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 3.10.0-1062.9.1.el7 (RHEL7.7)
8 * Server known to build on the following kernels (others may also work):
9 2.6.32-696.18.7.el6 (RHEL6.9)
10 3.10.0-862.14.4.el7 (RHEL7.5)
11 4.4.120-92.70 (SLES12 SP2)
12 4.4.180-94.97 (SLES12 SP3)
13 4.4.0-142 (Ubuntu 16.04)
14 4.15.0-45 (Ubuntu 18.04)
15 vanilla linux 4.6.7 (ZFS only)
16 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
17 * Client primary unpatched kernels built and tested during release cycle:
18 3.10.0-957.27.2.el7 (RHEL7.6)
19 3.10.0-1062.9.1.el7 (RHEL7.7)
20 4.18.0-80.11.2.el8 (RHEL8.0)
21 4.4.180-94.97 (SLES12 SP3)
22 4.12.14-95.45 (SLES12 SP4)
23 4.15.0-45 (Ubuntu 18.04)
24 * Client known to build on unpatched kernels (others may also work):
25 2.6.32-696.18.7.el6 (RHEL6.9)
26 2.6.32-754.6.3.el6 (RHEL6.10)
27 3.10.0-862.14.4.el7 (RHEL7.5)
28 4.14.0-49.13.1.el7a (RHEL7.5)
29 4.14.0-115.2.2.el7a (RHEL7.6)
30 4.18.0-147.3.1.el8 (RHEL8.1)
31 4.4.120-92.70 (SLES12 SP2)
32 4.4.0-142 (Ubuntu 16.04)
33 vanilla linux 4.15.0 (ZFS only)
34 * Recommended e2fsprogs version: 1.45.2.wc1 or newer
35 * Recommended ZFS / SPL version: 0.7.13
36 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
37 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
38 stack overflow. For more information, please refer to bugzilla 17630.
39 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
40 the re-exporting nodes to be mounted with "32bitapi" mount option
41 * Note for Ubuntu 14.04.5:
42 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
43 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
47 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
48 for currently supported client and server kernel versions.
49 * Server primary kernels built and tested during release cycle:
50 3.10.0-957.27.2.el7 (RHEL7.6)
51 3.10.0-1062.9.1.el7 (RHEL7.7)
52 * Server known to build on the following kernels (others may also work):
53 2.6.32-696.18.7.el6 (RHEL6.9)
54 3.10.0-862.14.4.el7 (RHEL7.5)
55 4.4.120-92.70 (SLES12 SP2)
56 4.4.180-94.97 (SLES12 SP3)
57 4.4.0-142 (Ubuntu 16.04)
58 4.15.0-45 (Ubuntu 18.04)
59 vanilla linux 4.6.7 (ZFS only)
60 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
61 * Client primary unpatched kernels built and tested during release cycle:
62 3.10.0-957.27.2.el7 (RHEL7.6)
63 3.10.0-1062.9.1.el7 (RHEL7.7)
64 4.18.0-80.11.2.el8 (RHEL8.0)
65 4.4.180-94.97 (SLES12 SP3)
66 4.12.14-95.29 (SLES12 SP4)
67 4.15.0-45 (Ubuntu 18.04)
68 * Client known to build on unpatched kernels (others may also work):
69 2.6.32-696.18.7.el6 (RHEL6.9)
70 2.6.32-754.6.3.el6 (RHEL6.10)
71 3.10.0-862.14.4.el7 (RHEL7.5)
72 4.14.0-49.13.1.el7a (RHEL7.5)
73 4.14.0-115.2.2.el7a (RHEL7.6)
74 4.18.0-147.3.1.el8 (RHEL8.1)
75 4.4.120-92.70 (SLES12 SP2)
76 4.4.0-142 (Ubuntu 16.04)
77 vanilla linux 4.15.0 (ZFS only)
78 * Recommended e2fsprogs version: 1.45.2.wc1 or newer
79 * Recommended ZFS / SPL version: 0.7.13
80 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
81 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
82 stack overflow. For more information, please refer to bugzilla 17630.
83 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
84 the re-exporting nodes to be mounted with "32bitapi" mount option
85 * Note for Ubuntu 14.04.5:
86 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
87 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
91 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
92 for currently supported client and server kernel versions.
93 * Server primary kernels built and tested during release cycle:
94 3.10.0-957.27.2.el7 (RHEL7.6)
95 3.10.0-1062.1.1.el7 (RHEL7.7)
96 * Server known to build on the following kernels (others may also work):
97 2.6.32-696.18.7.el6 (RHEL6.9)
98 3.10.0-862.14.4.el7 (RHEL7.5)
99 4.4.120-92.70 (SLES12 SP2)
100 4.4.180-94.97 (SLES12 SP3)
101 4.4.0-142 (Ubuntu 16.04)
102 4.15.0-45 (Ubuntu 18.04)
103 vanilla linux 4.6.7 (ZFS only)
104 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
105 * Client primary unpatched kernels built and tested during release cycle:
106 3.10.0-957.27.2.el7 (RHEL7.6)
107 3.10.0-1062.1.1.el7 (RHEL7.7)
108 4.18.0-80.7.1.el8 (RHEL8.0)
109 4.4.180-94.97 (SLES12 SP3)
110 4.12.14-95.29 (SLES12 SP4)
111 4.15.0-45 (Ubuntu 18.04)
112 * Client known to build on unpatched kernels (others may also work):
113 2.6.32-696.18.7.el6 (RHEL6.9)
114 2.6.32-754.6.3.el6 (RHEL6.10)
115 3.10.0-862.14.4.el7 (RHEL7.5)
116 4.14.0-49.13.1.el7a (RHEL7.5)
117 4.14.0-115.2.2.el7a (RHEL7.6)
118 4.4.120-92.70 (SLES12 SP2)
119 4.4.0-142 (Ubuntu 16.04)
120 vanilla linux 4.15.0 (ZFS only)
121 * Recommended e2fsprogs version: 1.45.2.wc1 or newer
122 * Recommended ZFS / SPL version: 0.7.13
123 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
124 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
125 stack overflow. For more information, please refer to bugzilla 17630.
126 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
127 the re-exporting nodes to be mounted with "32bitapi" mount option
128 * Note for Ubuntu 14.04.5:
129 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
130 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
134 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
135 for currently supported client and server kernel versions.
136 * Server known to build on the following kernels (others may also work):
137 2.6.32-696.18.7.el6 (RHEL6.9)
138 3.10.0-862.14.4.el7 (RHEL7.5)
139 3.10.0-957.10.1.el7 (RHEL7.6)
140 4.4.120-92.70 (SLES12 SP2)
141 4.4.162-94.69 (SLES12 SP3)
142 4.4.0-142 (Ubuntu 16.04)
143 4.15.0-45 (Ubuntu 18.04)
144 vanilla linux 4.6.7 (ZFS only)
145 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
146 * Client known to build on unpatched kernels (others may also work):
147 2.6.32-696.18.7.el6 (RHEL6.9)
148 2.6.32-754.6.3.el6 (RHEL6.10)
149 3.10.0-862.14.4.el7 (RHEL7.5)
150 3.10.0-957.10.1.el7 (RHEL7.6)
151 4.14.0-49.13.1.el7a (RHEL7.5)
152 4.14.0-115.2.2.el7a (RHEL7.6)
153 4.4.120-92.70 (SLES12 SP2)
154 4.4.162-94.69 (SLES12 SP3)
155 4.12.14-95.13 (SLES12 SP4)
156 4.4.0-142 (Ubuntu 16.04)
157 4.15.0-45 (Ubuntu 18.04)
158 vanilla linux 4.15.0 (ZFS only)
159 * Recommended e2fsprogs version: 1.44.5.wc1 or newer
160 * Recommended ZFS / SPL version: 0.7.13
161 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
162 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
163 stack overflow. For more information, please refer to bugzilla 17630.
164 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
165 the re-exporting nodes to be mounted with "32bitapi" mount option
166 * Note for Ubuntu 14.04.5:
167 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
168 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
173 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
174 for currently supported client and server kernel versions.
175 * Server known to build on the following kernels (others may also work):
176 2.6.32-696.18.7.el6 (RHEL6.9)
177 3.10.0-862.14.4.el7 (RHEL7.5)
178 3.10.0-957.10.1.el7 (RHEL7.6)
179 4.4.120-92.70 (SLES12 SP2)
180 4.4.162-94.69 (SLES12 SP3)
181 4.4.0-142 (Ubuntu 16.04)
182 4.15.0-45 (Ubuntu 18.04)
183 vanilla linux 4.6.7 (ZFS only)
184 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
185 * Client known to build on unpatched kernels (others may also work):
186 2.6.32-696.18.7.el6 (RHEL6.9)
187 2.6.32-754.6.3.el6 (RHEL6.10)
188 3.10.0-862.14.4.el7 (RHEL7.5)
189 3.10.0-957.10.1.el7 (RHEL7.6)
190 4.14.0-49.13.1.el7a (RHEL7.5)
191 4.14.0-115.2.2.el7a (RHEL7.6)
192 4.4.120-92.70 (SLES12 SP2)
193 4.4.162-94.69 (SLES12 SP3)
194 4.12.14-95.13 (SLES12 SP4)
195 4.4.0-142 (Ubuntu 16.04)
196 4.15.0-45 (Ubuntu 18.04)
197 vanilla linux 4.15.0 (ZFS only)
198 * Recommended e2fsprogs version: 1.44.5.wc1 or newer
199 * Recommended ZFS / SPL version: 0.7.13
200 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
201 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
202 stack overflow. For more information, please refer to bugzilla 17630.
203 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
204 the re-exporting nodes to be mounted with "32bitapi" mount option
205 * Note for Ubuntu 14.04.5:
206 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
207 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
211 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
212 for currently supported client and server kernel versions.
213 * Server known to build on the following kernels (others may also work):
214 2.6.32-696.18.7.el6 (RHEL6.9)
215 3.10.0-862.14.4.el7 (RHEL7.5)
216 3.10.0-957.el7 (RHEL7.6)
217 4.4.120-92.70 (SLES12 SP2)
218 4.4.162-94.69 (SLES12 SP3)
219 4.4.0-131 (Ubuntu 16.04)
220 4.15.0-32 (Ubuntu 18.04)
221 vanilla linux 4.6.7 (ZFS only)
222 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
223 * Client known to build on unpatched kernels (others may also work):
224 2.6.32-696.18.7.el6 (RHEL6.9)
225 2.6.32-754.6.3.el6 (RHEL6.10)
226 3.10.0-862.14.4.el7 (RHEL7.5)
227 3.10.0-957.el7 (RHEL7.6)
228 4.14.0-49.13.1.el7a (RHEL7.5)
229 4.14.0-115.2.2.el7a (RHEL7.6)
230 4.4.120-92.70 (SLES12 SP2)
231 4.4.162-94.69 (SLES12 SP3)
232 4.4.0-131 (Ubuntu 16.04)
233 4.15.0-32 (Ubuntu 18.04)
234 vanilla linux 4.15.0 (ZFS only)
235 * Recommended e2fsprogs version: 1.44.3.wc1 or newer
236 * Recommended ZFS / SPL version: 0.7.9
237 * Tested with ZFS / SPL version: 0.7.9
238 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
239 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
240 stack overflow. For more information, please refer to bugzilla 17630.
241 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
242 the re-exporting nodes to be mounted with "32bitapi" mount option
243 * Note for Ubuntu 14.04.5:
244 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
245 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
247 --------------------------------------------------------------------------------
250 2018-04-03 Intel Corporation
252 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
253 for currently supported client and server kernel versions.
254 * Server known to build on patched kernels:
255 2.6.32-431.29.2.el6 (RHEL6.5)
256 2.6.32-504.30.3.el6 (RHEL6.6)
257 2.6.32-573.26.1.el6 (RHEL6.7)
258 2.6.32-642.15.1.el6 (RHEL6.8)
259 2.6.32-696.18.7.el6 (RHEL6.9)
260 3.10.0-693.21.1.el7 (RHEL7.4)
261 3.0.101-0.47.71 (SLES11 SP3)
262 3.0.101-107 (SLES11 SP4)
263 3.12.74-60.64.40 (SLES12 SP1)
264 4.4.114-92.64 (SLES12 SP2)
265 4.4.114-94.11 (SLES12 SP3)
266 3.13.0-101 (Ubuntu 14.04, ZFS only)
267 4.4.0-85.108 (Ubuntu 14.04.5 LTS)
268 4.4.0-116 (Ubuntu 16.04)
269 vanilla linux 4.6.7 (ZFS only)
270 * Client known to build on unpatched kernels:
271 2.6.32-431.29.2.el6 (RHEL6.5)
272 2.6.32-504.30.3.el6 (RHEL6.6)
273 2.6.32-573.26.1.el6 (RHEL6.7)
274 2.6.32-642.15.1.el6 (RHEL6.8)
275 2.6.32-696.18.7.el6 (RHEL6.9)
276 3.10.0-693.21.1.el7 (RHEL7.4)
277 3.0.101-0.47.71 (SLES11 SP3)
278 3.0.101-107 (SLES11 SP4)
279 3.12.74-60.64.40 (SLES12 SP1)
280 4.4.114-92.64 (SLES12 SP2)
281 4.4.114-94.11 (SLES12 SP3)
282 3.13.0-101 (Ubuntu 14.04)
283 4.4.0-85.108 (Ubuntu 14.04.5 LTS)
284 4.4.0-116 (Ubuntu 16.04)
286 * Recommended e2fsprogs version: 1.42.13.wc6 or newer
287 * Recommended ZFS / SPL version: 0.7.6
288 * Tested with ZFS / SPL version: 0.7.6
289 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
290 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
291 stack overflow. For more information, please refer to bugzilla 17630.
292 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
293 the re-exporting nodes to be mounted with "32bitapi" mount option
294 * Note for Ubuntu 14.04.5:
295 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
296 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
298 --------------------------------------------------------------------------------
300 2017-07-05 Intel Corporation
302 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
303 for currently supported client and server kernel versions.
304 * Server known to build on patched kernels:
305 2.6.32-431.29.2.el6 (RHEL6.5)
306 2.6.32-504.30.3.el6 (RHEL6.6)
307 2.6.32-573.26.1.el6 (RHEL6.7)
308 2.6.32-642.15.1.el6 (RHEL6.8)
309 2.6.32-696.3.1.el6 (RHEL6.9)
310 3.10.0-514.21.1.el7 (RHEL7.3)
311 3.0.101-0.47.71 (SLES11 SP3)
312 3.0.101-100 (SLES11 SP4)
313 3.12.74-60.64.40 (SLES12 SP1)
314 4.4.59-92.17 (SLES12 SP2)
315 vanilla linux 4.6.7 (ZFS only)
316 * Client known to build on unpatched kernels:
317 2.6.32-431.29.2.el6 (RHEL6.5)
318 2.6.32-504.30.3.el6 (RHEL6.6)
319 2.6.32-573.26.1.el6 (RHEL6.7)
320 2.6.32-642.15.1.el6 (RHEL6.8)
321 2.6.32-696.3.1.el6 (RHEL6.9)
322 3.10.0-514.21.1.el7 (RHEL7.3)
323 3.0.101-0.47.71 (SLES11 SP3)
324 3.0.101-100 (SLES11 SP4)
325 3.12.74-60.64.40 (SLES12 SP1)
326 4.4.59-92.17 (SLES12 SP2)
328 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
329 * Recommended ZFS / SPL version: 0.6.5.9
330 * Tested with ZFS / SPL version: 0.6.5.9
331 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
332 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
333 stack overflow. For more information, please refer to bugzilla 17630.
334 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
335 the re-exporting nodes to be mounted with "32bitapi" mount option
337 --------------------------------------------------------------------------------
339 12-07-2016 Intel Corporation
341 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
342 for currently supported client and server kernel versions.
343 * Server known to build on patched kernels:
344 2.6.32-431.29.2.el6 (RHEL6.5)
345 2.6.32-504.30.3.el6 (RHEL6.6)
346 2.6.32-573.26.1.el6 (RHEL6.7)
347 2.6.32-642.11.1.el6 (RHEL6.8)
348 3.10.0-514.el7 (RHEL7.3)
349 3.0.101-0.47.71 (SLES11 SP3)
350 3.0.101-80 (SLES11 SP4)
351 3.12.62-60.64.8 (SLES12 SP1)
352 vanilla linux 4.5.7 (ZFS only)
353 * Client known to build on unpatched kernels:
354 2.6.32-431.29.2.el6 (RHEL6.5)
355 2.6.32-504.30.3.el6 (RHEL6.6)
356 2.6.32-573.26.1.el6 (RHEL6.7)
357 2.6.32-642.11.1.el6 (RHEL6.8)
358 3.10.0-514.el7 (RHEL7.3)
359 3.0.101-0.47.71 (SLES11 SP3)
360 3.0.101-80 (SLES11 SP4)
361 3.12.62-60.64.8 (SLES12 SP1)
363 * Recommended e2fsprogs version: 1.42.13.wc5 or newer
364 * Recommended ZFS / SPL version: 0.6.5.8
365 * Tested with ZFS / SPL version: 0.6.5.8
366 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
367 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
368 stack overflow. For more information, please refer to bugzilla 17630.
369 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
370 the re-exporting nodes to be mounted with "32bitapi" mount option
372 --------------------------------------------------------------------------------
374 02-29-2016 Intel Corporation
376 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
377 for currently supported client and server kernel versions.
378 * Server known to build on patched kernels:
379 2.6.32-431.29.2.el6 (RHEL6.5)
380 2.6.32-504.30.3.el6 (RHEL6.6)
381 2.6.32-573.12.1.el6 (RHEL6.7)
382 3.10.0-327.3.1.el7 (RHEL7.2)
383 3.0.101-0.47.71 (SLES11 SP3)
384 3.0.101-68 (SLES11 SP4)
385 vanilla linux 4.2.1 (ZFS only)
386 * Client known to build on unpatched kernels:
387 2.6.32-431.29.2.el6 (RHEL6.5)
388 2.6.32-504.30.3.el6 (RHEL6.6)
389 2.6.32-573.12.1.el6 (RHEL6.7)
390 3.10.0-327.3.1.el7 (RHEL7.2)
391 3.0.101-0.47.71 (SLES11 SP3)
392 3.0.101-68 (SLES11 SP4)
395 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
396 * Recommended ZFS / SPL version: 0.6.4.2
397 * Tested with ZFS / SPL version: 0.6.4.2
398 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
399 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
400 stack overflow. For more information, please refer to bugzilla 17630.
401 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
402 the re-exporting nodes to be mounted with "32bitapi" mount option
404 --------------------------------------------------------------------------------
406 03-10-2015 Intel Corporation
408 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
409 for currently supported client and server kernel versions.
410 * Server known to build on patched kernels:
411 2.6.32-431.29.2.el6 (RHEL6.5)
412 2.6.32-504.8.1.el6 (RHEL6.6)
413 3.0.101-0.46 (SLES11 SP3)
414 * Client known to build on unpatched kernels:
415 2.6.32-431.29.2.el6 (RHEL6.5)
416 2.6.32-504.8.1.el6 (RHEL6.6)
417 3.10.0-123.20.1.el7 (RHEL7)
418 3.0.101-0.46 (SLES11 SP3)
419 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
420 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
421 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
422 stack overflow. For more information, please refer to bugzilla 17630.
423 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
424 the re-exporting nodes to be mounted with "32bitapi" mount option
426 Severity : enhancement
428 Description: control OST-index in IDIF via ROCOMPAT flag.
429 Details : Introduce new flag OBD_ROCOMPAT_IDX_IN_IDIF that is stored in the
430 last_rcvd file. For new formatted OST device, it will be auto set;
431 for the case of upgrading from old OST device, you can enable it
432 via the lproc interface osd-ldiskfs.index_in_idif. With such flag
433 enabled, for new created OST-object, its IDIF-in-LMA will contain
434 the OST-index; for the existing OST-object, the OSD will convert
435 old format IDIF as new format IDIF with OST-index stored in the
436 LMA EA when accessing such OST-object or via OI scrub. Once such
437 flag is enabled, it cannot be reverted back, so the system cannot
438 be downgraded to the orignal incompatible version.
440 --------------------------------------------------------------------------------
442 07-30-2014 Intel Corporation
444 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
445 for currently supported client and server kernel versions.
446 * Server known to build on patched kernels:
447 2.6.32-431.20.3.el6 (RHEL6.5)
448 3.0.101-0.31 (SLES11 SP3)
449 * Client known to build on unpatched kernels:
450 2.6.32-431.20.3.el6 (RHEL6.5)
451 3.0.101-0.31 (SLES11 SP3)
454 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
455 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
456 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
457 stack overflow. For more information, please refer to bugzilla 17630.
458 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
459 the re-exporting nodes to be mounted with "32bitapi" mount option
461 --------------------------------------------------------------------------------
463 10-31-2013 Intel Corporation
465 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
466 * Server builds for kernels:
467 2.6.32-358.18.1.el6 (RHEL6.4)
468 * Client builds for unpatched kernels:
469 2.6.32-358.18.1.el6 (RHEL6.4)
470 3.0.92-0.8 (SLES11 SP3)
472 * Recommended e2fsprogs version: 1.42.7.wc2 or newer
473 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
474 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
475 stack overflow. For more information, please refer to bugzilla 17630.
476 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
477 the re-exporting nodes to be mounted with "32bitapi" mount option
479 --------------------------------------------------------------------------------
481 05-30-2013 Intel Corporation
483 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
484 * Server builds for kernels:
485 2.6.32-358.6.2.el6 (RHEL6.4)
486 * Client builds for unpatched kernels:
487 2.6.32-358.6.2.el6 (RHEL6.4)
488 3.0.74-0.6.6 (SLES11 SP2)
490 * Recommended e2fsprogs version: 1.42.7.wc1 or newer
491 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
492 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
493 stack overflow. For more information, please refer to bugzilla 17630.
494 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
495 the re-exporting nodes to be mounted with "32bitapi" mount option
497 --------------------------------------------------------------------------------
499 9-30-2012 Whamcloud, Inc.
501 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
502 * Server support for kernels:
503 2.6.32-279.5.1.el6 (RHEL6)
504 * Client support for unpatched kernels:
505 2.6.32-279.5.1.el6 (RHEL6)
507 2.6.18-238.19.1.el5 (RHEL5)
508 2.6.16.60-0.69.1 (SLES10)
509 * Recommended e2fsprogs version: 1.41.90.wc4
510 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
511 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
512 stack overflow. For more information, please refer to bugzilla 17630.
513 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
514 the re-exporting nodes to be mounted with "32bitapi" mount option
516 Severity : enhancement
518 Description: Update RHEL6.3 kernel to 2.6.32-279.5.1.el6.
520 Severity : enhancement
522 Description: Update RHEL6.3 kernel to 2.6.32-279.2.1.el6.
524 Severity : enhancement
526 Description: Update RHEL6.2 kernel to 2.6.32-220.17.1.el6
528 Severity : enhancement
530 Description: Update RHEL6.2 kernel to 2.6.32-220.13.1.el6
532 Severity : enhancement
534 Description: Update RHEL6.2 kernel to 2.6.32-220.7.1.el6
536 --------------------------------------------------------------------------------
538 04-30-2012 Whamcloud, Inc.
540 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
541 * Server support for kernels:
542 2.6.32-220.4.2.el6 (RHEL6)
543 * Client support for unpatched kernels:
544 2.6.32-220.4.2.el6 (RHEL6)
546 2.6.18-238.19.1.el5 (RHEL5)
547 2.6.16.60-0.69.1 (SLES10)
548 * Recommended e2fsprogs version: 1.41.90.wc4
549 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
550 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
551 stack overflow. For more information, please refer to bugzilla 17630.
552 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
553 the re-exporting nodes to be mounted with "32bitapi" mount option
555 Severity : enhancement
557 Description: Update RHEL6.2 kernel to 2.6.32-220.4.2.el6
559 Severity : enhancement
561 Description: Update RHEL6.2 kernel to 2.6.32-220.4.1.el6
563 Severity : enhancement
565 Description: Update RHEL6.2 kernel to 2.6.32-220.el6
567 Severity : enhancement
569 Description: Update RHEL6.1 kernel to 2.6.32-131.17.1.el6
571 --------------------------------------------------------------------------------
573 09-30-2011 Whamcloud, Inc.
575 * Server support for kernels:
576 2.6.18-238.19.1.el5 (RHEL5)
577 2.6.32-131.6.1.el6 (RHEL6)
578 * Client support for unpatched kernels:
579 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
580 2.6.18-238.19.1.el5 (RHEL5)
581 2.6.32-131.6.1.el6 (RHEL6)
582 2.6.29.4-167.fc11 (FC11)
583 2.6.16.60-0.69.1 (SLES10)
585 * Recommended e2fsprogs version: 1.41.90.wc3
586 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
587 Lustre filesystem with 4K stack may cause a stack overflow. For more
588 information, please refer to bugzilla 17630.
589 * NFSv4 reexport requires Lustre client on the re-exporting nodes to be
590 mounted with "32bitapi" mount option
592 Severity : enhancement
593 Description: Update RHEL6.1 kernel to 2.6.32-131.6.1.el6
595 Severity : enhancement
596 Description: Update RHEL5.6 kernel to 2.6.18-238.19.1.el5
598 Severity : enhancement
599 Description: Update RHEL5.6 kernel to 2.6.18-238.9.1.el5
602 Frequency : only with SLES10
604 Description: Use OFED "KMP" provided by Novell
605 Details : SLES10 SP3 ships with OFED in a separate "KMP" package.
606 Lustre is now built against this package. That means you need to
607 install the ofed-kmp package from Novell for the patchless client.
608 Note that the ofed-kmp that Novell ships may not exactly match the
609 kernel version but should still be compatible.
611 Severity : enhancement
613 Description: Update RHEL5.5 kernel to 2.6.18-194.17.1.el5
614 Update OEL5.5 kernel to 2.6.18-194.17.1.0.1.el5.
616 --------------------------------------------------------------------------------
618 2010-07-15 Oracle, Inc.
620 * Server support for kernels:
621 2.6.18-164.11.1.el5 (RHEL 5)
622 2.6.18-164.11.1.0.1.el5 (OEL 5)
623 * Client support for unpatched kernels:
624 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
625 2.6.18-164.11.1.el5 (RHEL 5),
626 2.6.18-164.11.1.0.1.el5 (OEL 5)
627 2.6.16.60-0.42.8 (SLES 10),
629 2.6.29.4-167.fc11 (FC11)
630 * Recommended e2fsprogs version: 1.41.10-sun2
632 Severity : enhancement
634 Description: Distribute lfs_migrate script with lustre
635 Details : The lfs_migrate mentioned in the manual is now included in the
636 lustre tree in lustre/scripts and will distributed in lustre RPMs.
