: liblustre clients evicted if not contacting servers
5473 Details : Don't put liblustre clients into the ping_evictor list, so
5474 they will not be evicted by the pinger ever.
5476 Severity : enhancement
5478 Description: Add ability to evict clients by NID from MDS.
5479 Details : By echoing "nid:$NID" string into
5480 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
5481 $NID would be instantly evicted from this MDS and from all active
5482 OSTs connected to it.
5486 Description: Do not query file size twice, somewhat slowing stat(2) calls.
5487 Details : lookup_it_finish() used to query file size from OSTs that was not
5492 Description: service threads change working directory to that of init
5493 Details : Starting lustre service threads may pin the working directory
5494 of the parent thread, making that filesystem busy. Threads
5495 now change to the working directory of init to avoid this.
5499 Frequency : during shutdown only
5500 Description: shutdown with a failed MDS or OST can cause unmount to hang
5501 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
5506 Frequency : 2.6 only
5507 Description: chmod/chown may include an extra supplementary group
5508 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
5509 supplementary group and if none is specified this is used.
5512 Bugzilla : 5479 (6816)
5514 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
5515 Details : If lookup is in progress on a file that is unlinked we might try
5516 to revalidate the inode and fail in revalidate after lookup is
5517 complete and ll_file_open() enqueues the open again but
5518 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
5521 Frequency : always, if lconf --abort_recovery used
5523 Description: lconf --abort_recovery fails with 'Operation not supported'
5524 Details : lconf was attempting to abort recovery on the MDT device and not
5527 ------------------------------------------------------------------------------
5529 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
5534 Frequency : rare (only unsupported configurations with a node running as an
5536 Bugzilla : 6514, 5137
5537 Description: Mounting a Lustre file system on a node running as an OST could
5539 Details : OSTs now preallocates memory needed to write out data at
5540 startup, instead of when needed, to avoid having to
5541 allocate memory in possibly low memory situations.
5542 Specifically, if the file system is mounted on on OST,
5543 memory pressure could force it to try to write out data,
5544 which it needed to allocate memory to do. Due to the low
5545 memory, it would be unable to do so and the node would
5546 become unresponsive.
5548 Severity : enhancement
5550 Description: Addition of lconf --service command line option
5551 Details : lconf now accepts a '--service <arg>' option, which is
5552 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
5554 Severity : enhancement
5556 Description: Failover mode is now the default for OSTs.
5557 Details : By default, OSTs will now run in failover mode. To return to
5558 the old behaviour, add '--failout' to the lmc line for OSTs.
5560 Severity : enhancement
5562 Description: Health checks are now provided for MDS and OSTs
5563 Details : Additional detailed health check information on MSD and OSTs
5564 is now provided through the procfs health_check value.
5567 Frequency : occasional, depends on IO load
5569 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
5570 after numerous create/delete cycles
5571 Details : The ext3 inode allocation policy would not allocate new inodes
5572 very well on the OSTs because there are no new directories
5573 being created. Instead we look for groups with free space if
5574 the parent directories are nearly full.
5579 Description: Network or server problems during mount may cause partially
5580 mounted clients instead of returning an error.
5581 Details : The config llog parsing code may overwrite the error return
5582 code during mount error handling, returning success instead
5588 Description: MDS can fail to allocate large reply buffers
5589 Details : After long uptimes the MDS can fail to allocate large reply
5590 buffers (e.g. zconf client mount config records) due to memory
5591 fragmentation or consumption by the buffer cache. Preallocate
5592 some large reply buffers so that these replies can be sent even
5593 under memory pressure.
5597 Frequency : rare (liblustre)
5598 Description: fsx running with liblustre complained that using truncate() to
5599 extend the file doesn't work. This patch corrects that issue.
5600 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
5601 ATTR_SIZE and lsm use in llu_setattr_raw.
5605 Frequency : rare, only 2.6 kernels
5606 Description: Unusual file access patterns on the MDS may result in inode
5607 data being lost in very rare circumstances.
5608 Details : Bad interaction between the ea-in-inode patch and the "no-read"
5609 code in the 2.6 kernel caused the inode and/or EA data not to
5610 be read from disk, causing single-file corruption.
5614 Frequency : rare, only 2.6 filesystems using extents
5615 Description: Heavy concurrent write and delete load may cause data corruption.
5616 Details : It was possible under high-load situations to have an extent
5617 metadata block in the block device cache from a just-unlinked
5618 file overwrite a newly-allocated data block. We now unmap any
5619 metadata buffers that alias just-allocated data blocks.
5623 Frequency : filesystems with default stripe_count larger than 77
5624 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
5625 Details : lconf specifies an inode size of 4096 bytes when the default
5626 stripe_count is larger than 77. This conflicts with the default
5627 inode density of 1 per 4096 bytes. Allocate smaller inodes in
5628 this case to avoid pinning too much memory for large EAs.
5630 ------------------------------------------------------------------------------
5632 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
5637 Frequency : rare (extremely heavy IO load with hundreds of clients)
5639 Description: Client is evicted, gets IO error writing to file
5640 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
5641 caused clients to be evicted for AST timeouts. The fixes in
5642 bug 5192 mean we no longer need to have such short AST timeouts
5643 so ldlm_timeout has been increased.
5646 Frequency : occasional during --force or --failover shutdown under load
5647 Bugzilla : 5949, 4834
5648 Description: Server oops/LBUG if stopped with --force or --failover under load
5649 Details : a collection of import/export refcount and cleanup ordering
5650 issues fixed for safer force cleanup
5653 Frequency : only filesystems larger than 120 OSTs
5654 Bugzilla : 5990, 6223
5655 Description: lfs getstripe would oops on a very large filesystem
5656 Details : lov_getconfig used kfree on vmalloc'd memory
5659 Frequency : only filesystems exporting via NFS to Solaris 10 clients
5660 Bugzilla : 6242, 6243
5661 Description: reading from files that had been truncated to a non-zero size
5662 but never opened returned no data
5663 Details : ll_file_read() reads zeros from no-object files to EOF
5668 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
5669 Details : There's little harm in aborting MDS/OSS recovery and letting it
5670 try again, so I removed the LASSERT and return an error instead.
5672 Severity : enhancement
5674 Description: New debugging infrastructure for tracking down data corruption
5675 Details : The I/O checksum code was replaced to: (a) control it at runtime,
5676 (b) cover more of the client-side code path, and (c) try to narrow
5677 down where problems occurred
5681 Bugzilla : 3819, 4364, 4397, 6313
5682 Description: Racing close and eviction MDS could cause assertion in mds_close
5683 Details : It was possible to get multiple mfd references during close and
5684 client eviction, leading to one thread referencing a freed mfd.
5686 Severity: : enhancement
5687 Bugzilla : 3262, 6359
5688 Description: Attempts to reconnect to servers are now more aggressive.
5689 Details : This builds on the enhanced upcall-less recovery that was added
5690 in 1.4.2. When trying to reconnect to servers, clients will
5691 now try each server in the failover group every 10 seconds. By
5692 default, clients would previously try one server every 25 seconds.