638 Severity : enhancement
640 Description: Failover nid list added to import proc file.
642 Severity : enhancement
644 Description: "som_preview" mount option required to use SOM
645 Details : The clients will need to use som_preview mount option
649 Description: Lustre does not do 1MB IOs to HW RAID
650 Details : Bump MAX_PHYS/HW_SEGMENTS and SG_ALL to 256 in the RHEL5 kernel.
652 Severity : enhancement
654 Description: lfs quota output is not very convenient for awk/sed-parsing
655 Details : Some positions in lfs quota output table could be empty or
656 non-empty which made it hard to parse it with scripts, now
657 a dash is put instead of space where there is not supposed
662 Description: Update to OFED 1.5.1
663 Details : OFED 1.5.1 will be used in environments that don't use
664 vendor supplied OFED.
668 Description: Remove lnet. prefix from lctl params display
669 Details : Also renamed lnet.memused to lnet_memused to
674 Description: Use conf_param -d to remove permanant settings.
678 Description: Add list_param -R
679 Details : lctl list_param now supports recursive mode.
682 Bugzilla : 21829, 22312
683 Description: Fix llobdstat.
684 Details : 1. Now updating the write_bytes counter
685 2. Port script fixes from 1.8.
689 Description: decrease the usage of memory on clients.
690 Details : 1. On clients, recycle dentries and inodes unused.
691 2. Delete the code related to ll_deathrow(att 6215 in bug 1443). It
696 Description: llobdstat fix and enhancement
697 Details : add a counter to set a limit to how many samples will be returned
698 fix a wildcard in the path to limit to obdfilter stats only
702 Description: lfs check only allowed for root
703 Details : Code cleanups and added a regresion test of non-root use of
708 Description: MDS Oopses in ptlrpc_dispatch_difficult_reply()/__list_add()
709 Details : Fix hr->hr_index overload.
713 Description: Fix FIEMAP functionality.
717 Description: FC11 patchless client support added
721 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and
722 OEL5.4 kernel to 2.6.18-164.11.1.0.1.el5.
726 Description: New acc-sm script and various test-framework enhancments.
730 Description: Previously landed patch was reverted.
734 Description: Update SLES10 kernel to 2.6.16.60-0.42.8 (patchless)
738 Description: Add ldiskfs maxdirsize mount option.
742 Description: Fix mkfs.lustre support for 16TB LUNs.
746 Description: Print mmp_check_interval
747 Details : print mmp_check_interval and make it possible to abort
748 mount operation in case it takes too long.
752 Description: Add cascading_rw to lustre mpi tests.
756 Description: Store SOM data in EA.
757 Details : Also several SOM related cleanups.
761 Description: Add refcount to md_open_data
765 Description: Use refcount on lov_request_set to prevent memory corruption.
769 Description: Statahead: use parent dir's i_mutex for synchronization.
773 Description: loadgen improvements.
777 Description: Enable peer health detection in ksocklnd and ko2iblnd.
781 Description: Control DCACHE_LUSTRE_INVALID flag with MDS_INODE_LOOKCUP lock
785 Description: Fix race on obd_devcie:md_stats
789 Description: Fix SLES 11 client builds.
790 Details : Resolved uninitalized pointer gcc warnings.
794 Description: Allow non-root access for "lfs check".
795 Details : Added a check in obd_class_ioctl() for OBD_IOC_PING_TARGET.
799 Description: Clear imp_force_reconnect correctly in ptlrpc_connect_interpret()
801 Severity : enhancement
803 Description: Update kernel to RHEL5.4 2.6.18-164.6.1.el5 and
804 OEL5 2.6.18-164.6.1.0.1.el5(Both in-kernel OFED enabled).
808 Description: Sanity test_27p fails.
809 Details : OBD_FAIL_OST_ENOSPC has been moved to filter_statfs().
810 Sanity tests 27 were cleaned up and improved.
814 Description: Unify req format on client/servers
815 Details : Use new req_capsule API [almost] everywhere instead of old PTLRPC
816 buffers and swabbers approach..
821 Description: drop join file stuff.
823 Severity : enhancement
825 Description: Add 2.6.27 and SLES11 (patchless client) support
829 Description: Fix in ptlrpc_expire_one_request() to print the signed time
832 Severity : enhancement
834 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
837 Severity : enhancement
839 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
841 Severity : enhancement
843 Description: correctly handle big reply message.
844 Details : send LNet event if reply is bigger than buffer and adjust this buffer
849 Description: Drop unnecessary __GFP_NOMEMALLOC flag from filter_get_page()
851 Severity : enhancement
853 Description: Add support for OEL5.
857 Description: Conf-sanity.sh 50g test - deactivated OST should not cause a panic.
859 Severity : enhancement
861 Description: "weak-modules" support
862 Details : Implement "weak-modules" support which enables kernel modules
863 to be used with any kernel that implements the same kABI. In
864 order to achieve this modules are now installed in
865 /lib/modules/$(uname -r)/updates/kernel on all distributions.
869 Description: client could not reconnect to OST because of an active request.
870 Details : The client now retries to reconnect to the same server, if a
871 connect request failed with EBUSY or -EAGAIN. Test case 35b
872 conf-sanity.sh has been created.
876 Description: Don't allow make backward step on assignin osc next id.
877 Details : race between allocation next id and ll_sync thread can be cause
878 of set wrong osc next id and can be kill valid ost objects.
882 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
886 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
890 Description: Error handling in osc_statfs_interpret() has been improved.
891 Details : Check in osc_statfs_interpret() for EBADR.
895 Description: Do not send statfs() requests to OSTs disabled by administrator.
896 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
898 Severity : enhancement
900 Description: Added sanity-quota test to for lfs(1) to work with numeric uid/gid.
902 Severity : enhancement
904 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
906 Severity : enhancement
908 Description: Enhancement for lfs(1) command to use numeric uid/gid.
912 Description: Improved cleanup procedure for conf-sanity test_46a.
916 Description: Do not start bulk transfer, if client tried to reconnect.
917 Details : Do not start bulk transfer, if client tried to reconnect,
918 but failed due to active rpcs. Check for exp_abort_active_req
919 has been implemented.
923 Description: Prevent inconsistences between linux and lustre mount structures.
924 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
925 Make the sleep interruptible.
927 Severity : enhancement
929 Description: provide server to client comms path
930 Details : server to client communications path via new LDLM_SET_INFO rpc
934 Description: truncate starts GFP_FS allocation under transaction causing deadlock
935 Details : ldiskfs_truncate calls grab_cache_page which may start page
936 allocation under an open transaction. This may lead to
937 calling prune_icache with consequent lustre reentrance.
941 Description: send correctly lsm on open replay
942 Details : MDS is trust to LSM size on replay open, but client can set wrong size
947 Description: enable client interface failover
948 Details : When a child reconnects from another NID, properly update export
949 nid hash position and ldlm reverse import.
951 Severity : enhancement
953 Description: Communicate OST degraded/readonly state via statfs to MDS
954 Details : Flags in the statfs returned from OSTs indicate whether the
955 OST is in a degraded RAID state, or if the filesystem has
956 turned read-only after a filesystem error is detected.
961 Description: don't panic if EPROTO was hit when reading symlink
962 Details : correctly handling request reference in error cases.
964 Severity : enhancement
966 Description: Add LustreNetLink, a kernel-userspace communcation path.
968 Severity : enhancement
970 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
975 Description: access to llog context before init.
976 Details : move handling CATALOGS file at osc layer and forbid access to llog
980 Frequency : rare, only when client is writing to OST on local node
982 Description: Avoid deadlock for local client writes
983 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
984 memory freeing context. This allows OST threads to set the
985 PF_MEMALLOC flag on task structures in order to allocate memory
986 from reserved pools and complete IO.
987 Use GFP_HIGHUSER for OST allocations for non-local client writes,
988 so that the OST threads generate memory pressure and allow
989 inactive pages to be reclaimed.
991 Severity : enhancement
993 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
998 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
999 Details : move ldlm namespace creation in setup phase to avoid grab
1000 _lprocfs_lock with cli_sem held.
1004 Description: Temporarily disable grant shrink.
1005 Details : Disable the feature for debugging.
1008 Frequency : only when formatting test filesystems
1010 Description: Unable to run several mkfs.lustre on loop devices at the same time
1011 Details : mkfs.lustre returns error 256 on the concurrent loop devices
1012 formatting. The solution is to proper handle the error.
1014 Severity : enhancement
1016 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
1018 Severity : enhancement
1020 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
1025 Description: resolve race between obd_disconnect and class_disconnect_exports
1026 Details : if obd_disconnect will be called to already disconnected export he
1027 forget release one reference and osc module can't unloaded.
1029 Severity : enhancement
1031 Description: Allow tuning service thread via /proc
1032 Details : For each service a new
1033 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
1034 created that can be used to set min/max thread counts, and get the
1035 current number of running threads.
1040 Descriptoin: don't return error if have partially created objects for file.
1041 Details : lov_update_create_set uses set->set_success as index for created
1042 objects, so if some requests will be failed, they will have hole
1043 at end of array and qos_shrink_lsm for allocate correct lsm.
1045 Severity : enhancement
1047 Description: Update OFED support to 1.4.1
1052 Description: Reduce small size read RPC
1053 Details : Set read-ahead limit for every file and only do read-ahead when
1054 available read-ahead pages are bigger than 1M to avoid small size
1057 Severity : enhancement
1059 Description: move AT tunable parameters for more consistent usage
1060 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
1062 Severity : enhancement
1064 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
1065 Details : allow skip disconnected ost for send statfs request and hide error
1071 Description: fix racy locking of mballoc block bitmaps causing BUG
1072 Details : The locking of the mballoc buddy bitmap and the in-memory
1073 block bitmap was using two different spin locks in some
1074 cases. This made it possible to incorrectly access the
1075 mballoc bitmap while another process was modifying it,
1076 causing a sanity assertion to fail. While no on-disk corruption
1077 was reported, there was some risk of this happening.
1080 Frequency : rare, on llog test 6
1082 Descriptoin: don't allow connect to already connected import
1083 Details : allowing connect to already connected import is hide connecting problem.
1088 Descriptoin: allow kill process which wait statahead result
1089 Details : for some reasons 'ls' can stick in waiting result from statahead,
1090 in this case need way for kill this process.
1092 Severity : enhancement
1094 Description: Add state history info file, enhance import info file
1095 Details : Track import connection state changes in a new osc/mdc proc file;
1096 add overview-type data to the osc/mdc import proc file.
1098 Severity : enhancement
1100 Description: MDS create should not wait for statfs RPC while holding DLM lock.
1102 Severity : enhancement
1104 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
1109 Descriptoin: OST grant too much space to client even there are not enough space.
1110 Details : Client will shrink its grant cache to OST if there are no write
1111 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
1112 this grant cache if there are already not enough avaible space
1113 (left_space < total_clients * 32M).
1116 Frequency : start MDS on uncleanly shutdowned MDS device
1118 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
1119 Details : stay in waiting mds<>ost recovery finished produce random bugs
1120 due race between two ll_sync thread for one lov target. send
1121 ACTIVATE event only if connect realy finished and import have
1125 Frequency : rare, connect and disconnect target at same time
1127 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
1128 Details : don't call obd_disconnect under lov_lock. this long time
1129 operation and can block ptlrpcd which answer to connect request.
1134 Descriptoin: don't lose wakeup for imp_recovery_waitq
1135 Details : recover_import_no_retry or invalidate_import and import_close can
1136 both sleep on imp_recovery_waitq, but we was send only one wakeup
1140 Frequency : always with long access acl
1142 Descriptoin: mds can't pack reply with long acl.
1143 Details : mds don't control size of acl but they limited by reint/getattr
1146 Severity : enhancement
1148 Description: Update to SLES10 kernel-2.6.16.60-0.33.
1150 Severity : enhancement
1152 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
1155 Frequency : start MDS on uncleanly shutdowned MDS device
1157 Descriptoin: aborting recovery hang on MDS
1158 Details : don't throttle destroy RPCs for the MDT.
1161 Frequency : on remount
1163 Description: external journal device not working after the remount
1164 Details : clear dev_rdonly flag for external journal devices in
1170 Description: shutdown vs evict race
1171 Details : client_disconnect_export vs connect request race.
1172 if client will evicted at this time - we start invalidate
1173 thread without referece to import and import can be freed
1177 Frequency : rare, need acl's on inode.
1179 Description: client can't handle ost additional correctly
1180 Details : if ost was added after client connected to mds client can have
1181 hit lnet_try_match_md ... to big messages to wide striped files.
1182 in this case need teach client to handle config events about add
1183 lov target and update client max ea size at that event.
1185 Severity : enhancement
1187 Description: Changelogs
1188 Details : Changelogs are a lightweight mechanism to track filesystem
1189 metadata and namespace changes. The changelog is recorded
1190 permanently on the MDTs, and is periodically "consumed" / purged
1191 when records are no longer needed.
1193 Severity : enhancement
1195 Description: compact fld format with extents
1196 Details : Store range of seq rather than every seq in FLD. Seq
1197 controller update FLD rather than clients. In Case of CMD, mdt0
1198 has FLD, all other metadata server act as non persistent proxy
1199 for FLD queries and cache fld entries in fld cache.
1204 Description: don't skip ost target if they assigned to file
1205 Details : Drop slow OSCs if we can, but not for requested start idx.
1206 This means "if OSC is slow and it is not the requested
1207 start OST, then it can be skipped, otherwise skip it only
1208 if it is inactive/recovering/out-of-space.
1212 Description: more cleanup in mds_lov
1213 Details : not send LOV EA under replay, we can't know about they size at this
1214 time. Don't allow client connect to mds before any ost connected,
1215 for avoid problems with LOV EA size and returning EIO to client.
1217 Severity : enhancement
1219 Description: Interoperability at server side (Disk interoperability)
1221 Severity : enhancement
1223 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
1225 Severity : enhancement
1227 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
1229 Severity : enhancement
1231 Description: New client IO stack (CLIO).
1233 Severity : enhancement
1235 Description: Commit on sharing. Eliminate inter-client dependencies between
1236 uncommitted transactions by doing transaction commits.
1237 Thereby clients may recovery independently.
1240 Frequency : Create a symlink file with a very long name
1242 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
1243 Details : If there is no extra space in the request for early cancels,
1244 ldlm_req_handles_avail() returns 0 instead of a negative value.
1246 Severity : enhancement
1248 Description: Add /proc entry for import status
1249 Details : The mdc, osc, and mgc import directories now have
1250 an import directory that contains useful import data for debugging
1251 connection problems.
1253 Severity : enhancement
1255 Description: Re-disable certain /proc logging
1256 Details : Enable and disable client's offset_stats, extents_stats and
1257 extents_stats_per_process stats logging on the fly.
1260 Frequency : Only on FC kernels 2.6.22+
1262 Description: oops in statahead
1263 Details : Do not drop reference count for the dentry from VFS when lookup,
1264 VFS will do that by itself.
1266 Severity : enhancement
1268 Description: Generic /proc file permissions
1269 Details : Set /Proc file permissions in a more generic way to enable non-
1270 root users operate on some /proc files.
1274 Description: Hitting mdc_commit_close() ASSERTION
1275 Details : Properly handle request reference release in
1276 ll_release_openhandle().
1280 Description: quota recovery deadlock during mds failover
1281 Details : This patch includes att18982, att18236, att18237 in bz14840.
1283 1. fix osts hang when mds does failover with quotaon
1284 2. prevent watchdog storm when osts threads wait for the
1289 Frequency : only patchless client
1290 Description: add workaround for race between add/remove dentry from hash
1292 Severity : enhancement
1294 Description: Allow OST glimpses to return PW locks
1298 Description: LBUG when llog conf file is full
1299 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
1303 Description: Prevent import from entering FULL state when server in recovery
1307 Description: service mount cannot take device name with ":"
1308 Details : Only when device name contains ":/" will mount treat it as
1314 Description: replace ptlrpcd with the statahead thread to interpret the async
1315 statahead RPC callback
1319 Frequency : on recovery
1320 Description: I/O failures after umount during fail back
1321 Details : if client reconnected to restarted server we need join to recovery
1322 instead of find server handler is changed and process self
1323 eviction with cancel all locks.
1325 Severity : enhancement
1327 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
1329 Severity : enhancement
1331 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
1333 Severity : enhancement
1335 Description: Upcall on Lustre log has been dumped
1336 Details : Allow for a user mode script to be called once a Lustre log has
1337 been dumped. It passes the filename of the dumped log to the
1338 script, the location of the script can be specified via
1339 /proc/sys/lnet/debug_log_upcall.
1344 Description: avoid idr_remove called for id which is not allocated.
1345 Details : Move assigment s_dev for clustered nfs to end of initialization,
1346 to avoid problem with error handling.
1351 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
1352 Details : When connection is reused this not moved from CONN_UNUSED_HASH
1353 into CONN_USED_HASH and this prodice warning when put connection
1354 again in unused hash.
1357 Severity : enhancement
1359 Description: File striping can now be set to use an arbitrary pool of OSTs.
1361 Severity : enhancement
1363 Description: Export bytes_read/bytes_write count on OSC/OST.
1367 Description: Early reply size mismatch, MGC loses connection
1368 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
1369 the connect flags are properly negotiated.
1374 Description: quotas are not honored with O_DIRECT
1375 Details : all writes with the flag O_DIRECT will use grants which leads to
1376 this problem. Now using OBD_BRW_SYNC to guard this.
1380 Description: add quota statistics
1381 Details : 1. sort out quota proc entries and proc code.
1382 2. add quota statistics
1384 Severity : enhancement
1386 Description: enable quota support for HEAD.
1390 Description: Properly propagate oinfo flags from lov to osc for statfs
1391 Details : restore missing copy oi_flags to lov requests.
1393 Severity : enhancement
1395 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
1396 : l_getgroups(8), lst(8), routerstat(8)
1397 Details : included man pages for llobdstat(8), llstat(8),
1398 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
1400 Severity : enhancement
1402 Description: configure's --enable-quota should check the
1403 : kernel .config for CONFIG_QUOTA
1404 Details : configure is terminated if --enable-quota is passed but
1405 : no quota support is in kernel
1409 Description: Remove portals compatibility
1410 Details : Remove portals compatibility, not interoperable with releases
1415 Description: Resolve device initialization race
1416 Details : Prevent proc handler from accessing devices added to the
1417 obd_devs array but yet be intialized.
1419 Severity : enhancement
1421 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
1423 Severity : enhancement
1425 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
1430 Description: Using wrong pointer in osc_brw_prep_request
1431 Details : Access to array[-1] can produce panic if kernel compiled with
1432 CONFIG_PAGE_ALLOC enabled
1436 Description: Client runs out of low memory
1437 Details : Consider only lowmem when counting initial number of llap pages
1441 Description: *optional* service tags registration
1442 Details : if the "service tags" package is installed on a Lustre node
1443 When the filesystem is mounted, a local-node service tag will
1444 be created. See http://inventory.sun.com/ for more information
1445 about the Service Tags asset management system.
1449 Description: Kernel BUG tries to release flock
1450 Details : Lustre does not destroy flock lock before last reference goes
1451 away. So always drop flock locks when client is evicted and
1452 perform unlock regardless of successfulness of speaking to MDS.
1456 Description: add refcount protection for osc callbacks, avoid panic on shutdown
1460 Description: sanity test 65a fails if stripecount of -1 is set
1461 Details : handle -1 striping on filesystem in ll_dirstripe_verify
1466 Description: ASSERTION(CheckWriteback(page,cmd)) failed
1467 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
1470 Severity : enhancement
1472 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
1476 Description: do not process already freed flock
1477 Details : flock can possibly be freed by another thread before it reaches
1478 to ldlm_flock_completion_ast.
1482 Description: LBUG during stress test
1483 Details : Need properly lock accesses the flock deadlock detection list.
1487 Description: oops in page fault handler
1488 Details : kernel page fault handler can return two special 'pages' in error
1489 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
1493 Description: timeout with invalidate import.
1494 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
1495 should be handled by ptlrpcd. This produce long age waiting and
1496 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
1498 Severity : enhancement
1500 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
1504 Description: enable MGS and MDT services start separately
1505 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
1506 a co-located MGS without starting the MGS, which is a complement
1507 to 'nosvc' mount option.
1511 Frequency : after recovery
1512 Description: precreate to many object's after del orphan.
1513 Details : del orphan st in oscc last_id == next_id and this triger growing
1514 count of precreated objects. Set flag LOW to skip increase count
1515 of precreated objects.
1519 Frequency : rare, on clear nid stats
1520 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
1521 Details : when clean nid stats sometimes try destroy live entry,
1522 and this produce panic in free.
1526 Description: Stack overflow during MDS log replay
1527 ease stack pressure by using a thread dealing llog_process.
1531 Description: wait until IO finished before start new when do lock cancel.
1532 Details : VM protocol want old IO finished before start new, in this case
1533 need wait until PG_writeback is cleared until check dirty flag and
1534 call writepages in lock cancel callback.
1536 Severity : enhancement
1538 Description: using special macro for print time and cleanup in includes.
1542 Description: mds_mfd_close() ASSERTION(rc == 0)
1543 Details : In mds_mfd_close(), we need protect inode's writecount change
1544 within its orphan write semaphore to prevent possible races.
1548 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
1552 Frequency : rare, on shutdown ost
1553 Description: don't hit live lock with umount ost.
1554 Details : shrink_dcache_parent can be in long loop with destroy dentries,
1555 use shrink_dcache_sb instead.
1559 Description: don't panic with use echo client
1560 Details : echo client pass NULL as client nid pointer and this produce null
1561 pointer dereference.
1565 Description: fix build on ppc32
1566 Details : compile code with -m64 flag produce wrong object file for ppc32.
1570 Description: add message levels for liblustreapi
1574 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
1575 Details : Move the 'good_osts' check before the 'total_bavail' check. This
1576 will result in an -EAGAIN and in the exit call path we call
1577 alloc_rr() which will with increasing aggressiveness attempt to
1578 aquire precreated objects on the minimum number of required OSCs.
1582 Description: Use old size assignment to avoid deadlock
1583 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
1584 the scheduling while holding a spinlock. We do not need locking
1585 for size in ll_update_inode() because size is only updated from
1586 the MDS for directories or files without objects, so there is no
1587 other place to do the update, and concurrent access to such inodes
1588 are protected by the inode lock.
1592 Description: resolve "_IOWR redefined" build error on SLES10
1596 Description: dump the memory debugging after all modules are unloaded to
1597 suppress false negative in conf_sanity test 39
1599 Severity : enhancement
1601 Description: build kernel-ib packages for OFED 1.3 in our release cycle
1606 Description: fix SLES kernel versioning
1607 Details : the kernel version for our SLES 10 kernel did not include a "-"
1608 before the "smp" at the end. while this was not a problem in
1609 general, it did mean that software trying to use the kernel
1610 version to try to detect a vendor specific kernel would fail.
1611 this was most evident by the OFED build scripts.
1615 Description: Don't update lov_desc members until making sure they are valid
1616 Details : When updating lov_desc members via proc fs, need fix their
1617 validities before doing the real update.
1621 Description: don't put request into delay list while invalidate in flight.
1622 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
1623 import in flight. this produce timeout for invalidate and sometimes
1624 can cause stale data.
1628 Frequency : on ppc only
1629 Description: not convert ost objects for directory because it's not exist.
1630 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
1634 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
1636 Severity : enhancement
1638 Description: Update to support 2.6.22.14 vanilla kernel.
1642 Frequency : rare, on recovery
1643 Description: read procfs can produce deadlock in some situation
1644 Details : Holding lprocfs lock which send rpc can produce block for destroy
1645 obd objects and this also block reconnect with -EALREADY. This
1646 isn't fix all lprocfs bugs - but make it rare.
1648 Severity : enhancement
1650 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
1653 Frequency : frequent on X2 node
1655 Description: mdc_set_open_replay_data LBUG
1656 Details : Set replay data for requests that are eligible for replay.
1660 Description: lustre_mgs: operation 101 on unconnected MGS
1661 Details : When MGC is disconnected from MGS long enough, MGS will evict the
1662 MGC, and late on MGC cannot successfully connect to MGS and a lot
1663 of the error messages complaining that MGS is not connected.
1667 Frequency : on network error
1668 Description: panic with double free request if network error
1669 Details : mdc_finish_enqueue is finish request if any network error ocuring,
1670 but it's true only for synchronus enqueue, for async enqueue
1671 (via ptlrpcd) this incorrect and ptlrpcd want finish request
1674 Severity : enhancement
1676 Description: client-side metadata stat-ahead during readdir(directory readahead)
1677 Details : perform client-side metadata stat-ahead when the client detects
1678 readdir and sequential stat of dir entries therein
1681 Frequency : on start mds
1683 Description: Implement get_info(last_id) in obdfilter.
1686 Frequency : occasional
1688 Description: Correctly check stale fid, not start epoch if ost not support SOM
1689 Details : open with flag O_CREATE need set old fid in op_fid3 because
1690 op_fid2 was overwritten with new generated fid, but mds can answer
1691 with one of these two fids and both is not stale. Setattr
1692 incorrectly started an epoch and assume will be called
1693 done_writing, but without SOM done_writing ever being called.
1696 Frequency : rare, depends on device drivers and load
1698 Description: MDS or OSS nodes crash due to stack overflow
1699 Details : Code changes in 1.8.0 increased the stack usage of some functions.
1700 In some cases, in conjunction with device drivers that use a lot
1701 of stack the MDS (or possibly OSS) service threads could overflow
1702 the stack. One change which was identified to consume additional
1703 stack has been reworked to avoid the extra stack usage.
1706 Frequency : occasional
1708 Description: Do not fail import if osc_interpret_create gets -EAGAIN
1709 Details : If osc_interpret_create got -EAGAIN it immediately exits and
1710 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
1711 oscc_has_objects and see OSC has no objests and call
1712 oscc_internal_create to resend create request.
1714 Severity : enhancement
1716 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
1718 Severity : enhancement
1720 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
1723 Frequency : very rare
1725 Description: Fix replay if there is an un-replied request and open
1726 Details : In some cases, older replay request will revert the
1727 mcd->mcd_last_xid on MDS which is used to record the client's
1728 latest sent request.
1730 Severity : enhancement
1732 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
1734 Severity : enhancement
1736 Description: Add rhel5 support to HEAD.
1738 Serverity : enhancement
1740 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
1745 Description: Don't allow skipping OSTs if index has been specified.
1746 Details : Don't allow skipping OSTs if index has been specified, make
1747 locking in internal create lots better.
1751 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
1752 Details : access to bitfield in structure is always rounded to long
1753 and this produce problem with not atomic change any bit.
1757 Description: Lustre make rpms failed.
1758 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
1759 builds Lustre rpms from tarball.
1762 Frequency : rare on shutdown ost
1764 Description: If llog cancel was not send before clean_exports phase, this can
1765 produce deadlock in llog code.