5697 Description: After recovery, certain operations trigger a failed
5698 assertion on a client.
5699 Details : Failing over an mds, using lconf -d --failover, while a
5700 client was doing a readdir() call would cause the client to
5701 LBUG after recovery completed and the readdir() was resent.
5703 Severity : enhancement
5705 Description: Default groups are now added by lconf
5706 Details : You can now run lconf --group <servicename> without having to
5707 manually add groups with lmc.
5710 Frequency : occasional
5712 Description: Nodes with an elan id of 0 trigger a failed assertion
5715 Frequency : always when accessing e.g. tty/console device nodes
5717 Description: tty and some other devices nodes cannot be used on lustre
5718 Details : file's private_data field is used by device data and lustre
5719 values in there got lost. New field was added to struct file to
5720 store fs-specific private data.
5723 Frequency : when exporting Lustre via NFS
5725 Description: NFSD failed occasionally when looking up a path component
5726 Details : NFSD is looking up ".." which was broken in ext3 directories
5727 that had grown large enough to become hashed.
5730 Frequency : Clusters with multiple interfaces not on the same subnet
5732 Description: Nodes will repeatedly try to reconnect to an interface which it
5733 cannot reach and report an error to the log.
5734 Details : Extra peer list entries will be created by lconf with some peers
5735 unreachable. lconf now validates the peer before adding it.
5738 Frequency : Only if a default stripe is set on the filesystem root.
5740 Description: Setting a default stripe on the filesystem root prevented the
5741 filesystem from being remounted.
5742 Details : The client was sending extra request flags in the root getattr
5743 request and did not allocate a reply buffer for the dir EA.
5746 Frequency : occasional, higher if lots of files are accessed by one client
5747 Bugzilla : 6159, 6097
5748 Description: Client trips assertion regarding lsm mismatch/magic
5749 Details : While revalidating inodes the VFS looks up inodes with ifind()
5750 and in rare cases can find an inode that is being freed.
5751 The ll_test_inode() code will free the lsm during ifind()
5752 when it finds an existing inode and then the VFS later attaches
5753 this free lsm to a new inode.
5757 Bugzilla : 6422, 7030
5758 Description: MDS deadlock between mkdir and client eviction
5759 Details : Creating a new file via mkdir or mknod (starting a transaction
5760 and getting the ns lock) can deadlock with client eviction
5761 (gets ns lock and trying to finish a synchronous transaction).
5764 Frequency : occasional
5765 Description: While starting a server, the fsfilt_ext3 module could not be
5767 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
5768 kernels. Previously, lconf would still use the ext3 name
5769 when trying to load modules. Now, it will correctly use
5770 ext3 on 2.4 and ldiskfs on 2.6.
5772 Severity : enhancement
5773 Description: The default stripe count has been changed to 1
5774 Details : The interpretation of the default stripe count (0, to lfs
5775 or lmc) has been changed to mean striping across a single
5776 OST, rather than all available. For general usage we have
5777 found a stripe count of 1 or 2 works best.
5779 Severity : enhancement
5780 Description: Add support for compiling against Cray portals.
5781 Details : Conditional compiling for some areas that are different
5785 Frequency : occasional
5786 Bugzilla : 6409, 6834
5787 Description: Creating files with an explicit stripe count may lead to
5788 a failed assertion on the MDS
5789 Details : If some OSTs are full or unavailable, creating files may
5790 trigger a failed assertion on the MDS. Now, Lustre will
5791 try to use other servers or return an error to the
5795 Frequency : occasional
5797 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
5798 caused lock timeout during readahead (since 1.4.2).
5799 Details : Processes doing readahead might match a lock that hasn't been
5800 granted yet if there are overlapping and conflicting lock
5801 requests. The readahead process waits on ungranted lock
5802 (original lock is CBPENDING), while OST waits for that process
5803 to cancel CBPENDING read lock and eventually evicts client.
5805 Severity : enhancement
5807 Description: Initial enabling of flock support for clients
5808 Details : Implements fcntl advisory locking and file status functions.
5809 This feature is provided as an optional mount flag (default
5810 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
5811 locking are implemented yet, and those that are are not guaranteed
5812 to be completely correct in production environments.
5813 mount -t lustre -o [flock|noflock] ...
5816 Frequency : occasional
5818 Description: OSTs running 2.4 kernels but with extents enabled might trip an
5819 assertion in the ext3 JBD (journaling) layer.
5820 Details : The b_committed_data struct is protected by the big kernel lock
5821 in 2.4 kernels, serializing journal_commit_transaction() and
5822 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
5823 access to this struct. In 2.6 kernels there is finer grained
5824 locking to improve SMP performance of the JBD layer.
5828 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
5833 Description: merge and cleanup kernel patches.
5836 -----------------------------------------------------------------------------
5838 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
5840 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
5841 versions of Lustre. Please update all servers and clients to
5842 version 1.4.2 or later at the same time. You must also run
5843 "lconf --write-conf {config}.xml" on the MDS while it is stopped
5844 to update the configuration logs.
5846 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
5847 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
5848 - swab llog records of type '0' so we get proper header size/idx (5861)
5849 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
5850 - fix rename of one directory over another leaking an inode (5953)
5851 - avoid SetPageDirty on 2.6 (5981)
5852 - don't re-add just-being-destroyed locks to the waiting list (5653)
5853 - when creating new directories, inherit the parent's custom
5854 striping settings if present parent (3048)
5855 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
5856 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
5857 - don't hold client locks on temporary worklist from l_lru (5666)
5858 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
5859 - automatically evict dead clients (5921)
5860 - Update file size properly in create+truncate+fstat case (6196)
5861 - Do not unhash mountpoint dentries, do not allow removal of
5863 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
5864 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
5865 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
5866 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
5867 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
5868 - Correctly handle reads of files with no objects (6243)
5869 - lctl recover will also mark a device active if deactivate used (5933)
5871 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
5872 - allow --write-conf on an MDS with different nettype than client (5619)
5873 - don't write config llogs to MDS for mounts not from that MDS (5617)
5874 - lconf should create multiple TCP connections from a client (5201)
5875 - init scripts are now turned off by default; run chkconfig --on
5876 lustre and chkconfig --on lustrefs to use them
5877 - upcalls are no longer needed for clients to recover to failover
5879 - add --abort-recovery option to lconf to abort recovery on device
5881 - add support for an arbitrary number of OSTs (3026)
5882 - Quota support protocol changes.
5883 - forward compatibility changes to wire structs (6007)
5884 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5885 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5886 - improved "lustre status" script
5887 - initialize blocksize for non-regular files (6062)
5888 - added --disable-server and --disable-client configure options (5782)
5889 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5890 - Vanilla 2.4.29 support
5891 - increase maximum number of obd devices to 520 (6242)
5892 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5893 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5894 SLES 9 2.6.5 kernels
5895 - limit stripes per file to 160 (the maximum EA size) (6093)
5897 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5900 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5901 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5902 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5903 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5904 - small ext3 extents cleanups and fixes (5733)
5905 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5906 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5907 - avoid needless client->OST connect, fix handle mismatch (5317)
5908 - fix DLM error path that led to out-of-sync client, long delays (5779)
5909 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5910 - fix several locking issues related to i_size (5492,5624,5654,5672)
5911 - don't move pending lock onto export if it is already evicted (5683)
5912 - fix kernel oops when creating .foo in unlinked directory (5548)
5913 - fix deadlock in obdfilter statistics vs. object create (5811)
5914 - use time_{before,after} to avoid timer jiffies wrap (5882)
5915 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5916 - Do not leak request if server was not able to process it (5154)
5917 - If mds_open unable to find parent dir, make that negative lookup(5154)
5918 - don't create new directories with extent-mapping (5909, 5936)
5920 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5921 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5922 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5923 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5924 - support for RHEL4 kernel on i686 (5773)
5925 - provide error messages when incompatible logs are encountered (5898)
5927 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5928 * version 1.4.0.10 (1.4.1 release candidate 1)
5930 - don't keep a lock reference when lock is not granted (4238)
5931 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5932 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5933 - drop import inflight refcount on signal_completed_replay error (5255)
5934 - unlock page after async write error during send (3677)
5935 - handle missing objects in filter_preprw_read properly (5265)
5936 - no transno return for symlink open, don't save no-trasno open (3440)
5937 - don't try to complete elan receive that already failed (4012)
5938 - free RPC server reply state on error (5406)
5939 - clean up thread from ptlrpc_start_thread() on error (5160)
5940 - readahead could read extra page into cache that wasn't ejected (5388)
5941 - prevent races in class_attach/setup/cleanup/detach (5260)
5942 - don't dereference de->d_inode after l_dput of de (5458)
5943 - use "int" for stripe value returned from lock_to_stripe (5544)
5944 - mballoc allocation and error-checking fixes in 2.6 (5504)
5945 - block device patches to fix I/O request sizes in 2.6 (5482)
5946 - look up hostnames for IB nals (5602)
5947 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5948 - don't start multiple acceptors for the same port (5277)
5949 - fix incorrect LASSERT in mds_getattr_name (5635)
5950 - export a proc file for general "ping" checking (5628)
5951 - fix "lfs check" to not block when the MDS is down (5628)
5953 - service request history (4965)
5954 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5955 - create an "evict_client" /proc entry on OSTs, like the MDS has
5956 - fix mount usage message, return errors per mount(8) (5168)
5957 - change grep [] to grep "[]" in tests so they work in more UMLs
5958 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5959 - remove ancient LOV_MAGIC_V0 EA support (5047)
5960 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5961 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5962 - allow readahead allocations to fail when low on memory (5383)
5963 - mmap locking landed again, after considerable improvement (2828)
5964 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
5966 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
5969 - send OST transaction number in read/write reply to free req (4966)
5970 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5971 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5972 - the watchdog thread now runs as interruptible (5246)
5973 - flock/lockf fixes (but it's still disabled, pending 5135)
5974 - don't use EXT3 constants in llite code (5094)
5975 - memory shortage at startup could cause assertion (5176)
5977 - reorganization of lov code
5978 - single portals codebase
5980 - add extents/mballoc support (5025)
5981 - direct I/O reads in the obdfilter (4048)
5982 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
5984 tbd Cluster File Systems, Inc. <info@clusterfs.com>
5987 - send OST transaction number in read/write reply to free req (4966)
5988 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5989 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5990 - the watchdog thread now runs as interruptible (5246)
5991 - handle missing objects in filter_preprw_read properly (5265)
5992 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5993 - drop import inflight refcount on signal_completed_replay error (5255)
5994 - unlock page after async write error during send (3677)
5995 - return original error code on reconstructed replies (3761)
5996 - no transno return for symlink open, don't save no-trasno open (3440)
5998 - add pid to ldlm debugging output (4922)
5999 - bump the watchdog timeouts -- we can't handle 30sec yet
6000 - extra debugging for orphan dentry/inode bug (5259)
6002 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
6005 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
6006 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
6007 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
6008 - fix NULL dereference in /proc/sys/portals/routes (4827)
6009 - allow failed mdc_close() operations to be interrupted (4561)
6010 - stop precreate on OST before MDS would time out on it (4778)
6011 - don't send partial-page writes before EOF from client (4410)
6012 - discard client grant for sub-page writes on large-page clients (4520)
6013 - don't free dentries not owned by NFS code, check generation (4806)
6014 - fix lsm leak if mds_create_objects() fails (4801)
6015 - limit debug_daemon file size, always print CERROR messages (4789)
6016 - use transno after validating reply (3892)
6017 - process timed out requests if import state changes (3754)
6018 - update mtime on OST during writes, return in glimpse (4829)
6019 - add mkfsoptions to LDAP (4679)
6020 - use ->max_readahead method instead of zapping global ra (5039)
6021 - don't interrupt __l_wait_event() during strace
6023 - add software watchdogs to catch hung threads quickly (4941)
6024 - make lustrefs init script start after nfs is mounted
6025 - fix CWARN/ERROR duplication (4930)
6026 - return async write errors to application if possible (2248)
6027 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
6028 - print NAL number in %x format (4645)
6029 - update barely-supported suse-2.4.21-171 series (4842)
6030 - support for sles 9 %post scripts
6031 - support for building 2.6 kernel-source packages
6032 - support for sles km_* packages
6034 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
6037 - ignore -ENOENT errors in osc_destroy (3639)
6038 - notify osc create thread that OSC is being cleaned up (4600)
6039 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
6040 - reconstruct ost_handle() like mds_handle() (4657)
6041 - create a new thread to do import eviction to avoid deadlock (3969)
6042 - let lconf resolve symlinked-to devices (4629)
6043 - don't unlink "objects" from directory with default EA (4554)
6044 - hold socknal file ref over connect in case target is down (4394)
6045 - allow more than 32000 subdirectories in a single directory (3244)
6046 - fix blocks count for O_DIRECT writes (3751)
6047 - OST returns ENOSPC from object create when no space left (4539)
6048 - don't send truncate RPC if file size isn't changing (4410)
6049 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
6050 - bind to privileged port in socknal and tcpnal (3689)
6052 - rate limit CERROR/CWARN console message to avoid overload (4519)
6053 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
6054 - basic mmap support (3918)
6055 - kernel patch series update from b1_4 (4711)