1766 Details : If llog thread has last reference to obd and call class_import_put
1767 this produce deadlock because llog_cleanup_commit_master wait when
1768 last llog_commit_thread exited, but this never success because was
1769 called from llog_commit_thread.
1773 Description: allow userland application know is lost one of stripes.
1774 Details : fill lvb_blocks with error code on ost and return it to
1775 application if error flag found.
1779 Description: NULL lov_tgts causing MDS oops
1780 Details : more safe checks for NULL lov_tgts for avoid oops.
1782 Severity : enhancement
1784 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
1788 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
1789 Details : If an OST is down the MDS will hang indefinitely in
1790 obd_statfs_rqset() waiting for the statfs data. While for
1791 MDS QOS usage of statfs, it should not stuck in waiting.
1793 Severity : enhancement
1795 Description: remote_acl support
1796 Details : Support ACL-based permission check for remote user.
1797 Support setfacl/getfacl for remote user with the utils
1798 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
1799 the system "{s,g}etfacl" utils.
1801 Severity : enhancement
1803 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
1805 Severity : enhancement
1807 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
1811 Description: make mgs_setparam() handle fsname containing dash
1812 Details : fsname containing a dash does not work with lctl conf_param
1814 Severity : enhancement
1816 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
1818 Severity : enhancement
1820 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
1824 Frequency : rare, at shutdown
1825 Description: access already free / zero obd_namespace.
1826 Details : if client_disconnect_export was called without force flag set,
1827 and exist connect request in flight, this can produce access to
1828 NULL pointer (or already free pointer) when connect_interpret
1829 store ocd flags in obd_namespace.
1833 Frequency : only at startup
1834 Description: not alloc memory with spinlock held.
1835 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
1836 if any spinlock held.
1838 Severity : enhancement
1840 Description: make lustre randomly fail allocating memory
1841 Details : Make lustre randomly failed allocating memory for testing purpose.
1843 Severity : enhancement
1845 Description: lost problems with lov objid file
1846 Details : Fixes some scability and access to not inited memory problems
1847 in work with lov objdid file.
1852 Description: lfs find does not continue on file error
1853 Details : Continue other files processing when a file/dir is absent.
1857 Description: Inconsistent usage of lustre_pack_reply()
1858 Details : Standardize the usage of lustre_pack_reply() such that it
1859 always generate a CERROR on failure.
1862 Frequency : occasional
1864 Description: MDS hang or stay in waiting lock
1865 Details : If client receive lock with CBPENDING flag ldlm need send lock
1866 cancel as separate rpc, to avoid situation when cancel request
1867 can't processed due all i/o threads stay in wait lock.
1871 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1875 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1879 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1883 Description: Update to RHEL4 latest kernel.
1885 Severity : enhancement
1887 Description: Build SLES10 patchless client fails
1888 Details : The configure was broken by run ./configure with
1889 --with-linux-obj=.... argument for patchless client. When the
1890 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1891 can't search header adequately. Use absolute path such as
1892 -I($LINUX)/include instead.
1896 Description: interrupt oig_wait produce painc on resend.
1897 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1898 private set, and this produce situation when rq_ptlrpcd_data not
1899 copyed to new allocated request and triggered LBUG on assert
1900 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1901 ptlrpcd set if request is changed and can be safety changed to use
1906 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1907 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1908 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1909 these macros and replace them with their difinitions instead.
1913 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1914 Details : Update which_patch & target file for SLES10 latest kernel.
1917 Description: Remove client patches from SLES 10 kernel.
1918 Details : This causes SLES 10 clients to behave as patchless clients
1919 even on a Lustre-patched (server) kernel.
1921 Severity : enhancement
1923 Description: self-adjustable client's lru lists
1924 Details : use adaptive algorithm for managing client cached locks lru
1925 lists according to current server load, other client's work
1926 pattern, memory activities, etc. Both, server and client
1927 side namespaces provide number of proc tunables for controlling
1930 Severity : enhancement
1932 Description: light-weight GSS support
1933 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
1934 and reduce performance overhead.
1936 Severity : enhancement
1938 Description: Linux keyring support
1939 Details : Support using service of Linux keyring for Lustre GSS internal
1940 context refresh/cache mechanism.
1944 Description: Fix errors in lfs documentation
1945 Details : Fixes man pages
1949 Description: don't use GFP_* in generic Lustre code.
1950 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
1954 Description: obdclass is limited by single OBD_ALLOC(idarray)
1955 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
1959 Description: warnings with build patchless client with vanila 2.6.19 and up
1960 Details : change old ctl_table style and replace ctl_table/ctl_table_header
1961 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
1965 Frequency : Only for SLES
1966 Description: sanity_quota fail test_1
1967 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
1968 file, which makes TSTID[2] inunique.
1972 Frequency : Only for SLES9
1973 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
1974 Details : Update target/ChangeLog/which_patch .
1978 Frequency : Only for test_10 in sanity-quota.sh
1979 Description: a bug in quota test code
1980 Details : checking a test flag in a wrong place
1984 Frequency : Only for test_13 in sanity-quota.sh
1985 Description: a bug in quota test script
1986 Details : shouldn't check overfull allocated quota
1990 Description: jbd statistics
1991 Details : Port older jbd statistics patch for sles10
1993 Severity : enhancement
1995 Description: Add printing inode info into message about error in writepage.
2000 Description: reply_lock_interpret crash due to race with it and lock cancel.
2001 Details : Do not replay locks that are being cancelled. Do not reference
2002 locks by their address during replay, just by their handle.
2006 Frequency : When flocks are used.
2007 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
2008 Details : Flock locks might destroy just granted lock if it could be merged
2009 with another existing flock, this is done in completion handler,
2010 so teach ldlm_cli_enquque_fini that this is a valid case for
2016 Description: Oops in read and write path when failing to allocate lock.
2017 Details : Check if lock allocation failed and return error back.
2021 Description: lstripe command fails for valid OST index
2022 Details : The stripe offset is compared to lov->desc.ld_tgt_count
2023 instead of lov->desc.ld_active_tgt_count.
2027 Description: sanity.sh failed test 103
2028 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
2033 Description: df doesn't work properly if diskfs blocksize != 4K
2034 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
2038 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
2039 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
2041 Severity : enhancement
2043 Description: Adaptive timeouts
2044 Details : RPC timeouts adapt to changing server load and network
2045 conditions to reduce resend attempts and improve recovery time.
2049 Description: llapi_file_create() does not allow some changes
2050 Details : add llapi_file_open() that allows specifying the mode and
2051 open flags, and also returns an open file handle.
2055 Description: merge and cleanup kernel patches.
2056 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
2060 Description: Add journal checksum support.(Kernel part)
2061 Details : The journal checksum feature adds two new flags i.e
2062 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
2063 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
2064 indicates that the commit block contains the checksum for
2065 the blocks described by the descriptor blocks. Now commit
2066 record can be sent to disk without waiting for descriptor
2067 blocks to be written to disk. This behavior is controlled
2068 using JBD2_FEATURE_ASYNC_COMMIT flag.
2072 Description: OSS needs mutliple precreate threads
2073 Details : Add ability to start more than one create thread per OSS.
2077 Description: Sanity.sh test_65k failed: setstripe should have succeeded
2078 Details : execute lfs setstripe on client
2082 Description: mds_obd_create error creating tmp object
2083 Details : When the user sets quota on root, llog will be affected and can't
2084 create files and write files.
2087 Frequency : Always on ia64 patchless client, and possibly others.
2089 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
2090 Details : This allows the patchless client to be loaded on architectures
2091 without this export.
2095 Description: RedHat Update kernel for RHEL5
2096 Details : Add the kernel config file for RHEL5.
2100 Description: RedHat Update kernel for RHEL5
2101 Details : Modify the kernel config file more closer RHEL5.
2105 Description: Build failure against Centos5 (RHEL5)
2106 Details : Define PAGE_SIZE when it isn't present.
2110 Description: test_7 on sanity-quota.sh can't be used on separate machines
2111 Details : add facet to handle it
2115 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
2116 Details : with NFS, the anon dentry's parent was set to itself in
2117 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
2118 corresponding dentry other than use rec->ur_name.
2120 Severity : enhancement
2122 Description: lfs setstripe enhancement
2123 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
2127 Description: enable data checksumming by default
2128 Details : enable checksum by default, allow --disable-checksum
2129 configure option and "-o nochecksum" mount option. Checksums
2130 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
2134 Description: lustre support for RHEL5
2135 Details : Add support for RHEL5.
2139 Description: Client eviction due to failover config
2140 Details : after a connection loss, the lustre client should attempt to
2141 reconnect to the last active server first before trying the
2142 other potential connections.
2146 Description: when mds and osts use different quota unit(32bit and 64bit),
2147 quota will be released repeatly.
2148 Details : void sending multiple quota reqs to mds, which will keep the
2149 status between the reqs.
2153 Description: osts not allocated evenly to files
2154 Details : change the condition to increase offset_idx
2157 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
2159 Description: Data corruption for OSTs that are formatted larger than 2TB
2161 Details : When generating the bio request for lustre file writes the
2162 sector number would overflow a temporary variable before being
2163 used for the IO. The data reads correctly from Lustre (which
2164 will overflow in a similar manner) but other file data or
2165 filesystem metadata may be corrupted in some cases.
2169 Description: Tune the kernel for good SCSI performance.
2170 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
2171 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
2176 Description: rewrite ext2-derived code in obdclass/uuid.c
2177 Details : rewrite inherited code (uuid parsing code from ext2 utils)
2178 from scratch preserving functionality.
2182 Description: "lfs find -obd UUID" prints directories
2183 Details : "lfs find -obd UUID" will return all directory names instead
2184 of just file names. It is incorrect because the directories
2185 do not reside on the OSTs.
2189 Description: use i_size_read and i_size_write in 2.6 port
2190 Details : replace inode->i_size access with i_size_read/write()
2193 Frequency : when removing large files
2195 Description: scheduling issue during removal of large Lustre files
2196 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
2197 It causes scheduling issues when removing large files (17TB in the
2201 Frequency : only with liblustre clients on XT3
2203 Description: evictions taking too long
2204 Details : allow llrd to evict clients directly on OSTs
2207 Frequency : only on ppc
2209 Description: /proc/fs/lustre/devices broken on ppc
2210 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
2211 We should make sure the type of 'index' is loff_t and then cast
2212 explicitly as needed below. Do not assign an explicitly cast
2216 Frequency : only for rhel5
2218 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
2219 Details : Modify the target file & which_kernel.
2221 Severity : enhancement
2223 Description: omit set fsid for export NFS
2224 Details : fix set/restore device id for avoid EMFILE error and mark lustre
2225 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
2229 Frequency : Always, for kernels after 2.6.16
2230 Description: Fix warning idr_remove called for id=.. which is not allocated.
2231 Details : Last kernels save old s_dev before kill super and not allow
2232 to restore from callback - restore it before call kill_anon_super.
2236 Description: buffer overruns could theoretically occur
2237 Details : llapi_semantic_traverse() modifies the "path" argument by
2238 appending values to the end of the origin string, and a
2239 overrun may occur. Adding buffer overrun check in liblustreapi.
2243 Description: Fix error on 'ls .' at the top of the Lustre mount.
2244 Details : Don't revalidate dentry if it is a root dentry.
2248 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
2249 Details : Modify vm-tunables-rhel4.patch.
2253 Description: Kernel config for 2.6.18-vanilla.
2254 Details : Modify targets/2.6-vanilla.target.in.
2255 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
2256 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
2257 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
2258 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
2262 Description: improve handling recoverable errors
2263 Details : if request processig with error which can be recoverable on server
2264 request should be resend, otherwise page released from cache and
2269 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
2270 Details : Modify target file & which_patch.
2271 A flaw was found in the IA32 system call emulation provided
2272 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
2273 value could be stored in the %RAX register, which could trigger an
2274 out-of-bounds system call table access. An untrusted local user
2275 could exploit this flaw to run code in the kernel
2276 (ie a root privilege escalation). (CVE-2007-4573).
2280 Description: O_DIRECT bypasses client statistics.
2281 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
2282 read_ahead_stats were updated. copied the stats section from
2283 osc_send_oap_rpc() into async_internal().
2287 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
2291 Description: change order of libsysio includes
2292 Details : '#include sysio.h' should always come before '#include xtio.h'
2296 Description: handle "serious error: objid * already exists" more gracefully
2297 Details : If LAST_ID value on disk is smaller than the objects existing in
2298 the O/0/d* directories, it indicates disk corruption and causes an
2299 LBUG(). If the object is 0-length, then we should use the existing
2300 object. This will help to avoid a full fsck in most cases.
2302 Severity : enhancement
2304 Description: adapt the lustre_config script to support the upgrade case
2305 Details : Add "-u" option for lustre_config script to support upgrading 1.4
2306 server targets to 1.6 in parallel.
2310 Description: To avoid grant space > avaible space when the disk is almost
2311 full. Without this patch you might see the error "grant XXXX >
2312 available" or some LBUG about grant, when the disk is almost
2314 Details : In filter_check_grant, for non_grant cache write, we should
2315 check the left space by if (*left > ungranted + bytes), instead
2316 of (*left > ungranted), because only we are sure the left space
2317 is enough for another "bytes", then the ungrant space should be
2318 increase. In client, we should update cl_avail_grant only there
2319 is OBD_MD_FLGRANT in the reply.
2323 Description: Update RHEL 4 kernel to fix local root privilege escalation.
2324 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
2325 described in CVE-2007-4573. This problem could allow untrusted
2326 local users to gain root access.
2329 Frequency : when using O_DIRECT and quotas
2331 Description: Incorrect file ownership on O_DIRECT output files
2332 Details : block usage reported by 'lfs quota' does not take into account
2333 files that have been written with O_DIRECT.
2338 Description: touch file failed when fs is not full
2339 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
2340 otherwise we can get ENOSP while fs is not full.
2344 Description: parallel lock callbacks
2345 Details : Instead of sending blocking and completion callbacks as separated
2346 requests, adding them to a set and sending in parallel.
2349 Frequency : only for Cray XT3
2350 Bugzilla : 12829/13455
2351 Description: Changing primary group doesn't change the group lustre assigns to
2353 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
2354 overridden with the primary group provided by the group upcall,
2355 whereas the supplied fsgid can be trusted if it is in the list of
2356 supplementary groups returned by the group upcall.
2358 Severity : enhancement
2360 Description: Allow masking D_WARNING, D_ERROR messages from console
2361 Details : Console messages can now be disabled via lnet.printk.
2365 Description: User code with malformed file open parameter crashes client node
2366 Details : Before packing join_file req, all the related reference should be
2367 checked carefully in case some malformed flags cause fake
2368 join_file req on client.
2372 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
2373 Details : ldlm_completion_ast() assumes that a lock is granted when the req
2374 mode is equal to the granted mode. However, it should also check
2375 that LDLM_FL_CP_REQD is not set.
2379 Description: Heavy nfs access might result in deadlocks
2380 Details : After ELC code landed, it is now improper to enqueue any mds
2381 locks under och_sem, because enqueue might want to decide to
2382 cancel open locks for same inode we are holding och_sem for.
2386 Description: Client eviction while running blogbench
2387 Details : A lot of unlink operations with concurrent I/O can lead to a
2388 deadlock causing evictions. To address the problem, the number of
2389 oustanding OST_DESTROY requests is now throttled to
2390 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
2391 callbacks are processed in priority.
2395 Description: enable ACLs on MDS by default
2396 Details : ACLs must be enabled on MDS by default.
2399 Frequency : PPC/PPC64 only
2401 Description: conflicts between asm-ppc64/types.h and lustre_types.h
2402 Details : fix duplicated definitions between asm-ppc64/types.h and
2403 lustre_types.h on PPC.
2406 Frequency : PPC/PPC64 only
2408 Description: asm-ppc/segment.h does not exist
2409 Details : fix compile issue on PPC.
2413 Description: better handle error messages in extents code
2416 Frequency : RHEL4 only
2418 Description: mkfs is very slow on IA64/RHEL4
2419 Details : A performance regression has been discovered in the MPT Fusion
2420 driver between versions 3.02.73rh and 3.02.99.00rh. As a
2421 consequence, we have downgraded the MPT Fusion driver in the RHEL4
2422 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
2424 Severity : enhancement
2426 Description: SNMP support enhancement
2427 Details : Adding total number of sampled request for an MDS node in snmp
2430 Severity : enhancement
2432 Description: Optimize ldlm waiting list processing for PR extent locks
2433 Details : When processing waiting list for read extent lock and meeting read
2434 lock that is same or wider to it that is not contended, skip
2435 processing rest of the list and immediatelly return current
2436 status of conflictness, since we are guaranteed there are no
2437 conflicting locks in the rest of the list.
2441 Description: Time out and refuse to reconnect
2442 Details : When the failover node is the primary node, it is possible
2443 to have two identical connections in imp_conn_list. We must
2444 compare not conn's pointers but NIDs, otherwise we can defeat
2445 connection throttling.
2449 Description: port llog fixes from b1_6 into HEAD
2450 Details : Port llog reference couting and some llog cleanups from b1_6
2451 (bug 10800) into HEAD, for protect from panic and access to already
2452 free llog structures.
2456 Description: Detect stride IO mode in read-ahead
2457 Details : When a client does stride read, read-ahead should detect that and
2458 read-ahead pages according to the detected stride pattern.
2462 Description: data checksumming impacts single node performance
2463 Details : add support for several checksum algorithm. Currently, only CRC32
2464 and Adler-32 are supported. The checksum type can be changed on
2465 the fly via /proc/fs/lustre/osc/*/checksum_type.
2469 Description: use adler32 for page checksums
2470 Details : when available, use the Adler-32 algorithm instead of CRC32 for
2475 Description: build for x2 fails
2476 Details : fix compile issue on Cray systems.
2480 Description: Properly match for duplicate locks
2481 Details : Due to different lock order from skiplists code, we need to
2482 traverse entire list for now
2485 Frequency : only on PPC/SLES10
2487 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
2488 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
2489 be defined. Add a hack in mkfs_lustre.c to work around this compile
2494 Description: LASSERT on MDS when client holding flock lock dies
2495 Details : ldlm pool logic depends on number of granted locks equal to
2496 number of released locks which is not true for flock locks, so
2497 just exclude such locks from consideration.
2501 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
2502 Details : Use fsfilt_sync() for both the whole filesystem sync and
2503 individual file sync to eliminate dangerous inode locking
2504 with I_LOCK that can lead to a deadlock.
2508 Description: performance in 1.6.3
2509 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
2510 sure that 1MB requests are not fragmented by the block layer.
2512 Severity : enhancement
2514 Description: organize the server-side client stats on per-nid basis
2515 Details : Change the structure of stats under obdfilter and mds to
2523 The "uuid"s file would list the uuids of _active_ exports.
2524 And the clear entry is to clear all stats and stale nids.
2526 Severity : enhancement
2528 Description: eliminate client locks in face of contention
2529 Details : file contention detection and lockless i/o implementation
2530 for contended files.
2534 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
2537 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
2539 Description: MDS may lose file striping (and hence file data) in some cases
2540 Details : If there are additional extended attributes stored on the MDS,
2541 in particular ACLs, SELinux, or user attributes (if user_xattr
2542 is specified for the client mount options) then there is a risk
2543 of attribute loss. Additionally, the Lustre file striping
2544 needs to be larger than default (e.g. striped over all OSTs),
2545 and an additional attribute must be stored initially in the
2546 inode and then increase in size enough to be moved to the
2547 external attribute block (e.g. ACL growing in size) for file
2552 Description: skiplist implementation simplification
2553 Details : skiplists are used to group compatible locks on granted list
2554 that was implemented as tracking first and last lock of each
2555 lock group the patch changes that to using doubly linked lists
2559 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
2560 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
2561 because LASSERT on both the data supplied by a client, and the
2562 data on disk is dangerous and incorrect.
2564 Severity : enhancement
2566 Description: Slow truncate/writes to huge files at high offsets.
2567 Details : Directly associate cached pages to lock that protect those pages,
2568 this allows us to quickly find what pages to write and remove
2569 once lock callback is received.
2573 Description: more ldlm soft lockups
2574 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
2575 starve other threads from the resource lock for a long time in
2576 case of long waiting queue, so change the debug level from
2577 D_OTHER to the less frequently used D_INFO.
2579 Severity : enhancement
2581 Description: add -gid, -group, -uid, -user options to lfs find
2585 Description: Hung threads in invalidate_inode_pages2_range
2586 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
2587 once one is completed. As a result, some RPCs are stuck in the
2588 queue and are never sent.
2592 Description: filter threads hungs on waiting journal commit
2593 Details : Cleanup filter group llog code, then only filter group llog will
2594 be only created in the MDS/OST syncing process.
2598 Description: Procfs and llog threads access destoryed import sometimes.
2599 Details : Sync the import destoryed process with procfs and llog threads by
2600 the import refcount and semaphore.
2602 Severity : enhancement
2604 Description: openlock cache of b1_6 port to HEAD
2609 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
2610 Details : If insertion of an extent fails, then discard the inode
2611 preallocation and free data blocks else it can lead to duplicate
2616 Description: don't always update ctime in ext3_xattr_set_handle()
2617 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
2618 In some cases the ctime should not be updated, for example for
2619 2.0->1.8 compatibility it is necessary to delete an xattr and it
2620 should not update the ctime.
2624 Bugzilla : 15713/16362
2625 Description: Assertion in iopen_connect_dentry in 1.6.3
2626 Details : looking up an inode via iopen with the wrong generation number can
2627 populate the dcache with a disconneced dentry while the inode
2628 number is in the process of being reallocated. This causes an
2629 assertion failure in iopen since the inode's dentry list contains
2630 both a connected and disconnected dentry.
2634 Description: assertion failure in ldlm_handle2lock()
2635 Details : fix a race between class_handle_unhash() and class_handle2object()
2636 introduced in lustre 1.6.5 by bug 13622.
2641 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
2642 Details : remove the limit of 256 scsi disks in the sd_iostat patch
2647 Description: oops in sd_iostats_seq_show()
2648 Details : unloading/reloading the scsi low level driver triggers a kernel
2649 bug when trying to access the sd iostat file.
2654 Description: Kernel panics during QLogic driver reload
2655 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
2656 patch, causing memory corruption.
2661 Description: journal_dev option does not work in b1_6
2662 Details : pass mount option during pre-mount.
2664 Severity : enhancement
2666 Description: Add a FIEMAP(FIle Extent MAP) ioctl
2667 Details : FIEMAP ioctl will allow an application to efficiently fetch the
2668 extent information of a file. It can be used to map logical blocks
2669 in a file to physical blocks in the block device.
2673 Description: LDLM soft lockups - improvement
2674 Details : It is be possible to send the lock handle along with each read
2675 or write request because the client is already doing a lock match
2676 itself so there isn't any reason the OST should have to re-do that
2680 Frequency : only on Cray X2
2682 Description: X2 build failures
2683 Details : fix build failures on Cray X2.
2687 Description: xid & resent requests
2688 Details : Initialize RPC XID from clock at startup (randomly if clock is
2691 Severity : enhancement
2693 Description: Add lustre_start utility to start or stop multiple Lustre servers
2698 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
2699 Details : In case of memory pressure, list_del() can be called twice on
2700 req->rq_history_list, causing a kernel oops.
2704 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
2705 Details : fix stack overflow in the distributed lock manager by defering
2706 export eviction after a failed AST to the elt thread instead of
2707 handling it in the dlm interpret routine.
2711 Description: Convert some comments to new format.
2712 Details : Update documenting comments to match doxygen conventions.
2716 Description: Grammar fixes.
2717 Details : A couple of trivial spelling fixes.
2721 Description: OSD_COUNTERS-mandatory
2722 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
2723 to simplify the code.
2727 Description: simplify cmm_device freeing logic.
2728 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
2732 Description: Add lockdep support to dt_object_operations locking interface.
2733 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
2734 parameter indicating lock ordering. Update mdd code to use new
2739 Description: Introduce failloc constants for lockless IO tests.
2740 Details : Add two new failloc constants to test lockless IO. Only one of
2741 them in implemented---another is checked in yet to be landed
2746 Description: Add lockdep support for inode mutex.
2747 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
2748 in the situations where Lustre has to lock more than one inode
2753 Description: Add optional invariants checking support.
2754 Details : Add new LINVRNT() macro, optional on new --enable-invariants
2755 configure switch. This macro is to be used for consistency and
2756 sanity checks that are too expensive to be left in `production'
2761 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
2762 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
2767 Description: Add lockdep support for ldlm_lock and ldlm_resource.
2768 Details : Use spin_lock_nested() in (the only) situation where more than
2769 one ldlm_lock is locked simultaneously. Also, fix possible
2770 dead-lock in ldlm_lock_change_resource() by enforcing particular
2775 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
2776 Details : Instead of specifying each ldlm_lock call-back through separate
2777 parameter, wrap them into struct ldlm_callback_suite.
2781 Description: Kill join_lru obd method and its callers.
2782 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
2783 in memory---a requirement implemented through ->o_join_lru() obd
2784 method in HEAD. Remove this method and its users.
2788 Description: Add asynchronous ldlm ENQUEUE completion handler.
2789 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
2790 case that stock ldlm_completion_ast() cannot handle as it waits
2791 until lock is granted. Introduce new ldlm_completion_ast_async()
2792 for this. Also comment ldlm_completion_ast().
2796 Description: ldlm_error <-> errno conversion.
2797 Details : Add functions to map (rather arbitrary) between LDLM error codes
2798 and standard errno values. CLIO needs this to prevent LDLM specific
2799 constants from escaping ldlm and osc.
2803 Description: Kill unused ldlm_handle2lock_ns() function.
2804 Details : Kill unused ldlm_handle2lock_ns() function.
2808 Description: Add lu_ref support to ldlm_lock
2809 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
2810 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
2811 and ldlm_resource. LDLM interface has to be changed, because code that
2812 releases a reference on a lock, has to "know" what reference this is.
2813 In the most frequent case
2815 lock = ldlm_handle2lock(handle);
2817 LDLM_LOCK_PUT(lock);
2819 no changes are required. When any other reference (received _not_ from
2820 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
2821 instead of LDLM_LOCK_PUT().
2823 Arguably, changes are pervasive, and interface requires some discipline
2824 for proper use. On the other hand, it was very instrumental in finding
2825 a few leaked lock references.
2829 Description: Add ldlm_lock_addref_try().
2830 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
2831 attempts to addref a lock that might be being canceled
2836 Description: Add ldlm_weigh_callback().
2837 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
2838 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
2839 hard-coded `number of pages' logic.
2843 Description: Add lockdep annotations to llog code.
2844 Details : Use appropriately tagged _nested() locking calls in the places
2845 where llog takes more than one ->lgh_lock lock.