6057 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
6060 - avoid crash during MDS cleanup with OST shut down (2775)
6061 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
6062 - don't use bad inodes on the MDS (3744)
6063 - dynamic object preallocation to improve recovery speed (4236)
6064 - don't hold spinlock over lock dumping or change debug flags (4401)
6065 - don't zero obd_dev when it is force cleaned (3651)
6066 - print grants to console if they go negative (4431)
6067 - "lctl deactivate" will stop automatic recovery attempts (3406)
6068 - look for existing locks in ldlm_handle_enqueue() (3764)
6069 - don't resolve lock handle twice in recovery avoiding race (4401)
6070 - revalidate should check working dir is a directory (4134)
6072 - don't always mark "slow" obdfilter messages as errors (4418)
6074 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
6077 - don't close LustreDB during write_conf until it is done (3860)
6078 - fix typo in lconf for_each_profile (3821)
6079 - allow dumping logs from multiple threads at one time (3820)
6080 - don't allow multiple threads in OSC recovery (3812)
6081 - fix debug_size parameters (3864)
6082 - fix mds_postrecov to initialize import for llog ctxt (3121)
6083 - replace config semaphore with spinlock (3306)
6084 - be sure to send a reply for a CANCEL rpc with bad export (3863)
6085 - don't allow enqueue to complete on a destroyed export (3822)
6086 - down write_lock before checking llog header bitmap (3825)
6087 - recover from lock replay timeout (3764)
6088 - up llog sem before sending rpc (3652)
6089 - reduce ns lock hold times when setting kms (3267)
6090 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
6091 - fix NULL deref and obd_dev leak on setup error (3312)
6092 - replace some LBUG about llog ops with error handling (3841)
6093 - don't match INVALID dentries from d_lookup and spin (3784)
6094 - hold dcache_lock while marking dentries INVALID and hashing (4255)
6095 - fix invalid assertion in ptlrpc_set_wait (3880)
6097 - add libwrap support for the TCP acceptor (3996)
6098 - add /proc/sys/portals/routes for non-root route listing (3994)
6099 - allow setting MDS UUID in .xml (2580)
6100 - print the stack of a process that LBUGs (4228)
6102 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
6105 - don't cleanup request in ll_file_open() on failed MDS open (3430)
6106 - make sure to unset replay flag from failed open requests (3440)
6107 - if default stripe count is 0, use OST count for inode size (3636)
6108 - update parent mtime/ctime on client for create/unlink (2611)
6109 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
6110 - free recovery state on server during a forced cleanup (3571)
6111 - unregister_reply for resent reqs (3063)
6112 - loop back devices mounting and status check on 2.6 (3563)
6113 - fix resource-creation race that can provoke i_size == 0 (3513)
6114 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
6115 - more debugging for page-accounting assertion (3746)
6116 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
6117 - avoid deadlock after precreation failure (3758)
6118 - fix race and lock order deadlock in orphan handling (3450, 3750)
6119 - add validity checks when grabbing inodes from l_ast_data (3599)
6121 - add /proc/.../recovery_status to obdfilter (3428)
6122 - lightweight CDEBUG infrastructure, debug daemon (3668)
6123 - change default OSC RPC parameters to be better on small clusters
6124 - turn off OST read cache for files smaller than 32MB
6125 - install man pages and include them in rpms (3100)
6126 - add new init script for (un)mounting lustre filesystems (2593)
6127 - run chkconfig in %post for init scripts (3701)
6128 - drop scimac NAL (unmaintained)
6130 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
6133 - clean kiobufs before and after use (3485)
6134 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
6135 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
6136 - change init script to fail more clearly if not run as root (1528)
6137 - allow clients to reconnect during replay (1742)
6138 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
6139 - don't do DNS lookups on NIDs too small for IP addresses (3442)
6140 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
6141 - fix cond_resched (3554)
6142 - only evict unfinished clients after recovery (3515)
6143 - allow bulk resend, prevent data loss (3570)
6144 - dynamic ptlrpc request buffer allocation (2102)
6145 - don't allow unlinking open directory if it isn't empty (2904)
6146 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
6147 - remove extraneous obd dereference causing LASSERT failure (3334)
6148 - don't use get_cycles() when creating temp. files on the mds (3156)
6149 - hold i_sem when setting i_size in ll_extent_lock() (3564)
6150 - handle EEXIST for set-stripe, set proper directory name (3336)
6152 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
6153 - fix ksocknal_fmb_callback() error messages (2918)
6155 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
6158 - don't copy lvb into (possibly NULL) reply on error (2983)
6159 - don't deref dentry after dput, don't free lvb on error (2922)
6160 - use the kms to determine writeback rpc length (2947)
6161 - increment oti_logcookies when osc is inactive (2948)
6162 - update client's i_blocks count via lvb messages (2543)
6163 - handle intent open/close of special files properly (1557)
6164 - mount MDS with errors=remount-ro, like obdfilter (2009)
6165 - initialize lock handle to avoid ASSERT on error cleanup (3057)
6166 - don't use cancelling-locks' kms values (2947)
6167 - use highest lock extent for kms, not last one (2925)
6168 - don't dereference ERR_PTR() dentry in error handling path (3107)
6169 - fix thread race in portals_debug_dumplog() (3122)
6170 - create lprocfs device entries at setup instead of at attach (1519)
6171 - common AST error handler, don't evict client on completion race (3145)
6172 - zero nameidata in detach_mnt in 2.6 (3118)
6173 - verify d_inode after revalidate_special is valid in 2.6 (3116)
6174 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
6175 - initialize RPC timeout timer earlier for 2.6 (3219)
6176 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
6177 - print nal/nid for unknown nid (3258)
6178 - additional checks for oscc recovery before doing precreate (3284)
6179 - fix ll_extent_lock() error return code for 64-bit systems (3043)
6180 - don't crash in mdc_close for bad permissions on open (3285)
6181 - zero i_rdev for non-device files (3147)
6182 - clear page->private before handing to FS, better assertion (3119)
6183 - tune the read pipeline (3236)
6184 - fix incorrect decref of invalidated dentry (2350)
6185 - provide read-ahead stats and refine rpc in flight stats (3328)
6186 - don't hold journal transaction open across create RPC (3313)
6187 - update atime on MDS at close time (3265)
6188 - close LDAP connection when recovering to avoid server load (3315)
6189 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
6190 - don't leak open file on MDS after open resend (3325)
6191 - serialize filter_precreate and filter_destroy_precreated (3329)
6192 - loop device shouldn't call sync_dev() for nul device (3092)
6193 - clear page cache after eviction (2766)
6194 - resynchronize MDS->OST in background (2824)
6195 - refuse to mount the same filesystem twice on same mountpoint (3394)
6196 - allow llmount to create routes for mounting behind routers (3320)
6197 - push lock cancellation to blocking thread for glimpse ASTs (3409)
6198 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
6199 - fix rare problem with rename on htree directories (3417)
6201 - allow default OST striping configuration per directory (1414)
6202 - fix compilation for qswnal for 2.6 kernels (3125)
6203 - increase maximum number of MDS request buffers for large systems
6204 - change liblustreapi to be useful for external progs like lfsck (3098)
6205 - increase local configuration timeout for slow disks (3353)
6206 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
6208 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
6211 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
6212 - don't overwrite extent policy data in reply if lock was blocked (2901)
6213 - drop filter export grants atomically with removal from device (2663)
6214 - del obd_self_export from work_list in class_disconnect_exports (2908)
6215 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
6216 - swab reply message in mdc_close, other PPC fixes (2464)
6217 - fix destroying of named logs (2325)
6218 - overwrite old logs when running lconf --write_conf (2264)
6219 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
6220 - fix race in target_handle_connect (2898)
6221 - mds_reint_create() should take same inode create lock (2926)
6222 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
6223 - don't close files for self_export to avoid uninitialized obd (2936)
6224 - allow MDS with the same name as client node (2939)
6225 - hold dentry reference for closed log files for unlink (2325)
6226 - reserve space for all logs during transactions (2059)
6227 - don't evict page beyond end of stripe extent (2925)
6228 - don't oops on a deleted current working directory (2399)
6229 - handle hard links to targets without a parent properly (2517)
6230 - don't dereference NULL lock when racing during eviction (2867)
6231 - don't grow lock extents when lots of conflicting locks (2919)
6233 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
6236 - account for cache space usage on clients to avoid data loss (974)
6237 - lfsck support in lustre kernel code (2349)
6238 - reduce journal credits needed for BRW writes (2370)
6239 - orphan handling to avoid losing space on client/server crashes
6240 - ptlrpcd can be blocked, stopping ALL progress (2477)
6241 - use lock value blocks to assist in proper KMS, faster stat (1021)
6242 - takes i_sem instead of DLM locks internally on obdfilter (2720)
6243 - recovery for initial connections (2355)
6244 - fixes for mds_cleanup_orphans (1934)
6245 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
6246 - block all file creations until orphan recovery completes (1901)
6247 - client remove rq_connection from request struct (2423)
6248 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
6249 - recovery timer race (2670)
6250 - mdc_close recovey bug (2532)
6251 - ptlrpc cleanup bug (2710)
6252 - mds timeout on local locks (2588)
6253 - namespace lock held during RPCs (2431)
6254 - handle interrupted sync write properly (2503)
6255 - don't try to handle a message that hasn't been replied to (2699)
6256 - client assert failure during cleanup after abort recovery (2701)
6257 - leak mdc device after failed mount (2712)
6258 - ptlrpc_check_set allows timedout requests to complete (2714)
6259 - wait for inflight reqs when ptlrpcd finishes (2710)
6260 - make sure unregistered services are removed from the srv_list
6261 - reset bulk XID's when resending them (caught by 1138 test)
6262 - unregister_bulk after timeout
6263 - fix lconf error (2694)
6264 - handle write after unfinished setstripe, stripe-only getstripe (2388)
6265 - readahead locks pages, leaves pending causing memory pressure (2673)
6266 - increase OST request buffers to 4096 on large machines (2729)
6267 - fix up permission of existing directories in simple_mkdir (2661)
6268 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
6269 - don't assign transno to errored transactions (2742)
6270 - don't delete objects on OST if given a bogus objid from MDS (2751)
6271 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
6272 - if rq_no_resend, then timeout request after recovery (2432)
6273 - fix MDS llog_logid record size, 64-bit array alignment (2733)
6274 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
6275 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
6276 - MDS recovery shouldn't delete objects that it has given out (2730)
6277 - if enqueue arrives after completion, don't clobber LVB (2819)
6278 - don't unlock pages twice when trigger_group_io returns error (2814)
6279 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
6280 - don't write pages to disk if there was an error (1450)
6281 - don't ping imports that have recovery disabled (2676)
6282 - take buffered bytes into account when balancing socknal conn (2817)
6283 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
6284 - reconnect unlink llog connection after MDS reconnects to OST (2816)
6285 - remove little-endian swabbing of llog records (1987)
6286 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
6287 - retry reposting request buffers if they fail (1191)
6288 - grow extent at grant time to avoid granting a revoked lock (2809)
6289 - lock revoke doesn't evict page if covered by a second lock (2765)
6290 - disable VM readahead to avoid reading outside lock extents (2805)
6292 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
6293 - updated kernel patches for hp-2.4.20 kernel (2681)
6295 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
6298 - fix truncated write corruption (2366)
6299 - fix for failed assertion in iopen_connect_dentry (1792,2517)
6301 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
6302 - ensure len(uuid) < 37 in lmc (1171)
6303 - fix ia64 OOPS in llog_test (2255)
6304 - zero end of page at obdfilter for partial page writes (2648)
6305 - don't leave stale dentries around after renames (bug 2428)
6306 - fix timeouts when evicting a client with a single lock held (2642)
6307 - set deadline for the initial HELLO message to drain (2634)
6308 - print out dotted-quad IP addresses in the socknal (2302)
6310 - additional debugging for MDS client eviction problem (2443)
6311 - fix mkfsoptions support for osts (2603, 2604)
6313 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
6316 - add series for the vanilla 2.6.0 kernel
6317 - add series for the vanilla 2.4.24 kernel
6318 - add series for a cray x86/64 UL kernel drop
6319 - fix xattr patches for the vanilla 2.4.19 series
6321 - generate true UUIDs in lmc (1171)
6322 - have portals stack dumping break in UML (2466)
6323 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
6324 - allocate with _NOFS in ldlm to avoid deadlock (1933)
6325 - wake callback waiting threads on client eviction (2460)
6326 - Add --ptldebug and --subsystem to lmc (1719)
6327 - update assertion to allow safe interrupt allocation
6328 - set rq_no_resend for cancel requests (2432)
6329 - recalculate ptlrpcd timeout after resend (2494)
6330 - call vfs_rmdir when removing pending directories (2368)
6331 - fix renaming a file to itself (2429)
6332 - lmc creates a default one-stripe lov (2454)
6333 - expand procfs space to handle large clusters (2326)
6334 - increase UML stack to avoid overflow
6335 - update lconf's list of debug and subsystem masks
6336 - fix lfs find --obd (2510)
6337 - /proc tunable for disabling filter read caching (2591)
6338 - stop rpm packages from altering slapd.conf (2301)
6339 - disable nagle in the socknal under 0conf (2578)
6340 - choose mds inode size based on stripe count (2572)
6341 - fix kernel-source rpm problems (2516)
6343 - add --disable-doc to avoid pdf generation (2421)
6344 - update documentation, tests, type-os, comments
6345 - avoid format warnings on ia64
6346 - remove the TOE NAL
6347 - tiny code cleanups by removing unused fields
6349 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
6352 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
6353 - remove the most blatant lies from BUILDING file (2371)
6354 - change default debug level to reasonable production setting
6355 - reduce client side cache size to reduce cache flush time
6356 - reduce max RPCs in flight to avoid unnecessary file fragmentation
6357 - make TCP zerocopy and pinger support enabled by default (2476)
6358 - sync writes completed after process exits caused crashes (2319)
6359 - maintain correct mount count on the MDS (2356)
6360 - backout 1557, because 2316 wasn't really fixed
6361 - better file I/O statistics gathering in /proc
6362 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
6363 - another kernel patch to fix zero-copy TCP function export
6364 - don't take duplicate lock when processing re-sent getattr (2420)
6365 - lctl uses obd_self_export instead of creating new conn (2353)
6366 - MDS/OST recovery case which requires object creation asserted (2425)
6367 - move lfs from /usr/sbin to /usr/bin in packages
6368 - fix race between mds_client_add and mds_client_free (2417)
6369 - use kmalloc instead of slabs in portals (2430)
6370 - don't create duplicate records when a failover MDS is present (2442)
6371 - remove unnecessary mount age check (2332)
6372 - don't remove directory inodes from locks prematurely (2451)
6373 - don't break if MDS service name is the same as hostname (2103)
6374 - fix races in client write RPC generation when cache full (2482)
6376 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
6379 - remove now-unused request->rq_obd (278)
6380 - if an allocation fails, print out how much memory we've used (1933)
6381 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
6382 - add the "configurable stack size" patch to most series files (1256)
6383 - ability to write large log records, for 100+ OST configs (2306)
6384 - fix NULL deref when filter_prep fails (2314)
6385 - fix operator precedence error in filter_sync
6386 - dynamic allocation of socknal TX descriptors (2315)
6387 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
6388 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
6389 - ensure that CWARN messages in Portals always get to the syslog
6390 - __init/__exit are not for prototype decls (ldlm_init/exit)
6391 - x86-64 compile warning fixes
6392 - fix gateway LMC keyword conflict (2318)
6393 - fix MDS lock inversions in getattr/reint paths (1844)
6394 - fix a rare lock re-ordering bug, which caused deadlock (2322)