2849 Description: Add loi_kms_set().
2850 Details : Wrap kms updates into a helper function.
2854 Description: Constify instances of struct lsm_operations.
2855 Details : Constify instances of struct lsm_operations.
2859 Description: lu_conf support.
2860 Details : On a server, a file system object is uniquely identified
2861 by a fid, which is sufficient to locate and load all object
2862 state (inode). On a client, on the other hand, more data are
2863 necessary instantiate an object. Change lu_object_find() and
2864 friends to take additional `lu_conf' argument describing object.
2865 Typically this includes layout information.
2869 Description: lu_context fixes.
2870 Details : Introduce new lu_context functions that are needed on the client
2871 side, where some system threads (ptlrpcd) are shared by multiple
2872 modules, and so cannot be stopped during module shutdown.
2876 Description: Add start and stop methods to lu_device_type_operations.
2877 Details : Introduce two new methods in lu_device_type_operations, that are
2878 invoked when first instance of a given type is created and last one
2879 is destroyed respectively. This is need by CLIO.
2883 Description: Add lu_ref support to struct lu_device.
2884 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
2885 track leaked references.
2889 Description: Introduce lu_kmem_descr.
2890 Details : lu_kmem_descr and its companion interface allow to create
2891 and destroy a number of kmem caches at once.
2895 Description: Fix lu_object finalization race.
2896 Details : Fix a race between lu_object_find() finding an object and its
2897 concurrent finalization. This race is (most likely) not possible
2898 on the server, but might happen on the client.
2902 Description: Introduce lu_ref interface.
2903 Details : lu_ref is a debugging module allowing to track references to
2904 a given object. It is quite cpu expensive, and has to be
2905 explicitly enabled with --enable-lu_ref. See usage description
2910 Description: Factor lu_site procfs stats into a separate function.
2911 Details : Separate lu_site stats printing code into a separate function
2912 that can be reused on a client.
2916 Description: Constify instances of struct {lu,dt,md}_device_operations.
2917 Details : Constify instances of struct {lu,dt,md}_device_operations.
2921 Description: Introduce struct md_site and move meta-data specific parts of
2922 struct lu_site here.
2923 Details : Move md-specific fields out of struct lu_site into special struct
2924 md_site, so that lu_site can be used on a client.
2928 Description: Kill mdd_lov_destroy().
2929 Details : Remove unused mdd code.
2933 Description: Add st_block checking to multistat.c.
2934 Details : Add st_block checking to multistat.c.
2938 Description: Add lu_ref support to struct obd_device.
2939 Details : Add lu_ref tracking to obd_device.
2943 Description: Kill obd_set_fail_loc().
2944 Details : Remove unused code.
2948 Description: Add special type for ptlrpc_request interpret functions.
2949 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
2950 there. Actual usage will be in CLIO.
2954 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
2955 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
2956 doesn't work with lockdep.
2960 Description: Add rwv.c test program.
2961 Details : New testing program exercising readv(2) and writev(2) (Qian).
2965 Description: Add sendfile.c test program.
2966 Details : New testing program exercising sendfile(2) (Jay).
2970 Description: Ratelimit a message that can be very frequent.
2971 Details : Ratelimit a memory allocation failure message that can
2976 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2977 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2981 Description: Liblustre build fixes.
2982 Details : Liblustre build fixes.
2986 Description: libcfs: add cfs_{need,cond}_resched() interface.
2987 Details : libcfs: add cfs_{need,cond}_resched() definition and
2988 implementations for Linux, NT, and liblustre.
2990 Severity : enhancement
2992 Description: More exported tunables for mballoc
2993 Details : Add support for tunable preallocation window and new tunables for
2994 large/small requests
2998 Description: Detect corruption of block bitmap and checking for preallocations
2999 Details : Checks validity of on-disk block bitmap. Also it does better
3000 checking of number of applied preallocations. When corruption is
3001 found, it turns filesystem readonly to prevent further corruptions.
3005 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
3006 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
3007 window to 0. If it does hit cache miss, and read-pattern is still
3008 stride-io mode, does not reset the stride window, but also does not
3009 increase the stride window length in this case.
3013 Frequency : only for big-endian servers
3014 Description: Check if system is big-endian while mounting fs with extents feature
3015 Details : Mounting a filesystem with extents feature will fail on big-endian
3016 systems since ext3-based ldiskfs is not supported on big-endian
3017 systems. This can be over-riden with "bigendian_extents" mount option.
3019 Severity : enhancement
3021 Description: The root squash functionality
3022 Details : A security feature, which is to prevent users from being able
3023 to mount lustre on their desktop, run as root, and delete
3024 all of the files in the filesystem. The goal is accomplished by
3025 remapping user id (UID) and group id (GID) of the root user to
3026 a UID and GID specified by the system administartor via Lustre
3027 configuration management server (MGS). The functionality also
3028 allows to specify sets of clients for which the remapping does
3033 Description: Excessive recovery window
3034 Details : With AT enabled, the recovery window can be excessively long (6000+
3035 seconds). To address this problem, we no longer use
3036 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
3037 timeout no longer depends on the service time, it is set to
3038 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
3039 time via pb_service_time.
3043 Description: Watchdog triggered on MDS failover
3044 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
3045 the OSTs know that the MDS "UUID" can be reused for the same export
3046 from a different NID, so we do not need to wait for the export to be
3050 Frequency : rare, only if using MMP with Linux RAID
3052 Description: MMP doesn't work with Linux RAID
3053 Details : While using HA for Lustre servers with Linux RAID, it is possible
3054 that MMP will not detect multiple mounts. To make this work we
3055 need to unplug the device queue in RAID when the MMP block is being
3056 written. Also while reading the MMP block, we should read it from
3057 disk and not the cached one.
3059 Severity : enhancement
3061 Description: open file using fid
3062 Details : A file can be opened using just its fid, like
3063 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
3066 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
3069 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
3070 Details : add proper locking for creation of jbd_4k slab cache
3074 Description: MMP check in ext3_remount() fails without displaying any error
3075 Details : When multiple mount protection fails during remount, proper error
3078 Severity : enhancement
3080 Description: Allow stripe size to be up to 4G-64k
3081 Details : Fix math logic to allow large stripe sizes.
3085 Description: add check for >8TB ldiskfs filesystems
3086 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
3087 Don't allow >8TB ldiskfs filesystems to be mounted without
3088 force_over_8tb mount option
3093 Descriptoin: OSS complains "too many blocks" and crashes
3094 Details : In filter_commitrw_write():
3095 1. lquota_chkquota();
3096 2. obdo_from_inode();
3097 3. lquota_pending_commit();
3098 "2." has changed uid and gid in this case, then caused
3099 lquota_chkquota() and lquota_pending_commit() to use
3100 different uid and gid. That is the root of the bug.
3104 Descriptoin: getstripe doesn't report stripe count = -1 sanely
3105 Details : Detect and print out the 'stripe count = -1' case properly.
3109 Description: Out or order replies might be lost on replay
3110 Details : In ptlrpc_retain_replayable_request if we cannot find retained
3111 request with tid smaller than one currently being added, add it
3112 to the start, not end of the list.
3114 --------------------------------------------------------------------------------
3116 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
3118 * Support for kernels:
3119 2.6.5-7.283 (SLES 9),
3120 2.6.9-55.EL (RHEL 4),
3121 2.6.16.46-0.14 (SLES 10),
3122 2.6.18.8 vanilla (kernel.org)
3123 * Client support for unpatched kernels:
3124 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
3125 2.6.16 - 2.6.22 vanilla (kernel.org)
3126 * Due to recently discovered recovery problems, we do not recommend
3127 using patchless RHEL 4 clients with this or any earlier release.
3128 * Recommended e2fsprogs version: 1.39.cfs8
3129 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
3130 * Starting with this release, the ldiskfs backing filesystem required
3131 by Lustre is now in its own package, lustre-ldiskfs. This package
3132 should be installed. It is versioned separately from Lustre and
3133 may be released separately in future.
3137 Description: block reactivating mgc import until all deactivates complete
3138 Details : Fix race when failing back MDT/MGS to itself (testing)
3141 Frequency : at statup only
3143 Description: mds_lov_synchronize race leads to various problems
3144 Details : simultaneous MDT->OST connections at startup can cause the
3145 sync to abort, leaving the OSC in a bad state.
3147 Severity : enhancement
3149 Description: add optional extra BUILD_VERSION info
3150 Details : add a new environment variable (namely LUSTRE_VERS) which allows
3151 to override the lustre version.
3154 Frequency : 2.6.18 servers only
3156 Description: ll_kern_mount() doesn't release the module reference
3157 Details : The ldiskfs module reference count never drops down to 0
3158 because ll_kern_mount() doesn't release the module reference.
3163 Description: server LBUG when using old ost_num_threads parameter
3164 Details : Accept the old ost_num_threads parameter but warn that it
3165 is deprecated, and fix an off-by-one error that caused an LBUG.
3170 Description: Transient SCSI error results in persistent IO issue
3171 Details : iobuf->dr_error is not reinitialized to 0 between two
3175 Frequency : sometimes when underlying device returns I/O errors
3177 Description: OSTs not going read-only during write failures
3178 Details : OSTs are not remounted read-only when the journal commit threads
3179 get I/O errors because fsfilt_ext3 calls journal_start/stop()
3180 instead of the ext3 wrappers.
3184 Description: poor connect scaling with increasing client count
3185 Details : Don't run filter_grant_sanity_check for more than 100 exports
3186 to improve scaling for large numbers of clients.
3189 Frequency : SLES10 only
3191 Description: sanity-quota.sh quotacheck failed: rc = -22
3192 Details : Quotas cannot be enabled on SLES10.
3195 Frequency : liblustre clients only
3197 Description: getdirentries does not give error when run on compute nodes
3198 Details : getdirentries does not fail when the size specified as an argument
3199 is too small to contain at least one entry
3201 Severity : enhancement
3203 Description: Add LNET router traceability for debug purposes
3204 Details : If a checksum failure occurs with a router as part of the
3205 IO path, the NID of the last router that forwarded the bulk data
3206 is printed so it can be identified.
3211 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
3212 Details : Due to a race condition, liblustre clients were occasionally
3213 evicted incorrectly.
3215 Severity : enhancement
3217 Description: lfs setstripe use optional parameters instead of postional
3220 Severity : enhancement
3222 Description: Nanosecond timestamp support for ldiskfs
3223 Details : The on-disk ldiskfs filesystem has added support for nanosecond
3224 resolution timestamps. There is not yet support for this at
3225 the Lustre filesystem level.
3228 Frequency : during server recovery
3230 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
3231 Details : request with rq_no_resend flag not awake l_wait_event if they get
3235 Frequency : nfs export on patchless client
3237 Description: connectathon hang when test nfs export over patchless client
3238 Details : Disconnected dentry cannot be found with lookup, so we do not need
3239 to unhash it or make it invalid
3242 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
3243 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
3244 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
3249 Description: open req refcounting wrong on reconnect
3250 Details : If reconnect happened between getting open reply from server and
3251 call to mdc_set_replay_data in ll_file_open, we will schedule
3252 replay for unreferenced request that we are about to free.
3253 Subsequent close will crash in variety of ways.
3254 Check that request is still eligible for replay in
3255 mdc_set_replay_data().
3260 Description: disable writes to filesystem when reading health_check file
3261 Details : the default for reading the health_check proc file has changed
3262 to NOT do a journal transaction and write to disk, because this
3263 can cause reads of the /proc file to hang and block HA state
3264 checking on a healthy but otherwise heavily loaded system. It
3265 is possible to return to the previous behaviour during configure
3266 with --enable-health-write.
3268 Severity : enhancement
3270 Description: 64-bit inode version
3271 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
3272 to the inode. This will be required for version-based recovery.
3277 Description: MDS fails to start if a duplicate client export is detected
3278 Details : in some rare cases it was possible for a client to connect to
3279 an MDS multiple times. Upon recovery the MDS would detect this
3280 and fail during startup. Handle this more gracefully.
3282 Severity : enhancement
3284 Description: Add -o localflock option to simulate old noflock
3286 Details : This will achieve local-only flock/fcntl locks
3292 Description: log_commit_thread vs filter_destroy race leads to crash
3293 Details : Take import reference before releasing llog record semaphore
3298 Description: Wrong request locking in request set processing
3299 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
3300 imp_delayed_list, in this place should be used imp_lock.
3303 Frequency : when reconnection
3305 Description: Grant Leak when osc reconnect to OST
3306 Details : When osc reconnect ost, OST(filter) should check whether it
3307 should grant more space to client by comparing fed_grant and
3308 cl_avail_grant, and return the granted space to client instead
3309 of "new granted" space, because client will call osc_init_grant
3310 to update the client grant space info.
3313 Frequency : when client reconnect to OST
3315 Description: Grant Leak when osc do resend and replay bulk write
3316 Details : When osc reconnect to OST, OST(filter)should clear grant info of
3317 bulk write request, because the grant info while be sync between
3318 OSC and OST when reconnect, and we should ignore the grant info
3319 these of resend/replay write req.
3324 Description: Grant space more than available left space sometimes.
3325 Details : When then OST is about to be full, if two bulk writing from
3326 different clients came to OST. Accord the avaliable space of the
3327 OST, the first req should be permitted, and the second one
3328 should be denied by ENOSPC. But if the seconde arrived before
3329 the first one is committed. The OST might wrongly permit second
3330 writing, which will cause grant space > available space.
3333 Frequency : when client is evicted
3335 Description: Grant might be wrongly erased when osc is evicted by OST
3336 Details : when the import is evicted by server, it will fork another
3337 thread ptlrpc_invalidate_import_thread to invalidate the
3338 import, where the grant will be set to 0. While the original
3339 thread will update the grant it got when connecting. So if
3340 the former happened latter, the grant will be wrongly errased
3341 because of this race.
3346 Description: Checking Stale with correct fid
3347 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
3348 to check whether it is stale, because sometimes, we want the
3349 enqueue happened anyway, and op_data.fid2 will not be initialized.
3351 Severity : enhancement
3353 Description: update patchless client
3354 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
3357 Frequency : only with 2.4 kernel
3359 Description: random memory corruption
3360 Details : size of struct ll_inode_info is to big for union inode.u and this
3361 can be cause of random memory corruption.
3366 Description: Memory leak in recovery
3367 Details : Lov_mds_md was not free in an error handler in mds_create_object.
3368 It should also check obd_fail before fsfilt_start, otherwise if
3369 fsfilt_start return -EROFS,(failover mds during mds recovery).
3370 then the req will return with repmsg->transno = 0 and rc = EROFS.
3371 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
3372 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
3373 be freed no matter whether fsfilt_commit success or not.
3376 Frequency : only with huge count clients
3378 Description: Prevents from taking the superblock lock in llap_from_page for
3380 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
3381 for llap_from_page call in ll_removepage prevents from taking the
3382 superblock lock for a soon died page.
3387 Description: Not check open intent error before release open handle
3388 Details : in some rare cases, the open intent error is not checked before
3389 release open handle, which may cause
3390 ASSERTION(open_req->rq_transno != 0), because it tries to release
3391 the failed open handle.
3396 Description: Set cat log bitmap only after create log success.
3397 Details : in some rare cases, the cat log bitmap is set too early. and it
3398 should be set only after create log success.
3402 Description: Accessing a block bevice can re-enable I/O when Lustre is
3403 tearing down a device.
3404 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
3409 Description: service threads may hog cpus when there are a lot of requests
3410 Details : Insert cond_resched to give other threads a chance to use some CPU
3415 Description: the cat log was not initialized in recovery
3416 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
3417 unlink log on mds will not be initialized until mds post
3418 recovery. And also in mds post recovery, the unlink log will
3419 initialization will be done asynchronausly, so there will be race
3420 between add unlink log and unlink log initialization.
3424 Description: brw_stats were being printed incorrectly
3425 Details : brw_stats were being printed as log2 but all of them were not
3426 recorded as log2. Also remove some code duplication arising from
3427 filter_tally_{read,write}.
3431 Frequency : rare, only in recovery.
3432 Description: ASSERTION(req->rq_type != LI_POISON) failed
3433 Details : imp_lock should be held while iterating over imp_sending_list for
3434 prevent destroy request after get timeout in ptlrpc_queue_wait.
3438 Description: replay-single.sh test 52 fails
3439 Details : A lock's skiplist need to be cleanup when it being unlinked
3440 from its resource list.
3444 Description: Short directio read returns full requested size rather than
3446 Details : Direct I/O operations should return actual amount of bytes
3447 transferred rather than requested size.
3449 Severity : enhancement
3451 Description: metadata RPC reduction (e.g. for rm performance)
3452 Details : decrease the amount of synchronous RPC between clients and servers
3453 by canceling conflicing lock before the operation on the client
3454 and packing thier handles into the main operation RPC to server.
3456 Severity : enhancement
3458 Description: Async OSC create to avoid the blocking unnecessarily.
3459 Details : If an OST has no remaining object, system will block on the
3460 creation when it needs to create a new object on this OST. Now,
3461 ways use pre-created objects when available, instead of blocking on
3462 an empty osc while others are not empty. If we must block, we
3463 block for the shortest possible period of time.
3467 Description: improve handling recoverable errors
3468 Details : if request processig with error which can be recoverable on server
3469 request should be resend, otherwise page released from cache and
3472 Severity : enhancement
3474 Description: refine locking for avoid write wrong info into lov_objid file
3475 Details : fix possible races with add new target and write/update data in
3478 --------------------------------------------------------------------------------
3480 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
3485 Frequency : on some architectures
3487 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
3488 Details : Uninitialized flags sometimes cause configuration commands to
3492 Frequency : patchless clients only
3494 Description: missing __iget() symbol export
3495 Details : The __iget() symbol export is missing. To avoid the need for
3496 this on patchless clients the deathrow inode reaper is turned
3497 off, and we depend on the VM to clean up old inodes. This
3498 dependency was during via the fix for bug 12181.
3500 --------------------------------------------------------------------------------
3502 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
3504 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
3505 INTEROPERATE with older versions automatically. In many cases a
3506 special upgrade step is needed. Please read the
3507 user documentation before upgrading any part of a live system.
3508 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
3509 version will not interoperate with 1.6 betas before beta5 (1.5.95).
3510 * WARNING: Lustre configuration and startup changes are required with
3511 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
3515 Severity : enhancement
3517 Description: MountConf
3518 Details : Lustre configuration is now managed via mkfs and mount
3519 commands instead of lmc and lconf. New obd types (MGS, MGC)
3520 are added for dynamic configuration management. See
3521 https://mail.clusterfs.com/wikis/lustre/MountConf for
3524 Severity : enhancement
3526 Description: dynamic OST addition
3527 Details : OSTs can now be added to a live filesystem
3529 Severity : enhancement
3531 Description: startup order invariance
3532 Details : MDTs and OSTs can be started in any order. Clients only
3533 require the MDT to complete startup.
3535 Severity : enhancement
3537 Description: parallel, asynchronous orphan cleanup
3538 Details : orphan cleanup is now performed in separate threads for each
3539 OST, allowing parallel non-blocking operation.
3541 Severity : enhancement
3543 Description: optimized stripe assignment
3544 Details : stripe assignments are now made based on ost space available,
3545 ost previous usage, and OSS previous usage, in order to try
3546 to optimize storage space and networking resources.
3548 Severity : enhancement
3550 Description: Permanently set tunables
3551 Details : All writable /proc/fs/lustre tunables can now be permanently
3552 set on a per-server basis, at mkfs time or on a live system.
3554 Severity : enhancement
3556 Description: Lustre message v2
3557 Details : Add lustre message format v2.
3559 Severity : enhancement
3561 Description: client OST exclusion list
3562 Details : Clients can be started with a list of OSTs that should be
3563 declared "inactive" for known non-responsive OSTs.
3567 Description: SPEC SFS validation failure on NFS v2 over lustre.
3568 Details : Changes the blocksize for regular files to be 2x RPC size,
3569 and not depend on stripe size.
3571 Severity : enhancement
3573 Description: Multiple MD RPCs in flight.
3574 Details : Further unserialise some read-only MDS RPCs - learn about intents.
3575 To avoid overly-overloading MDS, introduce a limit on number of
3576 MDS RPCs in flight for a single client and add /proc controls
3577 to adjust this limit.
3579 Severity : enhancement
3581 Description: client read/write statistics
3582 Details : Add client read/write call usage stats for performance
3583 analysis of user processes.
3584 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
3585 file access. extents_stats shows chunk size distribution.
3586 extents_stats_per_process show chunk size distribution per
3589 Severity : enhancement
3591 Description: mds statistics
3592 Details : Add detailed mds operations statistics in
3593 /proc/fs/lustre/mds/*/stats.
3597 Description: Failure of copying files with lustre special EAs.
3598 Details : Client side always return success for setxattr call for lustre
3599 special xattr (currently only "trusted.lov").
3604 Description: Refcount LNET uuids
3605 Details : The global LNET uuid list grew linearly with every startup;
3606 refcount repeated list entries instead of always adding to
3609 Severity : enhancement
3611 Description: Dynamic service threads
3612 Details : Within a small range, start extra service threads
3613 automatically when the request queue builds up.
3616 Frequency : mixed-endian client/server environments
3618 Description: mixed-endian crashes
3619 Details : The new msg_v2 system had some failures in mixed-endian
3622 Severity : enhancement
3624 Description: Easy OST removal
3625 Details : OSTs can be permanently deactivated with e.g. 'lctl
3626 conf_param lustre-OST0001.osc.active=0'
3628 Severity : enhancement
3630 Description: MGS proc entries
3631 Details : Added basic proc entries for the MGS showing what filesystems
3634 Severity : enhancement
3636 Description: provide MGS failover
3637 Details : Added config lock reacquisition after MGS server failover.
3639 Severity : enhancement
3641 Description: add Linux 2.4 support
3642 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
3646 Description: a large application tries to do I/O to the same resource and dies
3647 in the middle of it.
3648 Details : Check the req->rq_arrival time after the call to
3649 ost_brw_lock_get(), but before we do anything about
3650 processing it & sending the BULK transfer request. This
3651 should help move old stale pending locks off the queue as
3652 quickly as obd_timeout.
3655 Frequency : when an incorrect nid is specified during startup
3657 Description: ptlrpc connect to non-existant node causes kernel crash
3658 Details : LNET can't be re-entered from an event callback, which
3659 happened when we expire a message after the export has been
3660 cleaned up. Instead, hand the zombie cleanup off to another
3663 Severity : enhancement
3665 Description: plain/inodebits lock performance improvement
3666 Details : Grouping plain/inodebits in granted list by their request modes
3667 and bits policy, thus improving the performance of search through
3671 Frequency : only if OST filesystem is corrupted
3673 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3674 Details : for a short time RPCs with bulk IO are in the replay list,
3675 but replay of bulk IOs is unimplemented. If the OST filesystem
3676 is corrupted due to disk cache incoherency and then replay is
3677 started it is possible to trip an assertion. Avoid putting
3678 committed RPCs into the replay list at all to avoid this issue.
3681 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3683 Description: System hang on startup
3684 Details : This bug allowed the liblustre (e.g. catamount) client to
3685 return to the app before handling all startup RPCs. This
3686 could leave the node unresponsive to lustre network traffic
3687 and manifested as a server ptllnd timeout.
3689 Severity : enhancement
3691 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3692 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
3693 boolean to control whether to print peer debug info when a
3694 client's RPC times out.
3697 Frequency : only for kernels with patches from Lustre below 1.4.3
3699 Description: Remove old rdonly API
3700 Details : Remove old rdonly API which unused from at least lustre 1.4.3
3703 Frequency : only for devices with external journals
3705 Description: Set external device read-only also
3706 Details : During a commanded failover stop, we set the disk device
3707 read-only while the server shuts down. We now also set any
3708 external journal device read-only at the same time.
3711 Frequency : when upgrading from 1.4 while trying to change parameters
3713 Description: The wrong (new) MDC name was used when setting parameters for
3714 upgraded MDT's. Also allows changing of OSC (and MDC)
3715 parameters if --writeconf is specified at tunefs upgrade time.
3718 Frequency : when setting specific ost indicies
3720 Description: QOS code breaks on skipped indicies
3721 Details : Add checks for missing OST indicies in the QOS code, so OSTs
3722 created with --index need not be sequential.
3724 Severity : enhancement
3726 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3727 Details : The uninit_groups feature works in conjunction with the kernel
3728 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3729 up the pass1 processing of e2fsck. This is a read-only feature
3730 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3731 mount filesystems that have had this feature enabled.
3733 Severity : enhancement
3735 Description: Improve multi-block allocation algorithm to avoid fragmentation
3736 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3737 allocation locality and avoid filesystem fragmentation.
3739 ------------------------------------------------------------------------------
3741 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
3743 * Support for kernels:
3744 2.4.21-47.0.1.EL (RHEL 3)
3745 2.6.5-7.283 (SLES 9)
3746 2.6.9-42.0.10.EL (RHEL 4)
3747 2.6.12.6 vanilla (kernel.org)
3748 2.6.16.27-0.9 (SLES 10)
3749 * Recommended e2fsprogs version: 1.39.cfs5
3751 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3755 Frequency : occasional, depends on client load and configuration
3756 Bugzilla : 12181, 12203
3757 Description: data loss for recently-modified files
3759 Details : In some cases it is possible that recently written or created
3760 files may not be written to disk in a timely manner (this should
3761 normally be within 30s unless client IO load is very high).
3762 The problem appears as zero-length files or files that are a
3763 multiple of 1MB in size after a client crash or client eviction
3764 that are missing data at the end of the file.
3766 This problem is more likely to be hit on clients where files are
3767 repeatedly created and unlinked in the same directory, clients
3768 have a large amount of RAM, have many CPUs, the filesystem has
3769 many OSTs, the clients are rebooted frequently, and/or the files
3770 are not accessed by other nodes after being written.
3772 The presence of the problem can be detected by looking at
3773 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
3774 smaller than the second (nr_unused) then dirty files will not
3775 be flushed automatically to disk. "sync; sleep 10" should be
3776 run several times on the node before unmounting it to update
3777 Lustre (this is also safe to run on nodes without this problem).
3779 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
3780 that can cause this same problem, so customers using that kernel
3781 also need to update the kernel in addition to Lustre. In order
3782 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
3784 It is normal that files written just before a client crash (less
3785 than 30s) may not yet have been flushed to disk, even for local
3789 Frequency : frequent on thin XT3 nodes
3791 Description: UUID collision on thin XT3 Linux nodes
3792 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
3793 randomly, since we relied on an insufficiently-seeded PRNG.
3798 Description: OSS hangs after "All ost request buffers busy"
3799 Details : A deadlock between quota and journal operations caused OSS
3800 hangs after printing "All ost request buffers busy."
3803 Frequency : always on liblustre builds
3805 Description: Cleanup compiler warnings on liblustre
3808 Frequency : always on liblustre builds on XT3
3810 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
3814 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
3815 > 32000 subdirectories
3816 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
3817 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
3818 subdirectory count crosses 32000. This will aid e2fsck to
3819 correctly handle more than 32000 subdirectories.
3822 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3824 Description: System hang on startup
3825 Details : This bug allowed the liblustre (e.g. catamount) client to
3826 return to the app before handling all startup RPCs. This
3827 could leave the node unresponsive to lustre network traffic
3828 and manifested as a server ptllnd timeout.
3830 Severity : enhancement
3832 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3833 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
3834 boolean to control whether to print peer debug info when a
3835 client's RPC times out.
3840 Description: make O_SYNC working on 2.6 kernels
3841 Details : 2.6 kernels use different method for mark pages for write,
3842 so need add a code to lustre for O_SYNC work.
3847 Description: Failure to close file and release space on NFS
3848 Details : Put inode details into lock acquired in ll_intent_file_open.
3849 Use mdc_intent_lock in ll_intent_open to properly
3850 detect all kind of errors unhandled by mdc_enqueue.
3855 Description: proc file read during shutdown sometimes raced obd removal,
3857 Details : Add lock to prevent obd access after proc file removal.
3860 Frequency : Only for files larger than 4GB on 32-bit clients.
3862 Description: improperly doing page alignment of locks
3863 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
3864 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3869 Description: RPCs being resent when they shouldn't be
3870 Details : Some RPCs that should not be resent are being resent. This
3871 can cause inconsistencies in the RPC state machine. Do not
3872 resend such requests.
3875 Frequency : rare, only with NFS export
3877 Description: Crash on NFS re-export node
3878 Details : under very unusual load conditions an assertion is hit in
3879 ll_intent_file_open()
3882 Frequency : only if OST filesystem is corrupted
3884 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3885 Details : for a short time RPCs with bulk IO are in the replay list,
3886 but replay of bulk IOs is unimplemented. If the OST filesystem
3887 is corrupted due to disk cache incoherency and then replay is
3888 started it is possible to trip an assertion. Avoid putting
3889 committed RPCs into the replay list at all to avoid this issue.
3894 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3895 Details : Large single O_DIRECT read and write calls can fail to allocate
3896 a sufficiently large buffer to process the request. In case of
3897 allocation failure the allocation is retried with a smaller
3898 buffer and broken into smaller requests.
3900 Severity : enhancement
3902 Description: Add -o localflock option to simulate old noflock behaviour.
3903 Details : This will achieve local-only flock/fcntl locks coherentness.
3908 Description: versioning check is incomplete
3909 Details : Checking the version difference of client vs. server, report
3910 error if the gap is too big.
3915 Description: add support PG_writeback bit
3916 Details : add support for PG_writeback bit for Lustre, for more carefull
3917 work with page cache in 2.6 kernel. This also fix some deadlocks
3918 and remove hack for work O_SYNC with 2.6 kernel.
3920 Severity : enhancement
3922 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3923 Details : The uninit_groups feature works in conjunction with the kernel
3924 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3925 up the pass1 processing of e2fsck. This is a read-only feature
3926 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3927 mount filesystems that have had this feature enabled.
3929 Severity : enhancement
3931 Description: Improve multi-block allocation algorithm to avoid fragmentation
3932 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3933 allocation locality and avoid filesystem fragmentation.
3935 ------------------------------------------------------------------------------
3937 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3939 * Support for kernels:
3940 2.6.9-42.0.3.EL (RHEL 4)
3941 2.6.5-7.276 (SLES 9)
3942 2.4.21-47.0.1.EL (RHEL 3)
3943 2.6.12.6 vanilla (kernel.org)
3944 2.6.16.21-0.8 (SLES10)
3945 * Recommended e2fsprogs version: 1.39.cfs2-0
3947 * The backwards-compatible /proc/sys/portals symlink has been removed
3948 in this release. Before upgrading, please ensure that you change
3949 any configuration scripts or /etc/sysctl.conf files that access
3950 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3951 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3952 in advance of the upgrade on any system running Lustre 1.4.6 or
3953 newer, since /proc/sys/lnet was added in that version.
3954 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3958 Frequency : only when quota is used
3960 Description: avoid scanning export list for quota master
3961 Details : Change the algorithms to avoid scanning export list in order
3962 to improve the efficiency.
3965 Frequency : MDS failover only, very rarely
3967 Description: "went back in time" messages on mds failover
3968 Details : The greatest transno may be lost when the current operation
3969 finishes with an error (transno==0) and the client's last_rcvd
3970 record is over-written. Save the greatest transno in the
3971 mds_last_transno for this case.
3974 Frequency : always for specific kernels and striping counts
3976 Description: client may get "Matching packet too big" without ACL support
3977 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
3978 when trying to access files in certain configurations. The
3979 clients should in fact be denied when mounting because they do
3980 not understand ACLs.
3983 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
3985 Description: many clients connecting with IO in progress causes connect timeouts
3986 Details : Avoid synchronous journal commits to avoid delays caused by many
3987 clients connecting/disconnecting when bulk IO is in progress.
3988 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
3989 OST_IO_PORTAL to avoid delays behind potentially many pending
3993 Frequency : occasionally with multiple writers to a single file
3995 Description: shared writes to file may result in wrong size reported by stat()
3996 Details : Allow growing of kms when extent lock is cancelled
3999 Frequency : always with random mmap IO to multi-striped file
4001 Description: mmap write might be lost if we are writing to a 'hole' in stripe
4002 Details : Only if the hole is at the end of OST object so that kms is too
4003 small. Fix is to increase kms accordingly in ll_nopage.
4006 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
4008 Description: writes to a missing object would leak memory on the OST
4009 Details : If there is an inconsistency between the MDS and OST filesystems,
4010 such that the MDS references an object that doesn't exist, writes
4011 to that object will leak memory due to incorrect cleanup in the
4012 error handling path, eventually running out of memory on the OST.
4017 Description: Creating too long symlink causes lustre errors
4018 Details : Check symlink and name lengths before sending requests to MDS.
4021 Frequency : only if flock is enabled (not on by default)
4023 Description: posix locks not released on fd closure on 2.6.9+
4024 Details : We failed to add posix locks to list of inode locks on 2.6.9+
4025 kernels, this caused such locks not to be released on fd close and
4026 then assertions on fs unmount about still used locks.
4029 Frequency : MDS failover only, very rarely
4031 Description: clients may get ASSERTION(granted_lock != NULL)
4032 Details : When request was taking a long time, and a client was resending
4033 a getattr by name lock request. The were multiple lock requests
4034 with the same client lock handle and
4035 mds_getattr_name->fixup_handle_for_resent_request found one of the
4036 lock handles but later failed with ASSERTION(granted_lock != NULL).
4041 Description: handle->h_buffer_credits > 0, assertion failure
4042 Details : h_buffer_credits is zero after truncate, causing assertion
4043 failure. This patch extends the transaction or creates a new
4047 Frequency : NFS re-export or patchless client
4048 Bugzilla : 11179, 10796
4049 Description: Crash on NFS re-export node (__d_move)
4050 Details : We do not want to hash the dentry if we don't have a lock.
4051 But if this dentry is later used in d_move, we'd hit uninitialised
4052 list head d_hash, so we just do this to init d_hash field but
4053 leave dentry unhashed.
4056 Frequency : NFS re-export or patchless client
4058 Description: NFS exports has problem with symbolic link
4059 Details : lustre client didn't properly install dentry when re-exported
4060 to NFS or running patchless client.
4063 Frequency : NFS re-export or patchless client
4065 Description: Various nfs/patchless fixes.
4066 Details : fixes reuse disconected alias for lookup process - this fixes
4067 warning "find_exported_dentry: npd != pd",
4068 fix permission error with open files at nfs.
4069 fix apply umask when do revalidate.
4072 Frequency : occasional
4074 Description: Crash on NFS re-export node
4075 Details : calling clear_page() on the wrong pointer triggered oops in
4076 generic_mapping_read().
4079 Frequency : rarely, using O_DIRECT IO
4081 Description: unaligned directio crashes client with LASSERT
4082 Details : check for unaligned buffers before trying any requests.
4085 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
4087 Description: stale data returned from RAID cache
4088 Details : If only a small amount of IO is done to the RAID device before
4089 reading it again it is possible to get stale data from the RAID
4090 cache instead of reading it from disk.
4093 Frequency : always for sles10 kernel
4095 Description: sles10 support
4096 Details : ll_follow_link: compile fixes and using of nd_set_link
4097 under newer kernels.
4100 Frequency : depends on arch, kernel and compiler version, always on sles10
4103 Description: recursive or deep enough symlinks cause stack overflow
4104 Details : getting rid of large stack-allocated variable in
4108 Frequency : depends on hardware
4110 Description: lustre write performance loss in the SLES10 kernel
4111 Details : the performance loss is caused by using of write barriers in the
4112 ext3 code. The SLES10 kernel turns barrier support on by
4113 default. The fix is to undo that change for ldiskfs.
4115 ------------------------------------------------------------------------------
4117 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
4119 * Support for kernels:
4120 2.6.9-42.0.3EL (RHEL 4)
4121 2.6.5-7.276 (SLES 9)
4122 2.4.21-47.0.1.EL (RHEL 3)
4123 2.6.12.6 vanilla (kernel.org)
4127 Frequency : quota enabled and large files being deleted
4129 Description: releasing more than 4GB of quota at once hangs OST
4130 Details : If a user deletes more than 4GB of files on a single OST it
4131 will cause the OST to spin in an infinite loop. Release
4132 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4137 Description: statfs data retrieved from /proc may be stale or zero
4138 Details : When reading per-device statfs data from /proc, in the
4139 {kbytes,files}_{total,free,avail} files, it may appear
4140 as zero or be out of date.
4143 Frequency : systems with MD RAID1 external journal devices
4145 Description: lconf's call to blkid is confused by RAID1 journal devices
4146 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4147 of returning all block devices that match the journal UUID.
4150 Frequency : always, for aggregate stripe size over 4GB
4152 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
4153 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
4154 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
4157 Frequency : always if "lfs find" used on a local file/directory
4159 Description: "lfs find" segfaults if used on a local file/directory
4160 Details : The case where a directory component was not specified wasn't
4161 handled correctly. Handle this properly.
4164 Frequency : always on ppc64
4166 Description: the write to an ext3 filesystem mounted with mballoc got stuck
4167 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4168 perform endianness conversion.
4171 Frequency : rarely (truncate to non-zero file size after write under load)
4172 Bugzilla : 10730, 10687
4173 Description: Files padded with zeros to next 4K multiple
4174 Details : With filesystems mounted using the "extents" option (2.6 kernels)
4175 it is possible that files that are truncated to a non-zero size
4176 immediately after being written are filled with zero bytes beyond
4177 the truncated size. No file data is lost.
4179 Severity : enhancement
4181 Description: Allow recovery/failover for liblustre clients.
4182 Details : liblustre clients were unaware of failover configurations until
4185 Severity : enhancement
4187 Description: user file locks should fail when not mounting with flock option
4188 Details : Set up an error-returning stub in ll_file_operations.lock field
4189 to prevent incorrect behaviour when client is mounted without
4190 flock option. Also, set up properly f_op->flock field for
4194 Frequency : always on ia64
4196 Description: "lfs df" loops on printing out MDS statfs information
4197 Details : The obd_ioctl_data was not initialized and in some systems
4198 this caused a failure during the ioctl that did not return
4199 an error. Initialize the struct and return an error on failure.
4202 Frequency : SLES 9 only
4204 Description: Error of copying files with lustre special EAs as root
4205 Details : Client side always return success for setxattr call for lustre
4206 special xattr (currently only "trusted.lov").
4209 Frequency : rarely on clusters with both ia64+i386 clients
4211 Description: ia64+i686 clients doing shared IO on the same file may LBUG
4212 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
4213 clients are doing concurrent writes to the same file it is
4214 possible that the ia64 clients may LASSERT because the OST
4215 extent locks are not PAGE_SIZE aligned. Ensure that grown
4216 locks are always aligned on the request boundary.
4219 Frequency : specific use, occasional
4221 Description: Overwriting in use executable truncates on-disk binary image
4222 Details : If one node attempts to overwrite an executable in use by
4223 another node, we now correctly return ETXTBSY instead of
4224 truncating the file.
4226 Severity : enhancement
4228 Description: Async OSC create to avoid the blocking unnecessarily.
4229 Details : If an OST has no remaining object, system will block on the creating
4230 when need to create a new object on this OST. Now, ways use
4231 pre-created objects when available, instead of blocking on an
4232 empty osc while others are not empty. If we must block, we block
4233 for the shortest possible period of time.
4238 Description: chmod on Lustre root is propagated to other clients
4239 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
4240 invalid by the follow_mount time.
4243 Frequency : liblustre clients only
4245 Description: Race in 'instant cancel' lock handling could lead to such locks
4246 never to be granted in case of SMP MDS
4247 Details : Do not destroy not yet granted but cbpending locks in
4251 Frequency : replay/resend of open
4253 Description: non null lock assetion failure in mds_intent_policy
4254 Details : Trying to replay/resend lockless open requests resulted in
4255 mds_open() returning 0 with no lock. Now it sets a flag if
4256 a lock is going to be returned.
4258 Severity : enhancement
4260 Description: Checksum enhancements
4261 Details : New checksum enhancements allow for resending RPCs that failed
4264 Severity : enhancement
4266 Description: Tunables on number of dirty pages in cacche
4267 Details : Allow to set limit on number of dirty pages cached.
4272 Description: client crash on unmount - lock still has references
4273 Details : In some error handling cases it was possible to leak a lock
4274 reference on a client while accessing a file. This was not
4275 harmful to the client during operation, but would cause the
4276 client to crash when the filesystem is unmounted.
4279 Frequency : specific case, rare
4281 Description: ETXTBSY on mds though file not in use by client
4282 Details : ETXTBSY is no longer incorrectly returned when attempting to
4283 chmod or chown a directory that the user previously tried to
4284 execute or a currently-executing binary.
4287 Frequency : extremely rare except on liblustre-based clients
4289 Description: Lustre space not freed when files are deleted
4290 Details : Clean up open-unlinked files after client eviction. Previously
4291 the unlink was skipped and the files remained as orphans.
4296 Description: OST failure "would be an LBUG" in waiting_locks_callback()
4297 Details : In some cases it was possible to send a blocking callback to
4298 a client doing a glimpse, even though that client didn't get
4299 a lock granted. When the glimpse lock is cancelled on the OST
4300 the freed lock is left on the waiting list and corrupted the list.
4303 Frequency : all core dumps
4305 Description: Broke core dumps to lustre
4306 Details : Negative dentry may be unhashed if parent does not have UPDATE
4307 lock, but some callers, e.g. do_coredump, expect dentry to be
4308 hashed after successful create, hash it in ll_create_it.
4310 ------------------------------------------------------------------------------
4312 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
4314 * Support for kernels:
4315 2.6.9-42.0.2.EL (RHEL 4)
4316 2.6.5-7.276 (SLES 9)
4317 2.4.21-40.EL (RHEL 3)
4318 2.6.12.6 vanilla (kernel.org)
4322 Frequency : always on RHEL 3
4324 Description: Number of open files grows over time
4325 Details : The number of open files grows over time, whether or not
4326 Lustre is started. This was due to a filp leak introduced
4327 by one of our kernel patches.
4329 ------------------------------------------------------------------------------
4331 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4333 * Support for kernels:
4334 2.6.9-42.EL (RHEL 4)
4335 2.6.5-7.276 (SLES 9)
4336 2.4.21-40.EL (RHEL 3)
4337 2.6.12.6 vanilla (kernel.org)
4342 Bugzilla : 5719, 9635, 9792, 9684
4343 Description: OST (or MDS) trips assertions in (re)connection under heavy load
4344 Details : If a server is under heavy load and cannot reply to new
4345 connection requests before the client resends the (re)connect,
4346 the connection handling code can behave badly if two service
4347 threads are concurrently handing separate (re)connections from
4348 the same client. Add better locking to the connection handling
4349 code, and ensure that only a single connection will be processed
4350 for a given client UUID, even if the lock is dropped.
4352 Severity : enhancement
4354 Description: add TCP zero-copy support to kernel
4355 Details : Add support to the kernel TCP stack to allow zero-copy bulk
4356 sends if the hardware supports scatter-gather and checksumming.
4357 This allows socklnd to do client-write and server-read more
4358 efficiently and reduce CPU utilization from skbuf copying.
4361 Frequency : only if NFS exporting from client
4363 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
4364 Details : If setting attributes on a file created under NFS that had
4365 never been opened it would be possible to oops the client
4366 if the file had no objects.
4369 Frequency : always for liblustre
4371 Description: liblustre client does MDS+OSTs setattr RPC for each write
4372 Details : When doing a write from a liblustre client, the client
4373 incorrectly issued an RPC to the MDS and each OST the file was
4374 striped over in order to update the timestamps. When writing
4375 with small chunks and many clients this could overwhelm the MDS
4376 with RPCs. In all cases it would slow down the write because
4377 these RPCs are unnecessary.
4379 Severity : enhancement
4381 Description: allow number of MDS service threads to be changed at module load
4382 Details : It is now possible to change the number of MDS service threads
4383 running. Adding "options mds mds_num_threads={N}" to the MDS's
4384 /etc/modprobe.conf will set the number of threads for the next
4385 time Lustre is restarted (assuming the "mds" module is also
4386 reloaded at that time). The default number of threads will
4387 stay the same, 32 for most systems.
4392 Description: OST crash if filesystem is unformatted or corrupt
4393 Details : If an OST is started on a device that has never been formatted
4394 or if the filesystem is corrupt and cannot even mount then the
4395 error handling cleanup routines would dereference a NULL pointer.
4400 Description: NULL pointer deref in llap_from_page.
4401 Details : get_cache_page_nowait can return a page with NULL (or otherwise
4402 incorrect) mapping if the page was truncated/reclaimed while it was
4403 searched for. Check for this condition and skip such pages when
4404 doing readahead. Introduce extra check to llap_from_page() to
4405 verify page->mapping->host is non-NULL (so page is not anonymous).
4408 Frequency : Sometimes when using sys_sendfile
4410 Description: "page not covered by a lock" warnings from ll_readpage
4411 Details : sendfile called ll_readpage without right page locks present.
4412 Now we introduced ll_file_sendfile that does necessary locking
4413 around call to generic_file_sendfile() much like we do in
4417 Frequency : with certain MDS communication failures at client mount time
4419 Description: NULL pointer deref after failed client mount
4420 Details : a client connection request may delayed by the network layer
4421 and not be sent until after the PTLRPC layer has timed out the
4422 request. If the client fails the mount immediately it will try
4423 to clean up before the network times out the request. Add a
4424 reference from the request import to the obd device and delay
4425 the cleanup until the network drops the request.
4428 Frequency : occasionally during client (re)connect
4430 Description: assertion failure during client (re)connect
4431 Details : processing a client connection request may be delayed by the
4432 client or server longer than the client connect timeout. This
4433 causes the client to resend the connection request. If the
4434 original connection request is replied in this interval, the
4435 client may trip an assertion failure in ptlrpc_connect_interpret()
4436 which thought it would be the only running connect process.
4439 Frequency : only with obd_echo servers and clients that are rebooted
4441 Description: kernel BUG accessing uninitialized data structure
4442 Details : When running an obd_echo server it did not start the ping_evictor
4443 thread, and when a client was evicted an uninitialized data
4444 structure was accessed. Start the ping_evictor in the RPC
4445 service startup instead of the OBD startup.
4447 Severity : enhancement
4448 Bugzilla : 10193 (patchless)
4449 Description: Remove dependency on various unexported kernel interfaces.
4450 Details : No longer need reparent_to_init, exit_mm, exit_files,
4451 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
4454 Frequency : rare (only users of deprecated and unsupported LDAP config)
4456 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
4457 Details : LDAP apparently contains 'lustreName' attributes instead of
4458 'name'. A simple remapping of the name is sufficient.
4461 Frequency : rare (only with non-default dump_on_timeout debug enabled)
4463 Description: waiting_locks_callback trips kernel BUG if client is evicted
4464 Details : Running with the dump_on_timeout debug flag turned on makes
4465 it possible that the waiting_locks_callback() can try to dump
4466 the Lustre kernel debug logs from an interrupt handler. Defer
4467 this log dumping to the expired_lock_main() thread.
4469 Severity : enhancement
4471 Description: Support NFS exporting on 2.6 kernels.
4472 Details : Implement non-rawops metadata methods for NFS server to use without
4473 changing NFS server code.
4476 Frequency : very rare (synthetic metadata workload only)
4478 Description: two racing renames might cause an MDS thread to deadlock
4479 Details : Running the "racer" program may cause one MDS thread to rename
4480 a file from being the source of a rename to being the target of
4481 a rename at exactly the same time that another thread is doing
4482 so, and the second thread has already enqueued these locks after
4483 doing a lookup of the target and is trying to relock them in
4484 order. Ensure that we don't try to re-lock the same resource.
4487 Frequency : only very large systems with liblustre clients
4489 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
4490 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
4491 all OSTs in parallel. This allows the eviction of stale liblustre
4492 clients to proceed much faster than if they were done in series,
4493 and also offers similar improvements for other set_info RPCs.
4498 Description: excessive CPU usage during initial read phase on client
4499 Details : During the initial read phase on a client, it would agressively
4500 retry readahead on the file, consuming too much CPU and impacting
4501 performance (since 1.4.5.8). Improve the readahead algorithm
4502 to avoid this, and also improve some other common cases (read
4503 of small files in particular, where "small" is files smaller than
4504 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
4509 Description: MDS crash when receiving packet with unknown intent.
4510 Details : Do not LBUG in unknown intent case, just return -EFAULT
4512 Severity : enhancement
4513 Bugzilla : 9293, 9385
4514 Description: MDS RPCs are serialised on client. This is unnecessary for some.
4515 Details : Do not serialize getattr (non-intent version) and statfs.
4518 Frequency : occasional, when OST network is overloaded/intermittent
4520 Description: client evicted by OST after bulk IO timeout
4521 Details : If a client sends a bulk IO request (read or write) the OST
4522 may evict the client if it is unresposive to its data GET/PUT
4523 request. This is incorrect if the network is overloaded (takes
4524 too long to transfer the RPC data) or dropped the OST GET/PUT
4525 request. There is no need to evict the client at all, since
4526 the pinger and/or lock callbacks will handle this, and the
4527 client can restart the bulk request.
4530 Frequency : Always when mmapping file with no objects
4532 Description: client crashes when mmapping file with no objects
4533 Details : Check that we actually have objects in a file before doing any
4534 operations on objects in ll_vm_open, ll_vm_close and
4540 Description: Request leak when working with deleted CWD
4541 Details : Introduce advanced request refcount tracking for requests
4542 referenced from lustre intent.
4544 Severity : Enhancement
4546 Description: Cache open file handles on client.
4547 Details : MDS now will return special lock along with openhandle, if
4548 requested and client is allowed to hold openhandle, even if unused,
4549 until such a lock is revoked. Helps NFS a lot, since NFS is opening
4550 closing files for every read/write openration.
4552 Severity : Enhancement
4554 Description: Cache open negative dentries on client when possible.
4555 Details : Guard negative dentries with UPDATE lock on parent dir, drop
4556 negative dentries on lock revocation.
4561 Description: Remounting a client read-only wasn't possible with a zconf mount
4562 Details : It wasn't possible to remount a client read-only with llmount.
4564 Severity : enhancement
4565 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
4566 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
4567 MPICH is included to add Lustre-specific ADIO interfaces.
4568 This is based closely on the UFS ADIO layer and only differs
4569 in file creation, in order to allow the OST striping to be set.
4570 This is user-contributed code and not supported by CFS.
4575 Description: extended inode attributes (immutable, append-only) work improperly
4576 when 2.4 and 2.6 kernels are used on client/server or vice versa
4577 Details : Introduce kernel-independent values for these flags.
4579 Severity : enhancement
4582 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
4583 Details : Many of the /proc/ tunables can only be tuned at a megabyte
4584 granularity. Now, Fractional MB granularity is be supported,
4585 this is very useful for low memory system.
4587 Severity : enhancement
4589 Description: Getattr by fid
4590 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
4591 locks, avoids extra getattr rpc requests to MDS, allows '/' to
4592 have locks and avoids getattr rpc requests for it on every stat.
4595 Frequency : Always, for filesystems larger than 2TB
4597 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4598 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
4599 filesystems larger than 2TB to be mounted with mballoc enabled.
4602 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
4604 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4605 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
4606 filesystems larger than 2TB is mounted then the kernel will
4607 silently corrupt the start of the filesystem. CONFIG_LBD is
4608 enabled for all CFS-supported kernels, but the possibility of
4609 this happening with a modified kernel config exists.
4611 Severity : enhancement
4613 Description: add client O_DIRECT support for 2.6 kernels
4614 Details : It is now possible to do O_DIRECT reads and writes to files
4615 in the Lustre client mountpoint on 2.6 kernel clients.
4617 Severity : enhancement
4619 Description: parallel glimpse, setattr, statfs, punch, destroy requests
4620 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
4621 parallel, not waiting for response from every OST before sending
4622 a rpc to the next OST.
4627 Description: setattr vs write race when updating file timestamps
4628 Details : Client processes that update a file timestamp into the past
4629 right after writing to the file (e.g. tar) it is possible that
4630 the updated file modification time can be reset to the current
4631 time due to a race between processing the setattr and write RPC.
4633 Severity : enhancement
4635 Description: Bring 'lfs find' closer in line with regular Linux find.
4636 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
4637 -print0 options and obtains all the needed info through the lustre
4640 Severity : enhancement
4642 Description: support up to 1024 configured devices on one node
4643 Details : change obd_dev array from statically allocated to dynamically
4644 allocated structs as they are first used to reduce memory usage
4649 Description: Flush dirty partially truncated pages during truncate
4650 Details : Immediatelly flush partially truncated pages in filter_setattr,
4651 this way we completely avoid having any pages in page cache on OST
4652 and can retire ugly workarounds during writes to flush such pages.
4657 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
4658 Details : protect inode from truncation within vfs_unlink() context
4659 just take a reference before calling vfs_unlink() and release it
4660 when parent's i_sem is free.
4665 Description: last_id value checked outside lock on OST caused LASSERT failure
4666 Details : If there were multiple MDS->OST object precreate requests in
4667 flight, it was possible that the OST's last object id was checked
4668 outside a lock and incorrectly tripped an assertion. Move checks
4669 inside locks, and discard old precreate requests.
4672 Frequency : always, if extents are used on OSTs
4674 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
4675 Details : OSTs using the extents format would not zero the high 16 bits of
4676 the index physical block number. This is not a problem for any
4677 OST filesystems smaller than 16TB, and no kernels support ext3
4678 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
4679 new/modified files) and can be fixed for existing filesystems
4680 with e2fsprogs-1.39-cfs1.