6395 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
6396 - DLM race condition prevented some lock evictions (2328)
6397 - ENOMEM detection and retry on socknal sends (2230)
6398 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
6399 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
6400 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
6401 - better error messages when a client is rejected during recovery (1505)
6402 - avoid cancelling locks which were never granted, after failure (2330)
6403 - fix i_sem/journal inversion in mds_client_add (2333)
6404 - fix truncate/getattr lock cycle deadlock (2334)
6405 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
6406 - fix two rare exit paths which could leak an l_lock() ref (2321)
6407 - fencepost error in MDS/OST orphan recovery (2226)
6408 - make log record alignment 8 bytes (1988)
6409 - lstripe now fails when requested offset > ost_count (2237)
6410 - ensure that all kernel series have a complete list.h (1607)
6411 - fix crashes in special-file operations (2316)
6412 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
6414 - allow configurable automake binary, for testing new versions
6415 - small update to the lfs documentation
6417 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
6419 * fix negative export reference count in fsfilt_sync (2312)
6421 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
6422 * release candidate 0.9.1
6424 - orphans are moved into the PENDING directory for possible recovery
6425 - replayed opens now open by fid for orphan/rename safety (1042)
6426 - last close of an orphan inode generates a transno (683)
6427 - chdir() and mount() now pin the directory entry (1020)
6428 - avoid CERROR in normal ll_setattr_raw() error case (1500)
6429 - discard very old requests without processing them (1502)
6430 - remove some common, well-understood CERRORs (1505)
6431 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
6432 - clear "grant" flags in OST replies until OST grant code lands (1644)
6433 - fix read performance by not clobbering i_blksize on client (1598)
6434 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
6435 - make LRU size a /proc tunable, clears locks when reduced (707)
6436 - fix some lprocfs rot that prevented ptlbd from loading (1732)
6437 - server locks take references on exports now (1558)
6438 - build fixes for 2.4.20-rh trees (1663)
6439 - return an error from lov_create if all OSCs are inactive (1751)
6440 - fix import levels when a reconnect happens without a timeout (1597)
6441 - exit early from mds_open if we get a lookup error (1749)
6442 - partial page read at EOF wouldn't wait for disk before sending (1642)
6443 - avoid NULL deref in obdfilter when reading page past EOF (1592)
6444 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
6445 - fix LBUG in ll_it_open_error with rc = -2 (1861)
6446 - write/truncate lock inversion (1639)
6447 - Don't auto-load obdclass, portals modules during cleanup (1495)
6448 - fix timestamps from jumping to "now" (1763)
6449 - extra journal assertions (1648)
6450 - add an extra multiunlink test (1771)
6451 - fix read_record/write_record API (1776)
6452 - fix leak of offset_extent, possible incorrect i_size later (1772)
6453 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
6454 - Debugging for the kqswnal_get_idle_tx problems (1820)
6455 - Allow recovery to be attempted multiple times (1536)
6456 - Write out MDS last_rcvd file after it is first created (1600)
6457 - Fix tx_descriptor leak in failed transmit situations (1827)
6458 - ext3 journaling fixes for assertion failure after IO error (1871)
6459 - class_export_put() on freed export after completion AST error (1896)
6460 - Fix revalidate looping in VFS (1322)
6461 - Don't access a freed export during MDS_REINT timeout (1521)
6462 - Add open-unlink recovery support on the MDS (1673,1764)
6463 - Return an error if no MDS data was read from last_rcvd (1946)
6464 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
6465 - Don't setup a disk device that doesn't match exported UUID (317)
6466 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
6467 - avoid committing NULL handle in force close
6468 - local.sh is now a one-stripe LOV configuration
6469 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
6470 - fix ext3 htree duplicate directory entry corruption (1516)
6471 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
6472 - update to new LOV EA format (2097)
6473 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
6475 2003-06-15 Phil Schwan <phil@clusterfs.com>
6478 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
6479 - per-import/export recovery handling (958, 931, 959)
6480 - multiple last-rcvd slots, for serving multiple FSes (949)
6481 - connections are again shared between multiple imp/exports (963, 964)
6482 - "umount -f" would hang if any requests needed to be sent (393, 978)
6483 - avoid pinning large req buffer by copying for queued messages (989)
6484 - add "uuid" to "lctl device" command to help upcalls (991)
6485 - "open" RPCs with transnos would confuse recovery counters (1037)
6486 - do proper endian conversion of all wire messages (288, 340, 891)
6487 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
6488 - call waiting locks callback from LDLM recovery thread (1127, 1151)
6489 - fix ptlrpc_connection leak in target_handle_connect (1174)
6490 - fix import refcounting bug in OST and MDS cleanup (1134)
6491 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
6492 - fix very unlikely obd_types race condition (501)
6493 - remove osc_open hack for echo_client (1187)
6494 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
6495 - a failure in read_inode2 leads to deadlock (1139)
6496 - cancel ack-locks as soon as transaction is committed (1072)
6497 - fix major leaks and crashes in the bulk I/O path (937, 1057)
6498 - make sure to commitrw after any preprw to avoid deadlock (1162)
6499 - failing to execute a file in a lustre FS would lock inode (1203)
6500 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
6501 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
6502 - obd_uuid2tgt would walk off the end of the list (1255)
6503 - on IA64 the osc would give portals incorrect bulk size (1258)
6504 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
6505 - fix lock inversion caused by new llite matching code (1282)
6506 - limit the number of dirty pages on a client to 10MB (1286)
6507 - timed out locks were not being corrected cancelled (1289)
6508 - fix O_DIRECT above 4GB on IA-32 (1292)
6509 * major user-visible changes
6510 - fail out/fail over policy now controlled by the upcall (993)
6512 - add OBD_PING to check server availability and failure (954)
6513 - lustre messages are now sent in sending host order (288, 340, 891)
6514 - add eadatalen to MDS getattr reply (340)
6515 - OST read replies may contain second buffer, with per-page status (593)
6517 2003-03-11 Phil Schwan <phil@clusterfs.com>
6520 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
6521 - failed open()s could cause deadlock; fixed (867, 869)
6522 - stop cancelling OST locks when files are closed (481)
6523 - overlapping XID spaces caused network corruption (851, 853)
6524 - fix unsafe fsfilt counter arithmetic; change to atomic_t
6525 - setattr_raw added, to do single-RPC, server-side setattrs
6526 - lmc/lconf syntax change for OST UUIDs
6527 - fix crashy race condition between ptlrpc_free_req and osc_close
6528 - don't use request in mdc_enqueue if we hit a timeout (889)
6529 - don't set the inode i_size for regular files from the MDS (896)
6530 - handle out of order completion AST (842)
6531 - don't LBUG if a lock request times out after receiving AST (913)
6532 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
6533 - if a bad lock AST arrives, send an error instead of dropping entirely
6534 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
6535 - fix leak in bulk IO when only partially completed (899, 900, 926)
6536 - fix O_DIRECT for ia64 (55)
6537 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
6538 - C-z after timeout could hang a process forever; fixed (977)
6540 - client-side I/O cache (678, 924, 929, 941, 970)
6542 - READPAGE and SETATTRs which don't take server-side locks get
6545 2003-02-11 Phil Schwan <phil@clusterfs.com>
6548 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
6549 - Fix fsfilt_extN_readpage() to read a full page of directory
6550 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
6551 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
6552 - Fix imbalanced LOV object allocation and out-of-bound access (469)
6553 - Most intent operations were removed, in favour of a new RPC mode
6554 that does a single RPC to the server and bypasses most of the VFS
6555 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
6556 - Aggressively cancel local locks on DLM servers
6557 - mds_reint_unlink sends EA to the client if it's the last nlink.