4685 Description: import connection selection may be incorrect if timer wraps
4686 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
4687 import connections to be ignored if the 32-bit jiffies counter
4688 wraps. Use a 64-bit jiffies counter.
4691 Frequency : very large clusters immediately after boot
4693 Description: LNET request buffers exhausted under heavy short-term load
4694 Details : If a large number of client requests are generated on a service
4695 that has previously never seen so many requests it is possible
4696 that the request buffer growth cannot keep up with the spike in
4697 demand. Instead of dropping incoming requests, they are held in
4698 the LND until the RPC service can accept more requests.
4701 Frequency : Sometimes during replay
4703 Description: Assertion failure in ll_local_open after replay.
4704 Details : If replay happened on an open request reply before we were able
4705 to set replay handler, reply will become not swabbed tripping the
4706 assertion in ll_local_open. Now we set the handler right after
4707 recognising of open request
4710 Frequency : very rare
4712 Description: kernel reports "badness in vsnprintf"
4713 Details : Reading from the "recovery_status" /proc file in small chunks
4714 may cause a negative length in lprocfs_obd_rd_recovery_status()
4715 call to vsnprintf() (which is otherwise harmless). Exit early
4716 if there is no more space in the output buffer.
4718 Severity : enhancement
4720 Description: clear OBD RPC statistics by writing to them
4721 Details : It is now possible to clear the OBD RPC statistics by writing
4722 to the "stats" file.
4727 Description: Client mtime is not the same on different clients after utimes
4728 Details : In some cases, the client was using the utimes() syscall on
4729 a file cached on another node. The clients now validate the
4730 ctime from the MDS + OSTs to determine which one is right.
4735 Description: Inability to activate failout mode
4736 Details : lconf script incorrectly assumed that in pythong string's numeric
4737 value is used in comparisons.
4740 Frequency : always with multiple stripes per file
4742 Description: Inefficient object allocation for mutli-stripe files
4743 Details : When selecting which OSTs to stripe files over, for files with
4744 a stripe count that divides evenly into the number of OSTs,
4745 the MDS is always picking the same starting OST for each file.
4746 Return the OST selection heuristic to the original design.
4751 Description: mount failures may take full timeout to return an error
4752 Details : Under some heavy load conditions it is possible that a
4753 failed mount can wait for the full obd_timeout interval,
4754 possibly several minutes, before reporting an error.
4755 Instead return an error as soon as the status is known.
4757 Frequency : quota enabled and large files being deleted
4759 Description: releasing more than 4GB of quota at once hangs OST
4760 Details : If a user deletes more than 4GB of files on a single OST it
4761 will cause the OST to spin in an infinite loop. Release
4762 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4767 Description: statfs data retrieved from /proc may be stale or zero
4768 Details : When reading per-device statfs data from /proc, in the
4769 {kbytes,files}_{total,free,avail} files, it may appear
4770 as zero or be out of date.
4773 Frequency : systems with MD RAID1 external journal devices
4775 Description: lconf's call to blkid is confused by RAID1 journal devices
4776 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4777 of returning all block devices that match the journal UUID.
4780 Frequency : always, for aggregate stripe size over 4GB
4782 Description: assertion fails when trying to use 4GB stripe size
4783 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
4784 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
4787 Frequency : always on ppc64
4789 Description: the first write on an ext3 filesystem with mballoc got stuck
4790 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4791 perform endianness conversion.
4793 ------------------------------------------------------------------------------
4795 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4797 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
4798 INTEROPERATE with older versions automatically. Please read the
4799 user documentation before upgrading any part of a live system.
4800 * WARNING: Lustre networking configuration changes are required with
4801 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
4804 * Support for kernels:
4805 2.6.9-22.0.2.EL (RHEL 4)
4806 2.6.5-7.244 (SLES 9)
4807 2.6.12.6 vanilla (kernel.org)
4810 Severity : enhancement
4811 Bugzilla : 7981/8208
4812 Description: Introduced Lustre Networking (LNET)
4813 Details : LNET is new networking infrastructure for Lustre, it includes
4814 a reorganized network configuration mode (see the user
4815 documentation for full details) as well as support for routing
4816 between different network fabrics. Lustre Networking Devices
4817 (LNDs) for the supported network fabrics have also been
4818 created for this new infrastructure.
4820 Severity : enhancement
4821 Description: Introduced Access control lists
4822 Details : clients can set ACLs on files and directories in order to have
4823 more fine-grained permissions than the standard Unix UGO+RWX.
4824 The MDS must be started with the "-o acl" mount option.
4826 Severity : enhancement
4827 Description: Introduced filesystem quotas
4828 Details : Administrators may now establish per-user quotas on the
4831 Severity : enhancement
4833 Description: Configuration change for the XT3
4834 The PTLLND is now used to run Lustre over Portals on the XT3
4835 The configure option(s) --with-cray-portals are no longer used.
4836 Rather --with-portals=<path-to-portals-includes> is used to
4837 enable building on the XT3. In addition to enable XT3 specific
4838 features the option --enable-cray-xt3 must be used.
4843 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
4844 Details : A race between allocating a new llcd and re-getting the llcd_lock
4845 allowed another thread to grab newly-allocated llcd.
4847 Severity : enhancement
4849 Description: 2.6 OST async journal commit and locking fix to improve performance
4850 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
4851 2.6 kernels are now async as they already were in 2.4 kernels so
4852 that they can commit concurrently with the network bulk transfer.
4853 For block-allocated files the filter allocation semaphore is held
4854 to avoid filesystem fragmentation during allocation. BKL lock
4855 removed for 2.6 xattr operations where it is no longer needed.
4860 Description: lconf incorrectly determined whether two IP networks could talk
4861 Details : In some more complicated routing and multiple-network
4862 configurations, lconf will avoid trying to make a network
4863 connection to a disjoint part of the IP space. It was doing the
4864 math incorrectly for one set of cases.
4869 Description: Fix for potential infinite loop processing records in an llog.
4870 Details : If an llog record is corrupted/zeroed, it is possible to loop
4871 forever in llog_process(). Validate the llog record length
4872 and skip the remainder of the block on error.
4875 Frequency : occasional (liblustre only)
4877 Description: liblustre could not open files whose last component is a symlink
4878 Details : sysio_path_walk() would incorrectly pass the open intent to
4879 intermediate path components.
4882 Frequency : rare (liblustre only with non-standard tuning)
4883 Bugzilla : 7201 (7350)
4884 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4885 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4886 after it has been released, only do this for referenced locks
4888 Severity : enhancement
4890 Description: specifying an (invalid) directory default stripe_size of -1
4891 would reset the directory default striping
4892 Details : stripe_size -1 was used internally to signal directory stripe
4893 removal, now use "all default" to signal dir stripe removal
4894 as a directory striping of "all default" is not useful
4897 Frequency : common for large clusters running liblustre clients
4899 Description: doing an ls when liblustre clients are running is slow
4900 Details : sending a glimpse AST to a liblustre client waits for every AST
4901 to time out, as liblustre clients will not respond. Since they
4902 cannot cache data we refresh the OST lock LVB from disk instead.
4904 Severity : enhancement
4906 Description: doing an ls at the same time as file IO can be slow
4907 Details : enqueue and other "small" requests can be blocked behind many
4908 large IO requests. Create a new OST IO portal for non-IO
4909 requests so they can be processed faster.
4912 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4914 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4915 created file with mode 000
4916 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4917 allow an ftruncate() as a non-root user to a file with mode 000.
4918 Lustre now respects this flag to disable mode checks when
4919 truncating a file owned by the user
4922 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4925 Description: Revoking locks from clients that went dead or catatonic might take
4927 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4928 cancellation of such locks instant on servers without waiting for
4929 any reply from clients. Clients drops these locks when cancel
4930 notification from server is received without replying.
4933 Frequency : liblustre-only, when liblustre client dies or becomes busy
4935 Description: Doing ls on Linux clients can take a long time with active
4937 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4938 granting such locks to it in the first place if possible. Locks
4939 are taken by proxy on the OST during the read or write and
4940 dropped immediately afterward. Add connect flags handling, do
4941 not grant locks to liblustre clients for glimpse ASTs.
4943 Severity : enhancement
4945 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4946 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4947 sequential reads and behaves badly with applications doing random
4948 reads. Improve it by reading ahead at least read region, and
4949 avoiding excessive large RPC for small reads.
4951 Severity : enhancement
4953 Description: Creating more than 1000 files for a single job may cause a load
4954 imbalance on the OSTs if there are also a large number of OSTs.
4955 Details : qos_prep_create() uses an OST index reseed value that is an
4956 even multiple of the number of available OSTs so that if the
4957 reseed happens in the middle of the object allocation it will
4958 still utilize the OSTs as uniformly as possible.
4963 Description: OST or MDS may oops in ping_evictor_main()
4964 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
4965 but doesn't restart at beginning of obd_exports_timed list
4968 Severity : enhancement
4970 Description: improve by-nid export eviction on the MDS and OST
4971 Details : allow multiple exports with the same NID to be evicted at one
4972 time without re-searching the exports list.
4975 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
4977 Description: MDS may oops in groups_free()
4978 Details : in rare race conditions a newly allocated group_info struct is
4979 freed again, and this can be NULL. The 2.4 compatibility code
4980 for groups_free() checked for a NULL pointer, but 2.6 did not.
4983 Frequency : common for liblustre clients doing little filesystem IO
4984 Bugzilla : 9352, 7313
4985 Description: server may evict liblustre clients accessing contended locks
4986 Details : if a client is granted a lock or receives a completion AST
4987 with a blocking AST already set it would not reply to the AST
4988 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
4991 Frequency : lfs setstripe, only systems with more than 160 OSTs
4993 Description: unable to set striping with a starting offset beyond OST 160
4994 Details : llapi_create_file() incorrectly limited the starting stripe
4995 index to the maximum single-file stripe count.
4998 Frequency : LDAP users only
5000 Description: lconf did not handle in-kernel recovery with LDAP properly
5001 Details : lconf/LustreDB get_refs() is searching the wrong namespace
5003 Severity : enhancement
5005 Description: bind OST threads to NUMA nodes to improve performance
5006 Details : all OST threads are uniformly bound to CPUs on a single NUMA
5007 node and do their allocations there to localize memory access
5009 Severity : enhancement
5011 Description: llmount can determine client NID directly from Myrinet (GM)
5012 Details : the client NID code from gmnalnid was moved directly into
5013 llmount, removing the need to use this or specifying the
5014 client NID explicitly when mounting GM clients with zeroconf
5017 Frequency : if client is started with down MDS
5019 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
5020 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
5021 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
5022 ptlrpc_import_delayed_req() because we didn't check if the
5023 request was interrupted, and we also didn't break out of the
5024 event loop if there was no timeout
5029 Description: data loss during non-page-aligned writes to a single file from
5030 both multiple nodes and multiple threads on one node at same time
5031 Details : updates to KMS and lsm weren't protected by common lock. Resulting
5032 inconsistency led to false short-reads, that were cached and later
5033 used by ->prepare_write() to fill in partially written page,
5034 leading to data loss.
5037 Frequency : always, if lconf --abort_recovery used
5039 Description: lconf --abort_recovery fails with 'Operation not supported'
5040 Details : lconf was attempting to abort recovery on the MDT device and not
5043 Severity : enhancement
5045 Description: remove cleanup logs
5046 Details : replace lconf-generated cleanup logs with lustre internal
5047 cleanup routines. Eliminates the need for client-cleanup and
5050 Severity : enhancement
5052 Description: add support for EAs (user and system) on lustre filesystems
5053 Details : it is now possible to store extended attributes in the Lustre
5054 client filesystem, and with the user_xattr mount option it
5055 is possible to allow users to store EAs on their files also
5057 Severity : enhancement
5059 Description: Add possibility (config option) to show minimal available OST free
5061 Details : When compiled with --enable-mindf configure option, statfs(2)
5062 (and so, df) will return least minimal free space available from
5063 all OSTs as amount of free space on FS, instead of summary of
5064 free spaces of all OSTs.
5066 Severity : enhancement
5068 Description: do not expand extent locks acquired on OST-side
5069 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
5070 by server: they are not cached anyway.
5073 Frequency : when mmap is used/binaries executed from Lustre
5075 Description: Unmmap pages before throwing them away from read cache.
5076 Details : llap_shrink cache now attempts to unmap pages before discarding
5077 them (if unmapping failed - do not discard). SLES9 kernel has
5078 extra checks that trigger if this unmapping is not done first.
5083 Description: lconf didn't resolve symlinks before checking to see whether a
5084 given mountpoint was already in use
5087 Frequency : when migrating failover services
5088 Bugzilla : 6395, 9514
5089 Description: When migrating a subset of services from a node (e.g. failback
5090 from a failover service node) the remaining services would
5091 time out and evict clients.
5092 Details : lconf --force (implied by --failover) sets the global obd_timeout
5093 to 5 seconds in order to quickly disconnect, but this caused
5094 other RPCs to time out too quickly. Do not change the global
5095 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
5097 Severity : enhancement
5098 Frequency : if MDS is started with down OST
5099 Bugzilla : 9439,5706
5100 Description: Allow startup/shutdown of an MDS without depending on the
5101 availability of the OSTs.
5102 Details : Asynchronously call mds_lov_synchronize during MDS startup.
5103 Add appropriate locking and lov-osc refcounts for safe
5104 cleaning. Add osc abort_inflight calls in case the
5105 synchronize never started.
5108 Frequency : occasional (Cray XT3 only)
5110 Description: root not authorized to access files in CRAY_PORTALS environment
5111 Details : The client process capabilities were not honoured on the MDS in
5112 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
5113 been accessed by an authorized user then root was able to access
5114 the file on the local client also. The root user capabilities
5115 are now allowed on the MDS, as this environment has secure UID.
5118 Frequency : occasional
5120 Description: ldiskfs "too long searching" message happens too often
5121 Details : A debugging message (otherwise harmless) prints too often on
5122 the OST console. This has been reduced to only happen when
5123 there are fragmentation problems on the filesystem.
5128 Description: Division by zero in statfs when all OSCs are inactive
5129 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
5130 lov_statfs divides by value returned by lov_get_stripecnt().
5134 Bugzilla : 9489, 3273
5135 Description: First write from each client to each OST was only 4kB in size,
5136 to initialize client writeback cache, which caused sub-optimal
5137 RPCs and poor layout on disk for the first written file.
5138 Details : Clients now request an initial cache grant at (re)connect time
5139 and so that they can start streaming writes to the cache right
5140 away and always do full-sized RPCs if there is enough data.
5141 If the OST is rebooted the client also re-establishes its grant
5142 so that client cached writes will be honoured under the grant.
5147 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
5148 Details : Now I/O RPCs go to different portal number and (presumably) fast
5149 lock requests (and glimses) and other RPCs get their own service
5150 threads pool that should be able to service those RPCs
5153 Severity : enhancement
5155 Description: Ability to exchange lustre version between client and servers and
5156 issue warnings at client side if client is too old. Also for
5157 liblustre clients there is ability to refuse connection of too old
5159 Details : New 'version' field is added to connect data structure that is
5160 filled with version info. That info is later checked by server and
5164 Frequency : rare, liblustre only.
5165 Bugzilla : 9296, 9581
5166 Description: Two simultaneous writes from liblustre at offset within same page
5167 might proceed at the same time overwriting eachother with stale
5169 Details : I/O lock withing llu_file_prwv was released too early, before data
5170 actually was hitting the wire. Extended lock-holding time until
5171 server acknowledges receiving data.
5174 Frequency : extremely rare. Never observed in practice.
5176 Description: avoid generating lustre_handle cookie of 0.
5177 Details : class_handle_hash() generates handle cookies by incrementing
5178 global counter, and can hit 0 occasionaly (this is unlikely, but
5179 not impossible, because initial value of cookie counter is
5180 selected randonly). Value of 0 is used as a sentinel meaning
5181 "unassigned handle" --- avoid it. Also coalesce two critical
5182 sections in this function into one.
5184 Severity : enhancement
5186 Description: allow liblustre clients to delegate truncate locking to OST
5187 Details : To avoid overhead of locking, liblustre client instructs OST to
5188 take extent lock in ost_punch() on client's behalf. New connection
5189 flag is added to handle backward compatibility.
5191 Severity : enhancement
5192 Bugzilla : 4928, 7341, 9758
5193 Description: allow number of OST service threads to be specified
5194 Details : a module parameter allows the number of OST service threads
5195 to be specified via "options ost ost_num_threads={N}" in the
5196 OSS's /etc/modules.conf or /etc/modprobe.conf.
5200 Bugzilla : 6146, 9635, 9895
5201 Description: servers crash with bad pointer in target_handle_connect()
5202 Details : In rare cases when a client is reconnecting it was possible that
5203 the connection request was the last reference for that export.
5204 We would temporarily drop the export reference and get a new
5205 one, but this may have been the last reference and the export
5206 was just destroyed. Get new reference before dropping old one.
5208 Severity : enhancement
5209 Frequency : if client is started with failover MDS
5211 Description: Allow multiple MDS hostnames in the mount command
5212 Details : Try to read the configuration from all specified MDS
5213 hostnames during a client mount in case the "primary"
5216 Severity : enhancement
5218 Description: Stop sending data to evicted clients as soon as possible.
5219 Details : Check if the client we are about to send or are sending data to
5220 was evicted already. (Check is done every second of waiting,
5221 for which l_wait_event interface was extended to allow checking
5222 of exit condition at specified intervals).
5225 Frequency : rare, normally only when NFS exporting is done from client
5227 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
5229 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
5232 Frequency : always, if srand() is called before liblustre initialization
5234 Description: Liblustre uses system PRNG disturbing its usage by user application
5235 Details : Introduce internal to lustre fast and high-quality PRNG for
5236 lustre usage and make liblustre and some other places in generic
5237 lustre code to use it.
5239 Severity : enhancement
5240 Bugzilla : 9477, 9557, 9870
5241 Description: Verify that the MDS configuration logs are updated when xml is
5242 Details : Check if the .xml configuration logs are newer than the config
5243 logs stored on the MDS and report an error if this is the case.
5244 Request --write-conf, or allow starting with --old_conf.
5246 Severity : enhancement
5248 Description: Handle symlinks in the path when checking if Lustre is mounted.
5249 Details : Resolve intermediate symlinks when checking if a client has
5250 mounted a filesystem to avoid duplicate client mounts.
5255 Description: lconf can hit an error exception but still return success.
5256 Details : The lconf command catches the Command error exception at the top
5257 level script context and will exit with the associated exit
5258 status, but doesn't ensure that this exit status is non-zero.
5263 Description: failure of ptlrpc thread startup can cause oops
5264 Details : Starting a ptlrpc service thread can fail if there are a large
5265 number of threads or the server memory is very fragmented.
5266 Handle this without oopsing.
5269 Frequency : always, only if liblustre and non-default acceptor port was used
5271 Description: liblustre cannot connect to servers with non-default acceptor port
5272 Details : tcpnal_set_default_params() was not called and was therefore
5273 ignoring the environment varaible TCPNAL_PORT, as well as other
5274 TCPNAL_ environment variables
5279 Description: two objects could be created on the same OST for a single file
5280 Details : If an OST is down, in some cases it was possible to create two
5281 objects on a single OST for a single file. No problems other
5282 than potential performance impact and spurious error messages.
5286 Bugzilla : 5681, 9562
5287 Description: Client may oops in ll_unhash_aliases
5288 Details : Client dcache may become inconsistent in race condition.
5289 In some cases "getcwd" can fail if the current directory is
5295 Description: Inode refcounting problems in NFS export code
5296 Details : link_raw functions used to call d_instantiate without obtaining
5297 extra inode reference first.
5301 Bugzilla : 9942, 9903
5302 Description: Referencing freed requests leading to crash, memleaks with NFS.
5303 Details : We used to require that call to ll_revalidate_it was always
5304 followed by ll_lookup_it. Also with revalidate_special() it is
5305 possible to call ll_revalidate_it() twice for the same dentry
5306 even if first occurence returned success. This fix changes semantic
5307 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
5308 reference on a request referred from the intent.
5309 ll_intent_release() then releases such a request.
5312 Frequency : rare, normally benchmark loads only
5314 Description: unlinked inodes were kept in memory on the client
5315 Details : If a client is repeatedly creating and unlinking files it
5316 can accumulate a lot of stale inodes in the inode slab cache.
5317 If there is no other client load running this can cause the
5318 client node to run out of memory. Instead flush old inodes
5319 from client cache that have the same inode number as a new inode.
5322 Frequency : SLES9 2.6.5 kernel and long filenames only
5323 Bugzilla : 9969, 10379
5324 Description: utime reports stale NFS file handle
5325 Details : SLES9 uses out-of-dentry names in some cases, which confused
5326 the lustre dentry revalidation. Change it to always use the
5330 Frequency : rare, unless heavy write-truncate concurrency is continuous
5331 Bugzilla : 4180, 6984, 7171, 9963, 9331
5332 Description: OST becomes very slow and/or deadlocked during object unlink
5333 Details : filter_destroy() was holding onto the parent directory lock
5334 while truncating+unlinking objects. For very large objects this
5335 may block other threads for a long time and slow overall OST
5336 responsiveness. It may also be possible to get a lock ordering
5337 deadlock in this case, or run out of journal credits because of
5338 the combined truncate+unlink. Solution is to do object truncate
5339 first in one transaction without parent lock, and then do the
5340 final unlink in a new transaction with the parent lock. This
5341 reduces the lock hold time dramatically.
5344 Frequency : rare, 2.4 kernels only
5346 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
5347 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
5348 before unmounting their filesystem in order to avoid deadlock.
5349 The kernel_locked() function in 2.4 kernels only checks whether
5350 the kernel lock is held, not whether it is this process that is
5351 holding it as 2.6 kernels do.
5356 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
5357 Details : The client ptlrpc code may be trying to reconnect to a down
5358 server before a previous connection attempt has timed out.
5359 Increase the reconnect interval to be longer than the connection
5360 timeout interval to avoid sending duplicate connections to
5364 Frequency : echo_client brw_test command
5366 Description: fix echo_client to work with OST preallocated code
5367 Details : OST preallocation code (5137) didn't take echo_client IO path
5368 into account: echo_client calls filter methods outside of any
5369 OST thread and, hence, there is no per-thread preallocated
5370 pages and buffers to use. Solution: hijack pga pages for IO. As
5371 a byproduct, this avoids unnecessary data copying.
5375 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
5376 Description: Client can oops in mdc_commit_close() after open replay
5377 Details : It was possible for the MDS to return an open request with no
5378 transaction number in mds_finish_transno() if the client was
5379 evicted, but without actually returning an error. Clients
5380 would later try to replay that open and may trip an assertion
5381 Simplify the client close codepath, and always return an error
5382 from the MDS in case the open is not successful.
5385 Frequency : rare, 2.6 OSTs only
5387 Description: OST may deadlock under high load on fragmented files
5388 Details : If there was a heavy load and highly-fragmented OST filesystems
5389 it was possible to have all the OST threads deadlock waiting on
5390 allocation of biovecs, because the biovecs were not released
5391 until the entire RPC IO was completed. Instead, release biovecs
5392 as soon as they are complete to ensure forward IO progress.
5394 Severity : enhancement
5396 Description: Support for specifying external journal device at mount
5397 Details : If an OST or MDS device is formatted with an external journal
5398 device, this device major/minor is stored in the ext3 superblock
5399 and may not be valid for failover. Allow detecting and
5400 specifying the external journal at mount time.
5405 Description: Mounting an MDS with pending unlinked files may cause oops
5406 Details : target_finish_recovery() calls mds_postrecov() which returned
5407 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
5408 considers this an error and immediately begins cleaning up the
5409 lov, just after starting the mds_lov process
5411 Severity : enhancement
5413 Description: Implement 'lfs df' to report actual free space on per-OST basis
5414 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
5415 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
5416 usage of objects; '-h' to report in human readable format.
5418 ------------------------------------------------------------------------------
5420 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
5427 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
5428 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
5429 at startup during orphan unlinking. Instead, skip buddy update
5430 before setup, as it will be regenerated after recovery is complete.
5435 Description: If an OST is inactive, its locks might reference stale inodes.
5436 Details : lov_change_cbdata() must iterate over all namespaces, even if
5437 they are inactive to clear inode references from the lock.
5439 Severity : enhancement
5440 Frequency : occasional, if non-standard max_dirty_mb used
5442 Description: Client will block write RPCs if not enough grant
5443 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
5444 then the client will block writes while waiting for another RPC
5445 to complete instead of consuming its dirty limit. With change
5446 we get improved performance when max_dirty_mb is small.
5448 Severity : enhancement
5449 Bugzilla : 3389, 6253
5450 Description: Add support for supplementary groups on the MDS.
5451 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
5452 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
5453 lookups for user supplementary groups into a cache.
5457 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
5458 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
5459 for liblustre O_EXCL is enforced by MDS.
5463 Description: readdir never returns NULL in liblustre.
5464 Details : Corrected llu_iop_getdirentries logic, to return offset of next
5465 dentry in struct dirent.
5469 Frequency : liblustre only, depends on application IO pattern
5470 Description: liblustre clients evicted if not contacting servers
5471 Details : Don't put liblustre clients into the ping_evictor list, so
5472 they will not be evicted by the pinger ever.
5474 Severity : enhancement
5476 Description: Add ability to evict clients by NID from MDS.
5477 Details : By echoing "nid:$NID" string into
5478 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
5479 $NID would be instantly evicted from this MDS and from all active
5480 OSTs connected to it.
5484 Description: Do not query file size twice, somewhat slowing stat(2) calls.
5485 Details : lookup_it_finish() used to query file size from OSTs that was not
5490 Description: service threads change working directory to that of init
5491 Details : Starting lustre service threads may pin the working directory
5492 of the parent thread, making that filesystem busy. Threads
5493 now change to the working directory of init to avoid this.
5497 Frequency : during shutdown only
5498 Description: shutdown with a failed MDS or OST can cause unmount to hang
5499 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
5504 Frequency : 2.6 only
5505 Description: chmod/chown may include an extra supplementary group
5506 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
5507 supplementary group and if none is specified this is used.
5510 Bugzilla : 5479 (6816)
5512 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
5513 Details : If lookup is in progress on a file that is unlinked we might try
5514 to revalidate the inode and fail in revalidate after lookup is
5515 complete and ll_file_open() enqueues the open again but
5516 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
5519 Frequency : always, if lconf --abort_recovery used
5521 Description: lconf --abort_recovery fails with 'Operation not supported'
5522 Details : lconf was attempting to abort recovery on the MDT device and not
5525 ------------------------------------------------------------------------------
5527 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
5532 Frequency : rare (only unsupported configurations with a node running as an
5534 Bugzilla : 6514, 5137
5535 Description: Mounting a Lustre file system on a node running as an OST could
5537 Details : OSTs now preallocates memory needed to write out data at
5538 startup, instead of when needed, to avoid having to
5539 allocate memory in possibly low memory situations.
5540 Specifically, if the file system is mounted on on OST,
5541 memory pressure could force it to try to write out data,
5542 which it needed to allocate memory to do. Due to the low
5543 memory, it would be unable to do so and the node would
5544 become unresponsive.
5546 Severity : enhancement
5548 Description: Addition of lconf --service command line option
5549 Details : lconf now accepts a '--service <arg>' option, which is
5550 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
5552 Severity : enhancement
5554 Description: Failover mode is now the default for OSTs.
5555 Details : By default, OSTs will now run in failover mode. To return to
5556 the old behaviour, add '--failout' to the lmc line for OSTs.
5558 Severity : enhancement
5560 Description: Health checks are now provided for MDS and OSTs
5561 Details : Additional detailed health check information on MSD and OSTs
5562 is now provided through the procfs health_check value.
5565 Frequency : occasional, depends on IO load
5567 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
5568 after numerous create/delete cycles
5569 Details : The ext3 inode allocation policy would not allocate new inodes
5570 very well on the OSTs because there are no new directories
5571 being created. Instead we look for groups with free space if
5572 the parent directories are nearly full.
5577 Description: Network or server problems during mount may cause partially
5578 mounted clients instead of returning an error.
5579 Details : The config llog parsing code may overwrite the error return
5580 code during mount error handling, returning success instead
5586 Description: MDS can fail to allocate large reply buffers
5587 Details : After long uptimes the MDS can fail to allocate large reply
5588 buffers (e.g. zconf client mount config records) due to memory
5589 fragmentation or consumption by the buffer cache. Preallocate
5590 some large reply buffers so that these replies can be sent even
5591 under memory pressure.
5595 Frequency : rare (liblustre)
5596 Description: fsx running with liblustre complained that using truncate() to
5597 extend the file doesn't work. This patch corrects that issue.
5598 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
5599 ATTR_SIZE and lsm use in llu_setattr_raw.
5603 Frequency : rare, only 2.6 kernels
5604 Description: Unusual file access patterns on the MDS may result in inode
5605 data being lost in very rare circumstances.
5606 Details : Bad interaction between the ea-in-inode patch and the "no-read"
5607 code in the 2.6 kernel caused the inode and/or EA data not to
5608 be read from disk, causing single-file corruption.
5612 Frequency : rare, only 2.6 filesystems using extents
5613 Description: Heavy concurrent write and delete load may cause data corruption.
5614 Details : It was possible under high-load situations to have an extent
5615 metadata block in the block device cache from a just-unlinked
5616 file overwrite a newly-allocated data block. We now unmap any
5617 metadata buffers that alias just-allocated data blocks.
5621 Frequency : filesystems with default stripe_count larger than 77
5622 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
5623 Details : lconf specifies an inode size of 4096 bytes when the default
5624 stripe_count is larger than 77. This conflicts with the default
5625 inode density of 1 per 4096 bytes. Allocate smaller inodes in
5626 this case to avoid pinning too much memory for large EAs.
5628 ------------------------------------------------------------------------------
5630 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
5635 Frequency : rare (extremely heavy IO load with hundreds of clients)
5637 Description: Client is evicted, gets IO error writing to file
5638 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
5639 caused clients to be evicted for AST timeouts. The fixes in
5640 bug 5192 mean we no longer need to have such short AST timeouts
5641 so ldlm_timeout has been increased.
5644 Frequency : occasional during --force or --failover shutdown under load
5645 Bugzilla : 5949, 4834
5646 Description: Server oops/LBUG if stopped with --force or --failover under load
5647 Details : a collection of import/export refcount and cleanup ordering
5648 issues fixed for safer force cleanup
5651 Frequency : only filesystems larger than 120 OSTs
5652 Bugzilla : 5990, 6223
5653 Description: lfs getstripe would oops on a very large filesystem
5654 Details : lov_getconfig used kfree on vmalloc'd memory
5657 Frequency : only filesystems exporting via NFS to Solaris 10 clients
5658 Bugzilla : 6242, 6243
5659 Description: reading from files that had been truncated to a non-zero size
5660 but never opened returned no data
5661 Details : ll_file_read() reads zeros from no-object files to EOF
5666 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
5667 Details : There's little harm in aborting MDS/OSS recovery and letting it
5668 try again, so I removed the LASSERT and return an error instead.
5670 Severity : enhancement
5672 Description: New debugging infrastructure for tracking down data corruption
5673 Details : The I/O checksum code was replaced to: (a) control it at runtime,
5674 (b) cover more of the client-side code path, and (c) try to narrow
5675 down where problems occurred
5679 Bugzilla : 3819, 4364, 4397, 6313
5680 Description: Racing close and eviction MDS could cause assertion in mds_close
5681 Details : It was possible to get multiple mfd references during close and
5682 client eviction, leading to one thread referencing a freed mfd.
5684 Severity: : enhancement
5685 Bugzilla : 3262, 6359
5686 Description: Attempts to reconnect to servers are now more aggressive.
5687 Details : This builds on the enhanced upcall-less recovery that was added
5688 in 1.4.2. When trying to reconnect to servers, clients will
5689 now try each server in the failover group every 10 seconds. By
5690 default, clients would previously try one server every 25 seconds.
5695 Description: After recovery, certain operations trigger a failed
5696 assertion on a client.
5697 Details : Failing over an mds, using lconf -d --failover, while a
5698 client was doing a readdir() call would cause the client to
5699 LBUG after recovery completed and the readdir() was resent.
5701 Severity : enhancement
5703 Description: Default groups are now added by lconf
5704 Details : You can now run lconf --group <servicename> without having to
5705 manually add groups with lmc.
5708 Frequency : occasional
5710 Description: Nodes with an elan id of 0 trigger a failed assertion
5713 Frequency : always when accessing e.g. tty/console device nodes
5715 Description: tty and some other devices nodes cannot be used on lustre
5716 Details : file's private_data field is used by device data and lustre
5717 values in there got lost. New field was added to struct file to
5718 store fs-specific private data.
5721 Frequency : when exporting Lustre via NFS
5723 Description: NFSD failed occasionally when looking up a path component
5724 Details : NFSD is looking up ".." which was broken in ext3 directories
5725 that had grown large enough to become hashed.
5728 Frequency : Clusters with multiple interfaces not on the same subnet
5730 Description: Nodes will repeatedly try to reconnect to an interface which it
5731 cannot reach and report an error to the log.
5732 Details : Extra peer list entries will be created by lconf with some peers
5733 unreachable. lconf now validates the peer before adding it.
5736 Frequency : Only if a default stripe is set on the filesystem root.
5738 Description: Setting a default stripe on the filesystem root prevented the
5739 filesystem from being remounted.
5740 Details : The client was sending extra request flags in the root getattr
5741 request and did not allocate a reply buffer for the dir EA.
5744 Frequency : occasional, higher if lots of files are accessed by one client
5745 Bugzilla : 6159, 6097
5746 Description: Client trips assertion regarding lsm mismatch/magic
5747 Details : While revalidating inodes the VFS looks up inodes with ifind()
5748 and in rare cases can find an inode that is being freed.
5749 The ll_test_inode() code will free the lsm during ifind()
5750 when it finds an existing inode and then the VFS later attaches
5751 this free lsm to a new inode.
5755 Bugzilla : 6422, 7030
5756 Description: MDS deadlock between mkdir and client eviction
5757 Details : Creating a new file via mkdir or mknod (starting a transaction
5758 and getting the ns lock) can deadlock with client eviction
5759 (gets ns lock and trying to finish a synchronous transaction).
5762 Frequency : occasional
5763 Description: While starting a server, the fsfilt_ext3 module could not be
5765 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
5766 kernels. Previously, lconf would still use the ext3 name
5767 when trying to load modules. Now, it will correctly use
5768 ext3 on 2.4 and ldiskfs on 2.6.
5770 Severity : enhancement
5771 Description: The default stripe count has been changed to 1
5772 Details : The interpretation of the default stripe count (0, to lfs
5773 or lmc) has been changed to mean striping across a single
5774 OST, rather than all available. For general usage we have
5775 found a stripe count of 1 or 2 works best.
5777 Severity : enhancement
5778 Description: Add support for compiling against Cray portals.
5779 Details : Conditional compiling for some areas that are different
5783 Frequency : occasional
5784 Bugzilla : 6409, 6834
5785 Description: Creating files with an explicit stripe count may lead to
5786 a failed assertion on the MDS
5787 Details : If some OSTs are full or unavailable, creating files may
5788 trigger a failed assertion on the MDS. Now, Lustre will
5789 try to use other servers or return an error to the
5793 Frequency : occasional
5795 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
5796 caused lock timeout during readahead (since 1.4.2).
5797 Details : Processes doing readahead might match a lock that hasn't been
5798 granted yet if there are overlapping and conflicting lock
5799 requests. The readahead process waits on ungranted lock
5800 (original lock is CBPENDING), while OST waits for that process
5801 to cancel CBPENDING read lock and eventually evicts client.
5803 Severity : enhancement
5805 Description: Initial enabling of flock support for clients
5806 Details : Implements fcntl advisory locking and file status functions.
5807 This feature is provided as an optional mount flag (default
5808 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
5809 locking are implemented yet, and those that are are not guaranteed
5810 to be completely correct in production environments.
5811 mount -t lustre -o [flock|noflock] ...
5814 Frequency : occasional
5816 Description: OSTs running 2.4 kernels but with extents enabled might trip an
5817 assertion in the ext3 JBD (journaling) layer.
5818 Details : The b_committed_data struct is protected by the big kernel lock
5819 in 2.4 kernels, serializing journal_commit_transaction() and
5820 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
5821 access to this struct. In 2.6 kernels there is finer grained
5822 locking to improve SMP performance of the JBD layer.
5826 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
5831 Description: merge and cleanup kernel patches.
5834 -----------------------------------------------------------------------------
5836 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
5838 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
5839 versions of Lustre. Please update all servers and clients to
5840 version 1.4.2 or later at the same time. You must also run
5841 "lconf --write-conf {config}.xml" on the MDS while it is stopped
5842 to update the configuration logs.
5844 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
5845 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
5846 - swab llog records of type '0' so we get proper header size/idx (5861)
5847 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
5848 - fix rename of one directory over another leaking an inode (5953)
5849 - avoid SetPageDirty on 2.6 (5981)
5850 - don't re-add just-being-destroyed locks to the waiting list (5653)
5851 - when creating new directories, inherit the parent's custom
5852 striping settings if present parent (3048)
5853 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
5854 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
5855 - don't hold client locks on temporary worklist from l_lru (5666)
5856 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
5857 - automatically evict dead clients (5921)
5858 - Update file size properly in create+truncate+fstat case (6196)
5859 - Do not unhash mountpoint dentries, do not allow removal of
5861 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
5862 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
5863 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
5864 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
5865 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
5866 - Correctly handle reads of files with no objects (6243)
5867 - lctl recover will also mark a device active if deactivate used (5933)
5869 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
5870 - allow --write-conf on an MDS with different nettype than client (5619)
5871 - don't write config llogs to MDS for mounts not from that MDS (5617)
5872 - lconf should create multiple TCP connections from a client (5201)
5873 - init scripts are now turned off by default; run chkconfig --on
5874 lustre and chkconfig --on lustrefs to use them
5875 - upcalls are no longer needed for clients to recover to failover
5877 - add --abort-recovery option to lconf to abort recovery on device
5879 - add support for an arbitrary number of OSTs (3026)
5880 - Quota support protocol changes.
5881 - forward compatibility changes to wire structs (6007)
5882 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5883 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5884 - improved "lustre status" script
5885 - initialize blocksize for non-regular files (6062)
5886 - added --disable-server and --disable-client configure options (5782)
5887 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5888 - Vanilla 2.4.29 support
5889 - increase maximum number of obd devices to 520 (6242)
5890 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5891 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5892 SLES 9 2.6.5 kernels
5893 - limit stripes per file to 160 (the maximum EA size) (6093)
5895 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5898 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5899 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5900 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5901 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5902 - small ext3 extents cleanups and fixes (5733)
5903 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5904 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5905 - avoid needless client->OST connect, fix handle mismatch (5317)
5906 - fix DLM error path that led to out-of-sync client, long delays (5779)
5907 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5908 - fix several locking issues related to i_size (5492,5624,5654,5672)
5909 - don't move pending lock onto export if it is already evicted (5683)
5910 - fix kernel oops when creating .foo in unlinked directory (5548)
5911 - fix deadlock in obdfilter statistics vs. object create (5811)
5912 - use time_{before,after} to avoid timer jiffies wrap (5882)
5913 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5914 - Do not leak request if server was not able to process it (5154)
5915 - If mds_open unable to find parent dir, make that negative lookup(5154)
5916 - don't create new directories with extent-mapping (5909, 5936)
5918 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5919 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5920 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5921 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5922 - support for RHEL4 kernel on i686 (5773)
5923 - provide error messages when incompatible logs are encountered (5898)
5925 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5926 * version 1.4.0.10 (1.4.1 release candidate 1)
5928 - don't keep a lock reference when lock is not granted (4238)
5929 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5930 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5931 - drop import inflight refcount on signal_completed_replay error (5255)
5932 - unlock page after async write error during send (3677)
5933 - handle missing objects in filter_preprw_read properly (5265)
5934 - no transno return for symlink open, don't save no-trasno open (3440)
5935 - don't try to complete elan receive that already failed (4012)
5936 - free RPC server reply state on error (5406)
5937 - clean up thread from ptlrpc_start_thread() on error (5160)
5938 - readahead could read extra page into cache that wasn't ejected (5388)
5939 - prevent races in class_attach/setup/cleanup/detach (5260)
5940 - don't dereference de->d_inode after l_dput of de (5458)
5941 - use "int" for stripe value returned from lock_to_stripe (5544)
5942 - mballoc allocation and error-checking fixes in 2.6 (5504)
5943 - block device patches to fix I/O request sizes in 2.6 (5482)
5944 - look up hostnames for IB nals (5602)
5945 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5946 - don't start multiple acceptors for the same port (5277)
5947 - fix incorrect LASSERT in mds_getattr_name (5635)
5948 - export a proc file for general "ping" checking (5628)
5949 - fix "lfs check" to not block when the MDS is down (5628)
5951 - service request history (4965)
5952 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5953 - create an "evict_client" /proc entry on OSTs, like the MDS has
5954 - fix mount usage message, return errors per mount(8) (5168)
5955 - change grep [] to grep "[]" in tests so they work in more UMLs
5956 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5957 - remove ancient LOV_MAGIC_V0 EA support (5047)
5958 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5959 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5960 - allow readahead allocations to fail when low on memory (5383)
5961 - mmap locking landed again, after considerable improvement (2828)
5962 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
5964 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
5967 - send OST transaction number in read/write reply to free req (4966)
5968 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5969 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5970 - the watchdog thread now runs as interruptible (5246)
5971 - flock/lockf fixes (but it's still disabled, pending 5135)
5972 - don't use EXT3 constants in llite code (5094)
5973 - memory shortage at startup could cause assertion (5176)
5975 - reorganization of lov code
5976 - single portals codebase
5978 - add extents/mballoc support (5025)
5979 - direct I/O reads in the obdfilter (4048)
5980 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
5982 tbd Cluster File Systems, Inc. <info@clusterfs.com>
5985 - send OST transaction number in read/write reply to free req (4966)
5986 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5987 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5988 - the watchdog thread now runs as interruptible (5246)
5989 - handle missing objects in filter_preprw_read properly (5265)
5990 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5991 - drop import inflight refcount on signal_completed_replay error (5255)
5992 - unlock page after async write error during send (3677)
5993 - return original error code on reconstructed replies (3761)
5994 - no transno return for symlink open, don't save no-trasno open (3440)
5996 - add pid to ldlm debugging output (4922)
5997 - bump the watchdog timeouts -- we can't handle 30sec yet
5998 - extra debugging for orphan dentry/inode bug (5259)
6000 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
6003 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
6004 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
6005 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
6006 - fix NULL dereference in /proc/sys/portals/routes (4827)
6007 - allow failed mdc_close() operations to be interrupted (4561)
6008 - stop precreate on OST before MDS would time out on it (4778)
6009 - don't send partial-page writes before EOF from client (4410)
6010 - discard client grant for sub-page writes on large-page clients (4520)
6011 - don't free dentries not owned by NFS code, check generation (4806)
6012 - fix lsm leak if mds_create_objects() fails (4801)
6013 - limit debug_daemon file size, always print CERROR messages (4789)
6014 - use transno after validating reply (3892)
6015 - process timed out requests if import state changes (3754)
6016 - update mtime on OST during writes, return in glimpse (4829)
6017 - add mkfsoptions to LDAP (4679)
6018 - use ->max_readahead method instead of zapping global ra (5039)
6019 - don't interrupt __l_wait_event() during strace
6021 - add software watchdogs to catch hung threads quickly (4941)
6022 - make lustrefs init script start after nfs is mounted
6023 - fix CWARN/ERROR duplication (4930)
6024 - return async write errors to application if possible (2248)
6025 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
6026 - print NAL number in %x format (4645)
6027 - update barely-supported suse-2.4.21-171 series (4842)
6028 - support for sles 9 %post scripts
6029 - support for building 2.6 kernel-source packages
6030 - support for sles km_* packages
6032 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
6035 - ignore -ENOENT errors in osc_destroy (3639)
6036 - notify osc create thread that OSC is being cleaned up (4600)
6037 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
6038 - reconstruct ost_handle() like mds_handle() (4657)
6039 - create a new thread to do import eviction to avoid deadlock (3969)
6040 - let lconf resolve symlinked-to devices (4629)
6041 - don't unlink "objects" from directory with default EA (4554)
6042 - hold socknal file ref over connect in case target is down (4394)
6043 - allow more than 32000 subdirectories in a single directory (3244)
6044 - fix blocks count for O_DIRECT writes (3751)
6045 - OST returns ENOSPC from object create when no space left (4539)
6046 - don't send truncate RPC if file size isn't changing (4410)
6047 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
6048 - bind to privileged port in socknal and tcpnal (3689)
6050 - rate limit CERROR/CWARN console message to avoid overload (4519)
6051 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
6052 - basic mmap support (3918)
6053 - kernel patch series update from b1_4 (4711)
6055 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
6058 - avoid crash during MDS cleanup with OST shut down (2775)
6059 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
6060 - don't use bad inodes on the MDS (3744)
6061 - dynamic object preallocation to improve recovery speed (4236)
6062 - don't hold spinlock over lock dumping or change debug flags (4401)
6063 - don't zero obd_dev when it is force cleaned (3651)
6064 - print grants to console if they go negative (4431)
6065 - "lctl deactivate" will stop automatic recovery attempts (3406)
6066 - look for existing locks in ldlm_handle_enqueue() (3764)
6067 - don't resolve lock handle twice in recovery avoiding race (4401)
6068 - revalidate should check working dir is a directory (4134)
6070 - don't always mark "slow" obdfilter messages as errors (4418)
6072 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
6075 - don't close LustreDB during write_conf until it is done (3860)
6076 - fix typo in lconf for_each_profile (3821)
6077 - allow dumping logs from multiple threads at one time (3820)
6078 - don't allow multiple threads in OSC recovery (3812)
6079 - fix debug_size parameters (3864)
6080 - fix mds_postrecov to initialize import for llog ctxt (3121)
6081 - replace config semaphore with spinlock (3306)
6082 - be sure to send a reply for a CANCEL rpc with bad export (3863)
6083 - don't allow enqueue to complete on a destroyed export (3822)
6084 - down write_lock before checking llog header bitmap (3825)
6085 - recover from lock replay timeout (3764)
6086 - up llog sem before sending rpc (3652)
6087 - reduce ns lock hold times when setting kms (3267)
6088 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
6089 - fix NULL deref and obd_dev leak on setup error (3312)
6090 - replace some LBUG about llog ops with error handling (3841)
6091 - don't match INVALID dentries from d_lookup and spin (3784)
6092 - hold dcache_lock while marking dentries INVALID and hashing (4255)
6093 - fix invalid assertion in ptlrpc_set_wait (3880)
6095 - add libwrap support for the TCP acceptor (3996)
6096 - add /proc/sys/portals/routes for non-root route listing (3994)
6097 - allow setting MDS UUID in .xml (2580)
6098 - print the stack of a process that LBUGs (4228)
6100 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
6103 - don't cleanup request in ll_file_open() on failed MDS open (3430)
6104 - make sure to unset replay flag from failed open requests (3440)
6105 - if default stripe count is 0, use OST count for inode size (3636)
6106 - update parent mtime/ctime on client for create/unlink (2611)
6107 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
6108 - free recovery state on server during a forced cleanup (3571)
6109 - unregister_reply for resent reqs (3063)
6110 - loop back devices mounting and status check on 2.6 (3563)
6111 - fix resource-creation race that can provoke i_size == 0 (3513)
6112 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
6113 - more debugging for page-accounting assertion (3746)
6114 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
6115 - avoid deadlock after precreation failure (3758)
6116 - fix race and lock order deadlock in orphan handling (3450, 3750)
6117 - add validity checks when grabbing inodes from l_ast_data (3599)
6119 - add /proc/.../recovery_status to obdfilter (3428)
6120 - lightweight CDEBUG infrastructure, debug daemon (3668)
6121 - change default OSC RPC parameters to be better on small clusters
6122 - turn off OST read cache for files smaller than 32MB
6123 - install man pages and include them in rpms (3100)
6124 - add new init script for (un)mounting lustre filesystems (2593)
6125 - run chkconfig in %post for init scripts (3701)
6126 - drop scimac NAL (unmaintained)
6128 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
6131 - clean kiobufs before and after use (3485)
6132 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
6133 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
6134 - change init script to fail more clearly if not run as root (1528)
6135 - allow clients to reconnect during replay (1742)
6136 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
6137 - don't do DNS lookups on NIDs too small for IP addresses (3442)
6138 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
6139 - fix cond_resched (3554)
6140 - only evict unfinished clients after recovery (3515)
6141 - allow bulk resend, prevent data loss (3570)
6142 - dynamic ptlrpc request buffer allocation (2102)
6143 - don't allow unlinking open directory if it isn't empty (2904)
6144 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
6145 - remove extraneous obd dereference causing LASSERT failure (3334)
6146 - don't use get_cycles() when creating temp. files on the mds (3156)
6147 - hold i_sem when setting i_size in ll_extent_lock() (3564)
6148 - handle EEXIST for set-stripe, set proper directory name (3336)
6150 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
6151 - fix ksocknal_fmb_callback() error messages (2918)
6153 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
6156 - don't copy lvb into (possibly NULL) reply on error (2983)
6157 - don't deref dentry after dput, don't free lvb on error (2922)
6158 - use the kms to determine writeback rpc length (2947)
6159 - increment oti_logcookies when osc is inactive (2948)
6160 - update client's i_blocks count via lvb messages (2543)
6161 - handle intent open/close of special files properly (1557)
6162 - mount MDS with errors=remount-ro, like obdfilter (2009)
6163 - initialize lock handle to avoid ASSERT on error cleanup (3057)
6164 - don't use cancelling-locks' kms values (2947)
6165 - use highest lock extent for kms, not last one (2925)
6166 - don't dereference ERR_PTR() dentry in error handling path (3107)
6167 - fix thread race in portals_debug_dumplog() (3122)
6168 - create lprocfs device entries at setup instead of at attach (1519)
6169 - common AST error handler, don't evict client on completion race (3145)
6170 - zero nameidata in detach_mnt in 2.6 (3118)
6171 - verify d_inode after revalidate_special is valid in 2.6 (3116)
6172 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
6173 - initialize RPC timeout timer earlier for 2.6 (3219)
6174 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
6175 - print nal/nid for unknown nid (3258)
6176 - additional checks for oscc recovery before doing precreate (3284)
6177 - fix ll_extent_lock() error return code for 64-bit systems (3043)
6178 - don't crash in mdc_close for bad permissions on open (3285)
6179 - zero i_rdev for non-device files (3147)
6180 - clear page->private before handing to FS, better assertion (3119)
6181 - tune the read pipeline (3236)
6182 - fix incorrect decref of invalidated dentry (2350)
6183 - provide read-ahead stats and refine rpc in flight stats (3328)
6184 - don't hold journal transaction open across create RPC (3313)
6185 - update atime on MDS at close time (3265)
6186 - close LDAP connection when recovering to avoid server load (3315)
6187 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
6188 - don't leak open file on MDS after open resend (3325)
6189 - serialize filter_precreate and filter_destroy_precreated (3329)
6190 - loop device shouldn't call sync_dev() for nul device (3092)
6191 - clear page cache after eviction (2766)
6192 - resynchronize MDS->OST in background (2824)
6193 - refuse to mount the same filesystem twice on same mountpoint (3394)
6194 - allow llmount to create routes for mounting behind routers (3320)
6195 - push lock cancellation to blocking thread for glimpse ASTs (3409)
6196 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
6197 - fix rare problem with rename on htree directories (3417)
6199 - allow default OST striping configuration per directory (1414)
6200 - fix compilation for qswnal for 2.6 kernels (3125)
6201 - increase maximum number of MDS request buffers for large systems
6202 - change liblustreapi to be useful for external progs like lfsck (3098)
6203 - increase local configuration timeout for slow disks (3353)
6204 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
6206 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
6209 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
6210 - don't overwrite extent policy data in reply if lock was blocked (2901)
6211 - drop filter export grants atomically with removal from device (2663)
6212 - del obd_self_export from work_list in class_disconnect_exports (2908)
6213 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
6214 - swab reply message in mdc_close, other PPC fixes (2464)
6215 - fix destroying of named logs (2325)
6216 - overwrite old logs when running lconf --write_conf (2264)
6217 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
6218 - fix race in target_handle_connect (2898)
6219 - mds_reint_create() should take same inode create lock (2926)
6220 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