6558 client uses that EA to unlink OST objects.
6559 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
6560 - recursive symlinks were fixed (439)
6561 - fixed NULL deref in DEBUG_REQ
6562 - filter_update_lastobjid no longer calls sync, which annoyed extN
6563 - fixed multi-client small-writes to a single file problem (445)
6564 - fixed mtime updates during file writes (607)
6565 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
6566 - fixed bug in obd_brw_read/write() (under guise of testing 367)
6567 - fixed Linux OST size reporting problem (444, 656)
6568 - OST now updates object mtime with writes or setattr (607, 619)
6569 - client verifies file size before zeroing page past EOF (445)
6570 - OST now writes last allocated objid to disk with allocation (108)
6571 - LOV on echo now works (409)
6573 - mds_reint_unlink sends a new buffer, with the EA included. this
6574 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
6575 set if a regular file was being unlinked, and it was the last link
6576 - use PtlGet from the target for bulk writes (315)
6577 - OST now updates object mtime with writes or setattr (607, 619)
6578 - LDLM now has a grant-time callback to revalidate locked items, if
6580 - Many MDS operations were reorganized to combat race conditions
6582 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
6583 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
6585 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
6588 - Fully reactivate OST imports after reconnection (512, others)
6589 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
6590 - More graceful error handling for truncating on dead OST (515)
6591 - Don't error out unless we're actually accessing dead stripes (474)
6592 - Fix garbage sizes when stripes are missing (410)
6593 - LRU counters were broken, causing constant lock purge (433, 432)
6594 - garbage on read from stripes with failed OSTs (441)
6595 - mark OSCs as active before reconnecting during recovery (438)
6596 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
6597 - lfind did not preserve OST order in output (443)
6598 - symlinks cause hung clients, incorrect data (439)
6599 - stop dereferencing request after dropping refcount (457)
6600 - don't LASSERT(spin_is_locked) on non-SMP (455)
6601 - fixes for many rename() bugs
6602 - fstat didn't correctly synchronize attributes (399)
6603 - server must handle lock cancellation during blocking AST prep (487)
6604 - bulk descriptors were free()d too soon (511)
6605 - fix paths in lconf, which would load incorrect modules (451, 507)
6606 - fix confusing lconf 'host not found' error message (386)
6607 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
6608 - fix race condition in mdc_blocking_ast() for inode access (526)
6609 - fix lov_unpackmd() unpacking wrong number of stripes (537)
6610 - fix lov_set_osc_active() marking wrong OSC inactive (440)
6611 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
6612 - fix multiple writes of stripe MD to MDS (358, maybe 519)
6613 - fix lstripe in several ways (kernel side) (527)
6614 - fix request leak in ldlm_cli_enqueue (262)
6615 - incorrect OSC was marked inactive after OST failure
6616 - call mds_fs_cleanup before unmounting filesystem (524)
6617 - fix races between taking ns_lock and ldlm_lock_change_resource
6618 - fix races updating LOV export open file list
6619 - fix lov_enqueue error path, avoid decref-ing bad lock handle
6620 - fix recovery NULL deref in ldlm_cli_cancel_unused
6621 - fix some DLM races by using new hash table for lock handles (419)
6622 - permit the client to specify desired inodes, at replay
6623 - duplicate requests when we queue them for replay reintegration
6624 - fix last_rcvd offset calculation
6625 - sync after each recovered transaction, so we always make progress
6626 - never, not always, ERESTART requests without transnos
6627 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
6629 - skip replay if the MDS says that the client is already connected
6630 - don't check for a recovery-enabled export to match lctl's UUID
6631 - don't INC_USE_COUNT for phantom exports
6632 - don't crash when cleaning up phantom exports (567)
6633 - don't double-finish or set replay data for errored mdc_open requests
6634 - abort requests when they time out, so we don't get old replies
6635 - send/receive replies for AST messages again
6636 - if the client says that it doesn't have the lock, cancel it on the
6638 - if we timeout during I/O, don't try to cancel an in-use lock; instead
6639 mark it as destroyed, it will all work out when decref is called
6640 - fix module use counts (22, 581)
6642 - ASTs now expect a reply (server cancels lock on error reply)
6644 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
6647 - fix many simultaneous client startup (392)
6648 - fix dentry->d_it clobbering
6649 - credentials weren't being shipped for readdir/getattr operations
6650 - remove invalid assertions triggered during some concurrent MD
6652 - proper Lustre versions added (336, 389)
6653 - fix memory leak for create error case (398)
6654 - fix LOV locking bug that would get cli/srv out of sync
6655 - fix echo client over LOV (409)
6656 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
6657 - fix double-O_EXCL intent crash (424)
6658 - avoid sending multiple lock CANCELs (352)
6660 - MDS can do multi-client recovery (modulo bugs in new code)
6662 - many updates, edits, cleanups
6664 2002-11-18 Phil Schwan <phil@clusterfs.com>
6667 - fix null d_it dereference (346)
6668 - fix full OST/dbench hang (333)
6669 - fix permission problem with file removal (286)
6670 - fix removal of OSCs from LOV when they fail
6671 - fix NULL deref during bulk timeout (214)
6672 - fix problems related to multiple filesystems on one MDS (241)
6673 - fixed serious subtle metadata locking bugs
6674 - free locks on clients when inodes are removed due to memory
6676 - fix inode pointer in lock data (285)
6677 - partial support for multiple MDS on a single host (241)
6678 - data locks weren't cancelled at clear_inode time (290, 311)
6679 - intent locks could lead to unbounded lock growth (205)
6680 - added a maximum lock count, an LRU list, and a flusher
6681 - fix multiple rename (365)
6682 - properly abstracted the echo client
6683 - OSC locked 1 byte too many; fixed
6684 - rewrote brw callback code:
6685 - fixed recovery bugs related to LOVs (306)
6686 - fixed too-many-pages-in-one-write crash (191)
6687 - fixed (again) crash in sync_io_timeout (214)
6688 - probably fixed callback-related race (385)
6690 - Add capability to MDS protocol
6691 - LDLM cancellations and callbacks on different portals
6693 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
6696 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
6697 - defer open object destruction to close time (601981, 138)
6698 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
6699 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
6700 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
6701 - don't call obd_setattr in ll_file_release for destroyed objects
6702 * protocol change to lustre_msg: move |version| and add |flags|
6703 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
6704 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
6705 * added replay of create, unlink, link and rename operations during
6706 MDS failover; recovery should be much more robust now
6707 * remove failed OSCs from LOVs (only lov_create uses this so far)
6708 * the lustre-HOWTO was brought (more) up to date (582544)
6710 2002-10-23 Phil Schwan <phil@clusterfs.com>
6713 - in-use dentries weren't being reused properly (617851)
6714 - prevent multiple LDLM setup (599178)
6715 - fix LOV size calculations for truncate (617853)
6716 - fix client handling of MDS intent errors (POSIX)
6717 - fix permission bug in lovstripe.c test (624321)
6718 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
6719 - truncate past end of file could corrupt data
6720 - proper cleanup after timeouts, crashes, etc (592524, 550815)
6721 - a race in recovery could return ETIMEDOUT to apps (623947)
6722 - building outside the source directory was fixed
6723 * the lustre-HOWTO was brought (more) up to date (582544)
6724 * major progress was made on recovery functionality
6726 2002-10-10 Phil Schwan <phil@clusterfs.com>
6729 - recovery deadlock fix
6730 - rm -rf causes LBUG fix (617817)
6731 - file open by multiple tasks fix (618962)
6732 - directory permissions bugs (602707 and 620007)
6733 - journal_stop fixed with locking (611313)
6734 - O_APPEND failures resolved (618273, perhaps 614459)
6735 - lconf PATH fix (619770)
6736 - IA64 build fix (621450)
6737 - RPC buffer sizes scale with amount of memory
6739 2002-10-01 Phil Schwan <phil@clusterfs.com>
6742 - locks would be cancelled without throwing away data pages,
6743 resulting in inconsistent data (605627)
6744 - inode attributes were not always being refreshed (605627, 612449)
6745 - lconf now continues to cleanup after lctl reports an error
6746 - MDS now enforces user permissions (602707)
6747 - lprocfs cleanup fixed, but not yet enabled (614157)
6748 - fixed infinite server hang, should a client not respond to an AST
6749 - avoid going into recovery if user calls readlink() with a buffer
6750 that's too small (613941)
6751 - AST RPCs no longer require replies (614867) -- this may be changed
6752 - don't crash server if client sends an IOV that's too big (611336)
6753 - fixed lock conversion deadlock (611892)
6754 - fixed the following of symlinks (614622)
6755 * recovery: the server can remove locks from a client that dies, other
6756 clients can make progress
6757 * more extN patch fixes
6758 * compile-time configurable ptlrpc buffer allocations
6760 - collaborative read cache document
6761 - Lustre Lite Performance CDR document-in-progress
6763 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6766 - fix typo in patch-2.4.18
6768 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6771 - clear ptlrpc request each time in handle_incoming_request()
6772 - unlink of files now destroys the object on the OST
6774 2002-09-19 Peter Braam <braam@clusterfs.com>
6776 * add hard link support
6777 * change obdfile creation method
6778 * kernel patch changed
6780 2002-09-19 Peter Braam <braam@clusterfs.com>
6783 - stack overflow bug in extN fixed
6785 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
6787 * documentation updates
6788 - add man pages for config tools
6789 - update tests/README to describe testing with new config tools
6790 - finish metadata API descriptions
6791 * bug fixes and cleanups
6792 - statfs workaround for 16TB limit
6793 - LOV stripe allocation improved, can stripe on subset of OSTs
6794 - LOV file size/IO offset was wrong for files > 4GB in size
6795 - object EA data was being dropped, caused files to be unreadable
6796 - memory overflow with non-LOV OST caused memory corruption
6797 - fixed regression tests to work with new config tools, obdfilter
6798 - fixed bug when directory size became larger than 1 block
6799 - fixed bug (for single client case) when PWD was deleted
6800 - invalidate local directory pages when doing intent-based ops
6801 - avoid LDLM oops when lock callback contained bad data
6803 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
6805 * documentation updates
6806 * bug fixes and cleanups
6807 - configuration tools
6810 - 64-bit compile warnings
6811 - 64-bit internal statfs data
6813 * test_brw on persistent OST devices
6815 * lprocfs (disabled)
6817 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
6819 * documentation updates
6820 * bug fixes and cleanups
6821 * configuration tools
6823 2002-08-30 Peter J. Braam <braam@clusterfs.com>
6826 * many small fixes to 0_5_4
6827 * io/network handling
6828 * thinkos in MDS operations
6830 2002-08-24 Peter J. Braam <braam@clusterfs.com>
6833 * crucial basic fixes to 0.5.3
6834 * IOR, Iozone work over Elan
6837 2002-08-07 Phil Schwan <phil@clusterfs.com>
6838 * version 0_5_3, our first alpha
6839 * we use the new Portals iovs
6840 * documentation updates
6841 * bug fixes and cleanups
6842 * small changes in the DLM wire protocol
6844 2002-07-25 Peter J. Braam <braam@clusterfs.com>
6845 * version 0_5_1 with some initial stability,
6846 * locking on MD and file I/O.
6847 * documentation updates
6848 * several bug fixes since 0.5.0
6849 * small changes in wire protocol
6851 2002-07-18 Phil Schwan <phil@clusterfs.com>
6853 * delivered as Lustre Light Alpha
6854 * fixed a crash after handling invalid MDS requests
6855 * fixed directory pages for architectures with non-4k pages sizes
6857 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
6858 * release version v0_4_4
6859 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
6861 * Fixes a number of interruption problems with OST operations.
6862 * Update documentation for portals header changes
6863 * Move all wire protocol structs/defines to lustre_idl.h
6864 * Fixes symlink length bug.
6865 * Add tcpdump to repository.
6867 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
6868 * release version v0_4_3
6869 * Fixes statfs for inodes on extN.
6870 * Fixes bug in runtests which would delete /etc/hosts.
6871 * Use 64-bit object IDs wherever possible (not into VFS though)
6872 Remove ost_get_info, which is unused by lustre, and out of date.
6874 2002-07-03 Peter Braam <braam@clusterfs.com>
6875 * release version v0_4_2 Fixes a lookup error (type not passed)
6876 * move forward to head of Portals
6877 * move forward to latest Lustre kernel
6879 2002-06-25 Peter Braam <braam@clusterfs.com>
6880 * release version v0_4_1. Hopefully stable on single node use.