6221 - don't close files for self_export to avoid uninitialized obd (2936)
6222 - allow MDS with the same name as client node (2939)
6223 - hold dentry reference for closed log files for unlink (2325)
6224 - reserve space for all logs during transactions (2059)
6225 - don't evict page beyond end of stripe extent (2925)
6226 - don't oops on a deleted current working directory (2399)
6227 - handle hard links to targets without a parent properly (2517)
6228 - don't dereference NULL lock when racing during eviction (2867)
6229 - don't grow lock extents when lots of conflicting locks (2919)
6231 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
6234 - account for cache space usage on clients to avoid data loss (974)
6235 - lfsck support in lustre kernel code (2349)
6236 - reduce journal credits needed for BRW writes (2370)
6237 - orphan handling to avoid losing space on client/server crashes
6238 - ptlrpcd can be blocked, stopping ALL progress (2477)
6239 - use lock value blocks to assist in proper KMS, faster stat (1021)
6240 - takes i_sem instead of DLM locks internally on obdfilter (2720)
6241 - recovery for initial connections (2355)
6242 - fixes for mds_cleanup_orphans (1934)
6243 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
6244 - block all file creations until orphan recovery completes (1901)
6245 - client remove rq_connection from request struct (2423)
6246 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
6247 - recovery timer race (2670)
6248 - mdc_close recovey bug (2532)
6249 - ptlrpc cleanup bug (2710)
6250 - mds timeout on local locks (2588)
6251 - namespace lock held during RPCs (2431)
6252 - handle interrupted sync write properly (2503)
6253 - don't try to handle a message that hasn't been replied to (2699)
6254 - client assert failure during cleanup after abort recovery (2701)
6255 - leak mdc device after failed mount (2712)
6256 - ptlrpc_check_set allows timedout requests to complete (2714)
6257 - wait for inflight reqs when ptlrpcd finishes (2710)
6258 - make sure unregistered services are removed from the srv_list
6259 - reset bulk XID's when resending them (caught by 1138 test)
6260 - unregister_bulk after timeout
6261 - fix lconf error (2694)
6262 - handle write after unfinished setstripe, stripe-only getstripe (2388)
6263 - readahead locks pages, leaves pending causing memory pressure (2673)
6264 - increase OST request buffers to 4096 on large machines (2729)
6265 - fix up permission of existing directories in simple_mkdir (2661)
6266 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
6267 - don't assign transno to errored transactions (2742)
6268 - don't delete objects on OST if given a bogus objid from MDS (2751)
6269 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
6270 - if rq_no_resend, then timeout request after recovery (2432)
6271 - fix MDS llog_logid record size, 64-bit array alignment (2733)
6272 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
6273 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
6274 - MDS recovery shouldn't delete objects that it has given out (2730)
6275 - if enqueue arrives after completion, don't clobber LVB (2819)
6276 - don't unlock pages twice when trigger_group_io returns error (2814)
6277 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
6278 - don't write pages to disk if there was an error (1450)
6279 - don't ping imports that have recovery disabled (2676)
6280 - take buffered bytes into account when balancing socknal conn (2817)
6281 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
6282 - reconnect unlink llog connection after MDS reconnects to OST (2816)
6283 - remove little-endian swabbing of llog records (1987)
6284 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
6285 - retry reposting request buffers if they fail (1191)
6286 - grow extent at grant time to avoid granting a revoked lock (2809)
6287 - lock revoke doesn't evict page if covered by a second lock (2765)
6288 - disable VM readahead to avoid reading outside lock extents (2805)
6290 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
6291 - updated kernel patches for hp-2.4.20 kernel (2681)
6293 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
6296 - fix truncated write corruption (2366)
6297 - fix for failed assertion in iopen_connect_dentry (1792,2517)
6299 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
6300 - ensure len(uuid) < 37 in lmc (1171)
6301 - fix ia64 OOPS in llog_test (2255)
6302 - zero end of page at obdfilter for partial page writes (2648)
6303 - don't leave stale dentries around after renames (bug 2428)
6304 - fix timeouts when evicting a client with a single lock held (2642)
6305 - set deadline for the initial HELLO message to drain (2634)
6306 - print out dotted-quad IP addresses in the socknal (2302)
6308 - additional debugging for MDS client eviction problem (2443)
6309 - fix mkfsoptions support for osts (2603, 2604)
6311 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
6314 - add series for the vanilla 2.6.0 kernel
6315 - add series for the vanilla 2.4.24 kernel
6316 - add series for a cray x86/64 UL kernel drop
6317 - fix xattr patches for the vanilla 2.4.19 series
6319 - generate true UUIDs in lmc (1171)
6320 - have portals stack dumping break in UML (2466)
6321 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
6322 - allocate with _NOFS in ldlm to avoid deadlock (1933)
6323 - wake callback waiting threads on client eviction (2460)
6324 - Add --ptldebug and --subsystem to lmc (1719)
6325 - update assertion to allow safe interrupt allocation
6326 - set rq_no_resend for cancel requests (2432)
6327 - recalculate ptlrpcd timeout after resend (2494)
6328 - call vfs_rmdir when removing pending directories (2368)
6329 - fix renaming a file to itself (2429)
6330 - lmc creates a default one-stripe lov (2454)
6331 - expand procfs space to handle large clusters (2326)
6332 - increase UML stack to avoid overflow
6333 - update lconf's list of debug and subsystem masks
6334 - fix lfs find --obd (2510)
6335 - /proc tunable for disabling filter read caching (2591)
6336 - stop rpm packages from altering slapd.conf (2301)
6337 - disable nagle in the socknal under 0conf (2578)
6338 - choose mds inode size based on stripe count (2572)
6339 - fix kernel-source rpm problems (2516)
6341 - add --disable-doc to avoid pdf generation (2421)
6342 - update documentation, tests, type-os, comments
6343 - avoid format warnings on ia64
6344 - remove the TOE NAL
6345 - tiny code cleanups by removing unused fields
6347 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
6350 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
6351 - remove the most blatant lies from BUILDING file (2371)
6352 - change default debug level to reasonable production setting
6353 - reduce client side cache size to reduce cache flush time
6354 - reduce max RPCs in flight to avoid unnecessary file fragmentation
6355 - make TCP zerocopy and pinger support enabled by default (2476)
6356 - sync writes completed after process exits caused crashes (2319)
6357 - maintain correct mount count on the MDS (2356)
6358 - backout 1557, because 2316 wasn't really fixed
6359 - better file I/O statistics gathering in /proc
6360 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
6361 - another kernel patch to fix zero-copy TCP function export
6362 - don't take duplicate lock when processing re-sent getattr (2420)
6363 - lctl uses obd_self_export instead of creating new conn (2353)
6364 - MDS/OST recovery case which requires object creation asserted (2425)
6365 - move lfs from /usr/sbin to /usr/bin in packages
6366 - fix race between mds_client_add and mds_client_free (2417)
6367 - use kmalloc instead of slabs in portals (2430)
6368 - don't create duplicate records when a failover MDS is present (2442)
6369 - remove unnecessary mount age check (2332)
6370 - don't remove directory inodes from locks prematurely (2451)
6371 - don't break if MDS service name is the same as hostname (2103)
6372 - fix races in client write RPC generation when cache full (2482)
6374 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
6377 - remove now-unused request->rq_obd (278)
6378 - if an allocation fails, print out how much memory we've used (1933)
6379 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
6380 - add the "configurable stack size" patch to most series files (1256)
6381 - ability to write large log records, for 100+ OST configs (2306)
6382 - fix NULL deref when filter_prep fails (2314)
6383 - fix operator precedence error in filter_sync
6384 - dynamic allocation of socknal TX descriptors (2315)
6385 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
6386 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
6387 - ensure that CWARN messages in Portals always get to the syslog
6388 - __init/__exit are not for prototype decls (ldlm_init/exit)
6389 - x86-64 compile warning fixes
6390 - fix gateway LMC keyword conflict (2318)
6391 - fix MDS lock inversions in getattr/reint paths (1844)
6392 - fix a rare lock re-ordering bug, which caused deadlock (2322)
6393 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
6394 - DLM race condition prevented some lock evictions (2328)
6395 - ENOMEM detection and retry on socknal sends (2230)
6396 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
6397 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
6398 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
6399 - better error messages when a client is rejected during recovery (1505)
6400 - avoid cancelling locks which were never granted, after failure (2330)
6401 - fix i_sem/journal inversion in mds_client_add (2333)
6402 - fix truncate/getattr lock cycle deadlock (2334)
6403 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
6404 - fix two rare exit paths which could leak an l_lock() ref (2321)
6405 - fencepost error in MDS/OST orphan recovery (2226)
6406 - make log record alignment 8 bytes (1988)
6407 - lstripe now fails when requested offset > ost_count (2237)
6408 - ensure that all kernel series have a complete list.h (1607)
6409 - fix crashes in special-file operations (2316)
6410 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
6412 - allow configurable automake binary, for testing new versions
6413 - small update to the lfs documentation
6415 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
6417 * fix negative export reference count in fsfilt_sync (2312)
6419 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
6420 * release candidate 0.9.1
6422 - orphans are moved into the PENDING directory for possible recovery
6423 - replayed opens now open by fid for orphan/rename safety (1042)
6424 - last close of an orphan inode generates a transno (683)
6425 - chdir() and mount() now pin the directory entry (1020)
6426 - avoid CERROR in normal ll_setattr_raw() error case (1500)
6427 - discard very old requests without processing them (1502)
6428 - remove some common, well-understood CERRORs (1505)
6429 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
6430 - clear "grant" flags in OST replies until OST grant code lands (1644)
6431 - fix read performance by not clobbering i_blksize on client (1598)
6432 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
6433 - make LRU size a /proc tunable, clears locks when reduced (707)
6434 - fix some lprocfs rot that prevented ptlbd from loading (1732)
6435 - server locks take references on exports now (1558)
6436 - build fixes for 2.4.20-rh trees (1663)
6437 - return an error from lov_create if all OSCs are inactive (1751)
6438 - fix import levels when a reconnect happens without a timeout (1597)
6439 - exit early from mds_open if we get a lookup error (1749)
6440 - partial page read at EOF wouldn't wait for disk before sending (1642)
6441 - avoid NULL deref in obdfilter when reading page past EOF (1592)
6442 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
6443 - fix LBUG in ll_it_open_error with rc = -2 (1861)
6444 - write/truncate lock inversion (1639)
6445 - Don't auto-load obdclass, portals modules during cleanup (1495)
6446 - fix timestamps from jumping to "now" (1763)
6447 - extra journal assertions (1648)
6448 - add an extra multiunlink test (1771)
6449 - fix read_record/write_record API (1776)
6450 - fix leak of offset_extent, possible incorrect i_size later (1772)
6451 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
6452 - Debugging for the kqswnal_get_idle_tx problems (1820)
6453 - Allow recovery to be attempted multiple times (1536)
6454 - Write out MDS last_rcvd file after it is first created (1600)
6455 - Fix tx_descriptor leak in failed transmit situations (1827)
6456 - ext3 journaling fixes for assertion failure after IO error (1871)
6457 - class_export_put() on freed export after completion AST error (1896)
6458 - Fix revalidate looping in VFS (1322)
6459 - Don't access a freed export during MDS_REINT timeout (1521)
6460 - Add open-unlink recovery support on the MDS (1673,1764)
6461 - Return an error if no MDS data was read from last_rcvd (1946)
6462 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
6463 - Don't setup a disk device that doesn't match exported UUID (317)
6464 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
6465 - avoid committing NULL handle in force close
6466 - local.sh is now a one-stripe LOV configuration
6467 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
6468 - fix ext3 htree duplicate directory entry corruption (1516)
6469 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
6470 - update to new LOV EA format (2097)
6471 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
6473 2003-06-15 Phil Schwan <phil@clusterfs.com>
6476 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
6477 - per-import/export recovery handling (958, 931, 959)
6478 - multiple last-rcvd slots, for serving multiple FSes (949)
6479 - connections are again shared between multiple imp/exports (963, 964)
6480 - "umount -f" would hang if any requests needed to be sent (393, 978)
6481 - avoid pinning large req buffer by copying for queued messages (989)
6482 - add "uuid" to "lctl device" command to help upcalls (991)
6483 - "open" RPCs with transnos would confuse recovery counters (1037)
6484 - do proper endian conversion of all wire messages (288, 340, 891)
6485 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
6486 - call waiting locks callback from LDLM recovery thread (1127, 1151)
6487 - fix ptlrpc_connection leak in target_handle_connect (1174)
6488 - fix import refcounting bug in OST and MDS cleanup (1134)
6489 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
6490 - fix very unlikely obd_types race condition (501)
6491 - remove osc_open hack for echo_client (1187)
6492 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
6493 - a failure in read_inode2 leads to deadlock (1139)
6494 - cancel ack-locks as soon as transaction is committed (1072)
6495 - fix major leaks and crashes in the bulk I/O path (937, 1057)
6496 - make sure to commitrw after any preprw to avoid deadlock (1162)
6497 - failing to execute a file in a lustre FS would lock inode (1203)
6498 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
6499 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
6500 - obd_uuid2tgt would walk off the end of the list (1255)
6501 - on IA64 the osc would give portals incorrect bulk size (1258)
6502 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
6503 - fix lock inversion caused by new llite matching code (1282)
6504 - limit the number of dirty pages on a client to 10MB (1286)
6505 - timed out locks were not being corrected cancelled (1289)
6506 - fix O_DIRECT above 4GB on IA-32 (1292)
6507 * major user-visible changes
6508 - fail out/fail over policy now controlled by the upcall (993)
6510 - add OBD_PING to check server availability and failure (954)
6511 - lustre messages are now sent in sending host order (288, 340, 891)
6512 - add eadatalen to MDS getattr reply (340)
6513 - OST read replies may contain second buffer, with per-page status (593)
6515 2003-03-11 Phil Schwan <phil@clusterfs.com>
6518 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
6519 - failed open()s could cause deadlock; fixed (867, 869)
6520 - stop cancelling OST locks when files are closed (481)
6521 - overlapping XID spaces caused network corruption (851, 853)
6522 - fix unsafe fsfilt counter arithmetic; change to atomic_t
6523 - setattr_raw added, to do single-RPC, server-side setattrs
6524 - lmc/lconf syntax change for OST UUIDs
6525 - fix crashy race condition between ptlrpc_free_req and osc_close
6526 - don't use request in mdc_enqueue if we hit a timeout (889)
6527 - don't set the inode i_size for regular files from the MDS (896)
6528 - handle out of order completion AST (842)
6529 - don't LBUG if a lock request times out after receiving AST (913)
6530 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
6531 - if a bad lock AST arrives, send an error instead of dropping entirely
6532 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
6533 - fix leak in bulk IO when only partially completed (899, 900, 926)
6534 - fix O_DIRECT for ia64 (55)
6535 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
6536 - C-z after timeout could hang a process forever; fixed (977)
6538 - client-side I/O cache (678, 924, 929, 941, 970)
6540 - READPAGE and SETATTRs which don't take server-side locks get
6543 2003-02-11 Phil Schwan <phil@clusterfs.com>
6546 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
6547 - Fix fsfilt_extN_readpage() to read a full page of directory
6548 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
6549 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
6550 - Fix imbalanced LOV object allocation and out-of-bound access (469)
6551 - Most intent operations were removed, in favour of a new RPC mode
6552 that does a single RPC to the server and bypasses most of the VFS
6553 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
6554 - Aggressively cancel local locks on DLM servers
6555 - mds_reint_unlink sends EA to the client if it's the last nlink.
6556 client uses that EA to unlink OST objects.
6557 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
6558 - recursive symlinks were fixed (439)
6559 - fixed NULL deref in DEBUG_REQ
6560 - filter_update_lastobjid no longer calls sync, which annoyed extN
6561 - fixed multi-client small-writes to a single file problem (445)
6562 - fixed mtime updates during file writes (607)
6563 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
6564 - fixed bug in obd_brw_read/write() (under guise of testing 367)
6565 - fixed Linux OST size reporting problem (444, 656)
6566 - OST now updates object mtime with writes or setattr (607, 619)
6567 - client verifies file size before zeroing page past EOF (445)
6568 - OST now writes last allocated objid to disk with allocation (108)
6569 - LOV on echo now works (409)
6571 - mds_reint_unlink sends a new buffer, with the EA included. this
6572 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
6573 set if a regular file was being unlinked, and it was the last link
6574 - use PtlGet from the target for bulk writes (315)
6575 - OST now updates object mtime with writes or setattr (607, 619)
6576 - LDLM now has a grant-time callback to revalidate locked items, if
6578 - Many MDS operations were reorganized to combat race conditions
6580 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
6581 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
6583 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
6586 - Fully reactivate OST imports after reconnection (512, others)
6587 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
6588 - More graceful error handling for truncating on dead OST (515)
6589 - Don't error out unless we're actually accessing dead stripes (474)
6590 - Fix garbage sizes when stripes are missing (410)
6591 - LRU counters were broken, causing constant lock purge (433, 432)
6592 - garbage on read from stripes with failed OSTs (441)
6593 - mark OSCs as active before reconnecting during recovery (438)
6594 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
6595 - lfind did not preserve OST order in output (443)
6596 - symlinks cause hung clients, incorrect data (439)
6597 - stop dereferencing request after dropping refcount (457)
6598 - don't LASSERT(spin_is_locked) on non-SMP (455)
6599 - fixes for many rename() bugs
6600 - fstat didn't correctly synchronize attributes (399)
6601 - server must handle lock cancellation during blocking AST prep (487)
6602 - bulk descriptors were free()d too soon (511)
6603 - fix paths in lconf, which would load incorrect modules (451, 507)
6604 - fix confusing lconf 'host not found' error message (386)
6605 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
6606 - fix race condition in mdc_blocking_ast() for inode access (526)
6607 - fix lov_unpackmd() unpacking wrong number of stripes (537)
6608 - fix lov_set_osc_active() marking wrong OSC inactive (440)
6609 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
6610 - fix multiple writes of stripe MD to MDS (358, maybe 519)
6611 - fix lstripe in several ways (kernel side) (527)
6612 - fix request leak in ldlm_cli_enqueue (262)
6613 - incorrect OSC was marked inactive after OST failure
6614 - call mds_fs_cleanup before unmounting filesystem (524)
6615 - fix races between taking ns_lock and ldlm_lock_change_resource
6616 - fix races updating LOV export open file list
6617 - fix lov_enqueue error path, avoid decref-ing bad lock handle
6618 - fix recovery NULL deref in ldlm_cli_cancel_unused
6619 - fix some DLM races by using new hash table for lock handles (419)
6620 - permit the client to specify desired inodes, at replay
6621 - duplicate requests when we queue them for replay reintegration
6622 - fix last_rcvd offset calculation
6623 - sync after each recovered transaction, so we always make progress
6624 - never, not always, ERESTART requests without transnos
6625 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
6627 - skip replay if the MDS says that the client is already connected
6628 - don't check for a recovery-enabled export to match lctl's UUID
6629 - don't INC_USE_COUNT for phantom exports
6630 - don't crash when cleaning up phantom exports (567)
6631 - don't double-finish or set replay data for errored mdc_open requests
6632 - abort requests when they time out, so we don't get old replies
6633 - send/receive replies for AST messages again
6634 - if the client says that it doesn't have the lock, cancel it on the
6636 - if we timeout during I/O, don't try to cancel an in-use lock; instead
6637 mark it as destroyed, it will all work out when decref is called
6638 - fix module use counts (22, 581)
6640 - ASTs now expect a reply (server cancels lock on error reply)
6642 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
6645 - fix many simultaneous client startup (392)
6646 - fix dentry->d_it clobbering
6647 - credentials weren't being shipped for readdir/getattr operations
6648 - remove invalid assertions triggered during some concurrent MD
6650 - proper Lustre versions added (336, 389)
6651 - fix memory leak for create error case (398)
6652 - fix LOV locking bug that would get cli/srv out of sync
6653 - fix echo client over LOV (409)
6654 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
6655 - fix double-O_EXCL intent crash (424)
6656 - avoid sending multiple lock CANCELs (352)
6658 - MDS can do multi-client recovery (modulo bugs in new code)
6660 - many updates, edits, cleanups
6662 2002-11-18 Phil Schwan <phil@clusterfs.com>
6665 - fix null d_it dereference (346)
6666 - fix full OST/dbench hang (333)
6667 - fix permission problem with file removal (286)
6668 - fix removal of OSCs from LOV when they fail
6669 - fix NULL deref during bulk timeout (214)
6670 - fix problems related to multiple filesystems on one MDS (241)
6671 - fixed serious subtle metadata locking bugs
6672 - free locks on clients when inodes are removed due to memory
6674 - fix inode pointer in lock data (285)
6675 - partial support for multiple MDS on a single host (241)
6676 - data locks weren't cancelled at clear_inode time (290, 311)
6677 - intent locks could lead to unbounded lock growth (205)
6678 - added a maximum lock count, an LRU list, and a flusher
6679 - fix multiple rename (365)
6680 - properly abstracted the echo client
6681 - OSC locked 1 byte too many; fixed
6682 - rewrote brw callback code:
6683 - fixed recovery bugs related to LOVs (306)
6684 - fixed too-many-pages-in-one-write crash (191)
6685 - fixed (again) crash in sync_io_timeout (214)
6686 - probably fixed callback-related race (385)
6688 - Add capability to MDS protocol
6689 - LDLM cancellations and callbacks on different portals
6691 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
6694 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
6695 - defer open object destruction to close time (601981, 138)
6696 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
6697 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
6698 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
6699 - don't call obd_setattr in ll_file_release for destroyed objects
6700 * protocol change to lustre_msg: move |version| and add |flags|
6701 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
6702 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
6703 * added replay of create, unlink, link and rename operations during
6704 MDS failover; recovery should be much more robust now
6705 * remove failed OSCs from LOVs (only lov_create uses this so far)
6706 * the lustre-HOWTO was brought (more) up to date (582544)
6708 2002-10-23 Phil Schwan <phil@clusterfs.com>
6711 - in-use dentries weren't being reused properly (617851)
6712 - prevent multiple LDLM setup (599178)
6713 - fix LOV size calculations for truncate (617853)
6714 - fix client handling of MDS intent errors (POSIX)
6715 - fix permission bug in lovstripe.c test (624321)
6716 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
6717 - truncate past end of file could corrupt data
6718 - proper cleanup after timeouts, crashes, etc (592524, 550815)
6719 - a race in recovery could return ETIMEDOUT to apps (623947)
6720 - building outside the source directory was fixed
6721 * the lustre-HOWTO was brought (more) up to date (582544)
6722 * major progress was made on recovery functionality
6724 2002-10-10 Phil Schwan <phil@clusterfs.com>
6727 - recovery deadlock fix
6728 - rm -rf causes LBUG fix (617817)
6729 - file open by multiple tasks fix (618962)
6730 - directory permissions bugs (602707 and 620007)
6731 - journal_stop fixed with locking (611313)
6732 - O_APPEND failures resolved (618273, perhaps 614459)
6733 - lconf PATH fix (619770)
6734 - IA64 build fix (621450)
6735 - RPC buffer sizes scale with amount of memory
6737 2002-10-01 Phil Schwan <phil@clusterfs.com>
6740 - locks would be cancelled without throwing away data pages,
6741 resulting in inconsistent data (605627)
6742 - inode attributes were not always being refreshed (605627, 612449)
6743 - lconf now continues to cleanup after lctl reports an error
6744 - MDS now enforces user permissions (602707)
6745 - lprocfs cleanup fixed, but not yet enabled (614157)
6746 - fixed infinite server hang, should a client not respond to an AST
6747 - avoid going into recovery if user calls readlink() with a buffer
6748 that's too small (613941)
6749 - AST RPCs no longer require replies (614867) -- this may be changed
6750 - don't crash server if client sends an IOV that's too big (611336)
6751 - fixed lock conversion deadlock (611892)
6752 - fixed the following of symlinks (614622)
6753 * recovery: the server can remove locks from a client that dies, other
6754 clients can make progress
6755 * more extN patch fixes
6756 * compile-time configurable ptlrpc buffer allocations
6758 - collaborative read cache document
6759 - Lustre Lite Performance CDR document-in-progress
6761 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6764 - fix typo in patch-2.4.18
6766 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6769 - clear ptlrpc request each time in handle_incoming_request()
6770 - unlink of files now destroys the object on the OST
6772 2002-09-19 Peter Braam <braam@clusterfs.com>
6774 * add hard link support
6775 * change obdfile creation method
6776 * kernel patch changed
6778 2002-09-19 Peter Braam <braam@clusterfs.com>
6781 - stack overflow bug in extN fixed
6783 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
6785 * documentation updates
6786 - add man pages for config tools
6787 - update tests/README to describe testing with new config tools
6788 - finish metadata API descriptions
6789 * bug fixes and cleanups
6790 - statfs workaround for 16TB limit
6791 - LOV stripe allocation improved, can stripe on subset of OSTs
6792 - LOV file size/IO offset was wrong for files > 4GB in size
6793 - object EA data was being dropped, caused files to be unreadable
6794 - memory overflow with non-LOV OST caused memory corruption
6795 - fixed regression tests to work with new config tools, obdfilter
6796 - fixed bug when directory size became larger than 1 block
6797 - fixed bug (for single client case) when PWD was deleted
6798 - invalidate local directory pages when doing intent-based ops
6799 - avoid LDLM oops when lock callback contained bad data
6801 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
6803 * documentation updates
6804 * bug fixes and cleanups
6805 - configuration tools
6808 - 64-bit compile warnings
6809 - 64-bit internal statfs data
6811 * test_brw on persistent OST devices
6813 * lprocfs (disabled)
6815 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
6817 * documentation updates
6818 * bug fixes and cleanups
6819 * configuration tools
6821 2002-08-30 Peter J. Braam <braam@clusterfs.com>
6824 * many small fixes to 0_5_4
6825 * io/network handling
6826 * thinkos in MDS operations
6828 2002-08-24 Peter J. Braam <braam@clusterfs.com>
6831 * crucial basic fixes to 0.5.3
6832 * IOR, Iozone work over Elan
6835 2002-08-07 Phil Schwan <phil@clusterfs.com>
6836 * version 0_5_3, our first alpha
6837 * we use the new Portals iovs
6838 * documentation updates
6839 * bug fixes and cleanups
6840 * small changes in the DLM wire protocol
6842 2002-07-25 Peter J. Braam <braam@clusterfs.com>
6843 * version 0_5_1 with some initial stability,
6844 * locking on MD and file I/O.
6845 * documentation updates
6846 * several bug fixes since 0.5.0
6847 * small changes in wire protocol
6849 2002-07-18 Phil Schwan <phil@clusterfs.com>
6851 * delivered as Lustre Light Alpha
6852 * fixed a crash after handling invalid MDS requests
6853 * fixed directory pages for architectures with non-4k pages sizes
6855 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
6856 * release version v0_4_4
6857 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
6859 * Fixes a number of interruption problems with OST operations.
6860 * Update documentation for portals header changes
6861 * Move all wire protocol structs/defines to lustre_idl.h
6862 * Fixes symlink length bug.
6863 * Add tcpdump to repository.
6865 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
6866 * release version v0_4_3
6867 * Fixes statfs for inodes on extN.
6868 * Fixes bug in runtests which would delete /etc/hosts.
6869 * Use 64-bit object IDs wherever possible (not into VFS though)
6870 Remove ost_get_info, which is unused by lustre, and out of date.
6872 2002-07-03 Peter Braam <braam@clusterfs.com>
6873 * release version v0_4_2 Fixes a lookup error (type not passed)
6874 * move forward to head of Portals
6875 * move forward to latest Lustre kernel
6877 2002-06-25 Peter Braam <braam@clusterfs.com>
6878 * release version v0_4_1. Hopefully stable on single node use.